?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
NET: wimax, fix use after free
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blob26147746c272c2112b7067aa6415800434586b50
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/smp_lock.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/proc_fs.h>
39 #include <linux/ctype.h>
40 #include <linux/log2.h>
41 #include <linux/crc16.h>
42 #include <asm/uaccess.h>
43
44 #include "ext4.h"
45 #include "ext4_jbd2.h"
46 #include "xattr.h"
47 #include "acl.h"
48 #include "mballoc.h"
49
50 #define CREATE_TRACE_POINTS
51 #include <trace/events/ext4.h>
52
53 struct proc_dir_entry *ext4_proc_root;
54 static struct kset *ext4_kset;
55
56 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
57 unsigned long journal_devnum);
58 static int ext4_commit_super(struct super_block *sb, int sync);
59 static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61 static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
63 static int ext4_sync_fs(struct super_block *sb, int wait);
64 static const char *ext4_decode_error(struct super_block *sb, int errno,
65 char nbuf[16]);
66 static int ext4_remount(struct super_block *sb, int *flags, char *data);
67 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
68 static int ext4_unfreeze(struct super_block *sb);
69 static void ext4_write_super(struct super_block *sb);
70 static int ext4_freeze(struct super_block *sb);
71 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
72 const char *dev_name, void *data, struct vfsmount *mnt);
73
74 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
75 static struct file_system_type ext3_fs_type = {
76 .owner = THIS_MODULE,
77 .name = "ext3",
78 .get_sb = ext4_get_sb,
79 .kill_sb = kill_block_super,
80 .fs_flags = FS_REQUIRES_DEV,
81 };
82 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
83 #else
84 #define IS_EXT3_SB(sb) (0)
85 #endif
86
87 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
88 struct ext4_group_desc *bg)
89 {
90 return le32_to_cpu(bg->bg_block_bitmap_lo) |
91 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
92 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
93 }
94
95 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
97 {
98 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
99 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
100 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
101 }
102
103 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
104 struct ext4_group_desc *bg)
105 {
106 return le32_to_cpu(bg->bg_inode_table_lo) |
107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
109 }
110
111 __u32 ext4_free_blks_count(struct super_block *sb,
112 struct ext4_group_desc *bg)
113 {
114 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
116 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
117 }
118
119 __u32 ext4_free_inodes_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
121 {
122 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
124 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
125 }
126
127 __u32 ext4_used_dirs_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
129 {
130 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
132 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
133 }
134
135 __u32 ext4_itable_unused_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
137 {
138 return le16_to_cpu(bg->bg_itable_unused_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
140 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
141 }
142
143 void ext4_block_bitmap_set(struct super_block *sb,
144 struct ext4_group_desc *bg, ext4_fsblk_t blk)
145 {
146 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
147 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
148 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
149 }
150
151 void ext4_inode_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
153 {
154 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
157 }
158
159 void ext4_inode_table_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
161 {
162 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
165 }
166
167 void ext4_free_blks_set(struct super_block *sb,
168 struct ext4_group_desc *bg, __u32 count)
169 {
170 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
173 }
174
175 void ext4_free_inodes_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
177 {
178 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
181 }
182
183 void ext4_used_dirs_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
185 {
186 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
189 }
190
191 void ext4_itable_unused_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
193 {
194 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
197 }
198
199
200 /* Just increment the non-pointer handle value */
201 static handle_t *ext4_get_nojournal(void)
202 {
203 handle_t *handle = current->journal_info;
204 unsigned long ref_cnt = (unsigned long)handle;
205
206 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
207
208 ref_cnt++;
209 handle = (handle_t *)ref_cnt;
210
211 current->journal_info = handle;
212 return handle;
213 }
214
215
216 /* Decrement the non-pointer handle value */
217 static void ext4_put_nojournal(handle_t *handle)
218 {
219 unsigned long ref_cnt = (unsigned long)handle;
220
221 BUG_ON(ref_cnt == 0);
222
223 ref_cnt--;
224 handle = (handle_t *)ref_cnt;
225
226 current->journal_info = handle;
227 }
228
229 /*
230 * Wrappers for jbd2_journal_start/end.
231 *
232 * The only special thing we need to do here is to make sure that all
233 * journal_end calls result in the superblock being marked dirty, so
234 * that sync() will call the filesystem's write_super callback if
235 * appropriate.
236 */
237 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
238 {
239 journal_t *journal;
240
241 if (sb->s_flags & MS_RDONLY)
242 return ERR_PTR(-EROFS);
243
244 vfs_check_frozen(sb, SB_FREEZE_TRANS);
245 /* Special case here: if the journal has aborted behind our
246 * backs (eg. EIO in the commit thread), then we still need to
247 * take the FS itself readonly cleanly. */
248 journal = EXT4_SB(sb)->s_journal;
249 if (journal) {
250 if (is_journal_aborted(journal)) {
251 ext4_abort(sb, "Detected aborted journal");
252 return ERR_PTR(-EROFS);
253 }
254 return jbd2_journal_start(journal, nblocks);
255 }
256 return ext4_get_nojournal();
257 }
258
259 /*
260 * The only special thing we need to do here is to make sure that all
261 * jbd2_journal_stop calls result in the superblock being marked dirty, so
262 * that sync() will call the filesystem's write_super callback if
263 * appropriate.
264 */
265 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
266 {
267 struct super_block *sb;
268 int err;
269 int rc;
270
271 if (!ext4_handle_valid(handle)) {
272 ext4_put_nojournal(handle);
273 return 0;
274 }
275 sb = handle->h_transaction->t_journal->j_private;
276 err = handle->h_err;
277 rc = jbd2_journal_stop(handle);
278
279 if (!err)
280 err = rc;
281 if (err)
282 __ext4_std_error(sb, where, line, err);
283 return err;
284 }
285
286 void ext4_journal_abort_handle(const char *caller, unsigned int line,
287 const char *err_fn, struct buffer_head *bh,
288 handle_t *handle, int err)
289 {
290 char nbuf[16];
291 const char *errstr = ext4_decode_error(NULL, err, nbuf);
292
293 BUG_ON(!ext4_handle_valid(handle));
294
295 if (bh)
296 BUFFER_TRACE(bh, "abort");
297
298 if (!handle->h_err)
299 handle->h_err = err;
300
301 if (is_handle_aborted(handle))
302 return;
303
304 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
305 caller, line, errstr, err_fn);
306
307 jbd2_journal_abort_handle(handle);
308 }
309
310 static void __save_error_info(struct super_block *sb, const char *func,
311 unsigned int line)
312 {
313 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
314
315 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
316 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
317 es->s_last_error_time = cpu_to_le32(get_seconds());
318 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
319 es->s_last_error_line = cpu_to_le32(line);
320 if (!es->s_first_error_time) {
321 es->s_first_error_time = es->s_last_error_time;
322 strncpy(es->s_first_error_func, func,
323 sizeof(es->s_first_error_func));
324 es->s_first_error_line = cpu_to_le32(line);
325 es->s_first_error_ino = es->s_last_error_ino;
326 es->s_first_error_block = es->s_last_error_block;
327 }
328 /*
329 * Start the daily error reporting function if it hasn't been
330 * started already
331 */
332 if (!es->s_error_count)
333 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
334 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
335 }
336
337 static void save_error_info(struct super_block *sb, const char *func,
338 unsigned int line)
339 {
340 __save_error_info(sb, func, line);
341 ext4_commit_super(sb, 1);
342 }
343
344
345 /* Deal with the reporting of failure conditions on a filesystem such as
346 * inconsistencies detected or read IO failures.
347 *
348 * On ext2, we can store the error state of the filesystem in the
349 * superblock. That is not possible on ext4, because we may have other
350 * write ordering constraints on the superblock which prevent us from
351 * writing it out straight away; and given that the journal is about to
352 * be aborted, we can't rely on the current, or future, transactions to
353 * write out the superblock safely.
354 *
355 * We'll just use the jbd2_journal_abort() error code to record an error in
356 * the journal instead. On recovery, the journal will complain about
357 * that error until we've noted it down and cleared it.
358 */
359
360 static void ext4_handle_error(struct super_block *sb)
361 {
362 if (sb->s_flags & MS_RDONLY)
363 return;
364
365 if (!test_opt(sb, ERRORS_CONT)) {
366 journal_t *journal = EXT4_SB(sb)->s_journal;
367
368 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
369 if (journal)
370 jbd2_journal_abort(journal, -EIO);
371 }
372 if (test_opt(sb, ERRORS_RO)) {
373 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
374 sb->s_flags |= MS_RDONLY;
375 }
376 if (test_opt(sb, ERRORS_PANIC))
377 panic("EXT4-fs (device %s): panic forced after error\n",
378 sb->s_id);
379 }
380
381 void __ext4_error(struct super_block *sb, const char *function,
382 unsigned int line, const char *fmt, ...)
383 {
384 va_list args;
385
386 va_start(args, fmt);
387 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
388 sb->s_id, function, line, current->comm);
389 vprintk(fmt, args);
390 printk("\n");
391 va_end(args);
392
393 ext4_handle_error(sb);
394 }
395
396 void ext4_error_inode(struct inode *inode, const char *function,
397 unsigned int line, ext4_fsblk_t block,
398 const char *fmt, ...)
399 {
400 va_list args;
401 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
402
403 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
404 es->s_last_error_block = cpu_to_le64(block);
405 save_error_info(inode->i_sb, function, line);
406 va_start(args, fmt);
407 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
408 inode->i_sb->s_id, function, line, inode->i_ino);
409 if (block)
410 printk("block %llu: ", block);
411 printk("comm %s: ", current->comm);
412 vprintk(fmt, args);
413 printk("\n");
414 va_end(args);
415
416 ext4_handle_error(inode->i_sb);
417 }
418
419 void ext4_error_file(struct file *file, const char *function,
420 unsigned int line, const char *fmt, ...)
421 {
422 va_list args;
423 struct ext4_super_block *es;
424 struct inode *inode = file->f_dentry->d_inode;
425 char pathname[80], *path;
426
427 es = EXT4_SB(inode->i_sb)->s_es;
428 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
429 save_error_info(inode->i_sb, function, line);
430 va_start(args, fmt);
431 path = d_path(&(file->f_path), pathname, sizeof(pathname));
432 if (!path)
433 path = "(unknown)";
434 printk(KERN_CRIT
435 "EXT4-fs error (device %s): %s:%d: inode #%lu "
436 "(comm %s path %s): ",
437 inode->i_sb->s_id, function, line, inode->i_ino,
438 current->comm, path);
439 vprintk(fmt, args);
440 printk("\n");
441 va_end(args);
442
443 ext4_handle_error(inode->i_sb);
444 }
445
446 static const char *ext4_decode_error(struct super_block *sb, int errno,
447 char nbuf[16])
448 {
449 char *errstr = NULL;
450
451 switch (errno) {
452 case -EIO:
453 errstr = "IO failure";
454 break;
455 case -ENOMEM:
456 errstr = "Out of memory";
457 break;
458 case -EROFS:
459 if (!sb || (EXT4_SB(sb)->s_journal &&
460 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
461 errstr = "Journal has aborted";
462 else
463 errstr = "Readonly filesystem";
464 break;
465 default:
466 /* If the caller passed in an extra buffer for unknown
467 * errors, textualise them now. Else we just return
468 * NULL. */
469 if (nbuf) {
470 /* Check for truncated error codes... */
471 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
472 errstr = nbuf;
473 }
474 break;
475 }
476
477 return errstr;
478 }
479
480 /* __ext4_std_error decodes expected errors from journaling functions
481 * automatically and invokes the appropriate error response. */
482
483 void __ext4_std_error(struct super_block *sb, const char *function,
484 unsigned int line, int errno)
485 {
486 char nbuf[16];
487 const char *errstr;
488
489 /* Special case: if the error is EROFS, and we're not already
490 * inside a transaction, then there's really no point in logging
491 * an error. */
492 if (errno == -EROFS && journal_current_handle() == NULL &&
493 (sb->s_flags & MS_RDONLY))
494 return;
495
496 errstr = ext4_decode_error(sb, errno, nbuf);
497 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
498 sb->s_id, function, line, errstr);
499 save_error_info(sb, function, line);
500
501 ext4_handle_error(sb);
502 }
503
504 /*
505 * ext4_abort is a much stronger failure handler than ext4_error. The
506 * abort function may be used to deal with unrecoverable failures such
507 * as journal IO errors or ENOMEM at a critical moment in log management.
508 *
509 * We unconditionally force the filesystem into an ABORT|READONLY state,
510 * unless the error response on the fs has been set to panic in which
511 * case we take the easy way out and panic immediately.
512 */
513
514 void __ext4_abort(struct super_block *sb, const char *function,
515 unsigned int line, const char *fmt, ...)
516 {
517 va_list args;
518
519 save_error_info(sb, function, line);
520 va_start(args, fmt);
521 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
522 function, line);
523 vprintk(fmt, args);
524 printk("\n");
525 va_end(args);
526
527 if ((sb->s_flags & MS_RDONLY) == 0) {
528 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
529 sb->s_flags |= MS_RDONLY;
530 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
531 if (EXT4_SB(sb)->s_journal)
532 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
533 save_error_info(sb, function, line);
534 }
535 if (test_opt(sb, ERRORS_PANIC))
536 panic("EXT4-fs panic from previous error\n");
537 }
538
539 void ext4_msg (struct super_block * sb, const char *prefix,
540 const char *fmt, ...)
541 {
542 va_list args;
543
544 va_start(args, fmt);
545 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
546 vprintk(fmt, args);
547 printk("\n");
548 va_end(args);
549 }
550
551 void __ext4_warning(struct super_block *sb, const char *function,
552 unsigned int line, const char *fmt, ...)
553 {
554 va_list args;
555
556 va_start(args, fmt);
557 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
558 sb->s_id, function, line);
559 vprintk(fmt, args);
560 printk("\n");
561 va_end(args);
562 }
563
564 void __ext4_grp_locked_error(const char *function, unsigned int line,
565 struct super_block *sb, ext4_group_t grp,
566 unsigned long ino, ext4_fsblk_t block,
567 const char *fmt, ...)
568 __releases(bitlock)
569 __acquires(bitlock)
570 {
571 va_list args;
572 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
573
574 es->s_last_error_ino = cpu_to_le32(ino);
575 es->s_last_error_block = cpu_to_le64(block);
576 __save_error_info(sb, function, line);
577 va_start(args, fmt);
578 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
579 sb->s_id, function, line, grp);
580 if (ino)
581 printk("inode %lu: ", ino);
582 if (block)
583 printk("block %llu:", (unsigned long long) block);
584 vprintk(fmt, args);
585 printk("\n");
586 va_end(args);
587
588 if (test_opt(sb, ERRORS_CONT)) {
589 ext4_commit_super(sb, 0);
590 return;
591 }
592
593 ext4_unlock_group(sb, grp);
594 ext4_handle_error(sb);
595 /*
596 * We only get here in the ERRORS_RO case; relocking the group
597 * may be dangerous, but nothing bad will happen since the
598 * filesystem will have already been marked read/only and the
599 * journal has been aborted. We return 1 as a hint to callers
600 * who might what to use the return value from
601 * ext4_grp_locked_error() to distinguish beween the
602 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
603 * aggressively from the ext4 function in question, with a
604 * more appropriate error code.
605 */
606 ext4_lock_group(sb, grp);
607 return;
608 }
609
610 void ext4_update_dynamic_rev(struct super_block *sb)
611 {
612 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
613
614 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
615 return;
616
617 ext4_warning(sb,
618 "updating to rev %d because of new feature flag, "
619 "running e2fsck is recommended",
620 EXT4_DYNAMIC_REV);
621
622 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
623 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
624 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
625 /* leave es->s_feature_*compat flags alone */
626 /* es->s_uuid will be set by e2fsck if empty */
627
628 /*
629 * The rest of the superblock fields should be zero, and if not it
630 * means they are likely already in use, so leave them alone. We
631 * can leave it up to e2fsck to clean up any inconsistencies there.
