?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
jbd2: enable journal clients to enable v2 checksumming
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blob8dfb42e380eab8a015440c7353a24423b927b8c9
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 void ext4_write_super(struct super_block *sb);
78 static int ext4_freeze(struct super_block *sb);
79 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 const char *dev_name, void *data);
81 static inline int ext2_feature_set_ok(struct super_block *sb);
82 static inline int ext3_feature_set_ok(struct super_block *sb);
83 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
84 static void ext4_destroy_lazyinit_thread(void);
85 static void ext4_unregister_li_request(struct super_block *sb);
86 static void ext4_clear_request_list(void);
87
88 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89 static struct file_system_type ext2_fs_type = {
90 .owner = THIS_MODULE,
91 .name = "ext2",
92 .mount = ext4_mount,
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
95 };
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #else
98 #define IS_EXT2_SB(sb) (0)
99 #endif
100
101
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
106 .mount = ext4_mount,
107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109 };
110 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111 #else
112 #define IS_EXT3_SB(sb) (0)
113 #endif
114
115 static int ext4_verify_csum_type(struct super_block *sb,
116 struct ext4_super_block *es)
117 {
118 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
119 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
120 return 1;
121
122 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
123 }
124
125 static __le32 ext4_superblock_csum(struct super_block *sb,
126 struct ext4_super_block *es)
127 {
128 struct ext4_sb_info *sbi = EXT4_SB(sb);
129 int offset = offsetof(struct ext4_super_block, s_checksum);
130 __u32 csum;
131
132 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
133
134 return cpu_to_le32(csum);
135 }
136
137 int ext4_superblock_csum_verify(struct super_block *sb,
138 struct ext4_super_block *es)
139 {
140 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
141 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
142 return 1;
143
144 return es->s_checksum == ext4_superblock_csum(sb, es);
145 }
146
147 void ext4_superblock_csum_set(struct super_block *sb,
148 struct ext4_super_block *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 * The only special thing we need to do here is to make sure that all
332 * journal_end calls result in the superblock being marked dirty, so
333 * that sync() will call the filesystem's write_super callback if
334 * appropriate.
335 *
336 * To avoid j_barrier hold in userspace when a user calls freeze(),
337 * ext4 prevents a new handle from being started by s_frozen, which
338 * is in an upper layer.
339 */
340 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
341 {
342 journal_t *journal;
343 handle_t *handle;
344
345 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
346 if (sb->s_flags & MS_RDONLY)
347 return ERR_PTR(-EROFS);
348
349 journal = EXT4_SB(sb)->s_journal;
350 handle = ext4_journal_current_handle();
351
352 /*
353 * If a handle has been started, it should be allowed to
354 * finish, otherwise deadlock could happen between freeze
355 * and others(e.g. truncate) due to the restart of the
356 * journal handle if the filesystem is forzen and active
357 * handles are not stopped.
358 */
359 if (!handle)
360 vfs_check_frozen(sb, SB_FREEZE_TRANS);
361
362 if (!journal)
363 return ext4_get_nojournal();
364 /*
365 * Special case here: if the journal has aborted behind our
366 * backs (eg. EIO in the commit thread), then we still need to
367 * take the FS itself readonly cleanly.
368 */
369 if (is_journal_aborted(journal)) {
370 ext4_abort(sb, "Detected aborted journal");
371 return ERR_PTR(-EROFS);
372 }
373 return jbd2_journal_start(journal, nblocks);
374 }
375
376 /*
377 * The only special thing we need to do here is to make sure that all
378 * jbd2_journal_stop calls result in the superblock being marked dirty, so
379 * that sync() will call the filesystem's write_super callback if
380 * appropriate.
381 */
382 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
383 {
384 struct super_block *sb;
385 int err;
386 int rc;
387
388 if (!ext4_handle_valid(handle)) {
389 ext4_put_nojournal(handle);
390 return 0;
391 }
392 sb = handle->h_transaction->t_journal->j_private;
393 err = handle->h_err;
394 rc = jbd2_journal_stop(handle);
395
396 if (!err)
397 err = rc;
398 if (err)
399 __ext4_std_error(sb, where, line, err);
400 return err;
401 }
402
403 void ext4_journal_abort_handle(const char *caller, unsigned int line,
404 const char *err_fn, struct buffer_head *bh,
405 handle_t *handle, int err)
406 {
407 char nbuf[16];
408 const char *errstr = ext4_decode_error(NULL, err, nbuf);
409
410 BUG_ON(!ext4_handle_valid(handle));
411
412 if (bh)
413 BUFFER_TRACE(bh, "abort");
414
415 if (!handle->h_err)
416 handle->h_err = err;
417
418 if (is_handle_aborted(handle))
419 return;
420
421 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
422 caller, line, errstr, err_fn);
423
424 jbd2_journal_abort_handle(handle);
425 }
426
427 static void __save_error_info(struct super_block *sb, const char *func,
428 unsigned int line)
429 {
430 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
431
432 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
433 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
434 es->s_last_error_time = cpu_to_le32(get_seconds());
435 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
436 es->s_last_error_line = cpu_to_le32(line);
437 if (!es->s_first_error_time) {
438 es->s_first_error_time = es->s_last_error_time;
439 strncpy(es->s_first_error_func, func,
440 sizeof(es->s_first_error_func));
441 es->s_first_error_line = cpu_to_le32(line);
442 es->s_first_error_ino = es->s_last_error_ino;
443 es->s_first_error_block = es->s_last_error_block;
444 }
445 /*
446 * Start the daily error reporting function if it hasn't been
447 * started already
448 */
449 if (!es->s_error_count)
450 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
451 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
452 }
453
454 static void save_error_info(struct super_block *sb, const char *func,
455 unsigned int line)
456 {
457 __save_error_info(sb, func, line);
458 ext4_commit_super(sb, 1);
459 }
460
461 /*
462 * The del_gendisk() function uninitializes the disk-specific data
463 * structures, including the bdi structure, without telling anyone
464 * else. Once this happens, any attempt to call mark_buffer_dirty()
465 * (for example, by ext4_commit_super), will cause a kernel OOPS.
466 * This is a kludge to prevent these oops until we can put in a proper
467 * hook in del_gendisk() to inform the VFS and file system layers.
468 */
469 static int block_device_ejected(struct super_block *sb)
470 {
471 struct inode *bd_inode = sb->s_bdev->bd_inode;
472 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
473
474 return bdi->dev == NULL;
475 }
476
477 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
478 {
479 struct super_block *sb = journal->j_private;
480 struct ext4_sb_info *sbi = EXT4_SB(sb);
481 int error = is_journal_aborted(journal);
482 struct ext4_journal_cb_entry *jce, *tmp;
483
484 spin_lock(&sbi->s_md_lock);
485 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
486 list_del_init(&jce->jce_list);
487 spin_unlock(&sbi->s_md_lock);
488 jce->jce_func(sb, jce, error);
489 spin_lock(&sbi->s_md_lock);
490 }
491 spin_unlock(&sbi->s_md_lock);
492 }
493
494 /* Deal with the reporting of failure conditions on a filesystem such as
495 * inconsistencies detected or read IO failures.
496 *
497 * On ext2, we can store the error state of the filesystem in the
498 * superblock. That is not possible on ext4, because we may have other
499 * write ordering constraints on the superblock which prevent us from
500 * writing it out straight away; and given that the journal is about to
501 * be aborted, we can't rely on the current, or future, transactions to
502 * write out the superblock safely.
503 *
504 * We'll just use the jbd2_journal_abort() error code to record an error in
505 * the journal instead. On recovery, the journal will complain about
506 * that error until we've noted it down and cleared it.
507 */
508
509 static void ext4_handle_error(struct super_block *sb)
510 {
511 if (sb->s_flags & MS_RDONLY)
512 return;
513
514 if (!test_opt(sb, ERRORS_CONT)) {
515 journal_t *journal = EXT4_SB(sb)->s_journal;
516
517 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
518 if (journal)
519 jbd2_journal_abort(journal, -EIO);
520 }
521 if (test_opt(sb, ERRORS_RO)) {
522 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
523 sb->s_flags |= MS_RDONLY;
524 }
525 if (test_opt(sb, ERRORS_PANIC))
526 panic("EXT4-fs (device %s): panic forced after error\n",
527 sb->s_id);
528 }
529
530 void __ext4_error(struct super_block *sb, const char *function,
531 unsigned int line, const char *fmt, ...)
532 {
533 struct va_format vaf;
534 va_list args;
535
536 va_start(args, fmt);
537 vaf.fmt = fmt;
538 vaf.va = &args;
539 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
540 sb->s_id, function, line, current->comm, &vaf);
541 va_end(args);
542
543 ext4_handle_error(sb);
544 }
545
546 void ext4_error_inode(struct inode *inode, const char *function,
547 unsigned int line, ext4_fsblk_t block,
548 const char *fmt, ...)
549 {
550 va_list args;
551 struct va_format vaf;
552 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
553
554 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
555 es->s_last_error_block = cpu_to_le64(block);
556 save_error_info(inode->i_sb, function, line);
557 va_start(args, fmt);
558 vaf.fmt = fmt;
559 vaf.va = &args;
560 if (block)
561 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
562 "inode #%lu: block %llu: comm %s: %pV\n",
563 inode->i_sb->s_id, function, line, inode->i_ino,
564 block, current->comm, &vaf);
565 else
566 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
567 "inode #%lu: comm %s: %pV\n",
568 inode->i_sb->s_id, function, line, inode->i_ino,
569 current->comm, &vaf);
570 va_end(args);
571
572 ext4_handle_error(inode->i_sb);
573 }
574
575 void ext4_error_file(struct file *file, const char *function,
576 unsigned int line, ext4_fsblk_t block,
577 const char *fmt, ...)
578 {
579 va_list args;
580 struct va_format vaf;
581 struct ext4_super_block *es;
582 struct inode *inode = file->f_dentry->d_inode;
583 char pathname[80], *path;
584
585 es = EXT4_SB(inode->i_sb)->s_es;
586 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
587 save_error_info(inode->i_sb, function, line);
588 path = d_path(&(file->f_path), pathname, sizeof(pathname));
589 if (IS_ERR(path))
590 path = "(unknown)";
591 va_start(args, fmt);
592 vaf.fmt = fmt;
593 vaf.va = &args;
594 if (block)
595 printk(KERN_CRIT
596 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
597 "block %llu: comm %s: path %s: %pV\n",
598 inode->i_sb->s_id, function, line, inode->i_ino,
599 block, current->comm, path, &vaf);
600 else
601 printk(KERN_CRIT
602 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
603 "comm %s: path %s: %pV\n",
604 inode->i_sb->s_id, function, line, inode->i_ino,
605 current->comm, path, &vaf);
606 va_end(args);
607
608 ext4_handle_error(inode->i_sb);
609 }
610
611 static const char *ext4_decode_error(struct super_block *sb, int errno,
612 char nbuf[16])
613 {
614 char *errstr = NULL;
615
616 switch (errno) {
617 case -EIO:
618 errstr = "IO failure";
619 break;
620 case -ENOMEM:
621 errstr = "Out of memory";
622 break;
623 case -EROFS:
624 if (!sb || (EXT4_SB(sb)->s_journal &&
625 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
626 errstr = "Journal has aborted";
627 else
628 errstr = "Readonly filesystem";
629 break;
630 default:
631 /* If the caller passed in an extra buffer for unknown
632 * errors, textualise them now. Else we just return
633 * NULL. */
634 if (nbuf) {
635 /* Check for truncated error codes... */
636 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
637 errstr = nbuf;
638 }
639 break;
640 }
641
642 return errstr;
643 }
644
645 /* __ext4_std_error decodes expected errors from journaling functions
646 * automatically and invokes the appropriate error response. */
647
648 void __ext4_std_error(struct super_block *sb, const char *function,
649 unsigned int line, int errno)
650 {
651 char nbuf[16];
652 const char *errstr;
653
654 /* Special case: if the error is EROFS, and we're not already
655 * inside a transaction, then there's really no point in logging
656 * an error. */
657 if (errno == -EROFS && journal_current_handle() == NULL &&
658 (sb->s_flags & MS_RDONLY))
659 return;
660
661 errstr = ext4_decode_error(sb, errno, nbuf);
662 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
663 sb->s_id, function, line, errstr);
664 save_error_info(sb, function, line);
665
666 ext4_handle_error(sb);
667 }
668
669 /*
670 * ext4_abort is a much stronger failure handler than ext4_error. The
671 * abort function may be used to deal with unrecoverable failures such
672 * as journal IO errors or ENOMEM at a critical moment in log management.
673 *
674 * We unconditionally force the filesystem into an ABORT|READONLY state,
675 * unless the error response on the fs has been set to panic in which
676 * case we take the easy way out and panic immediately.
677 */
678
679 void __ext4_abort(struct super_block *sb, const char *function,
680 unsigned int line, const char *fmt, ...)
681 {
682 va_list args;
683
684 save_error_info(sb, function, line);
685 va_start(args, fmt);
686 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
687 function, line);
688 vprintk(fmt, args);
689 printk("\n");
690 va_end(args);
691
692 if ((sb->s_flags & MS_RDONLY) == 0) {
693 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
694 sb->s_flags |= MS_RDONLY;
695 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
696 if (EXT4_SB(sb)->s_journal)
697 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
698 save_error_info(sb, function, line);
699 }
700 if (test_opt(sb, ERRORS_PANIC))
701 panic("EXT4-fs panic from previous error\n");
702 }
703
704 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
705 {
706 struct va_format vaf;
707 va_list args;
708
709 va_start(args, fmt);
710 vaf.fmt = fmt;
711 vaf.va = &args;
712 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
713 va_end(args);
714 }
715
716 void __ext4_warning(struct super_block *sb, const char *function,
717 unsigned int line, const char *fmt, ...)
718 {
719 struct va_format vaf;
720 va_list args;
721
722 va_start(args, fmt);
723 vaf.fmt = fmt;
724 vaf.va = &args;
725 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
726 sb->s_id, function, line, &vaf);
727 va_end(args);
728 }
729
730 void __ext4_grp_locked_error(const char *function, unsigned int line,
731 struct super_block *sb, ext4_group_t grp,
732 unsigned long ino, ext4_fsblk_t block,
733 const char *fmt, ...)
734 __releases(bitlock)
735 __acquires(bitlock)
736 {
737 struct va_format vaf;
738 va_list args;
739 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
740
741 es->s_last_error_ino = cpu_to_le32(ino);
742 es->s_last_error_block = cpu_to_le64(block);
743 __save_error_info(sb, function, line);
744
745 va_start(args, fmt);
746
747 vaf.fmt = fmt;
748 vaf.va = &args;
749 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
750 sb->s_id, function, line, grp);
751 if (ino)
752 printk(KERN_CONT "inode %lu: ", ino);
753 if (block)
754 printk(KERN_CONT "block %llu:", (unsigned long long) block);
755 printk(KERN_CONT "%pV\n", &vaf);
756 va_end(args);
757
758 if (test_opt(sb, ERRORS_CONT)) {
759 ext4_commit_super(sb, 0);
760 return;
761 }
762
763 ext4_unlock_group(sb, grp);
764 ext4_handle_error(sb);
765 /*
766 * We only get here in the ERRORS_RO case; relocking the group
767 * may be dangerous, but nothing bad will happen since the
768 * filesystem will have already been marked read/only and the
769 * journal has been aborted. We return 1 as a hint to callers
770 * who might what to use the return value from
771 * ext4_grp_locked_error() to distinguish between the
772 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
773 * aggressively from the ext4 function in question, with a
774 * more appropriate error code.
775 */
776 ext4_lock_group(sb, grp);
777 return;
778 }
779
780 void ext4_update_dynamic_rev(struct super_block *sb)
781 {
782 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
783
784 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
785 return;
786
787 ext4_warning(sb,
788 "updating to rev %d because of new feature flag, "
789 "running e2fsck is recommended",
790 EXT4_DYNAMIC_REV);
791
792 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
793 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
794 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
795 /* leave es->s_feature_*compat flags alone */
796 /* es->s_uuid will be set by e2fsck if empty */
797
798 /*
799 * The rest of the superblock fields should be zero, and if not it
800 * means they are likely already in use, so leave them alone. We
801 * can leave it up to e2fsck to clean up any inconsistencies there.
802 */
803 }
804
805 /*
806 * Open the external journal device
807 */
808 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
809 {
810 struct block_device *bdev;
811 char b[BDEVNAME_SIZE];
812
813 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
814 if (IS_ERR(bdev))
815 goto fail;
816 return bdev;
817
818 fail:
819 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
820 __bdevname(dev, b), PTR_ERR(bdev));
821 return NULL;
822 }
823
824 /*
825 * Release the journal device
826 */
827 static int ext4_blkdev_put(struct block_device *bdev)
828 {
829 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
830 }
831
832 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
833 {
834 struct block_device *bdev;
835 int ret = -ENODEV;
836
837 bdev = sbi->journal_bdev;
838 if (bdev) {
839 ret = ext4_blkdev_put(bdev);
840 sbi->journal_bdev = NULL;
841 }
842 return ret;
843 }
844
845 static inline struct inode *orphan_list_entry(struct list_head *l)
846 {
847 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
848 }
849
850 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
851 {
852 struct list_head *l;
853
854 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
855 le32_to_cpu(sbi->s_es->s_last_orphan));
856
857 printk(KERN_ERR "sb_info orphan list:\n");
858 list_for_each(l, &sbi->s_orphan) {
859 struct inode *inode = orphan_list_entry(l);
860 printk(KERN_ERR " "
861 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
862 inode->i_sb->s_id, inode->i_ino, inode,
863 inode->i_mode, inode->i_nlink,
864 NEXT_ORPHAN(inode));
865 }
866 }
867
868 static void ext4_put_super(struct super_block *sb)
869 {
870 struct ext4_sb_info *sbi = EXT4_SB(sb);
871 struct ext4_super_block *es = sbi->s_es;
872 int i, err;
873
874 ext4_unregister_li_request(sb);
875 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
876
877 flush_workqueue(sbi->dio_unwritten_wq);
878 destroy_workqueue(sbi->dio_unwritten_wq);
879
880 lock_super(sb);
881 if (sbi->s_journal) {
882 err = jbd2_journal_destroy(sbi->s_journal);
883 sbi->s_journal = NULL;
884 if (err < 0)
885 ext4_abort(sb, "Couldn't clean up the journal");
886 }
887
888 del_timer(&sbi->s_err_report);
889 ext4_release_system_zone(sb);
890 ext4_mb_release(sb);
891 ext4_ext_release(sb);
892 ext4_xattr_put_super(sb);
893
894 if (!(sb->s_flags & MS_RDONLY)) {
895 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
896 es->s_state = cpu_to_le16(sbi->s_mount_state);
897 }
898 if (sb->s_dirt || !(sb->s_flags & MS_RDONLY))
899 ext4_commit_super(sb, 1);
900
901 if (sbi->s_proc) {
902 remove_proc_entry("options", sbi->s_proc);
903 remove_proc_entry(sb->s_id, ext4_proc_root);
904 }
905 kobject_del(&sbi->s_kobj);
906
907 for (i = 0; i < sbi->s_gdb_count; i++)
908 brelse(sbi->s_group_desc[i]);
909 ext4_kvfree(sbi->s_group_desc);
910 ext4_kvfree(sbi->s_flex_groups);
911 percpu_counter_destroy(&sbi->s_freeclusters_counter);
912 percpu_counter_destroy(&sbi->s_freeinodes_counter);
913 percpu_counter_destroy(&sbi->s_dirs_counter);
914 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
915 brelse(sbi->s_sbh);
916 #ifdef CONFIG_QUOTA
917 for (i = 0; i < MAXQUOTAS; i++)
918 kfree(sbi->s_qf_names[i]);
919 #endif
920
921 /* Debugging code just in case the in-memory inode orphan list
922 * isn't empty. The on-disk one can be non-empty if we've
923 * detected an error and taken the fs readonly, but the
924 * in-memory list had better be clean by this point. */
925 if (!list_empty(&sbi->s_orphan))
926 dump_orphan_list(sb, sbi);
927 J_ASSERT(list_empty(&sbi->s_orphan));
928
929 invalidate_bdev(sb->s_bdev);
930 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
931 /*
932 * Invalidate the journal device's buffers. We don't want them
933 * floating about in memory - the physical journal device may
934 * hotswapped, and it breaks the `ro-after' testing code.
935 */
936 sync_blockdev(sbi->journal_bdev);
937 invalidate_bdev(sbi->journal_bdev);
938 ext4_blkdev_remove(sbi);
939 }
940 if (sbi->s_mmp_tsk)
941 kthread_stop(sbi->s_mmp_tsk);
942 sb->s_fs_info = NULL;
943 /*
944 * Now that we are completely done shutting down the
945 * superblock, we need to actually destroy the kobject.
946 */
947 unlock_super(sb);
948 kobject_put(&sbi->s_kobj);
949 wait_for_completion(&sbi->s_kobj_unregister);
950 if (sbi->s_chksum_driver)
951 crypto_free_shash(sbi->s_chksum_driver);
952 kfree(sbi->s_blockgroup_lock);
953 kfree(sbi);
954 }
955
956 static struct kmem_cache *ext4_inode_cachep;
957
958 /*
959 * Called inside transaction, so use GFP_NOFS
960 */
961 static struct inode *ext4_alloc_inode(struct super_block *sb)
962 {
963 struct ext4_inode_info *ei;
964
965 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
966 if (!ei)
967 return NULL;
968
969 ei->vfs_inode.i_version = 1;
970 ei->vfs_inode.i_data.writeback_index = 0;
971 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
972 INIT_LIST_HEAD(&ei->i_prealloc_list);
973 spin_lock_init(&ei->i_prealloc_lock);
974 ei->i_reserved_data_blocks = 0;
975 ei->i_reserved_meta_blocks = 0;
976 ei->i_allocated_meta_blocks = 0;
977 ei->i_da_metadata_calc_len = 0;
978 spin_lock_init(&(ei->i_block_reservation_lock));
979 #ifdef CONFIG_QUOTA
980 ei->i_reserved_quota = 0;
981 #endif
982 ei->jinode = NULL;
983 INIT_LIST_HEAD(&ei->i_completed_io_list);
984 spin_lock_init(&ei->i_completed_io_lock);
985 ei->cur_aio_dio = NULL;
986 ei->i_sync_tid = 0;
987 ei->i_datasync_tid = 0;
988 atomic_set(&ei->i_ioend_count, 0);
989 atomic_set(&ei->i_aiodio_unwritten, 0);
990
991 return &ei->vfs_inode;
992 }
993
994 static int ext4_drop_inode(struct inode *inode)
995 {
996 int drop = generic_drop_inode(inode);
997
998 trace_ext4_drop_inode(inode, drop);
999 return drop;
1000 }
1001
1002 static void ext4_i_callback(struct rcu_head *head)
1003 {
1004 struct inode *inode = container_of(head, struct inode, i_rcu);
1005 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1006 }
1007
1008 static void ext4_destroy_inode(struct inode *inode)
1009 {
1010 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1011 ext4_msg(inode->i_sb, KERN_ERR,
1012 "Inode %lu (%p): orphan list check failed!",
1013 inode->i_ino, EXT4_I(inode));
1014 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1015 EXT4_I(inode), sizeof(struct ext4_inode_info),
1016 true);
1017 dump_stack();
1018 }
1019 call_rcu(&inode->i_rcu, ext4_i_callback);
1020 }
1021
1022 static void init_once(void *foo)
1023 {
1024 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1025
1026 INIT_LIST_HEAD(&ei->i_orphan);
1027 #ifdef CONFIG_EXT4_FS_XATTR
1028 init_rwsem(&ei->xattr_sem);
1029 #endif
1030 init_rwsem(&ei->i_data_sem);
1031 inode_init_once(&ei->vfs_inode);
1032 }
1033
1034 static int init_inodecache(void)
1035 {
1036 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1037 sizeof(struct ext4_inode_info),
1038 0, (SLAB_RECLAIM_ACCOUNT|
1039 SLAB_MEM_SPREAD),
1040 init_once);
1041 if (ext4_inode_cachep == NULL)
1042 return -ENOMEM;
1043 return 0;
1044 }
1045
1046 static void destroy_inodecache(void)
1047 {
1048 kmem_cache_destroy(ext4_inode_cachep);
1049 }
1050
1051 void ext4_clear_inode(struct inode *inode)
1052 {
1053 invalidate_inode_buffers(inode);
1054 end_writeback(inode);
1055 dquot_drop(inode);
1056 ext4_discard_preallocations(inode);
1057 if (EXT4_I(inode)->jinode) {
1058 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1059 EXT4_I(inode)->jinode);
1060 jbd2_free_inode(EXT4_I(inode)->jinode);
1061 EXT4_I(inode)->jinode = NULL;
1062 }
1063 }
1064
1065 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1066 u64 ino, u32 generation)
1067 {
1068 struct inode *inode;
1069
1070 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1071 return ERR_PTR(-ESTALE);
1072 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1073 return ERR_PTR(-ESTALE);
1074
1075 /* iget isn't really right if the inode is currently unallocated!!
1076 *
1077 * ext4_read_inode will return a bad_inode if the inode had been
1078 * deleted, so we should be safe.
1079 *
1080 * Currently we don't know the generation for parent directory, so
1081 * a generation of 0 means "accept any"
1082 */
1083 inode = ext4_iget(sb, ino);
1084 if (IS_ERR(inode))
1085 return ERR_CAST(inode);
1086 if (generation && inode->i_generation != generation) {
1087 iput(inode);
1088 return ERR_PTR(-ESTALE);
1089 }
1090
1091 return inode;
1092 }
1093
1094 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1095 int fh_len, int fh_type)
1096 {
1097 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1098 ext4_nfs_get_inode);
1099 }
1100
1101 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1102 int fh_len, int fh_type)
1103 {
1104 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1105 ext4_nfs_get_inode);
1106 }
1107
1108 /*
1109 * Try to release metadata pages (indirect blocks, directories) which are
1110 * mapped via the block device. Since these pages could have journal heads
1111 * which would prevent try_to_free_buffers() from freeing them, we must use
1112 * jbd2 layer's try_to_free_buffers() function to release them.
1113 */
1114 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1115 gfp_t wait)
1116 {
1117 journal_t *journal = EXT4_SB(sb)->s_journal;
1118
1119 WARN_ON(PageChecked(page));
1120 if (!page_has_buffers(page))
1121 return 0;
1122 if (journal)
1123 return jbd2_journal_try_to_free_buffers(journal, page,
1124 wait & ~__GFP_WAIT);
1125 return try_to_free_buffers(page);
1126 }
1127
1128 #ifdef CONFIG_QUOTA
1129 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1130 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1131
1132 static int ext4_write_dquot(struct dquot *dquot);
1133 static int ext4_acquire_dquot(struct dquot *dquot);
1134 static int ext4_release_dquot(struct dquot *dquot);
1135 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1136 static int ext4_write_info(struct super_block *sb, int type);
1137 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1138 struct path *path);
1139 static int ext4_quota_off(struct super_block *sb, int type);
1140 static int ext4_quota_on_mount(struct super_block *sb, int type);
1141 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1142 size_t len, loff_t off);
1143 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1144 const char *data, size_t len, loff_t off);
1145
1146 static const struct dquot_operations ext4_quota_operations = {
1147 .get_reserved_space = ext4_get_reserved_space,
1148 .write_dquot = ext4_write_dquot,
1149 .acquire_dquot = ext4_acquire_dquot,
1150 .release_dquot = ext4_release_dquot,
1151 .mark_dirty = ext4_mark_dquot_dirty,
1152 .write_info = ext4_write_info,
1153 .alloc_dquot = dquot_alloc,
1154 .destroy_dquot = dquot_destroy,
1155 };
1156
1157 static const struct quotactl_ops ext4_qctl_operations = {
1158 .quota_on = ext4_quota_on,
1159 .quota_off = ext4_quota_off,
1160 .quota_sync = dquot_quota_sync,
1161 .get_info = dquot_get_dqinfo,
1162 .set_info = dquot_set_dqinfo,
1163 .get_dqblk = dquot_get_dqblk,
1164 .set_dqblk = dquot_set_dqblk
1165 };
1166 #endif
1167
1168 static const struct super_operations ext4_sops = {
1169 .alloc_inode = ext4_alloc_inode,
1170 .destroy_inode = ext4_destroy_inode,
1171 .write_inode = ext4_write_inode,
1172 .dirty_inode = ext4_dirty_inode,
1173 .drop_inode = ext4_drop_inode,
1174 .evict_inode = ext4_evict_inode,
1175 .put_super = ext4_put_super,
1176 .sync_fs = ext4_sync_fs,
1177 .freeze_fs = ext4_freeze,
1178 .unfreeze_fs = ext4_unfreeze,
1179 .statfs = ext4_statfs,
1180 .remount_fs = ext4_remount,
1181 .show_options = ext4_show_options,
1182 #ifdef CONFIG_QUOTA
1183 .quota_read = ext4_quota_read,
1184 .quota_write = ext4_quota_write,
1185 #endif
1186 .bdev_try_to_free_page = bdev_try_to_free_page,
1187 };
1188
1189 static const struct super_operations ext4_nojournal_sops = {
1190 .alloc_inode = ext4_alloc_inode,
1191 .destroy_inode = ext4_destroy_inode,
1192 .write_inode = ext4_write_inode,
1193 .dirty_inode = ext4_dirty_inode,
1194 .drop_inode = ext4_drop_inode,
1195 .evict_inode = ext4_evict_inode,
1196 .write_super = ext4_write_super,
1197 .put_super = ext4_put_super,
1198 .statfs = ext4_statfs,
1199 .remount_fs = ext4_remount,
1200 .show_options = ext4_show_options,
1201 #ifdef CONFIG_QUOTA
1202 .quota_read = ext4_quota_read,
1203 .quota_write = ext4_quota_write,
1204 #endif
1205 .bdev_try_to_free_page = bdev_try_to_free_page,
1206 };
1207
1208 static const struct export_operations ext4_export_ops = {
1209 .fh_to_dentry = ext4_fh_to_dentry,
1210 .fh_to_parent = ext4_fh_to_parent,
1211 .get_parent = ext4_get_parent,
1212 };
1213
1214 enum {
1215 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1216 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1217 Opt_nouid32, Opt_debug, Opt_removed,
1218 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1219 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1220 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1221 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1222 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1223 Opt_data_err_abort, Opt_data_err_ignore,
1224 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1225 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1226 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1227 Opt_usrquota, Opt_grpquota, Opt_i_version,
1228 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1229 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1230 Opt_inode_readahead_blks, Opt_journal_ioprio,
1231 Opt_dioread_nolock, Opt_dioread_lock,
1232 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1233 };
1234
1235 static const match_table_t tokens = {
1236 {Opt_bsd_df, "bsddf"},
1237 {Opt_minix_df, "minixdf"},
1238 {Opt_grpid, "grpid"},
1239 {Opt_grpid, "bsdgroups"},
1240 {Opt_nogrpid, "nogrpid"},
1241 {Opt_nogrpid, "sysvgroups"},
1242 {Opt_resgid, "resgid=%u"},
1243 {Opt_resuid, "resuid=%u"},
1244 {Opt_sb, "sb=%u"},
1245 {Opt_err_cont, "errors=continue"},
1246 {Opt_err_panic, "errors=panic"},
1247 {Opt_err_ro, "errors=remount-ro"},
1248 {Opt_nouid32, "nouid32"},
1249 {Opt_debug, "debug"},
1250 {Opt_removed, "oldalloc"},
1251 {Opt_removed, "orlov"},
1252 {Opt_user_xattr, "user_xattr"},
1253 {Opt_nouser_xattr, "nouser_xattr"},
1254 {Opt_acl, "acl"},
1255 {Opt_noacl, "noacl"},
1256 {Opt_noload, "norecovery"},
1257 {Opt_noload, "noload"},
1258 {Opt_removed, "nobh"},
1259 {Opt_removed, "bh"},
1260 {Opt_commit, "commit=%u"},
1261 {Opt_min_batch_time, "min_batch_time=%u"},
1262 {Opt_max_batch_time, "max_batch_time=%u"},
1263 {Opt_journal_dev, "journal_dev=%u"},
1264 {Opt_journal_checksum, "journal_checksum"},
1265 {Opt_journal_async_commit, "journal_async_commit"},
1266 {Opt_abort, "abort"},
1267 {Opt_data_journal, "data=journal"},
1268 {Opt_data_ordered, "data=ordered"},
1269 {Opt_data_writeback, "data=writeback"},
1270 {Opt_data_err_abort, "data_err=abort"},
1271 {Opt_data_err_ignore, "data_err=ignore"},
1272 {Opt_offusrjquota, "usrjquota="},
1273 {Opt_usrjquota, "usrjquota=%s"},
1274 {Opt_offgrpjquota, "grpjquota="},
1275 {Opt_grpjquota, "grpjquota=%s"},
1276 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1277 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1278 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1279 {Opt_grpquota, "grpquota"},
1280 {Opt_noquota, "noquota"},
1281 {Opt_quota, "quota"},
1282 {Opt_usrquota, "usrquota"},
1283 {Opt_barrier, "barrier=%u"},
1284 {Opt_barrier, "barrier"},
1285 {Opt_nobarrier, "nobarrier"},
1286 {Opt_i_version, "i_version"},
1287 {Opt_stripe, "stripe=%u"},
1288 {Opt_delalloc, "delalloc"},
1289 {Opt_nodelalloc, "nodelalloc"},
1290 {Opt_mblk_io_submit, "mblk_io_submit"},
1291 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1292 {Opt_block_validity, "block_validity"},
1293 {Opt_noblock_validity, "noblock_validity"},
1294 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1295 {Opt_journal_ioprio, "journal_ioprio=%u"},
1296 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1297 {Opt_auto_da_alloc, "auto_da_alloc"},
1298 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1299 {Opt_dioread_nolock, "dioread_nolock"},
1300 {Opt_dioread_lock, "dioread_lock"},
1301 {Opt_discard, "discard"},
1302 {Opt_nodiscard, "nodiscard"},
1303 {Opt_init_itable, "init_itable=%u"},
1304 {Opt_init_itable, "init_itable"},
1305 {Opt_noinit_itable, "noinit_itable"},
1306 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1307 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1308 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1309 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1310 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1311 {Opt_err, NULL},
1312 };
1313
1314 static ext4_fsblk_t get_sb_block(void **data)
1315 {
1316 ext4_fsblk_t sb_block;
1317 char *options = (char *) *data;
1318
1319 if (!options || strncmp(options, "sb=", 3) != 0)
1320 return 1; /* Default location */
1321
1322 options += 3;
1323 /* TODO: use simple_strtoll with >32bit ext4 */
1324 sb_block = simple_strtoul(options, &options, 0);
1325 if (*options && *options != ',') {
1326 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1327 (char *) *data);
1328 return 1;
1329 }
1330 if (*options == ',')
1331 options++;
1332 *data = (void *) options;
1333
1334 return sb_block;
1335 }
1336
1337 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1338 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1339 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1340
1341 #ifdef CONFIG_QUOTA
1342 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1343 {
1344 struct ext4_sb_info *sbi = EXT4_SB(sb);
1345 char *qname;
1346
1347 if (sb_any_quota_loaded(sb) &&
1348 !sbi->s_qf_names[qtype]) {
1349 ext4_msg(sb, KERN_ERR,
1350 "Cannot change journaled "
1351 "quota options when quota turned on");
1352 return -1;
1353 }
1354 qname = match_strdup(args);
1355 if (!qname) {
1356 ext4_msg(sb, KERN_ERR,
1357 "Not enough memory for storing quotafile name");
1358 return -1;
1359 }
1360 if (sbi->s_qf_names[qtype] &&
1361 strcmp(sbi->s_qf_names[qtype], qname)) {
1362 ext4_msg(sb, KERN_ERR,
1363 "%s quota file already specified", QTYPE2NAME(qtype));
1364 kfree(qname);
1365 return -1;
1366 }
1367 sbi->s_qf_names[qtype] = qname;
1368 if (strchr(sbi->s_qf_names[qtype], '/')) {
1369 ext4_msg(sb, KERN_ERR,
1370 "quotafile must be on filesystem root");
1371 kfree(sbi->s_qf_names[qtype]);
1372 sbi->s_qf_names[qtype] = NULL;
1373 return -1;
1374 }
1375 set_opt(sb, QUOTA);
1376 return 1;
1377 }
1378
1379 static int clear_qf_name(struct super_block *sb, int qtype)
1380 {
1381
1382 struct ext4_sb_info *sbi = EXT4_SB(sb);
1383
1384 if (sb_any_quota_loaded(sb) &&
1385 sbi->s_qf_names[qtype]) {
1386 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1387 " when quota turned on");
1388 return -1;
1389 }
1390 /*
1391 * The space will be released later when all options are confirmed
1392 * to be correct
1393 */
1394 sbi->s_qf_names[qtype] = NULL;
1395 return 1;
1396 }
1397 #endif
1398
1399 #define MOPT_SET 0x0001
1400 #define MOPT_CLEAR 0x0002
1401 #define MOPT_NOSUPPORT 0x0004
1402 #define MOPT_EXPLICIT 0x0008
1403 #define MOPT_CLEAR_ERR 0x0010
1404 #define MOPT_GTE0 0x0020
1405 #ifdef CONFIG_QUOTA
1406 #define MOPT_Q 0
1407 #define MOPT_QFMT 0x0040
1408 #else
1409 #define MOPT_Q MOPT_NOSUPPORT
1410 #define MOPT_QFMT MOPT_NOSUPPORT
1411 #endif
1412 #define MOPT_DATAJ 0x0080
1413
1414 static const struct mount_opts {
1415 int token;
1416 int mount_opt;
1417 int flags;
1418 } ext4_mount_opts[] = {
1419 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1420 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1421 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1422 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1423 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1424 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1425 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1426 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1427 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1428 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1429 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1430 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1431 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1432 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1433 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1434 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1435 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1436 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1437 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1438 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1439 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1440 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1441 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1442 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1443 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1444 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1445 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1446 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1447 {Opt_commit, 0, MOPT_GTE0},
1448 {Opt_max_batch_time, 0, MOPT_GTE0},
1449 {Opt_min_batch_time, 0, MOPT_GTE0},
1450 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1451 {Opt_init_itable, 0, MOPT_GTE0},
1452 {Opt_stripe, 0, MOPT_GTE0},
1453 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1454 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1455 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1456 #ifdef CONFIG_EXT4_FS_XATTR
1457 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1458 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1459 #else
1460 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1461 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
1462 #endif
1463 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1464 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1465 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1466 #else
1467 {Opt_acl, 0, MOPT_NOSUPPORT},
1468 {Opt_noacl, 0, MOPT_NOSUPPORT},
1469 #endif
1470 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1471 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1472 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1473 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1474 MOPT_SET | MOPT_Q},
1475 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1476 MOPT_SET | MOPT_Q},
1477 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1478 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1479 {Opt_usrjquota, 0, MOPT_Q},
1480 {Opt_grpjquota, 0, MOPT_Q},
1481 {Opt_offusrjquota, 0, MOPT_Q},
1482 {Opt_offgrpjquota, 0, MOPT_Q},
1483 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1484 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1485 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1486 {Opt_err, 0, 0}
1487 };
1488
1489 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1490 substring_t *args, unsigned long *journal_devnum,
1491 unsigned int *journal_ioprio, int is_remount)
1492 {
1493 struct ext4_sb_info *sbi = EXT4_SB(sb);
1494 const struct mount_opts *m;
1495 int arg = 0;
1496
1497 #ifdef CONFIG_QUOTA
1498 if (token == Opt_usrjquota)
1499 return set_qf_name(sb, USRQUOTA, &args[0]);
1500 else if (token == Opt_grpjquota)
1501 return set_qf_name(sb, GRPQUOTA, &args[0]);
1502 else if (token == Opt_offusrjquota)
1503 return clear_qf_name(sb, USRQUOTA);
1504 else if (token == Opt_offgrpjquota)
1505 return clear_qf_name(sb, GRPQUOTA);
1506 #endif
1507 if (args->from && match_int(args, &arg))
1508 return -1;
1509 switch (token) {
1510 case Opt_noacl:
1511 case Opt_nouser_xattr:
1512 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1513 break;
1514 case Opt_sb:
1515 return 1; /* handled by get_sb_block() */
1516 case Opt_removed:
1517 ext4_msg(sb, KERN_WARNING,
1518 "Ignoring removed %s option", opt);
1519 return 1;
1520 case Opt_resuid:
1521 sbi->s_resuid = arg;
1522 return 1;
1523 case Opt_resgid:
1524 sbi->s_resgid = arg;
1525 return 1;
1526 case Opt_abort:
1527 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1528 return 1;
1529 case Opt_i_version:
1530 sb->s_flags |= MS_I_VERSION;
1531 return 1;
1532 case Opt_journal_dev:
1533 if (is_remount) {
1534 ext4_msg(sb, KERN_ERR,
1535 "Cannot specify journal on remount");
1536 return -1;
1537 }
1538 *journal_devnum = arg;
1539 return 1;
1540 case Opt_journal_ioprio:
1541 if (arg < 0 || arg > 7)
1542 return -1;
1543 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1544 return 1;
1545 }
1546
1547 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1548 if (token != m->token)
1549 continue;
1550 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1551 return -1;
1552 if (m->flags & MOPT_EXPLICIT)
1553 set_opt2(sb, EXPLICIT_DELALLOC);
1554 if (m->flags & MOPT_CLEAR_ERR)
1555 clear_opt(sb, ERRORS_MASK);
1556 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1557 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1558 "options when quota turned on");
1559 return -1;
1560 }
1561
1562 if (m->flags & MOPT_NOSUPPORT) {
1563 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1564 } else if (token == Opt_commit) {
1565 if (arg == 0)
1566 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1567 sbi->s_commit_interval = HZ * arg;
1568 } else if (token == Opt_max_batch_time) {
1569 if (arg == 0)
1570 arg = EXT4_DEF_MAX_BATCH_TIME;
1571 sbi->s_max_batch_time = arg;
1572 } else if (token == Opt_min_batch_time) {
1573 sbi->s_min_batch_time = arg;
1574 } else if (token == Opt_inode_readahead_blks) {
1575 if (arg > (1 << 30))
1576 return -1;
1577 if (arg && !is_power_of_2(arg)) {
1578 ext4_msg(sb, KERN_ERR,
1579 "EXT4-fs: inode_readahead_blks"
1580 " must be a power of 2");
1581 return -1;
1582 }
1583 sbi->s_inode_readahead_blks = arg;
1584 } else if (token == Opt_init_itable) {
1585 set_opt(sb, INIT_INODE_TABLE);
1586 if (!args->from)
1587 arg = EXT4_DEF_LI_WAIT_MULT;
1588 sbi->s_li_wait_mult = arg;
1589 } else if (token == Opt_stripe) {
1590 sbi->s_stripe = arg;
1591 } else if (m->flags & MOPT_DATAJ) {
1592 if (is_remount) {
1593 if (!sbi->s_journal)
1594 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1595 else if (test_opt(sb, DATA_FLAGS) !=
1596 m->mount_opt) {
1597 ext4_msg(sb, KERN_ERR,
1598 "Cannot change data mode on remount");
1599 return -1;
1600 }
1601 } else {
1602 clear_opt(sb, DATA_FLAGS);
1603 sbi->s_mount_opt |= m->mount_opt;
1604 }
1605 #ifdef CONFIG_QUOTA
1606 } else if (m->flags & MOPT_QFMT) {
1607 if (sb_any_quota_loaded(sb) &&
1608 sbi->s_jquota_fmt != m->mount_opt) {
1609 ext4_msg(sb, KERN_ERR, "Cannot "
1610 "change journaled quota options "
1611 "when quota turned on");
1612 return -1;
1613 }
1614 sbi->s_jquota_fmt = m->mount_opt;
1615 #endif
1616 } else {
1617 if (!args->from)
1618 arg = 1;
1619 if (m->flags & MOPT_CLEAR)
1620 arg = !arg;
1621 else if (unlikely(!(m->flags & MOPT_SET))) {
1622 ext4_msg(sb, KERN_WARNING,
1623 "buggy handling of option %s", opt);
1624 WARN_ON(1);
1625 return -1;
1626 }
1627 if (arg != 0)
1628 sbi->s_mount_opt |= m->mount_opt;
1629 else
1630 sbi->s_mount_opt &= ~m->mount_opt;
1631 }
1632 return 1;
1633 }
1634 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1635 "or missing value", opt);
1636 return -1;
1637 }
1638
1639 static int parse_options(char *options, struct super_block *sb,
1640 unsigned long *journal_devnum,
1641 unsigned int *journal_ioprio,
1642 int is_remount)
1643 {
1644 #ifdef CONFIG_QUOTA
1645 struct ext4_sb_info *sbi = EXT4_SB(sb);
1646 #endif
1647 char *p;
1648 substring_t args[MAX_OPT_ARGS];
1649 int token;
1650
1651 if (!options)
1652 return 1;
1653
1654 while ((p = strsep(&options, ",")) != NULL) {
1655 if (!*p)
1656 continue;
1657 /*
1658 * Initialize args struct so we know whether arg was
1659 * found; some options take optional arguments.
1660 */
1661 args[0].to = args[0].from = 0;
1662 token = match_token(p, tokens, args);
1663 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1664 journal_ioprio, is_remount) < 0)
1665 return 0;
1666 }
1667 #ifdef CONFIG_QUOTA
1668 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1669 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1670 clear_opt(sb, USRQUOTA);
1671
1672 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1673 clear_opt(sb, GRPQUOTA);
1674
1675 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1676 ext4_msg(sb, KERN_ERR, "old and new quota "
1677 "format mixing");
1678 return 0;
1679 }
1680
1681 if (!sbi->s_jquota_fmt) {
1682 ext4_msg(sb, KERN_ERR, "journaled quota format "
1683 "not specified");
1684 return 0;
1685 }
1686 } else {
1687 if (sbi->s_jquota_fmt) {
1688 ext4_msg(sb, KERN_ERR, "journaled quota format "
1689 "specified with no journaling "
1690 "enabled");
1691 return 0;
1692 }
1693 }
1694 #endif
1695 return 1;
1696 }
1697
1698 static inline void ext4_show_quota_options(struct seq_file *seq,
1699 struct super_block *sb)
1700 {
1701 #if defined(CONFIG_QUOTA)
1702 struct ext4_sb_info *sbi = EXT4_SB(sb);
1703
1704 if (sbi->s_jquota_fmt) {
1705 char *fmtname = "";
1706
1707 switch (sbi->s_jquota_fmt) {
1708 case QFMT_VFS_OLD:
1709 fmtname = "vfsold";
1710 break;
1711 case QFMT_VFS_V0:
1712 fmtname = "vfsv0";
1713 break;
1714 case QFMT_VFS_V1:
1715 fmtname = "vfsv1";
1716 break;
1717 }
1718 seq_printf(seq, ",jqfmt=%s", fmtname);
1719 }
1720
1721 if (sbi->s_qf_names[USRQUOTA])
1722 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1723
1724 if (sbi->s_qf_names[GRPQUOTA])
1725 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1726
1727 if (test_opt(sb, USRQUOTA))
1728 seq_puts(seq, ",usrquota");
1729
1730 if (test_opt(sb, GRPQUOTA))
1731 seq_puts(seq, ",grpquota");
1732 #endif
1733 }
1734
1735 static const char *token2str(int token)
1736 {
1737 static const struct match_token *t;
1738
1739 for (t = tokens; t->token != Opt_err; t++)
1740 if (t->token == token && !strchr(t->pattern, '='))
1741 break;
1742 return t->pattern;
1743 }
1744
1745 /*
1746 * Show an option if
1747 * - it's set to a non-default value OR
1748 * - if the per-sb default is different from the global default
1749 */
1750 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1751 int nodefs)
1752 {
1753 struct ext4_sb_info *sbi = EXT4_SB(sb);
1754 struct ext4_super_block *es = sbi->s_es;
1755 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1756 const struct mount_opts *m;
1757 char sep = nodefs ? '\n' : ',';
1758
1759 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1760 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1761
1762 if (sbi->s_sb_block != 1)
1763 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1764
1765 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1766 int want_set = m->flags & MOPT_SET;
1767 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1768 (m->flags & MOPT_CLEAR_ERR))
1769 continue;
1770 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1771 continue; /* skip if same as the default */
1772 if ((want_set &&
1773 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1774 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1775 continue; /* select Opt_noFoo vs Opt_Foo */
1776 SEQ_OPTS_PRINT("%s", token2str(m->token));
1777 }
1778
1779 if (nodefs || sbi->s_resuid != EXT4_DEF_RESUID ||
1780 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1781 SEQ_OPTS_PRINT("resuid=%u", sbi->s_resuid);
1782 if (nodefs || sbi->s_resgid != EXT4_DEF_RESGID ||
1783 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1784 SEQ_OPTS_PRINT("resgid=%u", sbi->s_resgid);
1785 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1786 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1787 SEQ_OPTS_PUTS("errors=remount-ro");
1788 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1789 SEQ_OPTS_PUTS("errors=continue");
1790 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1791 SEQ_OPTS_PUTS("errors=panic");
1792 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1793 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1794 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1795 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1796 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1797 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1798 if (sb->s_flags & MS_I_VERSION)
1799 SEQ_OPTS_PUTS("i_version");
1800 if (nodefs || sbi->s_stripe)
1801 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1802 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1803 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1804 SEQ_OPTS_PUTS("data=journal");
1805 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1806 SEQ_OPTS_PUTS("data=ordered");
1807 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1808 SEQ_OPTS_PUTS("data=writeback");
1809 }
1810 if (nodefs ||
1811 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1812 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1813 sbi->s_inode_readahead_blks);
1814
1815 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1816 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1817 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1818
1819 ext4_show_quota_options(seq, sb);
1820 return 0;
1821 }
1822
1823 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1824 {
1825 return _ext4_show_options(seq, root->d_sb, 0);
1826 }
1827
1828 static int options_seq_show(struct seq_file *seq, void *offset)
1829 {
1830 struct super_block *sb = seq->private;
1831 int rc;
1832
1833 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1834 rc = _ext4_show_options(seq, sb, 1);
1835 seq_puts(seq, "\n");
1836 return rc;
1837 }
1838
1839 static int options_open_fs(struct inode *inode, struct file *file)
1840 {
1841 return single_open(file, options_seq_show, PDE(inode)->data);
1842 }
1843
1844 static const struct file_operations ext4_seq_options_fops = {
1845 .owner = THIS_MODULE,
1846 .open = options_open_fs,
1847 .read = seq_read,
1848 .llseek = seq_lseek,
1849 .release = single_release,
1850 };
1851
1852 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1853 int read_only)
1854 {
1855 struct ext4_sb_info *sbi = EXT4_SB(sb);
1856 int res = 0;
1857
1858 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1859 ext4_msg(sb, KERN_ERR, "revision level too high, "
1860 "forcing read-only mode");
1861 res = MS_RDONLY;
1862 }
1863 if (read_only)
1864 goto done;
1865 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1866 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1867 "running e2fsck is recommended");
1868 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1869 ext4_msg(sb, KERN_WARNING,
1870 "warning: mounting fs with errors, "
1871 "running e2fsck is recommended");
1872 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1873 le16_to_cpu(es->s_mnt_count) >=
1874 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1875 ext4_msg(sb, KERN_WARNING,
1876 "warning: maximal mount count reached, "
1877 "running e2fsck is recommended");
1878 else if (le32_to_cpu(es->s_checkinterval) &&
1879 (le32_to_cpu(es->s_lastcheck) +
1880 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1881 ext4_msg(sb, KERN_WARNING,
1882 "warning: checktime reached, "
1883 "running e2fsck is recommended");
1884 if (!sbi->s_journal)
1885 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1886 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1887 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1888 le16_add_cpu(&es->s_mnt_count, 1);
1889 es->s_mtime = cpu_to_le32(get_seconds());
1890 ext4_update_dynamic_rev(sb);
1891 if (sbi->s_journal)
1892 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1893
1894 ext4_commit_super(sb, 1);
1895 done:
1896 if (test_opt(sb, DEBUG))
1897 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1898 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1899 sb->s_blocksize,
1900 sbi->s_groups_count,
1901 EXT4_BLOCKS_PER_GROUP(sb),
1902 EXT4_INODES_PER_GROUP(sb),
1903 sbi->s_mount_opt, sbi->s_mount_opt2);
1904
1905 cleancache_init_fs(sb);
1906 return res;
1907 }
1908
1909 static int ext4_fill_flex_info(struct super_block *sb)
1910 {
1911 struct ext4_sb_info *sbi = EXT4_SB(sb);
1912 struct ext4_group_desc *gdp = NULL;
1913 ext4_group_t flex_group_count;
1914 ext4_group_t flex_group;
1915 unsigned int groups_per_flex = 0;
1916 size_t size;
1917 int i;
1918
1919 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1920 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1921 sbi->s_log_groups_per_flex = 0;
1922 return 1;
1923 }
1924 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1925
1926 /* We allocate both existing and potentially added groups */
1927 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1928 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1929 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1930 size = flex_group_count * sizeof(struct flex_groups);
1931 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
1932 if (sbi->s_flex_groups == NULL) {
1933 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1934 flex_group_count);
1935 goto failed;
1936 }
1937
1938 for (i = 0; i < sbi->s_groups_count; i++) {
1939 gdp = ext4_get_group_desc(sb, i, NULL);
1940
1941 flex_group = ext4_flex_group(sbi, i);
1942 atomic_add(ext4_free_inodes_count(sb, gdp),
1943 &sbi->s_flex_groups[flex_group].free_inodes);
1944 atomic_add(ext4_free_group_clusters(sb, gdp),
1945 &sbi->s_flex_groups[flex_group].free_clusters);
1946 atomic_add(ext4_used_dirs_count(sb, gdp),
1947 &sbi->s_flex_groups[flex_group].used_dirs);
1948 }
1949
1950 return 1;
1951 failed:
1952 return 0;
1953 }
1954
1955 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1956 struct ext4_group_desc *gdp)
1957 {
1958 int offset;
1959 __u16 crc = 0;
1960 __le32 le_group = cpu_to_le32(block_group);
1961
1962 if ((sbi->s_es->s_feature_ro_compat &
1963 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1964 /* Use new metadata_csum algorithm */
1965 __u16 old_csum;
1966 __u32 csum32;
1967
1968 old_csum = gdp->bg_checksum;
1969 gdp->bg_checksum = 0;
1970 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1971 sizeof(le_group));
1972 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1973 sbi->s_desc_size);
1974 gdp->bg_checksum = old_csum;
1975
1976 crc = csum32 & 0xFFFF;
1977 goto out;
1978 }
1979
1980 /* old crc16 code */
1981 offset = offsetof(struct ext4_group_desc, bg_checksum);
1982
1983 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1984 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1985 crc = crc16(crc, (__u8 *)gdp, offset);
1986 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1987 /* for checksum of struct ext4_group_desc do the rest...*/
1988 if ((sbi->s_es->s_feature_incompat &
1989 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1990 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1991 crc = crc16(crc, (__u8 *)gdp + offset,
1992 le16_to_cpu(sbi->s_es->s_desc_size) -
1993 offset);
1994
1995 out:
1996 return cpu_to_le16(crc);
1997 }
1998
1999 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2000 struct ext4_group_desc *gdp)
2001 {
2002 if (ext4_has_group_desc_csum(sb) &&
2003 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2004 block_group, gdp)))
2005 return 0;
2006
2007 return 1;
2008 }
2009
2010 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2011 struct ext4_group_desc *gdp)
2012 {
2013 if (!ext4_has_group_desc_csum(sb))
2014 return;
2015 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2016 }
2017
2018 /* Called at mount-time, super-block is locked */
2019 static int ext4_check_descriptors(struct super_block *sb,
2020 ext4_group_t *first_not_zeroed)
2021 {
2022 struct ext4_sb_info *sbi = EXT4_SB(sb);
2023 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2024 ext4_fsblk_t last_block;
2025 ext4_fsblk_t block_bitmap;
2026 ext4_fsblk_t inode_bitmap;
2027 ext4_fsblk_t inode_table;
2028 int flexbg_flag = 0;
2029 ext4_group_t i, grp = sbi->s_groups_count;
2030
2031 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2032 flexbg_flag = 1;
2033
2034 ext4_debug("Checking group descriptors");
2035
2036 for (i = 0; i < sbi->s_groups_count; i++) {
2037 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2038
2039 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2040 last_block = ext4_blocks_count(sbi->s_es) - 1;
2041 else
2042 last_block = first_block +
2043 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2044
2045 if ((grp == sbi->s_groups_count) &&
2046 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2047 grp = i;
2048
2049 block_bitmap = ext4_block_bitmap(sb, gdp);
2050 if (block_bitmap < first_block || block_bitmap > last_block) {
2051 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2052 "Block bitmap for group %u not in group "
2053 "(block %llu)!", i, block_bitmap);
2054 return 0;
2055 }
2056 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2057 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2058 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2059 "Inode bitmap for group %u not in group "
2060 "(block %llu)!", i, inode_bitmap);
2061 return 0;
2062 }
2063 inode_table = ext4_inode_table(sb, gdp);
2064 if (inode_table < first_block ||
2065 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2066 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2067 "Inode table for group %u not in group "
2068 "(block %llu)!", i, inode_table);
2069 return 0;
2070 }
2071 ext4_lock_group(sb, i);
2072 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2073 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2074 "Checksum for group %u failed (%u!=%u)",
2075 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2076 gdp)), le16_to_cpu(gdp->bg_checksum));
2077 if (!(sb->s_flags & MS_RDONLY)) {
2078 ext4_unlock_group(sb, i);
2079 return 0;
2080 }
2081 }
2082 ext4_unlock_group(sb, i);
2083 if (!flexbg_flag)
2084 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2085 }
2086 if (NULL != first_not_zeroed)
2087 *first_not_zeroed = grp;
2088
2089 ext4_free_blocks_count_set(sbi->s_es,
2090 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2091 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2092 return 1;
2093 }
2094
2095 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2096 * the superblock) which were deleted from all directories, but held open by
2097 * a process at the time of a crash. We walk the list and try to delete these
2098 * inodes at recovery time (only with a read-write filesystem).
2099 *
2100 * In order to keep the orphan inode chain consistent during traversal (in
2101 * case of crash during recovery), we link each inode into the superblock
2102 * orphan list_head and handle it the same way as an inode deletion during
2103 * normal operation (which journals the operations for us).
2104 *
2105 * We only do an iget() and an iput() on each inode, which is very safe if we
2106 * accidentally point at an in-use or already deleted inode. The worst that
2107 * can happen in this case is that we get a "bit already cleared" message from
2108 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2109 * e2fsck was run on this filesystem, and it must have already done the orphan
2110 * inode cleanup for us, so we can safely abort without any further action.
2111 */
2112 static void ext4_orphan_cleanup(struct super_block *sb,
2113 struct ext4_super_block *es)
2114 {
2115 unsigned int s_flags = sb->s_flags;
2116 int nr_orphans = 0, nr_truncates = 0;
2117 #ifdef CONFIG_QUOTA
2118 int i;
2119 #endif
2120 if (!es->s_last_orphan) {
2121 jbd_debug(4, "no orphan inodes to clean up\n");
2122 return;
2123 }
2124
2125 if (bdev_read_only(sb->s_bdev)) {
2126 ext4_msg(sb, KERN_ERR, "write access "
2127 "unavailable, skipping orphan cleanup");
2128 return;
2129 }
2130
2131 /* Check if feature set would not allow a r/w mount */
2132 if (!ext4_feature_set_ok(sb, 0)) {
2133 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2134 "unknown ROCOMPAT features");
2135 return;
2136 }
2137
2138 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2139 if (es->s_last_orphan)
2140 jbd_debug(1, "Errors on filesystem, "
2141 "clearing orphan list.\n");
2142 es->s_last_orphan = 0;
2143 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2144 return;
2145 }
2146
2147 if (s_flags & MS_RDONLY) {
2148 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2149 sb->s_flags &= ~MS_RDONLY;
2150 }
2151 #ifdef CONFIG_QUOTA
2152 /* Needed for iput() to work correctly and not trash data */
2153 sb->s_flags |= MS_ACTIVE;
2154 /* Turn on quotas so that they are updated correctly */
2155 for (i = 0; i < MAXQUOTAS; i++) {
2156 if (EXT4_SB(sb)->s_qf_names[i]) {
2157 int ret = ext4_quota_on_mount(sb, i);
2158 if (ret < 0)
2159 ext4_msg(sb, KERN_ERR,
2160 "Cannot turn on journaled "
2161 "quota: error %d", ret);
2162 }
2163 }
2164 #endif
2165
2166 while (es->s_last_orphan) {
2167 struct inode *inode;
2168
2169 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2170 if (IS_ERR(inode)) {
2171 es->s_last_orphan = 0;
2172 break;
2173 }
2174
2175 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2176 dquot_initialize(inode);
2177 if (inode->i_nlink) {
2178 ext4_msg(sb, KERN_DEBUG,
2179 "%s: truncating inode %lu to %lld bytes",
2180 __func__, inode->i_ino, inode->i_size);
2181 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2182 inode->i_ino, inode->i_size);
2183 ext4_truncate(inode);
2184 nr_truncates++;
2185 } else {
2186 ext4_msg(sb, KERN_DEBUG,
2187 "%s: deleting unreferenced inode %lu",
2188 __func__, inode->i_ino);
2189 jbd_debug(2, "deleting unreferenced inode %lu\n",
2190 inode->i_ino);
2191 nr_orphans++;
2192 }
2193 iput(inode); /* The delete magic happens here! */
2194 }
2195
2196 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2197
2198 if (nr_orphans)
2199 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2200 PLURAL(nr_orphans));
2201 if (nr_truncates)
2202 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2203 PLURAL(nr_truncates));
2204 #ifdef CONFIG_QUOTA
2205 /* Turn quotas off */
2206 for (i = 0; i < MAXQUOTAS; i++) {
2207 if (sb_dqopt(sb)->files[i])
2208 dquot_quota_off(sb, i);
2209 }
2210 #endif
2211 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2212 }
2213
2214 /*
2215 * Maximal extent format file size.
2216 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2217 * extent format containers, within a sector_t, and within i_blocks
2218 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2219 * so that won't be a limiting factor.
2220 *
2221 * However there is other limiting factor. We do store extents in the form
2222 * of starting block and length, hence the resulting length of the extent
2223 * covering maximum file size must fit into on-disk format containers as
2224 * well. Given that length is always by 1 unit bigger than max unit (because
2225 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2226 *
2227 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2228 */
2229 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2230 {
2231 loff_t res;
2232 loff_t upper_limit = MAX_LFS_FILESIZE;
2233
2234 /* small i_blocks in vfs inode? */
2235 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2236 /*
2237 * CONFIG_LBDAF is not enabled implies the inode
2238 * i_block represent total blocks in 512 bytes
2239 * 32 == size of vfs inode i_blocks * 8
2240 */
2241 upper_limit = (1LL << 32) - 1;
2242
2243 /* total blocks in file system block size */
2244 upper_limit >>= (blkbits - 9);
2245 upper_limit <<= blkbits;
2246 }
2247
2248 /*
2249 * 32-bit extent-start container, ee_block. We lower the maxbytes
2250 * by one fs block, so ee_len can cover the extent of maximum file
2251 * size
2252 */
2253 res = (1LL << 32) - 1;
2254 res <<= blkbits;
2255
2256 /* Sanity check against vm- & vfs- imposed limits */
2257 if (res > upper_limit)
2258 res = upper_limit;
2259
2260 return res;
2261 }
2262
2263 /*
2264 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2265 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2266 * We need to be 1 filesystem block less than the 2^48 sector limit.
2267 */
2268 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2269 {
2270 loff_t res = EXT4_NDIR_BLOCKS;
2271 int meta_blocks;
2272 loff_t upper_limit;
2273 /* This is calculated to be the largest file size for a dense, block
2274 * mapped file such that the file's total number of 512-byte sectors,
2275 * including data and all indirect blocks, does not exceed (2^48 - 1).
2276 *
2277 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2278 * number of 512-byte sectors of the file.
2279 */
2280
2281 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2282 /*
2283 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2284 * the inode i_block field represents total file blocks in
2285 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2286 */
2287 upper_limit = (1LL << 32) - 1;
2288
2289 /* total blocks in file system block size */
2290 upper_limit >>= (bits - 9);
2291
2292 } else {
2293 /*
2294 * We use 48 bit ext4_inode i_blocks
2295 * With EXT4_HUGE_FILE_FL set the i_blocks
2296 * represent total number of blocks in
2297 * file system block size
2298 */
2299 upper_limit = (1LL << 48) - 1;
2300
2301 }
2302
2303 /* indirect blocks */
2304 meta_blocks = 1;
2305 /* double indirect blocks */
2306 meta_blocks += 1 + (1LL << (bits-2));
2307 /* tripple indirect blocks */
2308 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2309
2310 upper_limit -= meta_blocks;
2311 upper_limit <<= bits;
2312
2313 res += 1LL << (bits-2);
2314 res += 1LL << (2*(bits-2));
2315 res += 1LL << (3*(bits-2));
2316 res <<= bits;
2317 if (res > upper_limit)
2318 res = upper_limit;
2319
2320 if (res > MAX_LFS_FILESIZE)
2321 res = MAX_LFS_FILESIZE;
2322
2323 return res;
2324 }
2325
2326 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2327 ext4_fsblk_t logical_sb_block, int nr)
2328 {
2329 struct ext4_sb_info *sbi = EXT4_SB(sb);
2330 ext4_group_t bg, first_meta_bg;
2331 int has_super = 0;
2332
2333 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2334
2335 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2336 nr < first_meta_bg)
2337 return logical_sb_block + nr + 1;
2338 bg = sbi->s_desc_per_block * nr;
2339 if (ext4_bg_has_super(sb, bg))
2340 has_super = 1;
2341
2342 return (has_super + ext4_group_first_block_no(sb, bg));
2343 }
2344
2345 /**
2346 * ext4_get_stripe_size: Get the stripe size.
2347 * @sbi: In memory super block info
2348 *
2349 * If we have specified it via mount option, then
2350 * use the mount option value. If the value specified at mount time is
2351 * greater than the blocks per group use the super block value.
2352 * If the super block value is greater than blocks per group return 0.
2353 * Allocator needs it be less than blocks per group.
2354 *
2355 */
2356 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2357 {
2358 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2359 unsigned long stripe_width =
2360 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2361 int ret;
2362
2363 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2364 ret = sbi->s_stripe;
2365 else if (stripe_width <= sbi->s_blocks_per_group)
2366 ret = stripe_width;
2367 else if (stride <= sbi->s_blocks_per_group)
2368 ret = stride;
2369 else
2370 ret = 0;
2371
2372 /*
2373 * If the stripe width is 1, this makes no sense and
2374 * we set it to 0 to turn off stripe handling code.
2375 */
2376 if (ret <= 1)
2377 ret = 0;
2378
2379 return ret;
2380 }
2381
2382 /* sysfs supprt */
2383
2384 struct ext4_attr {
2385 struct attribute attr;
2386 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2387 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2388 const char *, size_t);
2389 int offset;
2390 };
2391
2392 static int parse_strtoul(const char *buf,
2393 unsigned long max, unsigned long *value)
2394 {
2395 char *endp;
2396
2397 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2398 endp = skip_spaces(endp);
2399 if (*endp || *value > max)
2400 return -EINVAL;
2401
2402 return 0;
2403 }
2404
2405 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2406 struct ext4_sb_info *sbi,
2407 char *buf)
2408 {
2409 return snprintf(buf, PAGE_SIZE, "%llu\n",
2410 (s64) EXT4_C2B(sbi,
2411 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2412 }
2413
2414 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2415 struct ext4_sb_info *sbi, char *buf)
2416 {
2417 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2418
2419 if (!sb->s_bdev->bd_part)
2420 return snprintf(buf, PAGE_SIZE, "0\n");
2421 return snprintf(buf, PAGE_SIZE, "%lu\n",
2422 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2423 sbi->s_sectors_written_start) >> 1);
2424 }
2425
2426 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2427 struct ext4_sb_info *sbi, char *buf)
2428 {
2429 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2430
2431 if (!sb->s_bdev->bd_part)
2432 return snprintf(buf, PAGE_SIZE, "0\n");
2433 return snprintf(buf, PAGE_SIZE, "%llu\n",
2434 (unsigned long long)(sbi->s_kbytes_written +
2435 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2436 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2437 }
2438
2439 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2440 struct ext4_sb_info *sbi,
2441 const char *buf, size_t count)
2442 {
2443 unsigned long t;
2444
2445 if (parse_strtoul(buf, 0x40000000, &t))
2446 return -EINVAL;
2447
2448 if (t && !is_power_of_2(t))
2449 return -EINVAL;
2450
2451 sbi->s_inode_readahead_blks = t;
2452 return count;
2453 }
2454
2455 static ssize_t sbi_ui_show(struct ext4_attr *a,
2456 struct ext4_sb_info *sbi, char *buf)
2457 {
2458 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2459
2460 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2461 }
2462
2463 static ssize_t sbi_ui_store(struct ext4_attr *a,
2464 struct ext4_sb_info *sbi,
2465 const char *buf, size_t count)
2466 {
2467 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2468 unsigned long t;
2469
2470 if (parse_strtoul(buf, 0xffffffff, &t))
2471 return -EINVAL;
2472 *ui = t;
2473 return count;
2474 }
2475
2476 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2477 static struct ext4_attr ext4_attr_##_name = { \
2478 .attr = {.name = __stringify(_name), .mode = _mode }, \
2479 .show = _show, \
2480 .store = _store, \
2481 .offset = offsetof(struct ext4_sb_info, _elname), \
2482 }
2483 #define EXT4_ATTR(name, mode, show, store) \
2484 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2485
2486 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2487 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2488 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2489 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2490 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2491 #define ATTR_LIST(name) &ext4_attr_##name.attr
2492
2493 EXT4_RO_ATTR(delayed_allocation_blocks);
2494 EXT4_RO_ATTR(session_write_kbytes);
2495 EXT4_RO_ATTR(lifetime_write_kbytes);
2496 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2497 inode_readahead_blks_store, s_inode_readahead_blks);
2498 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2499 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2500 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2501 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2502 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2503 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2504 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2505 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2506
2507 static struct attribute *ext4_attrs[] = {
2508 ATTR_LIST(delayed_allocation_blocks),
2509 ATTR_LIST(session_write_kbytes),
2510 ATTR_LIST(lifetime_write_kbytes),
2511 ATTR_LIST(inode_readahead_blks),
2512 ATTR_LIST(inode_goal),
2513 ATTR_LIST(mb_stats),
2514 ATTR_LIST(mb_max_to_scan),
2515 ATTR_LIST(mb_min_to_scan),
2516 ATTR_LIST(mb_order2_req),
2517 ATTR_LIST(mb_stream_req),
2518 ATTR_LIST(mb_group_prealloc),
2519 ATTR_LIST(max_writeback_mb_bump),
2520 NULL,
2521 };
2522
2523 /* Features this copy of ext4 supports */
2524 EXT4_INFO_ATTR(lazy_itable_init);
2525 EXT4_INFO_ATTR(batched_discard);
2526
2527 static struct attribute *ext4_feat_attrs[] = {
2528 ATTR_LIST(lazy_itable_init),
2529 ATTR_LIST(batched_discard),
2530 NULL,
2531 };
2532
2533 static ssize_t ext4_attr_show(struct kobject *kobj,
2534 struct attribute *attr, char *buf)
2535 {
2536 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2537 s_kobj);
2538 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2539
2540 return a->show ? a->show(a, sbi, buf) : 0;
2541 }
2542
2543 static ssize_t ext4_attr_store(struct kobject *kobj,
2544 struct attribute *attr,
2545 const char *buf, size_t len)
2546 {
2547 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2548 s_kobj);
2549 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2550
2551 return a->store ? a->store(a, sbi, buf, len) : 0;
2552 }
2553
2554 static void ext4_sb_release(struct kobject *kobj)
2555 {
2556 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2557 s_kobj);
2558 complete(&sbi->s_kobj_unregister);
2559 }
2560
2561 static const struct sysfs_ops ext4_attr_ops = {
2562 .show = ext4_attr_show,
2563 .store = ext4_attr_store,
2564 };
2565
2566 static struct kobj_type ext4_ktype = {
2567 .default_attrs = ext4_attrs,
2568 .sysfs_ops = &ext4_attr_ops,
2569 .release = ext4_sb_release,
2570 };
2571
2572 static void ext4_feat_release(struct kobject *kobj)
2573 {
2574 complete(&ext4_feat->f_kobj_unregister);
2575 }
2576
2577 static struct kobj_type ext4_feat_ktype = {
2578 .default_attrs = ext4_feat_attrs,
2579 .sysfs_ops = &ext4_attr_ops,
2580 .release = ext4_feat_release,
2581 };
2582
2583 /*
2584 * Check whether this filesystem can be mounted based on
2585 * the features present and the RDONLY/RDWR mount requested.
2586 * Returns 1 if this filesystem can be mounted as requested,
2587 * 0 if it cannot be.
2588 */
2589 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2590 {
2591 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2592 ext4_msg(sb, KERN_ERR,
2593 "Couldn't mount because of "
2594 "unsupported optional features (%x)",
2595 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2596 ~EXT4_FEATURE_INCOMPAT_SUPP));
2597 return 0;
2598 }
2599
2600 if (readonly)
2601 return 1;
2602
2603 /* Check that feature set is OK for a read-write mount */
2604 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2605 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2606 "unsupported optional features (%x)",
2607 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2608 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2609 return 0;
2610 }
2611 /*
2612 * Large file size enabled file system can only be mounted
2613 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2614 */
2615 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2616 if (sizeof(blkcnt_t) < sizeof(u64)) {
2617 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2618 "cannot be mounted RDWR without "
2619 "CONFIG_LBDAF");
2620 return 0;
2621 }
2622 }
2623 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2624 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2625 ext4_msg(sb, KERN_ERR,
2626 "Can't support bigalloc feature without "
2627 "extents feature\n");
2628 return 0;
2629 }
2630 return 1;
2631 }
2632
2633 /*
2634 * This function is called once a day if we have errors logged
2635 * on the file system
2636 */
2637 static void print_daily_error_info(unsigned long arg)
2638 {
2639 struct super_block *sb = (struct super_block *) arg;
2640 struct ext4_sb_info *sbi;
2641 struct ext4_super_block *es;
2642
2643 sbi = EXT4_SB(sb);
2644 es = sbi->s_es;
2645
2646 if (es->s_error_count)
2647 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2648 le32_to_cpu(es->s_error_count));
2649 if (es->s_first_error_time) {
2650 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2651 sb->s_id, le32_to_cpu(es->s_first_error_time),
2652 (int) sizeof(es->s_first_error_func),
2653 es->s_first_error_func,
2654 le32_to_cpu(es->s_first_error_line));
2655 if (es->s_first_error_ino)
2656 printk(": inode %u",
2657 le32_to_cpu(es->s_first_error_ino));
2658 if (es->s_first_error_block)
2659 printk(": block %llu", (unsigned long long)
2660 le64_to_cpu(es->s_first_error_block));
2661 printk("\n");
2662 }
2663 if (es->s_last_error_time) {
2664 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2665 sb->s_id, le32_to_cpu(es->s_last_error_time),
2666 (int) sizeof(es->s_last_error_func),
2667 es->s_last_error_func,
2668 le32_to_cpu(es->s_last_error_line));
2669 if (es->s_last_error_ino)
2670 printk(": inode %u",
2671 le32_to_cpu(es->s_last_error_ino));
2672 if (es->s_last_error_block)
2673 printk(": block %llu", (unsigned long long)
2674 le64_to_cpu(es->s_last_error_block));
2675 printk("\n");
2676 }
2677 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2678 }
2679
2680 /* Find next suitable group and run ext4_init_inode_table */
2681 static int ext4_run_li_request(struct ext4_li_request *elr)
2682 {
2683 struct ext4_group_desc *gdp = NULL;
2684 ext4_group_t group, ngroups;
2685 struct super_block *sb;
2686 unsigned long timeout = 0;
2687 int ret = 0;
2688
2689 sb = elr->lr_super;
2690 ngroups = EXT4_SB(sb)->s_groups_count;
2691
2692 for (group = elr->lr_next_group; group < ngroups; group++) {
2693 gdp = ext4_get_group_desc(sb, group, NULL);
2694 if (!gdp) {
2695 ret = 1;
2696 break;
2697 }
2698
2699 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2700 break;
2701 }
2702
2703 if (group == ngroups)
2704 ret = 1;
2705
2706 if (!ret) {
2707 timeout = jiffies;
2708 ret = ext4_init_inode_table(sb, group,
2709 elr->lr_timeout ? 0 : 1);
2710 if (elr->lr_timeout == 0) {
2711 timeout = (jiffies - timeout) *
2712 elr->lr_sbi->s_li_wait_mult;
2713 elr->lr_timeout = timeout;
2714 }
2715 elr->lr_next_sched = jiffies + elr->lr_timeout;
2716 elr->lr_next_group = group + 1;
2717 }
2718
2719 return ret;
2720 }
2721
2722 /*
2723 * Remove lr_request from the list_request and free the
2724 * request structure. Should be called with li_list_mtx held
2725 */
2726 static void ext4_remove_li_request(struct ext4_li_request *elr)
2727 {
2728 struct ext4_sb_info *sbi;
2729
2730 if (!elr)
2731 return;
2732
2733 sbi = elr->lr_sbi;
2734
2735 list_del(&elr->lr_request);
2736 sbi->s_li_request = NULL;
2737 kfree(elr);
2738 }
2739
2740 static void ext4_unregister_li_request(struct super_block *sb)
2741 {
2742 mutex_lock(&ext4_li_mtx);
2743 if (!ext4_li_info) {
2744 mutex_unlock(&ext4_li_mtx);
2745 return;
2746 }
2747
2748 mutex_lock(&ext4_li_info->li_list_mtx);
2749 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2750 mutex_unlock(&ext4_li_info->li_list_mtx);
2751 mutex_unlock(&ext4_li_mtx);
2752 }
2753
2754 static struct task_struct *ext4_lazyinit_task;
2755
2756 /*
2757 * This is the function where ext4lazyinit thread lives. It walks
2758 * through the request list searching for next scheduled filesystem.
2759 * When such a fs is found, run the lazy initialization request
2760 * (ext4_rn_li_request) and keep track of the time spend in this
2761 * function. Based on that time we compute next schedule time of
2762 * the request. When walking through the list is complete, compute
2763 * next waking time and put itself into sleep.
2764 */
2765 static int ext4_lazyinit_thread(void *arg)
2766 {
2767 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2768 struct list_head *pos, *n;
2769 struct ext4_li_request *elr;
2770 unsigned long next_wakeup, cur;
2771
2772 BUG_ON(NULL == eli);
2773
2774 cont_thread:
2775 while (true) {
2776 next_wakeup = MAX_JIFFY_OFFSET;
2777
2778 mutex_lock(&eli->li_list_mtx);
2779 if (list_empty(&eli->li_request_list)) {
2780 mutex_unlock(&eli->li_list_mtx);
2781 goto exit_thread;
2782 }
2783
2784 list_for_each_safe(pos, n, &eli->li_request_list) {
2785 elr = list_entry(pos, struct ext4_li_request,
2786 lr_request);
2787
2788 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2789 if (ext4_run_li_request(elr) != 0) {
2790 /* error, remove the lazy_init job */
2791 ext4_remove_li_request(elr);
2792 continue;
2793 }
2794 }
2795
2796 if (time_before(elr->lr_next_sched, next_wakeup))
2797 next_wakeup = elr->lr_next_sched;
2798 }
2799 mutex_unlock(&eli->li_list_mtx);
2800
2801 try_to_freeze();
2802
2803 cur = jiffies;
2804 if ((time_after_eq(cur, next_wakeup)) ||
2805 (MAX_JIFFY_OFFSET == next_wakeup)) {
2806 cond_resched();
2807 continue;
2808 }
2809
2810 schedule_timeout_interruptible(next_wakeup - cur);
2811
2812 if (kthread_should_stop()) {
2813 ext4_clear_request_list();
2814 goto exit_thread;
2815 }
2816 }
2817
2818 exit_thread:
2819 /*
2820 * It looks like the request list is empty, but we need
2821 * to check it under the li_list_mtx lock, to prevent any
2822 * additions into it, and of course we should lock ext4_li_mtx
2823 * to atomically free the list and ext4_li_info, because at
2824 * this point another ext4 filesystem could be registering
2825 * new one.
2826 */
2827 mutex_lock(&ext4_li_mtx);
2828 mutex_lock(&eli->li_list_mtx);
2829 if (!list_empty(&eli->li_request_list)) {
2830 mutex_unlock(&eli->li_list_mtx);
2831 mutex_unlock(&ext4_li_mtx);
2832 goto cont_thread;
2833 }
2834 mutex_unlock(&eli->li_list_mtx);
2835 kfree(ext4_li_info);
2836 ext4_li_info = NULL;
2837 mutex_unlock(&ext4_li_mtx);
2838
2839 return 0;
2840 }
2841
2842 static void ext4_clear_request_list(void)
2843 {
2844 struct list_head *pos, *n;
2845 struct ext4_li_request *elr;
2846
2847 mutex_lock(&ext4_li_info->li_list_mtx);
2848 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2849 elr = list_entry(pos, struct ext4_li_request,
2850 lr_request);
2851 ext4_remove_li_request(elr);
2852 }
2853 mutex_unlock(&ext4_li_info->li_list_mtx);
2854 }
2855
2856 static int ext4_run_lazyinit_thread(void)
2857 {
2858 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2859 ext4_li_info, "ext4lazyinit");
2860 if (IS_ERR(ext4_lazyinit_task)) {
2861 int err = PTR_ERR(ext4_lazyinit_task);
2862 ext4_clear_request_list();
2863 kfree(ext4_li_info);
2864 ext4_li_info = NULL;
2865 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2866 "initialization thread\n",
2867 err);
2868 return err;
2869 }
2870 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2871 return 0;
2872 }
2873
2874 /*
2875 * Check whether it make sense to run itable init. thread or not.
2876 * If there is at least one uninitialized inode table, return
2877 * corresponding group number, else the loop goes through all
2878 * groups and return total number of groups.
2879 */
2880 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2881 {
2882 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2883 struct ext4_group_desc *gdp = NULL;
2884
2885 for (group = 0; group < ngroups; group++) {
2886 gdp = ext4_get_group_desc(sb, group, NULL);
2887 if (!gdp)
2888 continue;
2889
2890 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2891 break;
2892 }
2893
2894 return group;
2895 }
2896
2897 static int ext4_li_info_new(void)
2898 {
2899 struct ext4_lazy_init *eli = NULL;
2900
2901 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2902 if (!eli)
2903 return -ENOMEM;
2904
2905 INIT_LIST_HEAD(&eli->li_request_list);
2906 mutex_init(&eli->li_list_mtx);
2907
2908 eli->li_state |= EXT4_LAZYINIT_QUIT;
2909
2910 ext4_li_info = eli;
2911
2912 return 0;
2913 }
2914
2915 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2916 ext4_group_t start)
2917 {
2918 struct ext4_sb_info *sbi = EXT4_SB(sb);
2919 struct ext4_li_request *elr;
2920 unsigned long rnd;
2921
2922 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2923 if (!elr)
2924 return NULL;
2925
2926 elr->lr_super = sb;
2927 elr->lr_sbi = sbi;
2928 elr->lr_next_group = start;
2929
2930 /*
2931 * Randomize first schedule time of the request to
2932 * spread the inode table initialization requests
2933 * better.
2934 */
2935 get_random_bytes(&rnd, sizeof(rnd));
2936 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2937 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2938
2939 return elr;
2940 }
2941
2942 static int ext4_register_li_request(struct super_block *sb,
2943 ext4_group_t first_not_zeroed)
2944 {
2945 struct ext4_sb_info *sbi = EXT4_SB(sb);
2946 struct ext4_li_request *elr;
2947 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2948 int ret = 0;
2949
2950 if (sbi->s_li_request != NULL) {
2951 /*
2952 * Reset timeout so it can be computed again, because
2953 * s_li_wait_mult might have changed.
2954 */
2955 sbi->s_li_request->lr_timeout = 0;
2956 return 0;
2957 }
2958
2959 if (first_not_zeroed == ngroups ||
2960 (sb->s_flags & MS_RDONLY) ||
2961 !test_opt(sb, INIT_INODE_TABLE))
2962 return 0;
2963
2964 elr = ext4_li_request_new(sb, first_not_zeroed);
2965 if (!elr)
2966 return -ENOMEM;
2967
2968 mutex_lock(&ext4_li_mtx);
2969
2970 if (NULL == ext4_li_info) {
2971 ret = ext4_li_info_new();
2972 if (ret)
2973 goto out;
2974 }
2975
2976 mutex_lock(&ext4_li_info->li_list_mtx);
2977 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2978 mutex_unlock(&ext4_li_info->li_list_mtx);
2979
2980 sbi->s_li_request = elr;
2981 /*
2982 * set elr to NULL here since it has been inserted to
2983 * the request_list and the removal and free of it is
2984 * handled by ext4_clear_request_list from now on.
2985 */
2986 elr = NULL;
2987
2988 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2989 ret = ext4_run_lazyinit_thread();
2990 if (ret)
2991 goto out;
2992 }
2993 out:
2994 mutex_unlock(&ext4_li_mtx);
2995 if (ret)
2996 kfree(elr);
2997 return ret;
2998 }
2999
3000 /*
3001 * We do not need to lock anything since this is called on
3002 * module unload.
3003 */
3004 static void ext4_destroy_lazyinit_thread(void)
3005 {
3006 /*
3007 * If thread exited earlier
3008 * there's nothing to be done.
3009 */
3010 if (!ext4_li_info || !ext4_lazyinit_task)
3011 return;
3012
3013 kthread_stop(ext4_lazyinit_task);
3014 }
3015
3016 static int set_journal_csum_feature_set(struct super_block *sb)
3017 {
3018 int ret = 1;
3019 int compat, incompat;
3020 struct ext4_sb_info *sbi = EXT4_SB(sb);
3021
3022 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3023 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3024 /* journal checksum v2 */
3025 compat = 0;
3026 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3027 } else {
3028 /* journal checksum v1 */
3029 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3030 incompat = 0;
3031 }
3032
3033 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3034 ret = jbd2_journal_set_features(sbi->s_journal,
3035 compat, 0,
3036 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3037 incompat);
3038 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3039 ret = jbd2_journal_set_features(sbi->s_journal,
3040 compat, 0,
3041 incompat);
3042 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3043 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3044 } else {
3045 jbd2_journal_clear_features(sbi->s_journal,
3046 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3047 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3048 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3049 }
3050
3051 return ret;
3052 }
3053
3054 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3055 {
3056 char *orig_data = kstrdup(data, GFP_KERNEL);
3057 struct buffer_head *bh;
3058 struct ext4_super_block *es = NULL;
3059 struct ext4_sb_info *sbi;
3060 ext4_fsblk_t block;
3061 ext4_fsblk_t sb_block = get_sb_block(&data);
3062 ext4_fsblk_t logical_sb_block;
3063 unsigned long offset = 0;
3064 unsigned long journal_devnum = 0;
3065 unsigned long def_mount_opts;
3066 struct inode *root;
3067 char *cp;
3068 const char *descr;
3069 int ret = -ENOMEM;
3070 int blocksize, clustersize;
3071 unsigned int db_count;
3072 unsigned int i;
3073 int needs_recovery, has_huge_files, has_bigalloc;
3074 __u64 blocks_count;
3075 int err;
3076 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3077 ext4_group_t first_not_zeroed;
3078
3079 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3080 if (!sbi)
3081 goto out_free_orig;
3082
3083 sbi->s_blockgroup_lock =
3084 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3085 if (!sbi->s_blockgroup_lock) {
3086 kfree(sbi);
3087 goto out_free_orig;
3088 }
3089 sb->s_fs_info = sbi;
3090 sbi->s_mount_opt = 0;
3091 sbi->s_resuid = EXT4_DEF_RESUID;
3092 sbi->s_resgid = EXT4_DEF_RESGID;
3093 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3094 sbi->s_sb_block = sb_block;
3095 if (sb->s_bdev->bd_part)
3096 sbi->s_sectors_written_start =
3097 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3098
3099 /* Cleanup superblock name */
3100 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3101 *cp = '!';
3102
3103 ret = -EINVAL;
3104 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3105 if (!blocksize) {
3106 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3107 goto out_fail;
3108 }
3109
3110 /*
3111 * The ext4 superblock will not be buffer aligned for other than 1kB
3112 * block sizes. We need to calculate the offset from buffer start.
3113 */
3114 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3115 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3116 offset = do_div(logical_sb_block, blocksize);
3117 } else {
3118 logical_sb_block = sb_block;
3119 }
3120
3121 if (!(bh = sb_bread(sb, logical_sb_block))) {
3122 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3123 goto out_fail;
3124 }
3125 /*
3126 * Note: s_es must be initialized as soon as possible because
3127 * some ext4 macro-instructions depend on its value
3128 */
3129 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3130 sbi->s_es = es;
3131 sb->s_magic = le16_to_cpu(es->s_magic);
3132 if (sb->s_magic != EXT4_SUPER_MAGIC)
3133 goto cantfind_ext4;
3134 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3135
3136 /* Warn if metadata_csum and gdt_csum are both set. */
3137 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3138 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3139 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3140 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3141 "redundant flags; please run fsck.");
3142
3143 /* Check for a known checksum algorithm */
3144 if (!ext4_verify_csum_type(sb, es)) {
3145 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3146 "unknown checksum algorithm.");
3147 silent = 1;
3148 goto cantfind_ext4;
3149 }
3150
3151 /* Load the checksum driver */
3152 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3153 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3154 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3155 if (IS_ERR(sbi->s_chksum_driver)) {
3156 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3157 ret = PTR_ERR(sbi->s_chksum_driver);
3158 sbi->s_chksum_driver = NULL;
3159 goto failed_mount;
3160 }
3161 }
3162
3163 /* Check superblock checksum */
3164 if (!ext4_superblock_csum_verify(sb, es)) {
3165 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3166 "invalid superblock checksum. Run e2fsck?");
3167 silent = 1;
3168 goto cantfind_ext4;
3169 }
3170
3171 /* Precompute checksum seed for all metadata */
3172 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3173 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3174 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3175 sizeof(es->s_uuid));
3176
3177 /* Set defaults before we parse the mount options */
3178 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3179 set_opt(sb, INIT_INODE_TABLE);
3180 if (def_mount_opts & EXT4_DEFM_DEBUG)
3181 set_opt(sb, DEBUG);
3182 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3183 set_opt(sb, GRPID);
3184 if (def_mount_opts & EXT4_DEFM_UID16)
3185 set_opt(sb, NO_UID32);
3186 /* xattr user namespace & acls are now defaulted on */
3187 #ifdef CONFIG_EXT4_FS_XATTR
3188 set_opt(sb, XATTR_USER);
3189 #endif
3190 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3191 set_opt(sb, POSIX_ACL);
3192 #endif
3193 set_opt(sb, MBLK_IO_SUBMIT);
3194 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3195 set_opt(sb, JOURNAL_DATA);
3196 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3197 set_opt(sb, ORDERED_DATA);
3198 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3199 set_opt(sb, WRITEBACK_DATA);
3200
3201 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3202 set_opt(sb, ERRORS_PANIC);
3203 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3204 set_opt(sb, ERRORS_CONT);
3205 else
3206 set_opt(sb, ERRORS_RO);
3207 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3208 set_opt(sb, BLOCK_VALIDITY);
3209 if (def_mount_opts & EXT4_DEFM_DISCARD)
3210 set_opt(sb, DISCARD);
3211
3212 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3213 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3214 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3215 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3216 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3217
3218 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3219 set_opt(sb, BARRIER);
3220
3221 /*
3222 * enable delayed allocation by default
3223 * Use -o nodelalloc to turn it off
3224 */
3225 if (!IS_EXT3_SB(sb) &&
3226 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3227 set_opt(sb, DELALLOC);
3228
3229 /*
3230 * set default s_li_wait_mult for lazyinit, for the case there is
3231 * no mount option specified.
3232 */
3233 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3234
3235 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3236 &journal_devnum, &journal_ioprio, 0)) {
3237 ext4_msg(sb, KERN_WARNING,
3238 "failed to parse options in superblock: %s",
3239 sbi->s_es->s_mount_opts);
3240 }
3241 sbi->s_def_mount_opt = sbi->s_mount_opt;
3242 if (!parse_options((char *) data, sb, &journal_devnum,
3243 &journal_ioprio, 0))
3244 goto failed_mount;
3245
3246 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3247 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3248 "with data=journal disables delayed "
3249 "allocation and O_DIRECT support!\n");
3250 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3251 ext4_msg(sb, KERN_ERR, "can't mount with "
3252 "both data=journal and delalloc");
3253 goto failed_mount;
3254 }
3255 if (test_opt(sb, DIOREAD_NOLOCK)) {
3256 ext4_msg(sb, KERN_ERR, "can't mount with "
3257 "both data=journal and delalloc");
3258 goto failed_mount;
3259 }
3260 if (test_opt(sb, DELALLOC))
3261 clear_opt(sb, DELALLOC);
3262 }
3263
3264 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3265 if (test_opt(sb, DIOREAD_NOLOCK)) {
3266 if (blocksize < PAGE_SIZE) {
3267 ext4_msg(sb, KERN_ERR, "can't mount with "
3268 "dioread_nolock if block size != PAGE_SIZE");
3269 goto failed_mount;
3270 }
3271 }
3272
3273 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3274 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3275
3276 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3277 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3278 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3279 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3280 ext4_msg(sb, KERN_WARNING,
3281 "feature flags set on rev 0 fs, "
3282 "running e2fsck is recommended");
3283
3284 if (IS_EXT2_SB(sb)) {
3285 if (ext2_feature_set_ok(sb))
3286 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3287 "using the ext4 subsystem");
3288 else {
3289 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3290 "to feature incompatibilities");
3291 goto failed_mount;
3292 }
3293 }
3294
3295 if (IS_EXT3_SB(sb)) {
3296 if (ext3_feature_set_ok(sb))
3297 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3298 "using the ext4 subsystem");
3299 else {
3300 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3301 "to feature incompatibilities");
3302 goto failed_mount;
3303 }
3304 }
3305
3306 /*
3307 * Check feature flags regardless of the revision level, since we
3308 * previously didn't change the revision level when setting the flags,
3309 * so there is a chance incompat flags are set on a rev 0 filesystem.
3310 */
3311 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3312 goto failed_mount;
3313
3314 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3315 blocksize > EXT4_MAX_BLOCK_SIZE) {
3316 ext4_msg(sb, KERN_ERR,
3317 "Unsupported filesystem blocksize %d", blocksize);
3318 goto failed_mount;
3319 }
3320
3321 if (sb->s_blocksize != blocksize) {
3322 /* Validate the filesystem blocksize */
3323 if (!sb_set_blocksize(sb, blocksize)) {
3324 ext4_msg(sb, KERN_ERR, "bad block size %d",
3325 blocksize);
3326 goto failed_mount;
3327 }
3328
3329 brelse(bh);
3330 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3331 offset = do_div(logical_sb_block, blocksize);
3332 bh = sb_bread(sb, logical_sb_block);
3333 if (!bh) {
3334 ext4_msg(sb, KERN_ERR,
3335 "Can't read superblock on 2nd try");
3336 goto failed_mount;
3337 }
3338 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3339 sbi->s_es = es;
3340 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3341 ext4_msg(sb, KERN_ERR,
3342 "Magic mismatch, very weird!");
3343 goto failed_mount;
3344 }
3345 }
3346
3347 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3348 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3349 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3350 has_huge_files);
3351 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3352
3353 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3354 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3355 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3356 } else {
3357 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3358 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3359 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3360 (!is_power_of_2(sbi->s_inode_size)) ||
3361 (sbi->s_inode_size > blocksize)) {
3362 ext4_msg(sb, KERN_ERR,
3363 "unsupported inode size: %d",
3364 sbi->s_inode_size);
3365 goto failed_mount;
3366 }
3367 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3368 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3369 }
3370
3371 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3372 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3373 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3374 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3375 !is_power_of_2(sbi->s_desc_size)) {
3376 ext4_msg(sb, KERN_ERR,
3377 "unsupported descriptor size %lu",
3378 sbi->s_desc_size);
3379 goto failed_mount;
3380 }
3381 } else
3382 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3383
3384 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3385 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3386 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3387 goto cantfind_ext4;
3388
3389 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3390 if (sbi->s_inodes_per_block == 0)
3391 goto cantfind_ext4;
3392 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3393 sbi->s_inodes_per_block;
3394 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3395 sbi->s_sbh = bh;
3396 sbi->s_mount_state = le16_to_cpu(es->s_state);
3397 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3398 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3399
3400 for (i = 0; i < 4; i++)
3401 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3402 sbi->s_def_hash_version = es->s_def_hash_version;
3403 i = le32_to_cpu(es->s_flags);
3404 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3405 sbi->s_hash_unsigned = 3;
3406 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3407 #ifdef __CHAR_UNSIGNED__
3408 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3409 sbi->s_hash_unsigned = 3;
3410 #else
3411 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3412 #endif
3413 }
3414
3415 /* Handle clustersize */
3416 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3417 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3418 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3419 if (has_bigalloc) {
3420 if (clustersize < blocksize) {
3421 ext4_msg(sb, KERN_ERR,
3422 "cluster size (%d) smaller than "
3423 "block size (%d)", clustersize, blocksize);
3424 goto failed_mount;
3425 }
3426 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3427 le32_to_cpu(es->s_log_block_size);
3428 sbi->s_clusters_per_group =
3429 le32_to_cpu(es->s_clusters_per_group);
3430 if (sbi->s_clusters_per_group > blocksize * 8) {
3431 ext4_msg(sb, KERN_ERR,
3432 "#clusters per group too big: %lu",
3433 sbi->s_clusters_per_group);
3434 goto failed_mount;
3435 }
3436 if (sbi->s_blocks_per_group !=
3437 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3438 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3439 "clusters per group (%lu) inconsistent",
3440 sbi->s_blocks_per_group,
3441 sbi->s_clusters_per_group);
3442 goto failed_mount;
3443 }
3444 } else {
3445 if (clustersize != blocksize) {
3446 ext4_warning(sb, "fragment/cluster size (%d) != "
3447 "block size (%d)", clustersize,
3448 blocksize);
3449 clustersize = blocksize;
3450 }
3451 if (sbi->s_blocks_per_group > blocksize * 8) {
3452 ext4_msg(sb, KERN_ERR,
3453 "#blocks per group too big: %lu",
3454 sbi->s_blocks_per_group);
3455 goto failed_mount;
3456 }
3457 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3458 sbi->s_cluster_bits = 0;
3459 }
3460 sbi->s_cluster_ratio = clustersize / blocksize;
3461
3462 if (sbi->s_inodes_per_group > blocksize * 8) {
3463 ext4_msg(sb, KERN_ERR,
3464 "#inodes per group too big: %lu",
3465 sbi->s_inodes_per_group);
3466 goto failed_mount;
3467 }
3468
3469 /*
3470 * Test whether we have more sectors than will fit in sector_t,
3471 * and whether the max offset is addressable by the page cache.
3472 */
3473 err = generic_check_addressable(sb->s_blocksize_bits,
3474 ext4_blocks_count(es));
3475 if (err) {
3476 ext4_msg(sb, KERN_ERR, "filesystem"
3477 " too large to mount safely on this system");
3478 if (sizeof(sector_t) < 8)
3479 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3480 ret = err;
3481 goto failed_mount;
3482 }
3483
3484 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3485 goto cantfind_ext4;
3486
3487 /* check blocks count against device size */
3488 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3489 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3490 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3491 "exceeds size of device (%llu blocks)",
3492 ext4_blocks_count(es), blocks_count);
3493 goto failed_mount;
3494 }
3495
3496 /*
3497 * It makes no sense for the first data block to be beyond the end
3498 * of the filesystem.
3499 */
3500 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3501 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3502 "block %u is beyond end of filesystem (%llu)",
3503 le32_to_cpu(es->s_first_data_block),
3504 ext4_blocks_count(es));
3505 goto failed_mount;
3506 }
3507 blocks_count = (ext4_blocks_count(es) -
3508 le32_to_cpu(es->s_first_data_block) +
3509 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3510 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3511 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3512 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3513 "(block count %llu, first data block %u, "
3514 "blocks per group %lu)", sbi->s_groups_count,
3515 ext4_blocks_count(es),
3516 le32_to_cpu(es->s_first_data_block),
3517 EXT4_BLOCKS_PER_GROUP(sb));
3518 goto failed_mount;
3519 }
3520 sbi->s_groups_count = blocks_count;
3521 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3522 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3523 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3524 EXT4_DESC_PER_BLOCK(sb);
3525 sbi->s_group_desc = ext4_kvmalloc(db_count *
3526 sizeof(struct buffer_head *),
3527 GFP_KERNEL);
3528 if (sbi->s_group_desc == NULL) {
3529 ext4_msg(sb, KERN_ERR, "not enough memory");
3530 goto failed_mount;
3531 }
3532
3533 if (ext4_proc_root)
3534 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3535
3536 if (sbi->s_proc)
3537 proc_create_data("options", S_IRUGO, sbi->s_proc,
3538 &ext4_seq_options_fops, sb);
3539
3540 bgl_lock_init(sbi->s_blockgroup_lock);
3541
3542 for (i = 0; i < db_count; i++) {
3543 block = descriptor_loc(sb, logical_sb_block, i);
3544 sbi->s_group_desc[i] = sb_bread(sb, block);
3545 if (!sbi->s_group_desc[i]) {
3546 ext4_msg(sb, KERN_ERR,
3547 "can't read group descriptor %d", i);
3548 db_count = i;
3549 goto failed_mount2;
3550 }
3551 }
3552 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3553 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3554 goto failed_mount2;
3555 }
3556 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3557 if (!ext4_fill_flex_info(sb)) {
3558 ext4_msg(sb, KERN_ERR,
3559 "unable to initialize "
3560 "flex_bg meta info!");
3561 goto failed_mount2;
3562 }
3563
3564 sbi->s_gdb_count = db_count;
3565 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3566 spin_lock_init(&sbi->s_next_gen_lock);
3567
3568 init_timer(&sbi->s_err_report);
3569 sbi->s_err_report.function = print_daily_error_info;
3570 sbi->s_err_report.data = (unsigned long) sb;
3571
3572 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3573 ext4_count_free_clusters(sb));
3574 if (!err) {
3575 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3576 ext4_count_free_inodes(sb));
3577 }
3578 if (!err) {
3579 err = percpu_counter_init(&sbi->s_dirs_counter,
3580 ext4_count_dirs(sb));
3581 }
3582 if (!err) {
3583 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3584 }
3585 if (err) {
3586 ext4_msg(sb, KERN_ERR, "insufficient memory");
3587 goto failed_mount3;
3588 }
3589
3590 sbi->s_stripe = ext4_get_stripe_size(sbi);
3591 sbi->s_max_writeback_mb_bump = 128;
3592
3593 /*
3594 * set up enough so that it can read an inode
3595 */
3596 if (!test_opt(sb, NOLOAD) &&
3597 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3598 sb->s_op = &ext4_sops;
3599 else
3600 sb->s_op = &ext4_nojournal_sops;
3601 sb->s_export_op = &ext4_export_ops;
3602 sb->s_xattr = ext4_xattr_handlers;
3603 #ifdef CONFIG_QUOTA
3604 sb->s_qcop = &ext4_qctl_operations;
3605 sb->dq_op = &ext4_quota_operations;
3606 #endif
3607 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3608
3609 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3610 mutex_init(&sbi->s_orphan_lock);
3611 sbi->s_resize_flags = 0;
3612
3613 sb->s_root = NULL;
3614
3615 needs_recovery = (es->s_last_orphan != 0 ||
3616 EXT4_HAS_INCOMPAT_FEATURE(sb,
3617 EXT4_FEATURE_INCOMPAT_RECOVER));
3618
3619 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3620 !(sb->s_flags & MS_RDONLY))
3621 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3622 goto failed_mount3;
3623
3624 /*
3625 * The first inode we look at is the journal inode. Don't try
3626 * root first: it may be modified in the journal!
3627 */
3628 if (!test_opt(sb, NOLOAD) &&
3629 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3630 if (ext4_load_journal(sb, es, journal_devnum))
3631 goto failed_mount3;
3632 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3633 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3634 ext4_msg(sb, KERN_ERR, "required journal recovery "
3635 "suppressed and not mounted read-only");
3636 goto failed_mount_wq;
3637 } else {
3638 clear_opt(sb, DATA_FLAGS);
3639 sbi->s_journal = NULL;
3640 needs_recovery = 0;
3641 goto no_journal;
3642 }
3643
3644 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3645 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3646 JBD2_FEATURE_INCOMPAT_64BIT)) {
3647 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3648 goto failed_mount_wq;
3649 }
3650
3651 if (!set_journal_csum_feature_set(sb)) {
3652 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3653 "feature set");
3654 goto failed_mount_wq;
3655 }
3656
3657 /* We have now updated the journal if required, so we can
3658 * validate the data journaling mode. */
3659 switch (test_opt(sb, DATA_FLAGS)) {
3660 case 0:
3661 /* No mode set, assume a default based on the journal
3662 * capabilities: ORDERED_DATA if the journal can
3663 * cope, else JOURNAL_DATA
3664 */
3665 if (jbd2_journal_check_available_features
3666 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3667 set_opt(sb, ORDERED_DATA);
3668 else
3669 set_opt(sb, JOURNAL_DATA);
3670 break;
3671
3672 case EXT4_MOUNT_ORDERED_DATA:
3673 case EXT4_MOUNT_WRITEBACK_DATA:
3674 if (!jbd2_journal_check_available_features
3675 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3676 ext4_msg(sb, KERN_ERR, "Journal does not support "
3677 "requested data journaling mode");
3678 goto failed_mount_wq;
3679 }
3680 default:
3681 break;
3682 }
3683 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3684
3685 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3686
3687 /*
3688 * The journal may have updated the bg summary counts, so we
3689 * need to update the global counters.
3690 */
3691 percpu_counter_set(&sbi->s_freeclusters_counter,
3692 ext4_count_free_clusters(sb));
3693 percpu_counter_set(&sbi->s_freeinodes_counter,
3694 ext4_count_free_inodes(sb));
3695 percpu_counter_set(&sbi->s_dirs_counter,
3696 ext4_count_dirs(sb));
3697 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3698
3699 no_journal:
3700 /*
3701 * The maximum number of concurrent works can be high and
3702 * concurrency isn't really necessary. Limit it to 1.
3703 */
3704 EXT4_SB(sb)->dio_unwritten_wq =
3705 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3706 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3707 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3708 goto failed_mount_wq;
3709 }
3710
3711 /*
3712 * The jbd2_journal_load will have done any necessary log recovery,
3713 * so we can safely mount the rest of the filesystem now.
3714 */
3715
3716 root = ext4_iget(sb, EXT4_ROOT_INO);
3717 if (IS_ERR(root)) {
3718 ext4_msg(sb, KERN_ERR, "get root inode failed");
3719 ret = PTR_ERR(root);
3720 root = NULL;
3721 goto failed_mount4;
3722 }
3723 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3724 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3725 iput(root);
3726 goto failed_mount4;
3727 }
3728 sb->s_root = d_make_root(root);
3729 if (!sb->s_root) {
3730 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3731 ret = -ENOMEM;
3732 goto failed_mount4;
3733 }
3734
3735 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3736
3737 /* determine the minimum size of new large inodes, if present */
3738 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3739 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3740 EXT4_GOOD_OLD_INODE_SIZE;
3741 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3742 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3743 if (sbi->s_want_extra_isize <
3744 le16_to_cpu(es->s_want_extra_isize))
3745 sbi->s_want_extra_isize =
3746 le16_to_cpu(es->s_want_extra_isize);
3747 if (sbi->s_want_extra_isize <
3748 le16_to_cpu(es->s_min_extra_isize))
3749 sbi->s_want_extra_isize =
3750 le16_to_cpu(es->s_min_extra_isize);
3751 }
3752 }
3753 /* Check if enough inode space is available */
3754 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3755 sbi->s_inode_size) {
3756 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3757 EXT4_GOOD_OLD_INODE_SIZE;
3758 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3759 "available");
3760 }
3761
3762 err = ext4_setup_system_zone(sb);
3763 if (err) {
3764 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3765 "zone (%d)", err);
3766 goto failed_mount4a;
3767 }
3768
3769 ext4_ext_init(sb);
3770 err = ext4_mb_init(sb, needs_recovery);
3771 if (err) {
3772 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3773 err);
3774 goto failed_mount5;
3775 }
3776
3777 err = ext4_register_li_request(sb, first_not_zeroed);
3778 if (err)
3779 goto failed_mount6;
3780
3781 sbi->s_kobj.kset = ext4_kset;
3782 init_completion(&sbi->s_kobj_unregister);
3783 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3784 "%s", sb->s_id);
3785 if (err)
3786 goto failed_mount7;
3787
3788 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3789 ext4_orphan_cleanup(sb, es);
3790 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3791 if (needs_recovery) {
3792 ext4_msg(sb, KERN_INFO, "recovery complete");
3793 ext4_mark_recovery_complete(sb, es);
3794 }
3795 if (EXT4_SB(sb)->s_journal) {
3796 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3797 descr = " journalled data mode";
3798 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3799 descr = " ordered data mode";
3800 else
3801 descr = " writeback data mode";
3802 } else
3803 descr = "out journal";
3804
3805 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3806 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3807 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3808
3809 if (es->s_error_count)
3810 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3811
3812 kfree(orig_data);
3813 return 0;
3814
3815 cantfind_ext4:
3816 if (!silent)
3817 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3818 goto failed_mount;
3819
3820 failed_mount7:
3821 ext4_unregister_li_request(sb);
3822 failed_mount6:
3823 ext4_mb_release(sb);
3824 failed_mount5:
3825 ext4_ext_release(sb);
3826 ext4_release_system_zone(sb);
3827 failed_mount4a:
3828 dput(sb->s_root);
3829 sb->s_root = NULL;
3830 failed_mount4:
3831 ext4_msg(sb, KERN_ERR, "mount failed");
3832 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3833 failed_mount_wq:
3834 if (sbi->s_journal) {
3835 jbd2_journal_destroy(sbi->s_journal);
3836 sbi->s_journal = NULL;
3837 }
3838 failed_mount3:
3839 del_timer(&sbi->s_err_report);
3840 if (sbi->s_flex_groups)
3841 ext4_kvfree(sbi->s_flex_groups);
3842 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3843 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3844 percpu_counter_destroy(&sbi->s_dirs_counter);
3845 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3846 if (sbi->s_mmp_tsk)
3847 kthread_stop(sbi->s_mmp_tsk);
3848 failed_mount2:
3849 for (i = 0; i < db_count; i++)
3850 brelse(sbi->s_group_desc[i]);
3851 ext4_kvfree(sbi->s_group_desc);
3852 failed_mount:
3853 if (sbi->s_chksum_driver)
3854 crypto_free_shash(sbi->s_chksum_driver);
3855 if (sbi->s_proc) {
3856 remove_proc_entry("options", sbi->s_proc);
3857 remove_proc_entry(sb->s_id, ext4_proc_root);
3858 }
3859 #ifdef CONFIG_QUOTA
3860 for (i = 0; i < MAXQUOTAS; i++)
3861 kfree(sbi->s_qf_names[i]);
3862 #endif
3863 ext4_blkdev_remove(sbi);
3864 brelse(bh);
3865 out_fail:
3866 sb->s_fs_info = NULL;
3867 kfree(sbi->s_blockgroup_lock);
3868 kfree(sbi);
3869 out_free_orig:
3870 kfree(orig_data);
3871 return ret;
3872 }
3873
3874 /*
3875 * Setup any per-fs journal parameters now. We'll do this both on
3876 * initial mount, once the journal has been initialised but before we've
3877 * done any recovery; and again on any subsequent remount.
3878 */
3879 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3880 {
3881 struct ext4_sb_info *sbi = EXT4_SB(sb);
3882
3883 journal->j_commit_interval = sbi->s_commit_interval;
3884 journal->j_min_batch_time = sbi->s_min_batch_time;
3885 journal->j_max_batch_time = sbi->s_max_batch_time;
3886
3887 write_lock(&journal->j_state_lock);
3888 if (test_opt(sb, BARRIER))
3889 journal->j_flags |= JBD2_BARRIER;
3890 else
3891 journal->j_flags &= ~JBD2_BARRIER;
3892 if (test_opt(sb, DATA_ERR_ABORT))
3893 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3894 else
3895 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3896 write_unlock(&journal->j_state_lock);
3897 }
3898
3899 static journal_t *ext4_get_journal(struct super_block *sb,
3900 unsigned int journal_inum)
3901 {
3902 struct inode *journal_inode;
3903 journal_t *journal;
3904
3905 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3906
3907 /* First, test for the existence of a valid inode on disk. Bad
3908 * things happen if we iget() an unused inode, as the subsequent
3909 * iput() will try to delete it. */
3910
3911 journal_inode = ext4_iget(sb, journal_inum);
3912 if (IS_ERR(journal_inode)) {
3913 ext4_msg(sb, KERN_ERR, "no journal found");
3914 return NULL;
3915 }
3916 if (!journal_inode->i_nlink) {
3917 make_bad_inode(journal_inode);
3918 iput(journal_inode);
3919 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3920 return NULL;
3921 }
3922
3923 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3924 journal_inode, journal_inode->i_size);
3925 if (!S_ISREG(journal_inode->i_mode)) {
3926 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3927 iput(journal_inode);
3928 return NULL;
3929 }
3930
3931 journal = jbd2_journal_init_inode(journal_inode);
3932 if (!journal) {
3933 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3934 iput(journal_inode);
3935 return NULL;
3936 }
3937 journal->j_private = sb;
3938 ext4_init_journal_params(sb, journal);
3939 return journal;
3940 }
3941
3942 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3943 dev_t j_dev)
3944 {
3945 struct buffer_head *bh;
3946 journal_t *journal;
3947 ext4_fsblk_t start;
3948 ext4_fsblk_t len;
3949 int hblock, blocksize;
3950 ext4_fsblk_t sb_block;
3951 unsigned long offset;
3952 struct ext4_super_block *es;
3953 struct block_device *bdev;
3954
3955 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3956
3957 bdev = ext4_blkdev_get(j_dev, sb);
3958 if (bdev == NULL)
3959 return NULL;
3960
3961 blocksize = sb->s_blocksize;
3962 hblock = bdev_logical_block_size(bdev);
3963 if (blocksize < hblock) {
3964 ext4_msg(sb, KERN_ERR,
3965 "blocksize too small for journal device");
3966 goto out_bdev;
3967 }
3968
3969 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3970 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3971 set_blocksize(bdev, blocksize);
3972 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3973 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3974 "external journal");
3975 goto out_bdev;
3976 }
3977
3978 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3979 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3980 !(le32_to_cpu(es->s_feature_incompat) &
3981 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3982 ext4_msg(sb, KERN_ERR, "external journal has "
3983 "bad superblock");
3984 brelse(bh);
3985 goto out_bdev;
3986 }
3987
3988 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3989 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3990 brelse(bh);
3991 goto out_bdev;
3992 }
3993
3994 len = ext4_blocks_count(es);
3995 start = sb_block + 1;
3996 brelse(bh); /* we're done with the superblock */
3997
3998 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3999 start, len, blocksize);
4000 if (!journal) {
4001 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4002 goto out_bdev;
4003 }
4004 journal->j_private = sb;
4005 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4006 wait_on_buffer(journal->j_sb_buffer);
4007 if (!buffer_uptodate(journal->j_sb_buffer)) {
4008 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4009 goto out_journal;
4010 }
4011 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4012 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4013 "user (unsupported) - %d",
4014 be32_to_cpu(journal->j_superblock->s_nr_users));
4015 goto out_journal;
4016 }
4017 EXT4_SB(sb)->journal_bdev = bdev;
4018 ext4_init_journal_params(sb, journal);
4019 return journal;
4020
4021 out_journal:
4022 jbd2_journal_destroy(journal);
4023 out_bdev:
4024 ext4_blkdev_put(bdev);
4025 return NULL;
4026 }
4027
4028 static int ext4_load_journal(struct super_block *sb,
4029 struct ext4_super_block *es,
4030 unsigned long journal_devnum)
4031 {
4032 journal_t *journal;
4033 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4034 dev_t journal_dev;
4035 int err = 0;
4036 int really_read_only;
4037
4038 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4039
4040 if (journal_devnum &&
4041 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4042 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4043 "numbers have changed");
4044 journal_dev = new_decode_dev(journal_devnum);
4045 } else
4046 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4047
4048 really_read_only = bdev_read_only(sb->s_bdev);
4049
4050 /*
4051 * Are we loading a blank journal or performing recovery after a
4052 * crash? For recovery, we need to check in advance whether we
4053 * can get read-write access to the device.
4054 */
4055 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4056 if (sb->s_flags & MS_RDONLY) {
4057 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4058 "required on readonly filesystem");
4059 if (really_read_only) {
4060 ext4_msg(sb, KERN_ERR, "write access "
4061 "unavailable, cannot proceed");
4062 return -EROFS;
4063 }
4064 ext4_msg(sb, KERN_INFO, "write access will "
4065 "be enabled during recovery");
4066 }
4067 }
4068
4069 if (journal_inum && journal_dev) {
4070 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4071 "and inode journals!");
4072 return -EINVAL;
4073 }
4074
4075 if (journal_inum) {
4076 if (!(journal = ext4_get_journal(sb, journal_inum)))
4077 return -EINVAL;
4078 } else {
4079 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4080 return -EINVAL;
4081 }
4082
4083 if (!(journal->j_flags & JBD2_BARRIER))
4084 ext4_msg(sb, KERN_INFO, "barriers disabled");
4085
4086 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4087 err = jbd2_journal_wipe(journal, !really_read_only);
4088 if (!err) {
4089 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4090 if (save)
4091 memcpy(save, ((char *) es) +
4092 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4093 err = jbd2_journal_load(journal);
4094 if (save)
4095 memcpy(((char *) es) + EXT4_S_ERR_START,
4096 save, EXT4_S_ERR_LEN);
4097 kfree(save);
4098 }
4099
4100 if (err) {
4101 ext4_msg(sb, KERN_ERR, "error loading journal");
4102 jbd2_journal_destroy(journal);
4103 return err;
4104 }
4105
4106 EXT4_SB(sb)->s_journal = journal;
4107 ext4_clear_journal_err(sb, es);
4108
4109 if (!really_read_only && journal_devnum &&
4110 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4111 es->s_journal_dev = cpu_to_le32(journal_devnum);
4112
4113 /* Make sure we flush the recovery flag to disk. */
4114 ext4_commit_super(sb, 1);
4115 }
4116
4117 return 0;
4118 }
4119
4120 static int ext4_commit_super(struct super_block *sb, int sync)
4121 {
4122 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4123 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4124 int error = 0;
4125
4126 if (!sbh || block_device_ejected(sb))
4127 return error;
4128 if (buffer_write_io_error(sbh)) {
4129 /*
4130 * Oh, dear. A previous attempt to write the
4131 * superblock failed. This could happen because the
4132 * USB device was yanked out. Or it could happen to
4133 * be a transient write error and maybe the block will
4134 * be remapped. Nothing we can do but to retry the
4135 * write and hope for the best.
4136 */
4137 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4138 "superblock detected");
4139 clear_buffer_write_io_error(sbh);
4140 set_buffer_uptodate(sbh);
4141 }
4142 /*
4143 * If the file system is mounted read-only, don't update the
4144 * superblock write time. This avoids updating the superblock
4145 * write time when we are mounting the root file system
4146 * read/only but we need to replay the journal; at that point,
4147 * for people who are east of GMT and who make their clock
4148 * tick in localtime for Windows bug-for-bug compatibility,
4149 * the clock is set in the future, and this will cause e2fsck
4150 * to complain and force a full file system check.
4151 */
4152 if (!(sb->s_flags & MS_RDONLY))
4153 es->s_wtime = cpu_to_le32(get_seconds());
4154 if (sb->s_bdev->bd_part)
4155 es->s_kbytes_written =
4156 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4157 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4158 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4159 else
4160 es->s_kbytes_written =
4161 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4162 ext4_free_blocks_count_set(es,
4163 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4164 &EXT4_SB(sb)->s_freeclusters_counter)));
4165 es->s_free_inodes_count =
4166 cpu_to_le32(percpu_counter_sum_positive(
4167 &EXT4_SB(sb)->s_freeinodes_counter));
4168 sb->s_dirt = 0;
4169 BUFFER_TRACE(sbh, "marking dirty");
4170 ext4_superblock_csum_set(sb, es);
4171 mark_buffer_dirty(sbh);
4172 if (sync) {
4173 error = sync_dirty_buffer(sbh);
4174 if (error)
4175 return error;
4176
4177 error = buffer_write_io_error(sbh);
4178 if (error) {
4179 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4180 "superblock");
4181 clear_buffer_write_io_error(sbh);
4182 set_buffer_uptodate(sbh);
4183 }
4184 }
4185 return error;
4186 }
4187
4188 /*
4189 * Have we just finished recovery? If so, and if we are mounting (or
4190 * remounting) the filesystem readonly, then we will end up with a
4191 * consistent fs on disk. Record that fact.
4192 */
4193 static void ext4_mark_recovery_complete(struct super_block *sb,
4194 struct ext4_super_block *es)
4195 {
4196 journal_t *journal = EXT4_SB(sb)->s_journal;
4197
4198 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4199 BUG_ON(journal != NULL);
4200 return;
4201 }
4202 jbd2_journal_lock_updates(journal);
4203 if (jbd2_journal_flush(journal) < 0)
4204 goto out;
4205
4206 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4207 sb->s_flags & MS_RDONLY) {
4208 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4209 ext4_commit_super(sb, 1);
4210 }
4211
4212 out:
4213 jbd2_journal_unlock_updates(journal);
4214 }
4215
4216 /*
4217 * If we are mounting (or read-write remounting) a filesystem whose journal
4218 * has recorded an error from a previous lifetime, move that error to the
4219 * main filesystem now.
4220 */
4221 static void ext4_clear_journal_err(struct super_block *sb,
4222 struct ext4_super_block *es)
4223 {
4224 journal_t *journal;
4225 int j_errno;
4226 const char *errstr;
4227
4228 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4229
4230 journal = EXT4_SB(sb)->s_journal;
4231
4232 /*
4233 * Now check for any error status which may have been recorded in the
4234 * journal by a prior ext4_error() or ext4_abort()
4235 */
4236
4237 j_errno = jbd2_journal_errno(journal);
4238 if (j_errno) {
4239 char nbuf[16];
4240
4241 errstr = ext4_decode_error(sb, j_errno, nbuf);
4242 ext4_warning(sb, "Filesystem error recorded "
4243 "from previous mount: %s", errstr);
4244 ext4_warning(sb, "Marking fs in need of filesystem check.");
4245
4246 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4247 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4248 ext4_commit_super(sb, 1);
4249
4250 jbd2_journal_clear_err(journal);
4251 }
4252 }
4253
4254 /*
4255 * Force the running and committing transactions to commit,
4256 * and wait on the commit.
4257 */
4258 int ext4_force_commit(struct super_block *sb)
4259 {
4260 journal_t *journal;
4261 int ret = 0;
4262
4263 if (sb->s_flags & MS_RDONLY)
4264 return 0;
4265
4266 journal = EXT4_SB(sb)->s_journal;
4267 if (journal) {
4268 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4269 ret = ext4_journal_force_commit(journal);
4270 }
4271
4272 return ret;
4273 }
4274
4275 static void ext4_write_super(struct super_block *sb)
4276 {
4277 lock_super(sb);
4278 ext4_commit_super(sb, 1);
4279 unlock_super(sb);
4280 }
4281
4282 static int ext4_sync_fs(struct super_block *sb, int wait)
4283 {
4284 int ret = 0;
4285 tid_t target;
4286 struct ext4_sb_info *sbi = EXT4_SB(sb);
4287
4288 trace_ext4_sync_fs(sb, wait);
4289 flush_workqueue(sbi->dio_unwritten_wq);
4290 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4291 if (wait)
4292 jbd2_log_wait_commit(sbi->s_journal, target);
4293 }
4294 return ret;
4295 }
4296
4297 /*
4298 * LVM calls this function before a (read-only) snapshot is created. This
4299 * gives us a chance to flush the journal completely and mark the fs clean.
4300 *
4301 * Note that only this function cannot bring a filesystem to be in a clean
4302 * state independently, because ext4 prevents a new handle from being started
4303 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4304 * the upper layer.
4305 */
4306 static int ext4_freeze(struct super_block *sb)
4307 {
4308 int error = 0;
4309 journal_t *journal;
4310
4311 if (sb->s_flags & MS_RDONLY)
4312 return 0;
4313
4314 journal = EXT4_SB(sb)->s_journal;
4315
4316 /* Now we set up the journal barrier. */
4317 jbd2_journal_lock_updates(journal);
4318
4319 /*
4320 * Don't clear the needs_recovery flag if we failed to flush
4321 * the journal.
4322 */
4323 error = jbd2_journal_flush(journal);
4324 if (error < 0)
4325 goto out;
4326
4327 /* Journal blocked and flushed, clear needs_recovery flag. */
4328 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4329 error = ext4_commit_super(sb, 1);
4330 out:
4331 /* we rely on s_frozen to stop further updates */
4332 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4333 return error;
4334 }
4335
4336 /*
4337 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4338 * flag here, even though the filesystem is not technically dirty yet.
4339 */
4340 static int ext4_unfreeze(struct super_block *sb)
4341 {
4342 if (sb->s_flags & MS_RDONLY)
4343 return 0;
4344
4345 lock_super(sb);
4346 /* Reset the needs_recovery flag before the fs is unlocked. */
4347 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4348 ext4_commit_super(sb, 1);
4349 unlock_super(sb);
4350 return 0;
4351 }
4352
4353 /*
4354 * Structure to save mount options for ext4_remount's benefit
4355 */
4356 struct ext4_mount_options {
4357 unsigned long s_mount_opt;
4358 unsigned long s_mount_opt2;
4359 uid_t s_resuid;
4360 gid_t s_resgid;
4361 unsigned long s_commit_interval;
4362 u32 s_min_batch_time, s_max_batch_time;
4363 #ifdef CONFIG_QUOTA
4364 int s_jquota_fmt;
4365 char *s_qf_names[MAXQUOTAS];
4366 #endif
4367 };
4368
4369 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4370 {
4371 struct ext4_super_block *es;
4372 struct ext4_sb_info *sbi = EXT4_SB(sb);
4373 unsigned long old_sb_flags;
4374 struct ext4_mount_options old_opts;
4375 int enable_quota = 0;
4376 ext4_group_t g;
4377 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4378 int err = 0;
4379 #ifdef CONFIG_QUOTA
4380 int i;
4381 #endif
4382 char *orig_data = kstrdup(data, GFP_KERNEL);
4383
4384 /* Store the original options */
4385 lock_super(sb);
4386 old_sb_flags = sb->s_flags;
4387 old_opts.s_mount_opt = sbi->s_mount_opt;
4388 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4389 old_opts.s_resuid = sbi->s_resuid;
4390 old_opts.s_resgid = sbi->s_resgid;
4391 old_opts.s_commit_interval = sbi->s_commit_interval;
4392 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4393 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4394 #ifdef CONFIG_QUOTA
4395 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4396 for (i = 0; i < MAXQUOTAS; i++)
4397 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4398 #endif
4399 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4400 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4401
4402 /*
4403 * Allow the "check" option to be passed as a remount option.
4404 */
4405 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4406 err = -EINVAL;
4407 goto restore_opts;
4408 }
4409
4410 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4411 ext4_abort(sb, "Abort forced by user");
4412
4413 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4414 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4415
4416 es = sbi->s_es;
4417
4418 if (sbi->s_journal) {
4419 ext4_init_journal_params(sb, sbi->s_journal);
4420 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4421 }
4422
4423 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4424 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4425 err = -EROFS;
4426 goto restore_opts;
4427 }
4428
4429 if (*flags & MS_RDONLY) {
4430 err = dquot_suspend(sb, -1);
4431 if (err < 0)
4432 goto restore_opts;
4433
4434 /*
4435 * First of all, the unconditional stuff we have to do
4436 * to disable replay of the journal when we next remount
4437 */
4438 sb->s_flags |= MS_RDONLY;
4439
4440 /*
4441 * OK, test if we are remounting a valid rw partition
4442 * readonly, and if so set the rdonly flag and then
4443 * mark the partition as valid again.
4444 */
4445 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4446 (sbi->s_mount_state & EXT4_VALID_FS))
4447 es->s_state = cpu_to_le16(sbi->s_mount_state);
4448
4449 if (sbi->s_journal)
4450 ext4_mark_recovery_complete(sb, es);
4451 } else {
4452 /* Make sure we can mount this feature set readwrite */
4453 if (!ext4_feature_set_ok(sb, 0)) {
4454 err = -EROFS;
4455 goto restore_opts;
4456 }
4457 /*
4458 * Make sure the group descriptor checksums
4459 * are sane. If they aren't, refuse to remount r/w.
4460 */
4461 for (g = 0; g < sbi->s_groups_count; g++) {
4462 struct ext4_group_desc *gdp =
4463 ext4_get_group_desc(sb, g, NULL);
4464
4465 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4466 ext4_msg(sb, KERN_ERR,
4467 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4468 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4469 le16_to_cpu(gdp->bg_checksum));
4470 err = -EINVAL;
4471 goto restore_opts;
4472 }
4473 }
4474
4475 /*
4476 * If we have an unprocessed orphan list hanging
4477 * around from a previously readonly bdev mount,
4478 * require a full umount/remount for now.
4479 */
4480 if (es->s_last_orphan) {
4481 ext4_msg(sb, KERN_WARNING, "Couldn't "
4482 "remount RDWR because of unprocessed "
4483 "orphan inode list. Please "
4484 "umount/remount instead");
4485 err = -EINVAL;
4486 goto restore_opts;
4487 }
4488
4489 /*
4490 * Mounting a RDONLY partition read-write, so reread
4491 * and store the current valid flag. (It may have
4492 * been changed by e2fsck since we originally mounted
4493 * the partition.)
4494 */
4495 if (sbi->s_journal)
4496 ext4_clear_journal_err(sb, es);
4497 sbi->s_mount_state = le16_to_cpu(es->s_state);
4498 if (!ext4_setup_super(sb, es, 0))
4499 sb->s_flags &= ~MS_RDONLY;
4500 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4501 EXT4_FEATURE_INCOMPAT_MMP))
4502 if (ext4_multi_mount_protect(sb,
4503 le64_to_cpu(es->s_mmp_block))) {
4504 err = -EROFS;
4505 goto restore_opts;
4506 }
4507 enable_quota = 1;
4508 }
4509 }
4510
4511 /*
4512 * Reinitialize lazy itable initialization thread based on
4513 * current settings
4514 */
4515 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4516 ext4_unregister_li_request(sb);
4517 else {
4518 ext4_group_t first_not_zeroed;
4519 first_not_zeroed = ext4_has_uninit_itable(sb);
4520 ext4_register_li_request(sb, first_not_zeroed);
4521 }
4522
4523 ext4_setup_system_zone(sb);
4524 if (sbi->s_journal == NULL)
4525 ext4_commit_super(sb, 1);
4526
4527 #ifdef CONFIG_QUOTA
4528 /* Release old quota file names */
4529 for (i = 0; i < MAXQUOTAS; i++)
4530 if (old_opts.s_qf_names[i] &&
4531 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4532 kfree(old_opts.s_qf_names[i]);
4533 #endif
4534 unlock_super(sb);
4535 if (enable_quota)
4536 dquot_resume(sb, -1);
4537
4538 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4539 kfree(orig_data);
4540 return 0;
4541
4542 restore_opts:
4543 sb->s_flags = old_sb_flags;
4544 sbi->s_mount_opt = old_opts.s_mount_opt;
4545 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4546 sbi->s_resuid = old_opts.s_resuid;
4547 sbi->s_resgid = old_opts.s_resgid;
4548 sbi->s_commit_interval = old_opts.s_commit_interval;
4549 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4550 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4551 #ifdef CONFIG_QUOTA
4552 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4553 for (i = 0; i < MAXQUOTAS; i++) {
4554 if (sbi->s_qf_names[i] &&
4555 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4556 kfree(sbi->s_qf_names[i]);
4557 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4558 }
4559 #endif
4560 unlock_super(sb);
4561 kfree(orig_data);
4562 return err;
4563 }
4564
4565 /*
4566 * Note: calculating the overhead so we can be compatible with
4567 * historical BSD practice is quite difficult in the face of
4568 * clusters/bigalloc. This is because multiple metadata blocks from
4569 * different block group can end up in the same allocation cluster.
4570 * Calculating the exact overhead in the face of clustered allocation
4571 * requires either O(all block bitmaps) in memory or O(number of block
4572 * groups**2) in time. We will still calculate the superblock for
4573 * older file systems --- and if we come across with a bigalloc file
4574 * system with zero in s_overhead_clusters the estimate will be close to
4575 * correct especially for very large cluster sizes --- but for newer
4576 * file systems, it's better to calculate this figure once at mkfs
4577 * time, and store it in the superblock. If the superblock value is
4578 * present (even for non-bigalloc file systems), we will use it.
4579 */
4580 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4581 {
4582 struct super_block *sb = dentry->d_sb;
4583 struct ext4_sb_info *sbi = EXT4_SB(sb);
4584 struct ext4_super_block *es = sbi->s_es;
4585 struct ext4_group_desc *gdp;
4586 u64 fsid;
4587 s64 bfree;
4588
4589 if (test_opt(sb, MINIX_DF)) {
4590 sbi->s_overhead_last = 0;
4591 } else if (es->s_overhead_clusters) {
4592 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4593 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4594 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4595 ext4_fsblk_t overhead = 0;
4596
4597 /*
4598 * Compute the overhead (FS structures). This is constant
4599 * for a given filesystem unless the number of block groups
4600 * changes so we cache the previous value until it does.
4601 */
4602
4603 /*
4604 * All of the blocks before first_data_block are
4605 * overhead
4606 */
4607 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4608
4609 /*
4610 * Add the overhead found in each block group
4611 */
4612 for (i = 0; i < ngroups; i++) {
4613 gdp = ext4_get_group_desc(sb, i, NULL);
4614 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4615 cond_resched();
4616 }
4617 sbi->s_overhead_last = overhead;
4618 smp_wmb();
4619 sbi->s_blocks_last = ext4_blocks_count(es);
4620 }
4621
4622 buf->f_type = EXT4_SUPER_MAGIC;
4623 buf->f_bsize = sb->s_blocksize;
4624 buf->f_blocks = (ext4_blocks_count(es) -
4625 EXT4_C2B(sbi, sbi->s_overhead_last));
4626 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4627 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4628 /* prevent underflow in case that few free space is available */
4629 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4630 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4631 if (buf->f_bfree < ext4_r_blocks_count(es))
4632 buf->f_bavail = 0;
4633 buf->f_files = le32_to_cpu(es->s_inodes_count);
4634 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4635 buf->f_namelen = EXT4_NAME_LEN;
4636 fsid = le64_to_cpup((void *)es->s_uuid) ^
4637 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4638 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4639 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4640
4641 return 0;
4642 }
4643
4644 /* Helper function for writing quotas on sync - we need to start transaction
4645 * before quota file is locked for write. Otherwise the are possible deadlocks:
4646 * Process 1 Process 2
4647 * ext4_create() quota_sync()
4648 * jbd2_journal_start() write_dquot()
4649 * dquot_initialize() down(dqio_mutex)
4650 * down(dqio_mutex) jbd2_journal_start()
4651 *
4652 */
4653
4654 #ifdef CONFIG_QUOTA
4655
4656 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4657 {
4658 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4659 }
4660
4661 static int ext4_write_dquot(struct dquot *dquot)
4662 {
4663 int ret, err;
4664 handle_t *handle;
4665 struct inode *inode;
4666
4667 inode = dquot_to_inode(dquot);
4668 handle = ext4_journal_start(inode,
4669 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4670 if (IS_ERR(handle))
4671 return PTR_ERR(handle);
4672 ret = dquot_commit(dquot);
4673 err = ext4_journal_stop(handle);
4674 if (!ret)
4675 ret = err;
4676 return ret;
4677 }
4678
4679 static int ext4_acquire_dquot(struct dquot *dquot)
4680 {
4681 int ret, err;
4682 handle_t *handle;
4683
4684 handle = ext4_journal_start(dquot_to_inode(dquot),
4685 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4686 if (IS_ERR(handle))
4687 return PTR_ERR(handle);
4688 ret = dquot_acquire(dquot);
4689 err = ext4_journal_stop(handle);
4690 if (!ret)
4691 ret = err;
4692 return ret;
4693 }
4694
4695 static int ext4_release_dquot(struct dquot *dquot)
4696 {
4697 int ret, err;
4698 handle_t *handle;
4699
4700 handle = ext4_journal_start(dquot_to_inode(dquot),
4701 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4702 if (IS_ERR(handle)) {
4703 /* Release dquot anyway to avoid endless cycle in dqput() */
4704 dquot_release(dquot);
4705 return PTR_ERR(handle);
4706 }
4707 ret = dquot_release(dquot);
4708 err = ext4_journal_stop(handle);
4709 if (!ret)
4710 ret = err;
4711 return ret;
4712 }
4713
4714 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4715 {
4716 /* Are we journaling quotas? */
4717 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4718 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4719 dquot_mark_dquot_dirty(dquot);
4720 return ext4_write_dquot(dquot);
4721 } else {
4722 return dquot_mark_dquot_dirty(dquot);
4723 }
4724 }
4725
4726 static int ext4_write_info(struct super_block *sb, int type)
4727 {
4728 int ret, err;
4729 handle_t *handle;
4730
4731 /* Data block + inode block */
4732 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4733 if (IS_ERR(handle))
4734 return PTR_ERR(handle);
4735 ret = dquot_commit_info(sb, type);
4736 err = ext4_journal_stop(handle);
4737 if (!ret)
4738 ret = err;
4739 return ret;
4740 }
4741
4742 /*
4743 * Turn on quotas during mount time - we need to find
4744 * the quota file and such...
4745 */
4746 static int ext4_quota_on_mount(struct super_block *sb, int type)
4747 {
4748 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4749 EXT4_SB(sb)->s_jquota_fmt, type);
4750 }
4751
4752 /*
4753 * Standard function to be called on quota_on
4754 */
4755 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4756 struct path *path)
4757 {
4758 int err;
4759
4760 if (!test_opt(sb, QUOTA))
4761 return -EINVAL;
4762
4763 /* Quotafile not on the same filesystem? */
4764 if (path->dentry->d_sb != sb)
4765 return -EXDEV;
4766 /* Journaling quota? */
4767 if (EXT4_SB(sb)->s_qf_names[type]) {
4768 /* Quotafile not in fs root? */
4769 if (path->dentry->d_parent != sb->s_root)
4770 ext4_msg(sb, KERN_WARNING,
4771 "Quota file not on filesystem root. "
4772 "Journaled quota will not work");
4773 }
4774
4775 /*
4776 * When we journal data on quota file, we have to flush journal to see
4777 * all updates to the file when we bypass pagecache...
4778 */
4779 if (EXT4_SB(sb)->s_journal &&
4780 ext4_should_journal_data(path->dentry->d_inode)) {
4781 /*
4782 * We don't need to lock updates but journal_flush() could
4783 * otherwise be livelocked...
4784 */
4785 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4786 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4787 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4788 if (err)
4789 return err;
4790 }
4791
4792 return dquot_quota_on(sb, type, format_id, path);
4793 }
4794
4795 static int ext4_quota_off(struct super_block *sb, int type)
4796 {
4797 struct inode *inode = sb_dqopt(sb)->files[type];
4798 handle_t *handle;
4799
4800 /* Force all delayed allocation blocks to be allocated.
4801 * Caller already holds s_umount sem */
4802 if (test_opt(sb, DELALLOC))
4803 sync_filesystem(sb);
4804
4805 if (!inode)
4806 goto out;
4807
4808 /* Update modification times of quota files when userspace can
4809 * start looking at them */
4810 handle = ext4_journal_start(inode, 1);
4811 if (IS_ERR(handle))
4812 goto out;
4813 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4814 ext4_mark_inode_dirty(handle, inode);
4815 ext4_journal_stop(handle);
4816
4817 out:
4818 return dquot_quota_off(sb, type);
4819 }
4820
4821 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4822 * acquiring the locks... As quota files are never truncated and quota code
4823 * itself serializes the operations (and no one else should touch the files)
4824 * we don't have to be afraid of races */
4825 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4826 size_t len, loff_t off)
4827 {
4828 struct inode *inode = sb_dqopt(sb)->files[type];
4829 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4830 int err = 0;
4831 int offset = off & (sb->s_blocksize - 1);
4832 int tocopy;
4833 size_t toread;
4834 struct buffer_head *bh;
4835 loff_t i_size = i_size_read(inode);
4836
4837 if (off > i_size)
4838 return 0;
4839 if (off+len > i_size)
4840 len = i_size-off;
4841 toread = len;
4842 while (toread > 0) {
4843 tocopy = sb->s_blocksize - offset < toread ?
4844 sb->s_blocksize - offset : toread;
4845 bh = ext4_bread(NULL, inode, blk, 0, &err);
4846 if (err)
4847 return err;
4848 if (!bh) /* A hole? */
4849 memset(data, 0, tocopy);
4850 else
4851 memcpy(data, bh->b_data+offset, tocopy);
4852 brelse(bh);
4853 offset = 0;
4854 toread -= tocopy;
4855 data += tocopy;
4856 blk++;
4857 }
4858 return len;
4859 }
4860
4861 /* Write to quotafile (we know the transaction is already started and has
4862 * enough credits) */
4863 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4864 const char *data, size_t len, loff_t off)
4865 {
4866 struct inode *inode = sb_dqopt(sb)->files[type];
4867 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4868 int err = 0;
4869 int offset = off & (sb->s_blocksize - 1);
4870 struct buffer_head *bh;
4871 handle_t *handle = journal_current_handle();
4872
4873 if (EXT4_SB(sb)->s_journal && !handle) {
4874 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4875 " cancelled because transaction is not started",
4876 (unsigned long long)off, (unsigned long long)len);
4877 return -EIO;
4878 }
4879 /*
4880 * Since we account only one data block in transaction credits,
4881 * then it is impossible to cross a block boundary.
4882 */
4883 if (sb->s_blocksize - offset < len) {
4884 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4885 " cancelled because not block aligned",
4886 (unsigned long long)off, (unsigned long long)len);
4887 return -EIO;
4888 }
4889
4890 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4891 bh = ext4_bread(handle, inode, blk, 1, &err);
4892 if (!bh)
4893 goto out;
4894 err = ext4_journal_get_write_access(handle, bh);
4895 if (err) {
4896 brelse(bh);
4897 goto out;
4898 }
4899 lock_buffer(bh);
4900 memcpy(bh->b_data+offset, data, len);
4901 flush_dcache_page(bh->b_page);
4902 unlock_buffer(bh);
4903 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4904 brelse(bh);
4905 out:
4906 if (err) {
4907 mutex_unlock(&inode->i_mutex);
4908 return err;
4909 }
4910 if (inode->i_size < off + len) {
4911 i_size_write(inode, off + len);
4912 EXT4_I(inode)->i_disksize = inode->i_size;
4913 ext4_mark_inode_dirty(handle, inode);
4914 }
4915 mutex_unlock(&inode->i_mutex);
4916 return len;
4917 }
4918
4919 #endif
4920
4921 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4922 const char *dev_name, void *data)
4923 {
4924 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4925 }
4926
4927 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4928 static inline void register_as_ext2(void)
4929 {
4930 int err = register_filesystem(&ext2_fs_type);
4931 if (err)
4932 printk(KERN_WARNING
4933 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4934 }
4935
4936 static inline void unregister_as_ext2(void)
4937 {
4938 unregister_filesystem(&ext2_fs_type);
4939 }
4940
4941 static inline int ext2_feature_set_ok(struct super_block *sb)
4942 {
4943 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4944 return 0;
4945 if (sb->s_flags & MS_RDONLY)
4946 return 1;
4947 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4948 return 0;
4949 return 1;
4950 }
4951 MODULE_ALIAS("ext2");
4952 #else
4953 static inline void register_as_ext2(void) { }
4954 static inline void unregister_as_ext2(void) { }
4955 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4956 #endif
4957
4958 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4959 static inline void register_as_ext3(void)
4960 {
4961 int err = register_filesystem(&ext3_fs_type);
4962 if (err)
4963 printk(KERN_WARNING
4964 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4965 }
4966
4967 static inline void unregister_as_ext3(void)
4968 {
4969 unregister_filesystem(&ext3_fs_type);
4970 }
4971
4972 static inline int ext3_feature_set_ok(struct super_block *sb)
4973 {
4974 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4975 return 0;
4976 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4977 return 0;
4978 if (sb->s_flags & MS_RDONLY)
4979 return 1;
4980 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4981 return 0;
4982 return 1;
4983 }
4984 MODULE_ALIAS("ext3");
4985 #else
4986 static inline void register_as_ext3(void) { }
4987 static inline void unregister_as_ext3(void) { }
4988 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
4989 #endif
4990
4991 static struct file_system_type ext4_fs_type = {
4992 .owner = THIS_MODULE,
4993 .name = "ext4",
4994 .mount = ext4_mount,
4995 .kill_sb = kill_block_super,
4996 .fs_flags = FS_REQUIRES_DEV,
4997 };
4998
4999 static int __init ext4_init_feat_adverts(void)
5000 {
5001 struct ext4_features *ef;
5002 int ret = -ENOMEM;
5003
5004 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5005 if (!ef)
5006 goto out;
5007
5008 ef->f_kobj.kset = ext4_kset;
5009 init_completion(&ef->f_kobj_unregister);
5010 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5011 "features");
5012 if (ret) {
5013 kfree(ef);
5014 goto out;
5015 }
5016
5017 ext4_feat = ef;
5018 ret = 0;
5019 out:
5020 return ret;
5021 }
5022
5023 static void ext4_exit_feat_adverts(void)
5024 {
5025 kobject_put(&ext4_feat->f_kobj);
5026 wait_for_completion(&ext4_feat->f_kobj_unregister);
5027 kfree(ext4_feat);
5028 }
5029
5030 /* Shared across all ext4 file systems */
5031 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5032 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5033
5034 static int __init ext4_init_fs(void)
5035 {
5036 int i, err;
5037
5038 ext4_li_info = NULL;
5039 mutex_init(&ext4_li_mtx);
5040
5041 ext4_check_flag_values();
5042
5043 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5044 mutex_init(&ext4__aio_mutex[i]);
5045 init_waitqueue_head(&ext4__ioend_wq[i]);
5046 }
5047
5048 err = ext4_init_pageio();
5049 if (err)
5050 return err;
5051 err = ext4_init_system_zone();
5052 if (err)
5053 goto out6;
5054 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5055 if (!ext4_kset)
5056 goto out5;
5057 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5058
5059 err = ext4_init_feat_adverts();
5060 if (err)
5061 goto out4;
5062
5063 err = ext4_init_mballoc();
5064 if (err)
5065 goto out3;
5066
5067 err = ext4_init_xattr();
5068 if (err)
5069 goto out2;
5070 err = init_inodecache();
5071 if (err)
5072 goto out1;
5073 register_as_ext3();
5074 register_as_ext2();
5075 err = register_filesystem(&ext4_fs_type);
5076 if (err)
5077 goto out;
5078
5079 return 0;
5080 out:
5081 unregister_as_ext2();
5082 unregister_as_ext3();
5083 destroy_inodecache();
5084 out1:
5085 ext4_exit_xattr();
5086 out2:
5087 ext4_exit_mballoc();
5088 out3:
5089 ext4_exit_feat_adverts();
5090 out4:
5091 if (ext4_proc_root)
5092 remove_proc_entry("fs/ext4", NULL);
5093 kset_unregister(ext4_kset);
5094 out5:
5095 ext4_exit_system_zone();
5096 out6:
5097 ext4_exit_pageio();
5098 return err;
5099 }
5100
5101 static void __exit ext4_exit_fs(void)
5102 {
5103 ext4_destroy_lazyinit_thread();
5104 unregister_as_ext2();
5105 unregister_as_ext3();
5106 unregister_filesystem(&ext4_fs_type);
5107 destroy_inodecache();
5108 ext4_exit_xattr();
5109 ext4_exit_mballoc();
5110 ext4_exit_feat_adverts();
5111 remove_proc_entry("fs/ext4", NULL);
5112 kset_unregister(ext4_kset);
5113 ext4_exit_system_zone();
5114 ext4_exit_pageio();
5115 }
5116
5117 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5118 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5119 MODULE_LICENSE("GPL");
5120 module_init(ext4_init_fs)
5121 module_exit(ext4_exit_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree