?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'linux-next' of git://git.kernel.org/pub/scm/linux/kernel/git/konrad...
[zen-stable.git] / fs / ext4 / super.c
blob3e1329e2f826132d8aec264116deb2c572be9ef7
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
435 /* Deal with the reporting of failure conditions on a filesystem such as
436 * inconsistencies detected or read IO failures.
437 *
438 * On ext2, we can store the error state of the filesystem in the
439 * superblock. That is not possible on ext4, because we may have other
440 * write ordering constraints on the superblock which prevent us from
441 * writing it out straight away; and given that the journal is about to
442 * be aborted, we can't rely on the current, or future, transactions to
443 * write out the superblock safely.
444 *
445 * We'll just use the jbd2_journal_abort() error code to record an error in
446 * the journal instead. On recovery, the journal will complain about
447 * that error until we've noted it down and cleared it.
448 */
449
450 static void ext4_handle_error(struct super_block *sb)
451 {
452 if (sb->s_flags & MS_RDONLY)
453 return;
454
455 if (!test_opt(sb, ERRORS_CONT)) {
456 journal_t *journal = EXT4_SB(sb)->s_journal;
457
458 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
459 if (journal)
460 jbd2_journal_abort(journal, -EIO);
461 }
462 if (test_opt(sb, ERRORS_RO)) {
463 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
464 sb->s_flags |= MS_RDONLY;
465 }
466 if (test_opt(sb, ERRORS_PANIC))
467 panic("EXT4-fs (device %s): panic forced after error\n",
468 sb->s_id);
469 }
470
471 void __ext4_error(struct super_block *sb, const char *function,
472 unsigned int line, const char *fmt, ...)
473 {
474 struct va_format vaf;
475 va_list args;
476
477 va_start(args, fmt);
478 vaf.fmt = fmt;
479 vaf.va = &args;
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
481 sb->s_id, function, line, current->comm, &vaf);
482 va_end(args);
483
484 ext4_handle_error(sb);
485 }
486
487 void ext4_error_inode(struct inode *inode, const char *function,
488 unsigned int line, ext4_fsblk_t block,
489 const char *fmt, ...)
490 {
491 va_list args;
492 struct va_format vaf;
493 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
494
495 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
496 es->s_last_error_block = cpu_to_le64(block);
497 save_error_info(inode->i_sb, function, line);
498 va_start(args, fmt);
499 vaf.fmt = fmt;
500 vaf.va = &args;
501 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
502 inode->i_sb->s_id, function, line, inode->i_ino);
503 if (block)
504 printk(KERN_CONT "block %llu: ", block);
505 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
506 va_end(args);
507
508 ext4_handle_error(inode->i_sb);
509 }
510
511 void ext4_error_file(struct file *file, const char *function,
512 unsigned int line, ext4_fsblk_t block,
513 const char *fmt, ...)
514 {
515 va_list args;
516 struct va_format vaf;
517 struct ext4_super_block *es;
518 struct inode *inode = file->f_dentry->d_inode;
519 char pathname[80], *path;
520
521 es = EXT4_SB(inode->i_sb)->s_es;
522 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
523 save_error_info(inode->i_sb, function, line);
524 path = d_path(&(file->f_path), pathname, sizeof(pathname));
525 if (IS_ERR(path))
526 path = "(unknown)";
527 printk(KERN_CRIT
528 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
529 inode->i_sb->s_id, function, line, inode->i_ino);
530 if (block)
531 printk(KERN_CONT "block %llu: ", block);
532 va_start(args, fmt);
533 vaf.fmt = fmt;
534 vaf.va = &args;
535 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
536 va_end(args);
537
538 ext4_handle_error(inode->i_sb);
539 }
540
541 static const char *ext4_decode_error(struct super_block *sb, int errno,
542 char nbuf[16])
543 {
544 char *errstr = NULL;
545
546 switch (errno) {
547 case -EIO:
548 errstr = "IO failure";
549 break;
550 case -ENOMEM:
551 errstr = "Out of memory";
552 break;
553 case -EROFS:
554 if (!sb || (EXT4_SB(sb)->s_journal &&
555 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
556 errstr = "Journal has aborted";
557 else
558 errstr = "Readonly filesystem";
559 break;
560 default:
561 /* If the caller passed in an extra buffer for unknown
562 * errors, textualise them now. Else we just return
563 * NULL. */
564 if (nbuf) {
565 /* Check for truncated error codes... */
566 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
567 errstr = nbuf;
568 }
569 break;
570 }
571
572 return errstr;
573 }
574
575 /* __ext4_std_error decodes expected errors from journaling functions
576 * automatically and invokes the appropriate error response. */
577
578 void __ext4_std_error(struct super_block *sb, const char *function,
579 unsigned int line, int errno)
580 {
581 char nbuf[16];
582 const char *errstr;
583
584 /* Special case: if the error is EROFS, and we're not already
585 * inside a transaction, then there's really no point in logging
586 * an error. */
587 if (errno == -EROFS && journal_current_handle() == NULL &&
588 (sb->s_flags & MS_RDONLY))
589 return;
590
591 errstr = ext4_decode_error(sb, errno, nbuf);
592 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
593 sb->s_id, function, line, errstr);
594 save_error_info(sb, function, line);
595
596 ext4_handle_error(sb);
597 }
598
599 /*
600 * ext4_abort is a much stronger failure handler than ext4_error. The
601 * abort function may be used to deal with unrecoverable failures such
602 * as journal IO errors or ENOMEM at a critical moment in log management.
603 *
604 * We unconditionally force the filesystem into an ABORT|READONLY state,
605 * unless the error response on the fs has been set to panic in which
606 * case we take the easy way out and panic immediately.
607 */
608
609 void __ext4_abort(struct super_block *sb, const char *function,
610 unsigned int line, const char *fmt, ...)
611 {
612 va_list args;
613
614 save_error_info(sb, function, line);
615 va_start(args, fmt);
616 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
617 function, line);
618 vprintk(fmt, args);
619 printk("\n");
620 va_end(args);
621
622 if ((sb->s_flags & MS_RDONLY) == 0) {
623 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
624 sb->s_flags |= MS_RDONLY;
625 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
626 if (EXT4_SB(sb)->s_journal)
627 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
628 save_error_info(sb, function, line);
629 }
630 if (test_opt(sb, ERRORS_PANIC))
631 panic("EXT4-fs panic from previous error\n");
632 }
633
634 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
635 {
636 struct va_format vaf;
637 va_list args;
638
639 va_start(args, fmt);
640 vaf.fmt = fmt;
641 vaf.va = &args;
642 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
643 va_end(args);
644 }
645
646 void __ext4_warning(struct super_block *sb, const char *function,
647 unsigned int line, const char *fmt, ...)
648 {
649 struct va_format vaf;
650 va_list args;
651
652 va_start(args, fmt);
653 vaf.fmt = fmt;
654 vaf.va = &args;
655 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
656 sb->s_id, function, line, &vaf);
657 va_end(args);
658 }
659
660 void __ext4_grp_locked_error(const char *function, unsigned int line,
661 struct super_block *sb, ext4_group_t grp,
662 unsigned long ino, ext4_fsblk_t block,
663 const char *fmt, ...)
664 __releases(bitlock)
665 __acquires(bitlock)
666 {
667 struct va_format vaf;
668 va_list args;
669 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
670
671 es->s_last_error_ino = cpu_to_le32(ino);
672 es->s_last_error_block = cpu_to_le64(block);
673 __save_error_info(sb, function, line);
674
675 va_start(args, fmt);
676
677 vaf.fmt = fmt;
678 vaf.va = &args;
679 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
680 sb->s_id, function, line, grp);
681 if (ino)
682 printk(KERN_CONT "inode %lu: ", ino);
683 if (block)
684 printk(KERN_CONT "block %llu:", (unsigned long long) block);
685 printk(KERN_CONT "%pV\n", &vaf);
686 va_end(args);
687
688 if (test_opt(sb, ERRORS_CONT)) {
689 ext4_commit_super(sb, 0);
690 return;
691 }
692
693 ext4_unlock_group(sb, grp);
694 ext4_handle_error(sb);
695 /*
696 * We only get here in the ERRORS_RO case; relocking the group
697 * may be dangerous, but nothing bad will happen since the
698 * filesystem will have already been marked read/only and the
699 * journal has been aborted. We return 1 as a hint to callers
700 * who might what to use the return value from
701 * ext4_grp_locked_error() to distinguish between the
702 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
703 * aggressively from the ext4 function in question, with a
704 * more appropriate error code.
705 */
706 ext4_lock_group(sb, grp);
707 return;
708 }
709
710 void ext4_update_dynamic_rev(struct super_block *sb)
711 {
712 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
713
714 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
715 return;
716
717 ext4_warning(sb,
718 "updating to rev %d because of new feature flag, "
719 "running e2fsck is recommended",
720 EXT4_DYNAMIC_REV);
721
722 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
723 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
724 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
725 /* leave es->s_feature_*compat flags alone */
726 /* es->s_uuid will be set by e2fsck if empty */
727
728 /*
729 * The rest of the superblock fields should be zero, and if not it
730 * means they are likely already in use, so leave them alone. We
731 * can leave it up to e2fsck to clean up any inconsistencies there.
732 */
733 }
734
735 /*
736 * Open the external journal device
737 */
738 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
739 {
740 struct block_device *bdev;
741 char b[BDEVNAME_SIZE];
742
743 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
744 if (IS_ERR(bdev))
745 goto fail;
746 return bdev;
747
748 fail:
749 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
750 __bdevname(dev, b), PTR_ERR(bdev));
751 return NULL;
752 }
753
754 /*
755 * Release the journal device
756 */
757 static int ext4_blkdev_put(struct block_device *bdev)
758 {
759 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
760 }
761
762 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
763 {
764 struct block_device *bdev;
765 int ret = -ENODEV;
766
767 bdev = sbi->journal_bdev;
768 if (bdev) {
769 ret = ext4_blkdev_put(bdev);
770 sbi->journal_bdev = NULL;
771 }
772 return ret;
773 }
774
775 static inline struct inode *orphan_list_entry(struct list_head *l)
776 {
777 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
778 }
779
780 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
781 {
782 struct list_head *l;
783
784 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
785 le32_to_cpu(sbi->s_es->s_last_orphan));
786
787 printk(KERN_ERR "sb_info orphan list:\n");
788 list_for_each(l, &sbi->s_orphan) {
789 struct inode *inode = orphan_list_entry(l);
790 printk(KERN_ERR " "
791 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
792 inode->i_sb->s_id, inode->i_ino, inode,
793 inode->i_mode, inode->i_nlink,
794 NEXT_ORPHAN(inode));
795 }
796 }
797
798 static void ext4_put_super(struct super_block *sb)
799 {
800 struct ext4_sb_info *sbi = EXT4_SB(sb);
801 struct ext4_super_block *es = sbi->s_es;
802 int i, err;
803
804 ext4_unregister_li_request(sb);
805 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
806
807 flush_workqueue(sbi->dio_unwritten_wq);
808 destroy_workqueue(sbi->dio_unwritten_wq);
809
810 lock_super(sb);
811 if (sb->s_dirt)
812 ext4_commit_super(sb, 1);
813
814 if (sbi->s_journal) {
815 err = jbd2_journal_destroy(sbi->s_journal);
816 sbi->s_journal = NULL;
817 if (err < 0)
818 ext4_abort(sb, "Couldn't clean up the journal");
819 }
820
821 del_timer(&sbi->s_err_report);
822 ext4_release_system_zone(sb);
823 ext4_mb_release(sb);
824 ext4_ext_release(sb);
825 ext4_xattr_put_super(sb);
826
827 if (!(sb->s_flags & MS_RDONLY)) {
828 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
829 es->s_state = cpu_to_le16(sbi->s_mount_state);
830 ext4_commit_super(sb, 1);
831 }
832 if (sbi->s_proc) {
833 remove_proc_entry(sb->s_id, ext4_proc_root);
834 }
835 kobject_del(&sbi->s_kobj);
836
837 for (i = 0; i < sbi->s_gdb_count; i++)
838 brelse(sbi->s_group_desc[i]);
839 ext4_kvfree(sbi->s_group_desc);
840 ext4_kvfree(sbi->s_flex_groups);
841 percpu_counter_destroy(&sbi->s_freeclusters_counter);
842 percpu_counter_destroy(&sbi->s_freeinodes_counter);
843 percpu_counter_destroy(&sbi->s_dirs_counter);
844 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
845 brelse(sbi->s_sbh);
846 #ifdef CONFIG_QUOTA
847 for (i = 0; i < MAXQUOTAS; i++)
848 kfree(sbi->s_qf_names[i]);
849 #endif
850
851 /* Debugging code just in case the in-memory inode orphan list
852 * isn't empty. The on-disk one can be non-empty if we've
853 * detected an error and taken the fs readonly, but the
854 * in-memory list had better be clean by this point. */
855 if (!list_empty(&sbi->s_orphan))
856 dump_orphan_list(sb, sbi);
857 J_ASSERT(list_empty(&sbi->s_orphan));
858
859 invalidate_bdev(sb->s_bdev);
860 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
861 /*
862 * Invalidate the journal device's buffers. We don't want them
863 * floating about in memory - the physical journal device may
864 * hotswapped, and it breaks the `ro-after' testing code.
865 */
866 sync_blockdev(sbi->journal_bdev);
867 invalidate_bdev(sbi->journal_bdev);
868 ext4_blkdev_remove(sbi);
869 }
870 if (sbi->s_mmp_tsk)
871 kthread_stop(sbi->s_mmp_tsk);
872 sb->s_fs_info = NULL;
873 /*
874 * Now that we are completely done shutting down the
875 * superblock, we need to actually destroy the kobject.
876 */
877 unlock_super(sb);
878 kobject_put(&sbi->s_kobj);
879 wait_for_completion(&sbi->s_kobj_unregister);
880 kfree(sbi->s_blockgroup_lock);
881 kfree(sbi);
882 }
883
884 static struct kmem_cache *ext4_inode_cachep;
885
886 /*
887 * Called inside transaction, so use GFP_NOFS
888 */
889 static struct inode *ext4_alloc_inode(struct super_block *sb)
890 {
891 struct ext4_inode_info *ei;
892
893 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
894 if (!ei)
895 return NULL;
896
897 ei->vfs_inode.i_version = 1;
898 ei->vfs_inode.i_data.writeback_index = 0;
899 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
900 INIT_LIST_HEAD(&ei->i_prealloc_list);
901 spin_lock_init(&ei->i_prealloc_lock);
902 ei->i_reserved_data_blocks = 0;
903 ei->i_reserved_meta_blocks = 0;
904 ei->i_allocated_meta_blocks = 0;
905 ei->i_da_metadata_calc_len = 0;
906 spin_lock_init(&(ei->i_block_reservation_lock));
907 #ifdef CONFIG_QUOTA
908 ei->i_reserved_quota = 0;
909 #endif
910 ei->jinode = NULL;
911 INIT_LIST_HEAD(&ei->i_completed_io_list);
912 spin_lock_init(&ei->i_completed_io_lock);
913 ei->cur_aio_dio = NULL;
914 ei->i_sync_tid = 0;
915 ei->i_datasync_tid = 0;
916 atomic_set(&ei->i_ioend_count, 0);
917 atomic_set(&ei->i_aiodio_unwritten, 0);
918
919 return &ei->vfs_inode;
920 }
921
922 static int ext4_drop_inode(struct inode *inode)
923 {
924 int drop = generic_drop_inode(inode);
925
926 trace_ext4_drop_inode(inode, drop);
927 return drop;
928 }
929
930 static void ext4_i_callback(struct rcu_head *head)
931 {
932 struct inode *inode = container_of(head, struct inode, i_rcu);
933 INIT_LIST_HEAD(&inode->i_dentry);
934 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
935 }
936
937 static void ext4_destroy_inode(struct inode *inode)
938 {
939 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
940 ext4_msg(inode->i_sb, KERN_ERR,
941 "Inode %lu (%p): orphan list check failed!",
942 inode->i_ino, EXT4_I(inode));
943 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
944 EXT4_I(inode), sizeof(struct ext4_inode_info),
945 true);
946 dump_stack();
947 }
948 call_rcu(&inode->i_rcu, ext4_i_callback);
949 }
950
951 static void init_once(void *foo)
952 {
953 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
954
955 INIT_LIST_HEAD(&ei->i_orphan);
956 #ifdef CONFIG_EXT4_FS_XATTR
957 init_rwsem(&ei->xattr_sem);
958 #endif
959 init_rwsem(&ei->i_data_sem);
960 inode_init_once(&ei->vfs_inode);
961 }
962
963 static int init_inodecache(void)
964 {
965 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
966 sizeof(struct ext4_inode_info),
967 0, (SLAB_RECLAIM_ACCOUNT|
968 SLAB_MEM_SPREAD),
969 init_once);
970 if (ext4_inode_cachep == NULL)
971 return -ENOMEM;
972 return 0;
973 }
974
975 static void destroy_inodecache(void)
976 {
977 kmem_cache_destroy(ext4_inode_cachep);
978 }
979
980 void ext4_clear_inode(struct inode *inode)
981 {
982 invalidate_inode_buffers(inode);
983 end_writeback(inode);
984 dquot_drop(inode);
985 ext4_discard_preallocations(inode);
986 if (EXT4_I(inode)->jinode) {
987 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
988 EXT4_I(inode)->jinode);
989 jbd2_free_inode(EXT4_I(inode)->jinode);
990 EXT4_I(inode)->jinode = NULL;
991 }
992 }
993
994 static inline void ext4_show_quota_options(struct seq_file *seq,
995 struct super_block *sb)
996 {
997 #if defined(CONFIG_QUOTA)
998 struct ext4_sb_info *sbi = EXT4_SB(sb);
999
1000 if (sbi->s_jquota_fmt) {
1001 char *fmtname = "";
1002
1003 switch (sbi->s_jquota_fmt) {
1004 case QFMT_VFS_OLD:
1005 fmtname = "vfsold";
1006 break;
1007 case QFMT_VFS_V0:
1008 fmtname = "vfsv0";
1009 break;
1010 case QFMT_VFS_V1:
1011 fmtname = "vfsv1";
1012 break;
1013 }
1014 seq_printf(seq, ",jqfmt=%s", fmtname);
1015 }
1016
1017 if (sbi->s_qf_names[USRQUOTA])
1018 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1019
1020 if (sbi->s_qf_names[GRPQUOTA])
1021 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1022
1023 if (test_opt(sb, USRQUOTA))
1024 seq_puts(seq, ",usrquota");
1025
1026 if (test_opt(sb, GRPQUOTA))
1027 seq_puts(seq, ",grpquota");
1028 #endif
1029 }
1030
1031 /*
1032 * Show an option if
1033 * - it's set to a non-default value OR
1034 * - if the per-sb default is different from the global default
1035 */
1036 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
1037 {
1038 int def_errors;
1039 unsigned long def_mount_opts;
1040 struct super_block *sb = vfs->mnt_sb;
1041 struct ext4_sb_info *sbi = EXT4_SB(sb);
1042 struct ext4_super_block *es = sbi->s_es;
1043
1044 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1045 def_errors = le16_to_cpu(es->s_errors);
1046
1047 if (sbi->s_sb_block != 1)
1048 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1049 if (test_opt(sb, MINIX_DF))
1050 seq_puts(seq, ",minixdf");
1051 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1052 seq_puts(seq, ",grpid");
1053 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1054 seq_puts(seq, ",nogrpid");
1055 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1056 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1057 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1058 }
1059 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1060 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1061 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1062 }
1063 if (test_opt(sb, ERRORS_RO)) {
1064 if (def_errors == EXT4_ERRORS_PANIC ||
1065 def_errors == EXT4_ERRORS_CONTINUE) {
1066 seq_puts(seq, ",errors=remount-ro");
1067 }
1068 }
1069 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1070 seq_puts(seq, ",errors=continue");
1071 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1072 seq_puts(seq, ",errors=panic");
1073 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1074 seq_puts(seq, ",nouid32");
1075 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1076 seq_puts(seq, ",debug");
1077 #ifdef CONFIG_EXT4_FS_XATTR
1078 if (test_opt(sb, XATTR_USER))
1079 seq_puts(seq, ",user_xattr");
1080 if (!test_opt(sb, XATTR_USER))
1081 seq_puts(seq, ",nouser_xattr");
1082 #endif
1083 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1084 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1085 seq_puts(seq, ",acl");
1086 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1087 seq_puts(seq, ",noacl");
1088 #endif
1089 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1090 seq_printf(seq, ",commit=%u",
1091 (unsigned) (sbi->s_commit_interval / HZ));
1092 }
1093 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1094 seq_printf(seq, ",min_batch_time=%u",
1095 (unsigned) sbi->s_min_batch_time);
1096 }
1097 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1098 seq_printf(seq, ",max_batch_time=%u",
1099 (unsigned) sbi->s_min_batch_time);
1100 }
1101
1102 /*
1103 * We're changing the default of barrier mount option, so
1104 * let's always display its mount state so it's clear what its
1105 * status is.
1106 */
1107 seq_puts(seq, ",barrier=");
1108 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1109 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1110 seq_puts(seq, ",journal_async_commit");
1111 else if (test_opt(sb, JOURNAL_CHECKSUM))
1112 seq_puts(seq, ",journal_checksum");
1113 if (test_opt(sb, I_VERSION))
1114 seq_puts(seq, ",i_version");
1115 if (!test_opt(sb, DELALLOC) &&
1116 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1117 seq_puts(seq, ",nodelalloc");
1118
1119 if (!test_opt(sb, MBLK_IO_SUBMIT))
1120 seq_puts(seq, ",nomblk_io_submit");
1121 if (sbi->s_stripe)
1122 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1123 /*
1124 * journal mode get enabled in different ways
1125 * So just print the value even if we didn't specify it
1126 */
1127 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1128 seq_puts(seq, ",data=journal");
1129 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1130 seq_puts(seq, ",data=ordered");
1131 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1132 seq_puts(seq, ",data=writeback");
1133
1134 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1135 seq_printf(seq, ",inode_readahead_blks=%u",
1136 sbi->s_inode_readahead_blks);
1137
1138 if (test_opt(sb, DATA_ERR_ABORT))
1139 seq_puts(seq, ",data_err=abort");
1140
1141 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1142 seq_puts(seq, ",noauto_da_alloc");
1143
1144 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1145 seq_puts(seq, ",discard");
1146
1147 if (test_opt(sb, NOLOAD))
1148 seq_puts(seq, ",norecovery");
1149
1150 if (test_opt(sb, DIOREAD_NOLOCK))
1151 seq_puts(seq, ",dioread_nolock");
1152
1153 if (test_opt(sb, BLOCK_VALIDITY) &&
1154 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1155 seq_puts(seq, ",block_validity");
1156
1157 if (!test_opt(sb, INIT_INODE_TABLE))
1158 seq_puts(seq, ",noinit_itable");
1159 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1160 seq_printf(seq, ",init_itable=%u",
1161 (unsigned) sbi->s_li_wait_mult);
1162
1163 ext4_show_quota_options(seq, sb);
1164
1165 return 0;
1166 }
1167
1168 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1169 u64 ino, u32 generation)
1170 {
1171 struct inode *inode;
1172
1173 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1174 return ERR_PTR(-ESTALE);
1175 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1176 return ERR_PTR(-ESTALE);
1177
1178 /* iget isn't really right if the inode is currently unallocated!!
1179 *
1180 * ext4_read_inode will return a bad_inode if the inode had been
1181 * deleted, so we should be safe.
1182 *
1183 * Currently we don't know the generation for parent directory, so
1184 * a generation of 0 means "accept any"
1185 */
1186 inode = ext4_iget(sb, ino);
1187 if (IS_ERR(inode))
1188 return ERR_CAST(inode);
1189 if (generation && inode->i_generation != generation) {
1190 iput(inode);
1191 return ERR_PTR(-ESTALE);
1192 }
1193
1194 return inode;
1195 }
1196
1197 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1198 int fh_len, int fh_type)
1199 {
1200 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1201 ext4_nfs_get_inode);
1202 }
1203
1204 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1205 int fh_len, int fh_type)
1206 {
1207 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1208 ext4_nfs_get_inode);
1209 }
1210
1211 /*
1212 * Try to release metadata pages (indirect blocks, directories) which are
1213 * mapped via the block device. Since these pages could have journal heads
1214 * which would prevent try_to_free_buffers() from freeing them, we must use
1215 * jbd2 layer's try_to_free_buffers() function to release them.
1216 */
1217 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1218 gfp_t wait)
1219 {
1220 journal_t *journal = EXT4_SB(sb)->s_journal;
1221
1222 WARN_ON(PageChecked(page));
1223 if (!page_has_buffers(page))
1224 return 0;
1225 if (journal)
1226 return jbd2_journal_try_to_free_buffers(journal, page,
1227 wait & ~__GFP_WAIT);
1228 return try_to_free_buffers(page);
1229 }
1230
1231 #ifdef CONFIG_QUOTA
1232 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1233 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1234
1235 static int ext4_write_dquot(struct dquot *dquot);
1236 static int ext4_acquire_dquot(struct dquot *dquot);
1237 static int ext4_release_dquot(struct dquot *dquot);
1238 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1239 static int ext4_write_info(struct super_block *sb, int type);
1240 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1241 struct path *path);
1242 static int ext4_quota_off(struct super_block *sb, int type);
1243 static int ext4_quota_on_mount(struct super_block *sb, int type);
1244 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1245 size_t len, loff_t off);
1246 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1247 const char *data, size_t len, loff_t off);
1248
1249 static const struct dquot_operations ext4_quota_operations = {
1250 .get_reserved_space = ext4_get_reserved_space,
1251 .write_dquot = ext4_write_dquot,
1252 .acquire_dquot = ext4_acquire_dquot,
1253 .release_dquot = ext4_release_dquot,
1254 .mark_dirty = ext4_mark_dquot_dirty,
1255 .write_info = ext4_write_info,
1256 .alloc_dquot = dquot_alloc,
1257 .destroy_dquot = dquot_destroy,
1258 };
1259
1260 static const struct quotactl_ops ext4_qctl_operations = {
1261 .quota_on = ext4_quota_on,
1262 .quota_off = ext4_quota_off,
1263 .quota_sync = dquot_quota_sync,
1264 .get_info = dquot_get_dqinfo,
1265 .set_info = dquot_set_dqinfo,
1266 .get_dqblk = dquot_get_dqblk,
1267 .set_dqblk = dquot_set_dqblk
1268 };
1269 #endif
1270
1271 static const struct super_operations ext4_sops = {
1272 .alloc_inode = ext4_alloc_inode,
1273 .destroy_inode = ext4_destroy_inode,
1274 .write_inode = ext4_write_inode,
1275 .dirty_inode = ext4_dirty_inode,
1276 .drop_inode = ext4_drop_inode,
1277 .evict_inode = ext4_evict_inode,
1278 .put_super = ext4_put_super,
1279 .sync_fs = ext4_sync_fs,
1280 .freeze_fs = ext4_freeze,
1281 .unfreeze_fs = ext4_unfreeze,
1282 .statfs = ext4_statfs,
1283 .remount_fs = ext4_remount,
1284 .show_options = ext4_show_options,
1285 #ifdef CONFIG_QUOTA
1286 .quota_read = ext4_quota_read,
1287 .quota_write = ext4_quota_write,
1288 #endif
1289 .bdev_try_to_free_page = bdev_try_to_free_page,
1290 };
1291
1292 static const struct super_operations ext4_nojournal_sops = {
1293 .alloc_inode = ext4_alloc_inode,
1294 .destroy_inode = ext4_destroy_inode,
1295 .write_inode = ext4_write_inode,
1296 .dirty_inode = ext4_dirty_inode,
1297 .drop_inode = ext4_drop_inode,
1298 .evict_inode = ext4_evict_inode,
1299 .write_super = ext4_write_super,
1300 .put_super = ext4_put_super,
1301 .statfs = ext4_statfs,
1302 .remount_fs = ext4_remount,
1303 .show_options = ext4_show_options,
1304 #ifdef CONFIG_QUOTA
1305 .quota_read = ext4_quota_read,
1306 .quota_write = ext4_quota_write,
1307 #endif
1308 .bdev_try_to_free_page = bdev_try_to_free_page,
1309 };
1310
1311 static const struct export_operations ext4_export_ops = {
1312 .fh_to_dentry = ext4_fh_to_dentry,
1313 .fh_to_parent = ext4_fh_to_parent,
1314 .get_parent = ext4_get_parent,
1315 };
1316
1317 enum {
1318 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1319 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1320 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1321 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1322 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1323 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1324 Opt_journal_update, Opt_journal_dev,
1325 Opt_journal_checksum, Opt_journal_async_commit,
1326 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1327 Opt_data_err_abort, Opt_data_err_ignore,
1328 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1329 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1330 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1331 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1332 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1333 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1334 Opt_inode_readahead_blks, Opt_journal_ioprio,
1335 Opt_dioread_nolock, Opt_dioread_lock,
1336 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1337 };
1338
1339 static const match_table_t tokens = {
1340 {Opt_bsd_df, "bsddf"},
1341 {Opt_minix_df, "minixdf"},
1342 {Opt_grpid, "grpid"},
1343 {Opt_grpid, "bsdgroups"},
1344 {Opt_nogrpid, "nogrpid"},
1345 {Opt_nogrpid, "sysvgroups"},
1346 {Opt_resgid, "resgid=%u"},
1347 {Opt_resuid, "resuid=%u"},
1348 {Opt_sb, "sb=%u"},
1349 {Opt_err_cont, "errors=continue"},
1350 {Opt_err_panic, "errors=panic"},
1351 {Opt_err_ro, "errors=remount-ro"},
1352 {Opt_nouid32, "nouid32"},
1353 {Opt_debug, "debug"},
1354 {Opt_oldalloc, "oldalloc"},
1355 {Opt_orlov, "orlov"},
1356 {Opt_user_xattr, "user_xattr"},
1357 {Opt_nouser_xattr, "nouser_xattr"},
1358 {Opt_acl, "acl"},
1359 {Opt_noacl, "noacl"},
1360 {Opt_noload, "noload"},
1361 {Opt_noload, "norecovery"},
1362 {Opt_nobh, "nobh"},
1363 {Opt_bh, "bh"},
1364 {Opt_commit, "commit=%u"},
1365 {Opt_min_batch_time, "min_batch_time=%u"},
1366 {Opt_max_batch_time, "max_batch_time=%u"},
1367 {Opt_journal_update, "journal=update"},
1368 {Opt_journal_dev, "journal_dev=%u"},
1369 {Opt_journal_checksum, "journal_checksum"},
1370 {Opt_journal_async_commit, "journal_async_commit"},
1371 {Opt_abort, "abort"},
1372 {Opt_data_journal, "data=journal"},
1373 {Opt_data_ordered, "data=ordered"},
1374 {Opt_data_writeback, "data=writeback"},
1375 {Opt_data_err_abort, "data_err=abort"},
1376 {Opt_data_err_ignore, "data_err=ignore"},
1377 {Opt_offusrjquota, "usrjquota="},
1378 {Opt_usrjquota, "usrjquota=%s"},
1379 {Opt_offgrpjquota, "grpjquota="},
1380 {Opt_grpjquota, "grpjquota=%s"},
1381 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1382 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1383 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1384 {Opt_grpquota, "grpquota"},
1385 {Opt_noquota, "noquota"},
1386 {Opt_quota, "quota"},
1387 {Opt_usrquota, "usrquota"},
1388 {Opt_barrier, "barrier=%u"},
1389 {Opt_barrier, "barrier"},
1390 {Opt_nobarrier, "nobarrier"},
1391 {Opt_i_version, "i_version"},
1392 {Opt_stripe, "stripe=%u"},
1393 {Opt_resize, "resize"},
1394 {Opt_delalloc, "delalloc"},
1395 {Opt_nodelalloc, "nodelalloc"},
1396 {Opt_mblk_io_submit, "mblk_io_submit"},
1397 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1398 {Opt_block_validity, "block_validity"},
1399 {Opt_noblock_validity, "noblock_validity"},
1400 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1401 {Opt_journal_ioprio, "journal_ioprio=%u"},
1402 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1403 {Opt_auto_da_alloc, "auto_da_alloc"},
1404 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1405 {Opt_dioread_nolock, "dioread_nolock"},
1406 {Opt_dioread_lock, "dioread_lock"},
1407 {Opt_discard, "discard"},
1408 {Opt_nodiscard, "nodiscard"},
1409 {Opt_init_itable, "init_itable=%u"},
1410 {Opt_init_itable, "init_itable"},
1411 {Opt_noinit_itable, "noinit_itable"},
1412 {Opt_err, NULL},
1413 };
1414
1415 static ext4_fsblk_t get_sb_block(void **data)
1416 {
1417 ext4_fsblk_t sb_block;
1418 char *options = (char *) *data;
1419
1420 if (!options || strncmp(options, "sb=", 3) != 0)
1421 return 1; /* Default location */
1422
1423 options += 3;
1424 /* TODO: use simple_strtoll with >32bit ext4 */
1425 sb_block = simple_strtoul(options, &options, 0);
1426 if (*options && *options != ',') {
1427 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1428 (char *) *data);
1429 return 1;
1430 }
1431 if (*options == ',')
1432 options++;
1433 *data = (void *) options;
1434
1435 return sb_block;
1436 }
1437
1438 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1439 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1440 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1441
1442 #ifdef CONFIG_QUOTA
1443 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1444 {
1445 struct ext4_sb_info *sbi = EXT4_SB(sb);
1446 char *qname;
1447
1448 if (sb_any_quota_loaded(sb) &&
1449 !sbi->s_qf_names[qtype]) {
1450 ext4_msg(sb, KERN_ERR,
1451 "Cannot change journaled "
1452 "quota options when quota turned on");
1453 return 0;
1454 }
1455 qname = match_strdup(args);
1456 if (!qname) {
1457 ext4_msg(sb, KERN_ERR,
1458 "Not enough memory for storing quotafile name");
1459 return 0;
1460 }
1461 if (sbi->s_qf_names[qtype] &&
1462 strcmp(sbi->s_qf_names[qtype], qname)) {
1463 ext4_msg(sb, KERN_ERR,
1464 "%s quota file already specified", QTYPE2NAME(qtype));
1465 kfree(qname);
1466 return 0;
1467 }
1468 sbi->s_qf_names[qtype] = qname;
1469 if (strchr(sbi->s_qf_names[qtype], '/')) {
1470 ext4_msg(sb, KERN_ERR,
1471 "quotafile must be on filesystem root");
1472 kfree(sbi->s_qf_names[qtype]);
1473 sbi->s_qf_names[qtype] = NULL;
1474 return 0;
1475 }
1476 set_opt(sb, QUOTA);
1477 return 1;
1478 }
1479
1480 static int clear_qf_name(struct super_block *sb, int qtype)
1481 {
1482
1483 struct ext4_sb_info *sbi = EXT4_SB(sb);
1484
1485 if (sb_any_quota_loaded(sb) &&
1486 sbi->s_qf_names[qtype]) {
1487 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1488 " when quota turned on");
1489 return 0;
1490 }
1491 /*
1492 * The space will be released later when all options are confirmed
1493 * to be correct
1494 */
1495 sbi->s_qf_names[qtype] = NULL;
1496 return 1;
1497 }
1498 #endif
1499
1500 static int parse_options(char *options, struct super_block *sb,
1501 unsigned long *journal_devnum,
1502 unsigned int *journal_ioprio,
1503 ext4_fsblk_t *n_blocks_count, int is_remount)
1504 {
1505 struct ext4_sb_info *sbi = EXT4_SB(sb);
1506 char *p;
1507 substring_t args[MAX_OPT_ARGS];
1508 int data_opt = 0;
1509 int option;
1510 #ifdef CONFIG_QUOTA
1511 int qfmt;
1512 #endif
1513
1514 if (!options)
1515 return 1;
1516
1517 while ((p = strsep(&options, ",")) != NULL) {
1518 int token;
1519 if (!*p)
1520 continue;
1521
1522 /*
1523 * Initialize args struct so we know whether arg was
1524 * found; some options take optional arguments.
1525 */
1526 args[0].to = args[0].from = NULL;
1527 token = match_token(p, tokens, args);
1528 switch (token) {
1529 case Opt_bsd_df:
1530 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1531 clear_opt(sb, MINIX_DF);
1532 break;
1533 case Opt_minix_df:
1534 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1535 set_opt(sb, MINIX_DF);
1536
1537 break;
1538 case Opt_grpid:
1539 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1540 set_opt(sb, GRPID);
1541
1542 break;
1543 case Opt_nogrpid:
1544 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1545 clear_opt(sb, GRPID);
1546
1547 break;
1548 case Opt_resuid:
1549 if (match_int(&args[0], &option))
1550 return 0;
1551 sbi->s_resuid = option;
1552 break;
1553 case Opt_resgid:
1554 if (match_int(&args[0], &option))
1555 return 0;
1556 sbi->s_resgid = option;
1557 break;
1558 case Opt_sb:
1559 /* handled by get_sb_block() instead of here */
1560 /* *sb_block = match_int(&args[0]); */
1561 break;
1562 case Opt_err_panic:
1563 clear_opt(sb, ERRORS_CONT);
1564 clear_opt(sb, ERRORS_RO);
1565 set_opt(sb, ERRORS_PANIC);
1566 break;
1567 case Opt_err_ro:
1568 clear_opt(sb, ERRORS_CONT);
1569 clear_opt(sb, ERRORS_PANIC);
1570 set_opt(sb, ERRORS_RO);
1571 break;
1572 case Opt_err_cont:
1573 clear_opt(sb, ERRORS_RO);
1574 clear_opt(sb, ERRORS_PANIC);
1575 set_opt(sb, ERRORS_CONT);
1576 break;
1577 case Opt_nouid32:
1578 set_opt(sb, NO_UID32);
1579 break;
1580 case Opt_debug:
1581 set_opt(sb, DEBUG);
1582 break;
1583 case Opt_oldalloc:
1584 ext4_msg(sb, KERN_WARNING,
1585 "Ignoring deprecated oldalloc option");
1586 break;
1587 case Opt_orlov:
1588 ext4_msg(sb, KERN_WARNING,
1589 "Ignoring deprecated orlov option");
1590 break;
1591 #ifdef CONFIG_EXT4_FS_XATTR
1592 case Opt_user_xattr:
1593 set_opt(sb, XATTR_USER);
1594 break;
1595 case Opt_nouser_xattr:
1596 clear_opt(sb, XATTR_USER);
1597 break;
1598 #else
1599 case Opt_user_xattr:
1600 case Opt_nouser_xattr:
1601 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1602 break;
1603 #endif
1604 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1605 case Opt_acl:
1606 set_opt(sb, POSIX_ACL);
1607 break;
1608 case Opt_noacl:
1609 clear_opt(sb, POSIX_ACL);
1610 break;
1611 #else
1612 case Opt_acl:
1613 case Opt_noacl:
1614 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1615 break;
1616 #endif
1617 case Opt_journal_update:
1618 /* @@@ FIXME */
1619 /* Eventually we will want to be able to create
1620 a journal file here. For now, only allow the
1621 user to specify an existing inode to be the
1622 journal file. */
1623 if (is_remount) {
1624 ext4_msg(sb, KERN_ERR,
1625 "Cannot specify journal on remount");
1626 return 0;
1627 }
1628 set_opt(sb, UPDATE_JOURNAL);
1629 break;
1630 case Opt_journal_dev:
1631 if (is_remount) {
1632 ext4_msg(sb, KERN_ERR,
1633 "Cannot specify journal on remount");
1634 return 0;
1635 }
1636 if (match_int(&args[0], &option))
1637 return 0;
1638 *journal_devnum = option;
1639 break;
1640 case Opt_journal_checksum:
1641 set_opt(sb, JOURNAL_CHECKSUM);
1642 break;
1643 case Opt_journal_async_commit:
1644 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1645 set_opt(sb, JOURNAL_CHECKSUM);
1646 break;
1647 case Opt_noload:
1648 set_opt(sb, NOLOAD);
1649 break;
1650 case Opt_commit:
1651 if (match_int(&args[0], &option))
1652 return 0;
1653 if (option < 0)
1654 return 0;
1655 if (option == 0)
1656 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1657 sbi->s_commit_interval = HZ * option;
1658 break;
1659 case Opt_max_batch_time:
1660 if (match_int(&args[0], &option))
1661 return 0;
1662 if (option < 0)
1663 return 0;
1664 if (option == 0)
1665 option = EXT4_DEF_MAX_BATCH_TIME;
1666 sbi->s_max_batch_time = option;
1667 break;
1668 case Opt_min_batch_time:
1669 if (match_int(&args[0], &option))
1670 return 0;
1671 if (option < 0)
1672 return 0;
1673 sbi->s_min_batch_time = option;
1674 break;
1675 case Opt_data_journal:
1676 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1677 goto datacheck;
1678 case Opt_data_ordered:
1679 data_opt = EXT4_MOUNT_ORDERED_DATA;
1680 goto datacheck;
1681 case Opt_data_writeback:
1682 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1683 datacheck:
1684 if (is_remount) {
1685 if (!sbi->s_journal)
1686 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1687 else if (test_opt(sb, DATA_FLAGS) != data_opt) {
1688 ext4_msg(sb, KERN_ERR,
1689 "Cannot change data mode on remount");
1690 return 0;
1691 }
1692 } else {
1693 clear_opt(sb, DATA_FLAGS);
1694 sbi->s_mount_opt |= data_opt;
1695 }
1696 break;
1697 case Opt_data_err_abort:
1698 set_opt(sb, DATA_ERR_ABORT);
1699 break;
1700 case Opt_data_err_ignore:
1701 clear_opt(sb, DATA_ERR_ABORT);
1702 break;
1703 #ifdef CONFIG_QUOTA
1704 case Opt_usrjquota:
1705 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1706 return 0;
1707 break;
1708 case Opt_grpjquota:
1709 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1710 return 0;
1711 break;
1712 case Opt_offusrjquota:
1713 if (!clear_qf_name(sb, USRQUOTA))
1714 return 0;
1715 break;
1716 case Opt_offgrpjquota:
1717 if (!clear_qf_name(sb, GRPQUOTA))
1718 return 0;
1719 break;
1720
1721 case Opt_jqfmt_vfsold:
1722 qfmt = QFMT_VFS_OLD;
1723 goto set_qf_format;
1724 case Opt_jqfmt_vfsv0:
1725 qfmt = QFMT_VFS_V0;
1726 goto set_qf_format;
1727 case Opt_jqfmt_vfsv1:
1728 qfmt = QFMT_VFS_V1;
1729 set_qf_format:
1730 if (sb_any_quota_loaded(sb) &&
1731 sbi->s_jquota_fmt != qfmt) {
1732 ext4_msg(sb, KERN_ERR, "Cannot change "
1733 "journaled quota options when "
1734 "quota turned on");
1735 return 0;
1736 }
1737 sbi->s_jquota_fmt = qfmt;
1738 break;
1739 case Opt_quota:
1740 case Opt_usrquota:
1741 set_opt(sb, QUOTA);
1742 set_opt(sb, USRQUOTA);
1743 break;
1744 case Opt_grpquota:
1745 set_opt(sb, QUOTA);
1746 set_opt(sb, GRPQUOTA);
1747 break;
1748 case Opt_noquota:
1749 if (sb_any_quota_loaded(sb)) {
1750 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1751 "options when quota turned on");
1752 return 0;
1753 }
1754 clear_opt(sb, QUOTA);
1755 clear_opt(sb, USRQUOTA);
1756 clear_opt(sb, GRPQUOTA);
1757 break;
1758 #else
1759 case Opt_quota:
1760 case Opt_usrquota:
1761 case Opt_grpquota:
1762 ext4_msg(sb, KERN_ERR,
1763 "quota options not supported");
1764 break;
1765 case Opt_usrjquota:
1766 case Opt_grpjquota:
1767 case Opt_offusrjquota:
1768 case Opt_offgrpjquota:
1769 case Opt_jqfmt_vfsold:
1770 case Opt_jqfmt_vfsv0:
1771 case Opt_jqfmt_vfsv1:
1772 ext4_msg(sb, KERN_ERR,
1773 "journaled quota options not supported");
1774 break;
1775 case Opt_noquota:
1776 break;
1777 #endif
1778 case Opt_abort:
1779 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1780 break;
1781 case Opt_nobarrier:
1782 clear_opt(sb, BARRIER);
1783 break;
1784 case Opt_barrier:
1785 if (args[0].from) {
1786 if (match_int(&args[0], &option))
1787 return 0;
1788 } else
1789 option = 1; /* No argument, default to 1 */
1790 if (option)
1791 set_opt(sb, BARRIER);
1792 else
1793 clear_opt(sb, BARRIER);
1794 break;
1795 case Opt_ignore:
1796 break;
1797 case Opt_resize:
1798 if (!is_remount) {
1799 ext4_msg(sb, KERN_ERR,
1800 "resize option only available "
1801 "for remount");
1802 return 0;
1803 }
1804 if (match_int(&args[0], &option) != 0)
1805 return 0;
1806 *n_blocks_count = option;
1807 break;
1808 case Opt_nobh:
1809 ext4_msg(sb, KERN_WARNING,
1810 "Ignoring deprecated nobh option");
1811 break;
1812 case Opt_bh:
1813 ext4_msg(sb, KERN_WARNING,
1814 "Ignoring deprecated bh option");
1815 break;
1816 case Opt_i_version:
1817 set_opt(sb, I_VERSION);
1818 sb->s_flags |= MS_I_VERSION;
1819 break;
1820 case Opt_nodelalloc:
1821 clear_opt(sb, DELALLOC);
1822 clear_opt2(sb, EXPLICIT_DELALLOC);
1823 break;
1824 case Opt_mblk_io_submit:
1825 set_opt(sb, MBLK_IO_SUBMIT);
1826 break;
1827 case Opt_nomblk_io_submit:
1828 clear_opt(sb, MBLK_IO_SUBMIT);
1829 break;
1830 case Opt_stripe:
1831 if (match_int(&args[0], &option))
1832 return 0;
1833 if (option < 0)
1834 return 0;
1835 sbi->s_stripe = option;
1836 break;
1837 case Opt_delalloc:
1838 set_opt(sb, DELALLOC);
1839 set_opt2(sb, EXPLICIT_DELALLOC);
1840 break;
1841 case Opt_block_validity:
1842 set_opt(sb, BLOCK_VALIDITY);
1843 break;
1844 case Opt_noblock_validity:
1845 clear_opt(sb, BLOCK_VALIDITY);
1846 break;
1847 case Opt_inode_readahead_blks:
1848 if (match_int(&args[0], &option))
1849 return 0;
1850 if (option < 0 || option > (1 << 30))
1851 return 0;
1852 if (option && !is_power_of_2(option)) {
1853 ext4_msg(sb, KERN_ERR,
1854 "EXT4-fs: inode_readahead_blks"
1855 " must be a power of 2");
1856 return 0;
1857 }
1858 sbi->s_inode_readahead_blks = option;
1859 break;
1860 case Opt_journal_ioprio:
1861 if (match_int(&args[0], &option))
1862 return 0;
1863 if (option < 0 || option > 7)
1864 break;
1865 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1866 option);
1867 break;
1868 case Opt_noauto_da_alloc:
1869 set_opt(sb, NO_AUTO_DA_ALLOC);
1870 break;
1871 case Opt_auto_da_alloc:
1872 if (args[0].from) {
1873 if (match_int(&args[0], &option))
1874 return 0;
1875 } else
1876 option = 1; /* No argument, default to 1 */
1877 if (option)
1878 clear_opt(sb, NO_AUTO_DA_ALLOC);
1879 else
1880 set_opt(sb,NO_AUTO_DA_ALLOC);
1881 break;
1882 case Opt_discard:
1883 set_opt(sb, DISCARD);
1884 break;
1885 case Opt_nodiscard:
1886 clear_opt(sb, DISCARD);
1887 break;
1888 case Opt_dioread_nolock:
1889 set_opt(sb, DIOREAD_NOLOCK);
1890 break;
1891 case Opt_dioread_lock:
1892 clear_opt(sb, DIOREAD_NOLOCK);
1893 break;
1894 case Opt_init_itable:
1895 set_opt(sb, INIT_INODE_TABLE);
1896 if (args[0].from) {
1897 if (match_int(&args[0], &option))
1898 return 0;
1899 } else
1900 option = EXT4_DEF_LI_WAIT_MULT;
1901 if (option < 0)
1902 return 0;
1903 sbi->s_li_wait_mult = option;
1904 break;
1905 case Opt_noinit_itable:
1906 clear_opt(sb, INIT_INODE_TABLE);
1907 break;
1908 default:
1909 ext4_msg(sb, KERN_ERR,
1910 "Unrecognized mount option \"%s\" "
1911 "or missing value", p);
1912 return 0;
1913 }
1914 }
1915 #ifdef CONFIG_QUOTA
1916 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1917 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1918 clear_opt(sb, USRQUOTA);
1919
1920 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1921 clear_opt(sb, GRPQUOTA);
1922
1923 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1924 ext4_msg(sb, KERN_ERR, "old and new quota "
1925 "format mixing");
1926 return 0;
1927 }
1928
1929 if (!sbi->s_jquota_fmt) {
1930 ext4_msg(sb, KERN_ERR, "journaled quota format "
1931 "not specified");
1932 return 0;
1933 }
1934 } else {
1935 if (sbi->s_jquota_fmt) {
1936 ext4_msg(sb, KERN_ERR, "journaled quota format "
1937 "specified with no journaling "
1938 "enabled");
1939 return 0;
1940 }
1941 }
1942 #endif
1943 return 1;
1944 }
1945
1946 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1947 int read_only)
1948 {
1949 struct ext4_sb_info *sbi = EXT4_SB(sb);
1950 int res = 0;
1951
1952 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1953 ext4_msg(sb, KERN_ERR, "revision level too high, "
1954 "forcing read-only mode");
1955 res = MS_RDONLY;
1956 }
1957 if (read_only)
1958 goto done;
1959 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1960 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1961 "running e2fsck is recommended");
1962 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1963 ext4_msg(sb, KERN_WARNING,
1964 "warning: mounting fs with errors, "
1965 "running e2fsck is recommended");
1966 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1967 le16_to_cpu(es->s_mnt_count) >=
1968 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1969 ext4_msg(sb, KERN_WARNING,
1970 "warning: maximal mount count reached, "
1971 "running e2fsck is recommended");
1972 else if (le32_to_cpu(es->s_checkinterval) &&
1973 (le32_to_cpu(es->s_lastcheck) +
1974 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1975 ext4_msg(sb, KERN_WARNING,
1976 "warning: checktime reached, "
1977 "running e2fsck is recommended");
1978 if (!sbi->s_journal)
1979 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1980 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1981 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1982 le16_add_cpu(&es->s_mnt_count, 1);
1983 es->s_mtime = cpu_to_le32(get_seconds());
1984 ext4_update_dynamic_rev(sb);
1985 if (sbi->s_journal)
1986 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1987
1988 ext4_commit_super(sb, 1);
1989 done:
1990 if (test_opt(sb, DEBUG))
1991 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1992 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1993 sb->s_blocksize,
1994 sbi->s_groups_count,
1995 EXT4_BLOCKS_PER_GROUP(sb),
1996 EXT4_INODES_PER_GROUP(sb),
1997 sbi->s_mount_opt, sbi->s_mount_opt2);
1998
1999 cleancache_init_fs(sb);
2000 return res;
2001 }
2002
2003 static int ext4_fill_flex_info(struct super_block *sb)
2004 {
2005 struct ext4_sb_info *sbi = EXT4_SB(sb);
2006 struct ext4_group_desc *gdp = NULL;
2007 ext4_group_t flex_group_count;
2008 ext4_group_t flex_group;
2009 int groups_per_flex = 0;
2010 size_t size;
2011 int i;
2012
2013 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2014 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
2015
2016 if (groups_per_flex < 2) {
2017 sbi->s_log_groups_per_flex = 0;
2018 return 1;
2019 }
2020
2021 /* We allocate both existing and potentially added groups */
2022 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2023 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2024 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2025 size = flex_group_count * sizeof(struct flex_groups);
2026 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2027 if (sbi->s_flex_groups == NULL) {
2028 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2029 flex_group_count);
2030 goto failed;
2031 }
2032
2033 for (i = 0; i < sbi->s_groups_count; i++) {
2034 gdp = ext4_get_group_desc(sb, i, NULL);
2035
2036 flex_group = ext4_flex_group(sbi, i);
2037 atomic_add(ext4_free_inodes_count(sb, gdp),
2038 &sbi->s_flex_groups[flex_group].free_inodes);
2039 atomic_add(ext4_free_group_clusters(sb, gdp),
2040 &sbi->s_flex_groups[flex_group].free_clusters);
2041 atomic_add(ext4_used_dirs_count(sb, gdp),
2042 &sbi->s_flex_groups[flex_group].used_dirs);
2043 }
2044
2045 return 1;
2046 failed:
2047 return 0;
2048 }
2049
2050 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2051 struct ext4_group_desc *gdp)
2052 {
2053 __u16 crc = 0;
2054
2055 if (sbi->s_es->s_feature_ro_compat &
2056 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2057 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2058 __le32 le_group = cpu_to_le32(block_group);
2059
2060 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2061 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2062 crc = crc16(crc, (__u8 *)gdp, offset);
2063 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2064 /* for checksum of struct ext4_group_desc do the rest...*/
2065 if ((sbi->s_es->s_feature_incompat &
2066 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2067 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2068 crc = crc16(crc, (__u8 *)gdp + offset,
2069 le16_to_cpu(sbi->s_es->s_desc_size) -
2070 offset);
2071 }
2072
2073 return cpu_to_le16(crc);
2074 }
2075
2076 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2077 struct ext4_group_desc *gdp)
2078 {
2079 if ((sbi->s_es->s_feature_ro_compat &
2080 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2081 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2082 return 0;
2083
2084 return 1;
2085 }
2086
2087 /* Called at mount-time, super-block is locked */
2088 static int ext4_check_descriptors(struct super_block *sb,
2089 ext4_group_t *first_not_zeroed)
2090 {
2091 struct ext4_sb_info *sbi = EXT4_SB(sb);
2092 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2093 ext4_fsblk_t last_block;
2094 ext4_fsblk_t block_bitmap;
2095 ext4_fsblk_t inode_bitmap;
2096 ext4_fsblk_t inode_table;
2097 int flexbg_flag = 0;
2098 ext4_group_t i, grp = sbi->s_groups_count;
2099
2100 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2101 flexbg_flag = 1;
2102
2103 ext4_debug("Checking group descriptors");
2104
2105 for (i = 0; i < sbi->s_groups_count; i++) {
2106 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2107
2108 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2109 last_block = ext4_blocks_count(sbi->s_es) - 1;
2110 else
2111 last_block = first_block +
2112 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2113
2114 if ((grp == sbi->s_groups_count) &&
2115 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2116 grp = i;
2117
2118 block_bitmap = ext4_block_bitmap(sb, gdp);
2119 if (block_bitmap < first_block || block_bitmap > last_block) {
2120 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2121 "Block bitmap for group %u not in group "
2122 "(block %llu)!", i, block_bitmap);
2123 return 0;
2124 }
2125 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2126 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2127 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2128 "Inode bitmap for group %u not in group "
2129 "(block %llu)!", i, inode_bitmap);
2130 return 0;
2131 }
2132 inode_table = ext4_inode_table(sb, gdp);
2133 if (inode_table < first_block ||
2134 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2135 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2136 "Inode table for group %u not in group "
2137 "(block %llu)!", i, inode_table);
2138 return 0;
2139 }
2140 ext4_lock_group(sb, i);
2141 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2142 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2143 "Checksum for group %u failed (%u!=%u)",
2144 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2145 gdp)), le16_to_cpu(gdp->bg_checksum));
2146 if (!(sb->s_flags & MS_RDONLY)) {
2147 ext4_unlock_group(sb, i);
2148 return 0;
2149 }
2150 }
2151 ext4_unlock_group(sb, i);
2152 if (!flexbg_flag)
2153 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2154 }
2155 if (NULL != first_not_zeroed)
2156 *first_not_zeroed = grp;
2157
2158 ext4_free_blocks_count_set(sbi->s_es,
2159 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2160 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2161 return 1;
2162 }
2163
2164 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2165 * the superblock) which were deleted from all directories, but held open by
2166 * a process at the time of a crash. We walk the list and try to delete these
2167 * inodes at recovery time (only with a read-write filesystem).
2168 *
2169 * In order to keep the orphan inode chain consistent during traversal (in
2170 * case of crash during recovery), we link each inode into the superblock
2171 * orphan list_head and handle it the same way as an inode deletion during
2172 * normal operation (which journals the operations for us).
2173 *
2174 * We only do an iget() and an iput() on each inode, which is very safe if we
2175 * accidentally point at an in-use or already deleted inode. The worst that
2176 * can happen in this case is that we get a "bit already cleared" message from
2177 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2178 * e2fsck was run on this filesystem, and it must have already done the orphan
2179 * inode cleanup for us, so we can safely abort without any further action.
2180 */
2181 static void ext4_orphan_cleanup(struct super_block *sb,
2182 struct ext4_super_block *es)
2183 {
2184 unsigned int s_flags = sb->s_flags;
2185 int nr_orphans = 0, nr_truncates = 0;
2186 #ifdef CONFIG_QUOTA
2187 int i;
2188 #endif
2189 if (!es->s_last_orphan) {
2190 jbd_debug(4, "no orphan inodes to clean up\n");
2191 return;
2192 }
2193
2194 if (bdev_read_only(sb->s_bdev)) {
2195 ext4_msg(sb, KERN_ERR, "write access "
2196 "unavailable, skipping orphan cleanup");
2197 return;
2198 }
2199
2200 /* Check if feature set would not allow a r/w mount */
2201 if (!ext4_feature_set_ok(sb, 0)) {
2202 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2203 "unknown ROCOMPAT features");
2204 return;
2205 }
2206
2207 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2208 if (es->s_last_orphan)
2209 jbd_debug(1, "Errors on filesystem, "
2210 "clearing orphan list.\n");
2211 es->s_last_orphan = 0;
2212 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2213 return;
2214 }
2215
2216 if (s_flags & MS_RDONLY) {
2217 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2218 sb->s_flags &= ~MS_RDONLY;
2219 }
2220 #ifdef CONFIG_QUOTA
2221 /* Needed for iput() to work correctly and not trash data */
2222 sb->s_flags |= MS_ACTIVE;
2223 /* Turn on quotas so that they are updated correctly */
2224 for (i = 0; i < MAXQUOTAS; i++) {
2225 if (EXT4_SB(sb)->s_qf_names[i]) {
2226 int ret = ext4_quota_on_mount(sb, i);
2227 if (ret < 0)
2228 ext4_msg(sb, KERN_ERR,
2229 "Cannot turn on journaled "
2230 "quota: error %d", ret);
2231 }
2232 }
2233 #endif
2234
2235 while (es->s_last_orphan) {
2236 struct inode *inode;
2237
2238 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2239 if (IS_ERR(inode)) {
2240 es->s_last_orphan = 0;
2241 break;
2242 }
2243
2244 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2245 dquot_initialize(inode);
2246 if (inode->i_nlink) {
2247 ext4_msg(sb, KERN_DEBUG,
2248 "%s: truncating inode %lu to %lld bytes",
2249 __func__, inode->i_ino, inode->i_size);
2250 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2251 inode->i_ino, inode->i_size);
2252 ext4_truncate(inode);
2253 nr_truncates++;
2254 } else {
2255 ext4_msg(sb, KERN_DEBUG,
2256 "%s: deleting unreferenced inode %lu",
2257 __func__, inode->i_ino);
2258 jbd_debug(2, "deleting unreferenced inode %lu\n",
2259 inode->i_ino);
2260 nr_orphans++;
2261 }
2262 iput(inode); /* The delete magic happens here! */
2263 }
2264
2265 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2266
2267 if (nr_orphans)
2268 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2269 PLURAL(nr_orphans));
2270 if (nr_truncates)
2271 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2272 PLURAL(nr_truncates));
2273 #ifdef CONFIG_QUOTA
2274 /* Turn quotas off */
2275 for (i = 0; i < MAXQUOTAS; i++) {
2276 if (sb_dqopt(sb)->files[i])
2277 dquot_quota_off(sb, i);
2278 }
2279 #endif
2280 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2281 }
2282
2283 /*
2284 * Maximal extent format file size.
2285 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2286 * extent format containers, within a sector_t, and within i_blocks
2287 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2288 * so that won't be a limiting factor.
2289 *
2290 * However there is other limiting factor. We do store extents in the form
2291 * of starting block and length, hence the resulting length of the extent
2292 * covering maximum file size must fit into on-disk format containers as
2293 * well. Given that length is always by 1 unit bigger than max unit (because
2294 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2295 *
2296 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2297 */
2298 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2299 {
2300 loff_t res;
2301 loff_t upper_limit = MAX_LFS_FILESIZE;
2302
2303 /* small i_blocks in vfs inode? */
2304 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2305 /*
2306 * CONFIG_LBDAF is not enabled implies the inode
2307 * i_block represent total blocks in 512 bytes
2308 * 32 == size of vfs inode i_blocks * 8
2309 */
2310 upper_limit = (1LL << 32) - 1;
2311
2312 /* total blocks in file system block size */
2313 upper_limit >>= (blkbits - 9);
2314 upper_limit <<= blkbits;
2315 }
2316
2317 /*
2318 * 32-bit extent-start container, ee_block. We lower the maxbytes
2319 * by one fs block, so ee_len can cover the extent of maximum file
2320 * size
2321 */
2322 res = (1LL << 32) - 1;
2323 res <<= blkbits;
2324
2325 /* Sanity check against vm- & vfs- imposed limits */
2326 if (res > upper_limit)
2327 res = upper_limit;
2328
2329 return res;
2330 }
2331
2332 /*
2333 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2334 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2335 * We need to be 1 filesystem block less than the 2^48 sector limit.
2336 */
2337 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2338 {
2339 loff_t res = EXT4_NDIR_BLOCKS;
2340 int meta_blocks;
2341 loff_t upper_limit;
2342 /* This is calculated to be the largest file size for a dense, block
2343 * mapped file such that the file's total number of 512-byte sectors,
2344 * including data and all indirect blocks, does not exceed (2^48 - 1).
2345 *
2346 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2347 * number of 512-byte sectors of the file.
2348 */
2349
2350 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2351 /*
2352 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2353 * the inode i_block field represents total file blocks in
2354 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2355 */
2356 upper_limit = (1LL << 32) - 1;
2357
2358 /* total blocks in file system block size */
2359 upper_limit >>= (bits - 9);
2360
2361 } else {
2362 /*
2363 * We use 48 bit ext4_inode i_blocks
2364 * With EXT4_HUGE_FILE_FL set the i_blocks
2365 * represent total number of blocks in
2366 * file system block size
2367 */
2368 upper_limit = (1LL << 48) - 1;
2369
2370 }
2371
2372 /* indirect blocks */
2373 meta_blocks = 1;
2374 /* double indirect blocks */
2375 meta_blocks += 1 + (1LL << (bits-2));
2376 /* tripple indirect blocks */
2377 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2378
2379 upper_limit -= meta_blocks;
2380 upper_limit <<= bits;
2381
2382 res += 1LL << (bits-2);
2383 res += 1LL << (2*(bits-2));
2384 res += 1LL << (3*(bits-2));
2385 res <<= bits;
2386 if (res > upper_limit)
2387 res = upper_limit;
2388
2389 if (res > MAX_LFS_FILESIZE)
2390 res = MAX_LFS_FILESIZE;
2391
2392 return res;
2393 }
2394
2395 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2396 ext4_fsblk_t logical_sb_block, int nr)
2397 {
2398 struct ext4_sb_info *sbi = EXT4_SB(sb);
2399 ext4_group_t bg, first_meta_bg;
2400 int has_super = 0;
2401
2402 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2403
2404 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2405 nr < first_meta_bg)
2406 return logical_sb_block + nr + 1;
2407 bg = sbi->s_desc_per_block * nr;
2408 if (ext4_bg_has_super(sb, bg))
2409 has_super = 1;
2410
2411 return (has_super + ext4_group_first_block_no(sb, bg));
2412 }
2413
2414 /**
2415 * ext4_get_stripe_size: Get the stripe size.
2416 * @sbi: In memory super block info
2417 *
2418 * If we have specified it via mount option, then
2419 * use the mount option value. If the value specified at mount time is
2420 * greater than the blocks per group use the super block value.
2421 * If the super block value is greater than blocks per group return 0.
2422 * Allocator needs it be less than blocks per group.
2423 *
2424 */
2425 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2426 {
2427 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2428 unsigned long stripe_width =
2429 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2430 int ret;
2431
2432 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2433 ret = sbi->s_stripe;
2434 else if (stripe_width <= sbi->s_blocks_per_group)
2435 ret = stripe_width;
2436 else if (stride <= sbi->s_blocks_per_group)
2437 ret = stride;
2438 else
2439 ret = 0;
2440
2441 /*
2442 * If the stripe width is 1, this makes no sense and
2443 * we set it to 0 to turn off stripe handling code.
2444 */
2445 if (ret <= 1)
2446 ret = 0;
2447
2448 return ret;
2449 }
2450
2451 /* sysfs supprt */
2452
2453 struct ext4_attr {
2454 struct attribute attr;
2455 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2456 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2457 const char *, size_t);
2458 int offset;
2459 };
2460
2461 static int parse_strtoul(const char *buf,
2462 unsigned long max, unsigned long *value)
2463 {
2464 char *endp;
2465
2466 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2467 endp = skip_spaces(endp);
2468 if (*endp || *value > max)
2469 return -EINVAL;
2470
2471 return 0;
2472 }
2473
2474 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2475 struct ext4_sb_info *sbi,
2476 char *buf)
2477 {
2478 return snprintf(buf, PAGE_SIZE, "%llu\n",
2479 (s64) EXT4_C2B(sbi,
2480 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2481 }
2482
2483 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2484 struct ext4_sb_info *sbi, char *buf)
2485 {
2486 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2487
2488 if (!sb->s_bdev->bd_part)
2489 return snprintf(buf, PAGE_SIZE, "0\n");
2490 return snprintf(buf, PAGE_SIZE, "%lu\n",
2491 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2492 sbi->s_sectors_written_start) >> 1);
2493 }
2494
2495 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2496 struct ext4_sb_info *sbi, char *buf)
2497 {
2498 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2499
2500 if (!sb->s_bdev->bd_part)
2501 return snprintf(buf, PAGE_SIZE, "0\n");
2502 return snprintf(buf, PAGE_SIZE, "%llu\n",
2503 (unsigned long long)(sbi->s_kbytes_written +
2504 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2505 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2506 }
2507
2508 static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2509 struct ext4_sb_info *sbi, char *buf)
2510 {
2511 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2512 }
2513
2514 static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2515 struct ext4_sb_info *sbi, char *buf)
2516 {
2517 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2518 }
2519
2520 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2521 struct ext4_sb_info *sbi,
2522 const char *buf, size_t count)
2523 {
2524 unsigned long t;
2525
2526 if (parse_strtoul(buf, 0x40000000, &t))
2527 return -EINVAL;
2528
2529 if (t && !is_power_of_2(t))
2530 return -EINVAL;
2531
2532 sbi->s_inode_readahead_blks = t;
2533 return count;
2534 }
2535
2536 static ssize_t sbi_ui_show(struct ext4_attr *a,
2537 struct ext4_sb_info *sbi, char *buf)
2538 {
2539 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2540
2541 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2542 }
2543
2544 static ssize_t sbi_ui_store(struct ext4_attr *a,
2545 struct ext4_sb_info *sbi,
2546 const char *buf, size_t count)
2547 {
2548 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2549 unsigned long t;
2550
2551 if (parse_strtoul(buf, 0xffffffff, &t))
2552 return -EINVAL;
2553 *ui = t;
2554 return count;
2555 }
2556
2557 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2558 static struct ext4_attr ext4_attr_##_name = { \
2559 .attr = {.name = __stringify(_name), .mode = _mode }, \
2560 .show = _show, \
2561 .store = _store, \
2562 .offset = offsetof(struct ext4_sb_info, _elname), \
2563 }
2564 #define EXT4_ATTR(name, mode, show, store) \
2565 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2566
2567 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2568 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2569 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2570 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2571 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2572 #define ATTR_LIST(name) &ext4_attr_##name.attr
2573
2574 EXT4_RO_ATTR(delayed_allocation_blocks);
2575 EXT4_RO_ATTR(session_write_kbytes);
2576 EXT4_RO_ATTR(lifetime_write_kbytes);
2577 EXT4_RO_ATTR(extent_cache_hits);
2578 EXT4_RO_ATTR(extent_cache_misses);
2579 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2580 inode_readahead_blks_store, s_inode_readahead_blks);
2581 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2582 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2583 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2584 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2585 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2586 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2587 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2588 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2589
2590 static struct attribute *ext4_attrs[] = {
2591 ATTR_LIST(delayed_allocation_blocks),
2592 ATTR_LIST(session_write_kbytes),
2593 ATTR_LIST(lifetime_write_kbytes),
2594 ATTR_LIST(extent_cache_hits),
2595 ATTR_LIST(extent_cache_misses),
2596 ATTR_LIST(inode_readahead_blks),
2597 ATTR_LIST(inode_goal),
2598 ATTR_LIST(mb_stats),
2599 ATTR_LIST(mb_max_to_scan),
2600 ATTR_LIST(mb_min_to_scan),
2601 ATTR_LIST(mb_order2_req),
2602 ATTR_LIST(mb_stream_req),
2603 ATTR_LIST(mb_group_prealloc),
2604 ATTR_LIST(max_writeback_mb_bump),
2605 NULL,
2606 };
2607
2608 /* Features this copy of ext4 supports */
2609 EXT4_INFO_ATTR(lazy_itable_init);
2610 EXT4_INFO_ATTR(batched_discard);
2611
2612 static struct attribute *ext4_feat_attrs[] = {
2613 ATTR_LIST(lazy_itable_init),
2614 ATTR_LIST(batched_discard),
2615 NULL,
2616 };
2617
2618 static ssize_t ext4_attr_show(struct kobject *kobj,
2619 struct attribute *attr, char *buf)
2620 {
2621 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2622 s_kobj);
2623 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2624
2625 return a->show ? a->show(a, sbi, buf) : 0;
2626 }
2627
2628 static ssize_t ext4_attr_store(struct kobject *kobj,
2629 struct attribute *attr,
2630 const char *buf, size_t len)
2631 {
2632 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2633 s_kobj);
2634 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2635
2636 return a->store ? a->store(a, sbi, buf, len) : 0;
2637 }
2638
2639 static void ext4_sb_release(struct kobject *kobj)
2640 {
2641 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2642 s_kobj);
2643 complete(&sbi->s_kobj_unregister);
2644 }
2645
2646 static const struct sysfs_ops ext4_attr_ops = {
2647 .show = ext4_attr_show,
2648 .store = ext4_attr_store,
2649 };
2650
2651 static struct kobj_type ext4_ktype = {
2652 .default_attrs = ext4_attrs,
2653 .sysfs_ops = &ext4_attr_ops,
2654 .release = ext4_sb_release,
2655 };
2656
2657 static void ext4_feat_release(struct kobject *kobj)
2658 {
2659 complete(&ext4_feat->f_kobj_unregister);
2660 }
2661
2662 static struct kobj_type ext4_feat_ktype = {
2663 .default_attrs = ext4_feat_attrs,
2664 .sysfs_ops = &ext4_attr_ops,
2665 .release = ext4_feat_release,
2666 };
2667
2668 /*
2669 * Check whether this filesystem can be mounted based on
2670 * the features present and the RDONLY/RDWR mount requested.
2671 * Returns 1 if this filesystem can be mounted as requested,
2672 * 0 if it cannot be.
2673 */
2674 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2675 {
2676 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2677 ext4_msg(sb, KERN_ERR,
2678 "Couldn't mount because of "
2679 "unsupported optional features (%x)",
2680 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2681 ~EXT4_FEATURE_INCOMPAT_SUPP));
2682 return 0;
2683 }
2684
2685 if (readonly)
2686 return 1;
2687
2688 /* Check that feature set is OK for a read-write mount */
2689 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2690 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2691 "unsupported optional features (%x)",
2692 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2693 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2694 return 0;
2695 }
2696 /*
2697 * Large file size enabled file system can only be mounted
2698 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2699 */
2700 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2701 if (sizeof(blkcnt_t) < sizeof(u64)) {
2702 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2703 "cannot be mounted RDWR without "
2704 "CONFIG_LBDAF");
2705 return 0;
2706 }
2707 }
2708 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2709 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2710 ext4_msg(sb, KERN_ERR,
2711 "Can't support bigalloc feature without "
2712 "extents feature\n");
2713 return 0;
2714 }
2715 return 1;
2716 }
2717
2718 /*
2719 * This function is called once a day if we have errors logged
2720 * on the file system
2721 */
2722 static void print_daily_error_info(unsigned long arg)
2723 {
2724 struct super_block *sb = (struct super_block *) arg;
2725 struct ext4_sb_info *sbi;
2726 struct ext4_super_block *es;
2727
2728 sbi = EXT4_SB(sb);
2729 es = sbi->s_es;
2730
2731 if (es->s_error_count)
2732 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2733 le32_to_cpu(es->s_error_count));
2734 if (es->s_first_error_time) {
2735 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2736 sb->s_id, le32_to_cpu(es->s_first_error_time),
2737 (int) sizeof(es->s_first_error_func),
2738 es->s_first_error_func,
2739 le32_to_cpu(es->s_first_error_line));
2740 if (es->s_first_error_ino)
2741 printk(": inode %u",
2742 le32_to_cpu(es->s_first_error_ino));
2743 if (es->s_first_error_block)
2744 printk(": block %llu", (unsigned long long)
2745 le64_to_cpu(es->s_first_error_block));
2746 printk("\n");
2747 }
2748 if (es->s_last_error_time) {
2749 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2750 sb->s_id, le32_to_cpu(es->s_last_error_time),
2751 (int) sizeof(es->s_last_error_func),
2752 es->s_last_error_func,
2753 le32_to_cpu(es->s_last_error_line));
2754 if (es->s_last_error_ino)
2755 printk(": inode %u",
2756 le32_to_cpu(es->s_last_error_ino));
2757 if (es->s_last_error_block)
2758 printk(": block %llu", (unsigned long long)
2759 le64_to_cpu(es->s_last_error_block));
2760 printk("\n");
2761 }
2762 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2763 }
2764
2765 /* Find next suitable group and run ext4_init_inode_table */
2766 static int ext4_run_li_request(struct ext4_li_request *elr)
2767 {
2768 struct ext4_group_desc *gdp = NULL;
2769 ext4_group_t group, ngroups;
2770 struct super_block *sb;
2771 unsigned long timeout = 0;
2772 int ret = 0;
2773
2774 sb = elr->lr_super;
2775 ngroups = EXT4_SB(sb)->s_groups_count;
2776
2777 for (group = elr->lr_next_group; group < ngroups; group++) {
2778 gdp = ext4_get_group_desc(sb, group, NULL);
2779 if (!gdp) {
2780 ret = 1;
2781 break;
2782 }
2783
2784 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2785 break;
2786 }
2787
2788 if (group == ngroups)
2789 ret = 1;
2790
2791 if (!ret) {
2792 timeout = jiffies;
2793 ret = ext4_init_inode_table(sb, group,
2794 elr->lr_timeout ? 0 : 1);
2795 if (elr->lr_timeout == 0) {
2796 timeout = (jiffies - timeout) *
2797 elr->lr_sbi->s_li_wait_mult;
2798 elr->lr_timeout = timeout;
2799 }
2800 elr->lr_next_sched = jiffies + elr->lr_timeout;
2801 elr->lr_next_group = group + 1;
2802 }
2803
2804 return ret;
2805 }
2806
2807 /*
2808 * Remove lr_request from the list_request and free the
2809 * request structure. Should be called with li_list_mtx held
2810 */
2811 static void ext4_remove_li_request(struct ext4_li_request *elr)
2812 {
2813 struct ext4_sb_info *sbi;
2814
2815 if (!elr)
2816 return;
2817
2818 sbi = elr->lr_sbi;
2819
2820 list_del(&elr->lr_request);
2821 sbi->s_li_request = NULL;
2822 kfree(elr);
2823 }
2824
2825 static void ext4_unregister_li_request(struct super_block *sb)
2826 {
2827 mutex_lock(&ext4_li_mtx);
2828 if (!ext4_li_info) {
2829 mutex_unlock(&ext4_li_mtx);
2830 return;
2831 }
2832
2833 mutex_lock(&ext4_li_info->li_list_mtx);
2834 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2835 mutex_unlock(&ext4_li_info->li_list_mtx);
2836 mutex_unlock(&ext4_li_mtx);
2837 }
2838
2839 static struct task_struct *ext4_lazyinit_task;
2840
2841 /*
2842 * This is the function where ext4lazyinit thread lives. It walks
2843 * through the request list searching for next scheduled filesystem.
2844 * When such a fs is found, run the lazy initialization request
2845 * (ext4_rn_li_request) and keep track of the time spend in this
2846 * function. Based on that time we compute next schedule time of
2847 * the request. When walking through the list is complete, compute
2848 * next waking time and put itself into sleep.
2849 */
2850 static int ext4_lazyinit_thread(void *arg)
2851 {
2852 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2853 struct list_head *pos, *n;
2854 struct ext4_li_request *elr;
2855 unsigned long next_wakeup, cur;
2856
2857 BUG_ON(NULL == eli);
2858
2859 cont_thread:
2860 while (true) {
2861 next_wakeup = MAX_JIFFY_OFFSET;
2862
2863 mutex_lock(&eli->li_list_mtx);
2864 if (list_empty(&eli->li_request_list)) {
2865 mutex_unlock(&eli->li_list_mtx);
2866 goto exit_thread;
2867 }
2868
2869 list_for_each_safe(pos, n, &eli->li_request_list) {
2870 elr = list_entry(pos, struct ext4_li_request,
2871 lr_request);
2872
2873 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2874 if (ext4_run_li_request(elr) != 0) {
2875 /* error, remove the lazy_init job */
2876 ext4_remove_li_request(elr);
2877 continue;
2878 }
2879 }
2880
2881 if (time_before(elr->lr_next_sched, next_wakeup))
2882 next_wakeup = elr->lr_next_sched;
2883 }
2884 mutex_unlock(&eli->li_list_mtx);
2885
2886 if (freezing(current))
2887 refrigerator();
2888
2889 cur = jiffies;
2890 if ((time_after_eq(cur, next_wakeup)) ||
2891 (MAX_JIFFY_OFFSET == next_wakeup)) {
2892 cond_resched();
2893 continue;
2894 }
2895
2896 schedule_timeout_interruptible(next_wakeup - cur);
2897
2898 if (kthread_should_stop()) {
2899 ext4_clear_request_list();
2900 goto exit_thread;
2901 }
2902 }
2903
2904 exit_thread:
2905 /*
2906 * It looks like the request list is empty, but we need
2907 * to check it under the li_list_mtx lock, to prevent any
2908 * additions into it, and of course we should lock ext4_li_mtx
2909 * to atomically free the list and ext4_li_info, because at
2910 * this point another ext4 filesystem could be registering
2911 * new one.
2912 */
2913 mutex_lock(&ext4_li_mtx);
2914 mutex_lock(&eli->li_list_mtx);
2915 if (!list_empty(&eli->li_request_list)) {
2916 mutex_unlock(&eli->li_list_mtx);
2917 mutex_unlock(&ext4_li_mtx);
2918 goto cont_thread;
2919 }
2920 mutex_unlock(&eli->li_list_mtx);
2921 kfree(ext4_li_info);
2922 ext4_li_info = NULL;
2923 mutex_unlock(&ext4_li_mtx);
2924
2925 return 0;
2926 }
2927
2928 static void ext4_clear_request_list(void)
2929 {
2930 struct list_head *pos, *n;
2931 struct ext4_li_request *elr;
2932
2933 mutex_lock(&ext4_li_info->li_list_mtx);
2934 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2935 elr = list_entry(pos, struct ext4_li_request,
2936 lr_request);
2937 ext4_remove_li_request(elr);
2938 }
2939 mutex_unlock(&ext4_li_info->li_list_mtx);
2940 }
2941
2942 static int ext4_run_lazyinit_thread(void)
2943 {
2944 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2945 ext4_li_info, "ext4lazyinit");
2946 if (IS_ERR(ext4_lazyinit_task)) {
2947 int err = PTR_ERR(ext4_lazyinit_task);
2948 ext4_clear_request_list();
2949 kfree(ext4_li_info);
2950 ext4_li_info = NULL;
2951 printk(KERN_CRIT "EXT4: error %d creating inode table "
2952 "initialization thread\n",
2953 err);
2954 return err;
2955 }
2956 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2957 return 0;
2958 }
2959
2960 /*
2961 * Check whether it make sense to run itable init. thread or not.
2962 * If there is at least one uninitialized inode table, return
2963 * corresponding group number, else the loop goes through all
2964 * groups and return total number of groups.
2965 */
2966 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2967 {
2968 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2969 struct ext4_group_desc *gdp = NULL;
2970
2971 for (group = 0; group < ngroups; group++) {
2972 gdp = ext4_get_group_desc(sb, group, NULL);
2973 if (!gdp)
2974 continue;
2975
2976 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2977 break;
2978 }
2979
2980 return group;
2981 }
2982
2983 static int ext4_li_info_new(void)
2984 {
2985 struct ext4_lazy_init *eli = NULL;
2986
2987 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2988 if (!eli)
2989 return -ENOMEM;
2990
2991 INIT_LIST_HEAD(&eli->li_request_list);
2992 mutex_init(&eli->li_list_mtx);
2993
2994 eli->li_state |= EXT4_LAZYINIT_QUIT;
2995
2996 ext4_li_info = eli;
2997
2998 return 0;
2999 }
3000
3001 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3002 ext4_group_t start)
3003 {
3004 struct ext4_sb_info *sbi = EXT4_SB(sb);
3005 struct ext4_li_request *elr;
3006 unsigned long rnd;
3007
3008 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3009 if (!elr)
3010 return NULL;
3011
3012 elr->lr_super = sb;
3013 elr->lr_sbi = sbi;
3014 elr->lr_next_group = start;
3015
3016 /*
3017 * Randomize first schedule time of the request to
3018 * spread the inode table initialization requests
3019 * better.
3020 */
3021 get_random_bytes(&rnd, sizeof(rnd));
3022 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3023 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3024
3025 return elr;
3026 }
3027
3028 static int ext4_register_li_request(struct super_block *sb,
3029 ext4_group_t first_not_zeroed)
3030 {
3031 struct ext4_sb_info *sbi = EXT4_SB(sb);
3032 struct ext4_li_request *elr;
3033 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3034 int ret = 0;
3035
3036 if (sbi->s_li_request != NULL) {
3037 /*
3038 * Reset timeout so it can be computed again, because
3039 * s_li_wait_mult might have changed.
3040 */
3041 sbi->s_li_request->lr_timeout = 0;
3042 return 0;
3043 }
3044
3045 if (first_not_zeroed == ngroups ||
3046 (sb->s_flags & MS_RDONLY) ||
3047 !test_opt(sb, INIT_INODE_TABLE))
3048 return 0;
3049
3050 elr = ext4_li_request_new(sb, first_not_zeroed);
3051 if (!elr)
3052 return -ENOMEM;
3053
3054 mutex_lock(&ext4_li_mtx);
3055
3056 if (NULL == ext4_li_info) {
3057 ret = ext4_li_info_new();
3058 if (ret)
3059 goto out;
3060 }
3061
3062 mutex_lock(&ext4_li_info->li_list_mtx);
3063 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3064 mutex_unlock(&ext4_li_info->li_list_mtx);
3065
3066 sbi->s_li_request = elr;
3067 /*
3068 * set elr to NULL here since it has been inserted to
3069 * the request_list and the removal and free of it is
3070 * handled by ext4_clear_request_list from now on.
3071 */
3072 elr = NULL;
3073
3074 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3075 ret = ext4_run_lazyinit_thread();
3076 if (ret)
3077 goto out;
3078 }
3079 out:
3080 mutex_unlock(&ext4_li_mtx);
3081 if (ret)
3082 kfree(elr);
3083 return ret;
3084 }
3085
3086 /*
3087 * We do not need to lock anything since this is called on
3088 * module unload.
3089 */
3090 static void ext4_destroy_lazyinit_thread(void)
3091 {
3092 /*
3093 * If thread exited earlier
3094 * there's nothing to be done.
3095 */
3096 if (!ext4_li_info || !ext4_lazyinit_task)
3097 return;
3098
3099 kthread_stop(ext4_lazyinit_task);
3100 }
3101
3102 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3103 {
3104 char *orig_data = kstrdup(data, GFP_KERNEL);
3105 struct buffer_head *bh;
3106 struct ext4_super_block *es = NULL;
3107 struct ext4_sb_info *sbi;
3108 ext4_fsblk_t block;
3109 ext4_fsblk_t sb_block = get_sb_block(&data);
3110 ext4_fsblk_t logical_sb_block;
3111 unsigned long offset = 0;
3112 unsigned long journal_devnum = 0;
3113 unsigned long def_mount_opts;
3114 struct inode *root;
3115 char *cp;
3116 const char *descr;
3117 int ret = -ENOMEM;
3118 int blocksize, clustersize;
3119 unsigned int db_count;
3120 unsigned int i;
3121 int needs_recovery, has_huge_files, has_bigalloc;
3122 __u64 blocks_count;
3123 int err;
3124 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3125 ext4_group_t first_not_zeroed;
3126
3127 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3128 if (!sbi)
3129 goto out_free_orig;
3130
3131 sbi->s_blockgroup_lock =
3132 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3133 if (!sbi->s_blockgroup_lock) {
3134 kfree(sbi);
3135 goto out_free_orig;
3136 }
3137 sb->s_fs_info = sbi;
3138 sbi->s_mount_opt = 0;
3139 sbi->s_resuid = EXT4_DEF_RESUID;
3140 sbi->s_resgid = EXT4_DEF_RESGID;
3141 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3142 sbi->s_sb_block = sb_block;
3143 if (sb->s_bdev->bd_part)
3144 sbi->s_sectors_written_start =
3145 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3146
3147 /* Cleanup superblock name */
3148 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3149 *cp = '!';
3150
3151 ret = -EINVAL;
3152 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3153 if (!blocksize) {
3154 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3155 goto out_fail;
3156 }
3157
3158 /*
3159 * The ext4 superblock will not be buffer aligned for other than 1kB
3160 * block sizes. We need to calculate the offset from buffer start.
3161 */
3162 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3163 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3164 offset = do_div(logical_sb_block, blocksize);
3165 } else {
3166 logical_sb_block = sb_block;
3167 }
3168
3169 if (!(bh = sb_bread(sb, logical_sb_block))) {
3170 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3171 goto out_fail;
3172 }
3173 /*
3174 * Note: s_es must be initialized as soon as possible because
3175 * some ext4 macro-instructions depend on its value
3176 */
3177 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3178 sbi->s_es = es;
3179 sb->s_magic = le16_to_cpu(es->s_magic);
3180 if (sb->s_magic != EXT4_SUPER_MAGIC)
3181 goto cantfind_ext4;
3182 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3183
3184 /* Set defaults before we parse the mount options */
3185 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3186 set_opt(sb, INIT_INODE_TABLE);
3187 if (def_mount_opts & EXT4_DEFM_DEBUG)
3188 set_opt(sb, DEBUG);
3189 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3190 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3191 "2.6.38");
3192 set_opt(sb, GRPID);
3193 }
3194 if (def_mount_opts & EXT4_DEFM_UID16)
3195 set_opt(sb, NO_UID32);
3196 /* xattr user namespace & acls are now defaulted on */
3197 #ifdef CONFIG_EXT4_FS_XATTR
3198 set_opt(sb, XATTR_USER);
3199 #endif
3200 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3201 set_opt(sb, POSIX_ACL);
3202 #endif
3203 set_opt(sb, MBLK_IO_SUBMIT);
3204 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3205 set_opt(sb, JOURNAL_DATA);
3206 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3207 set_opt(sb, ORDERED_DATA);
3208 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3209 set_opt(sb, WRITEBACK_DATA);
3210
3211 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3212 set_opt(sb, ERRORS_PANIC);
3213 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3214 set_opt(sb, ERRORS_CONT);
3215 else
3216 set_opt(sb, ERRORS_RO);
3217 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3218 set_opt(sb, BLOCK_VALIDITY);
3219 if (def_mount_opts & EXT4_DEFM_DISCARD)
3220 set_opt(sb, DISCARD);
3221
3222 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3223 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3224 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3225 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3226 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3227
3228 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3229 set_opt(sb, BARRIER);
3230
3231 /*
3232 * enable delayed allocation by default
3233 * Use -o nodelalloc to turn it off
3234 */
3235 if (!IS_EXT3_SB(sb) &&
3236 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3237 set_opt(sb, DELALLOC);
3238
3239 /*
3240 * set default s_li_wait_mult for lazyinit, for the case there is
3241 * no mount option specified.
3242 */
3243 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3244
3245 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3246 &journal_devnum, &journal_ioprio, NULL, 0)) {
3247 ext4_msg(sb, KERN_WARNING,
3248 "failed to parse options in superblock: %s",
3249 sbi->s_es->s_mount_opts);
3250 }
3251 if (!parse_options((char *) data, sb, &journal_devnum,
3252 &journal_ioprio, NULL, 0))
3253 goto failed_mount;
3254
3255 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3256 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3257 "with data=journal disables delayed "
3258 "allocation and O_DIRECT support!\n");
3259 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3260 ext4_msg(sb, KERN_ERR, "can't mount with "
3261 "both data=journal and delalloc");
3262 goto failed_mount;
3263 }
3264 if (test_opt(sb, DIOREAD_NOLOCK)) {
3265 ext4_msg(sb, KERN_ERR, "can't mount with "
3266 "both data=journal and delalloc");
3267 goto failed_mount;
3268 }
3269 if (test_opt(sb, DELALLOC))
3270 clear_opt(sb, DELALLOC);
3271 }
3272
3273 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3274 if (test_opt(sb, DIOREAD_NOLOCK)) {
3275 if (blocksize < PAGE_SIZE) {
3276 ext4_msg(sb, KERN_ERR, "can't mount with "
3277 "dioread_nolock if block size != PAGE_SIZE");
3278 goto failed_mount;
3279 }
3280 }
3281
3282 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3283 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3284
3285 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3286 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3287 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3288 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3289 ext4_msg(sb, KERN_WARNING,
3290 "feature flags set on rev 0 fs, "
3291 "running e2fsck is recommended");
3292
3293 if (IS_EXT2_SB(sb)) {
3294 if (ext2_feature_set_ok(sb))
3295 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3296 "using the ext4 subsystem");
3297 else {
3298 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3299 "to feature incompatibilities");
3300 goto failed_mount;
3301 }
3302 }
3303
3304 if (IS_EXT3_SB(sb)) {
3305 if (ext3_feature_set_ok(sb))
3306 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3307 "using the ext4 subsystem");
3308 else {
3309 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3310 "to feature incompatibilities");
3311 goto failed_mount;
3312 }
3313 }
3314
3315 /*
3316 * Check feature flags regardless of the revision level, since we
3317 * previously didn't change the revision level when setting the flags,
3318 * so there is a chance incompat flags are set on a rev 0 filesystem.
3319 */
3320 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3321 goto failed_mount;
3322
3323 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3324 blocksize > EXT4_MAX_BLOCK_SIZE) {
3325 ext4_msg(sb, KERN_ERR,
3326 "Unsupported filesystem blocksize %d", blocksize);
3327 goto failed_mount;
3328 }
3329
3330 if (sb->s_blocksize != blocksize) {
3331 /* Validate the filesystem blocksize */
3332 if (!sb_set_blocksize(sb, blocksize)) {
3333 ext4_msg(sb, KERN_ERR, "bad block size %d",
3334 blocksize);
3335 goto failed_mount;
3336 }
3337
3338 brelse(bh);
3339 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3340 offset = do_div(logical_sb_block, blocksize);
3341 bh = sb_bread(sb, logical_sb_block);
3342 if (!bh) {
3343 ext4_msg(sb, KERN_ERR,
3344 "Can't read superblock on 2nd try");
3345 goto failed_mount;
3346 }
3347 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3348 sbi->s_es = es;
3349 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3350 ext4_msg(sb, KERN_ERR,
3351 "Magic mismatch, very weird!");
3352 goto failed_mount;
3353 }
3354 }
3355
3356 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3357 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3358 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3359 has_huge_files);
3360 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3361
3362 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3363 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3364 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3365 } else {
3366 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3367 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3368 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3369 (!is_power_of_2(sbi->s_inode_size)) ||
3370 (sbi->s_inode_size > blocksize)) {
3371 ext4_msg(sb, KERN_ERR,
3372 "unsupported inode size: %d",
3373 sbi->s_inode_size);
3374 goto failed_mount;
3375 }
3376 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3377 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3378 }
3379
3380 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3381 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3382 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3383 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3384 !is_power_of_2(sbi->s_desc_size)) {
3385 ext4_msg(sb, KERN_ERR,
3386 "unsupported descriptor size %lu",
3387 sbi->s_desc_size);
3388 goto failed_mount;
3389 }
3390 } else
3391 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3392
3393 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3394 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3395 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3396 goto cantfind_ext4;
3397
3398 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3399 if (sbi->s_inodes_per_block == 0)
3400 goto cantfind_ext4;
3401 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3402 sbi->s_inodes_per_block;
3403 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3404 sbi->s_sbh = bh;
3405 sbi->s_mount_state = le16_to_cpu(es->s_state);
3406 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3407 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3408
3409 for (i = 0; i < 4; i++)
3410 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3411 sbi->s_def_hash_version = es->s_def_hash_version;
3412 i = le32_to_cpu(es->s_flags);
3413 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3414 sbi->s_hash_unsigned = 3;
3415 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3416 #ifdef __CHAR_UNSIGNED__
3417 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3418 sbi->s_hash_unsigned = 3;
3419 #else
3420 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3421 #endif
3422 sb->s_dirt = 1;
3423 }
3424
3425 /* Handle clustersize */
3426 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3427 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3428 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3429 if (has_bigalloc) {
3430 if (clustersize < blocksize) {
3431 ext4_msg(sb, KERN_ERR,
3432 "cluster size (%d) smaller than "
3433 "block size (%d)", clustersize, blocksize);
3434 goto failed_mount;
3435 }
3436 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3437 le32_to_cpu(es->s_log_block_size);
3438 sbi->s_clusters_per_group =
3439 le32_to_cpu(es->s_clusters_per_group);
3440 if (sbi->s_clusters_per_group > blocksize * 8) {
3441 ext4_msg(sb, KERN_ERR,
3442 "#clusters per group too big: %lu",
3443 sbi->s_clusters_per_group);
3444 goto failed_mount;
3445 }
3446 if (sbi->s_blocks_per_group !=
3447 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3448 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3449 "clusters per group (%lu) inconsistent",
3450 sbi->s_blocks_per_group,
3451 sbi->s_clusters_per_group);
3452 goto failed_mount;
3453 }
3454 } else {
3455 if (clustersize != blocksize) {
3456 ext4_warning(sb, "fragment/cluster size (%d) != "
3457 "block size (%d)", clustersize,
3458 blocksize);
3459 clustersize = blocksize;
3460 }
3461 if (sbi->s_blocks_per_group > blocksize * 8) {
3462 ext4_msg(sb, KERN_ERR,
3463 "#blocks per group too big: %lu",
3464 sbi->s_blocks_per_group);
3465 goto failed_mount;
3466 }
3467 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3468 sbi->s_cluster_bits = 0;
3469 }
3470 sbi->s_cluster_ratio = clustersize / blocksize;
3471
3472 if (sbi->s_inodes_per_group > blocksize * 8) {
3473 ext4_msg(sb, KERN_ERR,
3474 "#inodes per group too big: %lu",
3475 sbi->s_inodes_per_group);
3476 goto failed_mount;
3477 }
3478
3479 /*
3480 * Test whether we have more sectors than will fit in sector_t,
3481 * and whether the max offset is addressable by the page cache.
3482 */
3483 err = generic_check_addressable(sb->s_blocksize_bits,
3484 ext4_blocks_count(es));
3485 if (err) {
3486 ext4_msg(sb, KERN_ERR, "filesystem"
3487 " too large to mount safely on this system");
3488 if (sizeof(sector_t) < 8)
3489 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3490 ret = err;
3491 goto failed_mount;
3492 }
3493
3494 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3495 goto cantfind_ext4;
3496
3497 /* check blocks count against device size */
3498 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3499 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3500 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3501 "exceeds size of device (%llu blocks)",
3502 ext4_blocks_count(es), blocks_count);
3503 goto failed_mount;
3504 }
3505
3506 /*
3507 * It makes no sense for the first data block to be beyond the end
3508 * of the filesystem.
3509 */
3510 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3511 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3512 "block %u is beyond end of filesystem (%llu)",
3513 le32_to_cpu(es->s_first_data_block),
3514 ext4_blocks_count(es));
3515 goto failed_mount;
3516 }
3517 blocks_count = (ext4_blocks_count(es) -
3518 le32_to_cpu(es->s_first_data_block) +
3519 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3520 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3521 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3522 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3523 "(block count %llu, first data block %u, "
3524 "blocks per group %lu)", sbi->s_groups_count,
3525 ext4_blocks_count(es),
3526 le32_to_cpu(es->s_first_data_block),
3527 EXT4_BLOCKS_PER_GROUP(sb));
3528 goto failed_mount;
3529 }
3530 sbi->s_groups_count = blocks_count;
3531 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3532 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3533 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3534 EXT4_DESC_PER_BLOCK(sb);
3535 sbi->s_group_desc = ext4_kvmalloc(db_count *
3536 sizeof(struct buffer_head *),
3537 GFP_KERNEL);
3538 if (sbi->s_group_desc == NULL) {
3539 ext4_msg(sb, KERN_ERR, "not enough memory");
3540 goto failed_mount;
3541 }
3542
3543 if (ext4_proc_root)
3544 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3545
3546 bgl_lock_init(sbi->s_blockgroup_lock);
3547
3548 for (i = 0; i < db_count; i++) {
3549 block = descriptor_loc(sb, logical_sb_block, i);
3550 sbi->s_group_desc[i] = sb_bread(sb, block);
3551 if (!sbi->s_group_desc[i]) {
3552 ext4_msg(sb, KERN_ERR,
3553 "can't read group descriptor %d", i);
3554 db_count = i;
3555 goto failed_mount2;
3556 }
3557 }
3558 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3559 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3560 goto failed_mount2;
3561 }
3562 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3563 if (!ext4_fill_flex_info(sb)) {
3564 ext4_msg(sb, KERN_ERR,
3565 "unable to initialize "
3566 "flex_bg meta info!");
3567 goto failed_mount2;
3568 }
3569
3570 sbi->s_gdb_count = db_count;
3571 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3572 spin_lock_init(&sbi->s_next_gen_lock);
3573
3574 init_timer(&sbi->s_err_report);
3575 sbi->s_err_report.function = print_daily_error_info;
3576 sbi->s_err_report.data = (unsigned long) sb;
3577
3578 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3579 ext4_count_free_clusters(sb));
3580 if (!err) {
3581 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3582 ext4_count_free_inodes(sb));
3583 }
3584 if (!err) {
3585 err = percpu_counter_init(&sbi->s_dirs_counter,
3586 ext4_count_dirs(sb));
3587 }
3588 if (!err) {
3589 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3590 }
3591 if (err) {
3592 ext4_msg(sb, KERN_ERR, "insufficient memory");
3593 goto failed_mount3;
3594 }
3595
3596 sbi->s_stripe = ext4_get_stripe_size(sbi);
3597 sbi->s_max_writeback_mb_bump = 128;
3598
3599 /*
3600 * set up enough so that it can read an inode
3601 */
3602 if (!test_opt(sb, NOLOAD) &&
3603 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3604 sb->s_op = &ext4_sops;
3605 else
3606 sb->s_op = &ext4_nojournal_sops;
3607 sb->s_export_op = &ext4_export_ops;
3608 sb->s_xattr = ext4_xattr_handlers;
3609 #ifdef CONFIG_QUOTA
3610 sb->s_qcop = &ext4_qctl_operations;
3611 sb->dq_op = &ext4_quota_operations;
3612 #endif
3613 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3614
3615 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3616 mutex_init(&sbi->s_orphan_lock);
3617 sbi->s_resize_flags = 0;
3618
3619 sb->s_root = NULL;
3620
3621 needs_recovery = (es->s_last_orphan != 0 ||
3622 EXT4_HAS_INCOMPAT_FEATURE(sb,
3623 EXT4_FEATURE_INCOMPAT_RECOVER));
3624
3625 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3626 !(sb->s_flags & MS_RDONLY))
3627 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3628 goto failed_mount3;
3629
3630 /*
3631 * The first inode we look at is the journal inode. Don't try
3632 * root first: it may be modified in the journal!
3633 */
3634 if (!test_opt(sb, NOLOAD) &&
3635 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3636 if (ext4_load_journal(sb, es, journal_devnum))
3637 goto failed_mount3;
3638 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3639 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3640 ext4_msg(sb, KERN_ERR, "required journal recovery "
3641 "suppressed and not mounted read-only");
3642 goto failed_mount_wq;
3643 } else {
3644 clear_opt(sb, DATA_FLAGS);
3645 sbi->s_journal = NULL;
3646 needs_recovery = 0;
3647 goto no_journal;
3648 }
3649
3650 if (ext4_blocks_count(es) > 0xffffffffULL &&
3651 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3652 JBD2_FEATURE_INCOMPAT_64BIT)) {
3653 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3654 goto failed_mount_wq;
3655 }
3656
3657 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3658 jbd2_journal_set_features(sbi->s_journal,
3659 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3660 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3661 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3662 jbd2_journal_set_features(sbi->s_journal,
3663 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3664 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3665 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3666 } else {
3667 jbd2_journal_clear_features(sbi->s_journal,
3668 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3669 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3670 }
3671
3672 /* We have now updated the journal if required, so we can
3673 * validate the data journaling mode. */
3674 switch (test_opt(sb, DATA_FLAGS)) {
3675 case 0:
3676 /* No mode set, assume a default based on the journal
3677 * capabilities: ORDERED_DATA if the journal can
3678 * cope, else JOURNAL_DATA
3679 */
3680 if (jbd2_journal_check_available_features
3681 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3682 set_opt(sb, ORDERED_DATA);
3683 else
3684 set_opt(sb, JOURNAL_DATA);
3685 break;
3686
3687 case EXT4_MOUNT_ORDERED_DATA:
3688 case EXT4_MOUNT_WRITEBACK_DATA:
3689 if (!jbd2_journal_check_available_features
3690 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3691 ext4_msg(sb, KERN_ERR, "Journal does not support "
3692 "requested data journaling mode");
3693 goto failed_mount_wq;
3694 }
3695 default:
3696 break;
3697 }
3698 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3699
3700 /*
3701 * The journal may have updated the bg summary counts, so we
3702 * need to update the global counters.
3703 */
3704 percpu_counter_set(&sbi->s_freeclusters_counter,
3705 ext4_count_free_clusters(sb));
3706 percpu_counter_set(&sbi->s_freeinodes_counter,
3707 ext4_count_free_inodes(sb));
3708 percpu_counter_set(&sbi->s_dirs_counter,
3709 ext4_count_dirs(sb));
3710 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3711
3712 no_journal:
3713 /*
3714 * The maximum number of concurrent works can be high and
3715 * concurrency isn't really necessary. Limit it to 1.
3716 */
3717 EXT4_SB(sb)->dio_unwritten_wq =
3718 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3719 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3720 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3721 goto failed_mount_wq;
3722 }
3723
3724 /*
3725 * The jbd2_journal_load will have done any necessary log recovery,
3726 * so we can safely mount the rest of the filesystem now.
3727 */
3728
3729 root = ext4_iget(sb, EXT4_ROOT_INO);
3730 if (IS_ERR(root)) {
3731 ext4_msg(sb, KERN_ERR, "get root inode failed");
3732 ret = PTR_ERR(root);
3733 root = NULL;
3734 goto failed_mount4;
3735 }
3736 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3737 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3738 goto failed_mount4;
3739 }
3740 sb->s_root = d_alloc_root(root);
3741 if (!sb->s_root) {
3742 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3743 ret = -ENOMEM;
3744 goto failed_mount4;
3745 }
3746
3747 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3748
3749 /* determine the minimum size of new large inodes, if present */
3750 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3751 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3752 EXT4_GOOD_OLD_INODE_SIZE;
3753 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3754 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3755 if (sbi->s_want_extra_isize <
3756 le16_to_cpu(es->s_want_extra_isize))
3757 sbi->s_want_extra_isize =
3758 le16_to_cpu(es->s_want_extra_isize);
3759 if (sbi->s_want_extra_isize <
3760 le16_to_cpu(es->s_min_extra_isize))
3761 sbi->s_want_extra_isize =
3762 le16_to_cpu(es->s_min_extra_isize);
3763 }
3764 }
3765 /* Check if enough inode space is available */
3766 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3767 sbi->s_inode_size) {
3768 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3769 EXT4_GOOD_OLD_INODE_SIZE;
3770 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3771 "available");
3772 }
3773
3774 err = ext4_setup_system_zone(sb);
3775 if (err) {
3776 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3777 "zone (%d)", err);
3778 goto failed_mount4;
3779 }
3780
3781 ext4_ext_init(sb);
3782 err = ext4_mb_init(sb, needs_recovery);
3783 if (err) {
3784 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3785 err);
3786 goto failed_mount5;
3787 }
3788
3789 err = ext4_register_li_request(sb, first_not_zeroed);
3790 if (err)
3791 goto failed_mount6;
3792
3793 sbi->s_kobj.kset = ext4_kset;
3794 init_completion(&sbi->s_kobj_unregister);
3795 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3796 "%s", sb->s_id);
3797 if (err)
3798 goto failed_mount7;
3799
3800 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3801 ext4_orphan_cleanup(sb, es);
3802 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3803 if (needs_recovery) {
3804 ext4_msg(sb, KERN_INFO, "recovery complete");
3805 ext4_mark_recovery_complete(sb, es);
3806 }
3807 if (EXT4_SB(sb)->s_journal) {
3808 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3809 descr = " journalled data mode";
3810 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3811 descr = " ordered data mode";
3812 else
3813 descr = " writeback data mode";
3814 } else
3815 descr = "out journal";
3816
3817 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3818 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3819 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3820
3821 if (es->s_error_count)
3822 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3823
3824 kfree(orig_data);
3825 return 0;
3826
3827 cantfind_ext4:
3828 if (!silent)
3829 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3830 goto failed_mount;
3831
3832 failed_mount7:
3833 ext4_unregister_li_request(sb);
3834 failed_mount6:
3835 ext4_ext_release(sb);
3836 failed_mount5:
3837 ext4_mb_release(sb);
3838 ext4_release_system_zone(sb);
3839 failed_mount4:
3840 iput(root);
3841 sb->s_root = NULL;
3842 ext4_msg(sb, KERN_ERR, "mount failed");
3843 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3844 failed_mount_wq:
3845 if (sbi->s_journal) {
3846 jbd2_journal_destroy(sbi->s_journal);
3847 sbi->s_journal = NULL;
3848 }
3849 failed_mount3:
3850 del_timer(&sbi->s_err_report);
3851 if (sbi->s_flex_groups)
3852 ext4_kvfree(sbi->s_flex_groups);
3853 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3854 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3855 percpu_counter_destroy(&sbi->s_dirs_counter);
3856 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3857 if (sbi->s_mmp_tsk)
3858 kthread_stop(sbi->s_mmp_tsk);
3859 failed_mount2:
3860 for (i = 0; i < db_count; i++)
3861 brelse(sbi->s_group_desc[i]);
3862 ext4_kvfree(sbi->s_group_desc);
3863 failed_mount:
3864 if (sbi->s_proc) {
3865 remove_proc_entry(sb->s_id, ext4_proc_root);
3866 }
3867 #ifdef CONFIG_QUOTA
3868 for (i = 0; i < MAXQUOTAS; i++)
3869 kfree(sbi->s_qf_names[i]);
3870 #endif
3871 ext4_blkdev_remove(sbi);
3872 brelse(bh);
3873 out_fail:
3874 sb->s_fs_info = NULL;
3875 kfree(sbi->s_blockgroup_lock);
3876 kfree(sbi);
3877 out_free_orig:
3878 kfree(orig_data);
3879 return ret;
3880 }
3881
3882 /*
3883 * Setup any per-fs journal parameters now. We'll do this both on
3884 * initial mount, once the journal has been initialised but before we've
3885 * done any recovery; and again on any subsequent remount.
3886 */
3887 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3888 {
3889 struct ext4_sb_info *sbi = EXT4_SB(sb);
3890
3891 journal->j_commit_interval = sbi->s_commit_interval;
3892 journal->j_min_batch_time = sbi->s_min_batch_time;
3893 journal->j_max_batch_time = sbi->s_max_batch_time;
3894
3895 write_lock(&journal->j_state_lock);
3896 if (test_opt(sb, BARRIER))
3897 journal->j_flags |= JBD2_BARRIER;
3898 else
3899 journal->j_flags &= ~JBD2_BARRIER;
3900 if (test_opt(sb, DATA_ERR_ABORT))
3901 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3902 else
3903 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3904 write_unlock(&journal->j_state_lock);
3905 }
3906
3907 static journal_t *ext4_get_journal(struct super_block *sb,
3908 unsigned int journal_inum)
3909 {
3910 struct inode *journal_inode;
3911 journal_t *journal;
3912
3913 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3914
3915 /* First, test for the existence of a valid inode on disk. Bad
3916 * things happen if we iget() an unused inode, as the subsequent
3917 * iput() will try to delete it. */
3918
3919 journal_inode = ext4_iget(sb, journal_inum);
3920 if (IS_ERR(journal_inode)) {
3921 ext4_msg(sb, KERN_ERR, "no journal found");
3922 return NULL;
3923 }
3924 if (!journal_inode->i_nlink) {
3925 make_bad_inode(journal_inode);
3926 iput(journal_inode);
3927 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3928 return NULL;
3929 }
3930
3931 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3932 journal_inode, journal_inode->i_size);
3933 if (!S_ISREG(journal_inode->i_mode)) {
3934 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3935 iput(journal_inode);
3936 return NULL;
3937 }
3938
3939 journal = jbd2_journal_init_inode(journal_inode);
3940 if (!journal) {
3941 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3942 iput(journal_inode);
3943 return NULL;
3944 }
3945 journal->j_private = sb;
3946 ext4_init_journal_params(sb, journal);
3947 return journal;
3948 }
3949
3950 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3951 dev_t j_dev)
3952 {
3953 struct buffer_head *bh;
3954 journal_t *journal;
3955 ext4_fsblk_t start;
3956 ext4_fsblk_t len;
3957 int hblock, blocksize;
3958 ext4_fsblk_t sb_block;
3959 unsigned long offset;
3960 struct ext4_super_block *es;
3961 struct block_device *bdev;
3962
3963 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3964
3965 bdev = ext4_blkdev_get(j_dev, sb);
3966 if (bdev == NULL)
3967 return NULL;
3968
3969 blocksize = sb->s_blocksize;
3970 hblock = bdev_logical_block_size(bdev);
3971 if (blocksize < hblock) {
3972 ext4_msg(sb, KERN_ERR,
3973 "blocksize too small for journal device");
3974 goto out_bdev;
3975 }
3976
3977 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3978 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3979 set_blocksize(bdev, blocksize);
3980 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3981 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3982 "external journal");
3983 goto out_bdev;
3984 }
3985
3986 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3987 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3988 !(le32_to_cpu(es->s_feature_incompat) &
3989 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3990 ext4_msg(sb, KERN_ERR, "external journal has "
3991 "bad superblock");
3992 brelse(bh);
3993 goto out_bdev;
3994 }
3995
3996 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3997 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3998 brelse(bh);
3999 goto out_bdev;
4000 }
4001
4002 len = ext4_blocks_count(es);
4003 start = sb_block + 1;
4004 brelse(bh); /* we're done with the superblock */
4005
4006 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4007 start, len, blocksize);
4008 if (!journal) {
4009 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4010 goto out_bdev;
4011 }
4012 journal->j_private = sb;
4013 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4014 wait_on_buffer(journal->j_sb_buffer);
4015 if (!buffer_uptodate(journal->j_sb_buffer)) {
4016 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4017 goto out_journal;
4018 }
4019 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4020 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4021 "user (unsupported) - %d",
4022 be32_to_cpu(journal->j_superblock->s_nr_users));
4023 goto out_journal;
4024 }
4025 EXT4_SB(sb)->journal_bdev = bdev;
4026 ext4_init_journal_params(sb, journal);
4027 return journal;
4028
4029 out_journal:
4030 jbd2_journal_destroy(journal);
4031 out_bdev:
4032 ext4_blkdev_put(bdev);
4033 return NULL;
4034 }
4035
4036 static int ext4_load_journal(struct super_block *sb,
4037 struct ext4_super_block *es,
4038 unsigned long journal_devnum)
4039 {
4040 journal_t *journal;
4041 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4042 dev_t journal_dev;
4043 int err = 0;
4044 int really_read_only;
4045
4046 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4047
4048 if (journal_devnum &&
4049 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4050 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4051 "numbers have changed");
4052 journal_dev = new_decode_dev(journal_devnum);
4053 } else
4054 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4055
4056 really_read_only = bdev_read_only(sb->s_bdev);
4057
4058 /*
4059 * Are we loading a blank journal or performing recovery after a
4060 * crash? For recovery, we need to check in advance whether we
4061 * can get read-write access to the device.
4062 */
4063 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4064 if (sb->s_flags & MS_RDONLY) {
4065 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4066 "required on readonly filesystem");
4067 if (really_read_only) {
4068 ext4_msg(sb, KERN_ERR, "write access "
4069 "unavailable, cannot proceed");
4070 return -EROFS;
4071 }
4072 ext4_msg(sb, KERN_INFO, "write access will "
4073 "be enabled during recovery");
4074 }
4075 }
4076
4077 if (journal_inum && journal_dev) {
4078 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4079 "and inode journals!");
4080 return -EINVAL;
4081 }
4082
4083 if (journal_inum) {
4084 if (!(journal = ext4_get_journal(sb, journal_inum)))
4085 return -EINVAL;
4086 } else {
4087 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4088 return -EINVAL;
4089 }
4090
4091 if (!(journal->j_flags & JBD2_BARRIER))
4092 ext4_msg(sb, KERN_INFO, "barriers disabled");
4093
4094 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
4095 err = jbd2_journal_update_format(journal);
4096 if (err) {
4097 ext4_msg(sb, KERN_ERR, "error updating journal");
4098 jbd2_journal_destroy(journal);
4099 return err;
4100 }
4101 }
4102
4103 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4104 err = jbd2_journal_wipe(journal, !really_read_only);
4105 if (!err) {
4106 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4107 if (save)
4108 memcpy(save, ((char *) es) +
4109 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4110 err = jbd2_journal_load(journal);
4111 if (save)
4112 memcpy(((char *) es) + EXT4_S_ERR_START,
4113 save, EXT4_S_ERR_LEN);
4114 kfree(save);
4115 }
4116
4117 if (err) {
4118 ext4_msg(sb, KERN_ERR, "error loading journal");
4119 jbd2_journal_destroy(journal);
4120 return err;
4121 }
4122
4123 EXT4_SB(sb)->s_journal = journal;
4124 ext4_clear_journal_err(sb, es);
4125
4126 if (!really_read_only && journal_devnum &&
4127 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4128 es->s_journal_dev = cpu_to_le32(journal_devnum);
4129
4130 /* Make sure we flush the recovery flag to disk. */
4131 ext4_commit_super(sb, 1);
4132 }
4133
4134 return 0;
4135 }
4136
4137 static int ext4_commit_super(struct super_block *sb, int sync)
4138 {
4139 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4140 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4141 int error = 0;
4142
4143 if (!sbh || block_device_ejected(sb))
4144 return error;
4145 if (buffer_write_io_error(sbh)) {
4146 /*
4147 * Oh, dear. A previous attempt to write the
4148 * superblock failed. This could happen because the
4149 * USB device was yanked out. Or it could happen to
4150 * be a transient write error and maybe the block will
4151 * be remapped. Nothing we can do but to retry the
4152 * write and hope for the best.
4153 */
4154 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4155 "superblock detected");
4156 clear_buffer_write_io_error(sbh);
4157 set_buffer_uptodate(sbh);
4158 }
4159 /*
4160 * If the file system is mounted read-only, don't update the
4161 * superblock write time. This avoids updating the superblock
4162 * write time when we are mounting the root file system
4163 * read/only but we need to replay the journal; at that point,
4164 * for people who are east of GMT and who make their clock
4165 * tick in localtime for Windows bug-for-bug compatibility,
4166 * the clock is set in the future, and this will cause e2fsck
4167 * to complain and force a full file system check.
4168 */
4169 if (!(sb->s_flags & MS_RDONLY))
4170 es->s_wtime = cpu_to_le32(get_seconds());
4171 if (sb->s_bdev->bd_part)
4172 es->s_kbytes_written =
4173 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4174 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4175 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4176 else
4177 es->s_kbytes_written =
4178 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4179 ext4_free_blocks_count_set(es,
4180 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4181 &EXT4_SB(sb)->s_freeclusters_counter)));
4182 es->s_free_inodes_count =
4183 cpu_to_le32(percpu_counter_sum_positive(
4184 &EXT4_SB(sb)->s_freeinodes_counter));
4185 sb->s_dirt = 0;
4186 BUFFER_TRACE(sbh, "marking dirty");
4187 mark_buffer_dirty(sbh);
4188 if (sync) {
4189 error = sync_dirty_buffer(sbh);
4190 if (error)
4191 return error;
4192
4193 error = buffer_write_io_error(sbh);
4194 if (error) {
4195 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4196 "superblock");
4197 clear_buffer_write_io_error(sbh);
4198 set_buffer_uptodate(sbh);
4199 }
4200 }
4201 return error;
4202 }
4203
4204 /*
4205 * Have we just finished recovery? If so, and if we are mounting (or
4206 * remounting) the filesystem readonly, then we will end up with a
4207 * consistent fs on disk. Record that fact.
4208 */
4209 static void ext4_mark_recovery_complete(struct super_block *sb,
4210 struct ext4_super_block *es)
4211 {
4212 journal_t *journal = EXT4_SB(sb)->s_journal;
4213
4214 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4215 BUG_ON(journal != NULL);
4216 return;
4217 }
4218 jbd2_journal_lock_updates(journal);
4219 if (jbd2_journal_flush(journal) < 0)
4220 goto out;
4221
4222 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4223 sb->s_flags & MS_RDONLY) {
4224 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4225 ext4_commit_super(sb, 1);
4226 }
4227
4228 out:
4229 jbd2_journal_unlock_updates(journal);
4230 }
4231
4232 /*
4233 * If we are mounting (or read-write remounting) a filesystem whose journal
4234 * has recorded an error from a previous lifetime, move that error to the
4235 * main filesystem now.
4236 */
4237 static void ext4_clear_journal_err(struct super_block *sb,
4238 struct ext4_super_block *es)
4239 {
4240 journal_t *journal;
4241 int j_errno;
4242 const char *errstr;
4243
4244 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4245
4246 journal = EXT4_SB(sb)->s_journal;
4247
4248 /*
4249 * Now check for any error status which may have been recorded in the
4250 * journal by a prior ext4_error() or ext4_abort()
4251 */
4252
4253 j_errno = jbd2_journal_errno(journal);
4254 if (j_errno) {
4255 char nbuf[16];
4256
4257 errstr = ext4_decode_error(sb, j_errno, nbuf);
4258 ext4_warning(sb, "Filesystem error recorded "
4259 "from previous mount: %s", errstr);
4260 ext4_warning(sb, "Marking fs in need of filesystem check.");
4261
4262 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4263 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4264 ext4_commit_super(sb, 1);
4265
4266 jbd2_journal_clear_err(journal);
4267 }
4268 }
4269
4270 /*
4271 * Force the running and committing transactions to commit,
4272 * and wait on the commit.
4273 */
4274 int ext4_force_commit(struct super_block *sb)
4275 {
4276 journal_t *journal;
4277 int ret = 0;
4278
4279 if (sb->s_flags & MS_RDONLY)
4280 return 0;
4281
4282 journal = EXT4_SB(sb)->s_journal;
4283 if (journal) {
4284 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4285 ret = ext4_journal_force_commit(journal);
4286 }
4287
4288 return ret;
4289 }
4290
4291 static void ext4_write_super(struct super_block *sb)
4292 {
4293 lock_super(sb);
4294 ext4_commit_super(sb, 1);
4295 unlock_super(sb);
4296 }
4297
4298 static int ext4_sync_fs(struct super_block *sb, int wait)
4299 {
4300 int ret = 0;
4301 tid_t target;
4302 struct ext4_sb_info *sbi = EXT4_SB(sb);
4303
4304 trace_ext4_sync_fs(sb, wait);
4305 flush_workqueue(sbi->dio_unwritten_wq);
4306 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4307 if (wait)
4308 jbd2_log_wait_commit(sbi->s_journal, target);
4309 }
4310 return ret;
4311 }
4312
4313 /*
4314 * LVM calls this function before a (read-only) snapshot is created. This
4315 * gives us a chance to flush the journal completely and mark the fs clean.
4316 *
4317 * Note that only this function cannot bring a filesystem to be in a clean
4318 * state independently, because ext4 prevents a new handle from being started
4319 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4320 * the upper layer.
4321 */
4322 static int ext4_freeze(struct super_block *sb)
4323 {
4324 int error = 0;
4325 journal_t *journal;
4326
4327 if (sb->s_flags & MS_RDONLY)
4328 return 0;
4329
4330 journal = EXT4_SB(sb)->s_journal;
4331
4332 /* Now we set up the journal barrier. */
4333 jbd2_journal_lock_updates(journal);
4334
4335 /*
4336 * Don't clear the needs_recovery flag if we failed to flush
4337 * the journal.
4338 */
4339 error = jbd2_journal_flush(journal);
4340 if (error < 0)
4341 goto out;
4342
4343 /* Journal blocked and flushed, clear needs_recovery flag. */
4344 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4345 error = ext4_commit_super(sb, 1);
4346 out:
4347 /* we rely on s_frozen to stop further updates */
4348 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4349 return error;
4350 }
4351
4352 /*
4353 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4354 * flag here, even though the filesystem is not technically dirty yet.
4355 */
4356 static int ext4_unfreeze(struct super_block *sb)
4357 {
4358 if (sb->s_flags & MS_RDONLY)
4359 return 0;
4360
4361 lock_super(sb);
4362 /* Reset the needs_recovery flag before the fs is unlocked. */
4363 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4364 ext4_commit_super(sb, 1);
4365 unlock_super(sb);
4366 return 0;
4367 }
4368
4369 /*
4370 * Structure to save mount options for ext4_remount's benefit
4371 */
4372 struct ext4_mount_options {
4373 unsigned long s_mount_opt;
4374 unsigned long s_mount_opt2;
4375 uid_t s_resuid;
4376 gid_t s_resgid;
4377 unsigned long s_commit_interval;
4378 u32 s_min_batch_time, s_max_batch_time;
4379 #ifdef CONFIG_QUOTA
4380 int s_jquota_fmt;
4381 char *s_qf_names[MAXQUOTAS];
4382 #endif
4383 };
4384
4385 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4386 {
4387 struct ext4_super_block *es;
4388 struct ext4_sb_info *sbi = EXT4_SB(sb);
4389 ext4_fsblk_t n_blocks_count = 0;
4390 unsigned long old_sb_flags;
4391 struct ext4_mount_options old_opts;
4392 int enable_quota = 0;
4393 ext4_group_t g;
4394 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4395 int err = 0;
4396 #ifdef CONFIG_QUOTA
4397 int i;
4398 #endif
4399 char *orig_data = kstrdup(data, GFP_KERNEL);
4400
4401 /* Store the original options */
4402 lock_super(sb);
4403 old_sb_flags = sb->s_flags;
4404 old_opts.s_mount_opt = sbi->s_mount_opt;
4405 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4406 old_opts.s_resuid = sbi->s_resuid;
4407 old_opts.s_resgid = sbi->s_resgid;
4408 old_opts.s_commit_interval = sbi->s_commit_interval;
4409 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4410 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4411 #ifdef CONFIG_QUOTA
4412 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4413 for (i = 0; i < MAXQUOTAS; i++)
4414 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4415 #endif
4416 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4417 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4418
4419 /*
4420 * Allow the "check" option to be passed as a remount option.
4421 */
4422 if (!parse_options(data, sb, NULL, &journal_ioprio,
4423 &n_blocks_count, 1)) {
4424 err = -EINVAL;
4425 goto restore_opts;
4426 }
4427
4428 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4429 ext4_abort(sb, "Abort forced by user");
4430
4431 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4432 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4433
4434 es = sbi->s_es;
4435
4436 if (sbi->s_journal) {
4437 ext4_init_journal_params(sb, sbi->s_journal);
4438 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4439 }
4440
4441 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4442 n_blocks_count > ext4_blocks_count(es)) {
4443 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4444 err = -EROFS;
4445 goto restore_opts;
4446 }
4447
4448 if (*flags & MS_RDONLY) {
4449 err = dquot_suspend(sb, -1);
4450 if (err < 0)
4451 goto restore_opts;
4452
4453 /*
4454 * First of all, the unconditional stuff we have to do
4455 * to disable replay of the journal when we next remount
4456 */
4457 sb->s_flags |= MS_RDONLY;
4458
4459 /*
4460 * OK, test if we are remounting a valid rw partition
4461 * readonly, and if so set the rdonly flag and then
4462 * mark the partition as valid again.
4463 */
4464 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4465 (sbi->s_mount_state & EXT4_VALID_FS))
4466 es->s_state = cpu_to_le16(sbi->s_mount_state);
4467
4468 if (sbi->s_journal)
4469 ext4_mark_recovery_complete(sb, es);
4470 } else {
4471 /* Make sure we can mount this feature set readwrite */
4472 if (!ext4_feature_set_ok(sb, 0)) {
4473 err = -EROFS;
4474 goto restore_opts;
4475 }
4476 /*
4477 * Make sure the group descriptor checksums
4478 * are sane. If they aren't, refuse to remount r/w.
4479 */
4480 for (g = 0; g < sbi->s_groups_count; g++) {
4481 struct ext4_group_desc *gdp =
4482 ext4_get_group_desc(sb, g, NULL);
4483
4484 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4485 ext4_msg(sb, KERN_ERR,
4486 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4487 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4488 le16_to_cpu(gdp->bg_checksum));
4489 err = -EINVAL;
4490 goto restore_opts;
4491 }
4492 }
4493
4494 /*
4495 * If we have an unprocessed orphan list hanging
4496 * around from a previously readonly bdev mount,
4497 * require a full umount/remount for now.
4498 */
4499 if (es->s_last_orphan) {
4500 ext4_msg(sb, KERN_WARNING, "Couldn't "
4501 "remount RDWR because of unprocessed "
4502 "orphan inode list. Please "
4503 "umount/remount instead");
4504 err = -EINVAL;
4505 goto restore_opts;
4506 }
4507
4508 /*
4509 * Mounting a RDONLY partition read-write, so reread
4510 * and store the current valid flag. (It may have
4511 * been changed by e2fsck since we originally mounted
4512 * the partition.)
4513 */
4514 if (sbi->s_journal)
4515 ext4_clear_journal_err(sb, es);
4516 sbi->s_mount_state = le16_to_cpu(es->s_state);
4517 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4518 goto restore_opts;
4519 if (!ext4_setup_super(sb, es, 0))
4520 sb->s_flags &= ~MS_RDONLY;
4521 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4522 EXT4_FEATURE_INCOMPAT_MMP))
4523 if (ext4_multi_mount_protect(sb,
4524 le64_to_cpu(es->s_mmp_block))) {
4525 err = -EROFS;
4526 goto restore_opts;
4527 }
4528 enable_quota = 1;
4529 }
4530 }
4531
4532 /*
4533 * Reinitialize lazy itable initialization thread based on
4534 * current settings
4535 */
4536 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4537 ext4_unregister_li_request(sb);
4538 else {
4539 ext4_group_t first_not_zeroed;
4540 first_not_zeroed = ext4_has_uninit_itable(sb);
4541 ext4_register_li_request(sb, first_not_zeroed);
4542 }
4543
4544 ext4_setup_system_zone(sb);
4545 if (sbi->s_journal == NULL)
4546 ext4_commit_super(sb, 1);
4547
4548 #ifdef CONFIG_QUOTA
4549 /* Release old quota file names */
4550 for (i = 0; i < MAXQUOTAS; i++)
4551 if (old_opts.s_qf_names[i] &&
4552 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4553 kfree(old_opts.s_qf_names[i]);
4554 #endif
4555 unlock_super(sb);
4556 if (enable_quota)
4557 dquot_resume(sb, -1);
4558
4559 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4560 kfree(orig_data);
4561 return 0;
4562
4563 restore_opts:
4564 sb->s_flags = old_sb_flags;
4565 sbi->s_mount_opt = old_opts.s_mount_opt;
4566 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4567 sbi->s_resuid = old_opts.s_resuid;
4568 sbi->s_resgid = old_opts.s_resgid;
4569 sbi->s_commit_interval = old_opts.s_commit_interval;
4570 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4571 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4572 #ifdef CONFIG_QUOTA
4573 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4574 for (i = 0; i < MAXQUOTAS; i++) {
4575 if (sbi->s_qf_names[i] &&
4576 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4577 kfree(sbi->s_qf_names[i]);
4578 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4579 }
4580 #endif
4581 unlock_super(sb);
4582 kfree(orig_data);
4583 return err;
4584 }
4585
4586 /*
4587 * Note: calculating the overhead so we can be compatible with
4588 * historical BSD practice is quite difficult in the face of
4589 * clusters/bigalloc. This is because multiple metadata blocks from
4590 * different block group can end up in the same allocation cluster.
4591 * Calculating the exact overhead in the face of clustered allocation
4592 * requires either O(all block bitmaps) in memory or O(number of block
4593 * groups**2) in time. We will still calculate the superblock for
4594 * older file systems --- and if we come across with a bigalloc file
4595 * system with zero in s_overhead_clusters the estimate will be close to
4596 * correct especially for very large cluster sizes --- but for newer
4597 * file systems, it's better to calculate this figure once at mkfs
4598 * time, and store it in the superblock. If the superblock value is
4599 * present (even for non-bigalloc file systems), we will use it.
4600 */
4601 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4602 {
4603 struct super_block *sb = dentry->d_sb;
4604 struct ext4_sb_info *sbi = EXT4_SB(sb);
4605 struct ext4_super_block *es = sbi->s_es;
4606 struct ext4_group_desc *gdp;
4607 u64 fsid;
4608 s64 bfree;
4609
4610 if (test_opt(sb, MINIX_DF)) {
4611 sbi->s_overhead_last = 0;
4612 } else if (es->s_overhead_clusters) {
4613 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4614 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4615 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4616 ext4_fsblk_t overhead = 0;
4617
4618 /*
4619 * Compute the overhead (FS structures). This is constant
4620 * for a given filesystem unless the number of block groups
4621 * changes so we cache the previous value until it does.
4622 */
4623
4624 /*
4625 * All of the blocks before first_data_block are
4626 * overhead
4627 */
4628 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4629
4630 /*
4631 * Add the overhead found in each block group
4632 */
4633 for (i = 0; i < ngroups; i++) {
4634 gdp = ext4_get_group_desc(sb, i, NULL);
4635 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4636 cond_resched();
4637 }
4638 sbi->s_overhead_last = overhead;
4639 smp_wmb();
4640 sbi->s_blocks_last = ext4_blocks_count(es);
4641 }
4642
4643 buf->f_type = EXT4_SUPER_MAGIC;
4644 buf->f_bsize = sb->s_blocksize;
4645 buf->f_blocks = (ext4_blocks_count(es) -
4646 EXT4_C2B(sbi, sbi->s_overhead_last));
4647 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4648 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4649 /* prevent underflow in case that few free space is available */
4650 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4651 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4652 if (buf->f_bfree < ext4_r_blocks_count(es))
4653 buf->f_bavail = 0;
4654 buf->f_files = le32_to_cpu(es->s_inodes_count);
4655 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4656 buf->f_namelen = EXT4_NAME_LEN;
4657 fsid = le64_to_cpup((void *)es->s_uuid) ^
4658 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4659 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4660 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4661
4662 return 0;
4663 }
4664
4665 /* Helper function for writing quotas on sync - we need to start transaction
4666 * before quota file is locked for write. Otherwise the are possible deadlocks:
4667 * Process 1 Process 2
4668 * ext4_create() quota_sync()
4669 * jbd2_journal_start() write_dquot()
4670 * dquot_initialize() down(dqio_mutex)
4671 * down(dqio_mutex) jbd2_journal_start()
4672 *
4673 */
4674
4675 #ifdef CONFIG_QUOTA
4676
4677 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4678 {
4679 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4680 }
4681
4682 static int ext4_write_dquot(struct dquot *dquot)
4683 {
4684 int ret, err;
4685 handle_t *handle;
4686 struct inode *inode;
4687
4688 inode = dquot_to_inode(dquot);
4689 handle = ext4_journal_start(inode,
4690 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4691 if (IS_ERR(handle))
4692 return PTR_ERR(handle);
4693 ret = dquot_commit(dquot);
4694 err = ext4_journal_stop(handle);
4695 if (!ret)
4696 ret = err;
4697 return ret;
4698 }
4699
4700 static int ext4_acquire_dquot(struct dquot *dquot)
4701 {
4702 int ret, err;
4703 handle_t *handle;
4704
4705 handle = ext4_journal_start(dquot_to_inode(dquot),
4706 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4707 if (IS_ERR(handle))
4708 return PTR_ERR(handle);
4709 ret = dquot_acquire(dquot);
4710 err = ext4_journal_stop(handle);
4711 if (!ret)
4712 ret = err;
4713 return ret;
4714 }
4715
4716 static int ext4_release_dquot(struct dquot *dquot)
4717 {
4718 int ret, err;
4719 handle_t *handle;
4720
4721 handle = ext4_journal_start(dquot_to_inode(dquot),
4722 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4723 if (IS_ERR(handle)) {
4724 /* Release dquot anyway to avoid endless cycle in dqput() */
4725 dquot_release(dquot);
4726 return PTR_ERR(handle);
4727 }
4728 ret = dquot_release(dquot);
4729 err = ext4_journal_stop(handle);
4730 if (!ret)
4731 ret = err;
4732 return ret;
4733 }
4734
4735 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4736 {
4737 /* Are we journaling quotas? */
4738 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4739 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4740 dquot_mark_dquot_dirty(dquot);
4741 return ext4_write_dquot(dquot);
4742 } else {
4743 return dquot_mark_dquot_dirty(dquot);
4744 }
4745 }
4746
4747 static int ext4_write_info(struct super_block *sb, int type)
4748 {
4749 int ret, err;
4750 handle_t *handle;
4751
4752 /* Data block + inode block */
4753 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4754 if (IS_ERR(handle))
4755 return PTR_ERR(handle);
4756 ret = dquot_commit_info(sb, type);
4757 err = ext4_journal_stop(handle);
4758 if (!ret)
4759 ret = err;
4760 return ret;
4761 }
4762
4763 /*
4764 * Turn on quotas during mount time - we need to find
4765 * the quota file and such...
4766 */
4767 static int ext4_quota_on_mount(struct super_block *sb, int type)
4768 {
4769 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4770 EXT4_SB(sb)->s_jquota_fmt, type);
4771 }
4772
4773 /*
4774 * Standard function to be called on quota_on
4775 */
4776 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4777 struct path *path)
4778 {
4779 int err;
4780
4781 if (!test_opt(sb, QUOTA))
4782 return -EINVAL;
4783
4784 /* Quotafile not on the same filesystem? */
4785 if (path->mnt->mnt_sb != sb)
4786 return -EXDEV;
4787 /* Journaling quota? */
4788 if (EXT4_SB(sb)->s_qf_names[type]) {
4789 /* Quotafile not in fs root? */
4790 if (path->dentry->d_parent != sb->s_root)
4791 ext4_msg(sb, KERN_WARNING,
4792 "Quota file not on filesystem root. "
4793 "Journaled quota will not work");
4794 }
4795
4796 /*
4797 * When we journal data on quota file, we have to flush journal to see
4798 * all updates to the file when we bypass pagecache...
4799 */
4800 if (EXT4_SB(sb)->s_journal &&
4801 ext4_should_journal_data(path->dentry->d_inode)) {
4802 /*
4803 * We don't need to lock updates but journal_flush() could
4804 * otherwise be livelocked...
4805 */
4806 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4807 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4808 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4809 if (err)
4810 return err;
4811 }
4812
4813 return dquot_quota_on(sb, type, format_id, path);
4814 }
4815
4816 static int ext4_quota_off(struct super_block *sb, int type)
4817 {
4818 struct inode *inode = sb_dqopt(sb)->files[type];
4819 handle_t *handle;
4820
4821 /* Force all delayed allocation blocks to be allocated.
4822 * Caller already holds s_umount sem */
4823 if (test_opt(sb, DELALLOC))
4824 sync_filesystem(sb);
4825
4826 if (!inode)
4827 goto out;
4828
4829 /* Update modification times of quota files when userspace can
4830 * start looking at them */
4831 handle = ext4_journal_start(inode, 1);
4832 if (IS_ERR(handle))
4833 goto out;
4834 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4835 ext4_mark_inode_dirty(handle, inode);
4836 ext4_journal_stop(handle);
4837
4838 out:
4839 return dquot_quota_off(sb, type);
4840 }
4841
4842 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4843 * acquiring the locks... As quota files are never truncated and quota code
4844 * itself serializes the operations (and no one else should touch the files)
4845 * we don't have to be afraid of races */
4846 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4847 size_t len, loff_t off)
4848 {
4849 struct inode *inode = sb_dqopt(sb)->files[type];
4850 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4851 int err = 0;
4852 int offset = off & (sb->s_blocksize - 1);
4853 int tocopy;
4854 size_t toread;
4855 struct buffer_head *bh;
4856 loff_t i_size = i_size_read(inode);
4857
4858 if (off > i_size)
4859 return 0;
4860 if (off+len > i_size)
4861 len = i_size-off;
4862 toread = len;
4863 while (toread > 0) {
4864 tocopy = sb->s_blocksize - offset < toread ?
4865 sb->s_blocksize - offset : toread;
4866 bh = ext4_bread(NULL, inode, blk, 0, &err);
4867 if (err)
4868 return err;
4869 if (!bh) /* A hole? */
4870 memset(data, 0, tocopy);
4871 else
4872 memcpy(data, bh->b_data+offset, tocopy);
4873 brelse(bh);
4874 offset = 0;
4875 toread -= tocopy;
4876 data += tocopy;
4877 blk++;
4878 }
4879 return len;
4880 }
4881
4882 /* Write to quotafile (we know the transaction is already started and has
4883 * enough credits) */
4884 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4885 const char *data, size_t len, loff_t off)
4886 {
4887 struct inode *inode = sb_dqopt(sb)->files[type];
4888 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4889 int err = 0;
4890 int offset = off & (sb->s_blocksize - 1);
4891 struct buffer_head *bh;
4892 handle_t *handle = journal_current_handle();
4893
4894 if (EXT4_SB(sb)->s_journal && !handle) {
4895 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4896 " cancelled because transaction is not started",
4897 (unsigned long long)off, (unsigned long long)len);
4898 return -EIO;
4899 }
4900 /*
4901 * Since we account only one data block in transaction credits,
4902 * then it is impossible to cross a block boundary.
4903 */
4904 if (sb->s_blocksize - offset < len) {
4905 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4906 " cancelled because not block aligned",
4907 (unsigned long long)off, (unsigned long long)len);
4908 return -EIO;
4909 }
4910
4911 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4912 bh = ext4_bread(handle, inode, blk, 1, &err);
4913 if (!bh)
4914 goto out;
4915 err = ext4_journal_get_write_access(handle, bh);
4916 if (err) {
4917 brelse(bh);
4918 goto out;
4919 }
4920 lock_buffer(bh);
4921 memcpy(bh->b_data+offset, data, len);
4922 flush_dcache_page(bh->b_page);
4923 unlock_buffer(bh);
4924 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4925 brelse(bh);
4926 out:
4927 if (err) {
4928 mutex_unlock(&inode->i_mutex);
4929 return err;
4930 }
4931 if (inode->i_size < off + len) {
4932 i_size_write(inode, off + len);
4933 EXT4_I(inode)->i_disksize = inode->i_size;
4934 ext4_mark_inode_dirty(handle, inode);
4935 }
4936 mutex_unlock(&inode->i_mutex);
4937 return len;
4938 }
4939
4940 #endif
4941
4942 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4943 const char *dev_name, void *data)
4944 {
4945 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4946 }
4947
4948 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4949 static inline void register_as_ext2(void)
4950 {
4951 int err = register_filesystem(&ext2_fs_type);
4952 if (err)
4953 printk(KERN_WARNING
4954 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4955 }
4956
4957 static inline void unregister_as_ext2(void)
4958 {
4959 unregister_filesystem(&ext2_fs_type);
4960 }
4961
4962 static inline int ext2_feature_set_ok(struct super_block *sb)
4963 {
4964 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4965 return 0;
4966 if (sb->s_flags & MS_RDONLY)
4967 return 1;
4968 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4969 return 0;
4970 return 1;
4971 }
4972 MODULE_ALIAS("ext2");
4973 #else
4974 static inline void register_as_ext2(void) { }
4975 static inline void unregister_as_ext2(void) { }
4976 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4977 #endif
4978
4979 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4980 static inline void register_as_ext3(void)
4981 {
4982 int err = register_filesystem(&ext3_fs_type);
4983 if (err)
4984 printk(KERN_WARNING
4985 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4986 }
4987
4988 static inline void unregister_as_ext3(void)
4989 {
4990 unregister_filesystem(&ext3_fs_type);
4991 }
4992
4993 static inline int ext3_feature_set_ok(struct super_block *sb)
4994 {
4995 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4996 return 0;
4997 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4998 return 0;
4999 if (sb->s_flags & MS_RDONLY)
5000 return 1;
5001 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5002 return 0;
5003 return 1;
5004 }
5005 MODULE_ALIAS("ext3");
5006 #else
5007 static inline void register_as_ext3(void) { }
5008 static inline void unregister_as_ext3(void) { }
5009 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5010 #endif
5011
5012 static struct file_system_type ext4_fs_type = {
5013 .owner = THIS_MODULE,
5014 .name = "ext4",
5015 .mount = ext4_mount,
5016 .kill_sb = kill_block_super,
5017 .fs_flags = FS_REQUIRES_DEV,
5018 };
5019
5020 static int __init ext4_init_feat_adverts(void)
5021 {
5022 struct ext4_features *ef;
5023 int ret = -ENOMEM;
5024
5025 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5026 if (!ef)
5027 goto out;
5028
5029 ef->f_kobj.kset = ext4_kset;
5030 init_completion(&ef->f_kobj_unregister);
5031 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5032 "features");
5033 if (ret) {
5034 kfree(ef);
5035 goto out;
5036 }
5037
5038 ext4_feat = ef;
5039 ret = 0;
5040 out:
5041 return ret;
5042 }
5043
5044 static void ext4_exit_feat_adverts(void)
5045 {
5046 kobject_put(&ext4_feat->f_kobj);
5047 wait_for_completion(&ext4_feat->f_kobj_unregister);
5048 kfree(ext4_feat);
5049 }
5050
5051 /* Shared across all ext4 file systems */
5052 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5053 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5054
5055 static int __init ext4_init_fs(void)
5056 {
5057 int i, err;
5058
5059 ext4_check_flag_values();
5060
5061 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5062 mutex_init(&ext4__aio_mutex[i]);
5063 init_waitqueue_head(&ext4__ioend_wq[i]);
5064 }
5065
5066 err = ext4_init_pageio();
5067 if (err)
5068 return err;
5069 err = ext4_init_system_zone();
5070 if (err)
5071 goto out6;
5072 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5073 if (!ext4_kset)
5074 goto out5;
5075 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5076
5077 err = ext4_init_feat_adverts();
5078 if (err)
5079 goto out4;
5080
5081 err = ext4_init_mballoc();
5082 if (err)
5083 goto out3;
5084
5085 err = ext4_init_xattr();
5086 if (err)
5087 goto out2;
5088 err = init_inodecache();
5089 if (err)
5090 goto out1;
5091 register_as_ext3();
5092 register_as_ext2();
5093 err = register_filesystem(&ext4_fs_type);
5094 if (err)
5095 goto out;
5096
5097 ext4_li_info = NULL;
5098 mutex_init(&ext4_li_mtx);
5099 return 0;
5100 out:
5101 unregister_as_ext2();
5102 unregister_as_ext3();
5103 destroy_inodecache();
5104 out1:
5105 ext4_exit_xattr();
5106 out2:
5107 ext4_exit_mballoc();
5108 out3:
5109 ext4_exit_feat_adverts();
5110 out4:
5111 if (ext4_proc_root)
5112 remove_proc_entry("fs/ext4", NULL);
5113 kset_unregister(ext4_kset);
5114 out5:
5115 ext4_exit_system_zone();
5116 out6:
5117 ext4_exit_pageio();
5118 return err;
5119 }
5120
5121 static void __exit ext4_exit_fs(void)
5122 {
5123 ext4_destroy_lazyinit_thread();
5124 unregister_as_ext2();
5125 unregister_as_ext3();
5126 unregister_filesystem(&ext4_fs_type);
5127 destroy_inodecache();
5128 ext4_exit_xattr();
5129 ext4_exit_mballoc();
5130 ext4_exit_feat_adverts();
5131 remove_proc_entry("fs/ext4", NULL);
5132 kset_unregister(ext4_kset);
5133 ext4_exit_system_zone();
5134 ext4_exit_pageio();
5135 }
5136
5137 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5138 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5139 MODULE_LICENSE("GPL");
5140 module_init(ext4_init_fs)
5141 module_exit(ext4_exit_fs)