632 */
633 }
634
635 /*
636 * Open the external journal device
637 */
638 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
639 {
640 struct block_device *bdev;
641 char b[BDEVNAME_SIZE];
642
643 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
644 if (IS_ERR(bdev))
645 goto fail;
646 return bdev;
647
648 fail:
649 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
650 __bdevname(dev, b), PTR_ERR(bdev));
651 return NULL;
652 }
653
654 /*
655 * Release the journal device
656 */
657 static int ext4_blkdev_put(struct block_device *bdev)
658 {
659 bd_release(bdev);
660 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
661 }
662
663 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
664 {
665 struct block_device *bdev;
666 int ret = -ENODEV;
667
668 bdev = sbi->journal_bdev;
669 if (bdev) {
670 ret = ext4_blkdev_put(bdev);
671 sbi->journal_bdev = NULL;
672 }
673 return ret;
674 }
675
676 static inline struct inode *orphan_list_entry(struct list_head *l)
677 {
678 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
679 }
680
681 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
682 {
683 struct list_head *l;
684
685 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
686 le32_to_cpu(sbi->s_es->s_last_orphan));
687
688 printk(KERN_ERR "sb_info orphan list:\n");
689 list_for_each(l, &sbi->s_orphan) {
690 struct inode *inode = orphan_list_entry(l);
691 printk(KERN_ERR " "
692 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
693 inode->i_sb->s_id, inode->i_ino, inode,
694 inode->i_mode, inode->i_nlink,
695 NEXT_ORPHAN(inode));
696 }
697 }
698
699 static void ext4_put_super(struct super_block *sb)
700 {
701 struct ext4_sb_info *sbi = EXT4_SB(sb);
702 struct ext4_super_block *es = sbi->s_es;
703 int i, err;
704
705 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
706
707 flush_workqueue(sbi->dio_unwritten_wq);
708 destroy_workqueue(sbi->dio_unwritten_wq);
709
710 lock_super(sb);
711 lock_kernel();
712 if (sb->s_dirt)
713 ext4_commit_super(sb, 1);
714
715 if (sbi->s_journal) {
716 err = jbd2_journal_destroy(sbi->s_journal);
717 sbi->s_journal = NULL;
718 if (err < 0)
719 ext4_abort(sb, "Couldn't clean up the journal");
720 }
721
722 ext4_release_system_zone(sb);
723 ext4_mb_release(sb);
724 ext4_ext_release(sb);
725 ext4_xattr_put_super(sb);
726
727 if (!(sb->s_flags & MS_RDONLY)) {
728 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
729 es->s_state = cpu_to_le16(sbi->s_mount_state);
730 ext4_commit_super(sb, 1);
731 }
732 if (sbi->s_proc) {
733 remove_proc_entry(sb->s_id, ext4_proc_root);
734 }
735 kobject_del(&sbi->s_kobj);
736
737 for (i = 0; i < sbi->s_gdb_count; i++)
738 brelse(sbi->s_group_desc[i]);
739 kfree(sbi->s_group_desc);
740 if (is_vmalloc_addr(sbi->s_flex_groups))
741 vfree(sbi->s_flex_groups);
742 else
743 kfree(sbi->s_flex_groups);
744 percpu_counter_destroy(&sbi->s_freeblocks_counter);
745 percpu_counter_destroy(&sbi->s_freeinodes_counter);
746 percpu_counter_destroy(&sbi->s_dirs_counter);
747 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
748 brelse(sbi->s_sbh);
749 #ifdef CONFIG_QUOTA
750 for (i = 0; i < MAXQUOTAS; i++)
751 kfree(sbi->s_qf_names[i]);
752 #endif
753
754 /* Debugging code just in case the in-memory inode orphan list
755 * isn't empty. The on-disk one can be non-empty if we've
756 * detected an error and taken the fs readonly, but the
757 * in-memory list had better be clean by this point. */
758 if (!list_empty(&sbi->s_orphan))
759 dump_orphan_list(sb, sbi);
760 J_ASSERT(list_empty(&sbi->s_orphan));
761
762 invalidate_bdev(sb->s_bdev);
763 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
764 /*
765 * Invalidate the journal device's buffers. We don't want them
766 * floating about in memory - the physical journal device may
767 * hotswapped, and it breaks the `ro-after' testing code.
768 */
769 sync_blockdev(sbi->journal_bdev);
770 invalidate_bdev(sbi->journal_bdev);
771 ext4_blkdev_remove(sbi);
772 }
773 sb->s_fs_info = NULL;
774 /*
775 * Now that we are completely done shutting down the
776 * superblock, we need to actually destroy the kobject.
777 */
778 unlock_kernel();
779 unlock_super(sb);
780 kobject_put(&sbi->s_kobj);
781 wait_for_completion(&sbi->s_kobj_unregister);
782 kfree(sbi->s_blockgroup_lock);
783 kfree(sbi);
784 }
785
786 static struct kmem_cache *ext4_inode_cachep;
787
788 /*
789 * Called inside transaction, so use GFP_NOFS
790 */
791 static struct inode *ext4_alloc_inode(struct super_block *sb)
792 {
793 struct ext4_inode_info *ei;
794
795 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
796 if (!ei)
797 return NULL;
798
799 ei->vfs_inode.i_version = 1;
800 ei->vfs_inode.i_data.writeback_index = 0;
801 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
802 INIT_LIST_HEAD(&ei->i_prealloc_list);
803 spin_lock_init(&ei->i_prealloc_lock);
804 /*
805 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
806 * therefore it can be null here. Don't check it, just initialize
807 * jinode.
808 */
809 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
810 ei->i_reserved_data_blocks = 0;
811 ei->i_reserved_meta_blocks = 0;
812 ei->i_allocated_meta_blocks = 0;
813 ei->i_da_metadata_calc_len = 0;
814 ei->i_delalloc_reserved_flag = 0;
815 spin_lock_init(&(ei->i_block_reservation_lock));
816 #ifdef CONFIG_QUOTA
817 ei->i_reserved_quota = 0;
818 #endif
819 INIT_LIST_HEAD(&ei->i_completed_io_list);
820 spin_lock_init(&ei->i_completed_io_lock);
821 ei->cur_aio_dio = NULL;
822 ei->i_sync_tid = 0;
823 ei->i_datasync_tid = 0;
824
825 return &ei->vfs_inode;
826 }
827
828 static void ext4_destroy_inode(struct inode *inode)
829 {
830 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
831 ext4_msg(inode->i_sb, KERN_ERR,
832 "Inode %lu (%p): orphan list check failed!",
833 inode->i_ino, EXT4_I(inode));
834 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
835 EXT4_I(inode), sizeof(struct ext4_inode_info),
836 true);
837 dump_stack();
838 }
839 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
840 }
841
842 static void init_once(void *foo)
843 {
844 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
845
846 INIT_LIST_HEAD(&ei->i_orphan);
847 #ifdef CONFIG_EXT4_FS_XATTR
848 init_rwsem(&ei->xattr_sem);
849 #endif
850 init_rwsem(&ei->i_data_sem);
851 inode_init_once(&ei->vfs_inode);
852 }
853
854 static int init_inodecache(void)
855 {
856 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
857 sizeof(struct ext4_inode_info),
858 0, (SLAB_RECLAIM_ACCOUNT|
859 SLAB_MEM_SPREAD),
860 init_once);
861 if (ext4_inode_cachep == NULL)
862 return -ENOMEM;
863 return 0;
864 }
865
866 static void destroy_inodecache(void)
867 {
868 kmem_cache_destroy(ext4_inode_cachep);
869 }
870
871 void ext4_clear_inode(struct inode *inode)
872 {
873 invalidate_inode_buffers(inode);
874 end_writeback(inode);
875 dquot_drop(inode);
876 ext4_discard_preallocations(inode);
877 if (EXT4_JOURNAL(inode))
878 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
879 &EXT4_I(inode)->jinode);
880 }
881
882 static inline void ext4_show_quota_options(struct seq_file *seq,
883 struct super_block *sb)
884 {
885 #if defined(CONFIG_QUOTA)
886 struct ext4_sb_info *sbi = EXT4_SB(sb);
887
888 if (sbi->s_jquota_fmt) {
889 char *fmtname = "";
890
891 switch (sbi->s_jquota_fmt) {
892 case QFMT_VFS_OLD:
893 fmtname = "vfsold";
894 break;
895 case QFMT_VFS_V0:
896 fmtname = "vfsv0";
897 break;
898 case QFMT_VFS_V1:
899 fmtname = "vfsv1";
900 break;
901 }
902 seq_printf(seq, ",jqfmt=%s", fmtname);
903 }
904
905 if (sbi->s_qf_names[USRQUOTA])
906 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
907
908 if (sbi->s_qf_names[GRPQUOTA])
909 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
910
911 if (test_opt(sb, USRQUOTA))
912 seq_puts(seq, ",usrquota");
913
914 if (test_opt(sb, GRPQUOTA))
915 seq_puts(seq, ",grpquota");
916 #endif
917 }
918
919 /*
920 * Show an option if
921 * - it's set to a non-default value OR
922 * - if the per-sb default is different from the global default
923 */
924 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
925 {
926 int def_errors;
927 unsigned long def_mount_opts;
928 struct super_block *sb = vfs->mnt_sb;
929 struct ext4_sb_info *sbi = EXT4_SB(sb);
930 struct ext4_super_block *es = sbi->s_es;
931
932 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
933 def_errors = le16_to_cpu(es->s_errors);
934
935 if (sbi->s_sb_block != 1)
936 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
937 if (test_opt(sb, MINIX_DF))
938 seq_puts(seq, ",minixdf");
939 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
940 seq_puts(seq, ",grpid");
941 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
942 seq_puts(seq, ",nogrpid");
943 if (sbi->s_resuid != EXT4_DEF_RESUID ||
944 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
945 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
946 }
947 if (sbi->s_resgid != EXT4_DEF_RESGID ||
948 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
949 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
950 }
951 if (test_opt(sb, ERRORS_RO)) {
952 if (def_errors == EXT4_ERRORS_PANIC ||
953 def_errors == EXT4_ERRORS_CONTINUE) {
954 seq_puts(seq, ",errors=remount-ro");
955 }
956 }
957 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
958 seq_puts(seq, ",errors=continue");
959 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
960 seq_puts(seq, ",errors=panic");
961 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
962 seq_puts(seq, ",nouid32");
963 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
964 seq_puts(seq, ",debug");
965 if (test_opt(sb, OLDALLOC))
966 seq_puts(seq, ",oldalloc");
967 #ifdef CONFIG_EXT4_FS_XATTR
968 if (test_opt(sb, XATTR_USER) &&
969 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
970 seq_puts(seq, ",user_xattr");
971 if (!test_opt(sb, XATTR_USER) &&
972 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
973 seq_puts(seq, ",nouser_xattr");
974 }
975 #endif
976 #ifdef CONFIG_EXT4_FS_POSIX_ACL
977 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
978 seq_puts(seq, ",acl");
979 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
980 seq_puts(seq, ",noacl");
981 #endif
982 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
983 seq_printf(seq, ",commit=%u",
984 (unsigned) (sbi->s_commit_interval / HZ));
985 }
986 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
987 seq_printf(seq, ",min_batch_time=%u",
988 (unsigned) sbi->s_min_batch_time);
989 }
990 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
991 seq_printf(seq, ",max_batch_time=%u",
992 (unsigned) sbi->s_min_batch_time);
993 }
994
995 /*
996 * We're changing the default of barrier mount option, so
997 * let's always display its mount state so it's clear what its
998 * status is.
999 */
1000 seq_puts(seq, ",barrier=");
1001 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1002 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1003 seq_puts(seq, ",journal_async_commit");
1004 else if (test_opt(sb, JOURNAL_CHECKSUM))
1005 seq_puts(seq, ",journal_checksum");
1006 if (test_opt(sb, I_VERSION))
1007 seq_puts(seq, ",i_version");
1008 if (!test_opt(sb, DELALLOC) &&
1009 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1010 seq_puts(seq, ",nodelalloc");
1011
1012 if (sbi->s_stripe)
1013 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1014 /*
1015 * journal mode get enabled in different ways
1016 * So just print the value even if we didn't specify it
1017 */
1018 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1019 seq_puts(seq, ",data=journal");
1020 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1021 seq_puts(seq, ",data=ordered");
1022 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1023 seq_puts(seq, ",data=writeback");
1024
1025 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1026 seq_printf(seq, ",inode_readahead_blks=%u",
1027 sbi->s_inode_readahead_blks);
1028
1029 if (test_opt(sb, DATA_ERR_ABORT))
1030 seq_puts(seq, ",data_err=abort");
1031
1032 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1033 seq_puts(seq, ",noauto_da_alloc");
1034
1035 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1036 seq_puts(seq, ",discard");
1037
1038 if (test_opt(sb, NOLOAD))
1039 seq_puts(seq, ",norecovery");
1040
1041 if (test_opt(sb, DIOREAD_NOLOCK))
1042 seq_puts(seq, ",dioread_nolock");
1043
1044 if (test_opt(sb, BLOCK_VALIDITY) &&
1045 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1046 seq_puts(seq, ",block_validity");
1047
1048 ext4_show_quota_options(seq, sb);
1049
1050 return 0;
1051 }
1052
1053 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1054 u64 ino, u32 generation)
1055 {
1056 struct inode *inode;
1057
1058 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1059 return ERR_PTR(-ESTALE);
1060 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1061 return ERR_PTR(-ESTALE);
1062
1063 /* iget isn't really right if the inode is currently unallocated!!
1064 *
1065 * ext4_read_inode will return a bad_inode if the inode had been
1066 * deleted, so we should be safe.
1067 *
1068 * Currently we don't know the generation for parent directory, so
1069 * a generation of 0 means "accept any"
1070 */
1071 inode = ext4_iget(sb, ino);
1072 if (IS_ERR(inode))
1073 return ERR_CAST(inode);
1074 if (generation && inode->i_generation != generation) {
1075 iput(inode);
1076 return ERR_PTR(-ESTALE);
1077 }
1078
1079 return inode;
1080 }
1081
1082 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1083 int fh_len, int fh_type)
1084 {
1085 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1086 ext4_nfs_get_inode);
1087 }
1088
1089 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1090 int fh_len, int fh_type)
1091 {
1092 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1093 ext4_nfs_get_inode);
1094 }
1095
1096 /*
1097 * Try to release metadata pages (indirect blocks, directories) which are
1098 * mapped via the block device. Since these pages could have journal heads
1099 * which would prevent try_to_free_buffers() from freeing them, we must use
1100 * jbd2 layer's try_to_free_buffers() function to release them.
1101 */
1102 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1103 gfp_t wait)
1104 {
1105 journal_t *journal = EXT4_SB(sb)->s_journal;
1106
1107 WARN_ON(PageChecked(page));
1108 if (!page_has_buffers(page))
1109 return 0;
1110 if (journal)
1111 return jbd2_journal_try_to_free_buffers(journal, page,
1112 wait & ~__GFP_WAIT);
1113 return try_to_free_buffers(page);
1114 }
1115
1116 #ifdef CONFIG_QUOTA
1117 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1118 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1119
1120 static int ext4_write_dquot(struct dquot *dquot);
1121 static int ext4_acquire_dquot(struct dquot *dquot);
1122 static int ext4_release_dquot(struct dquot *dquot);
1123 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1124 static int ext4_write_info(struct super_block *sb, int type);
1125 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1126 char *path);
1127 static int ext4_quota_off(struct super_block *sb, int type);
1128 static int ext4_quota_on_mount(struct super_block *sb, int type);
1129 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1130 size_t len, loff_t off);
1131 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1132 const char *data, size_t len, loff_t off);
1133
1134 static const struct dquot_operations ext4_quota_operations = {
1135 #ifdef CONFIG_QUOTA
1136 .get_reserved_space = ext4_get_reserved_space,
1137 #endif
1138 .write_dquot = ext4_write_dquot,
1139 .acquire_dquot = ext4_acquire_dquot,
1140 .release_dquot = ext4_release_dquot,
1141 .mark_dirty = ext4_mark_dquot_dirty,
1142 .write_info = ext4_write_info,
1143 .alloc_dquot = dquot_alloc,
1144 .destroy_dquot = dquot_destroy,
1145 };
1146
1147 static const struct quotactl_ops ext4_qctl_operations = {
1148 .quota_on = ext4_quota_on,
1149 .quota_off = ext4_quota_off,
1150 .quota_sync = dquot_quota_sync,
1151 .get_info = dquot_get_dqinfo,
1152 .set_info = dquot_set_dqinfo,
1153 .get_dqblk = dquot_get_dqblk,
1154 .set_dqblk = dquot_set_dqblk
1155 };
1156 #endif
1157
1158 static const struct super_operations ext4_sops = {
1159 .alloc_inode = ext4_alloc_inode,
1160 .destroy_inode = ext4_destroy_inode,
1161 .write_inode = ext4_write_inode,
1162 .dirty_inode = ext4_dirty_inode,
1163 .evict_inode = ext4_evict_inode,
1164 .put_super = ext4_put_super,
1165 .sync_fs = ext4_sync_fs,
1166 .freeze_fs = ext4_freeze,
1167 .unfreeze_fs = ext4_unfreeze,
1168 .statfs = ext4_statfs,
1169 .remount_fs = ext4_remount,
1170 .show_options = ext4_show_options,
1171 #ifdef CONFIG_QUOTA
1172 .quota_read = ext4_quota_read,
1173 .quota_write = ext4_quota_write,
1174 #endif
1175 .bdev_try_to_free_page = bdev_try_to_free_page,
1176 };
1177
1178 static const struct super_operations ext4_nojournal_sops = {
1179 .alloc_inode = ext4_alloc_inode,
1180 .destroy_inode = ext4_destroy_inode,
1181 .write_inode = ext4_write_inode,
1182 .dirty_inode = ext4_dirty_inode,
1183 .evict_inode = ext4_evict_inode,
1184 .write_super = ext4_write_super,
1185 .put_super = ext4_put_super,
1186 .statfs = ext4_statfs,
1187 .remount_fs = ext4_remount,
1188 .show_options = ext4_show_options,
1189 #ifdef CONFIG_QUOTA
1190 .quota_read = ext4_quota_read,
1191 .quota_write = ext4_quota_write,
1192 #endif
1193 .bdev_try_to_free_page = bdev_try_to_free_page,
1194 };
1195
1196 static const struct export_operations ext4_export_ops = {
1197 .fh_to_dentry = ext4_fh_to_dentry,
1198 .fh_to_parent = ext4_fh_to_parent,
1199 .get_parent = ext4_get_parent,
1200 };
1201
1202 enum {
1203 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1204 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1205 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1206 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1207 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1208 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1209 Opt_journal_update, Opt_journal_dev,
1210 Opt_journal_checksum, Opt_journal_async_commit,
1211 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1212 Opt_data_err_abort, Opt_data_err_ignore,
1213 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1214 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1215 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1216 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1217 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1218 Opt_block_validity, Opt_noblock_validity,
1219 Opt_inode_readahead_blks, Opt_journal_ioprio,
1220 Opt_dioread_nolock, Opt_dioread_lock,
1221 Opt_discard, Opt_nodiscard,
1222 };
1223
1224 static const match_table_t tokens = {
1225 {Opt_bsd_df, "bsddf"},
1226 {Opt_minix_df, "minixdf"},
1227 {Opt_grpid, "grpid"},
1228 {Opt_grpid, "bsdgroups"},
1229 {Opt_nogrpid, "nogrpid"},
1230 {Opt_nogrpid, "sysvgroups"},
1231 {Opt_resgid, "resgid=%u"},
1232 {Opt_resuid, "resuid=%u"},
1233 {Opt_sb, "sb=%u"},
1234 {Opt_err_cont, "errors=continue"},
1235 {Opt_err_panic, "errors=panic"},
1236 {Opt_err_ro, "errors=remount-ro"},
1237 {Opt_nouid32, "nouid32"},
1238 {Opt_debug, "debug"},
1239 {Opt_oldalloc, "oldalloc"},
1240 {Opt_orlov, "orlov"},
1241 {Opt_user_xattr, "user_xattr"},
1242 {Opt_nouser_xattr, "nouser_xattr"},
1243 {Opt_acl, "acl"},
1244 {Opt_noacl, "noacl"},
1245 {Opt_noload, "noload"},
1246 {Opt_noload, "norecovery"},
1247 {Opt_nobh, "nobh"},
1248 {Opt_bh, "bh"},
1249 {Opt_commit, "commit=%u"},
1250 {Opt_min_batch_time, "min_batch_time=%u"},
1251 {Opt_max_batch_time, "max_batch_time=%u"},
1252 {Opt_journal_update, "journal=update"},
1253 {Opt_journal_dev, "journal_dev=%u"},
1254 {Opt_journal_checksum, "journal_checksum"},
1255 {Opt_journal_async_commit, "journal_async_commit"},
1256 {Opt_abort, "abort"},
1257 {Opt_data_journal, "data=journal"},
1258 {Opt_data_ordered, "data=ordered"},
1259 {Opt_data_writeback, "data=writeback"},
1260 {Opt_data_err_abort, "data_err=abort"},
1261 {Opt_data_err_ignore, "data_err=ignore"},
1262 {Opt_offusrjquota, "usrjquota="},
1263 {Opt_usrjquota, "usrjquota=%s"},
1264 {Opt_offgrpjquota, "grpjquota="},
1265 {Opt_grpjquota, "grpjquota=%s"},
1266 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1267 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1268 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1269 {Opt_grpquota, "grpquota"},
1270 {Opt_noquota, "noquota"},
1271 {Opt_quota, "quota"},
1272 {Opt_usrquota, "usrquota"},
1273 {Opt_barrier, "barrier=%u"},
1274 {Opt_barrier, "barrier"},
1275 {Opt_nobarrier, "nobarrier"},
1276 {Opt_i_version, "i_version"},
1277 {Opt_stripe, "stripe=%u"},
1278 {Opt_resize, "resize"},
1279 {Opt_delalloc, "delalloc"},
1280 {Opt_nodelalloc, "nodelalloc"},
1281 {Opt_block_validity, "block_validity"},
1282 {Opt_noblock_validity, "noblock_validity"},
1283 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1284 {Opt_journal_ioprio, "journal_ioprio=%u"},
1285 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1286 {Opt_auto_da_alloc, "auto_da_alloc"},
1287 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1288 {Opt_dioread_nolock, "dioread_nolock"},
1289 {Opt_dioread_lock, "dioread_lock"},
1290 {Opt_discard, "discard"},
1291 {Opt_nodiscard, "nodiscard"},
1292 {Opt_err, NULL},
1293 };
1294
1295 static ext4_fsblk_t get_sb_block(void **data)
1296 {
1297 ext4_fsblk_t sb_block;
1298 char *options = (char *) *data;
1299
1300 if (!options || strncmp(options, "sb=", 3) != 0)
1301 return 1; /* Default location */
1302
1303 options += 3;
1304 /* TODO: use simple_strtoll with >32bit ext4 */
1305 sb_block = simple_strtoul(options, &options, 0);
1306 if (*options && *options != ',') {
1307 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1308 (char *) *data);
1309 return 1;
1310 }
1311 if (*options == ',')
1312 options++;
1313 *data = (void *) options;
1314
1315 return sb_block;
1316 }
1317
1318 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1319 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1320 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1321
1322 #ifdef CONFIG_QUOTA
1323 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1324 {
1325 struct ext4_sb_info *sbi = EXT4_SB(sb);
1326 char *qname;
1327
1328 if (sb_any_quota_loaded(sb) &&
1329 !sbi->s_qf_names[qtype]) {
1330 ext4_msg(sb, KERN_ERR,
1331 "Cannot change journaled "
1332 "quota options when quota turned on");
1333 return 0;
1334 }
1335 qname = match_strdup(args);
1336 if (!qname) {
1337 ext4_msg(sb, KERN_ERR,
1338 "Not enough memory for storing quotafile name");
1339 return 0;
1340 }
1341 if (sbi->s_qf_names[qtype] &&
1342 strcmp(sbi->s_qf_names[qtype], qname)) {
1343 ext4_msg(sb, KERN_ERR,
1344 "%s quota file already specified", QTYPE2NAME(qtype));
1345 kfree(qname);
1346 return 0;
1347 }
1348 sbi->s_qf_names[qtype] = qname;
1349 if (strchr(sbi->s_qf_names[qtype], '/')) {
1350 ext4_msg(sb, KERN_ERR,
1351 "quotafile must be on filesystem root");
1352 kfree(sbi->s_qf_names[qtype]);
1353 sbi->s_qf_names[qtype] = NULL;
1354 return 0;
1355 }
1356 set_opt(sbi->s_mount_opt, QUOTA);
1357 return 1;
1358 }
1359
1360 static int clear_qf_name(struct super_block *sb, int qtype)
1361 {
1362
1363 struct ext4_sb_info *sbi = EXT4_SB(sb);
1364
1365 if (sb_any_quota_loaded(sb) &&
1366 sbi->s_qf_names[qtype]) {
1367 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1368 " when quota turned on");
1369 return 0;
1370 }
1371 /*
1372 * The space will be released later when all options are confirmed
1373 * to be correct
1374 */
1375 sbi->s_qf_names[qtype] = NULL;
1376 return 1;
1377 }
1378 #endif
1379
1380 static int parse_options(char *options, struct super_block *sb,
1381 unsigned long *journal_devnum,
1382 unsigned int *journal_ioprio,
1383 ext4_fsblk_t *n_blocks_count, int is_remount)
1384 {
1385 struct ext4_sb_info *sbi = EXT4_SB(sb);
1386 char *p;
1387 substring_t args[MAX_OPT_ARGS];
1388 int data_opt = 0;
1389 int option;
1390 #ifdef CONFIG_QUOTA
1391 int qfmt;
1392 #endif
1393
1394 if (!options)
1395 return 1;
1396
1397 while ((p = strsep(&options, ",")) != NULL) {
1398 int token;
1399 if (!*p)
1400 continue;
1401
1402 /*
1403 * Initialize args struct so we know whether arg was
1404 * found; some options take optional arguments.
1405 */
1406 args[0].to = args[0].from = 0;
1407 token = match_token(p, tokens, args);
1408 switch (token) {
1409 case Opt_bsd_df:
1410 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1411 clear_opt(sbi->s_mount_opt, MINIX_DF);
1412 break;
1413 case Opt_minix_df:
1414 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1415 set_opt(sbi->s_mount_opt, MINIX_DF);
1416
1417 break;
1418 case Opt_grpid:
1419 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1420 set_opt(sbi->s_mount_opt, GRPID);
1421
1422 break;
1423 case Opt_nogrpid:
1424 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1425 clear_opt(sbi->s_mount_opt, GRPID);
1426
1427 break;
1428 case Opt_resuid:
1429 if (match_int(&args[0], &option))
1430 return 0;
1431 sbi->s_resuid = option;
1432 break;
1433 case Opt_resgid:
1434 if (match_int(&args[0], &option))
1435 return 0;
1436 sbi->s_resgid = option;
1437 break;
1438 case Opt_sb:
1439 /* handled by get_sb_block() instead of here */
1440 /* *sb_block = match_int(&args[0]); */
1441 break;
1442 case Opt_err_panic:
1443 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1444 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1445 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1446 break;
1447 case Opt_err_ro:
1448 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1449 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1450 set_opt(sbi->s_mount_opt, ERRORS_RO);
1451 break;
1452 case Opt_err_cont:
1453 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1454 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1455 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1456 break;
1457 case Opt_nouid32:
1458 set_opt(sbi->s_mount_opt, NO_UID32);
1459 break;
1460 case Opt_debug:
1461 set_opt(sbi->s_mount_opt, DEBUG);
1462 break;
1463 case Opt_oldalloc:
1464 set_opt(sbi->s_mount_opt, OLDALLOC);
1465 break;
1466 case Opt_orlov:
1467 clear_opt(sbi->s_mount_opt, OLDALLOC);
1468 break;
1469 #ifdef CONFIG_EXT4_FS_XATTR
1470 case Opt_user_xattr:
1471 set_opt(sbi->s_mount_opt, XATTR_USER);
1472 break;
1473 case Opt_nouser_xattr:
1474 clear_opt(sbi->s_mount_opt, XATTR_USER);
1475 break;
1476 #else
1477 case Opt_user_xattr:
1478 case Opt_nouser_xattr:
1479 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1480 break;
1481 #endif
1482 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1483 case Opt_acl:
1484 set_opt(sbi->s_mount_opt, POSIX_ACL);
1485 break;
1486 case Opt_noacl:
1487 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1488 break;
1489 #else
1490 case Opt_acl:
1491 case Opt_noacl:
1492 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1493 break;
1494 #endif
1495 case Opt_journal_update:
1496 /* @@@ FIXME */
1497 /* Eventually we will want to be able to create
1498 a journal file here. For now, only allow the
1499 user to specify an existing inode to be the
1500 journal file. */
1501 if (is_remount) {
1502 ext4_msg(sb, KERN_ERR,
1503 "Cannot specify journal on remount");
1504 return 0;
1505 }
1506 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1507 break;
1508 case Opt_journal_dev:
1509 if (is_remount) {
1510 ext4_msg(sb, KERN_ERR,
1511 "Cannot specify journal on remount");
1512 return 0;
1513 }
1514 if (match_int(&args[0], &option))
1515 return 0;
1516 *journal_devnum = option;
1517 break;
1518 case Opt_journal_checksum:
1519 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1520 break;
1521 case Opt_journal_async_commit:
1522 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1523 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1524 break;
1525 case Opt_noload:
1526 set_opt(sbi->s_mount_opt, NOLOAD);
1527 break;
1528 case Opt_commit:
1529 if (match_int(&args[0], &option))
1530 return 0;
1531 if (option < 0)
1532 return 0;
1533 if (option == 0)
1534 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1535 sbi->s_commit_interval = HZ * option;
1536 break;
1537 case Opt_max_batch_time:
1538 if (match_int(&args[0], &option))
1539 return 0;
1540 if (option < 0)
1541 return 0;
1542 if (option == 0)
1543 option = EXT4_DEF_MAX_BATCH_TIME;
1544 sbi->s_max_batch_time = option;
1545 break;
1546 case Opt_min_batch_time:
1547 if (match_int(&args[0], &option))
1548 return 0;
1549 if (option < 0)
1550 return 0;
1551 sbi->s_min_batch_time = option;
1552 break;
1553 case Opt_data_journal:
1554 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1555 goto datacheck;
1556 case Opt_data_ordered:
1557 data_opt = EXT4_MOUNT_ORDERED_DATA;
1558 goto datacheck;
1559 case Opt_data_writeback:
1560 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1561 datacheck:
1562 if (is_remount) {
1563 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1564 ext4_msg(sb, KERN_ERR,
1565 "Cannot change data mode on remount");
1566 return 0;
1567 }
1568 } else {
1569 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1570 sbi->s_mount_opt |= data_opt;
1571 }
1572 break;
1573 case Opt_data_err_abort:
1574 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1575 break;
1576 case Opt_data_err_ignore:
1577 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1578 break;
1579 #ifdef CONFIG_QUOTA
1580 case Opt_usrjquota:
1581 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1582 return 0;
1583 break;
1584 case Opt_grpjquota:
1585 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1586 return 0;
1587 break;
1588 case Opt_offusrjquota:
1589 if (!clear_qf_name(sb, USRQUOTA))
1590 return 0;
1591 break;
1592 case Opt_offgrpjquota:
1593 if (!clear_qf_name(sb, GRPQUOTA))
1594 return 0;
1595 break;
1596
1597 case Opt_jqfmt_vfsold:
1598 qfmt = QFMT_VFS_OLD;
1599 goto set_qf_format;
1600 case Opt_jqfmt_vfsv0:
1601 qfmt = QFMT_VFS_V0;
1602 goto set_qf_format;
1603 case Opt_jqfmt_vfsv1:
1604 qfmt = QFMT_VFS_V1;
1605 set_qf_format:
1606 if (sb_any_quota_loaded(sb) &&
1607 sbi->s_jquota_fmt != qfmt) {
1608 ext4_msg(sb, KERN_ERR, "Cannot change "
1609 "journaled quota options when "
1610 "quota turned on");
1611 return 0;
1612 }
1613 sbi->s_jquota_fmt = qfmt;
1614 break;
1615 case Opt_quota:
1616 case Opt_usrquota:
1617 set_opt(sbi->s_mount_opt, QUOTA);
1618 set_opt(sbi->s_mount_opt, USRQUOTA);
1619 break;
1620 case Opt_grpquota:
1621 set_opt(sbi->s_mount_opt, QUOTA);
1622 set_opt(sbi->s_mount_opt, GRPQUOTA);
1623 break;
1624 case Opt_noquota:
1625 if (sb_any_quota_loaded(sb)) {
1626 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1627 "options when quota turned on");
1628 return 0;
1629 }
1630 clear_opt(sbi->s_mount_opt, QUOTA);
1631 clear_opt(sbi->s_mount_opt, USRQUOTA);
1632 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1633 break;
1634 #else
1635 case Opt_quota:
1636 case Opt_usrquota:
1637 case Opt_grpquota:
1638 ext4_msg(sb, KERN_ERR,
1639 "quota options not supported");
1640 break;
1641 case Opt_usrjquota:
1642 case Opt_grpjquota:
1643 case Opt_offusrjquota:
1644 case Opt_offgrpjquota:
1645 case Opt_jqfmt_vfsold:
1646 case Opt_jqfmt_vfsv0:
1647 case Opt_jqfmt_vfsv1:
1648 ext4_msg(sb, KERN_ERR,
1649 "journaled quota options not supported");
1650 break;
1651 case Opt_noquota:
1652 break;
1653 #endif
1654 case Opt_abort:
1655 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1656 break;
1657 case Opt_nobarrier:
1658 clear_opt(sbi->s_mount_opt, BARRIER);
1659 break;
1660 case Opt_barrier:
1661 if (args[0].from) {
1662 if (match_int(&args[0], &option))
1663 return 0;
1664 } else
1665 option = 1; /* No argument, default to 1 */
1666 if (option)
1667 set_opt(sbi->s_mount_opt, BARRIER);
1668 else
1669 clear_opt(sbi->s_mount_opt, BARRIER);
1670 break;
1671 case Opt_ignore:
1672 break;
1673 case Opt_resize:
1674 if (!is_remount) {
1675 ext4_msg(sb, KERN_ERR,
1676 "resize option only available "
1677 "for remount");
1678 return 0;
1679 }
1680 if (match_int(&args[0], &option) != 0)
1681 return 0;
1682 *n_blocks_count = option;
1683 break;
1684 case Opt_nobh:
1685 ext4_msg(sb, KERN_WARNING,
1686 "Ignoring deprecated nobh option");
1687 break;
1688 case Opt_bh:
1689 ext4_msg(sb, KERN_WARNING,
1690 "Ignoring deprecated bh option");
1691 break;
1692 case Opt_i_version:
1693 set_opt(sbi->s_mount_opt, I_VERSION);
1694 sb->s_flags |= MS_I_VERSION;
1695 break;
1696 case Opt_nodelalloc:
1697 clear_opt(sbi->s_mount_opt, DELALLOC);
1698 break;
1699 case Opt_stripe:
1700 if (match_int(&args[0], &option))
1701 return 0;
1702 if (option < 0)
1703 return 0;
1704 sbi->s_stripe = option;
1705 break;
1706 case Opt_delalloc:
1707 set_opt(sbi->s_mount_opt, DELALLOC);
1708 break;
1709 case Opt_block_validity:
1710 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1711 break;
1712 case Opt_noblock_validity:
1713 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1714 break;
1715 case Opt_inode_readahead_blks:
1716 if (match_int(&args[0], &option))
1717 return 0;
1718 if (option < 0 || option > (1 << 30))
1719 return 0;
1720 if (!is_power_of_2(option)) {
1721 ext4_msg(sb, KERN_ERR,
1722 "EXT4-fs: inode_readahead_blks"
1723 " must be a power of 2");
1724 return 0;
1725 }
1726 sbi->s_inode_readahead_blks = option;
1727 break;
1728 case Opt_journal_ioprio:
1729 if (match_int(&args[0], &option))
1730 return 0;
1731 if (option < 0 || option > 7)
1732 break;
1733 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1734 option);
1735 break;
1736 case Opt_noauto_da_alloc:
1737 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1738 break;
1739 case Opt_auto_da_alloc:
1740 if (args[0].from) {
1741 if (match_int(&args[0], &option))
1742 return 0;
1743 } else
1744 option = 1; /* No argument, default to 1 */
1745 if (option)
1746 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1747 else
1748 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1749 break;
1750 case Opt_discard:
1751 set_opt(sbi->s_mount_opt, DISCARD);
1752 break;
1753 case Opt_nodiscard:
1754 clear_opt(sbi->s_mount_opt, DISCARD);
1755 break;
1756 case Opt_dioread_nolock:
1757 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1758 break;
1759 case Opt_dioread_lock:
1760 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1761 break;
1762 default:
1763 ext4_msg(sb, KERN_ERR,
1764 "Unrecognized mount option \"%s\" "
1765 "or missing value", p);
1766 return 0;
1767 }
1768 }
1769 #ifdef CONFIG_QUOTA
1770 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1771 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1772 clear_opt(sbi->s_mount_opt, USRQUOTA);
1773
1774 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1775 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1776
1777 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1778 ext4_msg(sb, KERN_ERR, "old and new quota "
1779 "format mixing");
1780 return 0;
1781 }
1782
1783 if (!sbi->s_jquota_fmt) {
1784 ext4_msg(sb, KERN_ERR, "journaled quota format "
1785 "not specified");
1786 return 0;
1787 }
1788 } else {
1789 if (sbi->s_jquota_fmt) {
1790 ext4_msg(sb, KERN_ERR, "journaled quota format "
1791 "specified with no journaling "
1792 "enabled");
1793 return 0;
1794 }
1795 }
1796 #endif
1797 return 1;
1798 }
1799
1800 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1801 int read_only)
1802 {
1803 struct ext4_sb_info *sbi = EXT4_SB(sb);
1804 int res = 0;
1805
1806 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1807 ext4_msg(sb, KERN_ERR, "revision level too high, "
1808 "forcing read-only mode");
1809 res = MS_RDONLY;
1810 }
1811 if (read_only)
1812 return res;
1813 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1814 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1815 "running e2fsck is recommended");
1816 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1817 ext4_msg(sb, KERN_WARNING,
1818 "warning: mounting fs with errors, "
1819 "running e2fsck is recommended");
1820 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1821 le16_to_cpu(es->s_mnt_count) >=
1822 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1823 ext4_msg(sb, KERN_WARNING,
1824 "warning: maximal mount count reached, "
1825 "running e2fsck is recommended");
1826 else if (le32_to_cpu(es->s_checkinterval) &&
1827 (le32_to_cpu(es->s_lastcheck) +
1828 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1829 ext4_msg(sb, KERN_WARNING,
1830 "warning: checktime reached, "
1831 "running e2fsck is recommended");
1832 if (!sbi->s_journal)
1833 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1834 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1835 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1836 le16_add_cpu(&es->s_mnt_count, 1);
1837 es->s_mtime = cpu_to_le32(get_seconds());
1838 ext4_update_dynamic_rev(sb);
1839 if (sbi->s_journal)
1840 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1841
1842 ext4_commit_super(sb, 1);
1843 if (test_opt(sb, DEBUG))
1844 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1845 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1846 sb->s_blocksize,
1847 sbi->s_groups_count,
1848 EXT4_BLOCKS_PER_GROUP(sb),
1849 EXT4_INODES_PER_GROUP(sb),
1850 sbi->s_mount_opt);
1851
1852 return res;
1853 }
1854
1855 static int ext4_fill_flex_info(struct super_block *sb)
1856 {
1857 struct ext4_sb_info *sbi = EXT4_SB(sb);
1858 struct ext4_group_desc *gdp = NULL;
1859 ext4_group_t flex_group_count;
1860 ext4_group_t flex_group;
1861 int groups_per_flex = 0;
1862 size_t size;
1863 int i;
1864
1865 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1866 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1867
1868 if (groups_per_flex < 2) {
1869 sbi->s_log_groups_per_flex = 0;
1870 return 1;
1871 }
1872
1873 /* We allocate both existing and potentially added groups */
1874 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1875 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1876 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1877 size = flex_group_count * sizeof(struct flex_groups);
1878 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1879 if (sbi->s_flex_groups == NULL) {
1880 sbi->s_flex_groups = vmalloc(size);
1881 if (sbi->s_flex_groups)
1882 memset(sbi->s_flex_groups, 0, size);
1883 }
1884 if (sbi->s_flex_groups == NULL) {
1885 ext4_msg(sb, KERN_ERR, "not enough memory for "
1886 "%u flex groups", flex_group_count);
1887 goto failed;
1888 }
1889
1890 for (i = 0; i < sbi->s_groups_count; i++) {
1891 gdp = ext4_get_group_desc(sb, i, NULL);
1892
1893 flex_group = ext4_flex_group(sbi, i);
1894 atomic_add(ext4_free_inodes_count(sb, gdp),
1895 &sbi->s_flex_groups[flex_group].free_inodes);
1896 atomic_add(ext4_free_blks_count(sb, gdp),
1897 &sbi->s_flex_groups[flex_group].free_blocks);
1898 atomic_add(ext4_used_dirs_count(sb, gdp),
1899 &sbi->s_flex_groups[flex_group].used_dirs);
1900 }
1901
1902 return 1;
1903 failed:
1904 return 0;
1905 }
1906
1907 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1908 struct ext4_group_desc *gdp)
1909 {
1910 __u16 crc = 0;
1911
1912 if (sbi->s_es->s_feature_ro_compat &
1913 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1914 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1915 __le32 le_group = cpu_to_le32(block_group);
1916
1917 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1918 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1919 crc = crc16(crc, (__u8 *)gdp, offset);
1920 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1921 /* for checksum of struct ext4_group_desc do the rest...*/
1922 if ((sbi->s_es->s_feature_incompat &
1923 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1924 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1925 crc = crc16(crc, (__u8 *)gdp + offset,
1926 le16_to_cpu(sbi->s_es->s_desc_size) -
1927 offset);
1928 }
1929
1930 return cpu_to_le16(crc);
1931 }
1932
1933 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1934 struct ext4_group_desc *gdp)
1935 {
1936 if ((sbi->s_es->s_feature_ro_compat &
1937 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1938 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1939 return 0;
1940
1941 return 1;
1942 }
1943
1944 /* Called at mount-time, super-block is locked */
1945 static int ext4_check_descriptors(struct super_block *sb)
1946 {
1947 struct ext4_sb_info *sbi = EXT4_SB(sb);
1948 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1949 ext4_fsblk_t last_block;
1950 ext4_fsblk_t block_bitmap;
1951 ext4_fsblk_t inode_bitmap;
1952 ext4_fsblk_t inode_table;
1953 int flexbg_flag = 0;
1954 ext4_group_t i;
1955
1956 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1957 flexbg_flag = 1;
1958
1959 ext4_debug("Checking group descriptors");
1960
1961 for (i = 0; i < sbi->s_groups_count; i++) {
1962 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1963
1964 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1965 last_block = ext4_blocks_count(sbi->s_es) - 1;
1966 else
1967 last_block = first_block +
1968 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1969
1970 block_bitmap = ext4_block_bitmap(sb, gdp);
1971 if (block_bitmap < first_block || block_bitmap > last_block) {
1972 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1973 "Block bitmap for group %u not in group "
1974 "(block %llu)!", i, block_bitmap);
1975 return 0;
1976 }
1977 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1978 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1979 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1980 "Inode bitmap for group %u not in group "
1981 "(block %llu)!", i, inode_bitmap);
1982 return 0;
1983 }
1984 inode_table = ext4_inode_table(sb, gdp);
1985 if (inode_table < first_block ||
1986 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1987 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1988 "Inode table for group %u not in group "
1989 "(block %llu)!", i, inode_table);
1990 return 0;
1991 }
1992 ext4_lock_group(sb, i);
1993 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1994 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1995 "Checksum for group %u failed (%u!=%u)",
1996 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1997 gdp)), le16_to_cpu(gdp->bg_checksum));
1998 if (!(sb->s_flags & MS_RDONLY)) {
1999 ext4_unlock_group(sb, i);
2000 return 0;
2001 }
2002 }
2003 ext4_unlock_group(sb, i);
2004 if (!flexbg_flag)
2005 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2006 }
2007
2008 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2009 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2010 return 1;
2011 }
2012
2013 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2014 * the superblock) which were deleted from all directories, but held open by
2015 * a process at the time of a crash. We walk the list and try to delete these
2016 * inodes at recovery time (only with a read-write filesystem).
2017 *
2018 * In order to keep the orphan inode chain consistent during traversal (in
2019 * case of crash during recovery), we link each inode into the superblock
2020 * orphan list_head and handle it the same way as an inode deletion during
2021 * normal operation (which journals the operations for us).
2022 *
2023 * We only do an iget() and an iput() on each inode, which is very safe if we
2024 * accidentally point at an in-use or already deleted inode. The worst that
2025 * can happen in this case is that we get a "bit already cleared" message from
2026 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2027 * e2fsck was run on this filesystem, and it must have already done the orphan
2028 * inode cleanup for us, so we can safely abort without any further action.
2029 */
2030 static void ext4_orphan_cleanup(struct super_block *sb,
2031 struct ext4_super_block *es)
2032 {
2033 unsigned int s_flags = sb->s_flags;
2034 int nr_orphans = 0, nr_truncates = 0;
2035 #ifdef CONFIG_QUOTA
2036 int i;
2037 #endif
2038 if (!es->s_last_orphan) {
2039 jbd_debug(4, "no orphan inodes to clean up\n");
2040 return;
2041 }
2042
2043 if (bdev_read_only(sb->s_bdev)) {
2044 ext4_msg(sb, KERN_ERR, "write access "
2045 "unavailable, skipping orphan cleanup");
2046 return;
2047 }
2048
2049 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2050 if (es->s_last_orphan)
2051 jbd_debug(1, "Errors on filesystem, "
2052 "clearing orphan list.\n");
2053 es->s_last_orphan = 0;
2054 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2055 return;
2056 }
2057
2058 if (s_flags & MS_RDONLY) {
2059 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2060 sb->s_flags &= ~MS_RDONLY;
2061 }
2062 #ifdef CONFIG_QUOTA
2063 /* Needed for iput() to work correctly and not trash data */
2064 sb->s_flags |= MS_ACTIVE;
2065 /* Turn on quotas so that they are updated correctly */
2066 for (i = 0; i < MAXQUOTAS; i++) {
2067 if (EXT4_SB(sb)->s_qf_names[i]) {
2068 int ret = ext4_quota_on_mount(sb, i);
2069 if (ret < 0)
2070 ext4_msg(sb, KERN_ERR,
2071 "Cannot turn on journaled "
2072 "quota: error %d", ret);
2073 }
2074 }
2075 #endif
2076
2077 while (es->s_last_orphan) {
2078 struct inode *inode;
2079
2080 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2081 if (IS_ERR(inode)) {
2082 es->s_last_orphan = 0;
2083 break;
2084 }
2085
2086 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2087 dquot_initialize(inode);
2088 if (inode->i_nlink) {
2089 ext4_msg(sb, KERN_DEBUG,
2090 "%s: truncating inode %lu to %lld bytes",
2091 __func__, inode->i_ino, inode->i_size);
2092 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2093 inode->i_ino, inode->i_size);
2094 ext4_truncate(inode);
2095 nr_truncates++;
2096 } else {
2097 ext4_msg(sb, KERN_DEBUG,
2098 "%s: deleting unreferenced inode %lu",
2099 __func__, inode->i_ino);
2100 jbd_debug(2, "deleting unreferenced inode %lu\n",
2101 inode->i_ino);
2102 nr_orphans++;
2103 }
2104 iput(inode); /* The delete magic happens here! */
2105 }
2106
2107 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2108
2109 if (nr_orphans)
2110 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2111 PLURAL(nr_orphans));
2112 if (nr_truncates)
2113 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2114 PLURAL(nr_truncates));
2115 #ifdef CONFIG_QUOTA
2116 /* Turn quotas off */
2117 for (i = 0; i < MAXQUOTAS; i++) {
2118 if (sb_dqopt(sb)->files[i])
2119 dquot_quota_off(sb, i);
2120 }
2121 #endif
2122 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2123 }
2124
2125 /*
2126 * Maximal extent format file size.
2127 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2128 * extent format containers, within a sector_t, and within i_blocks
2129 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2130 * so that won't be a limiting factor.
2131 *
2132 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2133 */
2134 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2135 {
2136 loff_t res;
2137 loff_t upper_limit = MAX_LFS_FILESIZE;
2138
2139 /* small i_blocks in vfs inode? */
2140 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2141 /*
2142 * CONFIG_LBDAF is not enabled implies the inode
2143 * i_block represent total blocks in 512 bytes
2144 * 32 == size of vfs inode i_blocks * 8
2145 */
2146 upper_limit = (1LL << 32) - 1;
2147
2148 /* total blocks in file system block size */
2149 upper_limit >>= (blkbits - 9);
2150 upper_limit <<= blkbits;
2151 }
2152
2153 /* 32-bit extent-start container, ee_block */
2154 res = 1LL << 32;
2155 res <<= blkbits;
2156 res -= 1;
2157
2158 /* Sanity check against vm- & vfs- imposed limits */
2159 if (res > upper_limit)
2160 res = upper_limit;
2161
2162 return res;
2163 }
2164
2165 /*
2166 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2167 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2168 * We need to be 1 filesystem block less than the 2^48 sector limit.
2169 */
2170 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2171 {
2172 loff_t res = EXT4_NDIR_BLOCKS;
2173 int meta_blocks;
2174 loff_t upper_limit;
2175 /* This is calculated to be the largest file size for a dense, block
2176 * mapped file such that the file's total number of 512-byte sectors,
2177 * including data and all indirect blocks, does not exceed (2^48 - 1).
2178 *
2179 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2180 * number of 512-byte sectors of the file.
2181 */
2182
2183 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2184 /*
2185 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2186 * the inode i_block field represents total file blocks in
2187 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2188 */
2189 upper_limit = (1LL << 32) - 1;
2190
2191 /* total blocks in file system block size */
2192 upper_limit >>= (bits - 9);
2193
2194 } else {
2195 /*
2196 * We use 48 bit ext4_inode i_blocks
2197 * With EXT4_HUGE_FILE_FL set the i_blocks
2198 * represent total number of blocks in
2199 * file system block size
2200 */
2201 upper_limit = (1LL << 48) - 1;
2202
2203 }
2204
2205 /* indirect blocks */
2206 meta_blocks = 1;
2207 /* double indirect blocks */
2208 meta_blocks += 1 + (1LL << (bits-2));
2209 /* tripple indirect blocks */
2210 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2211
2212 upper_limit -= meta_blocks;
2213 upper_limit <<= bits;
2214
2215 res += 1LL << (bits-2);
2216 res += 1LL << (2*(bits-2));
2217 res += 1LL << (3*(bits-2));
2218 res <<= bits;
2219 if (res > upper_limit)
2220 res = upper_limit;
2221
2222 if (res > MAX_LFS_FILESIZE)
2223 res = MAX_LFS_FILESIZE;
2224
2225 return res;
2226 }
2227
2228 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2229 ext4_fsblk_t logical_sb_block, int nr)
2230 {
2231 struct ext4_sb_info *sbi = EXT4_SB(sb);
2232 ext4_group_t bg, first_meta_bg;
2233 int has_super = 0;
2234
2235 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2236
2237 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2238 nr < first_meta_bg)
2239 return logical_sb_block + nr + 1;
2240 bg = sbi->s_desc_per_block * nr;
2241 if (ext4_bg_has_super(sb, bg))
2242 has_super = 1;
2243
2244 return (has_super + ext4_group_first_block_no(sb, bg));
2245 }
2246
2247 /**
2248 * ext4_get_stripe_size: Get the stripe size.
2249 * @sbi: In memory super block info
2250 *
2251 * If we have specified it via mount option, then
2252 * use the mount option value. If the value specified at mount time is
2253 * greater than the blocks per group use the super block value.
2254 * If the super block value is greater than blocks per group return 0.
2255 * Allocator needs it be less than blocks per group.
2256 *
2257 */
2258 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2259 {
2260 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2261 unsigned long stripe_width =
2262 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2263
2264 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2265 return sbi->s_stripe;
2266
2267 if (stripe_width <= sbi->s_blocks_per_group)
2268 return stripe_width;
2269
2270 if (stride <= sbi->s_blocks_per_group)
2271 return stride;
2272
2273 return 0;
2274 }
2275
2276 /* sysfs supprt */
2277
2278 struct ext4_attr {
2279 struct attribute attr;
2280 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2281 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2282 const char *, size_t);
2283 int offset;
2284 };
2285
2286 static int parse_strtoul(const char *buf,
2287 unsigned long max, unsigned long *value)
2288 {
2289 char *endp;
2290
2291 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2292 endp = skip_spaces(endp);
2293 if (*endp || *value > max)
2294 return -EINVAL;
2295
2296 return 0;
2297 }
2298
2299 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2300 struct ext4_sb_info *sbi,
2301 char *buf)
2302 {
2303 return snprintf(buf, PAGE_SIZE, "%llu\n",
2304 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2305 }
2306
2307 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2308 struct ext4_sb_info *sbi, char *buf)
2309 {
2310 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2311
2312 if (!sb->s_bdev->bd_part)
2313 return snprintf(buf, PAGE_SIZE, "0\n");
2314 return snprintf(buf, PAGE_SIZE, "%lu\n",
2315 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2316 sbi->s_sectors_written_start) >> 1);
2317 }
2318
2319 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2320 struct ext4_sb_info *sbi, char *buf)
2321 {
2322 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2323
2324 if (!sb->s_bdev->bd_part)
2325 return snprintf(buf, PAGE_SIZE, "0\n");
2326 return snprintf(buf, PAGE_SIZE, "%llu\n",
2327 (unsigned long long)(sbi->s_kbytes_written +
2328 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2329 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2330 }
2331
2332 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2333 struct ext4_sb_info *sbi,
2334 const char *buf, size_t count)
2335 {
2336 unsigned long t;
2337
2338 if (parse_strtoul(buf, 0x40000000, &t))
2339 return -EINVAL;
2340
2341 if (!is_power_of_2(t))
2342 return -EINVAL;
2343
2344 sbi->s_inode_readahead_blks = t;
2345 return count;
2346 }
2347
2348 static ssize_t sbi_ui_show(struct ext4_attr *a,
2349 struct ext4_sb_info *sbi, char *buf)
2350 {
2351 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2352
2353 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2354 }
2355
2356 static ssize_t sbi_ui_store(struct ext4_attr *a,
2357 struct ext4_sb_info *sbi,
2358 const char *buf, size_t count)
2359 {
2360 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2361 unsigned long t;
2362
2363 if (parse_strtoul(buf, 0xffffffff, &t))
2364 return -EINVAL;
2365 *ui = t;
2366 return count;
2367 }
2368
2369 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2370 static struct ext4_attr ext4_attr_##_name = { \
2371 .attr = {.name = __stringify(_name), .mode = _mode }, \
2372 .show = _show, \
2373 .store = _store, \
2374 .offset = offsetof(struct ext4_sb_info, _elname), \
2375 }
2376 #define EXT4_ATTR(name, mode, show, store) \
2377 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2378
2379 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2380 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2381 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2382 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2383 #define ATTR_LIST(name) &ext4_attr_##name.attr
2384
2385 EXT4_RO_ATTR(delayed_allocation_blocks);
2386 EXT4_RO_ATTR(session_write_kbytes);
2387 EXT4_RO_ATTR(lifetime_write_kbytes);
2388 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2389 inode_readahead_blks_store, s_inode_readahead_blks);
2390 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2391 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2392 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2393 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2394 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2395 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2396 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2397 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2398
2399 static struct attribute *ext4_attrs[] = {
2400 ATTR_LIST(delayed_allocation_blocks),
2401 ATTR_LIST(session_write_kbytes),
2402 ATTR_LIST(lifetime_write_kbytes),
2403 ATTR_LIST(inode_readahead_blks),
2404 ATTR_LIST(inode_goal),
2405 ATTR_LIST(mb_stats),
2406 ATTR_LIST(mb_max_to_scan),
2407 ATTR_LIST(mb_min_to_scan),
2408 ATTR_LIST(mb_order2_req),
2409 ATTR_LIST(mb_stream_req),
2410 ATTR_LIST(mb_group_prealloc),
2411 ATTR_LIST(max_writeback_mb_bump),
2412 NULL,
2413 };
2414
2415 static ssize_t ext4_attr_show(struct kobject *kobj,
2416 struct attribute *attr, char *buf)
2417 {
2418 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2419 s_kobj);
2420 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2421
2422 return a->show ? a->show(a, sbi, buf) : 0;
2423 }
2424
2425 static ssize_t ext4_attr_store(struct kobject *kobj,
2426 struct attribute *attr,
2427 const char *buf, size_t len)
2428 {
2429 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2430 s_kobj);
2431 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2432
2433 return a->store ? a->store(a, sbi, buf, len) : 0;
2434 }
2435
2436 static void ext4_sb_release(struct kobject *kobj)
2437 {
2438 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2439 s_kobj);
2440 complete(&sbi->s_kobj_unregister);
2441 }
2442
2443
2444 static const struct sysfs_ops ext4_attr_ops = {
2445 .show = ext4_attr_show,
2446 .store = ext4_attr_store,
2447 };
2448
2449 static struct kobj_type ext4_ktype = {
2450 .default_attrs = ext4_attrs,
2451 .sysfs_ops = &ext4_attr_ops,
2452 .release = ext4_sb_release,
2453 };
2454
2455 /*
2456 * Check whether this filesystem can be mounted based on
2457 * the features present and the RDONLY/RDWR mount requested.
2458 * Returns 1 if this filesystem can be mounted as requested,
2459 * 0 if it cannot be.
2460 */
2461 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2462 {
2463 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2464 ext4_msg(sb, KERN_ERR,
2465 "Couldn't mount because of "
2466 "unsupported optional features (%x)",
2467 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2468 ~EXT4_FEATURE_INCOMPAT_SUPP));
2469 return 0;
2470 }
2471
2472 if (readonly)
2473 return 1;
2474
2475 /* Check that feature set is OK for a read-write mount */
2476 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2477 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2478 "unsupported optional features (%x)",
2479 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2480 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2481 return 0;
2482 }
2483 /*
2484 * Large file size enabled file system can only be mounted
2485 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2486 */
2487 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2488 if (sizeof(blkcnt_t) < sizeof(u64)) {
2489 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2490 "cannot be mounted RDWR without "
2491 "CONFIG_LBDAF");
2492 return 0;
2493 }
2494 }
2495 return 1;
2496 }
2497
2498 /*
2499 * This function is called once a day if we have errors logged
2500 * on the file system
2501 */
2502 static void print_daily_error_info(unsigned long arg)
2503 {
2504 struct super_block *sb = (struct super_block *) arg;
2505 struct ext4_sb_info *sbi;
2506 struct ext4_super_block *es;
2507
2508 sbi = EXT4_SB(sb);
2509 es = sbi->s_es;
2510
2511 if (es->s_error_count)
2512 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2513 le32_to_cpu(es->s_error_count));
2514 if (es->s_first_error_time) {
2515 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2516 sb->s_id, le32_to_cpu(es->s_first_error_time),
2517 (int) sizeof(es->s_first_error_func),
2518 es->s_first_error_func,
2519 le32_to_cpu(es->s_first_error_line));
2520 if (es->s_first_error_ino)
2521 printk(": inode %u",
2522 le32_to_cpu(es->s_first_error_ino));
2523 if (es->s_first_error_block)
2524 printk(": block %llu", (unsigned long long)
2525 le64_to_cpu(es->s_first_error_block));
2526 printk("\n");
2527 }
2528 if (es->s_last_error_time) {
2529 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2530 sb->s_id, le32_to_cpu(es->s_last_error_time),
2531 (int) sizeof(es->s_last_error_func),
2532 es->s_last_error_func,
2533 le32_to_cpu(es->s_last_error_line));
2534 if (es->s_last_error_ino)
2535 printk(": inode %u",
2536 le32_to_cpu(es->s_last_error_ino));
2537 if (es->s_last_error_block)
2538 printk(": block %llu", (unsigned long long)
2539 le64_to_cpu(es->s_last_error_block));
2540 printk("\n");
2541 }
2542 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2543 }
2544
2545 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2546 __releases(kernel_lock)
2547 __acquires(kernel_lock)
2548 {
2549 char *orig_data = kstrdup(data, GFP_KERNEL);
2550 struct buffer_head *bh;
2551 struct ext4_super_block *es = NULL;
2552 struct ext4_sb_info *sbi;
2553 ext4_fsblk_t block;
2554 ext4_fsblk_t sb_block = get_sb_block(&data);
2555 ext4_fsblk_t logical_sb_block;
2556 unsigned long offset = 0;
2557 unsigned long journal_devnum = 0;
2558 unsigned long def_mount_opts;
2559 struct inode *root;
2560 char *cp;
2561 const char *descr;
2562 int ret = -ENOMEM;
2563 int blocksize;
2564 unsigned int db_count;
2565 unsigned int i;
2566 int needs_recovery, has_huge_files;
2567 __u64 blocks_count;
2568 int err;
2569 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2570
2571 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2572 if (!sbi)
2573 goto out_free_orig;
2574
2575 sbi->s_blockgroup_lock =
2576 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2577 if (!sbi->s_blockgroup_lock) {
2578 kfree(sbi);
2579 goto out_free_orig;
2580 }
2581 sb->s_fs_info = sbi;
2582 sbi->s_mount_opt = 0;
2583 sbi->s_resuid = EXT4_DEF_RESUID;
2584 sbi->s_resgid = EXT4_DEF_RESGID;
2585 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2586 sbi->s_sb_block = sb_block;
2587 if (sb->s_bdev->bd_part)
2588 sbi->s_sectors_written_start =
2589 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2590
2591 unlock_kernel();
2592
2593 /* Cleanup superblock name */
2594 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2595 *cp = '!';
2596
2597 ret = -EINVAL;
2598 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2599 if (!blocksize) {
2600 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2601 goto out_fail;
2602 }
2603
2604 /*
2605 * The ext4 superblock will not be buffer aligned for other than 1kB
2606 * block sizes. We need to calculate the offset from buffer start.
2607 */
2608 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2609 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2610 offset = do_div(logical_sb_block, blocksize);
2611 } else {
2612 logical_sb_block = sb_block;
2613 }
2614
2615 if (!(bh = sb_bread(sb, logical_sb_block))) {
2616 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2617 goto out_fail;
2618 }
2619 /*
2620 * Note: s_es must be initialized as soon as possible because
2621 * some ext4 macro-instructions depend on its value
2622 */
2623 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2624 sbi->s_es = es;
2625 sb->s_magic = le16_to_cpu(es->s_magic);
2626 if (sb->s_magic != EXT4_SUPER_MAGIC)
2627 goto cantfind_ext4;
2628 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2629
2630 /* Set defaults before we parse the mount options */
2631 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2632 if (def_mount_opts & EXT4_DEFM_DEBUG)
2633 set_opt(sbi->s_mount_opt, DEBUG);
2634 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2635 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2636 "2.6.38");
2637 set_opt(sbi->s_mount_opt, GRPID);
2638 }
2639 if (def_mount_opts & EXT4_DEFM_UID16)
2640 set_opt(sbi->s_mount_opt, NO_UID32);
2641 #ifdef CONFIG_EXT4_FS_XATTR
2642 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2643 set_opt(sbi->s_mount_opt, XATTR_USER);
2644 #endif
2645 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2646 if (def_mount_opts & EXT4_DEFM_ACL)
2647 set_opt(sbi->s_mount_opt, POSIX_ACL);
2648 #endif
2649 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2650 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2651 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2652 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2653 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2654 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2655
2656 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2657 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2658 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2659 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2660 else
2661 set_opt(sbi->s_mount_opt, ERRORS_RO);
2662 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
2663 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
2664 if (def_mount_opts & EXT4_DEFM_DISCARD)
2665 set_opt(sbi->s_mount_opt, DISCARD);
2666
2667 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2668 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2669 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2670 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2671 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2672
2673 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
2674 set_opt(sbi->s_mount_opt, BARRIER);
2675
2676 /*
2677 * enable delayed allocation by default
2678 * Use -o nodelalloc to turn it off
2679 */
2680 if (!IS_EXT3_SB(sb) &&
2681 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
2682 set_opt(sbi->s_mount_opt, DELALLOC);
2683
2684 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
2685 &journal_devnum, &journal_ioprio, NULL, 0)) {
2686 ext4_msg(sb, KERN_WARNING,
2687 "failed to parse options in superblock: %s",
2688 sbi->s_es->s_mount_opts);
2689 }
2690 if (!parse_options((char *) data, sb, &journal_devnum,
2691 &journal_ioprio, NULL, 0))
2692 goto failed_mount;
2693
2694 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2695 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2696
2697 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2698 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2699 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2700 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2701 ext4_msg(sb, KERN_WARNING,
2702 "feature flags set on rev 0 fs, "
2703 "running e2fsck is recommended");
2704
2705 /*
2706 * Check feature flags regardless of the revision level, since we
2707 * previously didn't change the revision level when setting the flags,
2708 * so there is a chance incompat flags are set on a rev 0 filesystem.
2709 */
2710 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2711 goto failed_mount;
2712
2713 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2714
2715 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2716 blocksize > EXT4_MAX_BLOCK_SIZE) {
2717 ext4_msg(sb, KERN_ERR,
2718 "Unsupported filesystem blocksize %d", blocksize);
2719 goto failed_mount;
2720 }
2721
2722 if (sb->s_blocksize != blocksize) {
2723 /* Validate the filesystem blocksize */
2724 if (!sb_set_blocksize(sb, blocksize)) {
2725 ext4_msg(sb, KERN_ERR, "bad block size %d",
2726 blocksize);
2727 goto failed_mount;
2728 }
2729
2730 brelse(bh);
2731 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2732 offset = do_div(logical_sb_block, blocksize);
2733 bh = sb_bread(sb, logical_sb_block);
2734 if (!bh) {
2735 ext4_msg(sb, KERN_ERR,
2736 "Can't read superblock on 2nd try");
2737 goto failed_mount;
2738 }
2739 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2740 sbi->s_es = es;
2741 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2742 ext4_msg(sb, KERN_ERR,
2743 "Magic mismatch, very weird!");
2744 goto failed_mount;
2745 }
2746 }
2747
2748 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2749 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2750 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2751 has_huge_files);
2752 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2753
2754 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2755 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2756 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2757 } else {
2758 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2759 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2760 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2761 (!is_power_of_2(sbi->s_inode_size)) ||
2762 (sbi->s_inode_size > blocksize)) {
2763 ext4_msg(sb, KERN_ERR,
2764 "unsupported inode size: %d",
2765 sbi->s_inode_size);
2766 goto failed_mount;
2767 }
2768 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2769 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2770 }
2771
2772 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2773 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2774 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2775 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2776 !is_power_of_2(sbi->s_desc_size)) {
2777 ext4_msg(sb, KERN_ERR,
2778 "unsupported descriptor size %lu",
2779 sbi->s_desc_size);
2780 goto failed_mount;
2781 }
2782 } else
2783 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2784
2785 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2786 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2787 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2788 goto cantfind_ext4;
2789
2790 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2791 if (sbi->s_inodes_per_block == 0)
2792 goto cantfind_ext4;
2793 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2794 sbi->s_inodes_per_block;
2795 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2796 sbi->s_sbh = bh;
2797 sbi->s_mount_state = le16_to_cpu(es->s_state);
2798 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2799 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2800
2801 for (i = 0; i < 4; i++)
2802 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2803 sbi->s_def_hash_version = es->s_def_hash_version;
2804 i = le32_to_cpu(es->s_flags);
2805 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2806 sbi->s_hash_unsigned = 3;
2807 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2808 #ifdef __CHAR_UNSIGNED__
2809 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2810 sbi->s_hash_unsigned = 3;
2811 #else
2812 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2813 #endif
2814 sb->s_dirt = 1;
2815 }
2816
2817 if (sbi->s_blocks_per_group > blocksize * 8) {
2818 ext4_msg(sb, KERN_ERR,
2819 "#blocks per group too big: %lu",
2820 sbi->s_blocks_per_group);
2821 goto failed_mount;
2822 }
2823 if (sbi->s_inodes_per_group > blocksize * 8) {
2824 ext4_msg(sb, KERN_ERR,
2825 "#inodes per group too big: %lu",
2826 sbi->s_inodes_per_group);
2827 goto failed_mount;
2828 }
2829
2830 /*
2831 * Test whether we have more sectors than will fit in sector_t,
2832 * and whether the max offset is addressable by the page cache.
2833 */
2834 if ((ext4_blocks_count(es) >
2835 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2836 (ext4_blocks_count(es) >
2837 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2838 ext4_msg(sb, KERN_ERR, "filesystem"
2839 " too large to mount safely on this system");
2840 if (sizeof(sector_t) < 8)
2841 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2842 ret = -EFBIG;
2843 goto failed_mount;
2844 }
2845
2846 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2847 goto cantfind_ext4;
2848
2849 /* check blocks count against device size */
2850 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2851 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2852 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2853 "exceeds size of device (%llu blocks)",
2854 ext4_blocks_count(es), blocks_count);
2855 goto failed_mount;
2856 }
2857
2858 /*
2859 * It makes no sense for the first data block to be beyond the end
2860 * of the filesystem.
2861 */
2862 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2863 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2864 "block %u is beyond end of filesystem (%llu)",
2865 le32_to_cpu(es->s_first_data_block),
2866 ext4_blocks_count(es));
2867 goto failed_mount;
2868 }
2869 blocks_count = (ext4_blocks_count(es) -
2870 le32_to_cpu(es->s_first_data_block) +
2871 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2872 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2873 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2874 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2875 "(block count %llu, first data block %u, "
2876 "blocks per group %lu)", sbi->s_groups_count,
2877 ext4_blocks_count(es),
2878 le32_to_cpu(es->s_first_data_block),
2879 EXT4_BLOCKS_PER_GROUP(sb));
2880 goto failed_mount;
2881 }
2882 sbi->s_groups_count = blocks_count;
2883 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2884 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2885 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2886 EXT4_DESC_PER_BLOCK(sb);
2887 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2888 GFP_KERNEL);
2889 if (sbi->s_group_desc == NULL) {
2890 ext4_msg(sb, KERN_ERR, "not enough memory");
2891 goto failed_mount;
2892 }
2893
2894 #ifdef CONFIG_PROC_FS
2895 if (ext4_proc_root)
2896 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2897 #endif
2898
2899 bgl_lock_init(sbi->s_blockgroup_lock);
2900
2901 for (i = 0; i < db_count; i++) {
2902 block = descriptor_loc(sb, logical_sb_block, i);
2903 sbi->s_group_desc[i] = sb_bread(sb, block);
2904 if (!sbi->s_group_desc[i]) {
2905 ext4_msg(sb, KERN_ERR,
2906 "can't read group descriptor %d", i);
2907 db_count = i;
2908 goto failed_mount2;
2909 }
2910 }
2911 if (!ext4_check_descriptors(sb)) {
2912 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2913 goto failed_mount2;
2914 }
2915 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2916 if (!ext4_fill_flex_info(sb)) {
2917 ext4_msg(sb, KERN_ERR,
2918 "unable to initialize "
2919 "flex_bg meta info!");
2920 goto failed_mount2;
2921 }
2922
2923 sbi->s_gdb_count = db_count;
2924 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2925 spin_lock_init(&sbi->s_next_gen_lock);
2926
2927 sbi->s_stripe = ext4_get_stripe_size(sbi);
2928 sbi->s_max_writeback_mb_bump = 128;
2929
2930 /*
2931 * set up enough so that it can read an inode
2932 */
2933 if (!test_opt(sb, NOLOAD) &&
2934 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2935 sb->s_op = &ext4_sops;
2936 else
2937 sb->s_op = &ext4_nojournal_sops;
2938 sb->s_export_op = &ext4_export_ops;
2939 sb->s_xattr = ext4_xattr_handlers;
2940 #ifdef CONFIG_QUOTA
2941 sb->s_qcop = &ext4_qctl_operations;
2942 sb->dq_op = &ext4_quota_operations;
2943 #endif
2944 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2945 mutex_init(&sbi->s_orphan_lock);
2946 mutex_init(&sbi->s_resize_lock);
2947
2948 sb->s_root = NULL;
2949
2950 needs_recovery = (es->s_last_orphan != 0 ||
2951 EXT4_HAS_INCOMPAT_FEATURE(sb,
2952 EXT4_FEATURE_INCOMPAT_RECOVER));
2953
2954 /*
2955 * The first inode we look at is the journal inode. Don't try
2956 * root first: it may be modified in the journal!
2957 */
2958 if (!test_opt(sb, NOLOAD) &&
2959 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2960 if (ext4_load_journal(sb, es, journal_devnum))
2961 goto failed_mount3;
2962 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2963 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2964 ext4_msg(sb, KERN_ERR, "required journal recovery "
2965 "suppressed and not mounted read-only");
2966 goto failed_mount_wq;
2967 } else {
2968 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2969 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2970 sbi->s_journal = NULL;
2971 needs_recovery = 0;
2972 goto no_journal;
2973 }
2974
2975 if (ext4_blocks_count(es) > 0xffffffffULL &&
2976 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2977 JBD2_FEATURE_INCOMPAT_64BIT)) {
2978 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2979 goto failed_mount_wq;
2980 }
2981
2982 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2983 jbd2_journal_set_features(sbi->s_journal,
2984 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2985 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2986 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2987 jbd2_journal_set_features(sbi->s_journal,
2988 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2989 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2990 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2991 } else {
2992 jbd2_journal_clear_features(sbi->s_journal,
2993 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2994 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2995 }
2996
2997 /* We have now updated the journal if required, so we can
2998 * validate the data journaling mode. */
2999 switch (test_opt(sb, DATA_FLAGS)) {
3000 case 0:
3001 /* No mode set, assume a default based on the journal
3002 * capabilities: ORDERED_DATA if the journal can
3003 * cope, else JOURNAL_DATA
3004 */
3005 if (jbd2_journal_check_available_features
3006 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3007 set_opt(sbi->s_mount_opt, ORDERED_DATA);
3008 else
3009 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
3010 break;
3011
3012 case EXT4_MOUNT_ORDERED_DATA:
3013 case EXT4_MOUNT_WRITEBACK_DATA:
3014 if (!jbd2_journal_check_available_features
3015 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3016 ext4_msg(sb, KERN_ERR, "Journal does not support "
3017 "requested data journaling mode");
3018 goto failed_mount_wq;
3019 }
3020 default:
3021 break;
3022 }
3023 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3024
3025 no_journal:
3026 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3027 ext4_count_free_blocks(sb));
3028 if (!err)
3029 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3030 ext4_count_free_inodes(sb));
3031 if (!err)
3032 err = percpu_counter_init(&sbi->s_dirs_counter,
3033 ext4_count_dirs(sb));
3034 if (!err)
3035 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3036 if (err) {
3037 ext4_msg(sb, KERN_ERR, "insufficient memory");
3038 goto failed_mount_wq;
3039 }
3040
3041 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3042 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3043 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3044 goto failed_mount_wq;
3045 }
3046
3047 /*
3048 * The jbd2_journal_load will have done any necessary log recovery,
3049 * so we can safely mount the rest of the filesystem now.
3050 */
3051
3052 root = ext4_iget(sb, EXT4_ROOT_INO);
3053 if (IS_ERR(root)) {
3054 ext4_msg(sb, KERN_ERR, "get root inode failed");
3055 ret = PTR_ERR(root);
3056 goto failed_mount4;
3057 }
3058 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3059 iput(root);
3060 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3061 goto failed_mount4;
3062 }
3063 sb->s_root = d_alloc_root(root);
3064 if (!sb->s_root) {
3065 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3066 iput(root);
3067 ret = -ENOMEM;
3068 goto failed_mount4;
3069 }
3070
3071 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3072
3073 /* determine the minimum size of new large inodes, if present */
3074 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3075 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3076 EXT4_GOOD_OLD_INODE_SIZE;
3077 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3078 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3079 if (sbi->s_want_extra_isize <
3080 le16_to_cpu(es->s_want_extra_isize))
3081 sbi->s_want_extra_isize =
3082 le16_to_cpu(es->s_want_extra_isize);
3083 if (sbi->s_want_extra_isize <
3084 le16_to_cpu(es->s_min_extra_isize))
3085 sbi->s_want_extra_isize =
3086 le16_to_cpu(es->s_min_extra_isize);
3087 }
3088 }
3089 /* Check if enough inode space is available */
3090 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3091 sbi->s_inode_size) {
3092 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3093 EXT4_GOOD_OLD_INODE_SIZE;
3094 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3095 "available");
3096 }
3097
3098 if (test_opt(sb, DELALLOC) &&
3099 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3100 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3101 "requested data journaling mode");
3102 clear_opt(sbi->s_mount_opt, DELALLOC);
3103 }
3104 if (test_opt(sb, DIOREAD_NOLOCK)) {
3105 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3106 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3107 "option - requested data journaling mode");
3108 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3109 }
3110 if (sb->s_blocksize < PAGE_SIZE) {
3111 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3112 "option - block size is too small");
3113 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3114 }
3115 }
3116
3117 err = ext4_setup_system_zone(sb);
3118 if (err) {
3119 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3120 "zone (%d)", err);
3121 goto failed_mount4;
3122 }
3123
3124 ext4_ext_init(sb);
3125 err = ext4_mb_init(sb, needs_recovery);
3126 if (err) {
3127 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3128 err);
3129 goto failed_mount4;
3130 }
3131
3132 sbi->s_kobj.kset = ext4_kset;
3133 init_completion(&sbi->s_kobj_unregister);
3134 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3135 "%s", sb->s_id);
3136 if (err) {
3137 ext4_mb_release(sb);
3138 ext4_ext_release(sb);
3139 goto failed_mount4;
3140 };
3141
3142 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3143 ext4_orphan_cleanup(sb, es);
3144 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3145 if (needs_recovery) {
3146 ext4_msg(sb, KERN_INFO, "recovery complete");
3147 ext4_mark_recovery_complete(sb, es);
3148 }
3149 if (EXT4_SB(sb)->s_journal) {
3150 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3151 descr = " journalled data mode";
3152 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3153 descr = " ordered data mode";
3154 else
3155 descr = " writeback data mode";
3156 } else
3157 descr = "out journal";
3158
3159 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3160 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3161 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3162
3163 init_timer(&sbi->s_err_report);
3164 sbi->s_err_report.function = print_daily_error_info;
3165 sbi->s_err_report.data = (unsigned long) sb;
3166 if (es->s_error_count)
3167 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3168
3169 lock_kernel();
3170 kfree(orig_data);
3171 return 0;
3172
3173 cantfind_ext4:
3174 if (!silent)
3175 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3176 goto failed_mount;
3177
3178 failed_mount4:
3179 ext4_msg(sb, KERN_ERR, "mount failed");
3180 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3181 failed_mount_wq:
3182 ext4_release_system_zone(sb);
3183 if (sbi->s_journal) {
3184 jbd2_journal_destroy(sbi->s_journal);
3185 sbi->s_journal = NULL;
3186 }
3187 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3188 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3189 percpu_counter_destroy(&sbi->s_dirs_counter);
3190 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3191 failed_mount3:
3192 if (sbi->s_flex_groups) {
3193 if (is_vmalloc_addr(sbi->s_flex_groups))
3194 vfree(sbi->s_flex_groups);
3195 else
3196 kfree(sbi->s_flex_groups);
3197 }
3198 failed_mount2:
3199 for (i = 0; i < db_count; i++)
3200 brelse(sbi->s_group_desc[i]);
3201 kfree(sbi->s_group_desc);
3202 failed_mount:
3203 if (sbi->s_proc) {
3204 remove_proc_entry(sb->s_id, ext4_proc_root);
3205 }
3206 #ifdef CONFIG_QUOTA
3207 for (i = 0; i < MAXQUOTAS; i++)
3208 kfree(sbi->s_qf_names[i]);
3209 #endif
3210 ext4_blkdev_remove(sbi);
3211 brelse(bh);
3212 out_fail:
3213 sb->s_fs_info = NULL;
3214 kfree(sbi->s_blockgroup_lock);
3215 kfree(sbi);
3216 lock_kernel();
3217 out_free_orig:
3218 kfree(orig_data);
3219 return ret;
3220 }
3221
3222 /*
3223 * Setup any per-fs journal parameters now. We'll do this both on
3224 * initial mount, once the journal has been initialised but before we've
3225 * done any recovery; and again on any subsequent remount.
3226 */
3227 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3228 {
3229 struct ext4_sb_info *sbi = EXT4_SB(sb);
3230
3231 journal->j_commit_interval = sbi->s_commit_interval;
3232 journal->j_min_batch_time = sbi->s_min_batch_time;
3233 journal->j_max_batch_time = sbi->s_max_batch_time;
3234
3235 write_lock(&journal->j_state_lock);
3236 if (test_opt(sb, BARRIER))
3237 journal->j_flags |= JBD2_BARRIER;
3238 else
3239 journal->j_flags &= ~JBD2_BARRIER;
3240 if (test_opt(sb, DATA_ERR_ABORT))
3241 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3242 else
3243 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3244 write_unlock(&journal->j_state_lock);
3245 }
3246
3247 static journal_t *ext4_get_journal(struct super_block *sb,
3248 unsigned int journal_inum)
3249 {
3250 struct inode *journal_inode;
3251 journal_t *journal;
3252
3253 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3254
3255 /* First, test for the existence of a valid inode on disk. Bad
3256 * things happen if we iget() an unused inode, as the subsequent
3257 * iput() will try to delete it. */
3258
3259 journal_inode = ext4_iget(sb, journal_inum);
3260 if (IS_ERR(journal_inode)) {
3261 ext4_msg(sb, KERN_ERR, "no journal found");
3262 return NULL;
3263 }
3264 if (!journal_inode->i_nlink) {
3265 make_bad_inode(journal_inode);
3266 iput(journal_inode);
3267 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3268 return NULL;
3269 }
3270
3271 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3272 journal_inode, journal_inode->i_size);
3273 if (!S_ISREG(journal_inode->i_mode)) {
3274 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3275 iput(journal_inode);
3276 return NULL;
3277 }
3278
3279 journal = jbd2_journal_init_inode(journal_inode);
3280 if (!journal) {
3281 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3282 iput(journal_inode);
3283 return NULL;
3284 }
3285 journal->j_private = sb;
3286 ext4_init_journal_params(sb, journal);
3287 return journal;
3288 }
3289
3290 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3291 dev_t j_dev)
3292 {
3293 struct buffer_head *bh;
3294 journal_t *journal;
3295 ext4_fsblk_t start;
3296 ext4_fsblk_t len;
3297 int hblock, blocksize;
3298 ext4_fsblk_t sb_block;
3299 unsigned long offset;
3300 struct ext4_super_block *es;
3301 struct block_device *bdev;
3302
3303 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3304
3305 bdev = ext4_blkdev_get(j_dev, sb);
3306 if (bdev == NULL)
3307 return NULL;
3308
3309 if (bd_claim(bdev, sb)) {
3310 ext4_msg(sb, KERN_ERR,
3311 "failed to claim external journal device");
3312 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3313 return NULL;
3314 }
3315
3316 blocksize = sb->s_blocksize;
3317 hblock = bdev_logical_block_size(bdev);
3318 if (blocksize < hblock) {
3319 ext4_msg(sb, KERN_ERR,
3320 "blocksize too small for journal device");
3321 goto out_bdev;
3322 }
3323
3324 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3325 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3326 set_blocksize(bdev, blocksize);
3327 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3328 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3329 "external journal");
3330 goto out_bdev;
3331 }
3332
3333 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3334 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3335 !(le32_to_cpu(es->s_feature_incompat) &
3336 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3337 ext4_msg(sb, KERN_ERR, "external journal has "
3338 "bad superblock");
3339 brelse(bh);
3340 goto out_bdev;
3341 }
3342
3343 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3344 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3345 brelse(bh);
3346 goto out_bdev;
3347 }
3348
3349 len = ext4_blocks_count(es);
3350 start = sb_block + 1;
3351 brelse(bh); /* we're done with the superblock */
3352
3353 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3354 start, len, blocksize);
3355 if (!journal) {
3356 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3357 goto out_bdev;
3358 }
3359 journal->j_private = sb;
3360 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3361 wait_on_buffer(journal->j_sb_buffer);
3362 if (!buffer_uptodate(journal->j_sb_buffer)) {
3363 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3364 goto out_journal;
3365 }
3366 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3367 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3368 "user (unsupported) - %d",
3369 be32_to_cpu(journal->j_superblock->s_nr_users));
3370 goto out_journal;
3371 }
3372 EXT4_SB(sb)->journal_bdev = bdev;
3373 ext4_init_journal_params(sb, journal);
3374 return journal;
3375
3376 out_journal:
3377 jbd2_journal_destroy(journal);
3378 out_bdev:
3379 ext4_blkdev_put(bdev);
3380 return NULL;
3381 }
3382
3383 static int ext4_load_journal(struct super_block *sb,
3384 struct ext4_super_block *es,
3385 unsigned long journal_devnum)
3386 {
3387 journal_t *journal;
3388 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3389 dev_t journal_dev;
3390 int err = 0;
3391 int really_read_only;
3392
3393 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3394
3395 if (journal_devnum &&
3396 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3397 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3398 "numbers have changed");
3399 journal_dev = new_decode_dev(journal_devnum);
3400 } else
3401 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3402
3403 really_read_only = bdev_read_only(sb->s_bdev);
3404
3405 /*
3406 * Are we loading a blank journal or performing recovery after a
3407 * crash? For recovery, we need to check in advance whether we
3408 * can get read-write access to the device.
3409 */
3410 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3411 if (sb->s_flags & MS_RDONLY) {
3412 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3413 "required on readonly filesystem");
3414 if (really_read_only) {
3415 ext4_msg(sb, KERN_ERR, "write access "
3416 "unavailable, cannot proceed");
3417 return -EROFS;
3418 }
3419 ext4_msg(sb, KERN_INFO, "write access will "
3420 "be enabled during recovery");
3421 }
3422 }
3423
3424 if (journal_inum && journal_dev) {
3425 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3426 "and inode journals!");
3427 return -EINVAL;
3428 }
3429
3430 if (journal_inum) {
3431 if (!(journal = ext4_get_journal(sb, journal_inum)))
3432 return -EINVAL;
3433 } else {
3434 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3435 return -EINVAL;
3436 }
3437
3438 if (!(journal->j_flags & JBD2_BARRIER))
3439 ext4_msg(sb, KERN_INFO, "barriers disabled");
3440
3441 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3442 err = jbd2_journal_update_format(journal);
3443 if (err) {
3444 ext4_msg(sb, KERN_ERR, "error updating journal");
3445 jbd2_journal_destroy(journal);
3446 return err;
3447 }
3448 }
3449
3450 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3451 err = jbd2_journal_wipe(journal, !really_read_only);
3452 if (!err) {
3453 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3454 if (save)
3455 memcpy(save, ((char *) es) +
3456 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3457 err = jbd2_journal_load(journal);
3458 if (save)
3459 memcpy(((char *) es) + EXT4_S_ERR_START,
3460 save, EXT4_S_ERR_LEN);
3461 kfree(save);
3462 }
3463
3464 if (err) {
3465 ext4_msg(sb, KERN_ERR, "error loading journal");
3466 jbd2_journal_destroy(journal);
3467 return err;
3468 }
3469
3470 EXT4_SB(sb)->s_journal = journal;
3471 ext4_clear_journal_err(sb, es);
3472
3473 if (journal_devnum &&
3474 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3475 es->s_journal_dev = cpu_to_le32(journal_devnum);
3476
3477 /* Make sure we flush the recovery flag to disk. */
3478 ext4_commit_super(sb, 1);
3479 }
3480
3481 return 0;
3482 }
3483
3484 static int ext4_commit_super(struct super_block *sb, int sync)
3485 {
3486 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3487 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3488 int error = 0;
3489
3490 if (!sbh)
3491 return error;
3492 if (buffer_write_io_error(sbh)) {
3493 /*
3494 * Oh, dear. A previous attempt to write the
3495 * superblock failed. This could happen because the
3496 * USB device was yanked out. Or it could happen to
3497 * be a transient write error and maybe the block will
3498 * be remapped. Nothing we can do but to retry the
3499 * write and hope for the best.
3500 */
3501 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3502 "superblock detected");
3503 clear_buffer_write_io_error(sbh);
3504 set_buffer_uptodate(sbh);
3505 }
3506 /*
3507 * If the file system is mounted read-only, don't update the
3508 * superblock write time. This avoids updating the superblock
3509 * write time when we are mounting the root file system
3510 * read/only but we need to replay the journal; at that point,
3511 * for people who are east of GMT and who make their clock
3512 * tick in localtime for Windows bug-for-bug compatibility,
3513 * the clock is set in the future, and this will cause e2fsck
3514 * to complain and force a full file system check.
3515 */
3516 if (!(sb->s_flags & MS_RDONLY))
3517 es->s_wtime = cpu_to_le32(get_seconds());
3518 if (sb->s_bdev->bd_part)
3519 es->s_kbytes_written =
3520 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3521 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3522 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3523 else
3524 es->s_kbytes_written =
3525 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
3526 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3527 &EXT4_SB(sb)->s_freeblocks_counter));
3528 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3529 &EXT4_SB(sb)->s_freeinodes_counter));
3530 sb->s_dirt = 0;
3531 BUFFER_TRACE(sbh, "marking dirty");
3532 mark_buffer_dirty(sbh);
3533 if (sync) {
3534 error = sync_dirty_buffer(sbh);
3535 if (error)
3536 return error;
3537
3538 error = buffer_write_io_error(sbh);
3539 if (error) {
3540 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3541 "superblock");
3542 clear_buffer_write_io_error(sbh);
3543 set_buffer_uptodate(sbh);
3544 }
3545 }
3546 return error;
3547 }
3548
3549 /*
3550 * Have we just finished recovery? If so, and if we are mounting (or
3551 * remounting) the filesystem readonly, then we will end up with a
3552 * consistent fs on disk. Record that fact.
3553 */
3554 static void ext4_mark_recovery_complete(struct super_block *sb,
3555 struct ext4_super_block *es)
3556 {
3557 journal_t *journal = EXT4_SB(sb)->s_journal;
3558
3559 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3560 BUG_ON(journal != NULL);
3561 return;
3562 }
3563 jbd2_journal_lock_updates(journal);
3564 if (jbd2_journal_flush(journal) < 0)
3565 goto out;
3566
3567 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3568 sb->s_flags & MS_RDONLY) {
3569 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3570 ext4_commit_super(sb, 1);
3571 }
3572
3573 out:
3574 jbd2_journal_unlock_updates(journal);
3575 }
3576
3577 /*
3578 * If we are mounting (or read-write remounting) a filesystem whose journal
3579 * has recorded an error from a previous lifetime, move that error to the
3580 * main filesystem now.
3581 */
3582 static void ext4_clear_journal_err(struct super_block *sb,
3583 struct ext4_super_block *es)
3584 {
3585 journal_t *journal;
3586 int j_errno;
3587 const char *errstr;
3588
3589 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3590
3591 journal = EXT4_SB(sb)->s_journal;
3592
3593 /*
3594 * Now check for any error status which may have been recorded in the
3595 * journal by a prior ext4_error() or ext4_abort()
3596 */
3597
3598 j_errno = jbd2_journal_errno(journal);
3599 if (j_errno) {
3600 char nbuf[16];
3601
3602 errstr = ext4_decode_error(sb, j_errno, nbuf);
3603 ext4_warning(sb, "Filesystem error recorded "
3604 "from previous mount: %s", errstr);
3605 ext4_warning(sb, "Marking fs in need of filesystem check.");
3606
3607 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3608 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3609 ext4_commit_super(sb, 1);
3610
3611 jbd2_journal_clear_err(journal);
3612 }
3613 }
3614
3615 /*
3616 * Force the running and committing transactions to commit,
3617 * and wait on the commit.
3618 */
3619 int ext4_force_commit(struct super_block *sb)
3620 {
3621 journal_t *journal;
3622 int ret = 0;
3623
3624 if (sb->s_flags & MS_RDONLY)
3625 return 0;
3626
3627 journal = EXT4_SB(sb)->s_journal;
3628 if (journal) {
3629 vfs_check_frozen(sb, SB_FREEZE_TRANS);
3630 ret = ext4_journal_force_commit(journal);
3631 }
3632
3633 return ret;
3634 }
3635
3636 static void ext4_write_super(struct super_block *sb)
3637 {
3638 lock_super(sb);
3639 ext4_commit_super(sb, 1);
3640 unlock_super(sb);
3641 }
3642
3643 static int ext4_sync_fs(struct super_block *sb, int wait)
3644 {
3645 int ret = 0;
3646 tid_t target;
3647 struct ext4_sb_info *sbi = EXT4_SB(sb);
3648
3649 trace_ext4_sync_fs(sb, wait);
3650 flush_workqueue(sbi->dio_unwritten_wq);
3651 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3652 if (wait)
3653 jbd2_log_wait_commit(sbi->s_journal, target);
3654 }
3655 return ret;
3656 }
3657
3658 /*
3659 * LVM calls this function before a (read-only) snapshot is created. This
3660 * gives us a chance to flush the journal completely and mark the fs clean.
3661 */
3662 static int ext4_freeze(struct super_block *sb)
3663 {
3664 int error = 0;
3665 journal_t *journal;
3666
3667 if (sb->s_flags & MS_RDONLY)
3668 return 0;
3669
3670 journal = EXT4_SB(sb)->s_journal;
3671
3672 /* Now we set up the journal barrier. */
3673 jbd2_journal_lock_updates(journal);
3674
3675 /*
3676 * Don't clear the needs_recovery flag if we failed to flush
3677 * the journal.
3678 */
3679 error = jbd2_journal_flush(journal);
3680 if (error < 0)
3681 goto out;
3682
3683 /* Journal blocked and flushed, clear needs_recovery flag. */
3684 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3685 error = ext4_commit_super(sb, 1);
3686 out:
3687 /* we rely on s_frozen to stop further updates */
3688 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3689 return error;
3690 }
3691
3692 /*
3693 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3694 * flag here, even though the filesystem is not technically dirty yet.
3695 */
3696 static int ext4_unfreeze(struct super_block *sb)
3697 {
3698 if (sb->s_flags & MS_RDONLY)
3699 return 0;
3700
3701 lock_super(sb);
3702 /* Reset the needs_recovery flag before the fs is unlocked. */
3703 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3704 ext4_commit_super(sb, 1);
3705 unlock_super(sb);
3706 return 0;
3707 }
3708
3709 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3710 {
3711 struct ext4_super_block *es;
3712 struct ext4_sb_info *sbi = EXT4_SB(sb);
3713 ext4_fsblk_t n_blocks_count = 0;
3714 unsigned long old_sb_flags;
3715 struct ext4_mount_options old_opts;
3716 int enable_quota = 0;
3717 ext4_group_t g;
3718 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3719 int err;
3720 #ifdef CONFIG_QUOTA
3721 int i;
3722 #endif
3723 char *orig_data = kstrdup(data, GFP_KERNEL);
3724
3725 lock_kernel();
3726
3727 /* Store the original options */
3728 lock_super(sb);
3729 old_sb_flags = sb->s_flags;
3730 old_opts.s_mount_opt = sbi->s_mount_opt;
3731 old_opts.s_resuid = sbi->s_resuid;
3732 old_opts.s_resgid = sbi->s_resgid;
3733 old_opts.s_commit_interval = sbi->s_commit_interval;
3734 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3735 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3736 #ifdef CONFIG_QUOTA
3737 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3738 for (i = 0; i < MAXQUOTAS; i++)
3739 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3740 #endif
3741 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3742 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3743
3744 /*
3745 * Allow the "check" option to be passed as a remount option.
3746 */
3747 if (!parse_options(data, sb, NULL, &journal_ioprio,
3748 &n_blocks_count, 1)) {
3749 err = -EINVAL;
3750 goto restore_opts;
3751 }
3752
3753 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3754 ext4_abort(sb, "Abort forced by user");
3755
3756 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3757 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3758
3759 es = sbi->s_es;
3760
3761 if (sbi->s_journal) {
3762 ext4_init_journal_params(sb, sbi->s_journal);
3763 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3764 }
3765
3766 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3767 n_blocks_count > ext4_blocks_count(es)) {
3768 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3769 err = -EROFS;
3770 goto restore_opts;
3771 }
3772
3773 if (*flags & MS_RDONLY) {
3774 err = dquot_suspend(sb, -1);
3775 if (err < 0)
3776 goto restore_opts;
3777
3778 /*
3779 * First of all, the unconditional stuff we have to do
3780 * to disable replay of the journal when we next remount
3781 */
3782 sb->s_flags |= MS_RDONLY;
3783
3784 /*
3785 * OK, test if we are remounting a valid rw partition
3786 * readonly, and if so set the rdonly flag and then
3787 * mark the partition as valid again.
3788 */
3789 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3790 (sbi->s_mount_state & EXT4_VALID_FS))
3791 es->s_state = cpu_to_le16(sbi->s_mount_state);
3792
3793 if (sbi->s_journal)
3794 ext4_mark_recovery_complete(sb, es);
3795 } else {
3796 /* Make sure we can mount this feature set readwrite */
3797 if (!ext4_feature_set_ok(sb, 0)) {
3798 err = -EROFS;
3799 goto restore_opts;
3800 }
3801 /*
3802 * Make sure the group descriptor checksums
3803 * are sane. If they aren't, refuse to remount r/w.
3804 */
3805 for (g = 0; g < sbi->s_groups_count; g++) {
3806 struct ext4_group_desc *gdp =
3807 ext4_get_group_desc(sb, g, NULL);
3808
3809 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3810 ext4_msg(sb, KERN_ERR,
3811 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3812 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3813 le16_to_cpu(gdp->bg_checksum));
3814 err = -EINVAL;
3815 goto restore_opts;
3816 }
3817 }
3818
3819 /*
3820 * If we have an unprocessed orphan list hanging
3821 * around from a previously readonly bdev mount,
3822 * require a full umount/remount for now.
3823 */
3824 if (es->s_last_orphan) {
3825 ext4_msg(sb, KERN_WARNING, "Couldn't "
3826 "remount RDWR because of unprocessed "
3827 "orphan inode list. Please "
3828 "umount/remount instead");
3829 err = -EINVAL;
3830 goto restore_opts;
3831 }
3832
3833 /*
3834 * Mounting a RDONLY partition read-write, so reread
3835 * and store the current valid flag. (It may have
3836 * been changed by e2fsck since we originally mounted
3837 * the partition.)
3838 */
3839 if (sbi->s_journal)
3840 ext4_clear_journal_err(sb, es);
3841 sbi->s_mount_state = le16_to_cpu(es->s_state);
3842 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3843 goto restore_opts;
3844 if (!ext4_setup_super(sb, es, 0))
3845 sb->s_flags &= ~MS_RDONLY;
3846 enable_quota = 1;
3847 }
3848 }
3849 ext4_setup_system_zone(sb);
3850 if (sbi->s_journal == NULL)
3851 ext4_commit_super(sb, 1);
3852
3853 #ifdef CONFIG_QUOTA
3854 /* Release old quota file names */
3855 for (i = 0; i < MAXQUOTAS; i++)
3856 if (old_opts.s_qf_names[i] &&
3857 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3858 kfree(old_opts.s_qf_names[i]);
3859 #endif
3860 unlock_super(sb);
3861 unlock_kernel();
3862 if (enable_quota)
3863 dquot_resume(sb, -1);
3864
3865 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
3866 kfree(orig_data);
3867 return 0;
3868
3869 restore_opts:
3870 sb->s_flags = old_sb_flags;
3871 sbi->s_mount_opt = old_opts.s_mount_opt;
3872 sbi->s_resuid = old_opts.s_resuid;
3873 sbi->s_resgid = old_opts.s_resgid;
3874 sbi->s_commit_interval = old_opts.s_commit_interval;
3875 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3876 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3877 #ifdef CONFIG_QUOTA
3878 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3879 for (i = 0; i < MAXQUOTAS; i++) {
3880 if (sbi->s_qf_names[i] &&
3881 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3882 kfree(sbi->s_qf_names[i]);
3883 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3884 }
3885 #endif
3886 unlock_super(sb);
3887 unlock_kernel();
3888 kfree(orig_data);
3889 return err;
3890 }
3891
3892 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3893 {
3894 struct super_block *sb = dentry->d_sb;
3895 struct ext4_sb_info *sbi = EXT4_SB(sb);
3896 struct ext4_super_block *es = sbi->s_es;
3897 u64 fsid;
3898
3899 if (test_opt(sb, MINIX_DF)) {
3900 sbi->s_overhead_last = 0;
3901 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3902 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3903 ext4_fsblk_t overhead = 0;
3904
3905 /*
3906 * Compute the overhead (FS structures). This is constant
3907 * for a given filesystem unless the number of block groups
3908 * changes so we cache the previous value until it does.
3909 */
3910
3911 /*
3912 * All of the blocks before first_data_block are
3913 * overhead
3914 */
3915 overhead = le32_to_cpu(es->s_first_data_block);
3916
3917 /*
3918 * Add the overhead attributed to the superblock and
3919 * block group descriptors. If the sparse superblocks
3920 * feature is turned on, then not all groups have this.
3921 */
3922 for (i = 0; i < ngroups; i++) {
3923 overhead += ext4_bg_has_super(sb, i) +
3924 ext4_bg_num_gdb(sb, i);
3925 cond_resched();
3926 }
3927
3928 /*
3929 * Every block group has an inode bitmap, a block
3930 * bitmap, and an inode table.
3931 */
3932 overhead += ngroups * (2 + sbi->s_itb_per_group);
3933 sbi->s_overhead_last = overhead;
3934 smp_wmb();
3935 sbi->s_blocks_last = ext4_blocks_count(es);
3936 }
3937
3938 buf->f_type = EXT4_SUPER_MAGIC;
3939 buf->f_bsize = sb->s_blocksize;
3940 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3941 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3942 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3943 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3944 if (buf->f_bfree < ext4_r_blocks_count(es))
3945 buf->f_bavail = 0;
3946 buf->f_files = le32_to_cpu(es->s_inodes_count);
3947 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3948 buf->f_namelen = EXT4_NAME_LEN;
3949 fsid = le64_to_cpup((void *)es->s_uuid) ^
3950 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3951 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3952 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3953
3954 return 0;
3955 }
3956
3957 /* Helper function for writing quotas on sync - we need to start transaction
3958 * before quota file is locked for write. Otherwise the are possible deadlocks:
3959 * Process 1 Process 2
3960 * ext4_create() quota_sync()
3961 * jbd2_journal_start() write_dquot()
3962 * dquot_initialize() down(dqio_mutex)
3963 * down(dqio_mutex) jbd2_journal_start()
3964 *
3965 */
3966
3967 #ifdef CONFIG_QUOTA
3968
3969 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3970 {
3971 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3972 }
3973
3974 static int ext4_write_dquot(struct dquot *dquot)
3975 {
3976 int ret, err;
3977 handle_t *handle;
3978 struct inode *inode;
3979
3980 inode = dquot_to_inode(dquot);
3981 handle = ext4_journal_start(inode,
3982 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3983 if (IS_ERR(handle))
3984 return PTR_ERR(handle);
3985 ret = dquot_commit(dquot);
3986 err = ext4_journal_stop(handle);
3987 if (!ret)
3988 ret = err;
3989 return ret;
3990 }
3991
3992 static int ext4_acquire_dquot(struct dquot *dquot)
3993 {
3994 int ret, err;
3995 handle_t *handle;
3996
3997 handle = ext4_journal_start(dquot_to_inode(dquot),
3998 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
3999 if (IS_ERR(handle))
4000 return PTR_ERR(handle);
4001 ret = dquot_acquire(dquot);
4002 err = ext4_journal_stop(handle);
4003 if (!ret)
4004 ret = err;
4005 return ret;
4006 }
4007
4008 static int ext4_release_dquot(struct dquot *dquot)
4009 {
4010 int ret, err;
4011 handle_t *handle;
4012
4013 handle = ext4_journal_start(dquot_to_inode(dquot),
4014 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4015 if (IS_ERR(handle)) {
4016 /* Release dquot anyway to avoid endless cycle in dqput() */
4017 dquot_release(dquot);
4018 return PTR_ERR(handle);
4019 }
4020 ret = dquot_release(dquot);
4021 err = ext4_journal_stop(handle);
4022 if (!ret)
4023 ret = err;
4024 return ret;
4025 }
4026
4027 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4028 {
4029 /* Are we journaling quotas? */
4030 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4031 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4032 dquot_mark_dquot_dirty(dquot);
4033 return ext4_write_dquot(dquot);
4034 } else {
4035 return dquot_mark_dquot_dirty(dquot);
4036 }
4037 }
4038
4039 static int ext4_write_info(struct super_block *sb, int type)
4040 {
4041 int ret, err;
4042 handle_t *handle;
4043
4044 /* Data block + inode block */
4045 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4046 if (IS_ERR(handle))
4047 return PTR_ERR(handle);
4048 ret = dquot_commit_info(sb, type);
4049 err = ext4_journal_stop(handle);
4050 if (!ret)
4051 ret = err;
4052 return ret;
4053 }
4054
4055 /*
4056 * Turn on quotas during mount time - we need to find
4057 * the quota file and such...
4058 */
4059 static int ext4_quota_on_mount(struct super_block *sb, int type)
4060 {
4061 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4062 EXT4_SB(sb)->s_jquota_fmt, type);
4063 }
4064
4065 /*
4066 * Standard function to be called on quota_on
4067 */
4068 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4069 char *name)
4070 {
4071 int err;
4072 struct path path;
4073
4074 if (!test_opt(sb, QUOTA))
4075 return -EINVAL;
4076
4077 err = kern_path(name, LOOKUP_FOLLOW, &path);
4078 if (err)
4079 return err;
4080
4081 /* Quotafile not on the same filesystem? */
4082 if (path.mnt->mnt_sb != sb) {
4083 path_put(&path);
4084 return -EXDEV;
4085 }
4086 /* Journaling quota? */
4087 if (EXT4_SB(sb)->s_qf_names[type]) {
4088 /* Quotafile not in fs root? */
4089 if (path.dentry->d_parent != sb->s_root)
4090 ext4_msg(sb, KERN_WARNING,
4091 "Quota file not on filesystem root. "
4092 "Journaled quota will not work");
4093 }
4094
4095 /*
4096 * When we journal data on quota file, we have to flush journal to see
4097 * all updates to the file when we bypass pagecache...
4098 */
4099 if (EXT4_SB(sb)->s_journal &&
4100 ext4_should_journal_data(path.dentry->d_inode)) {
4101 /*
4102 * We don't need to lock updates but journal_flush() could
4103 * otherwise be livelocked...
4104 */
4105 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4106 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4107 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4108 if (err) {
4109 path_put(&path);
4110 return err;
4111 }
4112 }
4113
4114 err = dquot_quota_on_path(sb, type, format_id, &path);
4115 path_put(&path);
4116 return err;
4117 }
4118
4119 static int ext4_quota_off(struct super_block *sb, int type)
4120 {
4121 /* Force all delayed allocation blocks to be allocated */
4122 if (test_opt(sb, DELALLOC)) {
4123 down_read(&sb->s_umount);
4124 sync_filesystem(sb);
4125 up_read(&sb->s_umount);
4126 }
4127
4128 return dquot_quota_off(sb, type);
4129 }
4130
4131 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4132 * acquiring the locks... As quota files are never truncated and quota code
4133 * itself serializes the operations (and noone else should touch the files)
4134 * we don't have to be afraid of races */
4135 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4136 size_t len, loff_t off)
4137 {
4138 struct inode *inode = sb_dqopt(sb)->files[type];
4139 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4140 int err = 0;
4141 int offset = off & (sb->s_blocksize - 1);
4142 int tocopy;
4143 size_t toread;
4144 struct buffer_head *bh;
4145 loff_t i_size = i_size_read(inode);
4146
4147 if (off > i_size)
4148 return 0;
4149 if (off+len > i_size)
4150 len = i_size-off;
4151 toread = len;
4152 while (toread > 0) {
4153 tocopy = sb->s_blocksize - offset < toread ?
4154 sb->s_blocksize - offset : toread;
4155 bh = ext4_bread(NULL, inode, blk, 0, &err);
4156 if (err)
4157 return err;
4158 if (!bh) /* A hole? */
4159 memset(data, 0, tocopy);
4160 else
4161 memcpy(data, bh->b_data+offset, tocopy);
4162 brelse(bh);
4163 offset = 0;
4164 toread -= tocopy;
4165 data += tocopy;
4166 blk++;
4167 }
4168 return len;
4169 }
4170
4171 /* Write to quotafile (we know the transaction is already started and has
4172 * enough credits) */
4173 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4174 const char *data, size_t len, loff_t off)
4175 {
4176 struct inode *inode = sb_dqopt(sb)->files[type];
4177 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4178 int err = 0;
4179 int offset = off & (sb->s_blocksize - 1);
4180 struct buffer_head *bh;
4181 handle_t *handle = journal_current_handle();
4182
4183 if (EXT4_SB(sb)->s_journal && !handle) {
4184 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4185 " cancelled because transaction is not started",
4186 (unsigned long long)off, (unsigned long long)len);
4187 return -EIO;
4188 }
4189 /*
4190 * Since we account only one data block in transaction credits,
4191 * then it is impossible to cross a block boundary.
4192 */
4193 if (sb->s_blocksize - offset < len) {
4194 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4195 " cancelled because not block aligned",
4196 (unsigned long long)off, (unsigned long long)len);
4197 return -EIO;
4198 }
4199
4200 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4201 bh = ext4_bread(handle, inode, blk, 1, &err);
4202 if (!bh)
4203 goto out;
4204 err = ext4_journal_get_write_access(handle, bh);
4205 if (err) {
4206 brelse(bh);
4207 goto out;
4208 }
4209 lock_buffer(bh);
4210 memcpy(bh->b_data+offset, data, len);
4211 flush_dcache_page(bh->b_page);
4212 unlock_buffer(bh);
4213 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4214 brelse(bh);
4215 out:
4216 if (err) {
4217 mutex_unlock(&inode->i_mutex);
4218 return err;
4219 }
4220 if (inode->i_size < off + len) {
4221 i_size_write(inode, off + len);
4222 EXT4_I(inode)->i_disksize = inode->i_size;
4223 }
4224 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4225 ext4_mark_inode_dirty(handle, inode);
4226 mutex_unlock(&inode->i_mutex);
4227 return len;
4228 }
4229
4230 #endif
4231
4232 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4233 const char *dev_name, void *data, struct vfsmount *mnt)
4234 {
4235 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4236 }
4237
4238 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4239 static struct file_system_type ext2_fs_type = {
4240 .owner = THIS_MODULE,
4241 .name = "ext2",
4242 .get_sb = ext4_get_sb,
4243 .kill_sb = kill_block_super,
4244 .fs_flags = FS_REQUIRES_DEV,
4245 };
4246
4247 static inline void register_as_ext2(void)
4248 {
4249 int err = register_filesystem(&ext2_fs_type);
4250 if (err)
4251 printk(KERN_WARNING
4252 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4253 }
4254
4255 static inline void unregister_as_ext2(void)
4256 {
4257 unregister_filesystem(&ext2_fs_type);
4258 }
4259 MODULE_ALIAS("ext2");
4260 #else
4261 static inline void register_as_ext2(void) { }
4262 static inline void unregister_as_ext2(void) { }
4263 #endif
4264
4265 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4266 static inline void register_as_ext3(void)
4267 {
4268 int err = register_filesystem(&ext3_fs_type);
4269 if (err)
4270 printk(KERN_WARNING
4271 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4272 }
4273
4274 static inline void unregister_as_ext3(void)
4275 {
4276 unregister_filesystem(&ext3_fs_type);
4277 }
4278 MODULE_ALIAS("ext3");
4279 #else
4280 static inline void register_as_ext3(void) { }
4281 static inline void unregister_as_ext3(void) { }
4282 #endif
4283
4284 static struct file_system_type ext4_fs_type = {
4285 .owner = THIS_MODULE,
4286 .name = "ext4",
4287 .get_sb = ext4_get_sb,
4288 .kill_sb = kill_block_super,
4289 .fs_flags = FS_REQUIRES_DEV,
4290 };
4291
4292 static int __init init_ext4_fs(void)
4293 {
4294 int err;
4295
4296 ext4_check_flag_values();
4297 err = init_ext4_system_zone();
4298 if (err)
4299 return err;
4300 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4301 if (!ext4_kset)
4302 goto out4;
4303 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4304 err = init_ext4_mballoc();
4305 if (err)
4306 goto out3;
4307
4308 err = init_ext4_xattr();
4309 if (err)
4310 goto out2;
4311 err = init_inodecache();
4312 if (err)
4313 goto out1;
4314 register_as_ext2();
4315 register_as_ext3();
4316 err = register_filesystem(&ext4_fs_type);
4317 if (err)
4318 goto out;
4319 return 0;
4320 out:
4321 unregister_as_ext2();
4322 unregister_as_ext3();
4323 destroy_inodecache();
4324 out1:
4325 exit_ext4_xattr();
4326 out2:
4327 exit_ext4_mballoc();
4328 out3:
4329 remove_proc_entry("fs/ext4", NULL);
4330 kset_unregister(ext4_kset);
4331 out4:
4332 exit_ext4_system_zone();
4333 return err;
4334 }
4335
4336 static void __exit exit_ext4_fs(void)
4337 {
4338 unregister_as_ext2();
4339 unregister_as_ext3();
4340 unregister_filesystem(&ext4_fs_type);
4341 destroy_inodecache();
4342 exit_ext4_xattr();
4343 exit_ext4_mballoc();
4344 remove_proc_entry("fs/ext4", NULL);
4345 kset_unregister(ext4_kset);
4346 exit_ext4_system_zone();
4347 }
4348
4349 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4350 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4351 MODULE_LICENSE("GPL");
4352 module_init(init_ext4_fs)
4353 module_exit(exit_ext4_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree