?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
spi-topcliff-pch: supports a spi mode setup and bit order setup by IO control
[zen-stable.git] / fs / ext4 / super.c
blob14dbc78cc4948c9834cb547307ae2b405729aad2
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 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
934 }
935
936 static void ext4_destroy_inode(struct inode *inode)
937 {
938 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
939 ext4_msg(inode->i_sb, KERN_ERR,
940 "Inode %lu (%p): orphan list check failed!",
941 inode->i_ino, EXT4_I(inode));
942 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
943 EXT4_I(inode), sizeof(struct ext4_inode_info),
944 true);
945 dump_stack();
946 }
947 call_rcu(&inode->i_rcu, ext4_i_callback);
948 }
949
950 static void init_once(void *foo)
951 {
952 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
953
954 INIT_LIST_HEAD(&ei->i_orphan);
955 #ifdef CONFIG_EXT4_FS_XATTR
956 init_rwsem(&ei->xattr_sem);
957 #endif
958 init_rwsem(&ei->i_data_sem);
959 inode_init_once(&ei->vfs_inode);
960 }
961
962 static int init_inodecache(void)
963 {
964 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
965 sizeof(struct ext4_inode_info),
966 0, (SLAB_RECLAIM_ACCOUNT|
967 SLAB_MEM_SPREAD),
968 init_once);
969 if (ext4_inode_cachep == NULL)
970 return -ENOMEM;
971 return 0;
972 }
973
974 static void destroy_inodecache(void)
975 {
976 kmem_cache_destroy(ext4_inode_cachep);
977 }
978
979 void ext4_clear_inode(struct inode *inode)
980 {
981 invalidate_inode_buffers(inode);
982 end_writeback(inode);
983 dquot_drop(inode);
984 ext4_discard_preallocations(inode);
985 if (EXT4_I(inode)->jinode) {
986 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
987 EXT4_I(inode)->jinode);
988 jbd2_free_inode(EXT4_I(inode)->jinode);
989 EXT4_I(inode)->jinode = NULL;
990 }
991 }
992
993 static inline void ext4_show_quota_options(struct seq_file *seq,
994 struct super_block *sb)
995 {
996 #if defined(CONFIG_QUOTA)
997 struct ext4_sb_info *sbi = EXT4_SB(sb);
998
999 if (sbi->s_jquota_fmt) {
1000 char *fmtname = "";
1001
1002 switch (sbi->s_jquota_fmt) {
1003 case QFMT_VFS_OLD:
1004 fmtname = "vfsold";
1005 break;
1006 case QFMT_VFS_V0:
1007 fmtname = "vfsv0";
1008 break;
1009 case QFMT_VFS_V1:
1010 fmtname = "vfsv1";
1011 break;
1012 }
1013 seq_printf(seq, ",jqfmt=%s", fmtname);
1014 }
1015
1016 if (sbi->s_qf_names[USRQUOTA])
1017 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1018
1019 if (sbi->s_qf_names[GRPQUOTA])
1020 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1021
1022 if (test_opt(sb, USRQUOTA))
1023 seq_puts(seq, ",usrquota");
1024
1025 if (test_opt(sb, GRPQUOTA))
1026 seq_puts(seq, ",grpquota");
1027 #endif
1028 }
1029
1030 /*
1031 * Show an option if
1032 * - it's set to a non-default value OR
1033 * - if the per-sb default is different from the global default
1034 */
1035 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1036 {
1037 int def_errors;
1038 unsigned long def_mount_opts;
1039 struct super_block *sb = root->d_sb;
1040 struct ext4_sb_info *sbi = EXT4_SB(sb);
1041 struct ext4_super_block *es = sbi->s_es;
1042
1043 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1044 def_errors = le16_to_cpu(es->s_errors);
1045
1046 if (sbi->s_sb_block != 1)
1047 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1048 if (test_opt(sb, MINIX_DF))
1049 seq_puts(seq, ",minixdf");
1050 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1051 seq_puts(seq, ",grpid");
1052 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1053 seq_puts(seq, ",nogrpid");
1054 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1055 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1056 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1057 }
1058 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1059 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1060 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1061 }
1062 if (test_opt(sb, ERRORS_RO)) {
1063 if (def_errors == EXT4_ERRORS_PANIC ||
1064 def_errors == EXT4_ERRORS_CONTINUE) {
1065 seq_puts(seq, ",errors=remount-ro");
1066 }
1067 }
1068 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1069 seq_puts(seq, ",errors=continue");
1070 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1071 seq_puts(seq, ",errors=panic");
1072 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1073 seq_puts(seq, ",nouid32");
1074 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1075 seq_puts(seq, ",debug");
1076 #ifdef CONFIG_EXT4_FS_XATTR
1077 if (test_opt(sb, XATTR_USER))
1078 seq_puts(seq, ",user_xattr");
1079 if (!test_opt(sb, XATTR_USER))
1080 seq_puts(seq, ",nouser_xattr");
1081 #endif
1082 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1083 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1084 seq_puts(seq, ",acl");
1085 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1086 seq_puts(seq, ",noacl");
1087 #endif
1088 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1089 seq_printf(seq, ",commit=%u",
1090 (unsigned) (sbi->s_commit_interval / HZ));
1091 }
1092 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1093 seq_printf(seq, ",min_batch_time=%u",
1094 (unsigned) sbi->s_min_batch_time);
1095 }
1096 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1097 seq_printf(seq, ",max_batch_time=%u",
1098 (unsigned) sbi->s_max_batch_time);
1099 }
1100
1101 /*
1102 * We're changing the default of barrier mount option, so
1103 * let's always display its mount state so it's clear what its
1104 * status is.
1105 */
1106 seq_puts(seq, ",barrier=");
1107 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1108 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1109 seq_puts(seq, ",journal_async_commit");
1110 else if (test_opt(sb, JOURNAL_CHECKSUM))
1111 seq_puts(seq, ",journal_checksum");
1112 if (test_opt(sb, I_VERSION))
1113 seq_puts(seq, ",i_version");
1114 if (!test_opt(sb, DELALLOC) &&
1115 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1116 seq_puts(seq, ",nodelalloc");
1117
1118 if (!test_opt(sb, MBLK_IO_SUBMIT))
1119 seq_puts(seq, ",nomblk_io_submit");
1120 if (sbi->s_stripe)
1121 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1122 /*
1123 * journal mode get enabled in different ways
1124 * So just print the value even if we didn't specify it
1125 */
1126 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1127 seq_puts(seq, ",data=journal");
1128 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1129 seq_puts(seq, ",data=ordered");
1130 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1131 seq_puts(seq, ",data=writeback");
1132
1133 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1134 seq_printf(seq, ",inode_readahead_blks=%u",
1135 sbi->s_inode_readahead_blks);
1136
1137 if (test_opt(sb, DATA_ERR_ABORT))
1138 seq_puts(seq, ",data_err=abort");
1139
1140 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1141 seq_puts(seq, ",noauto_da_alloc");
1142
1143 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1144 seq_puts(seq, ",discard");
1145
1146 if (test_opt(sb, NOLOAD))
1147 seq_puts(seq, ",norecovery");
1148
1149 if (test_opt(sb, DIOREAD_NOLOCK))
1150 seq_puts(seq, ",dioread_nolock");
1151
1152 if (test_opt(sb, BLOCK_VALIDITY) &&
1153 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1154 seq_puts(seq, ",block_validity");
1155
1156 if (!test_opt(sb, INIT_INODE_TABLE))
1157 seq_puts(seq, ",noinit_itable");
1158 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1159 seq_printf(seq, ",init_itable=%u",
1160 (unsigned) sbi->s_li_wait_mult);
1161
1162 ext4_show_quota_options(seq, sb);
1163
1164 return 0;
1165 }
1166
1167 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1168 u64 ino, u32 generation)
1169 {
1170 struct inode *inode;
1171
1172 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1173 return ERR_PTR(-ESTALE);
1174 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1175 return ERR_PTR(-ESTALE);
1176
1177 /* iget isn't really right if the inode is currently unallocated!!
1178 *
1179 * ext4_read_inode will return a bad_inode if the inode had been
1180 * deleted, so we should be safe.
1181 *
1182 * Currently we don't know the generation for parent directory, so
1183 * a generation of 0 means "accept any"
1184 */
1185 inode = ext4_iget(sb, ino);
1186 if (IS_ERR(inode))
1187 return ERR_CAST(inode);
1188 if (generation && inode->i_generation != generation) {
1189 iput(inode);
1190 return ERR_PTR(-ESTALE);
1191 }
1192
1193 return inode;
1194 }
1195
1196 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1197 int fh_len, int fh_type)
1198 {
1199 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1200 ext4_nfs_get_inode);
1201 }
1202
1203 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1204 int fh_len, int fh_type)
1205 {
1206 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1207 ext4_nfs_get_inode);
1208 }
1209
1210 /*
1211 * Try to release metadata pages (indirect blocks, directories) which are
1212 * mapped via the block device. Since these pages could have journal heads
1213 * which would prevent try_to_free_buffers() from freeing them, we must use
1214 * jbd2 layer's try_to_free_buffers() function to release them.
1215 */
1216 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1217 gfp_t wait)
1218 {
1219 journal_t *journal = EXT4_SB(sb)->s_journal;
1220
1221 WARN_ON(PageChecked(page));
1222 if (!page_has_buffers(page))
1223 return 0;
1224 if (journal)
1225 return jbd2_journal_try_to_free_buffers(journal, page,
1226 wait & ~__GFP_WAIT);
1227 return try_to_free_buffers(page);
1228 }
1229
1230 #ifdef CONFIG_QUOTA
1231 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1232 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1233
1234 static int ext4_write_dquot(struct dquot *dquot);
1235 static int ext4_acquire_dquot(struct dquot *dquot);
1236 static int ext4_release_dquot(struct dquot *dquot);
1237 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1238 static int ext4_write_info(struct super_block *sb, int type);
1239 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1240 struct path *path);
1241 static int ext4_quota_off(struct super_block *sb, int type);
1242 static int ext4_quota_on_mount(struct super_block *sb, int type);
1243 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1244 size_t len, loff_t off);
1245 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1246 const char *data, size_t len, loff_t off);
1247
1248 static const struct dquot_operations ext4_quota_operations = {
1249 .get_reserved_space = ext4_get_reserved_space,
1250 .write_dquot = ext4_write_dquot,
1251 .acquire_dquot = ext4_acquire_dquot,
1252 .release_dquot = ext4_release_dquot,
1253 .mark_dirty = ext4_mark_dquot_dirty,
1254 .write_info = ext4_write_info,
1255 .alloc_dquot = dquot_alloc,
1256 .destroy_dquot = dquot_destroy,
1257 };
1258
1259 static const struct quotactl_ops ext4_qctl_operations = {
1260 .quota_on = ext4_quota_on,
1261 .quota_off = ext4_quota_off,
1262 .quota_sync = dquot_quota_sync,
1263 .get_info = dquot_get_dqinfo,
1264 .set_info = dquot_set_dqinfo,
1265 .get_dqblk = dquot_get_dqblk,
1266 .set_dqblk = dquot_set_dqblk
1267 };
1268 #endif
1269
1270 static const struct super_operations ext4_sops = {
1271 .alloc_inode = ext4_alloc_inode,
1272 .destroy_inode = ext4_destroy_inode,
1273 .write_inode = ext4_write_inode,
1274 .dirty_inode = ext4_dirty_inode,
1275 .drop_inode = ext4_drop_inode,
1276 .evict_inode = ext4_evict_inode,
1277 .put_super = ext4_put_super,
1278 .sync_fs = ext4_sync_fs,
1279 .freeze_fs = ext4_freeze,
1280 .unfreeze_fs = ext4_unfreeze,
1281 .statfs = ext4_statfs,
1282 .remount_fs = ext4_remount,
1283 .show_options = ext4_show_options,
1284 #ifdef CONFIG_QUOTA
1285 .quota_read = ext4_quota_read,
1286 .quota_write = ext4_quota_write,
1287 #endif
1288 .bdev_try_to_free_page = bdev_try_to_free_page,
1289 };
1290
1291 static const struct super_operations ext4_nojournal_sops = {
1292 .alloc_inode = ext4_alloc_inode,
1293 .destroy_inode = ext4_destroy_inode,
1294 .write_inode = ext4_write_inode,
1295 .dirty_inode = ext4_dirty_inode,
1296 .drop_inode = ext4_drop_inode,
1297 .evict_inode = ext4_evict_inode,
1298 .write_super = ext4_write_super,
1299 .put_super = ext4_put_super,
1300 .statfs = ext4_statfs,
1301 .remount_fs = ext4_remount,
1302 .show_options = ext4_show_options,
1303 #ifdef CONFIG_QUOTA
1304 .quota_read = ext4_quota_read,
1305 .quota_write = ext4_quota_write,
1306 #endif
1307 .bdev_try_to_free_page = bdev_try_to_free_page,
1308 };
1309
1310 static const struct export_operations ext4_export_ops = {
1311 .fh_to_dentry = ext4_fh_to_dentry,
1312 .fh_to_parent = ext4_fh_to_parent,
1313 .get_parent = ext4_get_parent,
1314 };
1315
1316 enum {
1317 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1318 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1319 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1320 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1321 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1322 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1323 Opt_journal_update, Opt_journal_dev,
1324 Opt_journal_checksum, Opt_journal_async_commit,
1325 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1326 Opt_data_err_abort, Opt_data_err_ignore,
1327 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1328 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1329 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1330 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1331 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1332 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1333 Opt_inode_readahead_blks, Opt_journal_ioprio,
1334 Opt_dioread_nolock, Opt_dioread_lock,
1335 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1336 };
1337
1338 static const match_table_t tokens = {
1339 {Opt_bsd_df, "bsddf"},
1340 {Opt_minix_df, "minixdf"},
1341 {Opt_grpid, "grpid"},
1342 {Opt_grpid, "bsdgroups"},
1343 {Opt_nogrpid, "nogrpid"},
1344 {Opt_nogrpid, "sysvgroups"},
1345 {Opt_resgid, "resgid=%u"},
1346 {Opt_resuid, "resuid=%u"},
1347 {Opt_sb, "sb=%u"},
1348 {Opt_err_cont, "errors=continue"},
1349 {Opt_err_panic, "errors=panic"},
1350 {Opt_err_ro, "errors=remount-ro"},
1351 {Opt_nouid32, "nouid32"},
1352 {Opt_debug, "debug"},
1353 {Opt_oldalloc, "oldalloc"},
1354 {Opt_orlov, "orlov"},
1355 {Opt_user_xattr, "user_xattr"},
1356 {Opt_nouser_xattr, "nouser_xattr"},
1357 {Opt_acl, "acl"},
1358 {Opt_noacl, "noacl"},
1359 {Opt_noload, "noload"},
1360 {Opt_noload, "norecovery"},
1361 {Opt_nobh, "nobh"},
1362 {Opt_bh, "bh"},
1363 {Opt_commit, "commit=%u"},
1364 {Opt_min_batch_time, "min_batch_time=%u"},
1365 {Opt_max_batch_time, "max_batch_time=%u"},
1366 {Opt_journal_update, "journal=update"},
1367 {Opt_journal_dev, "journal_dev=%u"},
1368 {Opt_journal_checksum, "journal_checksum"},
1369 {Opt_journal_async_commit, "journal_async_commit"},
1370 {Opt_abort, "abort"},
1371 {Opt_data_journal, "data=journal"},
1372 {Opt_data_ordered, "data=ordered"},
1373 {Opt_data_writeback, "data=writeback"},
1374 {Opt_data_err_abort, "data_err=abort"},
1375 {Opt_data_err_ignore, "data_err=ignore"},
1376 {Opt_offusrjquota, "usrjquota="},
1377 {Opt_usrjquota, "usrjquota=%s"},
1378 {Opt_offgrpjquota, "grpjquota="},
1379 {Opt_grpjquota, "grpjquota=%s"},
1380 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1381 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1382 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1383 {Opt_grpquota, "grpquota"},
1384 {Opt_noquota, "noquota"},
1385 {Opt_quota, "quota"},
1386 {Opt_usrquota, "usrquota"},
1387 {Opt_barrier, "barrier=%u"},
1388 {Opt_barrier, "barrier"},
1389 {Opt_nobarrier, "nobarrier"},
1390 {Opt_i_version, "i_version"},
1391 {Opt_stripe, "stripe=%u"},
1392 {Opt_resize, "resize"},
1393 {Opt_delalloc, "delalloc"},
1394 {Opt_nodelalloc, "nodelalloc"},
1395 {Opt_mblk_io_submit, "mblk_io_submit"},
1396 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1397 {Opt_block_validity, "block_validity"},
1398 {Opt_noblock_validity, "noblock_validity"},
1399 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1400 {Opt_journal_ioprio, "journal_ioprio=%u"},
1401 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1402 {Opt_auto_da_alloc, "auto_da_alloc"},
1403 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1404 {Opt_dioread_nolock, "dioread_nolock"},
1405 {Opt_dioread_lock, "dioread_lock"},
1406 {Opt_discard, "discard"},
1407 {Opt_nodiscard, "nodiscard"},
1408 {Opt_init_itable, "init_itable=%u"},
1409 {Opt_init_itable, "init_itable"},
1410 {Opt_noinit_itable, "noinit_itable"},
1411 {Opt_err, NULL},
1412 };
1413
1414 static ext4_fsblk_t get_sb_block(void **data)
1415 {
1416 ext4_fsblk_t sb_block;
1417 char *options = (char *) *data;
1418
1419 if (!options || strncmp(options, "sb=", 3) != 0)
1420 return 1; /* Default location */
1421
1422 options += 3;
1423 /* TODO: use simple_strtoll with >32bit ext4 */
1424 sb_block = simple_strtoul(options, &options, 0);
1425 if (*options && *options != ',') {
1426 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1427 (char *) *data);
1428 return 1;
1429 }
1430 if (*options == ',')
1431 options++;
1432 *data = (void *) options;
1433
1434 return sb_block;
1435 }
1436
1437 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1438 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1439 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1440
1441 #ifdef CONFIG_QUOTA
1442 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1443 {
1444 struct ext4_sb_info *sbi = EXT4_SB(sb);
1445 char *qname;
1446
1447 if (sb_any_quota_loaded(sb) &&
1448 !sbi->s_qf_names[qtype]) {
1449 ext4_msg(sb, KERN_ERR,
1450 "Cannot change journaled "
1451 "quota options when quota turned on");
1452 return 0;
1453 }
1454 qname = match_strdup(args);
1455 if (!qname) {
1456 ext4_msg(sb, KERN_ERR,
1457 "Not enough memory for storing quotafile name");
1458 return 0;
1459 }
1460 if (sbi->s_qf_names[qtype] &&
1461 strcmp(sbi->s_qf_names[qtype], qname)) {
1462 ext4_msg(sb, KERN_ERR,
1463 "%s quota file already specified", QTYPE2NAME(qtype));
1464 kfree(qname);
1465 return 0;
1466 }
1467 sbi->s_qf_names[qtype] = qname;
1468 if (strchr(sbi->s_qf_names[qtype], '/')) {
1469 ext4_msg(sb, KERN_ERR,
1470 "quotafile must be on filesystem root");
1471 kfree(sbi->s_qf_names[qtype]);
1472 sbi->s_qf_names[qtype] = NULL;
1473 return 0;
1474 }
1475 set_opt(sb, QUOTA);
1476 return 1;
1477 }
1478
1479 static int clear_qf_name(struct super_block *sb, int qtype)
1480 {
1481
1482 struct ext4_sb_info *sbi = EXT4_SB(sb);
1483
1484 if (sb_any_quota_loaded(sb) &&
1485 sbi->s_qf_names[qtype]) {
1486 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1487 " when quota turned on");
1488 return 0;
1489 }
1490 /*
1491 * The space will be released later when all options are confirmed
1492 * to be correct
1493 */
1494 sbi->s_qf_names[qtype] = NULL;
1495 return 1;
1496 }
1497 #endif
1498
1499 static int parse_options(char *options, struct super_block *sb,
1500 unsigned long *journal_devnum,
1501 unsigned int *journal_ioprio,
1502 ext4_fsblk_t *n_blocks_count, int is_remount)
1503 {
1504 struct ext4_sb_info *sbi = EXT4_SB(sb);
1505 char *p;
1506 substring_t args[MAX_OPT_ARGS];
1507 int data_opt = 0;
1508 int option;
1509 #ifdef CONFIG_QUOTA
1510 int qfmt;
1511 #endif
1512
1513 if (!options)
1514 return 1;
1515
1516 while ((p = strsep(&options, ",")) != NULL) {
1517 int token;
1518 if (!*p)
1519 continue;
1520
1521 /*
1522 * Initialize args struct so we know whether arg was
1523 * found; some options take optional arguments.
1524 */
1525 args[0].to = args[0].from = NULL;
1526 token = match_token(p, tokens, args);
1527 switch (token) {
1528 case Opt_bsd_df:
1529 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1530 clear_opt(sb, MINIX_DF);
1531 break;
1532 case Opt_minix_df:
1533 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1534 set_opt(sb, MINIX_DF);
1535
1536 break;
1537 case Opt_grpid:
1538 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1539 set_opt(sb, GRPID);
1540
1541 break;
1542 case Opt_nogrpid:
1543 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1544 clear_opt(sb, GRPID);
1545
1546 break;
1547 case Opt_resuid:
1548 if (match_int(&args[0], &option))
1549 return 0;
1550 sbi->s_resuid = option;
1551 break;
1552 case Opt_resgid:
1553 if (match_int(&args[0], &option))
1554 return 0;
1555 sbi->s_resgid = option;
1556 break;
1557 case Opt_sb:
1558 /* handled by get_sb_block() instead of here */
1559 /* *sb_block = match_int(&args[0]); */
1560 break;
1561 case Opt_err_panic:
1562 clear_opt(sb, ERRORS_CONT);
1563 clear_opt(sb, ERRORS_RO);
1564 set_opt(sb, ERRORS_PANIC);
1565 break;
1566 case Opt_err_ro:
1567 clear_opt(sb, ERRORS_CONT);
1568 clear_opt(sb, ERRORS_PANIC);
1569 set_opt(sb, ERRORS_RO);
1570 break;
1571 case Opt_err_cont:
1572 clear_opt(sb, ERRORS_RO);
1573 clear_opt(sb, ERRORS_PANIC);
1574 set_opt(sb, ERRORS_CONT);
1575 break;
1576 case Opt_nouid32:
1577 set_opt(sb, NO_UID32);
1578 break;
1579 case Opt_debug:
1580 set_opt(sb, DEBUG);
1581 break;
1582 case Opt_oldalloc:
1583 ext4_msg(sb, KERN_WARNING,
1584 "Ignoring deprecated oldalloc option");
1585 break;
1586 case Opt_orlov:
1587 ext4_msg(sb, KERN_WARNING,
1588 "Ignoring deprecated orlov option");
1589 break;
1590 #ifdef CONFIG_EXT4_FS_XATTR
1591 case Opt_user_xattr:
1592 set_opt(sb, XATTR_USER);
1593 break;
1594 case Opt_nouser_xattr:
1595 clear_opt(sb, XATTR_USER);
1596 break;
1597 #else
1598 case Opt_user_xattr:
1599 case Opt_nouser_xattr:
1600 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1601 break;
1602 #endif
1603 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1604 case Opt_acl:
1605 set_opt(sb, POSIX_ACL);
1606 break;
1607 case Opt_noacl:
1608 clear_opt(sb, POSIX_ACL);
1609 break;
1610 #else
1611 case Opt_acl:
1612 case Opt_noacl:
1613 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1614 break;
1615 #endif
1616 case Opt_journal_update:
1617 /* @@@ FIXME */
1618 /* Eventually we will want to be able to create
1619 a journal file here. For now, only allow the
1620 user to specify an existing inode to be the
1621 journal file. */
1622 if (is_remount) {
1623 ext4_msg(sb, KERN_ERR,
1624 "Cannot specify journal on remount");
1625 return 0;
1626 }
1627 set_opt(sb, UPDATE_JOURNAL);
1628 break;
1629 case Opt_journal_dev:
1630 if (is_remount) {
1631 ext4_msg(sb, KERN_ERR,
1632 "Cannot specify journal on remount");
1633 return 0;
1634 }
1635 if (match_int(&args[0], &option))
1636 return 0;
1637 *journal_devnum = option;
1638 break;
1639 case Opt_journal_checksum:
1640 set_opt(sb, JOURNAL_CHECKSUM);
1641 break;
1642 case Opt_journal_async_commit:
1643 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1644 set_opt(sb, JOURNAL_CHECKSUM);
1645 break;
1646 case Opt_noload:
1647 set_opt(sb, NOLOAD);
1648 break;
1649 case Opt_commit:
1650 if (match_int(&args[0], &option))
1651 return 0;
1652 if (option < 0)
1653 return 0;
1654 if (option == 0)
1655 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1656 sbi->s_commit_interval = HZ * option;
1657 break;
1658 case Opt_max_batch_time:
1659 if (match_int(&args[0], &option))
1660 return 0;
1661 if (option < 0)
1662 return 0;
1663 if (option == 0)
1664 option = EXT4_DEF_MAX_BATCH_TIME;
1665 sbi->s_max_batch_time = option;
1666 break;
1667 case Opt_min_batch_time:
1668 if (match_int(&args[0], &option))
1669 return 0;
1670 if (option < 0)
1671 return 0;
1672 sbi->s_min_batch_time = option;
1673 break;
1674 case Opt_data_journal:
1675 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1676 goto datacheck;
1677 case Opt_data_ordered:
1678 data_opt = EXT4_MOUNT_ORDERED_DATA;
1679 goto datacheck;
1680 case Opt_data_writeback:
1681 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1682 datacheck:
1683 if (is_remount) {
1684 if (!sbi->s_journal)
1685 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1686 else if (test_opt(sb, DATA_FLAGS) != data_opt) {
1687 ext4_msg(sb, KERN_ERR,
1688 "Cannot change data mode on remount");
1689 return 0;
1690 }
1691 } else {
1692 clear_opt(sb, DATA_FLAGS);
1693 sbi->s_mount_opt |= data_opt;
1694 }
1695 break;
1696 case Opt_data_err_abort:
1697 set_opt(sb, DATA_ERR_ABORT);
1698 break;
1699 case Opt_data_err_ignore:
1700 clear_opt(sb, DATA_ERR_ABORT);
1701 break;
1702 #ifdef CONFIG_QUOTA
1703 case Opt_usrjquota:
1704 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1705 return 0;
1706 break;
1707 case Opt_grpjquota:
1708 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1709 return 0;
1710 break;
1711 case Opt_offusrjquota:
1712 if (!clear_qf_name(sb, USRQUOTA))
1713 return 0;
1714 break;
1715 case Opt_offgrpjquota:
1716 if (!clear_qf_name(sb, GRPQUOTA))
1717 return 0;
1718 break;
1719
1720 case Opt_jqfmt_vfsold:
1721 qfmt = QFMT_VFS_OLD;
1722 goto set_qf_format;
1723 case Opt_jqfmt_vfsv0:
1724 qfmt = QFMT_VFS_V0;
1725 goto set_qf_format;
1726 case Opt_jqfmt_vfsv1:
1727 qfmt = QFMT_VFS_V1;
1728 set_qf_format:
1729 if (sb_any_quota_loaded(sb) &&
1730 sbi->s_jquota_fmt != qfmt) {
1731 ext4_msg(sb, KERN_ERR, "Cannot change "
1732 "journaled quota options when "
1733 "quota turned on");
1734 return 0;
1735 }
1736 sbi->s_jquota_fmt = qfmt;
1737 break;
1738 case Opt_quota:
1739 case Opt_usrquota:
1740 set_opt(sb, QUOTA);
1741 set_opt(sb, USRQUOTA);
1742 break;
1743 case Opt_grpquota:
1744 set_opt(sb, QUOTA);
1745 set_opt(sb, GRPQUOTA);
1746 break;
1747 case Opt_noquota:
1748 if (sb_any_quota_loaded(sb)) {
1749 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1750 "options when quota turned on");
1751 return 0;
1752 }
1753 clear_opt(sb, QUOTA);
1754 clear_opt(sb, USRQUOTA);
1755 clear_opt(sb, GRPQUOTA);
1756 break;
1757 #else
1758 case Opt_quota:
1759 case Opt_usrquota:
1760 case Opt_grpquota:
1761 ext4_msg(sb, KERN_ERR,
1762 "quota options not supported");
1763 break;
1764 case Opt_usrjquota:
1765 case Opt_grpjquota:
1766 case Opt_offusrjquota:
1767 case Opt_offgrpjquota:
1768 case Opt_jqfmt_vfsold:
1769 case Opt_jqfmt_vfsv0:
1770 case Opt_jqfmt_vfsv1:
1771 ext4_msg(sb, KERN_ERR,
1772 "journaled quota options not supported");
1773 break;
1774 case Opt_noquota:
1775 break;
1776 #endif
1777 case Opt_abort:
1778 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1779 break;
1780 case Opt_nobarrier:
1781 clear_opt(sb, BARRIER);
1782 break;
1783 case Opt_barrier:
1784 if (args[0].from) {
1785 if (match_int(&args[0], &option))
1786 return 0;
1787 } else
1788 option = 1; /* No argument, default to 1 */
1789 if (option)
1790 set_opt(sb, BARRIER);
1791 else
1792 clear_opt(sb, BARRIER);
1793 break;
1794 case Opt_ignore:
1795 break;
1796 case Opt_resize:
1797 if (!is_remount) {
1798 ext4_msg(sb, KERN_ERR,
1799 "resize option only available "
1800 "for remount");
1801 return 0;
1802 }
1803 if (match_int(&args[0], &option) != 0)
1804 return 0;
1805 *n_blocks_count = option;
1806 break;
1807 case Opt_nobh:
1808 ext4_msg(sb, KERN_WARNING,
1809 "Ignoring deprecated nobh option");
1810 break;
1811 case Opt_bh:
1812 ext4_msg(sb, KERN_WARNING,
1813 "Ignoring deprecated bh option");
1814 break;
1815 case Opt_i_version:
1816 set_opt(sb, I_VERSION);
1817 sb->s_flags |= MS_I_VERSION;
1818 break;
1819 case Opt_nodelalloc:
1820 clear_opt(sb, DELALLOC);
1821 clear_opt2(sb, EXPLICIT_DELALLOC);
1822 break;
1823 case Opt_mblk_io_submit:
1824 set_opt(sb, MBLK_IO_SUBMIT);
1825 break;
1826 case Opt_nomblk_io_submit:
1827 clear_opt(sb, MBLK_IO_SUBMIT);
1828 break;
1829 case Opt_stripe:
1830 if (match_int(&args[0], &option))
1831 return 0;
1832 if (option < 0)
1833 return 0;
1834 sbi->s_stripe = option;
1835 break;
1836 case Opt_delalloc:
1837 set_opt(sb, DELALLOC);
1838 set_opt2(sb, EXPLICIT_DELALLOC);
1839 break;
1840 case Opt_block_validity:
1841 set_opt(sb, BLOCK_VALIDITY);
1842 break;
1843 case Opt_noblock_validity:
1844 clear_opt(sb, BLOCK_VALIDITY);
1845 break;
1846 case Opt_inode_readahead_blks:
1847 if (match_int(&args[0], &option))
1848 return 0;
1849 if (option < 0 || option > (1 << 30))
1850 return 0;
1851 if (option && !is_power_of_2(option)) {
1852 ext4_msg(sb, KERN_ERR,
1853 "EXT4-fs: inode_readahead_blks"
1854 " must be a power of 2");
1855 return 0;
1856 }
1857 sbi->s_inode_readahead_blks = option;
1858 break;
1859 case Opt_journal_ioprio:
1860 if (match_int(&args[0], &option))
1861 return 0;
1862 if (option < 0 || option > 7)
1863 break;
1864 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1865 option);
1866 break;
1867 case Opt_noauto_da_alloc:
1868 set_opt(sb, NO_AUTO_DA_ALLOC);
1869 break;
1870 case Opt_auto_da_alloc:
1871 if (args[0].from) {
1872 if (match_int(&args[0], &option))
1873 return 0;
1874 } else
1875 option = 1; /* No argument, default to 1 */
1876 if (option)
1877 clear_opt(sb, NO_AUTO_DA_ALLOC);
1878 else
1879 set_opt(sb,NO_AUTO_DA_ALLOC);
1880 break;
1881 case Opt_discard:
1882 set_opt(sb, DISCARD);
1883 break;
1884 case Opt_nodiscard:
1885 clear_opt(sb, DISCARD);
1886 break;
1887 case Opt_dioread_nolock:
1888 set_opt(sb, DIOREAD_NOLOCK);
1889 break;
1890 case Opt_dioread_lock:
1891 clear_opt(sb, DIOREAD_NOLOCK);
1892 break;
1893 case Opt_init_itable:
1894 set_opt(sb, INIT_INODE_TABLE);
1895 if (args[0].from) {
1896 if (match_int(&args[0], &option))
1897 return 0;
1898 } else
1899 option = EXT4_DEF_LI_WAIT_MULT;
1900 if (option < 0)
1901 return 0;
1902 sbi->s_li_wait_mult = option;
1903 break;
1904 case Opt_noinit_itable:
1905 clear_opt(sb, INIT_INODE_TABLE);
1906 break;
1907 default:
1908 ext4_msg(sb, KERN_ERR,
1909 "Unrecognized mount option \"%s\" "
1910 "or missing value", p);
1911 return 0;
1912 }
1913 }
1914 #ifdef CONFIG_QUOTA
1915 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1916 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1917 clear_opt(sb, USRQUOTA);
1918
1919 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1920 clear_opt(sb, GRPQUOTA);
1921
1922 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1923 ext4_msg(sb, KERN_ERR, "old and new quota "
1924 "format mixing");
1925 return 0;
1926 }
1927
1928 if (!sbi->s_jquota_fmt) {
1929 ext4_msg(sb, KERN_ERR, "journaled quota format "
1930 "not specified");
1931 return 0;
1932 }
1933 } else {
1934 if (sbi->s_jquota_fmt) {
1935 ext4_msg(sb, KERN_ERR, "journaled quota format "
1936 "specified with no journaling "
1937 "enabled");
1938 return 0;
1939 }
1940 }
1941 #endif
1942 return 1;
1943 }
1944
1945 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1946 int read_only)
1947 {
1948 struct ext4_sb_info *sbi = EXT4_SB(sb);
1949 int res = 0;
1950
1951 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1952 ext4_msg(sb, KERN_ERR, "revision level too high, "
1953 "forcing read-only mode");
1954 res = MS_RDONLY;
1955 }
1956 if (read_only)
1957 goto done;
1958 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1959 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1960 "running e2fsck is recommended");
1961 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1962 ext4_msg(sb, KERN_WARNING,
1963 "warning: mounting fs with errors, "
1964 "running e2fsck is recommended");
1965 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1966 le16_to_cpu(es->s_mnt_count) >=
1967 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1968 ext4_msg(sb, KERN_WARNING,
1969 "warning: maximal mount count reached, "
1970 "running e2fsck is recommended");
1971 else if (le32_to_cpu(es->s_checkinterval) &&
1972 (le32_to_cpu(es->s_lastcheck) +
1973 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1974 ext4_msg(sb, KERN_WARNING,
1975 "warning: checktime reached, "
1976 "running e2fsck is recommended");
1977 if (!sbi->s_journal)
1978 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1979 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1980 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1981 le16_add_cpu(&es->s_mnt_count, 1);
1982 es->s_mtime = cpu_to_le32(get_seconds());
1983 ext4_update_dynamic_rev(sb);
1984 if (sbi->s_journal)
1985 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1986
1987 ext4_commit_super(sb, 1);
1988 done:
1989 if (test_opt(sb, DEBUG))
1990 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1991 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1992 sb->s_blocksize,
1993 sbi->s_groups_count,
1994 EXT4_BLOCKS_PER_GROUP(sb),
1995 EXT4_INODES_PER_GROUP(sb),
1996 sbi->s_mount_opt, sbi->s_mount_opt2);
1997
1998 cleancache_init_fs(sb);
1999 return res;
2000 }
2001
2002 static int ext4_fill_flex_info(struct super_block *sb)
2003 {
2004 struct ext4_sb_info *sbi = EXT4_SB(sb);
2005 struct ext4_group_desc *gdp = NULL;
2006 ext4_group_t flex_group_count;
2007 ext4_group_t flex_group;
2008 unsigned int groups_per_flex = 0;
2009 size_t size;
2010 int i;
2011
2012 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2013 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2014 sbi->s_log_groups_per_flex = 0;
2015 return 1;
2016 }
2017 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
2018
2019 /* We allocate both existing and potentially added groups */
2020 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2021 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2022 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2023 size = flex_group_count * sizeof(struct flex_groups);
2024 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2025 if (sbi->s_flex_groups == NULL) {
2026 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2027 flex_group_count);
2028 goto failed;
2029 }
2030
2031 for (i = 0; i < sbi->s_groups_count; i++) {
2032 gdp = ext4_get_group_desc(sb, i, NULL);
2033
2034 flex_group = ext4_flex_group(sbi, i);
2035 atomic_add(ext4_free_inodes_count(sb, gdp),
2036 &sbi->s_flex_groups[flex_group].free_inodes);
2037 atomic_add(ext4_free_group_clusters(sb, gdp),
2038 &sbi->s_flex_groups[flex_group].free_clusters);
2039 atomic_add(ext4_used_dirs_count(sb, gdp),
2040 &sbi->s_flex_groups[flex_group].used_dirs);
2041 }
2042
2043 return 1;
2044 failed:
2045 return 0;
2046 }
2047
2048 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2049 struct ext4_group_desc *gdp)
2050 {
2051 __u16 crc = 0;
2052
2053 if (sbi->s_es->s_feature_ro_compat &
2054 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2055 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2056 __le32 le_group = cpu_to_le32(block_group);
2057
2058 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2059 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2060 crc = crc16(crc, (__u8 *)gdp, offset);
2061 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2062 /* for checksum of struct ext4_group_desc do the rest...*/
2063 if ((sbi->s_es->s_feature_incompat &
2064 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2065 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2066 crc = crc16(crc, (__u8 *)gdp + offset,
2067 le16_to_cpu(sbi->s_es->s_desc_size) -
2068 offset);
2069 }
2070
2071 return cpu_to_le16(crc);
2072 }
2073
2074 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2075 struct ext4_group_desc *gdp)
2076 {
2077 if ((sbi->s_es->s_feature_ro_compat &
2078 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2079 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2080 return 0;
2081
2082 return 1;
2083 }
2084
2085 /* Called at mount-time, super-block is locked */
2086 static int ext4_check_descriptors(struct super_block *sb,
2087 ext4_group_t *first_not_zeroed)
2088 {
2089 struct ext4_sb_info *sbi = EXT4_SB(sb);
2090 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2091 ext4_fsblk_t last_block;
2092 ext4_fsblk_t block_bitmap;
2093 ext4_fsblk_t inode_bitmap;
2094 ext4_fsblk_t inode_table;
2095 int flexbg_flag = 0;
2096 ext4_group_t i, grp = sbi->s_groups_count;
2097
2098 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2099 flexbg_flag = 1;
2100
2101 ext4_debug("Checking group descriptors");
2102
2103 for (i = 0; i < sbi->s_groups_count; i++) {
2104 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2105
2106 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2107 last_block = ext4_blocks_count(sbi->s_es) - 1;
2108 else
2109 last_block = first_block +
2110 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2111
2112 if ((grp == sbi->s_groups_count) &&
2113 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2114 grp = i;
2115
2116 block_bitmap = ext4_block_bitmap(sb, gdp);
2117 if (block_bitmap < first_block || block_bitmap > last_block) {
2118 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2119 "Block bitmap for group %u not in group "
2120 "(block %llu)!", i, block_bitmap);
2121 return 0;
2122 }
2123 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2124 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2125 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2126 "Inode bitmap for group %u not in group "
2127 "(block %llu)!", i, inode_bitmap);
2128 return 0;
2129 }
2130 inode_table = ext4_inode_table(sb, gdp);
2131 if (inode_table < first_block ||
2132 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2133 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2134 "Inode table for group %u not in group "
2135 "(block %llu)!", i, inode_table);
2136 return 0;
2137 }
2138 ext4_lock_group(sb, i);
2139 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2140 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2141 "Checksum for group %u failed (%u!=%u)",
2142 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2143 gdp)), le16_to_cpu(gdp->bg_checksum));
2144 if (!(sb->s_flags & MS_RDONLY)) {
2145 ext4_unlock_group(sb, i);
2146 return 0;
2147 }
2148 }
2149 ext4_unlock_group(sb, i);
2150 if (!flexbg_flag)
2151 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2152 }
2153 if (NULL != first_not_zeroed)
2154 *first_not_zeroed = grp;
2155
2156 ext4_free_blocks_count_set(sbi->s_es,
2157 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2158 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2159 return 1;
2160 }
2161
2162 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2163 * the superblock) which were deleted from all directories, but held open by
2164 * a process at the time of a crash. We walk the list and try to delete these
2165 * inodes at recovery time (only with a read-write filesystem).
2166 *
2167 * In order to keep the orphan inode chain consistent during traversal (in
2168 * case of crash during recovery), we link each inode into the superblock
2169 * orphan list_head and handle it the same way as an inode deletion during
2170 * normal operation (which journals the operations for us).
2171 *
2172 * We only do an iget() and an iput() on each inode, which is very safe if we
2173 * accidentally point at an in-use or already deleted inode. The worst that
2174 * can happen in this case is that we get a "bit already cleared" message from
2175 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2176 * e2fsck was run on this filesystem, and it must have already done the orphan
2177 * inode cleanup for us, so we can safely abort without any further action.
2178 */
2179 static void ext4_orphan_cleanup(struct super_block *sb,
2180 struct ext4_super_block *es)
2181 {
2182 unsigned int s_flags = sb->s_flags;
2183 int nr_orphans = 0, nr_truncates = 0;
2184 #ifdef CONFIG_QUOTA
2185 int i;
2186 #endif
2187 if (!es->s_last_orphan) {
2188 jbd_debug(4, "no orphan inodes to clean up\n");
2189 return;
2190 }
2191
2192 if (bdev_read_only(sb->s_bdev)) {
2193 ext4_msg(sb, KERN_ERR, "write access "
2194 "unavailable, skipping orphan cleanup");
2195 return;
2196 }
2197
2198 /* Check if feature set would not allow a r/w mount */
2199 if (!ext4_feature_set_ok(sb, 0)) {
2200 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2201 "unknown ROCOMPAT features");
2202 return;
2203 }
2204
2205 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2206 if (es->s_last_orphan)
2207 jbd_debug(1, "Errors on filesystem, "
2208 "clearing orphan list.\n");
2209 es->s_last_orphan = 0;
2210 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2211 return;
2212 }
2213
2214 if (s_flags & MS_RDONLY) {
2215 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2216 sb->s_flags &= ~MS_RDONLY;
2217 }
2218 #ifdef CONFIG_QUOTA
2219 /* Needed for iput() to work correctly and not trash data */
2220 sb->s_flags |= MS_ACTIVE;
2221 /* Turn on quotas so that they are updated correctly */
2222 for (i = 0; i < MAXQUOTAS; i++) {
2223 if (EXT4_SB(sb)->s_qf_names[i]) {
2224 int ret = ext4_quota_on_mount(sb, i);
2225 if (ret < 0)
2226 ext4_msg(sb, KERN_ERR,
2227 "Cannot turn on journaled "
2228 "quota: error %d", ret);
2229 }
2230 }
2231 #endif
2232
2233 while (es->s_last_orphan) {
2234 struct inode *inode;
2235
2236 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2237 if (IS_ERR(inode)) {
2238 es->s_last_orphan = 0;
2239 break;
2240 }
2241
2242 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2243 dquot_initialize(inode);
2244 if (inode->i_nlink) {
2245 ext4_msg(sb, KERN_DEBUG,
2246 "%s: truncating inode %lu to %lld bytes",
2247 __func__, inode->i_ino, inode->i_size);
2248 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2249 inode->i_ino, inode->i_size);
2250 ext4_truncate(inode);
2251 nr_truncates++;
2252 } else {
2253 ext4_msg(sb, KERN_DEBUG,
2254 "%s: deleting unreferenced inode %lu",
2255 __func__, inode->i_ino);
2256 jbd_debug(2, "deleting unreferenced inode %lu\n",
2257 inode->i_ino);
2258 nr_orphans++;
2259 }
2260 iput(inode); /* The delete magic happens here! */
2261 }
2262
2263 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2264
2265 if (nr_orphans)
2266 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2267 PLURAL(nr_orphans));
2268 if (nr_truncates)
2269 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2270 PLURAL(nr_truncates));
2271 #ifdef CONFIG_QUOTA
2272 /* Turn quotas off */
2273 for (i = 0; i < MAXQUOTAS; i++) {
2274 if (sb_dqopt(sb)->files[i])
2275 dquot_quota_off(sb, i);
2276 }
2277 #endif
2278 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2279 }
2280
2281 /*
2282 * Maximal extent format file size.
2283 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2284 * extent format containers, within a sector_t, and within i_blocks
2285 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2286 * so that won't be a limiting factor.
2287 *
2288 * However there is other limiting factor. We do store extents in the form
2289 * of starting block and length, hence the resulting length of the extent
2290 * covering maximum file size must fit into on-disk format containers as
2291 * well. Given that length is always by 1 unit bigger than max unit (because
2292 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2293 *
2294 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2295 */
2296 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2297 {
2298 loff_t res;
2299 loff_t upper_limit = MAX_LFS_FILESIZE;
2300
2301 /* small i_blocks in vfs inode? */
2302 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2303 /*
2304 * CONFIG_LBDAF is not enabled implies the inode
2305 * i_block represent total blocks in 512 bytes
2306 * 32 == size of vfs inode i_blocks * 8
2307 */
2308 upper_limit = (1LL << 32) - 1;
2309
2310 /* total blocks in file system block size */
2311 upper_limit >>= (blkbits - 9);
2312 upper_limit <<= blkbits;
2313 }
2314
2315 /*
2316 * 32-bit extent-start container, ee_block. We lower the maxbytes
2317 * by one fs block, so ee_len can cover the extent of maximum file
2318 * size
2319 */
2320 res = (1LL << 32) - 1;
2321 res <<= blkbits;
2322
2323 /* Sanity check against vm- & vfs- imposed limits */
2324 if (res > upper_limit)
2325 res = upper_limit;
2326
2327 return res;
2328 }
2329
2330 /*
2331 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2332 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2333 * We need to be 1 filesystem block less than the 2^48 sector limit.
2334 */
2335 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2336 {
2337 loff_t res = EXT4_NDIR_BLOCKS;
2338 int meta_blocks;
2339 loff_t upper_limit;
2340 /* This is calculated to be the largest file size for a dense, block
2341 * mapped file such that the file's total number of 512-byte sectors,
2342 * including data and all indirect blocks, does not exceed (2^48 - 1).
2343 *
2344 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2345 * number of 512-byte sectors of the file.
2346 */
2347
2348 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2349 /*
2350 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2351 * the inode i_block field represents total file blocks in
2352 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2353 */
2354 upper_limit = (1LL << 32) - 1;
2355
2356 /* total blocks in file system block size */
2357 upper_limit >>= (bits - 9);
2358
2359 } else {
2360 /*
2361 * We use 48 bit ext4_inode i_blocks
2362 * With EXT4_HUGE_FILE_FL set the i_blocks
2363 * represent total number of blocks in
2364 * file system block size
2365 */
2366 upper_limit = (1LL << 48) - 1;
2367
2368 }
2369
2370 /* indirect blocks */
2371 meta_blocks = 1;
2372 /* double indirect blocks */
2373 meta_blocks += 1 + (1LL << (bits-2));
2374 /* tripple indirect blocks */
2375 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2376
2377 upper_limit -= meta_blocks;
2378 upper_limit <<= bits;
2379
2380 res += 1LL << (bits-2);
2381 res += 1LL << (2*(bits-2));
2382 res += 1LL << (3*(bits-2));
2383 res <<= bits;
2384 if (res > upper_limit)
2385 res = upper_limit;
2386
2387 if (res > MAX_LFS_FILESIZE)
2388 res = MAX_LFS_FILESIZE;
2389
2390 return res;
2391 }
2392
2393 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2394 ext4_fsblk_t logical_sb_block, int nr)
2395 {
2396 struct ext4_sb_info *sbi = EXT4_SB(sb);
2397 ext4_group_t bg, first_meta_bg;
2398 int has_super = 0;
2399
2400 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2401
2402 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2403 nr < first_meta_bg)
2404 return logical_sb_block + nr + 1;
2405 bg = sbi->s_desc_per_block * nr;
2406 if (ext4_bg_has_super(sb, bg))
2407 has_super = 1;
2408
2409 return (has_super + ext4_group_first_block_no(sb, bg));
2410 }
2411
2412 /**
2413 * ext4_get_stripe_size: Get the stripe size.
2414 * @sbi: In memory super block info
2415 *
2416 * If we have specified it via mount option, then
2417 * use the mount option value. If the value specified at mount time is
2418 * greater than the blocks per group use the super block value.
2419 * If the super block value is greater than blocks per group return 0.
2420 * Allocator needs it be less than blocks per group.
2421 *
2422 */
2423 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2424 {
2425 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2426 unsigned long stripe_width =
2427 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2428 int ret;
2429
2430 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2431 ret = sbi->s_stripe;
2432 else if (stripe_width <= sbi->s_blocks_per_group)
2433 ret = stripe_width;
2434 else if (stride <= sbi->s_blocks_per_group)
2435 ret = stride;
2436 else
2437 ret = 0;
2438
2439 /*
2440 * If the stripe width is 1, this makes no sense and
2441 * we set it to 0 to turn off stripe handling code.
2442 */
2443 if (ret <= 1)
2444 ret = 0;
2445
2446 return ret;
2447 }
2448
2449 /* sysfs supprt */
2450
2451 struct ext4_attr {
2452 struct attribute attr;
2453 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2454 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2455 const char *, size_t);
2456 int offset;
2457 };
2458
2459 static int parse_strtoul(const char *buf,
2460 unsigned long max, unsigned long *value)
2461 {
2462 char *endp;
2463
2464 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2465 endp = skip_spaces(endp);
2466 if (*endp || *value > max)
2467 return -EINVAL;
2468
2469 return 0;
2470 }
2471
2472 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2473 struct ext4_sb_info *sbi,
2474 char *buf)
2475 {
2476 return snprintf(buf, PAGE_SIZE, "%llu\n",
2477 (s64) EXT4_C2B(sbi,
2478 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2479 }
2480
2481 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2482 struct ext4_sb_info *sbi, char *buf)
2483 {
2484 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2485
2486 if (!sb->s_bdev->bd_part)
2487 return snprintf(buf, PAGE_SIZE, "0\n");
2488 return snprintf(buf, PAGE_SIZE, "%lu\n",
2489 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2490 sbi->s_sectors_written_start) >> 1);
2491 }
2492
2493 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2494 struct ext4_sb_info *sbi, char *buf)
2495 {
2496 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2497
2498 if (!sb->s_bdev->bd_part)
2499 return snprintf(buf, PAGE_SIZE, "0\n");
2500 return snprintf(buf, PAGE_SIZE, "%llu\n",
2501 (unsigned long long)(sbi->s_kbytes_written +
2502 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2503 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2504 }
2505
2506 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2507 struct ext4_sb_info *sbi,
2508 const char *buf, size_t count)
2509 {
2510 unsigned long t;
2511
2512 if (parse_strtoul(buf, 0x40000000, &t))
2513 return -EINVAL;
2514
2515 if (t && !is_power_of_2(t))
2516 return -EINVAL;
2517
2518 sbi->s_inode_readahead_blks = t;
2519 return count;
2520 }
2521
2522 static ssize_t sbi_ui_show(struct ext4_attr *a,
2523 struct ext4_sb_info *sbi, char *buf)
2524 {
2525 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2526
2527 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2528 }
2529
2530 static ssize_t sbi_ui_store(struct ext4_attr *a,
2531 struct ext4_sb_info *sbi,
2532 const char *buf, size_t count)
2533 {
2534 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2535 unsigned long t;
2536
2537 if (parse_strtoul(buf, 0xffffffff, &t))
2538 return -EINVAL;
2539 *ui = t;
2540 return count;
2541 }
2542
2543 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2544 static struct ext4_attr ext4_attr_##_name = { \
2545 .attr = {.name = __stringify(_name), .mode = _mode }, \
2546 .show = _show, \
2547 .store = _store, \
2548 .offset = offsetof(struct ext4_sb_info, _elname), \
2549 }
2550 #define EXT4_ATTR(name, mode, show, store) \
2551 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2552
2553 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2554 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2555 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2556 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2557 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2558 #define ATTR_LIST(name) &ext4_attr_##name.attr
2559
2560 EXT4_RO_ATTR(delayed_allocation_blocks);
2561 EXT4_RO_ATTR(session_write_kbytes);
2562 EXT4_RO_ATTR(lifetime_write_kbytes);
2563 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2564 inode_readahead_blks_store, s_inode_readahead_blks);
2565 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2566 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2567 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2568 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2569 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2570 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2571 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2572 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2573
2574 static struct attribute *ext4_attrs[] = {
2575 ATTR_LIST(delayed_allocation_blocks),
2576 ATTR_LIST(session_write_kbytes),
2577 ATTR_LIST(lifetime_write_kbytes),
2578 ATTR_LIST(inode_readahead_blks),
2579 ATTR_LIST(inode_goal),
2580 ATTR_LIST(mb_stats),
2581 ATTR_LIST(mb_max_to_scan),
2582 ATTR_LIST(mb_min_to_scan),
2583 ATTR_LIST(mb_order2_req),
2584 ATTR_LIST(mb_stream_req),
2585 ATTR_LIST(mb_group_prealloc),
2586 ATTR_LIST(max_writeback_mb_bump),
2587 NULL,
2588 };
2589
2590 /* Features this copy of ext4 supports */
2591 EXT4_INFO_ATTR(lazy_itable_init);
2592 EXT4_INFO_ATTR(batched_discard);
2593
2594 static struct attribute *ext4_feat_attrs[] = {
2595 ATTR_LIST(lazy_itable_init),
2596 ATTR_LIST(batched_discard),
2597 NULL,
2598 };
2599
2600 static ssize_t ext4_attr_show(struct kobject *kobj,
2601 struct attribute *attr, char *buf)
2602 {
2603 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2604 s_kobj);
2605 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2606
2607 return a->show ? a->show(a, sbi, buf) : 0;
2608 }
2609
2610 static ssize_t ext4_attr_store(struct kobject *kobj,
2611 struct attribute *attr,
2612 const char *buf, size_t len)
2613 {
2614 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2615 s_kobj);
2616 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2617
2618 return a->store ? a->store(a, sbi, buf, len) : 0;
2619 }
2620
2621 static void ext4_sb_release(struct kobject *kobj)
2622 {
2623 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2624 s_kobj);
2625 complete(&sbi->s_kobj_unregister);
2626 }
2627
2628 static const struct sysfs_ops ext4_attr_ops = {
2629 .show = ext4_attr_show,
2630 .store = ext4_attr_store,
2631 };
2632
2633 static struct kobj_type ext4_ktype = {
2634 .default_attrs = ext4_attrs,
2635 .sysfs_ops = &ext4_attr_ops,
2636 .release = ext4_sb_release,
2637 };
2638
2639 static void ext4_feat_release(struct kobject *kobj)
2640 {
2641 complete(&ext4_feat->f_kobj_unregister);
2642 }
2643
2644 static struct kobj_type ext4_feat_ktype = {
2645 .default_attrs = ext4_feat_attrs,
2646 .sysfs_ops = &ext4_attr_ops,
2647 .release = ext4_feat_release,
2648 };
2649
2650 /*
2651 * Check whether this filesystem can be mounted based on
2652 * the features present and the RDONLY/RDWR mount requested.
2653 * Returns 1 if this filesystem can be mounted as requested,
2654 * 0 if it cannot be.
2655 */
2656 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2657 {
2658 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2659 ext4_msg(sb, KERN_ERR,
2660 "Couldn't mount because of "
2661 "unsupported optional features (%x)",
2662 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2663 ~EXT4_FEATURE_INCOMPAT_SUPP));
2664 return 0;
2665 }
2666
2667 if (readonly)
2668 return 1;
2669
2670 /* Check that feature set is OK for a read-write mount */
2671 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2672 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2673 "unsupported optional features (%x)",
2674 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2675 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2676 return 0;
2677 }
2678 /*
2679 * Large file size enabled file system can only be mounted
2680 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2681 */
2682 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2683 if (sizeof(blkcnt_t) < sizeof(u64)) {
2684 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2685 "cannot be mounted RDWR without "
2686 "CONFIG_LBDAF");
2687 return 0;
2688 }
2689 }
2690 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2691 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2692 ext4_msg(sb, KERN_ERR,
2693 "Can't support bigalloc feature without "
2694 "extents feature\n");
2695 return 0;
2696 }
2697 return 1;
2698 }
2699
2700 /*
2701 * This function is called once a day if we have errors logged
2702 * on the file system
2703 */
2704 static void print_daily_error_info(unsigned long arg)
2705 {
2706 struct super_block *sb = (struct super_block *) arg;
2707 struct ext4_sb_info *sbi;
2708 struct ext4_super_block *es;
2709
2710 sbi = EXT4_SB(sb);
2711 es = sbi->s_es;
2712
2713 if (es->s_error_count)
2714 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2715 le32_to_cpu(es->s_error_count));
2716 if (es->s_first_error_time) {
2717 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2718 sb->s_id, le32_to_cpu(es->s_first_error_time),
2719 (int) sizeof(es->s_first_error_func),
2720 es->s_first_error_func,
2721 le32_to_cpu(es->s_first_error_line));
2722 if (es->s_first_error_ino)
2723 printk(": inode %u",
2724 le32_to_cpu(es->s_first_error_ino));
2725 if (es->s_first_error_block)
2726 printk(": block %llu", (unsigned long long)
2727 le64_to_cpu(es->s_first_error_block));
2728 printk("\n");
2729 }
2730 if (es->s_last_error_time) {
2731 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2732 sb->s_id, le32_to_cpu(es->s_last_error_time),
2733 (int) sizeof(es->s_last_error_func),
2734 es->s_last_error_func,
2735 le32_to_cpu(es->s_last_error_line));
2736 if (es->s_last_error_ino)
2737 printk(": inode %u",
2738 le32_to_cpu(es->s_last_error_ino));
2739 if (es->s_last_error_block)
2740 printk(": block %llu", (unsigned long long)
2741 le64_to_cpu(es->s_last_error_block));
2742 printk("\n");
2743 }
2744 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2745 }
2746
2747 /* Find next suitable group and run ext4_init_inode_table */
2748 static int ext4_run_li_request(struct ext4_li_request *elr)
2749 {
2750 struct ext4_group_desc *gdp = NULL;
2751 ext4_group_t group, ngroups;
2752 struct super_block *sb;
2753 unsigned long timeout = 0;
2754 int ret = 0;
2755
2756 sb = elr->lr_super;
2757 ngroups = EXT4_SB(sb)->s_groups_count;
2758
2759 for (group = elr->lr_next_group; group < ngroups; group++) {
2760 gdp = ext4_get_group_desc(sb, group, NULL);
2761 if (!gdp) {
2762 ret = 1;
2763 break;
2764 }
2765
2766 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2767 break;
2768 }
2769
2770 if (group == ngroups)
2771 ret = 1;
2772
2773 if (!ret) {
2774 timeout = jiffies;
2775 ret = ext4_init_inode_table(sb, group,
2776 elr->lr_timeout ? 0 : 1);
2777 if (elr->lr_timeout == 0) {
2778 timeout = (jiffies - timeout) *
2779 elr->lr_sbi->s_li_wait_mult;
2780 elr->lr_timeout = timeout;
2781 }
2782 elr->lr_next_sched = jiffies + elr->lr_timeout;
2783 elr->lr_next_group = group + 1;
2784 }
2785
2786 return ret;
2787 }
2788
2789 /*
2790 * Remove lr_request from the list_request and free the
2791 * request structure. Should be called with li_list_mtx held
2792 */
2793 static void ext4_remove_li_request(struct ext4_li_request *elr)
2794 {
2795 struct ext4_sb_info *sbi;
2796
2797 if (!elr)
2798 return;
2799
2800 sbi = elr->lr_sbi;
2801
2802 list_del(&elr->lr_request);
2803 sbi->s_li_request = NULL;
2804 kfree(elr);
2805 }
2806
2807 static void ext4_unregister_li_request(struct super_block *sb)
2808 {
2809 mutex_lock(&ext4_li_mtx);
2810 if (!ext4_li_info) {
2811 mutex_unlock(&ext4_li_mtx);
2812 return;
2813 }
2814
2815 mutex_lock(&ext4_li_info->li_list_mtx);
2816 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2817 mutex_unlock(&ext4_li_info->li_list_mtx);
2818 mutex_unlock(&ext4_li_mtx);
2819 }
2820
2821 static struct task_struct *ext4_lazyinit_task;
2822
2823 /*
2824 * This is the function where ext4lazyinit thread lives. It walks
2825 * through the request list searching for next scheduled filesystem.
2826 * When such a fs is found, run the lazy initialization request
2827 * (ext4_rn_li_request) and keep track of the time spend in this
2828 * function. Based on that time we compute next schedule time of
2829 * the request. When walking through the list is complete, compute
2830 * next waking time and put itself into sleep.
2831 */
2832 static int ext4_lazyinit_thread(void *arg)
2833 {
2834 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2835 struct list_head *pos, *n;
2836 struct ext4_li_request *elr;
2837 unsigned long next_wakeup, cur;
2838
2839 BUG_ON(NULL == eli);
2840
2841 cont_thread:
2842 while (true) {
2843 next_wakeup = MAX_JIFFY_OFFSET;
2844
2845 mutex_lock(&eli->li_list_mtx);
2846 if (list_empty(&eli->li_request_list)) {
2847 mutex_unlock(&eli->li_list_mtx);
2848 goto exit_thread;
2849 }
2850
2851 list_for_each_safe(pos, n, &eli->li_request_list) {
2852 elr = list_entry(pos, struct ext4_li_request,
2853 lr_request);
2854
2855 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2856 if (ext4_run_li_request(elr) != 0) {
2857 /* error, remove the lazy_init job */
2858 ext4_remove_li_request(elr);
2859 continue;
2860 }
2861 }
2862
2863 if (time_before(elr->lr_next_sched, next_wakeup))
2864 next_wakeup = elr->lr_next_sched;
2865 }
2866 mutex_unlock(&eli->li_list_mtx);
2867
2868 try_to_freeze();
2869
2870 cur = jiffies;
2871 if ((time_after_eq(cur, next_wakeup)) ||
2872 (MAX_JIFFY_OFFSET == next_wakeup)) {
2873 cond_resched();
2874 continue;
2875 }
2876
2877 schedule_timeout_interruptible(next_wakeup - cur);
2878
2879 if (kthread_should_stop()) {
2880 ext4_clear_request_list();
2881 goto exit_thread;
2882 }
2883 }
2884
2885 exit_thread:
2886 /*
2887 * It looks like the request list is empty, but we need
2888 * to check it under the li_list_mtx lock, to prevent any
2889 * additions into it, and of course we should lock ext4_li_mtx
2890 * to atomically free the list and ext4_li_info, because at
2891 * this point another ext4 filesystem could be registering
2892 * new one.
2893 */
2894 mutex_lock(&ext4_li_mtx);
2895 mutex_lock(&eli->li_list_mtx);
2896 if (!list_empty(&eli->li_request_list)) {
2897 mutex_unlock(&eli->li_list_mtx);
2898 mutex_unlock(&ext4_li_mtx);
2899 goto cont_thread;
2900 }
2901 mutex_unlock(&eli->li_list_mtx);
2902 kfree(ext4_li_info);
2903 ext4_li_info = NULL;
2904 mutex_unlock(&ext4_li_mtx);
2905
2906 return 0;
2907 }
2908
2909 static void ext4_clear_request_list(void)
2910 {
2911 struct list_head *pos, *n;
2912 struct ext4_li_request *elr;
2913
2914 mutex_lock(&ext4_li_info->li_list_mtx);
2915 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2916 elr = list_entry(pos, struct ext4_li_request,
2917 lr_request);
2918 ext4_remove_li_request(elr);
2919 }
2920 mutex_unlock(&ext4_li_info->li_list_mtx);
2921 }
2922
2923 static int ext4_run_lazyinit_thread(void)
2924 {
2925 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2926 ext4_li_info, "ext4lazyinit");
2927 if (IS_ERR(ext4_lazyinit_task)) {
2928 int err = PTR_ERR(ext4_lazyinit_task);
2929 ext4_clear_request_list();
2930 kfree(ext4_li_info);
2931 ext4_li_info = NULL;
2932 printk(KERN_CRIT "EXT4: error %d creating inode table "
2933 "initialization thread\n",
2934 err);
2935 return err;
2936 }
2937 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2938 return 0;
2939 }
2940
2941 /*
2942 * Check whether it make sense to run itable init. thread or not.
2943 * If there is at least one uninitialized inode table, return
2944 * corresponding group number, else the loop goes through all
2945 * groups and return total number of groups.
2946 */
2947 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2948 {
2949 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2950 struct ext4_group_desc *gdp = NULL;
2951
2952 for (group = 0; group < ngroups; group++) {
2953 gdp = ext4_get_group_desc(sb, group, NULL);
2954 if (!gdp)
2955 continue;
2956
2957 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2958 break;
2959 }
2960
2961 return group;
2962 }
2963
2964 static int ext4_li_info_new(void)
2965 {
2966 struct ext4_lazy_init *eli = NULL;
2967
2968 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2969 if (!eli)
2970 return -ENOMEM;
2971
2972 INIT_LIST_HEAD(&eli->li_request_list);
2973 mutex_init(&eli->li_list_mtx);
2974
2975 eli->li_state |= EXT4_LAZYINIT_QUIT;
2976
2977 ext4_li_info = eli;
2978
2979 return 0;
2980 }
2981
2982 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2983 ext4_group_t start)
2984 {
2985 struct ext4_sb_info *sbi = EXT4_SB(sb);
2986 struct ext4_li_request *elr;
2987 unsigned long rnd;
2988
2989 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2990 if (!elr)
2991 return NULL;
2992
2993 elr->lr_super = sb;
2994 elr->lr_sbi = sbi;
2995 elr->lr_next_group = start;
2996
2997 /*
2998 * Randomize first schedule time of the request to
2999 * spread the inode table initialization requests
3000 * better.
3001 */
3002 get_random_bytes(&rnd, sizeof(rnd));
3003 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3004 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3005
3006 return elr;
3007 }
3008
3009 static int ext4_register_li_request(struct super_block *sb,
3010 ext4_group_t first_not_zeroed)
3011 {
3012 struct ext4_sb_info *sbi = EXT4_SB(sb);
3013 struct ext4_li_request *elr;
3014 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3015 int ret = 0;
3016
3017 if (sbi->s_li_request != NULL) {
3018 /*
3019 * Reset timeout so it can be computed again, because
3020 * s_li_wait_mult might have changed.
3021 */
3022 sbi->s_li_request->lr_timeout = 0;
3023 return 0;
3024 }
3025
3026 if (first_not_zeroed == ngroups ||
3027 (sb->s_flags & MS_RDONLY) ||
3028 !test_opt(sb, INIT_INODE_TABLE))
3029 return 0;
3030
3031 elr = ext4_li_request_new(sb, first_not_zeroed);
3032 if (!elr)
3033 return -ENOMEM;
3034
3035 mutex_lock(&ext4_li_mtx);
3036
3037 if (NULL == ext4_li_info) {
3038 ret = ext4_li_info_new();
3039 if (ret)
3040 goto out;
3041 }
3042
3043 mutex_lock(&ext4_li_info->li_list_mtx);
3044 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3045 mutex_unlock(&ext4_li_info->li_list_mtx);
3046
3047 sbi->s_li_request = elr;
3048 /*
3049 * set elr to NULL here since it has been inserted to
3050 * the request_list and the removal and free of it is
3051 * handled by ext4_clear_request_list from now on.
3052 */
3053 elr = NULL;
3054
3055 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3056 ret = ext4_run_lazyinit_thread();
3057 if (ret)
3058 goto out;
3059 }
3060 out:
3061 mutex_unlock(&ext4_li_mtx);
3062 if (ret)
3063 kfree(elr);
3064 return ret;
3065 }
3066
3067 /*
3068 * We do not need to lock anything since this is called on
3069 * module unload.
3070 */
3071 static void ext4_destroy_lazyinit_thread(void)
3072 {
3073 /*
3074 * If thread exited earlier
3075 * there's nothing to be done.
3076 */
3077 if (!ext4_li_info || !ext4_lazyinit_task)
3078 return;
3079
3080 kthread_stop(ext4_lazyinit_task);
3081 }
3082
3083 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3084 {
3085 char *orig_data = kstrdup(data, GFP_KERNEL);
3086 struct buffer_head *bh;
3087 struct ext4_super_block *es = NULL;
3088 struct ext4_sb_info *sbi;
3089 ext4_fsblk_t block;
3090 ext4_fsblk_t sb_block = get_sb_block(&data);
3091 ext4_fsblk_t logical_sb_block;
3092 unsigned long offset = 0;
3093 unsigned long journal_devnum = 0;
3094 unsigned long def_mount_opts;
3095 struct inode *root;
3096 char *cp;
3097 const char *descr;
3098 int ret = -ENOMEM;
3099 int blocksize, clustersize;
3100 unsigned int db_count;
3101 unsigned int i;
3102 int needs_recovery, has_huge_files, has_bigalloc;
3103 __u64 blocks_count;
3104 int err;
3105 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3106 ext4_group_t first_not_zeroed;
3107
3108 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3109 if (!sbi)
3110 goto out_free_orig;
3111
3112 sbi->s_blockgroup_lock =
3113 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3114 if (!sbi->s_blockgroup_lock) {
3115 kfree(sbi);
3116 goto out_free_orig;
3117 }
3118 sb->s_fs_info = sbi;
3119 sbi->s_mount_opt = 0;
3120 sbi->s_resuid = EXT4_DEF_RESUID;
3121 sbi->s_resgid = EXT4_DEF_RESGID;
3122 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3123 sbi->s_sb_block = sb_block;
3124 if (sb->s_bdev->bd_part)
3125 sbi->s_sectors_written_start =
3126 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3127
3128 /* Cleanup superblock name */
3129 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3130 *cp = '!';
3131
3132 ret = -EINVAL;
3133 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3134 if (!blocksize) {
3135 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3136 goto out_fail;
3137 }
3138
3139 /*
3140 * The ext4 superblock will not be buffer aligned for other than 1kB
3141 * block sizes. We need to calculate the offset from buffer start.
3142 */
3143 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3144 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3145 offset = do_div(logical_sb_block, blocksize);
3146 } else {
3147 logical_sb_block = sb_block;
3148 }
3149
3150 if (!(bh = sb_bread(sb, logical_sb_block))) {
3151 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3152 goto out_fail;
3153 }
3154 /*
3155 * Note: s_es must be initialized as soon as possible because
3156 * some ext4 macro-instructions depend on its value
3157 */
3158 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3159 sbi->s_es = es;
3160 sb->s_magic = le16_to_cpu(es->s_magic);
3161 if (sb->s_magic != EXT4_SUPER_MAGIC)
3162 goto cantfind_ext4;
3163 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3164
3165 /* Set defaults before we parse the mount options */
3166 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3167 set_opt(sb, INIT_INODE_TABLE);
3168 if (def_mount_opts & EXT4_DEFM_DEBUG)
3169 set_opt(sb, DEBUG);
3170 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3171 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3172 "2.6.38");
3173 set_opt(sb, GRPID);
3174 }
3175 if (def_mount_opts & EXT4_DEFM_UID16)
3176 set_opt(sb, NO_UID32);
3177 /* xattr user namespace & acls are now defaulted on */
3178 #ifdef CONFIG_EXT4_FS_XATTR
3179 set_opt(sb, XATTR_USER);
3180 #endif
3181 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3182 set_opt(sb, POSIX_ACL);
3183 #endif
3184 set_opt(sb, MBLK_IO_SUBMIT);
3185 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3186 set_opt(sb, JOURNAL_DATA);
3187 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3188 set_opt(sb, ORDERED_DATA);
3189 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3190 set_opt(sb, WRITEBACK_DATA);
3191
3192 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3193 set_opt(sb, ERRORS_PANIC);
3194 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3195 set_opt(sb, ERRORS_CONT);
3196 else
3197 set_opt(sb, ERRORS_RO);
3198 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3199 set_opt(sb, BLOCK_VALIDITY);
3200 if (def_mount_opts & EXT4_DEFM_DISCARD)
3201 set_opt(sb, DISCARD);
3202
3203 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3204 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3205 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3206 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3207 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3208
3209 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3210 set_opt(sb, BARRIER);
3211
3212 /*
3213 * enable delayed allocation by default
3214 * Use -o nodelalloc to turn it off
3215 */
3216 if (!IS_EXT3_SB(sb) &&
3217 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3218 set_opt(sb, DELALLOC);
3219
3220 /*
3221 * set default s_li_wait_mult for lazyinit, for the case there is
3222 * no mount option specified.
3223 */
3224 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3225
3226 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3227 &journal_devnum, &journal_ioprio, NULL, 0)) {
3228 ext4_msg(sb, KERN_WARNING,
3229 "failed to parse options in superblock: %s",
3230 sbi->s_es->s_mount_opts);
3231 }
3232 if (!parse_options((char *) data, sb, &journal_devnum,
3233 &journal_ioprio, NULL, 0))
3234 goto failed_mount;
3235
3236 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3237 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3238 "with data=journal disables delayed "
3239 "allocation and O_DIRECT support!\n");
3240 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3241 ext4_msg(sb, KERN_ERR, "can't mount with "
3242 "both data=journal and delalloc");
3243 goto failed_mount;
3244 }
3245 if (test_opt(sb, DIOREAD_NOLOCK)) {
3246 ext4_msg(sb, KERN_ERR, "can't mount with "
3247 "both data=journal and delalloc");
3248 goto failed_mount;
3249 }
3250 if (test_opt(sb, DELALLOC))
3251 clear_opt(sb, DELALLOC);
3252 }
3253
3254 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3255 if (test_opt(sb, DIOREAD_NOLOCK)) {
3256 if (blocksize < PAGE_SIZE) {
3257 ext4_msg(sb, KERN_ERR, "can't mount with "
3258 "dioread_nolock if block size != PAGE_SIZE");
3259 goto failed_mount;
3260 }
3261 }
3262
3263 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3264 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3265
3266 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3267 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3268 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3269 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3270 ext4_msg(sb, KERN_WARNING,
3271 "feature flags set on rev 0 fs, "
3272 "running e2fsck is recommended");
3273
3274 if (IS_EXT2_SB(sb)) {
3275 if (ext2_feature_set_ok(sb))
3276 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3277 "using the ext4 subsystem");
3278 else {
3279 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3280 "to feature incompatibilities");
3281 goto failed_mount;
3282 }
3283 }
3284
3285 if (IS_EXT3_SB(sb)) {
3286 if (ext3_feature_set_ok(sb))
3287 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3288 "using the ext4 subsystem");
3289 else {
3290 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3291 "to feature incompatibilities");
3292 goto failed_mount;
3293 }
3294 }
3295
3296 /*
3297 * Check feature flags regardless of the revision level, since we
3298 * previously didn't change the revision level when setting the flags,
3299 * so there is a chance incompat flags are set on a rev 0 filesystem.
3300 */
3301 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3302 goto failed_mount;
3303
3304 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3305 blocksize > EXT4_MAX_BLOCK_SIZE) {
3306 ext4_msg(sb, KERN_ERR,
3307 "Unsupported filesystem blocksize %d", blocksize);
3308 goto failed_mount;
3309 }
3310
3311 if (sb->s_blocksize != blocksize) {
3312 /* Validate the filesystem blocksize */
3313 if (!sb_set_blocksize(sb, blocksize)) {
3314 ext4_msg(sb, KERN_ERR, "bad block size %d",
3315 blocksize);
3316 goto failed_mount;
3317 }
3318
3319 brelse(bh);
3320 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3321 offset = do_div(logical_sb_block, blocksize);
3322 bh = sb_bread(sb, logical_sb_block);
3323 if (!bh) {
3324 ext4_msg(sb, KERN_ERR,
3325 "Can't read superblock on 2nd try");
3326 goto failed_mount;
3327 }
3328 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3329 sbi->s_es = es;
3330 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3331 ext4_msg(sb, KERN_ERR,
3332 "Magic mismatch, very weird!");
3333 goto failed_mount;
3334 }
3335 }
3336
3337 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3338 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3339 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3340 has_huge_files);
3341 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3342
3343 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3344 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3345 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3346 } else {
3347 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3348 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3349 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3350 (!is_power_of_2(sbi->s_inode_size)) ||
3351 (sbi->s_inode_size > blocksize)) {
3352 ext4_msg(sb, KERN_ERR,
3353 "unsupported inode size: %d",
3354 sbi->s_inode_size);
3355 goto failed_mount;
3356 }
3357 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3358 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3359 }
3360
3361 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3362 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3363 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3364 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3365 !is_power_of_2(sbi->s_desc_size)) {
3366 ext4_msg(sb, KERN_ERR,
3367 "unsupported descriptor size %lu",
3368 sbi->s_desc_size);
3369 goto failed_mount;
3370 }
3371 } else
3372 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3373
3374 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3375 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3376 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3377 goto cantfind_ext4;
3378
3379 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3380 if (sbi->s_inodes_per_block == 0)
3381 goto cantfind_ext4;
3382 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3383 sbi->s_inodes_per_block;
3384 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3385 sbi->s_sbh = bh;
3386 sbi->s_mount_state = le16_to_cpu(es->s_state);
3387 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3388 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3389
3390 for (i = 0; i < 4; i++)
3391 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3392 sbi->s_def_hash_version = es->s_def_hash_version;
3393 i = le32_to_cpu(es->s_flags);
3394 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3395 sbi->s_hash_unsigned = 3;
3396 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3397 #ifdef __CHAR_UNSIGNED__
3398 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3399 sbi->s_hash_unsigned = 3;
3400 #else
3401 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3402 #endif
3403 sb->s_dirt = 1;
3404 }
3405
3406 /* Handle clustersize */
3407 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3408 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3409 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3410 if (has_bigalloc) {
3411 if (clustersize < blocksize) {
3412 ext4_msg(sb, KERN_ERR,
3413 "cluster size (%d) smaller than "
3414 "block size (%d)", clustersize, blocksize);
3415 goto failed_mount;
3416 }
3417 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3418 le32_to_cpu(es->s_log_block_size);
3419 sbi->s_clusters_per_group =
3420 le32_to_cpu(es->s_clusters_per_group);
3421 if (sbi->s_clusters_per_group > blocksize * 8) {
3422 ext4_msg(sb, KERN_ERR,
3423 "#clusters per group too big: %lu",
3424 sbi->s_clusters_per_group);
3425 goto failed_mount;
3426 }
3427 if (sbi->s_blocks_per_group !=
3428 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3429 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3430 "clusters per group (%lu) inconsistent",
3431 sbi->s_blocks_per_group,
3432 sbi->s_clusters_per_group);
3433 goto failed_mount;
3434 }
3435 } else {
3436 if (clustersize != blocksize) {
3437 ext4_warning(sb, "fragment/cluster size (%d) != "
3438 "block size (%d)", clustersize,
3439 blocksize);
3440 clustersize = blocksize;
3441 }
3442 if (sbi->s_blocks_per_group > blocksize * 8) {
3443 ext4_msg(sb, KERN_ERR,
3444 "#blocks per group too big: %lu",
3445 sbi->s_blocks_per_group);
3446 goto failed_mount;
3447 }
3448 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3449 sbi->s_cluster_bits = 0;
3450 }
3451 sbi->s_cluster_ratio = clustersize / blocksize;
3452
3453 if (sbi->s_inodes_per_group > blocksize * 8) {
3454 ext4_msg(sb, KERN_ERR,
3455 "#inodes per group too big: %lu",
3456 sbi->s_inodes_per_group);
3457 goto failed_mount;
3458 }
3459
3460 /*
3461 * Test whether we have more sectors than will fit in sector_t,
3462 * and whether the max offset is addressable by the page cache.
3463 */
3464 err = generic_check_addressable(sb->s_blocksize_bits,
3465 ext4_blocks_count(es));
3466 if (err) {
3467 ext4_msg(sb, KERN_ERR, "filesystem"
3468 " too large to mount safely on this system");
3469 if (sizeof(sector_t) < 8)
3470 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3471 ret = err;
3472 goto failed_mount;
3473 }
3474
3475 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3476 goto cantfind_ext4;
3477
3478 /* check blocks count against device size */
3479 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3480 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3481 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3482 "exceeds size of device (%llu blocks)",
3483 ext4_blocks_count(es), blocks_count);
3484 goto failed_mount;
3485 }
3486
3487 /*
3488 * It makes no sense for the first data block to be beyond the end
3489 * of the filesystem.
3490 */
3491 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3492 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3493 "block %u is beyond end of filesystem (%llu)",
3494 le32_to_cpu(es->s_first_data_block),
3495 ext4_blocks_count(es));
3496 goto failed_mount;
3497 }
3498 blocks_count = (ext4_blocks_count(es) -
3499 le32_to_cpu(es->s_first_data_block) +
3500 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3501 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3502 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3503 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3504 "(block count %llu, first data block %u, "
3505 "blocks per group %lu)", sbi->s_groups_count,
3506 ext4_blocks_count(es),
3507 le32_to_cpu(es->s_first_data_block),
3508 EXT4_BLOCKS_PER_GROUP(sb));
3509 goto failed_mount;
3510 }
3511 sbi->s_groups_count = blocks_count;
3512 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3513 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3514 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3515 EXT4_DESC_PER_BLOCK(sb);
3516 sbi->s_group_desc = ext4_kvmalloc(db_count *
3517 sizeof(struct buffer_head *),
3518 GFP_KERNEL);
3519 if (sbi->s_group_desc == NULL) {
3520 ext4_msg(sb, KERN_ERR, "not enough memory");
3521 goto failed_mount;
3522 }
3523
3524 if (ext4_proc_root)
3525 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3526
3527 bgl_lock_init(sbi->s_blockgroup_lock);
3528
3529 for (i = 0; i < db_count; i++) {
3530 block = descriptor_loc(sb, logical_sb_block, i);
3531 sbi->s_group_desc[i] = sb_bread(sb, block);
3532 if (!sbi->s_group_desc[i]) {
3533 ext4_msg(sb, KERN_ERR,
3534 "can't read group descriptor %d", i);
3535 db_count = i;
3536 goto failed_mount2;
3537 }
3538 }
3539 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3540 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3541 goto failed_mount2;
3542 }
3543 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3544 if (!ext4_fill_flex_info(sb)) {
3545 ext4_msg(sb, KERN_ERR,
3546 "unable to initialize "
3547 "flex_bg meta info!");
3548 goto failed_mount2;
3549 }
3550
3551 sbi->s_gdb_count = db_count;
3552 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3553 spin_lock_init(&sbi->s_next_gen_lock);
3554
3555 init_timer(&sbi->s_err_report);
3556 sbi->s_err_report.function = print_daily_error_info;
3557 sbi->s_err_report.data = (unsigned long) sb;
3558
3559 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3560 ext4_count_free_clusters(sb));
3561 if (!err) {
3562 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3563 ext4_count_free_inodes(sb));
3564 }
3565 if (!err) {
3566 err = percpu_counter_init(&sbi->s_dirs_counter,
3567 ext4_count_dirs(sb));
3568 }
3569 if (!err) {
3570 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3571 }
3572 if (err) {
3573 ext4_msg(sb, KERN_ERR, "insufficient memory");
3574 goto failed_mount3;
3575 }
3576
3577 sbi->s_stripe = ext4_get_stripe_size(sbi);
3578 sbi->s_max_writeback_mb_bump = 128;
3579
3580 /*
3581 * set up enough so that it can read an inode
3582 */
3583 if (!test_opt(sb, NOLOAD) &&
3584 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3585 sb->s_op = &ext4_sops;
3586 else
3587 sb->s_op = &ext4_nojournal_sops;
3588 sb->s_export_op = &ext4_export_ops;
3589 sb->s_xattr = ext4_xattr_handlers;
3590 #ifdef CONFIG_QUOTA
3591 sb->s_qcop = &ext4_qctl_operations;
3592 sb->dq_op = &ext4_quota_operations;
3593 #endif
3594 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3595
3596 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3597 mutex_init(&sbi->s_orphan_lock);
3598 sbi->s_resize_flags = 0;
3599
3600 sb->s_root = NULL;
3601
3602 needs_recovery = (es->s_last_orphan != 0 ||
3603 EXT4_HAS_INCOMPAT_FEATURE(sb,
3604 EXT4_FEATURE_INCOMPAT_RECOVER));
3605
3606 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3607 !(sb->s_flags & MS_RDONLY))
3608 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3609 goto failed_mount3;
3610
3611 /*
3612 * The first inode we look at is the journal inode. Don't try
3613 * root first: it may be modified in the journal!
3614 */
3615 if (!test_opt(sb, NOLOAD) &&
3616 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3617 if (ext4_load_journal(sb, es, journal_devnum))
3618 goto failed_mount3;
3619 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3620 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3621 ext4_msg(sb, KERN_ERR, "required journal recovery "
3622 "suppressed and not mounted read-only");
3623 goto failed_mount_wq;
3624 } else {
3625 clear_opt(sb, DATA_FLAGS);
3626 sbi->s_journal = NULL;
3627 needs_recovery = 0;
3628 goto no_journal;
3629 }
3630
3631 if (ext4_blocks_count(es) > 0xffffffffULL &&
3632 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3633 JBD2_FEATURE_INCOMPAT_64BIT)) {
3634 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3635 goto failed_mount_wq;
3636 }
3637
3638 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3639 jbd2_journal_set_features(sbi->s_journal,
3640 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3641 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3642 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3643 jbd2_journal_set_features(sbi->s_journal,
3644 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3645 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3646 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3647 } else {
3648 jbd2_journal_clear_features(sbi->s_journal,
3649 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3650 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3651 }
3652
3653 /* We have now updated the journal if required, so we can
3654 * validate the data journaling mode. */
3655 switch (test_opt(sb, DATA_FLAGS)) {
3656 case 0:
3657 /* No mode set, assume a default based on the journal
3658 * capabilities: ORDERED_DATA if the journal can
3659 * cope, else JOURNAL_DATA
3660 */
3661 if (jbd2_journal_check_available_features
3662 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3663 set_opt(sb, ORDERED_DATA);
3664 else
3665 set_opt(sb, JOURNAL_DATA);
3666 break;
3667
3668 case EXT4_MOUNT_ORDERED_DATA:
3669 case EXT4_MOUNT_WRITEBACK_DATA:
3670 if (!jbd2_journal_check_available_features
3671 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3672 ext4_msg(sb, KERN_ERR, "Journal does not support "
3673 "requested data journaling mode");
3674 goto failed_mount_wq;
3675 }
3676 default:
3677 break;
3678 }
3679 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3680
3681 /*
3682 * The journal may have updated the bg summary counts, so we
3683 * need to update the global counters.
3684 */
3685 percpu_counter_set(&sbi->s_freeclusters_counter,
3686 ext4_count_free_clusters(sb));
3687 percpu_counter_set(&sbi->s_freeinodes_counter,
3688 ext4_count_free_inodes(sb));
3689 percpu_counter_set(&sbi->s_dirs_counter,
3690 ext4_count_dirs(sb));
3691 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3692
3693 no_journal:
3694 /*
3695 * The maximum number of concurrent works can be high and
3696 * concurrency isn't really necessary. Limit it to 1.
3697 */
3698 EXT4_SB(sb)->dio_unwritten_wq =
3699 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3700 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3701 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3702 goto failed_mount_wq;
3703 }
3704
3705 /*
3706 * The jbd2_journal_load will have done any necessary log recovery,
3707 * so we can safely mount the rest of the filesystem now.
3708 */
3709
3710 root = ext4_iget(sb, EXT4_ROOT_INO);
3711 if (IS_ERR(root)) {
3712 ext4_msg(sb, KERN_ERR, "get root inode failed");
3713 ret = PTR_ERR(root);
3714 root = NULL;
3715 goto failed_mount4;
3716 }
3717 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3718 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3719 iput(root);
3720 goto failed_mount4;
3721 }
3722 sb->s_root = d_alloc_root(root);
3723 if (!sb->s_root) {
3724 iput(root);
3725 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3726 ret = -ENOMEM;
3727 goto failed_mount4;
3728 }
3729
3730 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3731
3732 /* determine the minimum size of new large inodes, if present */
3733 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3734 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3735 EXT4_GOOD_OLD_INODE_SIZE;
3736 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3737 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3738 if (sbi->s_want_extra_isize <
3739 le16_to_cpu(es->s_want_extra_isize))
3740 sbi->s_want_extra_isize =
3741 le16_to_cpu(es->s_want_extra_isize);
3742 if (sbi->s_want_extra_isize <
3743 le16_to_cpu(es->s_min_extra_isize))
3744 sbi->s_want_extra_isize =
3745 le16_to_cpu(es->s_min_extra_isize);
3746 }
3747 }
3748 /* Check if enough inode space is available */
3749 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3750 sbi->s_inode_size) {
3751 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3752 EXT4_GOOD_OLD_INODE_SIZE;
3753 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3754 "available");
3755 }
3756
3757 err = ext4_setup_system_zone(sb);
3758 if (err) {
3759 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3760 "zone (%d)", err);
3761 goto failed_mount4a;
3762 }
3763
3764 ext4_ext_init(sb);
3765 err = ext4_mb_init(sb, needs_recovery);
3766 if (err) {
3767 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3768 err);
3769 goto failed_mount5;
3770 }
3771
3772 err = ext4_register_li_request(sb, first_not_zeroed);
3773 if (err)
3774 goto failed_mount6;
3775
3776 sbi->s_kobj.kset = ext4_kset;
3777 init_completion(&sbi->s_kobj_unregister);
3778 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3779 "%s", sb->s_id);
3780 if (err)
3781 goto failed_mount7;
3782
3783 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3784 ext4_orphan_cleanup(sb, es);
3785 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3786 if (needs_recovery) {
3787 ext4_msg(sb, KERN_INFO, "recovery complete");
3788 ext4_mark_recovery_complete(sb, es);
3789 }
3790 if (EXT4_SB(sb)->s_journal) {
3791 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3792 descr = " journalled data mode";
3793 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3794 descr = " ordered data mode";
3795 else
3796 descr = " writeback data mode";
3797 } else
3798 descr = "out journal";
3799
3800 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3801 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3802 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3803
3804 if (es->s_error_count)
3805 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3806
3807 kfree(orig_data);
3808 return 0;
3809
3810 cantfind_ext4:
3811 if (!silent)
3812 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3813 goto failed_mount;
3814
3815 failed_mount7:
3816 ext4_unregister_li_request(sb);
3817 failed_mount6:
3818 ext4_mb_release(sb);
3819 failed_mount5:
3820 ext4_ext_release(sb);
3821 ext4_release_system_zone(sb);
3822 failed_mount4a:
3823 dput(sb->s_root);
3824 sb->s_root = NULL;
3825 failed_mount4:
3826 ext4_msg(sb, KERN_ERR, "mount failed");
3827 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3828 failed_mount_wq:
3829 if (sbi->s_journal) {
3830 jbd2_journal_destroy(sbi->s_journal);
3831 sbi->s_journal = NULL;
3832 }
3833 failed_mount3:
3834 del_timer(&sbi->s_err_report);
3835 if (sbi->s_flex_groups)
3836 ext4_kvfree(sbi->s_flex_groups);
3837 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3838 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3839 percpu_counter_destroy(&sbi->s_dirs_counter);
3840 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3841 if (sbi->s_mmp_tsk)
3842 kthread_stop(sbi->s_mmp_tsk);
3843 failed_mount2:
3844 for (i = 0; i < db_count; i++)
3845 brelse(sbi->s_group_desc[i]);
3846 ext4_kvfree(sbi->s_group_desc);
3847 failed_mount:
3848 if (sbi->s_proc) {
3849 remove_proc_entry(sb->s_id, ext4_proc_root);
3850 }
3851 #ifdef CONFIG_QUOTA
3852 for (i = 0; i < MAXQUOTAS; i++)
3853 kfree(sbi->s_qf_names[i]);
3854 #endif
3855 ext4_blkdev_remove(sbi);
3856 brelse(bh);
3857 out_fail:
3858 sb->s_fs_info = NULL;
3859 kfree(sbi->s_blockgroup_lock);
3860 kfree(sbi);
3861 out_free_orig:
3862 kfree(orig_data);
3863 return ret;
3864 }
3865
3866 /*
3867 * Setup any per-fs journal parameters now. We'll do this both on
3868 * initial mount, once the journal has been initialised but before we've
3869 * done any recovery; and again on any subsequent remount.
3870 */
3871 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3872 {
3873 struct ext4_sb_info *sbi = EXT4_SB(sb);
3874
3875 journal->j_commit_interval = sbi->s_commit_interval;
3876 journal->j_min_batch_time = sbi->s_min_batch_time;
3877 journal->j_max_batch_time = sbi->s_max_batch_time;
3878
3879 write_lock(&journal->j_state_lock);
3880 if (test_opt(sb, BARRIER))
3881 journal->j_flags |= JBD2_BARRIER;
3882 else
3883 journal->j_flags &= ~JBD2_BARRIER;
3884 if (test_opt(sb, DATA_ERR_ABORT))
3885 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3886 else
3887 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3888 write_unlock(&journal->j_state_lock);
3889 }
3890
3891 static journal_t *ext4_get_journal(struct super_block *sb,
3892 unsigned int journal_inum)
3893 {
3894 struct inode *journal_inode;
3895 journal_t *journal;
3896
3897 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3898
3899 /* First, test for the existence of a valid inode on disk. Bad
3900 * things happen if we iget() an unused inode, as the subsequent
3901 * iput() will try to delete it. */
3902
3903 journal_inode = ext4_iget(sb, journal_inum);
3904 if (IS_ERR(journal_inode)) {
3905 ext4_msg(sb, KERN_ERR, "no journal found");
3906 return NULL;
3907 }
3908 if (!journal_inode->i_nlink) {
3909 make_bad_inode(journal_inode);
3910 iput(journal_inode);
3911 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3912 return NULL;
3913 }
3914
3915 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3916 journal_inode, journal_inode->i_size);
3917 if (!S_ISREG(journal_inode->i_mode)) {
3918 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3919 iput(journal_inode);
3920 return NULL;
3921 }
3922
3923 journal = jbd2_journal_init_inode(journal_inode);
3924 if (!journal) {
3925 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3926 iput(journal_inode);
3927 return NULL;
3928 }
3929 journal->j_private = sb;
3930 ext4_init_journal_params(sb, journal);
3931 return journal;
3932 }
3933
3934 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3935 dev_t j_dev)
3936 {
3937 struct buffer_head *bh;
3938 journal_t *journal;
3939 ext4_fsblk_t start;
3940 ext4_fsblk_t len;
3941 int hblock, blocksize;
3942 ext4_fsblk_t sb_block;
3943 unsigned long offset;
3944 struct ext4_super_block *es;
3945 struct block_device *bdev;
3946
3947 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3948
3949 bdev = ext4_blkdev_get(j_dev, sb);
3950 if (bdev == NULL)
3951 return NULL;
3952
3953 blocksize = sb->s_blocksize;
3954 hblock = bdev_logical_block_size(bdev);
3955 if (blocksize < hblock) {
3956 ext4_msg(sb, KERN_ERR,
3957 "blocksize too small for journal device");
3958 goto out_bdev;
3959 }
3960
3961 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3962 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3963 set_blocksize(bdev, blocksize);
3964 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3965 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3966 "external journal");
3967 goto out_bdev;
3968 }
3969
3970 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3971 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3972 !(le32_to_cpu(es->s_feature_incompat) &
3973 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3974 ext4_msg(sb, KERN_ERR, "external journal has "
3975 "bad superblock");
3976 brelse(bh);
3977 goto out_bdev;
3978 }
3979
3980 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3981 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3982 brelse(bh);
3983 goto out_bdev;
3984 }
3985
3986 len = ext4_blocks_count(es);
3987 start = sb_block + 1;
3988 brelse(bh); /* we're done with the superblock */
3989
3990 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3991 start, len, blocksize);
3992 if (!journal) {
3993 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3994 goto out_bdev;
3995 }
3996 journal->j_private = sb;
3997 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3998 wait_on_buffer(journal->j_sb_buffer);
3999 if (!buffer_uptodate(journal->j_sb_buffer)) {
4000 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4001 goto out_journal;
4002 }
4003 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4004 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4005 "user (unsupported) - %d",
4006 be32_to_cpu(journal->j_superblock->s_nr_users));
4007 goto out_journal;
4008 }
4009 EXT4_SB(sb)->journal_bdev = bdev;
4010 ext4_init_journal_params(sb, journal);
4011 return journal;
4012
4013 out_journal:
4014 jbd2_journal_destroy(journal);
4015 out_bdev:
4016 ext4_blkdev_put(bdev);
4017 return NULL;
4018 }
4019
4020 static int ext4_load_journal(struct super_block *sb,
4021 struct ext4_super_block *es,
4022 unsigned long journal_devnum)
4023 {
4024 journal_t *journal;
4025 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4026 dev_t journal_dev;
4027 int err = 0;
4028 int really_read_only;
4029
4030 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4031
4032 if (journal_devnum &&
4033 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4034 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4035 "numbers have changed");
4036 journal_dev = new_decode_dev(journal_devnum);
4037 } else
4038 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4039
4040 really_read_only = bdev_read_only(sb->s_bdev);
4041
4042 /*
4043 * Are we loading a blank journal or performing recovery after a
4044 * crash? For recovery, we need to check in advance whether we
4045 * can get read-write access to the device.
4046 */
4047 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4048 if (sb->s_flags & MS_RDONLY) {
4049 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4050 "required on readonly filesystem");
4051 if (really_read_only) {
4052 ext4_msg(sb, KERN_ERR, "write access "
4053 "unavailable, cannot proceed");
4054 return -EROFS;
4055 }
4056 ext4_msg(sb, KERN_INFO, "write access will "
4057 "be enabled during recovery");
4058 }
4059 }
4060
4061 if (journal_inum && journal_dev) {
4062 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4063 "and inode journals!");
4064 return -EINVAL;
4065 }
4066
4067 if (journal_inum) {
4068 if (!(journal = ext4_get_journal(sb, journal_inum)))
4069 return -EINVAL;
4070 } else {
4071 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4072 return -EINVAL;
4073 }
4074
4075 if (!(journal->j_flags & JBD2_BARRIER))
4076 ext4_msg(sb, KERN_INFO, "barriers disabled");
4077
4078 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
4079 err = jbd2_journal_update_format(journal);
4080 if (err) {
4081 ext4_msg(sb, KERN_ERR, "error updating journal");
4082 jbd2_journal_destroy(journal);
4083 return err;
4084 }
4085 }
4086
4087 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4088 err = jbd2_journal_wipe(journal, !really_read_only);
4089 if (!err) {
4090 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4091 if (save)
4092 memcpy(save, ((char *) es) +
4093 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4094 err = jbd2_journal_load(journal);
4095 if (save)
4096 memcpy(((char *) es) + EXT4_S_ERR_START,
4097 save, EXT4_S_ERR_LEN);
4098 kfree(save);
4099 }
4100
4101 if (err) {
4102 ext4_msg(sb, KERN_ERR, "error loading journal");
4103 jbd2_journal_destroy(journal);
4104 return err;
4105 }
4106
4107 EXT4_SB(sb)->s_journal = journal;
4108 ext4_clear_journal_err(sb, es);
4109
4110 if (!really_read_only && journal_devnum &&
4111 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4112 es->s_journal_dev = cpu_to_le32(journal_devnum);
4113
4114 /* Make sure we flush the recovery flag to disk. */
4115 ext4_commit_super(sb, 1);
4116 }
4117
4118 return 0;
4119 }
4120
4121 static int ext4_commit_super(struct super_block *sb, int sync)
4122 {
4123 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4124 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4125 int error = 0;
4126
4127 if (!sbh || block_device_ejected(sb))
4128 return error;
4129 if (buffer_write_io_error(sbh)) {
4130 /*
4131 * Oh, dear. A previous attempt to write the
4132 * superblock failed. This could happen because the
4133 * USB device was yanked out. Or it could happen to
4134 * be a transient write error and maybe the block will
4135 * be remapped. Nothing we can do but to retry the
4136 * write and hope for the best.
4137 */
4138 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4139 "superblock detected");
4140 clear_buffer_write_io_error(sbh);
4141 set_buffer_uptodate(sbh);
4142 }
4143 /*
4144 * If the file system is mounted read-only, don't update the
4145 * superblock write time. This avoids updating the superblock
4146 * write time when we are mounting the root file system
4147 * read/only but we need to replay the journal; at that point,
4148 * for people who are east of GMT and who make their clock
4149 * tick in localtime for Windows bug-for-bug compatibility,
4150 * the clock is set in the future, and this will cause e2fsck
4151 * to complain and force a full file system check.
4152 */
4153 if (!(sb->s_flags & MS_RDONLY))
4154 es->s_wtime = cpu_to_le32(get_seconds());
4155 if (sb->s_bdev->bd_part)
4156 es->s_kbytes_written =
4157 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4158 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4159 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4160 else
4161 es->s_kbytes_written =
4162 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4163 ext4_free_blocks_count_set(es,
4164 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4165 &EXT4_SB(sb)->s_freeclusters_counter)));
4166 es->s_free_inodes_count =
4167 cpu_to_le32(percpu_counter_sum_positive(
4168 &EXT4_SB(sb)->s_freeinodes_counter));
4169 sb->s_dirt = 0;
4170 BUFFER_TRACE(sbh, "marking dirty");
4171 mark_buffer_dirty(sbh);
4172 if (sync) {
4173 error = sync_dirty_buffer(sbh);
4174 if (error)
4175 return error;
4176
4177 error = buffer_write_io_error(sbh);
4178 if (error) {
4179 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4180 "superblock");
4181 clear_buffer_write_io_error(sbh);
4182 set_buffer_uptodate(sbh);
4183 }
4184 }
4185 return error;
4186 }
4187
4188 /*
4189 * Have we just finished recovery? If so, and if we are mounting (or
4190 * remounting) the filesystem readonly, then we will end up with a
4191 * consistent fs on disk. Record that fact.
4192 */
4193 static void ext4_mark_recovery_complete(struct super_block *sb,
4194 struct ext4_super_block *es)
4195 {
4196 journal_t *journal = EXT4_SB(sb)->s_journal;
4197
4198 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4199 BUG_ON(journal != NULL);
4200 return;
4201 }
4202 jbd2_journal_lock_updates(journal);
4203 if (jbd2_journal_flush(journal) < 0)
4204 goto out;
4205
4206 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4207 sb->s_flags & MS_RDONLY) {
4208 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4209 ext4_commit_super(sb, 1);
4210 }
4211
4212 out:
4213 jbd2_journal_unlock_updates(journal);
4214 }
4215
4216 /*
4217 * If we are mounting (or read-write remounting) a filesystem whose journal
4218 * has recorded an error from a previous lifetime, move that error to the
4219 * main filesystem now.
4220 */
4221 static void ext4_clear_journal_err(struct super_block *sb,
4222 struct ext4_super_block *es)
4223 {
4224 journal_t *journal;
4225 int j_errno;
4226 const char *errstr;
4227
4228 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4229
4230 journal = EXT4_SB(sb)->s_journal;
4231
4232 /*
4233 * Now check for any error status which may have been recorded in the
4234 * journal by a prior ext4_error() or ext4_abort()
4235 */
4236
4237 j_errno = jbd2_journal_errno(journal);
4238 if (j_errno) {
4239 char nbuf[16];
4240
4241 errstr = ext4_decode_error(sb, j_errno, nbuf);
4242 ext4_warning(sb, "Filesystem error recorded "
4243 "from previous mount: %s", errstr);
4244 ext4_warning(sb, "Marking fs in need of filesystem check.");
4245
4246 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4247 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4248 ext4_commit_super(sb, 1);
4249
4250 jbd2_journal_clear_err(journal);
4251 }
4252 }
4253
4254 /*
4255 * Force the running and committing transactions to commit,
4256 * and wait on the commit.
4257 */
4258 int ext4_force_commit(struct super_block *sb)
4259 {
4260 journal_t *journal;
4261 int ret = 0;
4262
4263 if (sb->s_flags & MS_RDONLY)
4264 return 0;
4265
4266 journal = EXT4_SB(sb)->s_journal;
4267 if (journal) {
4268 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4269 ret = ext4_journal_force_commit(journal);
4270 }
4271
4272 return ret;
4273 }
4274
4275 static void ext4_write_super(struct super_block *sb)
4276 {
4277 lock_super(sb);
4278 ext4_commit_super(sb, 1);
4279 unlock_super(sb);
4280 }
4281
4282 static int ext4_sync_fs(struct super_block *sb, int wait)
4283 {
4284 int ret = 0;
4285 tid_t target;
4286 struct ext4_sb_info *sbi = EXT4_SB(sb);
4287
4288 trace_ext4_sync_fs(sb, wait);
4289 flush_workqueue(sbi->dio_unwritten_wq);
4290 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4291 if (wait)
4292 jbd2_log_wait_commit(sbi->s_journal, target);
4293 }
4294 return ret;
4295 }
4296
4297 /*
4298 * LVM calls this function before a (read-only) snapshot is created. This
4299 * gives us a chance to flush the journal completely and mark the fs clean.
4300 *
4301 * Note that only this function cannot bring a filesystem to be in a clean
4302 * state independently, because ext4 prevents a new handle from being started
4303 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4304 * the upper layer.
4305 */
4306 static int ext4_freeze(struct super_block *sb)
4307 {
4308 int error = 0;
4309 journal_t *journal;
4310
4311 if (sb->s_flags & MS_RDONLY)
4312 return 0;
4313
4314 journal = EXT4_SB(sb)->s_journal;
4315
4316 /* Now we set up the journal barrier. */
4317 jbd2_journal_lock_updates(journal);
4318
4319 /*
4320 * Don't clear the needs_recovery flag if we failed to flush
4321 * the journal.
4322 */
4323 error = jbd2_journal_flush(journal);
4324 if (error < 0)
4325 goto out;
4326
4327 /* Journal blocked and flushed, clear needs_recovery flag. */
4328 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4329 error = ext4_commit_super(sb, 1);
4330 out:
4331 /* we rely on s_frozen to stop further updates */
4332 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4333 return error;
4334 }
4335
4336 /*
4337 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4338 * flag here, even though the filesystem is not technically dirty yet.
4339 */
4340 static int ext4_unfreeze(struct super_block *sb)
4341 {
4342 if (sb->s_flags & MS_RDONLY)
4343 return 0;
4344
4345 lock_super(sb);
4346 /* Reset the needs_recovery flag before the fs is unlocked. */
4347 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4348 ext4_commit_super(sb, 1);
4349 unlock_super(sb);
4350 return 0;
4351 }
4352
4353 /*
4354 * Structure to save mount options for ext4_remount's benefit
4355 */
4356 struct ext4_mount_options {
4357 unsigned long s_mount_opt;
4358 unsigned long s_mount_opt2;
4359 uid_t s_resuid;
4360 gid_t s_resgid;
4361 unsigned long s_commit_interval;
4362 u32 s_min_batch_time, s_max_batch_time;
4363 #ifdef CONFIG_QUOTA
4364 int s_jquota_fmt;
4365 char *s_qf_names[MAXQUOTAS];
4366 #endif
4367 };
4368
4369 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4370 {
4371 struct ext4_super_block *es;
4372 struct ext4_sb_info *sbi = EXT4_SB(sb);
4373 ext4_fsblk_t n_blocks_count = 0;
4374 unsigned long old_sb_flags;
4375 struct ext4_mount_options old_opts;
4376 int enable_quota = 0;
4377 ext4_group_t g;
4378 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4379 int err = 0;
4380 #ifdef CONFIG_QUOTA
4381 int i;
4382 #endif
4383 char *orig_data = kstrdup(data, GFP_KERNEL);
4384
4385 /* Store the original options */
4386 lock_super(sb);
4387 old_sb_flags = sb->s_flags;
4388 old_opts.s_mount_opt = sbi->s_mount_opt;
4389 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4390 old_opts.s_resuid = sbi->s_resuid;
4391 old_opts.s_resgid = sbi->s_resgid;
4392 old_opts.s_commit_interval = sbi->s_commit_interval;
4393 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4394 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4395 #ifdef CONFIG_QUOTA
4396 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4397 for (i = 0; i < MAXQUOTAS; i++)
4398 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4399 #endif
4400 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4401 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4402
4403 /*
4404 * Allow the "check" option to be passed as a remount option.
4405 */
4406 if (!parse_options(data, sb, NULL, &journal_ioprio,
4407 &n_blocks_count, 1)) {
4408 err = -EINVAL;
4409 goto restore_opts;
4410 }
4411
4412 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4413 ext4_abort(sb, "Abort forced by user");
4414
4415 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4416 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4417
4418 es = sbi->s_es;
4419
4420 if (sbi->s_journal) {
4421 ext4_init_journal_params(sb, sbi->s_journal);
4422 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4423 }
4424
4425 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4426 n_blocks_count > ext4_blocks_count(es)) {
4427 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4428 err = -EROFS;
4429 goto restore_opts;
4430 }
4431
4432 if (*flags & MS_RDONLY) {
4433 err = dquot_suspend(sb, -1);
4434 if (err < 0)
4435 goto restore_opts;
4436
4437 /*
4438 * First of all, the unconditional stuff we have to do
4439 * to disable replay of the journal when we next remount
4440 */
4441 sb->s_flags |= MS_RDONLY;
4442
4443 /*
4444 * OK, test if we are remounting a valid rw partition
4445 * readonly, and if so set the rdonly flag and then
4446 * mark the partition as valid again.
4447 */
4448 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4449 (sbi->s_mount_state & EXT4_VALID_FS))
4450 es->s_state = cpu_to_le16(sbi->s_mount_state);
4451
4452 if (sbi->s_journal)
4453 ext4_mark_recovery_complete(sb, es);
4454 } else {
4455 /* Make sure we can mount this feature set readwrite */
4456 if (!ext4_feature_set_ok(sb, 0)) {
4457 err = -EROFS;
4458 goto restore_opts;
4459 }
4460 /*
4461 * Make sure the group descriptor checksums
4462 * are sane. If they aren't, refuse to remount r/w.
4463 */
4464 for (g = 0; g < sbi->s_groups_count; g++) {
4465 struct ext4_group_desc *gdp =
4466 ext4_get_group_desc(sb, g, NULL);
4467
4468 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4469 ext4_msg(sb, KERN_ERR,
4470 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4471 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4472 le16_to_cpu(gdp->bg_checksum));
4473 err = -EINVAL;
4474 goto restore_opts;
4475 }
4476 }
4477
4478 /*
4479 * If we have an unprocessed orphan list hanging
4480 * around from a previously readonly bdev mount,
4481 * require a full umount/remount for now.
4482 */
4483 if (es->s_last_orphan) {
4484 ext4_msg(sb, KERN_WARNING, "Couldn't "
4485 "remount RDWR because of unprocessed "
4486 "orphan inode list. Please "
4487 "umount/remount instead");
4488 err = -EINVAL;
4489 goto restore_opts;
4490 }
4491
4492 /*
4493 * Mounting a RDONLY partition read-write, so reread
4494 * and store the current valid flag. (It may have
4495 * been changed by e2fsck since we originally mounted
4496 * the partition.)
4497 */
4498 if (sbi->s_journal)
4499 ext4_clear_journal_err(sb, es);
4500 sbi->s_mount_state = le16_to_cpu(es->s_state);
4501 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4502 goto restore_opts;
4503 if (!ext4_setup_super(sb, es, 0))
4504 sb->s_flags &= ~MS_RDONLY;
4505 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4506 EXT4_FEATURE_INCOMPAT_MMP))
4507 if (ext4_multi_mount_protect(sb,
4508 le64_to_cpu(es->s_mmp_block))) {
4509 err = -EROFS;
4510 goto restore_opts;
4511 }
4512 enable_quota = 1;
4513 }
4514 }
4515
4516 /*
4517 * Reinitialize lazy itable initialization thread based on
4518 * current settings
4519 */
4520 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4521 ext4_unregister_li_request(sb);
4522 else {
4523 ext4_group_t first_not_zeroed;
4524 first_not_zeroed = ext4_has_uninit_itable(sb);
4525 ext4_register_li_request(sb, first_not_zeroed);
4526 }
4527
4528 ext4_setup_system_zone(sb);
4529 if (sbi->s_journal == NULL)
4530 ext4_commit_super(sb, 1);
4531
4532 #ifdef CONFIG_QUOTA
4533 /* Release old quota file names */
4534 for (i = 0; i < MAXQUOTAS; i++)
4535 if (old_opts.s_qf_names[i] &&
4536 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4537 kfree(old_opts.s_qf_names[i]);
4538 #endif
4539 unlock_super(sb);
4540 if (enable_quota)
4541 dquot_resume(sb, -1);
4542
4543 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4544 kfree(orig_data);
4545 return 0;
4546
4547 restore_opts:
4548 sb->s_flags = old_sb_flags;
4549 sbi->s_mount_opt = old_opts.s_mount_opt;
4550 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4551 sbi->s_resuid = old_opts.s_resuid;
4552 sbi->s_resgid = old_opts.s_resgid;
4553 sbi->s_commit_interval = old_opts.s_commit_interval;
4554 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4555 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4556 #ifdef CONFIG_QUOTA
4557 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4558 for (i = 0; i < MAXQUOTAS; i++) {
4559 if (sbi->s_qf_names[i] &&
4560 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4561 kfree(sbi->s_qf_names[i]);
4562 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4563 }
4564 #endif
4565 unlock_super(sb);
4566 kfree(orig_data);
4567 return err;
4568 }
4569
4570 /*
4571 * Note: calculating the overhead so we can be compatible with
4572 * historical BSD practice is quite difficult in the face of
4573 * clusters/bigalloc. This is because multiple metadata blocks from
4574 * different block group can end up in the same allocation cluster.
4575 * Calculating the exact overhead in the face of clustered allocation
4576 * requires either O(all block bitmaps) in memory or O(number of block
4577 * groups**2) in time. We will still calculate the superblock for
4578 * older file systems --- and if we come across with a bigalloc file
4579 * system with zero in s_overhead_clusters the estimate will be close to
4580 * correct especially for very large cluster sizes --- but for newer
4581 * file systems, it's better to calculate this figure once at mkfs
4582 * time, and store it in the superblock. If the superblock value is
4583 * present (even for non-bigalloc file systems), we will use it.
4584 */
4585 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4586 {
4587 struct super_block *sb = dentry->d_sb;
4588 struct ext4_sb_info *sbi = EXT4_SB(sb);
4589 struct ext4_super_block *es = sbi->s_es;
4590 struct ext4_group_desc *gdp;
4591 u64 fsid;
4592 s64 bfree;
4593
4594 if (test_opt(sb, MINIX_DF)) {
4595 sbi->s_overhead_last = 0;
4596 } else if (es->s_overhead_clusters) {
4597 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4598 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4599 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4600 ext4_fsblk_t overhead = 0;
4601
4602 /*
4603 * Compute the overhead (FS structures). This is constant
4604 * for a given filesystem unless the number of block groups
4605 * changes so we cache the previous value until it does.
4606 */
4607
4608 /*
4609 * All of the blocks before first_data_block are
4610 * overhead
4611 */
4612 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4613
4614 /*
4615 * Add the overhead found in each block group
4616 */
4617 for (i = 0; i < ngroups; i++) {
4618 gdp = ext4_get_group_desc(sb, i, NULL);
4619 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4620 cond_resched();
4621 }
4622 sbi->s_overhead_last = overhead;
4623 smp_wmb();
4624 sbi->s_blocks_last = ext4_blocks_count(es);
4625 }
4626
4627 buf->f_type = EXT4_SUPER_MAGIC;
4628 buf->f_bsize = sb->s_blocksize;
4629 buf->f_blocks = (ext4_blocks_count(es) -
4630 EXT4_C2B(sbi, sbi->s_overhead_last));
4631 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4632 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4633 /* prevent underflow in case that few free space is available */
4634 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4635 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4636 if (buf->f_bfree < ext4_r_blocks_count(es))
4637 buf->f_bavail = 0;
4638 buf->f_files = le32_to_cpu(es->s_inodes_count);
4639 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4640 buf->f_namelen = EXT4_NAME_LEN;
4641 fsid = le64_to_cpup((void *)es->s_uuid) ^
4642 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4643 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4644 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4645
4646 return 0;
4647 }
4648
4649 /* Helper function for writing quotas on sync - we need to start transaction
4650 * before quota file is locked for write. Otherwise the are possible deadlocks:
4651 * Process 1 Process 2
4652 * ext4_create() quota_sync()
4653 * jbd2_journal_start() write_dquot()
4654 * dquot_initialize() down(dqio_mutex)
4655 * down(dqio_mutex) jbd2_journal_start()
4656 *
4657 */
4658
4659 #ifdef CONFIG_QUOTA
4660
4661 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4662 {
4663 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4664 }
4665
4666 static int ext4_write_dquot(struct dquot *dquot)
4667 {
4668 int ret, err;
4669 handle_t *handle;
4670 struct inode *inode;
4671
4672 inode = dquot_to_inode(dquot);
4673 handle = ext4_journal_start(inode,
4674 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4675 if (IS_ERR(handle))
4676 return PTR_ERR(handle);
4677 ret = dquot_commit(dquot);
4678 err = ext4_journal_stop(handle);
4679 if (!ret)
4680 ret = err;
4681 return ret;
4682 }
4683
4684 static int ext4_acquire_dquot(struct dquot *dquot)
4685 {
4686 int ret, err;
4687 handle_t *handle;
4688
4689 handle = ext4_journal_start(dquot_to_inode(dquot),
4690 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4691 if (IS_ERR(handle))
4692 return PTR_ERR(handle);
4693 ret = dquot_acquire(dquot);
4694 err = ext4_journal_stop(handle);
4695 if (!ret)
4696 ret = err;
4697 return ret;
4698 }
4699
4700 static int ext4_release_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_DEL_BLOCKS(dquot->dq_sb));
4707 if (IS_ERR(handle)) {
4708 /* Release dquot anyway to avoid endless cycle in dqput() */
4709 dquot_release(dquot);
4710 return PTR_ERR(handle);
4711 }
4712 ret = dquot_release(dquot);
4713 err = ext4_journal_stop(handle);
4714 if (!ret)
4715 ret = err;
4716 return ret;
4717 }
4718
4719 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4720 {
4721 /* Are we journaling quotas? */
4722 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4723 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4724 dquot_mark_dquot_dirty(dquot);
4725 return ext4_write_dquot(dquot);
4726 } else {
4727 return dquot_mark_dquot_dirty(dquot);
4728 }
4729 }
4730
4731 static int ext4_write_info(struct super_block *sb, int type)
4732 {
4733 int ret, err;
4734 handle_t *handle;
4735
4736 /* Data block + inode block */
4737 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4738 if (IS_ERR(handle))
4739 return PTR_ERR(handle);
4740 ret = dquot_commit_info(sb, type);
4741 err = ext4_journal_stop(handle);
4742 if (!ret)
4743 ret = err;
4744 return ret;
4745 }
4746
4747 /*
4748 * Turn on quotas during mount time - we need to find
4749 * the quota file and such...
4750 */
4751 static int ext4_quota_on_mount(struct super_block *sb, int type)
4752 {
4753 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4754 EXT4_SB(sb)->s_jquota_fmt, type);
4755 }
4756
4757 /*
4758 * Standard function to be called on quota_on
4759 */
4760 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4761 struct path *path)
4762 {
4763 int err;
4764
4765 if (!test_opt(sb, QUOTA))
4766 return -EINVAL;
4767
4768 /* Quotafile not on the same filesystem? */
4769 if (path->dentry->d_sb != sb)
4770 return -EXDEV;
4771 /* Journaling quota? */
4772 if (EXT4_SB(sb)->s_qf_names[type]) {
4773 /* Quotafile not in fs root? */
4774 if (path->dentry->d_parent != sb->s_root)
4775 ext4_msg(sb, KERN_WARNING,
4776 "Quota file not on filesystem root. "
4777 "Journaled quota will not work");
4778 }
4779
4780 /*
4781 * When we journal data on quota file, we have to flush journal to see
4782 * all updates to the file when we bypass pagecache...
4783 */
4784 if (EXT4_SB(sb)->s_journal &&
4785 ext4_should_journal_data(path->dentry->d_inode)) {
4786 /*
4787 * We don't need to lock updates but journal_flush() could
4788 * otherwise be livelocked...
4789 */
4790 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4791 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4792 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4793 if (err)
4794 return err;
4795 }
4796
4797 return dquot_quota_on(sb, type, format_id, path);
4798 }
4799
4800 static int ext4_quota_off(struct super_block *sb, int type)
4801 {
4802 struct inode *inode = sb_dqopt(sb)->files[type];
4803 handle_t *handle;
4804
4805 /* Force all delayed allocation blocks to be allocated.
4806 * Caller already holds s_umount sem */
4807 if (test_opt(sb, DELALLOC))
4808 sync_filesystem(sb);
4809
4810 if (!inode)
4811 goto out;
4812
4813 /* Update modification times of quota files when userspace can
4814 * start looking at them */
4815 handle = ext4_journal_start(inode, 1);
4816 if (IS_ERR(handle))
4817 goto out;
4818 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4819 ext4_mark_inode_dirty(handle, inode);
4820 ext4_journal_stop(handle);
4821
4822 out:
4823 return dquot_quota_off(sb, type);
4824 }
4825
4826 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4827 * acquiring the locks... As quota files are never truncated and quota code
4828 * itself serializes the operations (and no one else should touch the files)
4829 * we don't have to be afraid of races */
4830 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4831 size_t len, loff_t off)
4832 {
4833 struct inode *inode = sb_dqopt(sb)->files[type];
4834 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4835 int err = 0;
4836 int offset = off & (sb->s_blocksize - 1);
4837 int tocopy;
4838 size_t toread;
4839 struct buffer_head *bh;
4840 loff_t i_size = i_size_read(inode);
4841
4842 if (off > i_size)
4843 return 0;
4844 if (off+len > i_size)
4845 len = i_size-off;
4846 toread = len;
4847 while (toread > 0) {
4848 tocopy = sb->s_blocksize - offset < toread ?
4849 sb->s_blocksize - offset : toread;
4850 bh = ext4_bread(NULL, inode, blk, 0, &err);
4851 if (err)
4852 return err;
4853 if (!bh) /* A hole? */
4854 memset(data, 0, tocopy);
4855 else
4856 memcpy(data, bh->b_data+offset, tocopy);
4857 brelse(bh);
4858 offset = 0;
4859 toread -= tocopy;
4860 data += tocopy;
4861 blk++;
4862 }
4863 return len;
4864 }
4865
4866 /* Write to quotafile (we know the transaction is already started and has
4867 * enough credits) */
4868 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4869 const char *data, size_t len, loff_t off)
4870 {
4871 struct inode *inode = sb_dqopt(sb)->files[type];
4872 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4873 int err = 0;
4874 int offset = off & (sb->s_blocksize - 1);
4875 struct buffer_head *bh;
4876 handle_t *handle = journal_current_handle();
4877
4878 if (EXT4_SB(sb)->s_journal && !handle) {
4879 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4880 " cancelled because transaction is not started",
4881 (unsigned long long)off, (unsigned long long)len);
4882 return -EIO;
4883 }
4884 /*
4885 * Since we account only one data block in transaction credits,
4886 * then it is impossible to cross a block boundary.
4887 */
4888 if (sb->s_blocksize - offset < len) {
4889 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4890 " cancelled because not block aligned",
4891 (unsigned long long)off, (unsigned long long)len);
4892 return -EIO;
4893 }
4894
4895 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4896 bh = ext4_bread(handle, inode, blk, 1, &err);
4897 if (!bh)
4898 goto out;
4899 err = ext4_journal_get_write_access(handle, bh);
4900 if (err) {
4901 brelse(bh);
4902 goto out;
4903 }
4904 lock_buffer(bh);
4905 memcpy(bh->b_data+offset, data, len);
4906 flush_dcache_page(bh->b_page);
4907 unlock_buffer(bh);
4908 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4909 brelse(bh);
4910 out:
4911 if (err) {
4912 mutex_unlock(&inode->i_mutex);
4913 return err;
4914 }
4915 if (inode->i_size < off + len) {
4916 i_size_write(inode, off + len);
4917 EXT4_I(inode)->i_disksize = inode->i_size;
4918 ext4_mark_inode_dirty(handle, inode);
4919 }
4920 mutex_unlock(&inode->i_mutex);
4921 return len;
4922 }
4923
4924 #endif
4925
4926 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4927 const char *dev_name, void *data)
4928 {
4929 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4930 }
4931
4932 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4933 static inline void register_as_ext2(void)
4934 {
4935 int err = register_filesystem(&ext2_fs_type);
4936 if (err)
4937 printk(KERN_WARNING
4938 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4939 }
4940
4941 static inline void unregister_as_ext2(void)
4942 {
4943 unregister_filesystem(&ext2_fs_type);
4944 }
4945
4946 static inline int ext2_feature_set_ok(struct super_block *sb)
4947 {
4948 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4949 return 0;
4950 if (sb->s_flags & MS_RDONLY)
4951 return 1;
4952 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4953 return 0;
4954 return 1;
4955 }
4956 MODULE_ALIAS("ext2");
4957 #else
4958 static inline void register_as_ext2(void) { }
4959 static inline void unregister_as_ext2(void) { }
4960 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4961 #endif
4962
4963 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4964 static inline void register_as_ext3(void)
4965 {
4966 int err = register_filesystem(&ext3_fs_type);
4967 if (err)
4968 printk(KERN_WARNING
4969 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4970 }
4971
4972 static inline void unregister_as_ext3(void)
4973 {
4974 unregister_filesystem(&ext3_fs_type);
4975 }
4976
4977 static inline int ext3_feature_set_ok(struct super_block *sb)
4978 {
4979 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4980 return 0;
4981 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4982 return 0;
4983 if (sb->s_flags & MS_RDONLY)
4984 return 1;
4985 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4986 return 0;
4987 return 1;
4988 }
4989 MODULE_ALIAS("ext3");
4990 #else
4991 static inline void register_as_ext3(void) { }
4992 static inline void unregister_as_ext3(void) { }
4993 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
4994 #endif
4995
4996 static struct file_system_type ext4_fs_type = {
4997 .owner = THIS_MODULE,
4998 .name = "ext4",
4999 .mount = ext4_mount,
5000 .kill_sb = kill_block_super,
5001 .fs_flags = FS_REQUIRES_DEV,
5002 };
5003
5004 static int __init ext4_init_feat_adverts(void)
5005 {
5006 struct ext4_features *ef;
5007 int ret = -ENOMEM;
5008
5009 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5010 if (!ef)
5011 goto out;
5012
5013 ef->f_kobj.kset = ext4_kset;
5014 init_completion(&ef->f_kobj_unregister);
5015 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5016 "features");
5017 if (ret) {
5018 kfree(ef);
5019 goto out;
5020 }
5021
5022 ext4_feat = ef;
5023 ret = 0;
5024 out:
5025 return ret;
5026 }
5027
5028 static void ext4_exit_feat_adverts(void)
5029 {
5030 kobject_put(&ext4_feat->f_kobj);
5031 wait_for_completion(&ext4_feat->f_kobj_unregister);
5032 kfree(ext4_feat);
5033 }
5034
5035 /* Shared across all ext4 file systems */
5036 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5037 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5038
5039 static int __init ext4_init_fs(void)
5040 {
5041 int i, err;
5042
5043 ext4_check_flag_values();
5044
5045 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5046 mutex_init(&ext4__aio_mutex[i]);
5047 init_waitqueue_head(&ext4__ioend_wq[i]);
5048 }
5049
5050 err = ext4_init_pageio();
5051 if (err)
5052 return err;
5053 err = ext4_init_system_zone();
5054 if (err)
5055 goto out6;
5056 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5057 if (!ext4_kset)
5058 goto out5;
5059 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5060
5061 err = ext4_init_feat_adverts();
5062 if (err)
5063 goto out4;
5064
5065 err = ext4_init_mballoc();
5066 if (err)
5067 goto out3;
5068
5069 err = ext4_init_xattr();
5070 if (err)
5071 goto out2;
5072 err = init_inodecache();
5073 if (err)
5074 goto out1;
5075 register_as_ext3();
5076 register_as_ext2();
5077 err = register_filesystem(&ext4_fs_type);
5078 if (err)
5079 goto out;
5080
5081 ext4_li_info = NULL;
5082 mutex_init(&ext4_li_mtx);
5083 return 0;
5084 out:
5085 unregister_as_ext2();
5086 unregister_as_ext3();
5087 destroy_inodecache();
5088 out1:
5089 ext4_exit_xattr();
5090 out2:
5091 ext4_exit_mballoc();
5092 out3:
5093 ext4_exit_feat_adverts();
5094 out4:
5095 if (ext4_proc_root)
5096 remove_proc_entry("fs/ext4", NULL);
5097 kset_unregister(ext4_kset);
5098 out5:
5099 ext4_exit_system_zone();
5100 out6:
5101 ext4_exit_pageio();
5102 return err;
5103 }
5104
5105 static void __exit ext4_exit_fs(void)
5106 {
5107 ext4_destroy_lazyinit_thread();
5108 unregister_as_ext2();
5109 unregister_as_ext3();
5110 unregister_filesystem(&ext4_fs_type);
5111 destroy_inodecache();
5112 ext4_exit_xattr();
5113 ext4_exit_mballoc();
5114 ext4_exit_feat_adverts();
5115 remove_proc_entry("fs/ext4", NULL);
5116 kset_unregister(ext4_kset);
5117 ext4_exit_system_zone();
5118 ext4_exit_pageio();
5119 }
5120
5121 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5122 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5123 MODULE_LICENSE("GPL");
5124 module_init(ext4_init_fs)
5125 module_exit(ext4_exit_fs)