?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ext4: use a table-driven handler for mount options
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blob61a65ee2c90521f48e3b4fe60866749aaf1522d5
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_commit_super(struct super_block *sb, int sync);
66 static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68 static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70 static int ext4_sync_fs(struct super_block *sb, int wait);
71 static const char *ext4_decode_error(struct super_block *sb, int errno,
72 char nbuf[16]);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static void ext4_write_super(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94 };
95 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96 #else
97 #define IS_EXT2_SB(sb) (0)
98 #endif
99
100
101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102 static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
104 .name = "ext3",
105 .mount = ext4_mount,
106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
108 };
109 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110 #else
111 #define IS_EXT3_SB(sb) (0)
112 #endif
113
114 void *ext4_kvmalloc(size_t size, gfp_t flags)
115 {
116 void *ret;
117
118 ret = kmalloc(size, flags);
119 if (!ret)
120 ret = __vmalloc(size, flags, PAGE_KERNEL);
121 return ret;
122 }
123
124 void *ext4_kvzalloc(size_t size, gfp_t flags)
125 {
126 void *ret;
127
128 ret = kzalloc(size, flags);
129 if (!ret)
130 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
131 return ret;
132 }
133
134 void ext4_kvfree(void *ptr)
135 {
136 if (is_vmalloc_addr(ptr))
137 vfree(ptr);
138 else
139 kfree(ptr);
140
141 }
142
143 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
144 struct ext4_group_desc *bg)
145 {
146 return le32_to_cpu(bg->bg_block_bitmap_lo) |
147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
148 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
149 }
150
151 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
152 struct ext4_group_desc *bg)
153 {
154 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
155 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
156 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
157 }
158
159 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
160 struct ext4_group_desc *bg)
161 {
162 return le32_to_cpu(bg->bg_inode_table_lo) |
163 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
164 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
165 }
166
167 __u32 ext4_free_group_clusters(struct super_block *sb,
168 struct ext4_group_desc *bg)
169 {
170 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
171 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
172 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
173 }
174
175 __u32 ext4_free_inodes_count(struct super_block *sb,
176 struct ext4_group_desc *bg)
177 {
178 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
179 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
180 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
181 }
182
183 __u32 ext4_used_dirs_count(struct super_block *sb,
184 struct ext4_group_desc *bg)
185 {
186 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
187 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
188 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
189 }
190
191 __u32 ext4_itable_unused_count(struct super_block *sb,
192 struct ext4_group_desc *bg)
193 {
194 return le16_to_cpu(bg->bg_itable_unused_lo) |
195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
196 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
197 }
198
199 void ext4_block_bitmap_set(struct super_block *sb,
200 struct ext4_group_desc *bg, ext4_fsblk_t blk)
201 {
202 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
205 }
206
207 void ext4_inode_bitmap_set(struct super_block *sb,
208 struct ext4_group_desc *bg, ext4_fsblk_t blk)
209 {
210 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
211 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
212 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
213 }
214
215 void ext4_inode_table_set(struct super_block *sb,
216 struct ext4_group_desc *bg, ext4_fsblk_t blk)
217 {
218 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
219 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
220 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
221 }
222
223 void ext4_free_group_clusters_set(struct super_block *sb,
224 struct ext4_group_desc *bg, __u32 count)
225 {
226 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
227 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
228 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
229 }
230
231 void ext4_free_inodes_set(struct super_block *sb,
232 struct ext4_group_desc *bg, __u32 count)
233 {
234 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
235 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
236 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
237 }
238
239 void ext4_used_dirs_set(struct super_block *sb,
240 struct ext4_group_desc *bg, __u32 count)
241 {
242 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
243 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
244 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
245 }
246
247 void ext4_itable_unused_set(struct super_block *sb,
248 struct ext4_group_desc *bg, __u32 count)
249 {
250 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
253 }
254
255
256 /* Just increment the non-pointer handle value */
257 static handle_t *ext4_get_nojournal(void)
258 {
259 handle_t *handle = current->journal_info;
260 unsigned long ref_cnt = (unsigned long)handle;
261
262 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
263
264 ref_cnt++;
265 handle = (handle_t *)ref_cnt;
266
267 current->journal_info = handle;
268 return handle;
269 }
270
271
272 /* Decrement the non-pointer handle value */
273 static void ext4_put_nojournal(handle_t *handle)
274 {
275 unsigned long ref_cnt = (unsigned long)handle;
276
277 BUG_ON(ref_cnt == 0);
278
279 ref_cnt--;
280 handle = (handle_t *)ref_cnt;
281
282 current->journal_info = handle;
283 }
284
285 /*
286 * Wrappers for jbd2_journal_start/end.
287 *
288 * The only special thing we need to do here is to make sure that all
289 * journal_end calls result in the superblock being marked dirty, so
290 * that sync() will call the filesystem's write_super callback if
291 * appropriate.
292 *
293 * To avoid j_barrier hold in userspace when a user calls freeze(),
294 * ext4 prevents a new handle from being started by s_frozen, which
295 * is in an upper layer.
296 */
297 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
298 {
299 journal_t *journal;
300 handle_t *handle;
301
302 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
303 if (sb->s_flags & MS_RDONLY)
304 return ERR_PTR(-EROFS);
305
306 journal = EXT4_SB(sb)->s_journal;
307 handle = ext4_journal_current_handle();
308
309 /*
310 * If a handle has been started, it should be allowed to
311 * finish, otherwise deadlock could happen between freeze
312 * and others(e.g. truncate) due to the restart of the
313 * journal handle if the filesystem is forzen and active
314 * handles are not stopped.
315 */
316 if (!handle)
317 vfs_check_frozen(sb, SB_FREEZE_TRANS);
318
319 if (!journal)
320 return ext4_get_nojournal();
321 /*
322 * Special case here: if the journal has aborted behind our
323 * backs (eg. EIO in the commit thread), then we still need to
324 * take the FS itself readonly cleanly.
325 */
326 if (is_journal_aborted(journal)) {
327 ext4_abort(sb, "Detected aborted journal");
328 return ERR_PTR(-EROFS);
329 }
330 return jbd2_journal_start(journal, nblocks);
331 }
332
333 /*
334 * The only special thing we need to do here is to make sure that all
335 * jbd2_journal_stop calls result in the superblock being marked dirty, so
336 * that sync() will call the filesystem's write_super callback if
337 * appropriate.
338 */
339 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
340 {
341 struct super_block *sb;
342 int err;
343 int rc;
344
345 if (!ext4_handle_valid(handle)) {
346 ext4_put_nojournal(handle);
347 return 0;
348 }
349 sb = handle->h_transaction->t_journal->j_private;
350 err = handle->h_err;
351 rc = jbd2_journal_stop(handle);
352
353 if (!err)
354 err = rc;
355 if (err)
356 __ext4_std_error(sb, where, line, err);
357 return err;
358 }
359
360 void ext4_journal_abort_handle(const char *caller, unsigned int line,
361 const char *err_fn, struct buffer_head *bh,
362 handle_t *handle, int err)
363 {
364 char nbuf[16];
365 const char *errstr = ext4_decode_error(NULL, err, nbuf);
366
367 BUG_ON(!ext4_handle_valid(handle));
368
369 if (bh)
370 BUFFER_TRACE(bh, "abort");
371
372 if (!handle->h_err)
373 handle->h_err = err;
374
375 if (is_handle_aborted(handle))
376 return;
377
378 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
379 caller, line, errstr, err_fn);
380
381 jbd2_journal_abort_handle(handle);
382 }
383
384 static void __save_error_info(struct super_block *sb, const char *func,
385 unsigned int line)
386 {
387 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
388
389 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
390 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
391 es->s_last_error_time = cpu_to_le32(get_seconds());
392 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
393 es->s_last_error_line = cpu_to_le32(line);
394 if (!es->s_first_error_time) {
395 es->s_first_error_time = es->s_last_error_time;
396 strncpy(es->s_first_error_func, func,
397 sizeof(es->s_first_error_func));
398 es->s_first_error_line = cpu_to_le32(line);
399 es->s_first_error_ino = es->s_last_error_ino;
400 es->s_first_error_block = es->s_last_error_block;
401 }
402 /*
403 * Start the daily error reporting function if it hasn't been
404 * started already
405 */
406 if (!es->s_error_count)
407 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
408 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
409 }
410
411 static void save_error_info(struct super_block *sb, const char *func,
412 unsigned int line)
413 {
414 __save_error_info(sb, func, line);
415 ext4_commit_super(sb, 1);
416 }
417
418 /*
419 * The del_gendisk() function uninitializes the disk-specific data
420 * structures, including the bdi structure, without telling anyone
421 * else. Once this happens, any attempt to call mark_buffer_dirty()
422 * (for example, by ext4_commit_super), will cause a kernel OOPS.
423 * This is a kludge to prevent these oops until we can put in a proper
424 * hook in del_gendisk() to inform the VFS and file system layers.
425 */
426 static int block_device_ejected(struct super_block *sb)
427 {
428 struct inode *bd_inode = sb->s_bdev->bd_inode;
429 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
430
431 return bdi->dev == NULL;
432 }
433
434 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
435 {
436 struct super_block *sb = journal->j_private;
437 struct ext4_sb_info *sbi = EXT4_SB(sb);
438 int error = is_journal_aborted(journal);
439 struct ext4_journal_cb_entry *jce, *tmp;
440
441 spin_lock(&sbi->s_md_lock);
442 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
443 list_del_init(&jce->jce_list);
444 spin_unlock(&sbi->s_md_lock);
445 jce->jce_func(sb, jce, error);
446 spin_lock(&sbi->s_md_lock);
447 }
448 spin_unlock(&sbi->s_md_lock);
449 }
450
451 /* Deal with the reporting of failure conditions on a filesystem such as
452 * inconsistencies detected or read IO failures.
453 *
454 * On ext2, we can store the error state of the filesystem in the
455 * superblock. That is not possible on ext4, because we may have other
456 * write ordering constraints on the superblock which prevent us from
457 * writing it out straight away; and given that the journal is about to
458 * be aborted, we can't rely on the current, or future, transactions to
459 * write out the superblock safely.
460 *
461 * We'll just use the jbd2_journal_abort() error code to record an error in
462 * the journal instead. On recovery, the journal will complain about
463 * that error until we've noted it down and cleared it.
464 */
465
466 static void ext4_handle_error(struct super_block *sb)
467 {
468 if (sb->s_flags & MS_RDONLY)
469 return;
470
471 if (!test_opt(sb, ERRORS_CONT)) {
472 journal_t *journal = EXT4_SB(sb)->s_journal;
473
474 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
475 if (journal)
476 jbd2_journal_abort(journal, -EIO);
477 }
478 if (test_opt(sb, ERRORS_RO)) {
479 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
480 sb->s_flags |= MS_RDONLY;
481 }
482 if (test_opt(sb, ERRORS_PANIC))
483 panic("EXT4-fs (device %s): panic forced after error\n",
484 sb->s_id);
485 }
486
487 void __ext4_error(struct super_block *sb, const char *function,
488 unsigned int line, const char *fmt, ...)
489 {
490 struct va_format vaf;
491 va_list args;
492
493 va_start(args, fmt);
494 vaf.fmt = fmt;
495 vaf.va = &args;
496 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
497 sb->s_id, function, line, current->comm, &vaf);
498 va_end(args);
499
500 ext4_handle_error(sb);
501 }
502
503 void ext4_error_inode(struct inode *inode, const char *function,
504 unsigned int line, ext4_fsblk_t block,
505 const char *fmt, ...)
506 {
507 va_list args;
508 struct va_format vaf;
509 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
510
511 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
512 es->s_last_error_block = cpu_to_le64(block);
513 save_error_info(inode->i_sb, function, line);
514 va_start(args, fmt);
515 vaf.fmt = fmt;
516 vaf.va = &args;
517 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
518 inode->i_sb->s_id, function, line, inode->i_ino);
519 if (block)
520 printk(KERN_CONT "block %llu: ", block);
521 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
522 va_end(args);
523
524 ext4_handle_error(inode->i_sb);
525 }
526
527 void ext4_error_file(struct file *file, const char *function,
528 unsigned int line, ext4_fsblk_t block,
529 const char *fmt, ...)
530 {
531 va_list args;
532 struct va_format vaf;
533 struct ext4_super_block *es;
534 struct inode *inode = file->f_dentry->d_inode;
535 char pathname[80], *path;
536
537 es = EXT4_SB(inode->i_sb)->s_es;
538 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
539 save_error_info(inode->i_sb, function, line);
540 path = d_path(&(file->f_path), pathname, sizeof(pathname));
541 if (IS_ERR(path))
542 path = "(unknown)";
543 printk(KERN_CRIT
544 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
545 inode->i_sb->s_id, function, line, inode->i_ino);
546 if (block)
547 printk(KERN_CONT "block %llu: ", block);
548 va_start(args, fmt);
549 vaf.fmt = fmt;
550 vaf.va = &args;
551 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
552 va_end(args);
553
554 ext4_handle_error(inode->i_sb);
555 }
556
557 static const char *ext4_decode_error(struct super_block *sb, int errno,
558 char nbuf[16])
559 {
560 char *errstr = NULL;
561
562 switch (errno) {
563 case -EIO:
564 errstr = "IO failure";
565 break;
566 case -ENOMEM:
567 errstr = "Out of memory";
568 break;
569 case -EROFS:
570 if (!sb || (EXT4_SB(sb)->s_journal &&
571 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
572 errstr = "Journal has aborted";
573 else
574 errstr = "Readonly filesystem";
575 break;
576 default:
577 /* If the caller passed in an extra buffer for unknown
578 * errors, textualise them now. Else we just return
579 * NULL. */
580 if (nbuf) {
581 /* Check for truncated error codes... */
582 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
583 errstr = nbuf;
584 }
585 break;
586 }
587
588 return errstr;
589 }
590
591 /* __ext4_std_error decodes expected errors from journaling functions
592 * automatically and invokes the appropriate error response. */
593
594 void __ext4_std_error(struct super_block *sb, const char *function,
595 unsigned int line, int errno)
596 {
597 char nbuf[16];
598 const char *errstr;
599
600 /* Special case: if the error is EROFS, and we're not already
601 * inside a transaction, then there's really no point in logging
602 * an error. */
603 if (errno == -EROFS && journal_current_handle() == NULL &&
604 (sb->s_flags & MS_RDONLY))
605 return;
606
607 errstr = ext4_decode_error(sb, errno, nbuf);
608 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
609 sb->s_id, function, line, errstr);
610 save_error_info(sb, function, line);
611
612 ext4_handle_error(sb);
613 }
614
615 /*
616 * ext4_abort is a much stronger failure handler than ext4_error. The
617 * abort function may be used to deal with unrecoverable failures such
618 * as journal IO errors or ENOMEM at a critical moment in log management.
619 *
620 * We unconditionally force the filesystem into an ABORT|READONLY state,
621 * unless the error response on the fs has been set to panic in which
622 * case we take the easy way out and panic immediately.
623 */
624
625 void __ext4_abort(struct super_block *sb, const char *function,
626 unsigned int line, const char *fmt, ...)
627 {
628 va_list args;
629
630 save_error_info(sb, function, line);
631 va_start(args, fmt);
632 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
633 function, line);
634 vprintk(fmt, args);
635 printk("\n");
636 va_end(args);
637
638 if ((sb->s_flags & MS_RDONLY) == 0) {
639 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
640 sb->s_flags |= MS_RDONLY;
641 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
642 if (EXT4_SB(sb)->s_journal)
643 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
644 save_error_info(sb, function, line);
645 }
646 if (test_opt(sb, ERRORS_PANIC))
647 panic("EXT4-fs panic from previous error\n");
648 }
649
650 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
651 {
652 struct va_format vaf;
653 va_list args;
654
655 va_start(args, fmt);
656 vaf.fmt = fmt;
657 vaf.va = &args;
658 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
659 va_end(args);
660 }
661
662 void __ext4_warning(struct super_block *sb, const char *function,
663 unsigned int line, const char *fmt, ...)
664 {
665 struct va_format vaf;
666 va_list args;
667
668 va_start(args, fmt);
669 vaf.fmt = fmt;
670 vaf.va = &args;
671 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
672 sb->s_id, function, line, &vaf);
673 va_end(args);
674 }
675
676 void __ext4_grp_locked_error(const char *function, unsigned int line,
677 struct super_block *sb, ext4_group_t grp,
678 unsigned long ino, ext4_fsblk_t block,
679 const char *fmt, ...)
680 __releases(bitlock)
681 __acquires(bitlock)
682 {
683 struct va_format vaf;
684 va_list args;
685 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
686
687 es->s_last_error_ino = cpu_to_le32(ino);
688 es->s_last_error_block = cpu_to_le64(block);
689 __save_error_info(sb, function, line);
690
691 va_start(args, fmt);
692
693 vaf.fmt = fmt;
694 vaf.va = &args;
695 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
696 sb->s_id, function, line, grp);
697 if (ino)
698 printk(KERN_CONT "inode %lu: ", ino);
699 if (block)
700 printk(KERN_CONT "block %llu:", (unsigned long long) block);
701 printk(KERN_CONT "%pV\n", &vaf);
702 va_end(args);
703
704 if (test_opt(sb, ERRORS_CONT)) {
705 ext4_commit_super(sb, 0);
706 return;
707 }
708
709 ext4_unlock_group(sb, grp);
710 ext4_handle_error(sb);
711 /*
712 * We only get here in the ERRORS_RO case; relocking the group
713 * may be dangerous, but nothing bad will happen since the
714 * filesystem will have already been marked read/only and the
715 * journal has been aborted. We return 1 as a hint to callers
716 * who might what to use the return value from
717 * ext4_grp_locked_error() to distinguish between the
718 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
719 * aggressively from the ext4 function in question, with a
720 * more appropriate error code.
721 */
722 ext4_lock_group(sb, grp);
723 return;
724 }
725
726 void ext4_update_dynamic_rev(struct super_block *sb)
727 {
728 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
729
730 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
731 return;
732
733 ext4_warning(sb,
734 "updating to rev %d because of new feature flag, "
735 "running e2fsck is recommended",
736 EXT4_DYNAMIC_REV);
737
738 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
739 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
740 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
741 /* leave es->s_feature_*compat flags alone */
742 /* es->s_uuid will be set by e2fsck if empty */
743
744 /*
745 * The rest of the superblock fields should be zero, and if not it
746 * means they are likely already in use, so leave them alone. We
747 * can leave it up to e2fsck to clean up any inconsistencies there.
748 */
749 }
750
751 /*
752 * Open the external journal device
753 */
754 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
755 {
756 struct block_device *bdev;
757 char b[BDEVNAME_SIZE];
758
759 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
760 if (IS_ERR(bdev))
761 goto fail;
762 return bdev;
763
764 fail:
765 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
766 __bdevname(dev, b), PTR_ERR(bdev));
767 return NULL;
768 }
769
770 /*
771 * Release the journal device
772 */
773 static int ext4_blkdev_put(struct block_device *bdev)
774 {
775 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
776 }
777
778 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
779 {
780 struct block_device *bdev;
781 int ret = -ENODEV;
782
783 bdev = sbi->journal_bdev;
784 if (bdev) {
785 ret = ext4_blkdev_put(bdev);
786 sbi->journal_bdev = NULL;
787 }
788 return ret;
789 }
790
791 static inline struct inode *orphan_list_entry(struct list_head *l)
792 {
793 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
794 }
795
796 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
797 {
798 struct list_head *l;
799
800 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
801 le32_to_cpu(sbi->s_es->s_last_orphan));
802
803 printk(KERN_ERR "sb_info orphan list:\n");
804 list_for_each(l, &sbi->s_orphan) {
805 struct inode *inode = orphan_list_entry(l);
806 printk(KERN_ERR " "
807 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
808 inode->i_sb->s_id, inode->i_ino, inode,
809 inode->i_mode, inode->i_nlink,
810 NEXT_ORPHAN(inode));
811 }
812 }
813
814 static void ext4_put_super(struct super_block *sb)
815 {
816 struct ext4_sb_info *sbi = EXT4_SB(sb);
817 struct ext4_super_block *es = sbi->s_es;
818 int i, err;
819
820 ext4_unregister_li_request(sb);
821 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
822
823 flush_workqueue(sbi->dio_unwritten_wq);
824 destroy_workqueue(sbi->dio_unwritten_wq);
825
826 lock_super(sb);
827 if (sb->s_dirt)
828 ext4_commit_super(sb, 1);
829
830 if (sbi->s_journal) {
831 err = jbd2_journal_destroy(sbi->s_journal);
832 sbi->s_journal = NULL;
833 if (err < 0)
834 ext4_abort(sb, "Couldn't clean up the journal");
835 }
836
837 del_timer(&sbi->s_err_report);
838 ext4_release_system_zone(sb);
839 ext4_mb_release(sb);
840 ext4_ext_release(sb);
841 ext4_xattr_put_super(sb);
842
843 if (!(sb->s_flags & MS_RDONLY)) {
844 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
845 es->s_state = cpu_to_le16(sbi->s_mount_state);
846 ext4_commit_super(sb, 1);
847 }
848 if (sbi->s_proc) {
849 remove_proc_entry(sb->s_id, ext4_proc_root);
850 }
851 kobject_del(&sbi->s_kobj);
852
853 for (i = 0; i < sbi->s_gdb_count; i++)
854 brelse(sbi->s_group_desc[i]);
855 ext4_kvfree(sbi->s_group_desc);
856 ext4_kvfree(sbi->s_flex_groups);
857 percpu_counter_destroy(&sbi->s_freeclusters_counter);
858 percpu_counter_destroy(&sbi->s_freeinodes_counter);
859 percpu_counter_destroy(&sbi->s_dirs_counter);
860 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
861 brelse(sbi->s_sbh);
862 #ifdef CONFIG_QUOTA
863 for (i = 0; i < MAXQUOTAS; i++)
864 kfree(sbi->s_qf_names[i]);
865 #endif
866
867 /* Debugging code just in case the in-memory inode orphan list
868 * isn't empty. The on-disk one can be non-empty if we've
869 * detected an error and taken the fs readonly, but the
870 * in-memory list had better be clean by this point. */
871 if (!list_empty(&sbi->s_orphan))
872 dump_orphan_list(sb, sbi);
873 J_ASSERT(list_empty(&sbi->s_orphan));
874
875 invalidate_bdev(sb->s_bdev);
876 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
877 /*
878 * Invalidate the journal device's buffers. We don't want them
879 * floating about in memory - the physical journal device may
880 * hotswapped, and it breaks the `ro-after' testing code.
881 */
882 sync_blockdev(sbi->journal_bdev);
883 invalidate_bdev(sbi->journal_bdev);
884 ext4_blkdev_remove(sbi);
885 }
886 if (sbi->s_mmp_tsk)
887 kthread_stop(sbi->s_mmp_tsk);
888 sb->s_fs_info = NULL;
889 /*
890 * Now that we are completely done shutting down the
891 * superblock, we need to actually destroy the kobject.
892 */
893 unlock_super(sb);
894 kobject_put(&sbi->s_kobj);
895 wait_for_completion(&sbi->s_kobj_unregister);
896 kfree(sbi->s_blockgroup_lock);
897 kfree(sbi);
898 }
899
900 static struct kmem_cache *ext4_inode_cachep;
901
902 /*
903 * Called inside transaction, so use GFP_NOFS
904 */
905 static struct inode *ext4_alloc_inode(struct super_block *sb)
906 {
907 struct ext4_inode_info *ei;
908
909 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
910 if (!ei)
911 return NULL;
912
913 ei->vfs_inode.i_version = 1;
914 ei->vfs_inode.i_data.writeback_index = 0;
915 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
916 INIT_LIST_HEAD(&ei->i_prealloc_list);
917 spin_lock_init(&ei->i_prealloc_lock);
918 ei->i_reserved_data_blocks = 0;
919 ei->i_reserved_meta_blocks = 0;
920 ei->i_allocated_meta_blocks = 0;
921 ei->i_da_metadata_calc_len = 0;
922 spin_lock_init(&(ei->i_block_reservation_lock));
923 #ifdef CONFIG_QUOTA
924 ei->i_reserved_quota = 0;
925 #endif
926 ei->jinode = NULL;
927 INIT_LIST_HEAD(&ei->i_completed_io_list);
928 spin_lock_init(&ei->i_completed_io_lock);
929 ei->cur_aio_dio = NULL;
930 ei->i_sync_tid = 0;
931 ei->i_datasync_tid = 0;
932 atomic_set(&ei->i_ioend_count, 0);
933 atomic_set(&ei->i_aiodio_unwritten, 0);
934
935 return &ei->vfs_inode;
936 }
937
938 static int ext4_drop_inode(struct inode *inode)
939 {
940 int drop = generic_drop_inode(inode);
941
942 trace_ext4_drop_inode(inode, drop);
943 return drop;
944 }
945
946 static void ext4_i_callback(struct rcu_head *head)
947 {
948 struct inode *inode = container_of(head, struct inode, i_rcu);
949 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
950 }
951
952 static void ext4_destroy_inode(struct inode *inode)
953 {
954 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
955 ext4_msg(inode->i_sb, KERN_ERR,
956 "Inode %lu (%p): orphan list check failed!",
957 inode->i_ino, EXT4_I(inode));
958 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
959 EXT4_I(inode), sizeof(struct ext4_inode_info),
960 true);
961 dump_stack();
962 }
963 call_rcu(&inode->i_rcu, ext4_i_callback);
964 }
965
966 static void init_once(void *foo)
967 {
968 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
969
970 INIT_LIST_HEAD(&ei->i_orphan);
971 #ifdef CONFIG_EXT4_FS_XATTR
972 init_rwsem(&ei->xattr_sem);
973 #endif
974 init_rwsem(&ei->i_data_sem);
975 inode_init_once(&ei->vfs_inode);
976 }
977
978 static int init_inodecache(void)
979 {
980 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
981 sizeof(struct ext4_inode_info),
982 0, (SLAB_RECLAIM_ACCOUNT|
983 SLAB_MEM_SPREAD),
984 init_once);
985 if (ext4_inode_cachep == NULL)
986 return -ENOMEM;
987 return 0;
988 }
989
990 static void destroy_inodecache(void)
991 {
992 kmem_cache_destroy(ext4_inode_cachep);
993 }
994
995 void ext4_clear_inode(struct inode *inode)
996 {
997 invalidate_inode_buffers(inode);
998 end_writeback(inode);
999 dquot_drop(inode);
1000 ext4_discard_preallocations(inode);
1001 if (EXT4_I(inode)->jinode) {
1002 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1003 EXT4_I(inode)->jinode);
1004 jbd2_free_inode(EXT4_I(inode)->jinode);
1005 EXT4_I(inode)->jinode = NULL;
1006 }
1007 }
1008
1009 static inline void ext4_show_quota_options(struct seq_file *seq,
1010 struct super_block *sb)
1011 {
1012 #if defined(CONFIG_QUOTA)
1013 struct ext4_sb_info *sbi = EXT4_SB(sb);
1014
1015 if (sbi->s_jquota_fmt) {
1016 char *fmtname = "";
1017
1018 switch (sbi->s_jquota_fmt) {
1019 case QFMT_VFS_OLD:
1020 fmtname = "vfsold";
1021 break;
1022 case QFMT_VFS_V0:
1023 fmtname = "vfsv0";
1024 break;
1025 case QFMT_VFS_V1:
1026 fmtname = "vfsv1";
1027 break;
1028 }
1029 seq_printf(seq, ",jqfmt=%s", fmtname);
1030 }
1031
1032 if (sbi->s_qf_names[USRQUOTA])
1033 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1034
1035 if (sbi->s_qf_names[GRPQUOTA])
1036 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1037
1038 if (test_opt(sb, USRQUOTA))
1039 seq_puts(seq, ",usrquota");
1040
1041 if (test_opt(sb, GRPQUOTA))
1042 seq_puts(seq, ",grpquota");
1043 #endif
1044 }
1045
1046 /*
1047 * Show an option if
1048 * - it's set to a non-default value OR
1049 * - if the per-sb default is different from the global default
1050 */
1051 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1052 {
1053 int def_errors;
1054 unsigned long def_mount_opts;
1055 struct super_block *sb = root->d_sb;
1056 struct ext4_sb_info *sbi = EXT4_SB(sb);
1057 struct ext4_super_block *es = sbi->s_es;
1058
1059 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1060 def_errors = le16_to_cpu(es->s_errors);
1061
1062 if (sbi->s_sb_block != 1)
1063 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1064 if (test_opt(sb, MINIX_DF))
1065 seq_puts(seq, ",minixdf");
1066 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1067 seq_puts(seq, ",grpid");
1068 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1069 seq_puts(seq, ",nogrpid");
1070 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1071 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1072 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1073 }
1074 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1075 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1076 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1077 }
1078 if (test_opt(sb, ERRORS_RO)) {
1079 if (def_errors == EXT4_ERRORS_PANIC ||
1080 def_errors == EXT4_ERRORS_CONTINUE) {
1081 seq_puts(seq, ",errors=remount-ro");
1082 }
1083 }
1084 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1085 seq_puts(seq, ",errors=continue");
1086 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1087 seq_puts(seq, ",errors=panic");
1088 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1089 seq_puts(seq, ",nouid32");
1090 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1091 seq_puts(seq, ",debug");
1092 #ifdef CONFIG_EXT4_FS_XATTR
1093 if (test_opt(sb, XATTR_USER))
1094 seq_puts(seq, ",user_xattr");
1095 if (!test_opt(sb, XATTR_USER))
1096 seq_puts(seq, ",nouser_xattr");
1097 #endif
1098 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1099 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1100 seq_puts(seq, ",acl");
1101 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1102 seq_puts(seq, ",noacl");
1103 #endif
1104 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1105 seq_printf(seq, ",commit=%u",
1106 (unsigned) (sbi->s_commit_interval / HZ));
1107 }
1108 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1109 seq_printf(seq, ",min_batch_time=%u",
1110 (unsigned) sbi->s_min_batch_time);
1111 }
1112 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1113 seq_printf(seq, ",max_batch_time=%u",
1114 (unsigned) sbi->s_max_batch_time);
1115 }
1116
1117 /*
1118 * We're changing the default of barrier mount option, so
1119 * let's always display its mount state so it's clear what its
1120 * status is.
1121 */
1122 seq_puts(seq, ",barrier=");
1123 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1124 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1125 seq_puts(seq, ",journal_async_commit");
1126 else if (test_opt(sb, JOURNAL_CHECKSUM))
1127 seq_puts(seq, ",journal_checksum");
1128 if (sb->s_flags & MS_I_VERSION)
1129 seq_puts(seq, ",i_version");
1130 if (!test_opt(sb, DELALLOC) &&
1131 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1132 seq_puts(seq, ",nodelalloc");
1133
1134 if (!test_opt(sb, MBLK_IO_SUBMIT))
1135 seq_puts(seq, ",nomblk_io_submit");
1136 if (sbi->s_stripe)
1137 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1138 /*
1139 * journal mode get enabled in different ways
1140 * So just print the value even if we didn't specify it
1141 */
1142 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1143 seq_puts(seq, ",data=journal");
1144 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1145 seq_puts(seq, ",data=ordered");
1146 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1147 seq_puts(seq, ",data=writeback");
1148
1149 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1150 seq_printf(seq, ",inode_readahead_blks=%u",
1151 sbi->s_inode_readahead_blks);
1152
1153 if (test_opt(sb, DATA_ERR_ABORT))
1154 seq_puts(seq, ",data_err=abort");
1155
1156 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1157 seq_puts(seq, ",noauto_da_alloc");
1158
1159 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1160 seq_puts(seq, ",discard");
1161
1162 if (test_opt(sb, NOLOAD))
1163 seq_puts(seq, ",norecovery");
1164
1165 if (test_opt(sb, DIOREAD_NOLOCK))
1166 seq_puts(seq, ",dioread_nolock");
1167
1168 if (test_opt(sb, BLOCK_VALIDITY) &&
1169 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1170 seq_puts(seq, ",block_validity");
1171
1172 if (!test_opt(sb, INIT_INODE_TABLE))
1173 seq_puts(seq, ",noinit_itable");
1174 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1175 seq_printf(seq, ",init_itable=%u",
1176 (unsigned) sbi->s_li_wait_mult);
1177
1178 ext4_show_quota_options(seq, sb);
1179
1180 return 0;
1181 }
1182
1183 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1184 u64 ino, u32 generation)
1185 {
1186 struct inode *inode;
1187
1188 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1189 return ERR_PTR(-ESTALE);
1190 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1191 return ERR_PTR(-ESTALE);
1192
1193 /* iget isn't really right if the inode is currently unallocated!!
1194 *
1195 * ext4_read_inode will return a bad_inode if the inode had been
1196 * deleted, so we should be safe.
1197 *
1198 * Currently we don't know the generation for parent directory, so
1199 * a generation of 0 means "accept any"
1200 */
1201 inode = ext4_iget(sb, ino);
1202 if (IS_ERR(inode))
1203 return ERR_CAST(inode);
1204 if (generation && inode->i_generation != generation) {
1205 iput(inode);
1206 return ERR_PTR(-ESTALE);
1207 }
1208
1209 return inode;
1210 }
1211
1212 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1213 int fh_len, int fh_type)
1214 {
1215 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1216 ext4_nfs_get_inode);
1217 }
1218
1219 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1220 int fh_len, int fh_type)
1221 {
1222 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1223 ext4_nfs_get_inode);
1224 }
1225
1226 /*
1227 * Try to release metadata pages (indirect blocks, directories) which are
1228 * mapped via the block device. Since these pages could have journal heads
1229 * which would prevent try_to_free_buffers() from freeing them, we must use
1230 * jbd2 layer's try_to_free_buffers() function to release them.
1231 */
1232 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1233 gfp_t wait)
1234 {
1235 journal_t *journal = EXT4_SB(sb)->s_journal;
1236
1237 WARN_ON(PageChecked(page));
1238 if (!page_has_buffers(page))
1239 return 0;
1240 if (journal)
1241 return jbd2_journal_try_to_free_buffers(journal, page,
1242 wait & ~__GFP_WAIT);
1243 return try_to_free_buffers(page);
1244 }
1245
1246 #ifdef CONFIG_QUOTA
1247 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1248 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1249
1250 static int ext4_write_dquot(struct dquot *dquot);
1251 static int ext4_acquire_dquot(struct dquot *dquot);
1252 static int ext4_release_dquot(struct dquot *dquot);
1253 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1254 static int ext4_write_info(struct super_block *sb, int type);
1255 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1256 struct path *path);
1257 static int ext4_quota_off(struct super_block *sb, int type);
1258 static int ext4_quota_on_mount(struct super_block *sb, int type);
1259 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1260 size_t len, loff_t off);
1261 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1262 const char *data, size_t len, loff_t off);
1263
1264 static const struct dquot_operations ext4_quota_operations = {
1265 .get_reserved_space = ext4_get_reserved_space,
1266 .write_dquot = ext4_write_dquot,
1267 .acquire_dquot = ext4_acquire_dquot,
1268 .release_dquot = ext4_release_dquot,
1269 .mark_dirty = ext4_mark_dquot_dirty,
1270 .write_info = ext4_write_info,
1271 .alloc_dquot = dquot_alloc,
1272 .destroy_dquot = dquot_destroy,
1273 };
1274
1275 static const struct quotactl_ops ext4_qctl_operations = {
1276 .quota_on = ext4_quota_on,
1277 .quota_off = ext4_quota_off,
1278 .quota_sync = dquot_quota_sync,
1279 .get_info = dquot_get_dqinfo,
1280 .set_info = dquot_set_dqinfo,
1281 .get_dqblk = dquot_get_dqblk,
1282 .set_dqblk = dquot_set_dqblk
1283 };
1284 #endif
1285
1286 static const struct super_operations ext4_sops = {
1287 .alloc_inode = ext4_alloc_inode,
1288 .destroy_inode = ext4_destroy_inode,
1289 .write_inode = ext4_write_inode,
1290 .dirty_inode = ext4_dirty_inode,
1291 .drop_inode = ext4_drop_inode,
1292 .evict_inode = ext4_evict_inode,
1293 .put_super = ext4_put_super,
1294 .sync_fs = ext4_sync_fs,
1295 .freeze_fs = ext4_freeze,
1296 .unfreeze_fs = ext4_unfreeze,
1297 .statfs = ext4_statfs,
1298 .remount_fs = ext4_remount,
1299 .show_options = ext4_show_options,
1300 #ifdef CONFIG_QUOTA
1301 .quota_read = ext4_quota_read,
1302 .quota_write = ext4_quota_write,
1303 #endif
1304 .bdev_try_to_free_page = bdev_try_to_free_page,
1305 };
1306
1307 static const struct super_operations ext4_nojournal_sops = {
1308 .alloc_inode = ext4_alloc_inode,
1309 .destroy_inode = ext4_destroy_inode,
1310 .write_inode = ext4_write_inode,
1311 .dirty_inode = ext4_dirty_inode,
1312 .drop_inode = ext4_drop_inode,
1313 .evict_inode = ext4_evict_inode,
1314 .write_super = ext4_write_super,
1315 .put_super = ext4_put_super,
1316 .statfs = ext4_statfs,
1317 .remount_fs = ext4_remount,
1318 .show_options = ext4_show_options,
1319 #ifdef CONFIG_QUOTA
1320 .quota_read = ext4_quota_read,
1321 .quota_write = ext4_quota_write,
1322 #endif
1323 .bdev_try_to_free_page = bdev_try_to_free_page,
1324 };
1325
1326 static const struct export_operations ext4_export_ops = {
1327 .fh_to_dentry = ext4_fh_to_dentry,
1328 .fh_to_parent = ext4_fh_to_parent,
1329 .get_parent = ext4_get_parent,
1330 };
1331
1332 enum {
1333 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1334 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1335 Opt_nouid32, Opt_debug, Opt_removed,
1336 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1337 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1338 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1339 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1340 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1341 Opt_data_err_abort, Opt_data_err_ignore,
1342 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1343 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1344 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1345 Opt_usrquota, Opt_grpquota, Opt_i_version,
1346 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1347 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1348 Opt_inode_readahead_blks, Opt_journal_ioprio,
1349 Opt_dioread_nolock, Opt_dioread_lock,
1350 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1351 };
1352
1353 static const match_table_t tokens = {
1354 {Opt_bsd_df, "bsddf"},
1355 {Opt_minix_df, "minixdf"},
1356 {Opt_grpid, "grpid"},
1357 {Opt_grpid, "bsdgroups"},
1358 {Opt_nogrpid, "nogrpid"},
1359 {Opt_nogrpid, "sysvgroups"},
1360 {Opt_resgid, "resgid=%u"},
1361 {Opt_resuid, "resuid=%u"},
1362 {Opt_sb, "sb=%u"},
1363 {Opt_err_cont, "errors=continue"},
1364 {Opt_err_panic, "errors=panic"},
1365 {Opt_err_ro, "errors=remount-ro"},
1366 {Opt_nouid32, "nouid32"},
1367 {Opt_debug, "debug"},
1368 {Opt_removed, "oldalloc"},
1369 {Opt_removed, "orlov"},
1370 {Opt_user_xattr, "user_xattr"},
1371 {Opt_nouser_xattr, "nouser_xattr"},
1372 {Opt_acl, "acl"},
1373 {Opt_noacl, "noacl"},
1374 {Opt_noload, "noload"},
1375 {Opt_noload, "norecovery"},
1376 {Opt_removed, "nobh"},
1377 {Opt_removed, "bh"},
1378 {Opt_commit, "commit=%u"},
1379 {Opt_min_batch_time, "min_batch_time=%u"},
1380 {Opt_max_batch_time, "max_batch_time=%u"},
1381 {Opt_journal_dev, "journal_dev=%u"},
1382 {Opt_journal_checksum, "journal_checksum"},
1383 {Opt_journal_async_commit, "journal_async_commit"},
1384 {Opt_abort, "abort"},
1385 {Opt_data_journal, "data=journal"},
1386 {Opt_data_ordered, "data=ordered"},
1387 {Opt_data_writeback, "data=writeback"},
1388 {Opt_data_err_abort, "data_err=abort"},
1389 {Opt_data_err_ignore, "data_err=ignore"},
1390 {Opt_offusrjquota, "usrjquota="},
1391 {Opt_usrjquota, "usrjquota=%s"},
1392 {Opt_offgrpjquota, "grpjquota="},
1393 {Opt_grpjquota, "grpjquota=%s"},
1394 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1395 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1396 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1397 {Opt_grpquota, "grpquota"},
1398 {Opt_noquota, "noquota"},
1399 {Opt_quota, "quota"},
1400 {Opt_usrquota, "usrquota"},
1401 {Opt_barrier, "barrier=%u"},
1402 {Opt_barrier, "barrier"},
1403 {Opt_nobarrier, "nobarrier"},
1404 {Opt_i_version, "i_version"},
1405 {Opt_stripe, "stripe=%u"},
1406 {Opt_delalloc, "delalloc"},
1407 {Opt_nodelalloc, "nodelalloc"},
1408 {Opt_mblk_io_submit, "mblk_io_submit"},
1409 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1410 {Opt_block_validity, "block_validity"},
1411 {Opt_noblock_validity, "noblock_validity"},
1412 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1413 {Opt_journal_ioprio, "journal_ioprio=%u"},
1414 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1415 {Opt_auto_da_alloc, "auto_da_alloc"},
1416 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1417 {Opt_dioread_nolock, "dioread_nolock"},
1418 {Opt_dioread_lock, "dioread_lock"},
1419 {Opt_discard, "discard"},
1420 {Opt_nodiscard, "nodiscard"},
1421 {Opt_init_itable, "init_itable=%u"},
1422 {Opt_init_itable, "init_itable"},
1423 {Opt_noinit_itable, "noinit_itable"},
1424 {Opt_err, NULL},
1425 };
1426
1427 static ext4_fsblk_t get_sb_block(void **data)
1428 {
1429 ext4_fsblk_t sb_block;
1430 char *options = (char *) *data;
1431
1432 if (!options || strncmp(options, "sb=", 3) != 0)
1433 return 1; /* Default location */
1434
1435 options += 3;
1436 /* TODO: use simple_strtoll with >32bit ext4 */
1437 sb_block = simple_strtoul(options, &options, 0);
1438 if (*options && *options != ',') {
1439 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1440 (char *) *data);
1441 return 1;
1442 }
1443 if (*options == ',')
1444 options++;
1445 *data = (void *) options;
1446
1447 return sb_block;
1448 }
1449
1450 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1451 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1452 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1453
1454 #ifdef CONFIG_QUOTA
1455 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1456 {
1457 struct ext4_sb_info *sbi = EXT4_SB(sb);
1458 char *qname;
1459
1460 if (sb_any_quota_loaded(sb) &&
1461 !sbi->s_qf_names[qtype]) {
1462 ext4_msg(sb, KERN_ERR,
1463 "Cannot change journaled "
1464 "quota options when quota turned on");
1465 return 0;
1466 }
1467 qname = match_strdup(args);
1468 if (!qname) {
1469 ext4_msg(sb, KERN_ERR,
1470 "Not enough memory for storing quotafile name");
1471 return 0;
1472 }
1473 if (sbi->s_qf_names[qtype] &&
1474 strcmp(sbi->s_qf_names[qtype], qname)) {
1475 ext4_msg(sb, KERN_ERR,
1476 "%s quota file already specified", QTYPE2NAME(qtype));
1477 kfree(qname);
1478 return 0;
1479 }
1480 sbi->s_qf_names[qtype] = qname;
1481 if (strchr(sbi->s_qf_names[qtype], '/')) {
1482 ext4_msg(sb, KERN_ERR,
1483 "quotafile must be on filesystem root");
1484 kfree(sbi->s_qf_names[qtype]);
1485 sbi->s_qf_names[qtype] = NULL;
1486 return 0;
1487 }
1488 set_opt(sb, QUOTA);
1489 return 1;
1490 }
1491
1492 static int clear_qf_name(struct super_block *sb, int qtype)
1493 {
1494
1495 struct ext4_sb_info *sbi = EXT4_SB(sb);
1496
1497 if (sb_any_quota_loaded(sb) &&
1498 sbi->s_qf_names[qtype]) {
1499 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1500 " when quota turned on");
1501 return 0;
1502 }
1503 /*
1504 * The space will be released later when all options are confirmed
1505 * to be correct
1506 */
1507 sbi->s_qf_names[qtype] = NULL;
1508 return 1;
1509 }
1510 #endif
1511
1512 #define MOPT_SET 0x0001
1513 #define MOPT_CLEAR 0x0002
1514 #define MOPT_NOSUPPORT 0x0004
1515 #define MOPT_EXPLICIT 0x0008
1516 #define MOPT_CLEAR_ERR 0x0010
1517 #define MOPT_GTE0 0x0020
1518 #ifdef CONFIG_QUOTA
1519 #define MOPT_Q 0
1520 #define MOPT_QFMT 0x0040
1521 #else
1522 #define MOPT_Q MOPT_NOSUPPORT
1523 #define MOPT_QFMT MOPT_NOSUPPORT
1524 #endif
1525 #define MOPT_DATAJ 0x0080
1526
1527 static const struct mount_opts {
1528 int token;
1529 int mount_opt;
1530 int flags;
1531 } ext4_mount_opts[] = {
1532 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1533 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1534 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1535 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1536 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1537 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1538 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1539 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1540 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1541 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1542 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1543 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1544 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1545 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1546 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1547 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1548 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1549 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1550 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1551 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1552 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1553 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1554 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1555 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1556 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1557 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1558 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1559 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1560 {Opt_commit, 0, MOPT_GTE0},
1561 {Opt_max_batch_time, 0, MOPT_GTE0},
1562 {Opt_min_batch_time, 0, MOPT_GTE0},
1563 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1564 {Opt_init_itable, 0, MOPT_GTE0},
1565 {Opt_stripe, 0, MOPT_GTE0},
1566 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1567 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1568 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1569 #ifdef CONFIG_EXT4_FS_XATTR
1570 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1571 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1572 #else
1573 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1574 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
1575 #endif
1576 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1577 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1578 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1579 #else
1580 {Opt_acl, 0, MOPT_NOSUPPORT},
1581 {Opt_noacl, 0, MOPT_NOSUPPORT},
1582 #endif
1583 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1584 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1585 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1586 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1587 MOPT_SET | MOPT_Q},
1588 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1589 MOPT_SET | MOPT_Q},
1590 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1591 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1592 {Opt_usrjquota, 0, MOPT_Q},
1593 {Opt_grpjquota, 0, MOPT_Q},
1594 {Opt_offusrjquota, 0, MOPT_Q},
1595 {Opt_offgrpjquota, 0, MOPT_Q},
1596 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1597 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1598 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1599 {Opt_err, 0, 0}
1600 };
1601
1602 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1603 substring_t *args, unsigned long *journal_devnum,
1604 unsigned int *journal_ioprio, int is_remount)
1605 {
1606 struct ext4_sb_info *sbi = EXT4_SB(sb);
1607 const struct mount_opts *m;
1608 int arg = 0;
1609
1610 if (args->from && match_int(args, &arg))
1611 return -1;
1612 switch (token) {
1613 case Opt_sb:
1614 return 1; /* handled by get_sb_block() */
1615 case Opt_removed:
1616 ext4_msg(sb, KERN_WARNING,
1617 "Ignoring removed %s option", opt);
1618 return 1;
1619 case Opt_resuid:
1620 sbi->s_resuid = arg;
1621 return 1;
1622 case Opt_resgid:
1623 sbi->s_resgid = arg;
1624 return 1;
1625 case Opt_abort:
1626 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1627 return 1;
1628 case Opt_i_version:
1629 sb->s_flags |= MS_I_VERSION;
1630 return 1;
1631 case Opt_journal_dev:
1632 if (is_remount) {
1633 ext4_msg(sb, KERN_ERR,
1634 "Cannot specify journal on remount");
1635 return -1;
1636 }
1637 *journal_devnum = arg;
1638 return 1;
1639 case Opt_journal_ioprio:
1640 if (arg < 0 || arg > 7)
1641 return -1;
1642 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1643 return 1;
1644 }
1645
1646 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1647 if (token != m->token)
1648 continue;
1649 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1650 return -1;
1651 if (m->flags & MOPT_EXPLICIT)
1652 set_opt2(sb, EXPLICIT_DELALLOC);
1653 if (m->flags & MOPT_CLEAR_ERR)
1654 clear_opt(sb, ERRORS_MASK);
1655 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1656 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1657 "options when quota turned on");
1658 return -1;
1659 }
1660
1661 if (m->flags & MOPT_NOSUPPORT) {
1662 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1663 } else if (token == Opt_commit) {
1664 if (arg == 0)
1665 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1666 sbi->s_commit_interval = HZ * arg;
1667 } else if (token == Opt_max_batch_time) {
1668 if (arg == 0)
1669 arg = EXT4_DEF_MAX_BATCH_TIME;
1670 sbi->s_max_batch_time = arg;
1671 } else if (token == Opt_min_batch_time) {
1672 sbi->s_min_batch_time = arg;
1673 } else if (token == Opt_inode_readahead_blks) {
1674 if (arg > (1 << 30))
1675 return -1;
1676 if (arg && !is_power_of_2(arg)) {
1677 ext4_msg(sb, KERN_ERR,
1678 "EXT4-fs: inode_readahead_blks"
1679 " must be a power of 2");
1680 return -1;
1681 }
1682 sbi->s_inode_readahead_blks = arg;
1683 } else if (token == Opt_init_itable) {
1684 set_opt(sb, INIT_INODE_TABLE);
1685 if (!args->from)
1686 arg = EXT4_DEF_LI_WAIT_MULT;
1687 sbi->s_li_wait_mult = arg;
1688 } else if (token == Opt_stripe) {
1689 sbi->s_stripe = arg;
1690 } else if (m->flags & MOPT_DATAJ) {
1691 if (is_remount) {
1692 if (!sbi->s_journal)
1693 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1694 else if (test_opt(sb, DATA_FLAGS) !=
1695 m->mount_opt) {
1696 ext4_msg(sb, KERN_ERR,
1697 "Cannot change data mode on remount");
1698 return -1;
1699 }
1700 } else {
1701 clear_opt(sb, DATA_FLAGS);
1702 sbi->s_mount_opt |= m->mount_opt;
1703 }
1704 #ifdef CONFIG_QUOTA
1705 } else if (token == Opt_usrjquota) {
1706 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1707 return -1;
1708 } else if (token == Opt_grpjquota) {
1709 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1710 return -1;
1711 } else if (token == Opt_offusrjquota) {
1712 if (!clear_qf_name(sb, USRQUOTA))
1713 return -1;
1714 } else if (token == Opt_offgrpjquota) {
1715 if (!clear_qf_name(sb, GRPQUOTA))
1716 return -1;
1717 } else if (m->flags & MOPT_QFMT) {
1718 if (sb_any_quota_loaded(sb) &&
1719 sbi->s_jquota_fmt != m->mount_opt) {
1720 ext4_msg(sb, KERN_ERR, "Cannot "
1721 "change journaled quota options "
1722 "when quota turned on");
1723 return -1;
1724 }
1725 sbi->s_jquota_fmt = m->mount_opt;
1726 #endif
1727 } else {
1728 if (!args->from)
1729 arg = 1;
1730 if (m->flags & MOPT_CLEAR)
1731 arg = !arg;
1732 else if (unlikely(!(m->flags & MOPT_SET))) {
1733 ext4_msg(sb, KERN_WARNING,
1734 "buggy handling of option %s", opt);
1735 WARN_ON(1);
1736 return -1;
1737 }
1738 if (arg != 0)
1739 sbi->s_mount_opt |= m->mount_opt;
1740 else
1741 sbi->s_mount_opt &= ~m->mount_opt;
1742 }
1743 return 1;
1744 }
1745 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1746 "or missing value", opt);
1747 return -1;
1748 }
1749
1750 static int parse_options(char *options, struct super_block *sb,
1751 unsigned long *journal_devnum,
1752 unsigned int *journal_ioprio,
1753 int is_remount)
1754 {
1755 struct ext4_sb_info *sbi = EXT4_SB(sb);
1756 char *p;
1757 substring_t args[MAX_OPT_ARGS];
1758 int token;
1759
1760 if (!options)
1761 return 1;
1762
1763 while ((p = strsep(&options, ",")) != NULL) {
1764 if (!*p)
1765 continue;
1766 /*
1767 * Initialize args struct so we know whether arg was
1768 * found; some options take optional arguments.
1769 */
1770 args[0].to = args[0].from = 0;
1771 token = match_token(p, tokens, args);
1772 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1773 journal_ioprio, is_remount) < 0)
1774 return 0;
1775 }
1776 #ifdef CONFIG_QUOTA
1777 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1778 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1779 clear_opt(sb, USRQUOTA);
1780
1781 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1782 clear_opt(sb, GRPQUOTA);
1783
1784 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1785 ext4_msg(sb, KERN_ERR, "old and new quota "
1786 "format mixing");
1787 return 0;
1788 }
1789
1790 if (!sbi->s_jquota_fmt) {
1791 ext4_msg(sb, KERN_ERR, "journaled quota format "
1792 "not specified");
1793 return 0;
1794 }
1795 } else {
1796 if (sbi->s_jquota_fmt) {
1797 ext4_msg(sb, KERN_ERR, "journaled quota format "
1798 "specified with no journaling "
1799 "enabled");
1800 return 0;
1801 }
1802 }
1803 #endif
1804 return 1;
1805 }
1806
1807 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1808 int read_only)
1809 {
1810 struct ext4_sb_info *sbi = EXT4_SB(sb);
1811 int res = 0;
1812
1813 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1814 ext4_msg(sb, KERN_ERR, "revision level too high, "
1815 "forcing read-only mode");
1816 res = MS_RDONLY;
1817 }
1818 if (read_only)
1819 goto done;
1820 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1821 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1822 "running e2fsck is recommended");
1823 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1824 ext4_msg(sb, KERN_WARNING,
1825 "warning: mounting fs with errors, "
1826 "running e2fsck is recommended");
1827 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1828 le16_to_cpu(es->s_mnt_count) >=
1829 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1830 ext4_msg(sb, KERN_WARNING,
1831 "warning: maximal mount count reached, "
1832 "running e2fsck is recommended");
1833 else if (le32_to_cpu(es->s_checkinterval) &&
1834 (le32_to_cpu(es->s_lastcheck) +
1835 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1836 ext4_msg(sb, KERN_WARNING,
1837 "warning: checktime reached, "
1838 "running e2fsck is recommended");
1839 if (!sbi->s_journal)
1840 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1841 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1842 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1843 le16_add_cpu(&es->s_mnt_count, 1);
1844 es->s_mtime = cpu_to_le32(get_seconds());
1845 ext4_update_dynamic_rev(sb);
1846 if (sbi->s_journal)
1847 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1848
1849 ext4_commit_super(sb, 1);
1850 done:
1851 if (test_opt(sb, DEBUG))
1852 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1853 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1854 sb->s_blocksize,
1855 sbi->s_groups_count,
1856 EXT4_BLOCKS_PER_GROUP(sb),
1857 EXT4_INODES_PER_GROUP(sb),
1858 sbi->s_mount_opt, sbi->s_mount_opt2);
1859
1860 cleancache_init_fs(sb);
1861 return res;
1862 }
1863
1864 static int ext4_fill_flex_info(struct super_block *sb)
1865 {
1866 struct ext4_sb_info *sbi = EXT4_SB(sb);
1867 struct ext4_group_desc *gdp = NULL;
1868 ext4_group_t flex_group_count;
1869 ext4_group_t flex_group;
1870 unsigned int groups_per_flex = 0;
1871 size_t size;
1872 int i;
1873
1874 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1875 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1876 sbi->s_log_groups_per_flex = 0;
1877 return 1;
1878 }
1879 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1880
1881 /* We allocate both existing and potentially added groups */
1882 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1883 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1884 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1885 size = flex_group_count * sizeof(struct flex_groups);
1886 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
1887 if (sbi->s_flex_groups == NULL) {
1888 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1889 flex_group_count);
1890 goto failed;
1891 }
1892
1893 for (i = 0; i < sbi->s_groups_count; i++) {
1894 gdp = ext4_get_group_desc(sb, i, NULL);
1895
1896 flex_group = ext4_flex_group(sbi, i);
1897 atomic_add(ext4_free_inodes_count(sb, gdp),
1898 &sbi->s_flex_groups[flex_group].free_inodes);
1899 atomic_add(ext4_free_group_clusters(sb, gdp),
1900 &sbi->s_flex_groups[flex_group].free_clusters);
1901 atomic_add(ext4_used_dirs_count(sb, gdp),
1902 &sbi->s_flex_groups[flex_group].used_dirs);
1903 }
1904
1905 return 1;
1906 failed:
1907 return 0;
1908 }
1909
1910 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1911 struct ext4_group_desc *gdp)
1912 {
1913 __u16 crc = 0;
1914
1915 if (sbi->s_es->s_feature_ro_compat &
1916 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1917 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1918 __le32 le_group = cpu_to_le32(block_group);
1919
1920 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1921 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1922 crc = crc16(crc, (__u8 *)gdp, offset);
1923 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1924 /* for checksum of struct ext4_group_desc do the rest...*/
1925 if ((sbi->s_es->s_feature_incompat &
1926 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1927 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1928 crc = crc16(crc, (__u8 *)gdp + offset,
1929 le16_to_cpu(sbi->s_es->s_desc_size) -
1930 offset);
1931 }
1932
1933 return cpu_to_le16(crc);
1934 }
1935
1936 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1937 struct ext4_group_desc *gdp)
1938 {
1939 if ((sbi->s_es->s_feature_ro_compat &
1940 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1941 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1942 return 0;
1943
1944 return 1;
1945 }
1946
1947 /* Called at mount-time, super-block is locked */
1948 static int ext4_check_descriptors(struct super_block *sb,
1949 ext4_group_t *first_not_zeroed)
1950 {
1951 struct ext4_sb_info *sbi = EXT4_SB(sb);
1952 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1953 ext4_fsblk_t last_block;
1954 ext4_fsblk_t block_bitmap;
1955 ext4_fsblk_t inode_bitmap;
1956 ext4_fsblk_t inode_table;
1957 int flexbg_flag = 0;
1958 ext4_group_t i, grp = sbi->s_groups_count;
1959
1960 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1961 flexbg_flag = 1;
1962
1963 ext4_debug("Checking group descriptors");
1964
1965 for (i = 0; i < sbi->s_groups_count; i++) {
1966 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1967
1968 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1969 last_block = ext4_blocks_count(sbi->s_es) - 1;
1970 else
1971 last_block = first_block +
1972 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1973
1974 if ((grp == sbi->s_groups_count) &&
1975 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
1976 grp = i;
1977
1978 block_bitmap = ext4_block_bitmap(sb, gdp);
1979 if (block_bitmap < first_block || block_bitmap > last_block) {
1980 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1981 "Block bitmap for group %u not in group "
1982 "(block %llu)!", i, block_bitmap);
1983 return 0;
1984 }
1985 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1986 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1987 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1988 "Inode bitmap for group %u not in group "
1989 "(block %llu)!", i, inode_bitmap);
1990 return 0;
1991 }
1992 inode_table = ext4_inode_table(sb, gdp);
1993 if (inode_table < first_block ||
1994 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1995 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1996 "Inode table for group %u not in group "
1997 "(block %llu)!", i, inode_table);
1998 return 0;
1999 }
2000 ext4_lock_group(sb, i);
2001 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2002 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2003 "Checksum for group %u failed (%u!=%u)",
2004 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2005 gdp)), le16_to_cpu(gdp->bg_checksum));
2006 if (!(sb->s_flags & MS_RDONLY)) {
2007 ext4_unlock_group(sb, i);
2008 return 0;
2009 }
2010 }
2011 ext4_unlock_group(sb, i);
2012 if (!flexbg_flag)
2013 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2014 }
2015 if (NULL != first_not_zeroed)
2016 *first_not_zeroed = grp;
2017
2018 ext4_free_blocks_count_set(sbi->s_es,
2019 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2020 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2021 return 1;
2022 }
2023
2024 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2025 * the superblock) which were deleted from all directories, but held open by
2026 * a process at the time of a crash. We walk the list and try to delete these
2027 * inodes at recovery time (only with a read-write filesystem).
2028 *
2029 * In order to keep the orphan inode chain consistent during traversal (in
2030 * case of crash during recovery), we link each inode into the superblock
2031 * orphan list_head and handle it the same way as an inode deletion during
2032 * normal operation (which journals the operations for us).
2033 *
2034 * We only do an iget() and an iput() on each inode, which is very safe if we
2035 * accidentally point at an in-use or already deleted inode. The worst that
2036 * can happen in this case is that we get a "bit already cleared" message from
2037 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2038 * e2fsck was run on this filesystem, and it must have already done the orphan
2039 * inode cleanup for us, so we can safely abort without any further action.
2040 */
2041 static void ext4_orphan_cleanup(struct super_block *sb,
2042 struct ext4_super_block *es)
2043 {
2044 unsigned int s_flags = sb->s_flags;
2045 int nr_orphans = 0, nr_truncates = 0;
2046 #ifdef CONFIG_QUOTA
2047 int i;
2048 #endif
2049 if (!es->s_last_orphan) {
2050 jbd_debug(4, "no orphan inodes to clean up\n");
2051 return;
2052 }
2053
2054 if (bdev_read_only(sb->s_bdev)) {
2055 ext4_msg(sb, KERN_ERR, "write access "
2056 "unavailable, skipping orphan cleanup");
2057 return;
2058 }
2059
2060 /* Check if feature set would not allow a r/w mount */
2061 if (!ext4_feature_set_ok(sb, 0)) {
2062 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2063 "unknown ROCOMPAT features");
2064 return;
2065 }
2066
2067 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2068 if (es->s_last_orphan)
2069 jbd_debug(1, "Errors on filesystem, "
2070 "clearing orphan list.\n");
2071 es->s_last_orphan = 0;
2072 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2073 return;
2074 }
2075
2076 if (s_flags & MS_RDONLY) {
2077 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2078 sb->s_flags &= ~MS_RDONLY;
2079 }
2080 #ifdef CONFIG_QUOTA
2081 /* Needed for iput() to work correctly and not trash data */
2082 sb->s_flags |= MS_ACTIVE;
2083 /* Turn on quotas so that they are updated correctly */
2084 for (i = 0; i < MAXQUOTAS; i++) {
2085 if (EXT4_SB(sb)->s_qf_names[i]) {
2086 int ret = ext4_quota_on_mount(sb, i);
2087 if (ret < 0)
2088 ext4_msg(sb, KERN_ERR,
2089 "Cannot turn on journaled "
2090 "quota: error %d", ret);
2091 }
2092 }
2093 #endif
2094
2095 while (es->s_last_orphan) {
2096 struct inode *inode;
2097
2098 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2099 if (IS_ERR(inode)) {
2100 es->s_last_orphan = 0;
2101 break;
2102 }
2103
2104 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2105 dquot_initialize(inode);
2106 if (inode->i_nlink) {
2107 ext4_msg(sb, KERN_DEBUG,
2108 "%s: truncating inode %lu to %lld bytes",
2109 __func__, inode->i_ino, inode->i_size);
2110 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2111 inode->i_ino, inode->i_size);
2112 ext4_truncate(inode);
2113 nr_truncates++;
2114 } else {
2115 ext4_msg(sb, KERN_DEBUG,
2116 "%s: deleting unreferenced inode %lu",
2117 __func__, inode->i_ino);
2118 jbd_debug(2, "deleting unreferenced inode %lu\n",
2119 inode->i_ino);
2120 nr_orphans++;
2121 }
2122 iput(inode); /* The delete magic happens here! */
2123 }
2124
2125 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2126
2127 if (nr_orphans)
2128 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2129 PLURAL(nr_orphans));
2130 if (nr_truncates)
2131 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2132 PLURAL(nr_truncates));
2133 #ifdef CONFIG_QUOTA
2134 /* Turn quotas off */
2135 for (i = 0; i < MAXQUOTAS; i++) {
2136 if (sb_dqopt(sb)->files[i])
2137 dquot_quota_off(sb, i);
2138 }
2139 #endif
2140 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2141 }
2142
2143 /*
2144 * Maximal extent format file size.
2145 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2146 * extent format containers, within a sector_t, and within i_blocks
2147 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2148 * so that won't be a limiting factor.
2149 *
2150 * However there is other limiting factor. We do store extents in the form
2151 * of starting block and length, hence the resulting length of the extent
2152 * covering maximum file size must fit into on-disk format containers as
2153 * well. Given that length is always by 1 unit bigger than max unit (because
2154 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2155 *
2156 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2157 */
2158 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2159 {
2160 loff_t res;
2161 loff_t upper_limit = MAX_LFS_FILESIZE;
2162
2163 /* small i_blocks in vfs inode? */
2164 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2165 /*
2166 * CONFIG_LBDAF is not enabled implies the inode
2167 * i_block represent total blocks in 512 bytes
2168 * 32 == size of vfs inode i_blocks * 8
2169 */
2170 upper_limit = (1LL << 32) - 1;
2171
2172 /* total blocks in file system block size */
2173 upper_limit >>= (blkbits - 9);
2174 upper_limit <<= blkbits;
2175 }
2176
2177 /*
2178 * 32-bit extent-start container, ee_block. We lower the maxbytes
2179 * by one fs block, so ee_len can cover the extent of maximum file
2180 * size
2181 */
2182 res = (1LL << 32) - 1;
2183 res <<= blkbits;
2184
2185 /* Sanity check against vm- & vfs- imposed limits */
2186 if (res > upper_limit)
2187 res = upper_limit;
2188
2189 return res;
2190 }
2191
2192 /*
2193 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2194 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2195 * We need to be 1 filesystem block less than the 2^48 sector limit.
2196 */
2197 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2198 {
2199 loff_t res = EXT4_NDIR_BLOCKS;
2200 int meta_blocks;
2201 loff_t upper_limit;
2202 /* This is calculated to be the largest file size for a dense, block
2203 * mapped file such that the file's total number of 512-byte sectors,
2204 * including data and all indirect blocks, does not exceed (2^48 - 1).
2205 *
2206 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2207 * number of 512-byte sectors of the file.
2208 */
2209
2210 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2211 /*
2212 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2213 * the inode i_block field represents total file blocks in
2214 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2215 */
2216 upper_limit = (1LL << 32) - 1;
2217
2218 /* total blocks in file system block size */
2219 upper_limit >>= (bits - 9);
2220
2221 } else {
2222 /*
2223 * We use 48 bit ext4_inode i_blocks
2224 * With EXT4_HUGE_FILE_FL set the i_blocks
2225 * represent total number of blocks in
2226 * file system block size
2227 */
2228 upper_limit = (1LL << 48) - 1;
2229
2230 }
2231
2232 /* indirect blocks */
2233 meta_blocks = 1;
2234 /* double indirect blocks */
2235 meta_blocks += 1 + (1LL << (bits-2));
2236 /* tripple indirect blocks */
2237 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2238
2239 upper_limit -= meta_blocks;
2240 upper_limit <<= bits;
2241
2242 res += 1LL << (bits-2);
2243 res += 1LL << (2*(bits-2));
2244 res += 1LL << (3*(bits-2));
2245 res <<= bits;
2246 if (res > upper_limit)
2247 res = upper_limit;
2248
2249 if (res > MAX_LFS_FILESIZE)
2250 res = MAX_LFS_FILESIZE;
2251
2252 return res;
2253 }
2254
2255 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2256 ext4_fsblk_t logical_sb_block, int nr)
2257 {
2258 struct ext4_sb_info *sbi = EXT4_SB(sb);
2259 ext4_group_t bg, first_meta_bg;
2260 int has_super = 0;
2261
2262 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2263
2264 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2265 nr < first_meta_bg)
2266 return logical_sb_block + nr + 1;
2267 bg = sbi->s_desc_per_block * nr;
2268 if (ext4_bg_has_super(sb, bg))
2269 has_super = 1;
2270
2271 return (has_super + ext4_group_first_block_no(sb, bg));
2272 }
2273
2274 /**
2275 * ext4_get_stripe_size: Get the stripe size.
2276 * @sbi: In memory super block info
2277 *
2278 * If we have specified it via mount option, then
2279 * use the mount option value. If the value specified at mount time is
2280 * greater than the blocks per group use the super block value.
2281 * If the super block value is greater than blocks per group return 0.
2282 * Allocator needs it be less than blocks per group.
2283 *
2284 */
2285 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2286 {
2287 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2288 unsigned long stripe_width =
2289 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2290 int ret;
2291
2292 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2293 ret = sbi->s_stripe;
2294 else if (stripe_width <= sbi->s_blocks_per_group)
2295 ret = stripe_width;
2296 else if (stride <= sbi->s_blocks_per_group)
2297 ret = stride;
2298 else
2299 ret = 0;
2300
2301 /*
2302 * If the stripe width is 1, this makes no sense and
2303 * we set it to 0 to turn off stripe handling code.
2304 */
2305 if (ret <= 1)
2306 ret = 0;
2307
2308 return ret;
2309 }
2310
2311 /* sysfs supprt */
2312
2313 struct ext4_attr {
2314 struct attribute attr;
2315 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2316 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2317 const char *, size_t);
2318 int offset;
2319 };
2320
2321 static int parse_strtoul(const char *buf,
2322 unsigned long max, unsigned long *value)
2323 {
2324 char *endp;
2325
2326 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2327 endp = skip_spaces(endp);
2328 if (*endp || *value > max)
2329 return -EINVAL;
2330
2331 return 0;
2332 }
2333
2334 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2335 struct ext4_sb_info *sbi,
2336 char *buf)
2337 {
2338 return snprintf(buf, PAGE_SIZE, "%llu\n",
2339 (s64) EXT4_C2B(sbi,
2340 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2341 }
2342
2343 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2344 struct ext4_sb_info *sbi, char *buf)
2345 {
2346 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2347
2348 if (!sb->s_bdev->bd_part)
2349 return snprintf(buf, PAGE_SIZE, "0\n");
2350 return snprintf(buf, PAGE_SIZE, "%lu\n",
2351 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2352 sbi->s_sectors_written_start) >> 1);
2353 }
2354
2355 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2356 struct ext4_sb_info *sbi, char *buf)
2357 {
2358 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2359
2360 if (!sb->s_bdev->bd_part)
2361 return snprintf(buf, PAGE_SIZE, "0\n");
2362 return snprintf(buf, PAGE_SIZE, "%llu\n",
2363 (unsigned long long)(sbi->s_kbytes_written +
2364 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2365 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2366 }
2367
2368 static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2369 struct ext4_sb_info *sbi, char *buf)
2370 {
2371 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2372 }
2373
2374 static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2375 struct ext4_sb_info *sbi, char *buf)
2376 {
2377 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2378 }
2379
2380 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2381 struct ext4_sb_info *sbi,
2382 const char *buf, size_t count)
2383 {
2384 unsigned long t;
2385
2386 if (parse_strtoul(buf, 0x40000000, &t))
2387 return -EINVAL;
2388
2389 if (t && !is_power_of_2(t))
2390 return -EINVAL;
2391
2392 sbi->s_inode_readahead_blks = t;
2393 return count;
2394 }
2395
2396 static ssize_t sbi_ui_show(struct ext4_attr *a,
2397 struct ext4_sb_info *sbi, char *buf)
2398 {
2399 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2400
2401 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2402 }
2403
2404 static ssize_t sbi_ui_store(struct ext4_attr *a,
2405 struct ext4_sb_info *sbi,
2406 const char *buf, size_t count)
2407 {
2408 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2409 unsigned long t;
2410
2411 if (parse_strtoul(buf, 0xffffffff, &t))
2412 return -EINVAL;
2413 *ui = t;
2414 return count;
2415 }
2416
2417 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2418 static struct ext4_attr ext4_attr_##_name = { \
2419 .attr = {.name = __stringify(_name), .mode = _mode }, \
2420 .show = _show, \
2421 .store = _store, \
2422 .offset = offsetof(struct ext4_sb_info, _elname), \
2423 }
2424 #define EXT4_ATTR(name, mode, show, store) \
2425 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2426
2427 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2428 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2429 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2430 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2431 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2432 #define ATTR_LIST(name) &ext4_attr_##name.attr
2433
2434 EXT4_RO_ATTR(delayed_allocation_blocks);
2435 EXT4_RO_ATTR(session_write_kbytes);
2436 EXT4_RO_ATTR(lifetime_write_kbytes);
2437 EXT4_RO_ATTR(extent_cache_hits);
2438 EXT4_RO_ATTR(extent_cache_misses);
2439 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2440 inode_readahead_blks_store, s_inode_readahead_blks);
2441 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2442 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2443 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2444 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2445 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2446 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2447 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2448 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2449
2450 static struct attribute *ext4_attrs[] = {
2451 ATTR_LIST(delayed_allocation_blocks),
2452 ATTR_LIST(session_write_kbytes),
2453 ATTR_LIST(lifetime_write_kbytes),
2454 ATTR_LIST(extent_cache_hits),
2455 ATTR_LIST(extent_cache_misses),
2456 ATTR_LIST(inode_readahead_blks),
2457 ATTR_LIST(inode_goal),
2458 ATTR_LIST(mb_stats),
2459 ATTR_LIST(mb_max_to_scan),
2460 ATTR_LIST(mb_min_to_scan),
2461 ATTR_LIST(mb_order2_req),
2462 ATTR_LIST(mb_stream_req),
2463 ATTR_LIST(mb_group_prealloc),
2464 ATTR_LIST(max_writeback_mb_bump),
2465 NULL,
2466 };
2467
2468 /* Features this copy of ext4 supports */
2469 EXT4_INFO_ATTR(lazy_itable_init);
2470 EXT4_INFO_ATTR(batched_discard);
2471
2472 static struct attribute *ext4_feat_attrs[] = {
2473 ATTR_LIST(lazy_itable_init),
2474 ATTR_LIST(batched_discard),
2475 NULL,
2476 };
2477
2478 static ssize_t ext4_attr_show(struct kobject *kobj,
2479 struct attribute *attr, char *buf)
2480 {
2481 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2482 s_kobj);
2483 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2484
2485 return a->show ? a->show(a, sbi, buf) : 0;
2486 }
2487
2488 static ssize_t ext4_attr_store(struct kobject *kobj,
2489 struct attribute *attr,
2490 const char *buf, size_t len)
2491 {
2492 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2493 s_kobj);
2494 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2495
2496 return a->store ? a->store(a, sbi, buf, len) : 0;
2497 }
2498
2499 static void ext4_sb_release(struct kobject *kobj)
2500 {
2501 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2502 s_kobj);
2503 complete(&sbi->s_kobj_unregister);
2504 }
2505
2506 static const struct sysfs_ops ext4_attr_ops = {
2507 .show = ext4_attr_show,
2508 .store = ext4_attr_store,
2509 };
2510
2511 static struct kobj_type ext4_ktype = {
2512 .default_attrs = ext4_attrs,
2513 .sysfs_ops = &ext4_attr_ops,
2514 .release = ext4_sb_release,
2515 };
2516
2517 static void ext4_feat_release(struct kobject *kobj)
2518 {
2519 complete(&ext4_feat->f_kobj_unregister);
2520 }
2521
2522 static struct kobj_type ext4_feat_ktype = {
2523 .default_attrs = ext4_feat_attrs,
2524 .sysfs_ops = &ext4_attr_ops,
2525 .release = ext4_feat_release,
2526 };
2527
2528 /*
2529 * Check whether this filesystem can be mounted based on
2530 * the features present and the RDONLY/RDWR mount requested.
2531 * Returns 1 if this filesystem can be mounted as requested,
2532 * 0 if it cannot be.
2533 */
2534 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2535 {
2536 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2537 ext4_msg(sb, KERN_ERR,
2538 "Couldn't mount because of "
2539 "unsupported optional features (%x)",
2540 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2541 ~EXT4_FEATURE_INCOMPAT_SUPP));
2542 return 0;
2543 }
2544
2545 if (readonly)
2546 return 1;
2547
2548 /* Check that feature set is OK for a read-write mount */
2549 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2550 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2551 "unsupported optional features (%x)",
2552 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2553 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2554 return 0;
2555 }
2556 /*
2557 * Large file size enabled file system can only be mounted
2558 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2559 */
2560 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2561 if (sizeof(blkcnt_t) < sizeof(u64)) {
2562 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2563 "cannot be mounted RDWR without "
2564 "CONFIG_LBDAF");
2565 return 0;
2566 }
2567 }
2568 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2569 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2570 ext4_msg(sb, KERN_ERR,
2571 "Can't support bigalloc feature without "
2572 "extents feature\n");
2573 return 0;
2574 }
2575 return 1;
2576 }
2577
2578 /*
2579 * This function is called once a day if we have errors logged
2580 * on the file system
2581 */
2582 static void print_daily_error_info(unsigned long arg)
2583 {
2584 struct super_block *sb = (struct super_block *) arg;
2585 struct ext4_sb_info *sbi;
2586 struct ext4_super_block *es;
2587
2588 sbi = EXT4_SB(sb);
2589 es = sbi->s_es;
2590
2591 if (es->s_error_count)
2592 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2593 le32_to_cpu(es->s_error_count));
2594 if (es->s_first_error_time) {
2595 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2596 sb->s_id, le32_to_cpu(es->s_first_error_time),
2597 (int) sizeof(es->s_first_error_func),
2598 es->s_first_error_func,
2599 le32_to_cpu(es->s_first_error_line));
2600 if (es->s_first_error_ino)
2601 printk(": inode %u",
2602 le32_to_cpu(es->s_first_error_ino));
2603 if (es->s_first_error_block)
2604 printk(": block %llu", (unsigned long long)
2605 le64_to_cpu(es->s_first_error_block));
2606 printk("\n");
2607 }
2608 if (es->s_last_error_time) {
2609 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2610 sb->s_id, le32_to_cpu(es->s_last_error_time),
2611 (int) sizeof(es->s_last_error_func),
2612 es->s_last_error_func,
2613 le32_to_cpu(es->s_last_error_line));
2614 if (es->s_last_error_ino)
2615 printk(": inode %u",
2616 le32_to_cpu(es->s_last_error_ino));
2617 if (es->s_last_error_block)
2618 printk(": block %llu", (unsigned long long)
2619 le64_to_cpu(es->s_last_error_block));
2620 printk("\n");
2621 }
2622 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2623 }
2624
2625 /* Find next suitable group and run ext4_init_inode_table */
2626 static int ext4_run_li_request(struct ext4_li_request *elr)
2627 {
2628 struct ext4_group_desc *gdp = NULL;
2629 ext4_group_t group, ngroups;
2630 struct super_block *sb;
2631 unsigned long timeout = 0;
2632 int ret = 0;
2633
2634 sb = elr->lr_super;
2635 ngroups = EXT4_SB(sb)->s_groups_count;
2636
2637 for (group = elr->lr_next_group; group < ngroups; group++) {
2638 gdp = ext4_get_group_desc(sb, group, NULL);
2639 if (!gdp) {
2640 ret = 1;
2641 break;
2642 }
2643
2644 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2645 break;
2646 }
2647
2648 if (group == ngroups)
2649 ret = 1;
2650
2651 if (!ret) {
2652 timeout = jiffies;
2653 ret = ext4_init_inode_table(sb, group,
2654 elr->lr_timeout ? 0 : 1);
2655 if (elr->lr_timeout == 0) {
2656 timeout = (jiffies - timeout) *
2657 elr->lr_sbi->s_li_wait_mult;
2658 elr->lr_timeout = timeout;
2659 }
2660 elr->lr_next_sched = jiffies + elr->lr_timeout;
2661 elr->lr_next_group = group + 1;
2662 }
2663
2664 return ret;
2665 }
2666
2667 /*
2668 * Remove lr_request from the list_request and free the
2669 * request structure. Should be called with li_list_mtx held
2670 */
2671 static void ext4_remove_li_request(struct ext4_li_request *elr)
2672 {
2673 struct ext4_sb_info *sbi;
2674
2675 if (!elr)
2676 return;
2677
2678 sbi = elr->lr_sbi;
2679
2680 list_del(&elr->lr_request);
2681 sbi->s_li_request = NULL;
2682 kfree(elr);
2683 }
2684
2685 static void ext4_unregister_li_request(struct super_block *sb)
2686 {
2687 mutex_lock(&ext4_li_mtx);
2688 if (!ext4_li_info) {
2689 mutex_unlock(&ext4_li_mtx);
2690 return;
2691 }
2692
2693 mutex_lock(&ext4_li_info->li_list_mtx);
2694 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2695 mutex_unlock(&ext4_li_info->li_list_mtx);
2696 mutex_unlock(&ext4_li_mtx);
2697 }
2698
2699 static struct task_struct *ext4_lazyinit_task;
2700
2701 /*
2702 * This is the function where ext4lazyinit thread lives. It walks
2703 * through the request list searching for next scheduled filesystem.
2704 * When such a fs is found, run the lazy initialization request
2705 * (ext4_rn_li_request) and keep track of the time spend in this
2706 * function. Based on that time we compute next schedule time of
2707 * the request. When walking through the list is complete, compute
2708 * next waking time and put itself into sleep.
2709 */
2710 static int ext4_lazyinit_thread(void *arg)
2711 {
2712 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2713 struct list_head *pos, *n;
2714 struct ext4_li_request *elr;
2715 unsigned long next_wakeup, cur;
2716
2717 BUG_ON(NULL == eli);
2718
2719 cont_thread:
2720 while (true) {
2721 next_wakeup = MAX_JIFFY_OFFSET;
2722
2723 mutex_lock(&eli->li_list_mtx);
2724 if (list_empty(&eli->li_request_list)) {
2725 mutex_unlock(&eli->li_list_mtx);
2726 goto exit_thread;
2727 }
2728
2729 list_for_each_safe(pos, n, &eli->li_request_list) {
2730 elr = list_entry(pos, struct ext4_li_request,
2731 lr_request);
2732
2733 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2734 if (ext4_run_li_request(elr) != 0) {
2735 /* error, remove the lazy_init job */
2736 ext4_remove_li_request(elr);
2737 continue;
2738 }
2739 }
2740
2741 if (time_before(elr->lr_next_sched, next_wakeup))
2742 next_wakeup = elr->lr_next_sched;
2743 }
2744 mutex_unlock(&eli->li_list_mtx);
2745
2746 try_to_freeze();
2747
2748 cur = jiffies;
2749 if ((time_after_eq(cur, next_wakeup)) ||
2750 (MAX_JIFFY_OFFSET == next_wakeup)) {
2751 cond_resched();
2752 continue;
2753 }
2754
2755 schedule_timeout_interruptible(next_wakeup - cur);
2756
2757 if (kthread_should_stop()) {
2758 ext4_clear_request_list();
2759 goto exit_thread;
2760 }
2761 }
2762
2763 exit_thread:
2764 /*
2765 * It looks like the request list is empty, but we need
2766 * to check it under the li_list_mtx lock, to prevent any
2767 * additions into it, and of course we should lock ext4_li_mtx
2768 * to atomically free the list and ext4_li_info, because at
2769 * this point another ext4 filesystem could be registering
2770 * new one.
2771 */
2772 mutex_lock(&ext4_li_mtx);
2773 mutex_lock(&eli->li_list_mtx);
2774 if (!list_empty(&eli->li_request_list)) {
2775 mutex_unlock(&eli->li_list_mtx);
2776 mutex_unlock(&ext4_li_mtx);
2777 goto cont_thread;
2778 }
2779 mutex_unlock(&eli->li_list_mtx);
2780 kfree(ext4_li_info);
2781 ext4_li_info = NULL;
2782 mutex_unlock(&ext4_li_mtx);
2783
2784 return 0;
2785 }
2786
2787 static void ext4_clear_request_list(void)
2788 {
2789 struct list_head *pos, *n;
2790 struct ext4_li_request *elr;
2791
2792 mutex_lock(&ext4_li_info->li_list_mtx);
2793 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2794 elr = list_entry(pos, struct ext4_li_request,
2795 lr_request);
2796 ext4_remove_li_request(elr);
2797 }
2798 mutex_unlock(&ext4_li_info->li_list_mtx);
2799 }
2800
2801 static int ext4_run_lazyinit_thread(void)
2802 {
2803 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2804 ext4_li_info, "ext4lazyinit");
2805 if (IS_ERR(ext4_lazyinit_task)) {
2806 int err = PTR_ERR(ext4_lazyinit_task);
2807 ext4_clear_request_list();
2808 kfree(ext4_li_info);
2809 ext4_li_info = NULL;
2810 printk(KERN_CRIT "EXT4: error %d creating inode table "
2811 "initialization thread\n",
2812 err);
2813 return err;
2814 }
2815 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2816 return 0;
2817 }
2818
2819 /*
2820 * Check whether it make sense to run itable init. thread or not.
2821 * If there is at least one uninitialized inode table, return
2822 * corresponding group number, else the loop goes through all
2823 * groups and return total number of groups.
2824 */
2825 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2826 {
2827 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2828 struct ext4_group_desc *gdp = NULL;
2829
2830 for (group = 0; group < ngroups; group++) {
2831 gdp = ext4_get_group_desc(sb, group, NULL);
2832 if (!gdp)
2833 continue;
2834
2835 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2836 break;
2837 }
2838
2839 return group;
2840 }
2841
2842 static int ext4_li_info_new(void)
2843 {
2844 struct ext4_lazy_init *eli = NULL;
2845
2846 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2847 if (!eli)
2848 return -ENOMEM;
2849
2850 INIT_LIST_HEAD(&eli->li_request_list);
2851 mutex_init(&eli->li_list_mtx);
2852
2853 eli->li_state |= EXT4_LAZYINIT_QUIT;
2854
2855 ext4_li_info = eli;
2856
2857 return 0;
2858 }
2859
2860 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2861 ext4_group_t start)
2862 {
2863 struct ext4_sb_info *sbi = EXT4_SB(sb);
2864 struct ext4_li_request *elr;
2865 unsigned long rnd;
2866
2867 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2868 if (!elr)
2869 return NULL;
2870
2871 elr->lr_super = sb;
2872 elr->lr_sbi = sbi;
2873 elr->lr_next_group = start;
2874
2875 /*
2876 * Randomize first schedule time of the request to
2877 * spread the inode table initialization requests
2878 * better.
2879 */
2880 get_random_bytes(&rnd, sizeof(rnd));
2881 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2882 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2883
2884 return elr;
2885 }
2886
2887 static int ext4_register_li_request(struct super_block *sb,
2888 ext4_group_t first_not_zeroed)
2889 {
2890 struct ext4_sb_info *sbi = EXT4_SB(sb);
2891 struct ext4_li_request *elr;
2892 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2893 int ret = 0;
2894
2895 if (sbi->s_li_request != NULL) {
2896 /*
2897 * Reset timeout so it can be computed again, because
2898 * s_li_wait_mult might have changed.
2899 */
2900 sbi->s_li_request->lr_timeout = 0;
2901 return 0;
2902 }
2903
2904 if (first_not_zeroed == ngroups ||
2905 (sb->s_flags & MS_RDONLY) ||
2906 !test_opt(sb, INIT_INODE_TABLE))
2907 return 0;
2908
2909 elr = ext4_li_request_new(sb, first_not_zeroed);
2910 if (!elr)
2911 return -ENOMEM;
2912
2913 mutex_lock(&ext4_li_mtx);
2914
2915 if (NULL == ext4_li_info) {
2916 ret = ext4_li_info_new();
2917 if (ret)
2918 goto out;
2919 }
2920
2921 mutex_lock(&ext4_li_info->li_list_mtx);
2922 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2923 mutex_unlock(&ext4_li_info->li_list_mtx);
2924
2925 sbi->s_li_request = elr;
2926 /*
2927 * set elr to NULL here since it has been inserted to
2928 * the request_list and the removal and free of it is
2929 * handled by ext4_clear_request_list from now on.
2930 */
2931 elr = NULL;
2932
2933 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2934 ret = ext4_run_lazyinit_thread();
2935 if (ret)
2936 goto out;
2937 }
2938 out:
2939 mutex_unlock(&ext4_li_mtx);
2940 if (ret)
2941 kfree(elr);
2942 return ret;
2943 }
2944
2945 /*
2946 * We do not need to lock anything since this is called on
2947 * module unload.
2948 */
2949 static void ext4_destroy_lazyinit_thread(void)
2950 {
2951 /*
2952 * If thread exited earlier
2953 * there's nothing to be done.
2954 */
2955 if (!ext4_li_info || !ext4_lazyinit_task)
2956 return;
2957
2958 kthread_stop(ext4_lazyinit_task);
2959 }
2960
2961 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2962 {
2963 char *orig_data = kstrdup(data, GFP_KERNEL);
2964 struct buffer_head *bh;
2965 struct ext4_super_block *es = NULL;
2966 struct ext4_sb_info *sbi;
2967 ext4_fsblk_t block;
2968 ext4_fsblk_t sb_block = get_sb_block(&data);
2969 ext4_fsblk_t logical_sb_block;
2970 unsigned long offset = 0;
2971 unsigned long journal_devnum = 0;
2972 unsigned long def_mount_opts;
2973 struct inode *root;
2974 char *cp;
2975 const char *descr;
2976 int ret = -ENOMEM;
2977 int blocksize, clustersize;
2978 unsigned int db_count;
2979 unsigned int i;
2980 int needs_recovery, has_huge_files, has_bigalloc;
2981 __u64 blocks_count;
2982 int err;
2983 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2984 ext4_group_t first_not_zeroed;
2985
2986 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2987 if (!sbi)
2988 goto out_free_orig;
2989
2990 sbi->s_blockgroup_lock =
2991 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2992 if (!sbi->s_blockgroup_lock) {
2993 kfree(sbi);
2994 goto out_free_orig;
2995 }
2996 sb->s_fs_info = sbi;
2997 sbi->s_mount_opt = 0;
2998 sbi->s_resuid = EXT4_DEF_RESUID;
2999 sbi->s_resgid = EXT4_DEF_RESGID;
3000 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3001 sbi->s_sb_block = sb_block;
3002 if (sb->s_bdev->bd_part)
3003 sbi->s_sectors_written_start =
3004 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3005
3006 /* Cleanup superblock name */
3007 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3008 *cp = '!';
3009
3010 ret = -EINVAL;
3011 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3012 if (!blocksize) {
3013 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3014 goto out_fail;
3015 }
3016
3017 /*
3018 * The ext4 superblock will not be buffer aligned for other than 1kB
3019 * block sizes. We need to calculate the offset from buffer start.
3020 */
3021 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3022 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3023 offset = do_div(logical_sb_block, blocksize);
3024 } else {
3025 logical_sb_block = sb_block;
3026 }
3027
3028 if (!(bh = sb_bread(sb, logical_sb_block))) {
3029 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3030 goto out_fail;
3031 }
3032 /*
3033 * Note: s_es must be initialized as soon as possible because
3034 * some ext4 macro-instructions depend on its value
3035 */
3036 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3037 sbi->s_es = es;
3038 sb->s_magic = le16_to_cpu(es->s_magic);
3039 if (sb->s_magic != EXT4_SUPER_MAGIC)
3040 goto cantfind_ext4;
3041 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3042
3043 /* Set defaults before we parse the mount options */
3044 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3045 set_opt(sb, INIT_INODE_TABLE);
3046 if (def_mount_opts & EXT4_DEFM_DEBUG)
3047 set_opt(sb, DEBUG);
3048 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3049 set_opt(sb, GRPID);
3050 if (def_mount_opts & EXT4_DEFM_UID16)
3051 set_opt(sb, NO_UID32);
3052 /* xattr user namespace & acls are now defaulted on */
3053 #ifdef CONFIG_EXT4_FS_XATTR
3054 set_opt(sb, XATTR_USER);
3055 #endif
3056 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3057 set_opt(sb, POSIX_ACL);
3058 #endif
3059 set_opt(sb, MBLK_IO_SUBMIT);
3060 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3061 set_opt(sb, JOURNAL_DATA);
3062 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3063 set_opt(sb, ORDERED_DATA);
3064 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3065 set_opt(sb, WRITEBACK_DATA);
3066
3067 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3068 set_opt(sb, ERRORS_PANIC);
3069 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3070 set_opt(sb, ERRORS_CONT);
3071 else
3072 set_opt(sb, ERRORS_RO);
3073 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3074 set_opt(sb, BLOCK_VALIDITY);
3075 if (def_mount_opts & EXT4_DEFM_DISCARD)
3076 set_opt(sb, DISCARD);
3077
3078 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3079 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3080 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3081 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3082 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3083
3084 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3085 set_opt(sb, BARRIER);
3086
3087 /*
3088 * enable delayed allocation by default
3089 * Use -o nodelalloc to turn it off
3090 */
3091 if (!IS_EXT3_SB(sb) &&
3092 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3093 set_opt(sb, DELALLOC);
3094
3095 /*
3096 * set default s_li_wait_mult for lazyinit, for the case there is
3097 * no mount option specified.
3098 */
3099 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3100
3101 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3102 &journal_devnum, &journal_ioprio, 0)) {
3103 ext4_msg(sb, KERN_WARNING,
3104 "failed to parse options in superblock: %s",
3105 sbi->s_es->s_mount_opts);
3106 }
3107 if (!parse_options((char *) data, sb, &journal_devnum,
3108 &journal_ioprio, 0))
3109 goto failed_mount;
3110
3111 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3112 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3113 "with data=journal disables delayed "
3114 "allocation and O_DIRECT support!\n");
3115 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3116 ext4_msg(sb, KERN_ERR, "can't mount with "
3117 "both data=journal and delalloc");
3118 goto failed_mount;
3119 }
3120 if (test_opt(sb, DIOREAD_NOLOCK)) {
3121 ext4_msg(sb, KERN_ERR, "can't mount with "
3122 "both data=journal and delalloc");
3123 goto failed_mount;
3124 }
3125 if (test_opt(sb, DELALLOC))
3126 clear_opt(sb, DELALLOC);
3127 }
3128
3129 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3130 if (test_opt(sb, DIOREAD_NOLOCK)) {
3131 if (blocksize < PAGE_SIZE) {
3132 ext4_msg(sb, KERN_ERR, "can't mount with "
3133 "dioread_nolock if block size != PAGE_SIZE");
3134 goto failed_mount;
3135 }
3136 }
3137
3138 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3139 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3140
3141 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3142 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3143 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3144 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3145 ext4_msg(sb, KERN_WARNING,
3146 "feature flags set on rev 0 fs, "
3147 "running e2fsck is recommended");
3148
3149 if (IS_EXT2_SB(sb)) {
3150 if (ext2_feature_set_ok(sb))
3151 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3152 "using the ext4 subsystem");
3153 else {
3154 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3155 "to feature incompatibilities");
3156 goto failed_mount;
3157 }
3158 }
3159
3160 if (IS_EXT3_SB(sb)) {
3161 if (ext3_feature_set_ok(sb))
3162 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3163 "using the ext4 subsystem");
3164 else {
3165 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3166 "to feature incompatibilities");
3167 goto failed_mount;
3168 }
3169 }
3170
3171 /*
3172 * Check feature flags regardless of the revision level, since we
3173 * previously didn't change the revision level when setting the flags,
3174 * so there is a chance incompat flags are set on a rev 0 filesystem.
3175 */
3176 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3177 goto failed_mount;
3178
3179 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3180 blocksize > EXT4_MAX_BLOCK_SIZE) {
3181 ext4_msg(sb, KERN_ERR,
3182 "Unsupported filesystem blocksize %d", blocksize);
3183 goto failed_mount;
3184 }
3185
3186 if (sb->s_blocksize != blocksize) {
3187 /* Validate the filesystem blocksize */
3188 if (!sb_set_blocksize(sb, blocksize)) {
3189 ext4_msg(sb, KERN_ERR, "bad block size %d",
3190 blocksize);
3191 goto failed_mount;
3192 }
3193
3194 brelse(bh);
3195 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3196 offset = do_div(logical_sb_block, blocksize);
3197 bh = sb_bread(sb, logical_sb_block);
3198 if (!bh) {
3199 ext4_msg(sb, KERN_ERR,
3200 "Can't read superblock on 2nd try");
3201 goto failed_mount;
3202 }
3203 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3204 sbi->s_es = es;
3205 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3206 ext4_msg(sb, KERN_ERR,
3207 "Magic mismatch, very weird!");
3208 goto failed_mount;
3209 }
3210 }
3211
3212 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3213 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3214 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3215 has_huge_files);
3216 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3217
3218 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3219 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3220 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3221 } else {
3222 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3223 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3224 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3225 (!is_power_of_2(sbi->s_inode_size)) ||
3226 (sbi->s_inode_size > blocksize)) {
3227 ext4_msg(sb, KERN_ERR,
3228 "unsupported inode size: %d",
3229 sbi->s_inode_size);
3230 goto failed_mount;
3231 }
3232 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3233 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3234 }
3235
3236 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3237 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3238 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3239 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3240 !is_power_of_2(sbi->s_desc_size)) {
3241 ext4_msg(sb, KERN_ERR,
3242 "unsupported descriptor size %lu",
3243 sbi->s_desc_size);
3244 goto failed_mount;
3245 }
3246 } else
3247 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3248
3249 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3250 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3251 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3252 goto cantfind_ext4;
3253
3254 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3255 if (sbi->s_inodes_per_block == 0)
3256 goto cantfind_ext4;
3257 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3258 sbi->s_inodes_per_block;
3259 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3260 sbi->s_sbh = bh;
3261 sbi->s_mount_state = le16_to_cpu(es->s_state);
3262 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3263 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3264
3265 for (i = 0; i < 4; i++)
3266 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3267 sbi->s_def_hash_version = es->s_def_hash_version;
3268 i = le32_to_cpu(es->s_flags);
3269 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3270 sbi->s_hash_unsigned = 3;
3271 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3272 #ifdef __CHAR_UNSIGNED__
3273 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3274 sbi->s_hash_unsigned = 3;
3275 #else
3276 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3277 #endif
3278 sb->s_dirt = 1;
3279 }
3280
3281 /* Handle clustersize */
3282 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3283 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3284 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3285 if (has_bigalloc) {
3286 if (clustersize < blocksize) {
3287 ext4_msg(sb, KERN_ERR,
3288 "cluster size (%d) smaller than "
3289 "block size (%d)", clustersize, blocksize);
3290 goto failed_mount;
3291 }
3292 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3293 le32_to_cpu(es->s_log_block_size);
3294 sbi->s_clusters_per_group =
3295 le32_to_cpu(es->s_clusters_per_group);
3296 if (sbi->s_clusters_per_group > blocksize * 8) {
3297 ext4_msg(sb, KERN_ERR,
3298 "#clusters per group too big: %lu",
3299 sbi->s_clusters_per_group);
3300 goto failed_mount;
3301 }
3302 if (sbi->s_blocks_per_group !=
3303 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3304 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3305 "clusters per group (%lu) inconsistent",
3306 sbi->s_blocks_per_group,
3307 sbi->s_clusters_per_group);
3308 goto failed_mount;
3309 }
3310 } else {
3311 if (clustersize != blocksize) {
3312 ext4_warning(sb, "fragment/cluster size (%d) != "
3313 "block size (%d)", clustersize,
3314 blocksize);
3315 clustersize = blocksize;
3316 }
3317 if (sbi->s_blocks_per_group > blocksize * 8) {
3318 ext4_msg(sb, KERN_ERR,
3319 "#blocks per group too big: %lu",
3320 sbi->s_blocks_per_group);
3321 goto failed_mount;
3322 }
3323 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3324 sbi->s_cluster_bits = 0;
3325 }
3326 sbi->s_cluster_ratio = clustersize / blocksize;
3327
3328 if (sbi->s_inodes_per_group > blocksize * 8) {
3329 ext4_msg(sb, KERN_ERR,
3330 "#inodes per group too big: %lu",
3331 sbi->s_inodes_per_group);
3332 goto failed_mount;
3333 }
3334
3335 /*
3336 * Test whether we have more sectors than will fit in sector_t,
3337 * and whether the max offset is addressable by the page cache.
3338 */
3339 err = generic_check_addressable(sb->s_blocksize_bits,
3340 ext4_blocks_count(es));
3341 if (err) {
3342 ext4_msg(sb, KERN_ERR, "filesystem"
3343 " too large to mount safely on this system");
3344 if (sizeof(sector_t) < 8)
3345 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3346 ret = err;
3347 goto failed_mount;
3348 }
3349
3350 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3351 goto cantfind_ext4;
3352
3353 /* check blocks count against device size */
3354 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3355 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3356 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3357 "exceeds size of device (%llu blocks)",
3358 ext4_blocks_count(es), blocks_count);
3359 goto failed_mount;
3360 }
3361
3362 /*
3363 * It makes no sense for the first data block to be beyond the end
3364 * of the filesystem.
3365 */
3366 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3367 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3368 "block %u is beyond end of filesystem (%llu)",
3369 le32_to_cpu(es->s_first_data_block),
3370 ext4_blocks_count(es));
3371 goto failed_mount;
3372 }
3373 blocks_count = (ext4_blocks_count(es) -
3374 le32_to_cpu(es->s_first_data_block) +
3375 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3376 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3377 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3378 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3379 "(block count %llu, first data block %u, "
3380 "blocks per group %lu)", sbi->s_groups_count,
3381 ext4_blocks_count(es),
3382 le32_to_cpu(es->s_first_data_block),
3383 EXT4_BLOCKS_PER_GROUP(sb));
3384 goto failed_mount;
3385 }
3386 sbi->s_groups_count = blocks_count;
3387 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3388 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3389 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3390 EXT4_DESC_PER_BLOCK(sb);
3391 sbi->s_group_desc = ext4_kvmalloc(db_count *
3392 sizeof(struct buffer_head *),
3393 GFP_KERNEL);
3394 if (sbi->s_group_desc == NULL) {
3395 ext4_msg(sb, KERN_ERR, "not enough memory");
3396 goto failed_mount;
3397 }
3398
3399 if (ext4_proc_root)
3400 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3401
3402 bgl_lock_init(sbi->s_blockgroup_lock);
3403
3404 for (i = 0; i < db_count; i++) {
3405 block = descriptor_loc(sb, logical_sb_block, i);
3406 sbi->s_group_desc[i] = sb_bread(sb, block);
3407 if (!sbi->s_group_desc[i]) {
3408 ext4_msg(sb, KERN_ERR,
3409 "can't read group descriptor %d", i);
3410 db_count = i;
3411 goto failed_mount2;
3412 }
3413 }
3414 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3415 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3416 goto failed_mount2;
3417 }
3418 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3419 if (!ext4_fill_flex_info(sb)) {
3420 ext4_msg(sb, KERN_ERR,
3421 "unable to initialize "
3422 "flex_bg meta info!");
3423 goto failed_mount2;
3424 }
3425
3426 sbi->s_gdb_count = db_count;
3427 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3428 spin_lock_init(&sbi->s_next_gen_lock);
3429
3430 init_timer(&sbi->s_err_report);
3431 sbi->s_err_report.function = print_daily_error_info;
3432 sbi->s_err_report.data = (unsigned long) sb;
3433
3434 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3435 ext4_count_free_clusters(sb));
3436 if (!err) {
3437 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3438 ext4_count_free_inodes(sb));
3439 }
3440 if (!err) {
3441 err = percpu_counter_init(&sbi->s_dirs_counter,
3442 ext4_count_dirs(sb));
3443 }
3444 if (!err) {
3445 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3446 }
3447 if (err) {
3448 ext4_msg(sb, KERN_ERR, "insufficient memory");
3449 goto failed_mount3;
3450 }
3451
3452 sbi->s_stripe = ext4_get_stripe_size(sbi);
3453 sbi->s_max_writeback_mb_bump = 128;
3454
3455 /*
3456 * set up enough so that it can read an inode
3457 */
3458 if (!test_opt(sb, NOLOAD) &&
3459 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3460 sb->s_op = &ext4_sops;
3461 else
3462 sb->s_op = &ext4_nojournal_sops;
3463 sb->s_export_op = &ext4_export_ops;
3464 sb->s_xattr = ext4_xattr_handlers;
3465 #ifdef CONFIG_QUOTA
3466 sb->s_qcop = &ext4_qctl_operations;
3467 sb->dq_op = &ext4_quota_operations;
3468 #endif
3469 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3470
3471 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3472 mutex_init(&sbi->s_orphan_lock);
3473 sbi->s_resize_flags = 0;
3474
3475 sb->s_root = NULL;
3476
3477 needs_recovery = (es->s_last_orphan != 0 ||
3478 EXT4_HAS_INCOMPAT_FEATURE(sb,
3479 EXT4_FEATURE_INCOMPAT_RECOVER));
3480
3481 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3482 !(sb->s_flags & MS_RDONLY))
3483 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3484 goto failed_mount3;
3485
3486 /*
3487 * The first inode we look at is the journal inode. Don't try
3488 * root first: it may be modified in the journal!
3489 */
3490 if (!test_opt(sb, NOLOAD) &&
3491 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3492 if (ext4_load_journal(sb, es, journal_devnum))
3493 goto failed_mount3;
3494 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3495 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3496 ext4_msg(sb, KERN_ERR, "required journal recovery "
3497 "suppressed and not mounted read-only");
3498 goto failed_mount_wq;
3499 } else {
3500 clear_opt(sb, DATA_FLAGS);
3501 sbi->s_journal = NULL;
3502 needs_recovery = 0;
3503 goto no_journal;
3504 }
3505
3506 if (ext4_blocks_count(es) > 0xffffffffULL &&
3507 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3508 JBD2_FEATURE_INCOMPAT_64BIT)) {
3509 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3510 goto failed_mount_wq;
3511 }
3512
3513 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3514 jbd2_journal_set_features(sbi->s_journal,
3515 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3516 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3517 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3518 jbd2_journal_set_features(sbi->s_journal,
3519 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3520 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3521 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3522 } else {
3523 jbd2_journal_clear_features(sbi->s_journal,
3524 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3525 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3526 }
3527
3528 /* We have now updated the journal if required, so we can
3529 * validate the data journaling mode. */
3530 switch (test_opt(sb, DATA_FLAGS)) {
3531 case 0:
3532 /* No mode set, assume a default based on the journal
3533 * capabilities: ORDERED_DATA if the journal can
3534 * cope, else JOURNAL_DATA
3535 */
3536 if (jbd2_journal_check_available_features
3537 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3538 set_opt(sb, ORDERED_DATA);
3539 else
3540 set_opt(sb, JOURNAL_DATA);
3541 break;
3542
3543 case EXT4_MOUNT_ORDERED_DATA:
3544 case EXT4_MOUNT_WRITEBACK_DATA:
3545 if (!jbd2_journal_check_available_features
3546 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3547 ext4_msg(sb, KERN_ERR, "Journal does not support "
3548 "requested data journaling mode");
3549 goto failed_mount_wq;
3550 }
3551 default:
3552 break;
3553 }
3554 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3555
3556 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3557
3558 /*
3559 * The journal may have updated the bg summary counts, so we
3560 * need to update the global counters.
3561 */
3562 percpu_counter_set(&sbi->s_freeclusters_counter,
3563 ext4_count_free_clusters(sb));
3564 percpu_counter_set(&sbi->s_freeinodes_counter,
3565 ext4_count_free_inodes(sb));
3566 percpu_counter_set(&sbi->s_dirs_counter,
3567 ext4_count_dirs(sb));
3568 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3569
3570 no_journal:
3571 /*
3572 * The maximum number of concurrent works can be high and
3573 * concurrency isn't really necessary. Limit it to 1.
3574 */
3575 EXT4_SB(sb)->dio_unwritten_wq =
3576 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3577 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3578 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3579 goto failed_mount_wq;
3580 }
3581
3582 /*
3583 * The jbd2_journal_load will have done any necessary log recovery,
3584 * so we can safely mount the rest of the filesystem now.
3585 */
3586
3587 root = ext4_iget(sb, EXT4_ROOT_INO);
3588 if (IS_ERR(root)) {
3589 ext4_msg(sb, KERN_ERR, "get root inode failed");
3590 ret = PTR_ERR(root);
3591 root = NULL;
3592 goto failed_mount4;
3593 }
3594 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3595 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3596 iput(root);
3597 goto failed_mount4;
3598 }
3599 sb->s_root = d_alloc_root(root);
3600 if (!sb->s_root) {
3601 iput(root);
3602 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3603 ret = -ENOMEM;
3604 goto failed_mount4;
3605 }
3606
3607 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3608
3609 /* determine the minimum size of new large inodes, if present */
3610 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3611 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3612 EXT4_GOOD_OLD_INODE_SIZE;
3613 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3614 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3615 if (sbi->s_want_extra_isize <
3616 le16_to_cpu(es->s_want_extra_isize))
3617 sbi->s_want_extra_isize =
3618 le16_to_cpu(es->s_want_extra_isize);
3619 if (sbi->s_want_extra_isize <
3620 le16_to_cpu(es->s_min_extra_isize))
3621 sbi->s_want_extra_isize =
3622 le16_to_cpu(es->s_min_extra_isize);
3623 }
3624 }
3625 /* Check if enough inode space is available */
3626 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3627 sbi->s_inode_size) {
3628 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3629 EXT4_GOOD_OLD_INODE_SIZE;
3630 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3631 "available");
3632 }
3633
3634 err = ext4_setup_system_zone(sb);
3635 if (err) {
3636 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3637 "zone (%d)", err);
3638 goto failed_mount4a;
3639 }
3640
3641 ext4_ext_init(sb);
3642 err = ext4_mb_init(sb, needs_recovery);
3643 if (err) {
3644 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3645 err);
3646 goto failed_mount5;
3647 }
3648
3649 err = ext4_register_li_request(sb, first_not_zeroed);
3650 if (err)
3651 goto failed_mount6;
3652
3653 sbi->s_kobj.kset = ext4_kset;
3654 init_completion(&sbi->s_kobj_unregister);
3655 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3656 "%s", sb->s_id);
3657 if (err)
3658 goto failed_mount7;
3659
3660 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3661 ext4_orphan_cleanup(sb, es);
3662 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3663 if (needs_recovery) {
3664 ext4_msg(sb, KERN_INFO, "recovery complete");
3665 ext4_mark_recovery_complete(sb, es);
3666 }
3667 if (EXT4_SB(sb)->s_journal) {
3668 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3669 descr = " journalled data mode";
3670 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3671 descr = " ordered data mode";
3672 else
3673 descr = " writeback data mode";
3674 } else
3675 descr = "out journal";
3676
3677 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3678 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3679 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3680
3681 if (es->s_error_count)
3682 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3683
3684 kfree(orig_data);
3685 return 0;
3686
3687 cantfind_ext4:
3688 if (!silent)
3689 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3690 goto failed_mount;
3691
3692 failed_mount7:
3693 ext4_unregister_li_request(sb);
3694 failed_mount6:
3695 ext4_mb_release(sb);
3696 failed_mount5:
3697 ext4_ext_release(sb);
3698 ext4_release_system_zone(sb);
3699 failed_mount4a:
3700 dput(sb->s_root);
3701 sb->s_root = NULL;
3702 failed_mount4:
3703 ext4_msg(sb, KERN_ERR, "mount failed");
3704 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3705 failed_mount_wq:
3706 if (sbi->s_journal) {
3707 jbd2_journal_destroy(sbi->s_journal);
3708 sbi->s_journal = NULL;
3709 }
3710 failed_mount3:
3711 del_timer(&sbi->s_err_report);
3712 if (sbi->s_flex_groups)
3713 ext4_kvfree(sbi->s_flex_groups);
3714 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3715 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3716 percpu_counter_destroy(&sbi->s_dirs_counter);
3717 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3718 if (sbi->s_mmp_tsk)
3719 kthread_stop(sbi->s_mmp_tsk);
3720 failed_mount2:
3721 for (i = 0; i < db_count; i++)
3722 brelse(sbi->s_group_desc[i]);
3723 ext4_kvfree(sbi->s_group_desc);
3724 failed_mount:
3725 if (sbi->s_proc) {
3726 remove_proc_entry(sb->s_id, ext4_proc_root);
3727 }
3728 #ifdef CONFIG_QUOTA
3729 for (i = 0; i < MAXQUOTAS; i++)
3730 kfree(sbi->s_qf_names[i]);
3731 #endif
3732 ext4_blkdev_remove(sbi);
3733 brelse(bh);
3734 out_fail:
3735 sb->s_fs_info = NULL;
3736 kfree(sbi->s_blockgroup_lock);
3737 kfree(sbi);
3738 out_free_orig:
3739 kfree(orig_data);
3740 return ret;
3741 }
3742
3743 /*
3744 * Setup any per-fs journal parameters now. We'll do this both on
3745 * initial mount, once the journal has been initialised but before we've
3746 * done any recovery; and again on any subsequent remount.
3747 */
3748 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3749 {
3750 struct ext4_sb_info *sbi = EXT4_SB(sb);
3751
3752 journal->j_commit_interval = sbi->s_commit_interval;
3753 journal->j_min_batch_time = sbi->s_min_batch_time;
3754 journal->j_max_batch_time = sbi->s_max_batch_time;
3755
3756 write_lock(&journal->j_state_lock);
3757 if (test_opt(sb, BARRIER))
3758 journal->j_flags |= JBD2_BARRIER;
3759 else
3760 journal->j_flags &= ~JBD2_BARRIER;
3761 if (test_opt(sb, DATA_ERR_ABORT))
3762 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3763 else
3764 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3765 write_unlock(&journal->j_state_lock);
3766 }
3767
3768 static journal_t *ext4_get_journal(struct super_block *sb,
3769 unsigned int journal_inum)
3770 {
3771 struct inode *journal_inode;
3772 journal_t *journal;
3773
3774 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3775
3776 /* First, test for the existence of a valid inode on disk. Bad
3777 * things happen if we iget() an unused inode, as the subsequent
3778 * iput() will try to delete it. */
3779
3780 journal_inode = ext4_iget(sb, journal_inum);
3781 if (IS_ERR(journal_inode)) {
3782 ext4_msg(sb, KERN_ERR, "no journal found");
3783 return NULL;
3784 }
3785 if (!journal_inode->i_nlink) {
3786 make_bad_inode(journal_inode);
3787 iput(journal_inode);
3788 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3789 return NULL;
3790 }
3791
3792 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3793 journal_inode, journal_inode->i_size);
3794 if (!S_ISREG(journal_inode->i_mode)) {
3795 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3796 iput(journal_inode);
3797 return NULL;
3798 }
3799
3800 journal = jbd2_journal_init_inode(journal_inode);
3801 if (!journal) {
3802 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3803 iput(journal_inode);
3804 return NULL;
3805 }
3806 journal->j_private = sb;
3807 ext4_init_journal_params(sb, journal);
3808 return journal;
3809 }
3810
3811 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3812 dev_t j_dev)
3813 {
3814 struct buffer_head *bh;
3815 journal_t *journal;
3816 ext4_fsblk_t start;
3817 ext4_fsblk_t len;
3818 int hblock, blocksize;
3819 ext4_fsblk_t sb_block;
3820 unsigned long offset;
3821 struct ext4_super_block *es;
3822 struct block_device *bdev;
3823
3824 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3825
3826 bdev = ext4_blkdev_get(j_dev, sb);
3827 if (bdev == NULL)
3828 return NULL;
3829
3830 blocksize = sb->s_blocksize;
3831 hblock = bdev_logical_block_size(bdev);
3832 if (blocksize < hblock) {
3833 ext4_msg(sb, KERN_ERR,
3834 "blocksize too small for journal device");
3835 goto out_bdev;
3836 }
3837
3838 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3839 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3840 set_blocksize(bdev, blocksize);
3841 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3842 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3843 "external journal");
3844 goto out_bdev;
3845 }
3846
3847 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3848 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3849 !(le32_to_cpu(es->s_feature_incompat) &
3850 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3851 ext4_msg(sb, KERN_ERR, "external journal has "
3852 "bad superblock");
3853 brelse(bh);
3854 goto out_bdev;
3855 }
3856
3857 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3858 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3859 brelse(bh);
3860 goto out_bdev;
3861 }
3862
3863 len = ext4_blocks_count(es);
3864 start = sb_block + 1;
3865 brelse(bh); /* we're done with the superblock */
3866
3867 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3868 start, len, blocksize);
3869 if (!journal) {
3870 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3871 goto out_bdev;
3872 }
3873 journal->j_private = sb;
3874 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3875 wait_on_buffer(journal->j_sb_buffer);
3876 if (!buffer_uptodate(journal->j_sb_buffer)) {
3877 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3878 goto out_journal;
3879 }
3880 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3881 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3882 "user (unsupported) - %d",
3883 be32_to_cpu(journal->j_superblock->s_nr_users));
3884 goto out_journal;
3885 }
3886 EXT4_SB(sb)->journal_bdev = bdev;
3887 ext4_init_journal_params(sb, journal);
3888 return journal;
3889
3890 out_journal:
3891 jbd2_journal_destroy(journal);
3892 out_bdev:
3893 ext4_blkdev_put(bdev);
3894 return NULL;
3895 }
3896
3897 static int ext4_load_journal(struct super_block *sb,
3898 struct ext4_super_block *es,
3899 unsigned long journal_devnum)
3900 {
3901 journal_t *journal;
3902 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3903 dev_t journal_dev;
3904 int err = 0;
3905 int really_read_only;
3906
3907 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3908
3909 if (journal_devnum &&
3910 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3911 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3912 "numbers have changed");
3913 journal_dev = new_decode_dev(journal_devnum);
3914 } else
3915 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3916
3917 really_read_only = bdev_read_only(sb->s_bdev);
3918
3919 /*
3920 * Are we loading a blank journal or performing recovery after a
3921 * crash? For recovery, we need to check in advance whether we
3922 * can get read-write access to the device.
3923 */
3924 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3925 if (sb->s_flags & MS_RDONLY) {
3926 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3927 "required on readonly filesystem");
3928 if (really_read_only) {
3929 ext4_msg(sb, KERN_ERR, "write access "
3930 "unavailable, cannot proceed");
3931 return -EROFS;
3932 }
3933 ext4_msg(sb, KERN_INFO, "write access will "
3934 "be enabled during recovery");
3935 }
3936 }
3937
3938 if (journal_inum && journal_dev) {
3939 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3940 "and inode journals!");
3941 return -EINVAL;
3942 }
3943
3944 if (journal_inum) {
3945 if (!(journal = ext4_get_journal(sb, journal_inum)))
3946 return -EINVAL;
3947 } else {
3948 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3949 return -EINVAL;
3950 }
3951
3952 if (!(journal->j_flags & JBD2_BARRIER))
3953 ext4_msg(sb, KERN_INFO, "barriers disabled");
3954
3955 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3956 err = jbd2_journal_wipe(journal, !really_read_only);
3957 if (!err) {
3958 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3959 if (save)
3960 memcpy(save, ((char *) es) +
3961 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3962 err = jbd2_journal_load(journal);
3963 if (save)
3964 memcpy(((char *) es) + EXT4_S_ERR_START,
3965 save, EXT4_S_ERR_LEN);
3966 kfree(save);
3967 }
3968
3969 if (err) {
3970 ext4_msg(sb, KERN_ERR, "error loading journal");
3971 jbd2_journal_destroy(journal);
3972 return err;
3973 }
3974
3975 EXT4_SB(sb)->s_journal = journal;
3976 ext4_clear_journal_err(sb, es);
3977
3978 if (!really_read_only && journal_devnum &&
3979 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3980 es->s_journal_dev = cpu_to_le32(journal_devnum);
3981
3982 /* Make sure we flush the recovery flag to disk. */
3983 ext4_commit_super(sb, 1);
3984 }
3985
3986 return 0;
3987 }
3988
3989 static int ext4_commit_super(struct super_block *sb, int sync)
3990 {
3991 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3992 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3993 int error = 0;
3994
3995 if (!sbh || block_device_ejected(sb))
3996 return error;
3997 if (buffer_write_io_error(sbh)) {
3998 /*
3999 * Oh, dear. A previous attempt to write the
4000 * superblock failed. This could happen because the
4001 * USB device was yanked out. Or it could happen to
4002 * be a transient write error and maybe the block will
4003 * be remapped. Nothing we can do but to retry the
4004 * write and hope for the best.
4005 */
4006 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4007 "superblock detected");
4008 clear_buffer_write_io_error(sbh);
4009 set_buffer_uptodate(sbh);
4010 }
4011 /*
4012 * If the file system is mounted read-only, don't update the
4013 * superblock write time. This avoids updating the superblock
4014 * write time when we are mounting the root file system
4015 * read/only but we need to replay the journal; at that point,
4016 * for people who are east of GMT and who make their clock
4017 * tick in localtime for Windows bug-for-bug compatibility,
4018 * the clock is set in the future, and this will cause e2fsck
4019 * to complain and force a full file system check.
4020 */
4021 if (!(sb->s_flags & MS_RDONLY))
4022 es->s_wtime = cpu_to_le32(get_seconds());
4023 if (sb->s_bdev->bd_part)
4024 es->s_kbytes_written =
4025 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4026 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4027 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4028 else
4029 es->s_kbytes_written =
4030 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4031 ext4_free_blocks_count_set(es,
4032 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4033 &EXT4_SB(sb)->s_freeclusters_counter)));
4034 es->s_free_inodes_count =
4035 cpu_to_le32(percpu_counter_sum_positive(
4036 &EXT4_SB(sb)->s_freeinodes_counter));
4037 sb->s_dirt = 0;
4038 BUFFER_TRACE(sbh, "marking dirty");
4039 mark_buffer_dirty(sbh);
4040 if (sync) {
4041 error = sync_dirty_buffer(sbh);
4042 if (error)
4043 return error;
4044
4045 error = buffer_write_io_error(sbh);
4046 if (error) {
4047 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4048 "superblock");
4049 clear_buffer_write_io_error(sbh);
4050 set_buffer_uptodate(sbh);
4051 }
4052 }
4053 return error;
4054 }
4055
4056 /*
4057 * Have we just finished recovery? If so, and if we are mounting (or
4058 * remounting) the filesystem readonly, then we will end up with a
4059 * consistent fs on disk. Record that fact.
4060 */
4061 static void ext4_mark_recovery_complete(struct super_block *sb,
4062 struct ext4_super_block *es)
4063 {
4064 journal_t *journal = EXT4_SB(sb)->s_journal;
4065
4066 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4067 BUG_ON(journal != NULL);
4068 return;
4069 }
4070 jbd2_journal_lock_updates(journal);
4071 if (jbd2_journal_flush(journal) < 0)
4072 goto out;
4073
4074 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4075 sb->s_flags & MS_RDONLY) {
4076 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4077 ext4_commit_super(sb, 1);
4078 }
4079
4080 out:
4081 jbd2_journal_unlock_updates(journal);
4082 }
4083
4084 /*
4085 * If we are mounting (or read-write remounting) a filesystem whose journal
4086 * has recorded an error from a previous lifetime, move that error to the
4087 * main filesystem now.
4088 */
4089 static void ext4_clear_journal_err(struct super_block *sb,
4090 struct ext4_super_block *es)
4091 {
4092 journal_t *journal;
4093 int j_errno;
4094 const char *errstr;
4095
4096 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4097
4098 journal = EXT4_SB(sb)->s_journal;
4099
4100 /*
4101 * Now check for any error status which may have been recorded in the
4102 * journal by a prior ext4_error() or ext4_abort()
4103 */
4104
4105 j_errno = jbd2_journal_errno(journal);
4106 if (j_errno) {
4107 char nbuf[16];
4108
4109 errstr = ext4_decode_error(sb, j_errno, nbuf);
4110 ext4_warning(sb, "Filesystem error recorded "
4111 "from previous mount: %s", errstr);
4112 ext4_warning(sb, "Marking fs in need of filesystem check.");
4113
4114 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4115 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4116 ext4_commit_super(sb, 1);
4117
4118 jbd2_journal_clear_err(journal);
4119 }
4120 }
4121
4122 /*
4123 * Force the running and committing transactions to commit,
4124 * and wait on the commit.
4125 */
4126 int ext4_force_commit(struct super_block *sb)
4127 {
4128 journal_t *journal;
4129 int ret = 0;
4130
4131 if (sb->s_flags & MS_RDONLY)
4132 return 0;
4133
4134 journal = EXT4_SB(sb)->s_journal;
4135 if (journal) {
4136 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4137 ret = ext4_journal_force_commit(journal);
4138 }
4139
4140 return ret;
4141 }
4142
4143 static void ext4_write_super(struct super_block *sb)
4144 {
4145 lock_super(sb);
4146 ext4_commit_super(sb, 1);
4147 unlock_super(sb);
4148 }
4149
4150 static int ext4_sync_fs(struct super_block *sb, int wait)
4151 {
4152 int ret = 0;
4153 tid_t target;
4154 struct ext4_sb_info *sbi = EXT4_SB(sb);
4155
4156 trace_ext4_sync_fs(sb, wait);
4157 flush_workqueue(sbi->dio_unwritten_wq);
4158 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4159 if (wait)
4160 jbd2_log_wait_commit(sbi->s_journal, target);
4161 }
4162 return ret;
4163 }
4164
4165 /*
4166 * LVM calls this function before a (read-only) snapshot is created. This
4167 * gives us a chance to flush the journal completely and mark the fs clean.
4168 *
4169 * Note that only this function cannot bring a filesystem to be in a clean
4170 * state independently, because ext4 prevents a new handle from being started
4171 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4172 * the upper layer.
4173 */
4174 static int ext4_freeze(struct super_block *sb)
4175 {
4176 int error = 0;
4177 journal_t *journal;
4178
4179 if (sb->s_flags & MS_RDONLY)
4180 return 0;
4181
4182 journal = EXT4_SB(sb)->s_journal;
4183
4184 /* Now we set up the journal barrier. */
4185 jbd2_journal_lock_updates(journal);
4186
4187 /*
4188 * Don't clear the needs_recovery flag if we failed to flush
4189 * the journal.
4190 */
4191 error = jbd2_journal_flush(journal);
4192 if (error < 0)
4193 goto out;
4194
4195 /* Journal blocked and flushed, clear needs_recovery flag. */
4196 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4197 error = ext4_commit_super(sb, 1);
4198 out:
4199 /* we rely on s_frozen to stop further updates */
4200 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4201 return error;
4202 }
4203
4204 /*
4205 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4206 * flag here, even though the filesystem is not technically dirty yet.
4207 */
4208 static int ext4_unfreeze(struct super_block *sb)
4209 {
4210 if (sb->s_flags & MS_RDONLY)
4211 return 0;
4212
4213 lock_super(sb);
4214 /* Reset the needs_recovery flag before the fs is unlocked. */
4215 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4216 ext4_commit_super(sb, 1);
4217 unlock_super(sb);
4218 return 0;
4219 }
4220
4221 /*
4222 * Structure to save mount options for ext4_remount's benefit
4223 */
4224 struct ext4_mount_options {
4225 unsigned long s_mount_opt;
4226 unsigned long s_mount_opt2;
4227 uid_t s_resuid;
4228 gid_t s_resgid;
4229 unsigned long s_commit_interval;
4230 u32 s_min_batch_time, s_max_batch_time;
4231 #ifdef CONFIG_QUOTA
4232 int s_jquota_fmt;
4233 char *s_qf_names[MAXQUOTAS];
4234 #endif
4235 };
4236
4237 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4238 {
4239 struct ext4_super_block *es;
4240 struct ext4_sb_info *sbi = EXT4_SB(sb);
4241 unsigned long old_sb_flags;
4242 struct ext4_mount_options old_opts;
4243 int enable_quota = 0;
4244 ext4_group_t g;
4245 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4246 int err = 0;
4247 #ifdef CONFIG_QUOTA
4248 int i;
4249 #endif
4250 char *orig_data = kstrdup(data, GFP_KERNEL);
4251
4252 /* Store the original options */
4253 lock_super(sb);
4254 old_sb_flags = sb->s_flags;
4255 old_opts.s_mount_opt = sbi->s_mount_opt;
4256 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4257 old_opts.s_resuid = sbi->s_resuid;
4258 old_opts.s_resgid = sbi->s_resgid;
4259 old_opts.s_commit_interval = sbi->s_commit_interval;
4260 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4261 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4262 #ifdef CONFIG_QUOTA
4263 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4264 for (i = 0; i < MAXQUOTAS; i++)
4265 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4266 #endif
4267 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4268 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4269
4270 /*
4271 * Allow the "check" option to be passed as a remount option.
4272 */
4273 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4274 err = -EINVAL;
4275 goto restore_opts;
4276 }
4277
4278 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4279 ext4_abort(sb, "Abort forced by user");
4280
4281 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4282 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4283
4284 es = sbi->s_es;
4285
4286 if (sbi->s_journal) {
4287 ext4_init_journal_params(sb, sbi->s_journal);
4288 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4289 }
4290
4291 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4292 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4293 err = -EROFS;
4294 goto restore_opts;
4295 }
4296
4297 if (*flags & MS_RDONLY) {
4298 err = dquot_suspend(sb, -1);
4299 if (err < 0)
4300 goto restore_opts;
4301
4302 /*
4303 * First of all, the unconditional stuff we have to do
4304 * to disable replay of the journal when we next remount
4305 */
4306 sb->s_flags |= MS_RDONLY;
4307
4308 /*
4309 * OK, test if we are remounting a valid rw partition
4310 * readonly, and if so set the rdonly flag and then
4311 * mark the partition as valid again.
4312 */
4313 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4314 (sbi->s_mount_state & EXT4_VALID_FS))
4315 es->s_state = cpu_to_le16(sbi->s_mount_state);
4316
4317 if (sbi->s_journal)
4318 ext4_mark_recovery_complete(sb, es);
4319 } else {
4320 /* Make sure we can mount this feature set readwrite */
4321 if (!ext4_feature_set_ok(sb, 0)) {
4322 err = -EROFS;
4323 goto restore_opts;
4324 }
4325 /*
4326 * Make sure the group descriptor checksums
4327 * are sane. If they aren't, refuse to remount r/w.
4328 */
4329 for (g = 0; g < sbi->s_groups_count; g++) {
4330 struct ext4_group_desc *gdp =
4331 ext4_get_group_desc(sb, g, NULL);
4332
4333 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4334 ext4_msg(sb, KERN_ERR,
4335 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4336 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4337 le16_to_cpu(gdp->bg_checksum));
4338 err = -EINVAL;
4339 goto restore_opts;
4340 }
4341 }
4342
4343 /*
4344 * If we have an unprocessed orphan list hanging
4345 * around from a previously readonly bdev mount,
4346 * require a full umount/remount for now.
4347 */
4348 if (es->s_last_orphan) {
4349 ext4_msg(sb, KERN_WARNING, "Couldn't "
4350 "remount RDWR because of unprocessed "
4351 "orphan inode list. Please "
4352 "umount/remount instead");
4353 err = -EINVAL;
4354 goto restore_opts;
4355 }
4356
4357 /*
4358 * Mounting a RDONLY partition read-write, so reread
4359 * and store the current valid flag. (It may have
4360 * been changed by e2fsck since we originally mounted
4361 * the partition.)
4362 */
4363 if (sbi->s_journal)
4364 ext4_clear_journal_err(sb, es);
4365 sbi->s_mount_state = le16_to_cpu(es->s_state);
4366 if (!ext4_setup_super(sb, es, 0))
4367 sb->s_flags &= ~MS_RDONLY;
4368 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4369 EXT4_FEATURE_INCOMPAT_MMP))
4370 if (ext4_multi_mount_protect(sb,
4371 le64_to_cpu(es->s_mmp_block))) {
4372 err = -EROFS;
4373 goto restore_opts;
4374 }
4375 enable_quota = 1;
4376 }
4377 }
4378
4379 /*
4380 * Reinitialize lazy itable initialization thread based on
4381 * current settings
4382 */
4383 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4384 ext4_unregister_li_request(sb);
4385 else {
4386 ext4_group_t first_not_zeroed;
4387 first_not_zeroed = ext4_has_uninit_itable(sb);
4388 ext4_register_li_request(sb, first_not_zeroed);
4389 }
4390
4391 ext4_setup_system_zone(sb);
4392 if (sbi->s_journal == NULL)
4393 ext4_commit_super(sb, 1);
4394
4395 #ifdef CONFIG_QUOTA
4396 /* Release old quota file names */
4397 for (i = 0; i < MAXQUOTAS; i++)
4398 if (old_opts.s_qf_names[i] &&
4399 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4400 kfree(old_opts.s_qf_names[i]);
4401 #endif
4402 unlock_super(sb);
4403 if (enable_quota)
4404 dquot_resume(sb, -1);
4405
4406 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4407 kfree(orig_data);
4408 return 0;
4409
4410 restore_opts:
4411 sb->s_flags = old_sb_flags;
4412 sbi->s_mount_opt = old_opts.s_mount_opt;
4413 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4414 sbi->s_resuid = old_opts.s_resuid;
4415 sbi->s_resgid = old_opts.s_resgid;
4416 sbi->s_commit_interval = old_opts.s_commit_interval;
4417 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4418 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4419 #ifdef CONFIG_QUOTA
4420 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4421 for (i = 0; i < MAXQUOTAS; i++) {
4422 if (sbi->s_qf_names[i] &&
4423 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4424 kfree(sbi->s_qf_names[i]);
4425 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4426 }
4427 #endif
4428 unlock_super(sb);
4429 kfree(orig_data);
4430 return err;
4431 }
4432
4433 /*
4434 * Note: calculating the overhead so we can be compatible with
4435 * historical BSD practice is quite difficult in the face of
4436 * clusters/bigalloc. This is because multiple metadata blocks from
4437 * different block group can end up in the same allocation cluster.
4438 * Calculating the exact overhead in the face of clustered allocation
4439 * requires either O(all block bitmaps) in memory or O(number of block
4440 * groups**2) in time. We will still calculate the superblock for
4441 * older file systems --- and if we come across with a bigalloc file
4442 * system with zero in s_overhead_clusters the estimate will be close to
4443 * correct especially for very large cluster sizes --- but for newer
4444 * file systems, it's better to calculate this figure once at mkfs
4445 * time, and store it in the superblock. If the superblock value is
4446 * present (even for non-bigalloc file systems), we will use it.
4447 */
4448 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4449 {
4450 struct super_block *sb = dentry->d_sb;
4451 struct ext4_sb_info *sbi = EXT4_SB(sb);
4452 struct ext4_super_block *es = sbi->s_es;
4453 struct ext4_group_desc *gdp;
4454 u64 fsid;
4455 s64 bfree;
4456
4457 if (test_opt(sb, MINIX_DF)) {
4458 sbi->s_overhead_last = 0;
4459 } else if (es->s_overhead_clusters) {
4460 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4461 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4462 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4463 ext4_fsblk_t overhead = 0;
4464
4465 /*
4466 * Compute the overhead (FS structures). This is constant
4467 * for a given filesystem unless the number of block groups
4468 * changes so we cache the previous value until it does.
4469 */
4470
4471 /*
4472 * All of the blocks before first_data_block are
4473 * overhead
4474 */
4475 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4476
4477 /*
4478 * Add the overhead found in each block group
4479 */
4480 for (i = 0; i < ngroups; i++) {
4481 gdp = ext4_get_group_desc(sb, i, NULL);
4482 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4483 cond_resched();
4484 }
4485 sbi->s_overhead_last = overhead;
4486 smp_wmb();
4487 sbi->s_blocks_last = ext4_blocks_count(es);
4488 }
4489
4490 buf->f_type = EXT4_SUPER_MAGIC;
4491 buf->f_bsize = sb->s_blocksize;
4492 buf->f_blocks = (ext4_blocks_count(es) -
4493 EXT4_C2B(sbi, sbi->s_overhead_last));
4494 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4495 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4496 /* prevent underflow in case that few free space is available */
4497 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4498 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4499 if (buf->f_bfree < ext4_r_blocks_count(es))
4500 buf->f_bavail = 0;
4501 buf->f_files = le32_to_cpu(es->s_inodes_count);
4502 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4503 buf->f_namelen = EXT4_NAME_LEN;
4504 fsid = le64_to_cpup((void *)es->s_uuid) ^
4505 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4506 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4507 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4508
4509 return 0;
4510 }
4511
4512 /* Helper function for writing quotas on sync - we need to start transaction
4513 * before quota file is locked for write. Otherwise the are possible deadlocks:
4514 * Process 1 Process 2
4515 * ext4_create() quota_sync()
4516 * jbd2_journal_start() write_dquot()
4517 * dquot_initialize() down(dqio_mutex)
4518 * down(dqio_mutex) jbd2_journal_start()
4519 *
4520 */
4521
4522 #ifdef CONFIG_QUOTA
4523
4524 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4525 {
4526 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4527 }
4528
4529 static int ext4_write_dquot(struct dquot *dquot)
4530 {
4531 int ret, err;
4532 handle_t *handle;
4533 struct inode *inode;
4534
4535 inode = dquot_to_inode(dquot);
4536 handle = ext4_journal_start(inode,
4537 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4538 if (IS_ERR(handle))
4539 return PTR_ERR(handle);
4540 ret = dquot_commit(dquot);
4541 err = ext4_journal_stop(handle);
4542 if (!ret)
4543 ret = err;
4544 return ret;
4545 }
4546
4547 static int ext4_acquire_dquot(struct dquot *dquot)
4548 {
4549 int ret, err;
4550 handle_t *handle;
4551
4552 handle = ext4_journal_start(dquot_to_inode(dquot),
4553 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4554 if (IS_ERR(handle))
4555 return PTR_ERR(handle);
4556 ret = dquot_acquire(dquot);
4557 err = ext4_journal_stop(handle);
4558 if (!ret)
4559 ret = err;
4560 return ret;
4561 }
4562
4563 static int ext4_release_dquot(struct dquot *dquot)
4564 {
4565 int ret, err;
4566 handle_t *handle;
4567
4568 handle = ext4_journal_start(dquot_to_inode(dquot),
4569 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4570 if (IS_ERR(handle)) {
4571 /* Release dquot anyway to avoid endless cycle in dqput() */
4572 dquot_release(dquot);
4573 return PTR_ERR(handle);
4574 }
4575 ret = dquot_release(dquot);
4576 err = ext4_journal_stop(handle);
4577 if (!ret)
4578 ret = err;
4579 return ret;
4580 }
4581
4582 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4583 {
4584 /* Are we journaling quotas? */
4585 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4586 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4587 dquot_mark_dquot_dirty(dquot);
4588 return ext4_write_dquot(dquot);
4589 } else {
4590 return dquot_mark_dquot_dirty(dquot);
4591 }
4592 }
4593
4594 static int ext4_write_info(struct super_block *sb, int type)
4595 {
4596 int ret, err;
4597 handle_t *handle;
4598
4599 /* Data block + inode block */
4600 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4601 if (IS_ERR(handle))
4602 return PTR_ERR(handle);
4603 ret = dquot_commit_info(sb, type);
4604 err = ext4_journal_stop(handle);
4605 if (!ret)
4606 ret = err;
4607 return ret;
4608 }
4609
4610 /*
4611 * Turn on quotas during mount time - we need to find
4612 * the quota file and such...
4613 */
4614 static int ext4_quota_on_mount(struct super_block *sb, int type)
4615 {
4616 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4617 EXT4_SB(sb)->s_jquota_fmt, type);
4618 }
4619
4620 /*
4621 * Standard function to be called on quota_on
4622 */
4623 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4624 struct path *path)
4625 {
4626 int err;
4627
4628 if (!test_opt(sb, QUOTA))
4629 return -EINVAL;
4630
4631 /* Quotafile not on the same filesystem? */
4632 if (path->dentry->d_sb != sb)
4633 return -EXDEV;
4634 /* Journaling quota? */
4635 if (EXT4_SB(sb)->s_qf_names[type]) {
4636 /* Quotafile not in fs root? */
4637 if (path->dentry->d_parent != sb->s_root)
4638 ext4_msg(sb, KERN_WARNING,
4639 "Quota file not on filesystem root. "
4640 "Journaled quota will not work");
4641 }
4642
4643 /*
4644 * When we journal data on quota file, we have to flush journal to see
4645 * all updates to the file when we bypass pagecache...
4646 */
4647 if (EXT4_SB(sb)->s_journal &&
4648 ext4_should_journal_data(path->dentry->d_inode)) {
4649 /*
4650 * We don't need to lock updates but journal_flush() could
4651 * otherwise be livelocked...
4652 */
4653 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4654 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4655 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4656 if (err)
4657 return err;
4658 }
4659
4660 return dquot_quota_on(sb, type, format_id, path);
4661 }
4662
4663 static int ext4_quota_off(struct super_block *sb, int type)
4664 {
4665 struct inode *inode = sb_dqopt(sb)->files[type];
4666 handle_t *handle;
4667
4668 /* Force all delayed allocation blocks to be allocated.
4669 * Caller already holds s_umount sem */
4670 if (test_opt(sb, DELALLOC))
4671 sync_filesystem(sb);
4672
4673 if (!inode)
4674 goto out;
4675
4676 /* Update modification times of quota files when userspace can
4677 * start looking at them */
4678 handle = ext4_journal_start(inode, 1);
4679 if (IS_ERR(handle))
4680 goto out;
4681 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4682 ext4_mark_inode_dirty(handle, inode);
4683 ext4_journal_stop(handle);
4684
4685 out:
4686 return dquot_quota_off(sb, type);
4687 }
4688
4689 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4690 * acquiring the locks... As quota files are never truncated and quota code
4691 * itself serializes the operations (and no one else should touch the files)
4692 * we don't have to be afraid of races */
4693 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4694 size_t len, loff_t off)
4695 {
4696 struct inode *inode = sb_dqopt(sb)->files[type];
4697 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4698 int err = 0;
4699 int offset = off & (sb->s_blocksize - 1);
4700 int tocopy;
4701 size_t toread;
4702 struct buffer_head *bh;
4703 loff_t i_size = i_size_read(inode);
4704
4705 if (off > i_size)
4706 return 0;
4707 if (off+len > i_size)
4708 len = i_size-off;
4709 toread = len;
4710 while (toread > 0) {
4711 tocopy = sb->s_blocksize - offset < toread ?
4712 sb->s_blocksize - offset : toread;
4713 bh = ext4_bread(NULL, inode, blk, 0, &err);
4714 if (err)
4715 return err;
4716 if (!bh) /* A hole? */
4717 memset(data, 0, tocopy);
4718 else
4719 memcpy(data, bh->b_data+offset, tocopy);
4720 brelse(bh);
4721 offset = 0;
4722 toread -= tocopy;
4723 data += tocopy;
4724 blk++;
4725 }
4726 return len;
4727 }
4728
4729 /* Write to quotafile (we know the transaction is already started and has
4730 * enough credits) */
4731 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4732 const char *data, size_t len, loff_t off)
4733 {
4734 struct inode *inode = sb_dqopt(sb)->files[type];
4735 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4736 int err = 0;
4737 int offset = off & (sb->s_blocksize - 1);
4738 struct buffer_head *bh;
4739 handle_t *handle = journal_current_handle();
4740
4741 if (EXT4_SB(sb)->s_journal && !handle) {
4742 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4743 " cancelled because transaction is not started",
4744 (unsigned long long)off, (unsigned long long)len);
4745 return -EIO;
4746 }
4747 /*
4748 * Since we account only one data block in transaction credits,
4749 * then it is impossible to cross a block boundary.
4750 */
4751 if (sb->s_blocksize - offset < len) {
4752 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4753 " cancelled because not block aligned",
4754 (unsigned long long)off, (unsigned long long)len);
4755 return -EIO;
4756 }
4757
4758 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4759 bh = ext4_bread(handle, inode, blk, 1, &err);
4760 if (!bh)
4761 goto out;
4762 err = ext4_journal_get_write_access(handle, bh);
4763 if (err) {
4764 brelse(bh);
4765 goto out;
4766 }
4767 lock_buffer(bh);
4768 memcpy(bh->b_data+offset, data, len);
4769 flush_dcache_page(bh->b_page);
4770 unlock_buffer(bh);
4771 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4772 brelse(bh);
4773 out:
4774 if (err) {
4775 mutex_unlock(&inode->i_mutex);
4776 return err;
4777 }
4778 if (inode->i_size < off + len) {
4779 i_size_write(inode, off + len);
4780 EXT4_I(inode)->i_disksize = inode->i_size;
4781 ext4_mark_inode_dirty(handle, inode);
4782 }
4783 mutex_unlock(&inode->i_mutex);
4784 return len;
4785 }
4786
4787 #endif
4788
4789 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4790 const char *dev_name, void *data)
4791 {
4792 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4793 }
4794
4795 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4796 static inline void register_as_ext2(void)
4797 {
4798 int err = register_filesystem(&ext2_fs_type);
4799 if (err)
4800 printk(KERN_WARNING
4801 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4802 }
4803
4804 static inline void unregister_as_ext2(void)
4805 {
4806 unregister_filesystem(&ext2_fs_type);
4807 }
4808
4809 static inline int ext2_feature_set_ok(struct super_block *sb)
4810 {
4811 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4812 return 0;
4813 if (sb->s_flags & MS_RDONLY)
4814 return 1;
4815 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4816 return 0;
4817 return 1;
4818 }
4819 MODULE_ALIAS("ext2");
4820 #else
4821 static inline void register_as_ext2(void) { }
4822 static inline void unregister_as_ext2(void) { }
4823 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4824 #endif
4825
4826 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4827 static inline void register_as_ext3(void)
4828 {
4829 int err = register_filesystem(&ext3_fs_type);
4830 if (err)
4831 printk(KERN_WARNING
4832 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4833 }
4834
4835 static inline void unregister_as_ext3(void)
4836 {
4837 unregister_filesystem(&ext3_fs_type);
4838 }
4839
4840 static inline int ext3_feature_set_ok(struct super_block *sb)
4841 {
4842 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4843 return 0;
4844 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4845 return 0;
4846 if (sb->s_flags & MS_RDONLY)
4847 return 1;
4848 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4849 return 0;
4850 return 1;
4851 }
4852 MODULE_ALIAS("ext3");
4853 #else
4854 static inline void register_as_ext3(void) { }
4855 static inline void unregister_as_ext3(void) { }
4856 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
4857 #endif
4858
4859 static struct file_system_type ext4_fs_type = {
4860 .owner = THIS_MODULE,
4861 .name = "ext4",
4862 .mount = ext4_mount,
4863 .kill_sb = kill_block_super,
4864 .fs_flags = FS_REQUIRES_DEV,
4865 };
4866
4867 static int __init ext4_init_feat_adverts(void)
4868 {
4869 struct ext4_features *ef;
4870 int ret = -ENOMEM;
4871
4872 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4873 if (!ef)
4874 goto out;
4875
4876 ef->f_kobj.kset = ext4_kset;
4877 init_completion(&ef->f_kobj_unregister);
4878 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4879 "features");
4880 if (ret) {
4881 kfree(ef);
4882 goto out;
4883 }
4884
4885 ext4_feat = ef;
4886 ret = 0;
4887 out:
4888 return ret;
4889 }
4890
4891 static void ext4_exit_feat_adverts(void)
4892 {
4893 kobject_put(&ext4_feat->f_kobj);
4894 wait_for_completion(&ext4_feat->f_kobj_unregister);
4895 kfree(ext4_feat);
4896 }
4897
4898 /* Shared across all ext4 file systems */
4899 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4900 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4901
4902 static int __init ext4_init_fs(void)
4903 {
4904 int i, err;
4905
4906 ext4_check_flag_values();
4907
4908 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4909 mutex_init(&ext4__aio_mutex[i]);
4910 init_waitqueue_head(&ext4__ioend_wq[i]);
4911 }
4912
4913 err = ext4_init_pageio();
4914 if (err)
4915 return err;
4916 err = ext4_init_system_zone();
4917 if (err)
4918 goto out6;
4919 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4920 if (!ext4_kset)
4921 goto out5;
4922 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4923
4924 err = ext4_init_feat_adverts();
4925 if (err)
4926 goto out4;
4927
4928 err = ext4_init_mballoc();
4929 if (err)
4930 goto out3;
4931
4932 err = ext4_init_xattr();
4933 if (err)
4934 goto out2;
4935 err = init_inodecache();
4936 if (err)
4937 goto out1;
4938 register_as_ext3();
4939 register_as_ext2();
4940 err = register_filesystem(&ext4_fs_type);
4941 if (err)
4942 goto out;
4943
4944 ext4_li_info = NULL;
4945 mutex_init(&ext4_li_mtx);
4946 return 0;
4947 out:
4948 unregister_as_ext2();
4949 unregister_as_ext3();
4950 destroy_inodecache();
4951 out1:
4952 ext4_exit_xattr();
4953 out2:
4954 ext4_exit_mballoc();
4955 out3:
4956 ext4_exit_feat_adverts();
4957 out4:
4958 if (ext4_proc_root)
4959 remove_proc_entry("fs/ext4", NULL);
4960 kset_unregister(ext4_kset);
4961 out5:
4962 ext4_exit_system_zone();
4963 out6:
4964 ext4_exit_pageio();
4965 return err;
4966 }
4967
4968 static void __exit ext4_exit_fs(void)
4969 {
4970 ext4_destroy_lazyinit_thread();
4971 unregister_as_ext2();
4972 unregister_as_ext3();
4973 unregister_filesystem(&ext4_fs_type);
4974 destroy_inodecache();
4975 ext4_exit_xattr();
4976 ext4_exit_mballoc();
4977 ext4_exit_feat_adverts();
4978 remove_proc_entry("fs/ext4", NULL);
4979 kset_unregister(ext4_kset);
4980 ext4_exit_system_zone();
4981 ext4_exit_pageio();
4982 }
4983
4984 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4985 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4986 MODULE_LICENSE("GPL");
4987 module_init(ext4_init_fs)
4988 module_exit(ext4_exit_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree