?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
RT-AC56 3.0.0.4.374.37 core
[tomato.git] / release / src-rt-6.x.4708 / linux / linux-2.6.36 / fs / ext4 / super.c
blob1f0d08c5be740143416daa260467c14c098a8f66
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) && \
75 defined(CONFIG_EXT4_USE_FOR_EXT23)
76 static struct file_system_type ext3_fs_type = {
77 .owner = THIS_MODULE,
78 .name = "ext3",
79 .get_sb = ext4_get_sb,
80 .kill_sb = kill_block_super,
81 .fs_flags = FS_REQUIRES_DEV,
82 };
83 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
84 #else
85 #define IS_EXT3_SB(sb) (0)
86 #endif
87
88 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
89 struct ext4_group_desc *bg)
90 {
91 return le32_to_cpu(bg->bg_block_bitmap_lo) |
92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
93 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
94 }
95
96 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
97 struct ext4_group_desc *bg)
98 {
99 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
101 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
102 }
103
104 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
105 struct ext4_group_desc *bg)
106 {
107 return le32_to_cpu(bg->bg_inode_table_lo) |
108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
109 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
110 }
111
112 __u32 ext4_free_blks_count(struct super_block *sb,
113 struct ext4_group_desc *bg)
114 {
115 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
117 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
118 }
119
120 __u32 ext4_free_inodes_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
122 {
123 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
125 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
126 }
127
128 __u32 ext4_used_dirs_count(struct super_block *sb,
129 struct ext4_group_desc *bg)
130 {
131 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
132 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
133 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
134 }
135
136 __u32 ext4_itable_unused_count(struct super_block *sb,
137 struct ext4_group_desc *bg)
138 {
139 return le16_to_cpu(bg->bg_itable_unused_lo) |
140 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
141 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
142 }
143
144 void ext4_block_bitmap_set(struct super_block *sb,
145 struct ext4_group_desc *bg, ext4_fsblk_t blk)
146 {
147 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
148 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
149 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
150 }
151
152 void ext4_inode_bitmap_set(struct super_block *sb,
153 struct ext4_group_desc *bg, ext4_fsblk_t blk)
154 {
155 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
158 }
159
160 void ext4_inode_table_set(struct super_block *sb,
161 struct ext4_group_desc *bg, ext4_fsblk_t blk)
162 {
163 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
166 }
167
168 void ext4_free_blks_set(struct super_block *sb,
169 struct ext4_group_desc *bg, __u32 count)
170 {
171 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
174 }
175
176 void ext4_free_inodes_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
178 {
179 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
182 }
183
184 void ext4_used_dirs_set(struct super_block *sb,
185 struct ext4_group_desc *bg, __u32 count)
186 {
187 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
188 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
189 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
190 }
191
192 void ext4_itable_unused_set(struct super_block *sb,
193 struct ext4_group_desc *bg, __u32 count)
194 {
195 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
196 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
197 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
198 }
199
200
201 /* Just increment the non-pointer handle value */
202 static handle_t *ext4_get_nojournal(void)
203 {
204 handle_t *handle = current->journal_info;
205 unsigned long ref_cnt = (unsigned long)handle;
206
207 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
208
209 ref_cnt++;
210 handle = (handle_t *)ref_cnt;
211
212 current->journal_info = handle;
213 return handle;
214 }
215
216
217 /* Decrement the non-pointer handle value */
218 static void ext4_put_nojournal(handle_t *handle)
219 {
220 unsigned long ref_cnt = (unsigned long)handle;
221
222 BUG_ON(ref_cnt == 0);
223
224 ref_cnt--;
225 handle = (handle_t *)ref_cnt;
226
227 current->journal_info = handle;
228 }
229
230 /*
231 * Wrappers for jbd2_journal_start/end.
232 *
233 * The only special thing we need to do here is to make sure that all
234 * journal_end calls result in the superblock being marked dirty, so
235 * that sync() will call the filesystem's write_super callback if
236 * appropriate.
237 */
238 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
239 {
240 journal_t *journal;
241
242 if (sb->s_flags & MS_RDONLY)
243 return ERR_PTR(-EROFS);
244
245 vfs_check_frozen(sb, SB_FREEZE_TRANS);
246 /* Special case here: if the journal has aborted behind our
247 * backs (eg. EIO in the commit thread), then we still need to
248 * take the FS itself readonly cleanly. */
249 journal = EXT4_SB(sb)->s_journal;
250 if (journal) {
251 if (is_journal_aborted(journal)) {
252 ext4_abort(sb, "Detected aborted journal");
253 return ERR_PTR(-EROFS);
254 }
255 return jbd2_journal_start(journal, nblocks);
256 }
257 return ext4_get_nojournal();
258 }
259
260 /*
261 * The only special thing we need to do here is to make sure that all
262 * jbd2_journal_stop calls result in the superblock being marked dirty, so
263 * that sync() will call the filesystem's write_super callback if
264 * appropriate.
265 */
266 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
267 {
268 struct super_block *sb;
269 int err;
270 int rc;
271
272 if (!ext4_handle_valid(handle)) {
273 ext4_put_nojournal(handle);
274 return 0;
275 }
276 sb = handle->h_transaction->t_journal->j_private;
277 err = handle->h_err;
278 rc = jbd2_journal_stop(handle);
279
280 if (!err)
281 err = rc;
282 if (err)
283 __ext4_std_error(sb, where, line, err);
284 return err;
285 }
286
287 void ext4_journal_abort_handle(const char *caller, unsigned int line,
288 const char *err_fn, struct buffer_head *bh,
289 handle_t *handle, int err)
290 {
291 char nbuf[16];
292 const char *errstr = ext4_decode_error(NULL, err, nbuf);
293
294 BUG_ON(!ext4_handle_valid(handle));
295
296 if (bh)
297 BUFFER_TRACE(bh, "abort");
298
299 if (!handle->h_err)
300 handle->h_err = err;
301
302 if (is_handle_aborted(handle))
303 return;
304
305 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
306 caller, line, errstr, err_fn);
307
308 jbd2_journal_abort_handle(handle);
309 }
310
311 static void __save_error_info(struct super_block *sb, const char *func,
312 unsigned int line)
313 {
314 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
315
316 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
317 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
318 es->s_last_error_time = cpu_to_le32(get_seconds());
319 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
320 es->s_last_error_line = cpu_to_le32(line);
321 if (!es->s_first_error_time) {
322 es->s_first_error_time = es->s_last_error_time;
323 strncpy(es->s_first_error_func, func,
324 sizeof(es->s_first_error_func));
325 es->s_first_error_line = cpu_to_le32(line);
326 es->s_first_error_ino = es->s_last_error_ino;
327 es->s_first_error_block = es->s_last_error_block;
328 }
329 /*
330 * Start the daily error reporting function if it hasn't been
331 * started already
332 */
333 if (!es->s_error_count)
334 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
335 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
336 }
337
338 static void save_error_info(struct super_block *sb, const char *func,
339 unsigned int line)
340 {
341 __save_error_info(sb, func, line);
342 ext4_commit_super(sb, 1);
343 }
344
345
346 /* Deal with the reporting of failure conditions on a filesystem such as
347 * inconsistencies detected or read IO failures.
348 *
349 * On ext2, we can store the error state of the filesystem in the
350 * superblock. That is not possible on ext4, because we may have other
351 * write ordering constraints on the superblock which prevent us from
352 * writing it out straight away; and given that the journal is about to
353 * be aborted, we can't rely on the current, or future, transactions to
354 * write out the superblock safely.
355 *
356 * We'll just use the jbd2_journal_abort() error code to record an error in
357 * the journal instead. On recovery, the journal will complain about
358 * that error until we've noted it down and cleared it.
359 */
360
361 static void ext4_handle_error(struct super_block *sb)
362 {
363 if (sb->s_flags & MS_RDONLY)
364 return;
365
366 if (!test_opt(sb, ERRORS_CONT)) {
367 journal_t *journal = EXT4_SB(sb)->s_journal;
368
369 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
370 if (journal)
371 jbd2_journal_abort(journal, -EIO);
372 }
373 if (test_opt(sb, ERRORS_RO)) {
374 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
375 sb->s_flags |= MS_RDONLY;
376 }
377 if (test_opt(sb, ERRORS_PANIC))
378 panic("EXT4-fs (device %s): panic forced after error\n",
379 sb->s_id);
380 }
381
382 void __ext4_error(struct super_block *sb, const char *function,
383 unsigned int line, const char *fmt, ...)
384 {
385 va_list args;
386
387 va_start(args, fmt);
388 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
389 sb->s_id, function, line, current->comm);
390 vprintk(fmt, args);
391 printk("\n");
392 va_end(args);
393
394 ext4_handle_error(sb);
395 }
396
397 void ext4_error_inode(struct inode *inode, const char *function,
398 unsigned int line, ext4_fsblk_t block,
399 const char *fmt, ...)
400 {
401 va_list args;
402 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
403
404 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
405 es->s_last_error_block = cpu_to_le64(block);
406 save_error_info(inode->i_sb, function, line);
407 va_start(args, fmt);
408 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
409 inode->i_sb->s_id, function, line, inode->i_ino);
410 if (block)
411 printk("block %llu: ", block);
412 printk("comm %s: ", current->comm);
413 vprintk(fmt, args);
414 printk("\n");
415 va_end(args);
416
417 ext4_handle_error(inode->i_sb);
418 }
419
420 void ext4_error_file(struct file *file, const char *function,
421 unsigned int line, const char *fmt, ...)
422 {
423 va_list args;
424 struct ext4_super_block *es;
425 struct inode *inode = file->f_dentry->d_inode;
426 char pathname[80], *path;
427
428 es = EXT4_SB(inode->i_sb)->s_es;
429 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
430 save_error_info(inode->i_sb, function, line);
431 va_start(args, fmt);
432 path = d_path(&(file->f_path), pathname, sizeof(pathname));
433 if (!path)
434 path = "(unknown)";
435 printk(KERN_CRIT
436 "EXT4-fs error (device %s): %s:%d: inode #%lu "
437 "(comm %s path %s): ",
438 inode->i_sb->s_id, function, line, inode->i_ino,
439 current->comm, path);
440 vprintk(fmt, args);
441 printk("\n");
442 va_end(args);
443
444 ext4_handle_error(inode->i_sb);
445 }
446
447 static const char *ext4_decode_error(struct super_block *sb, int errno,
448 char nbuf[16])
449 {
450 char *errstr = NULL;
451
452 switch (errno) {
453 case -EIO:
454 errstr = "IO failure";
455 break;
456 case -ENOMEM:
457 errstr = "Out of memory";
458 break;
459 case -EROFS:
460 if (!sb || (EXT4_SB(sb)->s_journal &&
461 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
462 errstr = "Journal has aborted";
463 else
464 errstr = "Readonly filesystem";
465 break;
466 default:
467 /* If the caller passed in an extra buffer for unknown
468 * errors, textualise them now. Else we just return
469 * NULL. */
470 if (nbuf) {
471 /* Check for truncated error codes... */
472 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
473 errstr = nbuf;
474 }
475 break;
476 }
477
478 return errstr;
479 }
480
481 /* __ext4_std_error decodes expected errors from journaling functions
482 * automatically and invokes the appropriate error response. */
483
484 void __ext4_std_error(struct super_block *sb, const char *function,
485 unsigned int line, int errno)
486 {
487 char nbuf[16];
488 const char *errstr;
489
490 /* Special case: if the error is EROFS, and we're not already
491 * inside a transaction, then there's really no point in logging
492 * an error. */
493 if (errno == -EROFS && journal_current_handle() == NULL &&
494 (sb->s_flags & MS_RDONLY))
495 return;
496
497 errstr = ext4_decode_error(sb, errno, nbuf);
498 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
499 sb->s_id, function, line, errstr);
500 save_error_info(sb, function, line);
501
502 ext4_handle_error(sb);
503 }
504
505 /*
506 * ext4_abort is a much stronger failure handler than ext4_error. The
507 * abort function may be used to deal with unrecoverable failures such
508 * as journal IO errors or ENOMEM at a critical moment in log management.
509 *
510 * We unconditionally force the filesystem into an ABORT|READONLY state,
511 * unless the error response on the fs has been set to panic in which
512 * case we take the easy way out and panic immediately.
513 */
514
515 void __ext4_abort(struct super_block *sb, const char *function,
516 unsigned int line, const char *fmt, ...)
517 {
518 va_list args;
519
520 save_error_info(sb, function, line);
521 va_start(args, fmt);
522 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
523 function, line);
524 vprintk(fmt, args);
525 printk("\n");
526 va_end(args);
527
528 if ((sb->s_flags & MS_RDONLY) == 0) {
529 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
530 sb->s_flags |= MS_RDONLY;
531 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
532 if (EXT4_SB(sb)->s_journal)
533 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
534 save_error_info(sb, function, line);
535 }
536 if (test_opt(sb, ERRORS_PANIC))
537 panic("EXT4-fs panic from previous error\n");
538 }
539
540 void ext4_msg (struct super_block * sb, const char *prefix,
541 const char *fmt, ...)
542 {
543 va_list args;
544
545 va_start(args, fmt);
546 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
547 vprintk(fmt, args);
548 printk("\n");
549 va_end(args);
550 }
551
552 void __ext4_warning(struct super_block *sb, const char *function,
553 unsigned int line, const char *fmt, ...)
554 {
555 va_list args;
556
557 va_start(args, fmt);
558 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
559 sb->s_id, function, line);
560 vprintk(fmt, args);
561 printk("\n");
562 va_end(args);
563 }
564
565 void __ext4_grp_locked_error(const char *function, unsigned int line,
566 struct super_block *sb, ext4_group_t grp,
567 unsigned long ino, ext4_fsblk_t block,
568 const char *fmt, ...)
569 __releases(bitlock)
570 __acquires(bitlock)
571 {
572 va_list args;
573 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
574
575 es->s_last_error_ino = cpu_to_le32(ino);
576 es->s_last_error_block = cpu_to_le64(block);
577 __save_error_info(sb, function, line);
578 va_start(args, fmt);
579 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
580 sb->s_id, function, line, grp);
581 if (ino)
582 printk("inode %lu: ", ino);
583 if (block)
584 printk("block %llu:", (unsigned long long) block);
585 vprintk(fmt, args);
586 printk("\n");
587 va_end(args);
588
589 if (test_opt(sb, ERRORS_CONT)) {
590 ext4_commit_super(sb, 0);
591 return;
592 }
593
594 ext4_unlock_group(sb, grp);
595 ext4_handle_error(sb);
596 /*
597 * We only get here in the ERRORS_RO case; relocking the group
598 * may be dangerous, but nothing bad will happen since the
599 * filesystem will have already been marked read/only and the
600 * journal has been aborted. We return 1 as a hint to callers
601 * who might what to use the return value from
602 * ext4_grp_locked_error() to distinguish beween the
603 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
604 * aggressively from the ext4 function in question, with a
605 * more appropriate error code.
606 */
607 ext4_lock_group(sb, grp);
608 return;
609 }
610
611 void ext4_update_dynamic_rev(struct super_block *sb)
612 {
613 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
614
615 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
616 return;
617
618 ext4_warning(sb,
619 "updating to rev %d because of new feature flag, "
620 "running e2fsck is recommended",
621 EXT4_DYNAMIC_REV);
622
623 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
624 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
625 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
626 /* leave es->s_feature_*compat flags alone */
627 /* es->s_uuid will be set by e2fsck if empty */
628
629 /*
630 * The rest of the superblock fields should be zero, and if not it
631 * means they are likely already in use, so leave them alone. We
632 * can leave it up to e2fsck to clean up any inconsistencies there.
633 */
634 }
635
636 /*
637 * Open the external journal device
638 */
639 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
640 {
641 struct block_device *bdev;
642 char b[BDEVNAME_SIZE];
643
644 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
645 if (IS_ERR(bdev))
646 goto fail;
647 return bdev;
648
649 fail:
650 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
651 __bdevname(dev, b), PTR_ERR(bdev));
652 return NULL;
653 }
654
655 /*
656 * Release the journal device
657 */
658 static int ext4_blkdev_put(struct block_device *bdev)
659 {
660 bd_release(bdev);
661 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
662 }
663
664 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
665 {
666 struct block_device *bdev;
667 int ret = -ENODEV;
668
669 bdev = sbi->journal_bdev;
670 if (bdev) {
671 ret = ext4_blkdev_put(bdev);
672 sbi->journal_bdev = NULL;
673 }
674 return ret;
675 }
676
677 static inline struct inode *orphan_list_entry(struct list_head *l)
678 {
679 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
680 }
681
682 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
683 {
684 struct list_head *l;
685
686 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
687 le32_to_cpu(sbi->s_es->s_last_orphan));
688
689 printk(KERN_ERR "sb_info orphan list:\n");
690 list_for_each(l, &sbi->s_orphan) {
691 struct inode *inode = orphan_list_entry(l);
692 printk(KERN_ERR " "
693 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
694 inode->i_sb->s_id, inode->i_ino, inode,
695 inode->i_mode, inode->i_nlink,
696 NEXT_ORPHAN(inode));
697 }
698 }
699
700 static void ext4_put_super(struct super_block *sb)
701 {
702 struct ext4_sb_info *sbi = EXT4_SB(sb);
703 struct ext4_super_block *es = sbi->s_es;
704 int i, err;
705
706 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
707
708 flush_workqueue(sbi->dio_unwritten_wq);
709 destroy_workqueue(sbi->dio_unwritten_wq);
710
711 lock_super(sb);
712 lock_kernel();
713 if (sb->s_dirt)
714 ext4_commit_super(sb, 1);
715
716 if (sbi->s_journal) {
717 err = jbd2_journal_destroy(sbi->s_journal);
718 sbi->s_journal = NULL;
719 if (err < 0)
720 ext4_abort(sb, "Couldn't clean up the journal");
721 }
722
723 del_timer(&sbi->s_err_report);
724 ext4_release_system_zone(sb);
725 ext4_mb_release(sb);
726 ext4_ext_release(sb);
727 ext4_xattr_put_super(sb);
728
729 if (!(sb->s_flags & MS_RDONLY)) {
730 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
731 es->s_state = cpu_to_le16(sbi->s_mount_state);
732 ext4_commit_super(sb, 1);
733 }
734 if (sbi->s_proc) {
735 remove_proc_entry(sb->s_id, ext4_proc_root);
736 }
737 kobject_del(&sbi->s_kobj);
738
739 for (i = 0; i < sbi->s_gdb_count; i++)
740 brelse(sbi->s_group_desc[i]);
741 kfree(sbi->s_group_desc);
742 if (is_vmalloc_addr(sbi->s_flex_groups))
743 vfree(sbi->s_flex_groups);
744 else
745 kfree(sbi->s_flex_groups);
746 percpu_counter_destroy(&sbi->s_freeblocks_counter);
747 percpu_counter_destroy(&sbi->s_freeinodes_counter);
748 percpu_counter_destroy(&sbi->s_dirs_counter);
749 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
750 brelse(sbi->s_sbh);
751 #ifdef CONFIG_QUOTA
752 for (i = 0; i < MAXQUOTAS; i++)
753 kfree(sbi->s_qf_names[i]);
754 #endif
755
756 /* Debugging code just in case the in-memory inode orphan list
757 * isn't empty. The on-disk one can be non-empty if we've
758 * detected an error and taken the fs readonly, but the
759 * in-memory list had better be clean by this point. */
760 if (!list_empty(&sbi->s_orphan))
761 dump_orphan_list(sb, sbi);
762 J_ASSERT(list_empty(&sbi->s_orphan));
763
764 invalidate_bdev(sb->s_bdev);
765 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
766 /*
767 * Invalidate the journal device's buffers. We don't want them
768 * floating about in memory - the physical journal device may
769 * hotswapped, and it breaks the `ro-after' testing code.
770 */
771 sync_blockdev(sbi->journal_bdev);
772 invalidate_bdev(sbi->journal_bdev);
773 ext4_blkdev_remove(sbi);
774 }
775 sb->s_fs_info = NULL;
776 /*
777 * Now that we are completely done shutting down the
778 * superblock, we need to actually destroy the kobject.
779 */
780 unlock_kernel();
781 unlock_super(sb);
782 kobject_put(&sbi->s_kobj);
783 wait_for_completion(&sbi->s_kobj_unregister);
784 kfree(sbi->s_blockgroup_lock);
785 kfree(sbi);
786 }
787
788 static struct kmem_cache *ext4_inode_cachep;
789
790 /*
791 * Called inside transaction, so use GFP_NOFS
792 */
793 static struct inode *ext4_alloc_inode(struct super_block *sb)
794 {
795 struct ext4_inode_info *ei;
796
797 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
798 if (!ei)
799 return NULL;
800
801 ei->vfs_inode.i_version = 1;
802 ei->vfs_inode.i_data.writeback_index = 0;
803 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
804 INIT_LIST_HEAD(&ei->i_prealloc_list);
805 spin_lock_init(&ei->i_prealloc_lock);
806 /*
807 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
808 * therefore it can be null here. Don't check it, just initialize
809 * jinode.
810 */
811 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
812 ei->i_reserved_data_blocks = 0;
813 ei->i_reserved_meta_blocks = 0;
814 ei->i_allocated_meta_blocks = 0;
815 ei->i_da_metadata_calc_len = 0;
816 ei->i_delalloc_reserved_flag = 0;
817 spin_lock_init(&(ei->i_block_reservation_lock));
818 #ifdef CONFIG_QUOTA
819 ei->i_reserved_quota = 0;
820 #endif
821 INIT_LIST_HEAD(&ei->i_completed_io_list);
822 spin_lock_init(&ei->i_completed_io_lock);
823 ei->cur_aio_dio = NULL;
824 ei->i_sync_tid = 0;
825 ei->i_datasync_tid = 0;
826
827 return &ei->vfs_inode;
828 }
829
830 static void ext4_destroy_inode(struct inode *inode)
831 {
832 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
833 ext4_msg(inode->i_sb, KERN_ERR,
834 "Inode %lu (%p): orphan list check failed!",
835 inode->i_ino, EXT4_I(inode));
836 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
837 EXT4_I(inode), sizeof(struct ext4_inode_info),
838 true);
839 dump_stack();
840 }
841 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
842 }
843
844 static void init_once(void *foo)
845 {
846 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
847
848 INIT_LIST_HEAD(&ei->i_orphan);
849 #ifdef CONFIG_EXT4_FS_XATTR
850 init_rwsem(&ei->xattr_sem);
851 #endif
852 init_rwsem(&ei->i_data_sem);
853 inode_init_once(&ei->vfs_inode);
854 }
855
856 static int init_inodecache(void)
857 {
858 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
859 sizeof(struct ext4_inode_info),
860 0, (SLAB_RECLAIM_ACCOUNT|
861 SLAB_MEM_SPREAD),
862 init_once);
863 if (ext4_inode_cachep == NULL)
864 return -ENOMEM;
865 return 0;
866 }
867
868 static void destroy_inodecache(void)
869 {
870 kmem_cache_destroy(ext4_inode_cachep);
871 }
872
873 void ext4_clear_inode(struct inode *inode)
874 {
875 invalidate_inode_buffers(inode);
876 end_writeback(inode);
877 dquot_drop(inode);
878 ext4_discard_preallocations(inode);
879 if (EXT4_JOURNAL(inode))
880 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
881 &EXT4_I(inode)->jinode);
882 }
883
884 static inline void ext4_show_quota_options(struct seq_file *seq,
885 struct super_block *sb)
886 {
887 #if defined(CONFIG_QUOTA)
888 struct ext4_sb_info *sbi = EXT4_SB(sb);
889
890 if (sbi->s_jquota_fmt) {
891 char *fmtname = "";
892
893 switch (sbi->s_jquota_fmt) {
894 case QFMT_VFS_OLD:
895 fmtname = "vfsold";
896 break;
897 case QFMT_VFS_V0:
898 fmtname = "vfsv0";
899 break;
900 case QFMT_VFS_V1:
901 fmtname = "vfsv1";
902 break;
903 }
904 seq_printf(seq, ",jqfmt=%s", fmtname);
905 }
906
907 if (sbi->s_qf_names[USRQUOTA])
908 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
909
910 if (sbi->s_qf_names[GRPQUOTA])
911 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
912
913 if (test_opt(sb, USRQUOTA))
914 seq_puts(seq, ",usrquota");
915
916 if (test_opt(sb, GRPQUOTA))
917 seq_puts(seq, ",grpquota");
918 #endif
919 }
920
921 /*
922 * Show an option if
923 * - it's set to a non-default value OR
924 * - if the per-sb default is different from the global default
925 */
926 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
927 {
928 int def_errors;
929 unsigned long def_mount_opts;
930 struct super_block *sb = vfs->mnt_sb;
931 struct ext4_sb_info *sbi = EXT4_SB(sb);
932 struct ext4_super_block *es = sbi->s_es;
933
934 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
935 def_errors = le16_to_cpu(es->s_errors);
936
937 if (sbi->s_sb_block != 1)
938 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
939 if (test_opt(sb, MINIX_DF))
940 seq_puts(seq, ",minixdf");
941 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
942 seq_puts(seq, ",grpid");
943 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
944 seq_puts(seq, ",nogrpid");
945 if (sbi->s_resuid != EXT4_DEF_RESUID ||
946 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
947 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
948 }
949 if (sbi->s_resgid != EXT4_DEF_RESGID ||
950 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
951 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
952 }
953 if (test_opt(sb, ERRORS_RO)) {
954 if (def_errors == EXT4_ERRORS_PANIC ||
955 def_errors == EXT4_ERRORS_CONTINUE) {
956 seq_puts(seq, ",errors=remount-ro");
957 }
958 }
959 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
960 seq_puts(seq, ",errors=continue");
961 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
962 seq_puts(seq, ",errors=panic");
963 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
964 seq_puts(seq, ",nouid32");
965 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
966 seq_puts(seq, ",debug");
967 if (test_opt(sb, OLDALLOC))
968 seq_puts(seq, ",oldalloc");
969 #ifdef CONFIG_EXT4_FS_XATTR
970 if (test_opt(sb, XATTR_USER) &&
971 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
972 seq_puts(seq, ",user_xattr");
973 if (!test_opt(sb, XATTR_USER) &&
974 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
975 seq_puts(seq, ",nouser_xattr");
976 }
977 #endif
978 #ifdef CONFIG_EXT4_FS_POSIX_ACL
979 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
980 seq_puts(seq, ",acl");
981 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
982 seq_puts(seq, ",noacl");
983 #endif
984 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
985 seq_printf(seq, ",commit=%u",
986 (unsigned) (sbi->s_commit_interval / HZ));
987 }
988 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
989 seq_printf(seq, ",min_batch_time=%u",
990 (unsigned) sbi->s_min_batch_time);
991 }
992 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
993 seq_printf(seq, ",max_batch_time=%u",
994 (unsigned) sbi->s_min_batch_time);
995 }
996
997 /*
998 * We're changing the default of barrier mount option, so
999 * let's always display its mount state so it's clear what its
1000 * status is.
1001 */
1002 seq_puts(seq, ",barrier=");
1003 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1004 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1005 seq_puts(seq, ",journal_async_commit");
1006 else if (test_opt(sb, JOURNAL_CHECKSUM))
1007 seq_puts(seq, ",journal_checksum");
1008 if (test_opt(sb, I_VERSION))
1009 seq_puts(seq, ",i_version");
1010 if (!test_opt(sb, DELALLOC) &&
1011 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1012 seq_puts(seq, ",nodelalloc");
1013
1014 if (sbi->s_stripe)
1015 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1016 /*
1017 * journal mode get enabled in different ways
1018 * So just print the value even if we didn't specify it
1019 */
1020 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1021 seq_puts(seq, ",data=journal");
1022 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1023 seq_puts(seq, ",data=ordered");
1024 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1025 seq_puts(seq, ",data=writeback");
1026
1027 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1028 seq_printf(seq, ",inode_readahead_blks=%u",
1029 sbi->s_inode_readahead_blks);
1030
1031 if (test_opt(sb, DATA_ERR_ABORT))
1032 seq_puts(seq, ",data_err=abort");
1033
1034 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1035 seq_puts(seq, ",noauto_da_alloc");
1036
1037 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1038 seq_puts(seq, ",discard");
1039
1040 if (test_opt(sb, NOLOAD))
1041 seq_puts(seq, ",norecovery");
1042
1043 if (test_opt(sb, DIOREAD_NOLOCK))
1044 seq_puts(seq, ",dioread_nolock");
1045
1046 if (test_opt(sb, BLOCK_VALIDITY) &&
1047 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1048 seq_puts(seq, ",block_validity");
1049
1050 ext4_show_quota_options(seq, sb);
1051
1052 return 0;
1053 }
1054
1055 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1056 u64 ino, u32 generation)
1057 {
1058 struct inode *inode;
1059
1060 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1061 return ERR_PTR(-ESTALE);
1062 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1063 return ERR_PTR(-ESTALE);
1064
1065 /* iget isn't really right if the inode is currently unallocated!!
1066 *
1067 * ext4_read_inode will return a bad_inode if the inode had been
1068 * deleted, so we should be safe.
1069 *
1070 * Currently we don't know the generation for parent directory, so
1071 * a generation of 0 means "accept any"
1072 */
1073 inode = ext4_iget(sb, ino);
1074 if (IS_ERR(inode))
1075 return ERR_CAST(inode);
1076 if (generation && inode->i_generation != generation) {
1077 iput(inode);
1078 return ERR_PTR(-ESTALE);
1079 }
1080
1081 return inode;
1082 }
1083
1084 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1085 int fh_len, int fh_type)
1086 {
1087 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1088 ext4_nfs_get_inode);
1089 }
1090
1091 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1092 int fh_len, int fh_type)
1093 {
1094 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1095 ext4_nfs_get_inode);
1096 }
1097
1098 /*
1099 * Try to release metadata pages (indirect blocks, directories) which are
1100 * mapped via the block device. Since these pages could have journal heads
1101 * which would prevent try_to_free_buffers() from freeing them, we must use
1102 * jbd2 layer's try_to_free_buffers() function to release them.
1103 */
1104 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1105 gfp_t wait)
1106 {
1107 journal_t *journal = EXT4_SB(sb)->s_journal;
1108
1109 WARN_ON(PageChecked(page));
1110 if (!page_has_buffers(page))
1111 return 0;
1112 if (journal)
1113 return jbd2_journal_try_to_free_buffers(journal, page,
1114 wait & ~__GFP_WAIT);
1115 return try_to_free_buffers(page);
1116 }
1117
1118 #ifdef CONFIG_QUOTA
1119 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1120 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1121
1122 static int ext4_write_dquot(struct dquot *dquot);
1123 static int ext4_acquire_dquot(struct dquot *dquot);
1124 static int ext4_release_dquot(struct dquot *dquot);
1125 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1126 static int ext4_write_info(struct super_block *sb, int type);
1127 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1128 char *path);
1129 static int ext4_quota_off(struct super_block *sb, int type);
1130 static int ext4_quota_on_mount(struct super_block *sb, int type);
1131 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1132 size_t len, loff_t off);
1133 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1134 const char *data, size_t len, loff_t off);
1135
1136 static const struct dquot_operations ext4_quota_operations = {
1137 #ifdef CONFIG_QUOTA
1138 .get_reserved_space = ext4_get_reserved_space,
1139 #endif
1140 .write_dquot = ext4_write_dquot,
1141 .acquire_dquot = ext4_acquire_dquot,
1142 .release_dquot = ext4_release_dquot,
1143 .mark_dirty = ext4_mark_dquot_dirty,
1144 .write_info = ext4_write_info,
1145 .alloc_dquot = dquot_alloc,
1146 .destroy_dquot = dquot_destroy,
1147 };
1148
1149 static const struct quotactl_ops ext4_qctl_operations = {
1150 .quota_on = ext4_quota_on,
1151 .quota_off = ext4_quota_off,
1152 .quota_sync = dquot_quota_sync,
1153 .get_info = dquot_get_dqinfo,
1154 .set_info = dquot_set_dqinfo,
1155 .get_dqblk = dquot_get_dqblk,
1156 .set_dqblk = dquot_set_dqblk
1157 };
1158 #endif
1159
1160 static const struct super_operations ext4_sops = {
1161 .alloc_inode = ext4_alloc_inode,
1162 .destroy_inode = ext4_destroy_inode,
1163 .write_inode = ext4_write_inode,
1164 .dirty_inode = ext4_dirty_inode,
1165 .evict_inode = ext4_evict_inode,
1166 .put_super = ext4_put_super,
1167 .sync_fs = ext4_sync_fs,
1168 .freeze_fs = ext4_freeze,
1169 .unfreeze_fs = ext4_unfreeze,
1170 .statfs = ext4_statfs,
1171 .remount_fs = ext4_remount,
1172 .show_options = ext4_show_options,
1173 #ifdef CONFIG_QUOTA
1174 .quota_read = ext4_quota_read,
1175 .quota_write = ext4_quota_write,
1176 #endif
1177 .bdev_try_to_free_page = bdev_try_to_free_page,
1178 };
1179
1180 static const struct super_operations ext4_nojournal_sops = {
1181 .alloc_inode = ext4_alloc_inode,
1182 .destroy_inode = ext4_destroy_inode,
1183 .write_inode = ext4_write_inode,
1184 .dirty_inode = ext4_dirty_inode,
1185 .evict_inode = ext4_evict_inode,
1186 .write_super = ext4_write_super,
1187 .put_super = ext4_put_super,
1188 .statfs = ext4_statfs,
1189 .remount_fs = ext4_remount,
1190 .show_options = ext4_show_options,
1191 #ifdef CONFIG_QUOTA
1192 .quota_read = ext4_quota_read,
1193 .quota_write = ext4_quota_write,
1194 #endif
1195 .bdev_try_to_free_page = bdev_try_to_free_page,
1196 };
1197
1198 static const struct export_operations ext4_export_ops = {
1199 .fh_to_dentry = ext4_fh_to_dentry,
1200 .fh_to_parent = ext4_fh_to_parent,
1201 .get_parent = ext4_get_parent,
1202 };
1203
1204 enum {
1205 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1206 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1207 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1208 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1209 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1210 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1211 Opt_journal_update, Opt_journal_dev,
1212 Opt_journal_checksum, Opt_journal_async_commit,
1213 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1214 Opt_data_err_abort, Opt_data_err_ignore,
1215 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1216 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1217 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1218 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1219 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
1220 Opt_block_validity, Opt_noblock_validity,
1221 Opt_inode_readahead_blks, Opt_journal_ioprio,
1222 Opt_dioread_nolock, Opt_dioread_lock,
1223 Opt_discard, Opt_nodiscard,
1224 };
1225
1226 static const match_table_t tokens = {
1227 {Opt_bsd_df, "bsddf"},
1228 {Opt_minix_df, "minixdf"},
1229 {Opt_grpid, "grpid"},
1230 {Opt_grpid, "bsdgroups"},
1231 {Opt_nogrpid, "nogrpid"},
1232 {Opt_nogrpid, "sysvgroups"},
1233 {Opt_resgid, "resgid=%u"},
1234 {Opt_resuid, "resuid=%u"},
1235 {Opt_sb, "sb=%u"},
1236 {Opt_err_cont, "errors=continue"},
1237 {Opt_err_panic, "errors=panic"},
1238 {Opt_err_ro, "errors=remount-ro"},
1239 {Opt_nouid32, "nouid32"},
1240 {Opt_debug, "debug"},
1241 {Opt_oldalloc, "oldalloc"},
1242 {Opt_orlov, "orlov"},
1243 {Opt_user_xattr, "user_xattr"},
1244 {Opt_nouser_xattr, "nouser_xattr"},
1245 {Opt_acl, "acl"},
1246 {Opt_noacl, "noacl"},
1247 {Opt_noload, "noload"},
1248 {Opt_noload, "norecovery"},
1249 {Opt_nobh, "nobh"},
1250 {Opt_bh, "bh"},
1251 {Opt_commit, "commit=%u"},
1252 {Opt_min_batch_time, "min_batch_time=%u"},
1253 {Opt_max_batch_time, "max_batch_time=%u"},
1254 {Opt_journal_update, "journal=update"},
1255 {Opt_journal_dev, "journal_dev=%u"},
1256 {Opt_journal_checksum, "journal_checksum"},
1257 {Opt_journal_async_commit, "journal_async_commit"},
1258 {Opt_abort, "abort"},
1259 {Opt_data_journal, "data=journal"},
1260 {Opt_data_ordered, "data=ordered"},
1261 {Opt_data_writeback, "data=writeback"},
1262 {Opt_data_err_abort, "data_err=abort"},
1263 {Opt_data_err_ignore, "data_err=ignore"},
1264 {Opt_offusrjquota, "usrjquota="},
1265 {Opt_usrjquota, "usrjquota=%s"},
1266 {Opt_offgrpjquota, "grpjquota="},
1267 {Opt_grpjquota, "grpjquota=%s"},
1268 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1269 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1270 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1271 {Opt_grpquota, "grpquota"},
1272 {Opt_noquota, "noquota"},
1273 {Opt_quota, "quota"},
1274 {Opt_usrquota, "usrquota"},
1275 {Opt_barrier, "barrier=%u"},
1276 {Opt_barrier, "barrier"},
1277 {Opt_nobarrier, "nobarrier"},
1278 {Opt_i_version, "i_version"},
1279 {Opt_stripe, "stripe=%u"},
1280 {Opt_resize, "resize"},
1281 {Opt_delalloc, "delalloc"},
1282 {Opt_nodelalloc, "nodelalloc"},
1283 {Opt_block_validity, "block_validity"},
1284 {Opt_noblock_validity, "noblock_validity"},
1285 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1286 {Opt_journal_ioprio, "journal_ioprio=%u"},
1287 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1288 {Opt_auto_da_alloc, "auto_da_alloc"},
1289 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1290 {Opt_dioread_nolock, "dioread_nolock"},
1291 {Opt_dioread_lock, "dioread_lock"},
1292 {Opt_discard, "discard"},
1293 {Opt_nodiscard, "nodiscard"},
1294 {Opt_err, NULL},
1295 };
1296
1297 static ext4_fsblk_t get_sb_block(void **data)
1298 {
1299 ext4_fsblk_t sb_block;
1300 char *options = (char *) *data;
1301
1302 if (!options || strncmp(options, "sb=", 3) != 0)
1303 return 1; /* Default location */
1304
1305 options += 3;
1306 /* TODO: use simple_strtoll with >32bit ext4 */
1307 sb_block = simple_strtoul(options, &options, 0);
1308 if (*options && *options != ',') {
1309 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1310 (char *) *data);
1311 return 1;
1312 }
1313 if (*options == ',')
1314 options++;
1315 *data = (void *) options;
1316
1317 return sb_block;
1318 }
1319
1320 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1321 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1322 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1323
1324 #ifdef CONFIG_QUOTA
1325 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1326 {
1327 struct ext4_sb_info *sbi = EXT4_SB(sb);
1328 char *qname;
1329
1330 if (sb_any_quota_loaded(sb) &&
1331 !sbi->s_qf_names[qtype]) {
1332 ext4_msg(sb, KERN_ERR,
1333 "Cannot change journaled "
1334 "quota options when quota turned on");
1335 return 0;
1336 }
1337 qname = match_strdup(args);
1338 if (!qname) {
1339 ext4_msg(sb, KERN_ERR,
1340 "Not enough memory for storing quotafile name");
1341 return 0;
1342 }
1343 if (sbi->s_qf_names[qtype] &&
1344 strcmp(sbi->s_qf_names[qtype], qname)) {
1345 ext4_msg(sb, KERN_ERR,
1346 "%s quota file already specified", QTYPE2NAME(qtype));
1347 kfree(qname);
1348 return 0;
1349 }
1350 sbi->s_qf_names[qtype] = qname;
1351 if (strchr(sbi->s_qf_names[qtype], '/')) {
1352 ext4_msg(sb, KERN_ERR,
1353 "quotafile must be on filesystem root");
1354 kfree(sbi->s_qf_names[qtype]);
1355 sbi->s_qf_names[qtype] = NULL;
1356 return 0;
1357 }
1358 set_opt(sbi->s_mount_opt, QUOTA);
1359 return 1;
1360 }
1361
1362 static int clear_qf_name(struct super_block *sb, int qtype)
1363 {
1364
1365 struct ext4_sb_info *sbi = EXT4_SB(sb);
1366
1367 if (sb_any_quota_loaded(sb) &&
1368 sbi->s_qf_names[qtype]) {
1369 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1370 " when quota turned on");
1371 return 0;
1372 }
1373 /*
1374 * The space will be released later when all options are confirmed
1375 * to be correct
1376 */
1377 sbi->s_qf_names[qtype] = NULL;
1378 return 1;
1379 }
1380 #endif
1381
1382 static int parse_options(char *options, struct super_block *sb,
1383 unsigned long *journal_devnum,
1384 unsigned int *journal_ioprio,
1385 ext4_fsblk_t *n_blocks_count, int is_remount)
1386 {
1387 struct ext4_sb_info *sbi = EXT4_SB(sb);
1388 char *p;
1389 substring_t args[MAX_OPT_ARGS];
1390 int data_opt = 0;
1391 int option;
1392 #ifdef CONFIG_QUOTA
1393 int qfmt;
1394 #endif
1395
1396 if (!options)
1397 return 1;
1398
1399 while ((p = strsep(&options, ",")) != NULL) {
1400 int token;
1401 if (!*p)
1402 continue;
1403
1404 /*
1405 * Initialize args struct so we know whether arg was
1406 * found; some options take optional arguments.
1407 */
1408 args[0].to = args[0].from = 0;
1409 token = match_token(p, tokens, args);
1410 switch (token) {
1411 case Opt_bsd_df:
1412 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1413 clear_opt(sbi->s_mount_opt, MINIX_DF);
1414 break;
1415 case Opt_minix_df:
1416 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1417 set_opt(sbi->s_mount_opt, MINIX_DF);
1418
1419 break;
1420 case Opt_grpid:
1421 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1422 set_opt(sbi->s_mount_opt, GRPID);
1423
1424 break;
1425 case Opt_nogrpid:
1426 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1427 clear_opt(sbi->s_mount_opt, GRPID);
1428
1429 break;
1430 case Opt_resuid:
1431 if (match_int(&args[0], &option))
1432 return 0;
1433 sbi->s_resuid = option;
1434 break;
1435 case Opt_resgid:
1436 if (match_int(&args[0], &option))
1437 return 0;
1438 sbi->s_resgid = option;
1439 break;
1440 case Opt_sb:
1441 /* handled by get_sb_block() instead of here */
1442 /* *sb_block = match_int(&args[0]); */
1443 break;
1444 case Opt_err_panic:
1445 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1446 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1447 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
1448 break;
1449 case Opt_err_ro:
1450 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1451 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1452 set_opt(sbi->s_mount_opt, ERRORS_RO);
1453 break;
1454 case Opt_err_cont:
1455 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1456 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1457 set_opt(sbi->s_mount_opt, ERRORS_CONT);
1458 break;
1459 case Opt_nouid32:
1460 set_opt(sbi->s_mount_opt, NO_UID32);
1461 break;
1462 case Opt_debug:
1463 set_opt(sbi->s_mount_opt, DEBUG);
1464 break;
1465 case Opt_oldalloc:
1466 set_opt(sbi->s_mount_opt, OLDALLOC);
1467 break;
1468 case Opt_orlov:
1469 clear_opt(sbi->s_mount_opt, OLDALLOC);
1470 break;
1471 #ifdef CONFIG_EXT4_FS_XATTR
1472 case Opt_user_xattr:
1473 set_opt(sbi->s_mount_opt, XATTR_USER);
1474 break;
1475 case Opt_nouser_xattr:
1476 clear_opt(sbi->s_mount_opt, XATTR_USER);
1477 break;
1478 #else
1479 case Opt_user_xattr:
1480 case Opt_nouser_xattr:
1481 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1482 break;
1483 #endif
1484 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1485 case Opt_acl:
1486 set_opt(sbi->s_mount_opt, POSIX_ACL);
1487 break;
1488 case Opt_noacl:
1489 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1490 break;
1491 #else
1492 case Opt_acl:
1493 case Opt_noacl:
1494 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1495 break;
1496 #endif
1497 case Opt_journal_update:
1498 /* Eventually we will want to be able to create
1499 a journal file here. For now, only allow the
1500 user to specify an existing inode to be the
1501 journal file. */
1502 if (is_remount) {
1503 ext4_msg(sb, KERN_ERR,
1504 "Cannot specify journal on remount");
1505 return 0;
1506 }
1507 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
1508 break;
1509 case Opt_journal_dev:
1510 if (is_remount) {
1511 ext4_msg(sb, KERN_ERR,
1512 "Cannot specify journal on remount");
1513 return 0;
1514 }
1515 if (match_int(&args[0], &option))
1516 return 0;
1517 *journal_devnum = option;
1518 break;
1519 case Opt_journal_checksum:
1520 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1521 break;
1522 case Opt_journal_async_commit:
1523 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1524 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1525 break;
1526 case Opt_noload:
1527 set_opt(sbi->s_mount_opt, NOLOAD);
1528 break;
1529 case Opt_commit:
1530 if (match_int(&args[0], &option))
1531 return 0;
1532 if (option < 0)
1533 return 0;
1534 if (option == 0)
1535 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1536 sbi->s_commit_interval = HZ * option;
1537 break;
1538 case Opt_max_batch_time:
1539 if (match_int(&args[0], &option))
1540 return 0;
1541 if (option < 0)
1542 return 0;
1543 if (option == 0)
1544 option = EXT4_DEF_MAX_BATCH_TIME;
1545 sbi->s_max_batch_time = option;
1546 break;
1547 case Opt_min_batch_time:
1548 if (match_int(&args[0], &option))
1549 return 0;
1550 if (option < 0)
1551 return 0;
1552 sbi->s_min_batch_time = option;
1553 break;
1554 case Opt_data_journal:
1555 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1556 goto datacheck;
1557 case Opt_data_ordered:
1558 data_opt = EXT4_MOUNT_ORDERED_DATA;
1559 goto datacheck;
1560 case Opt_data_writeback:
1561 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1562 datacheck:
1563 if (is_remount) {
1564 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1565 ext4_msg(sb, KERN_ERR,
1566 "Cannot change data mode on remount");
1567 return 0;
1568 }
1569 } else {
1570 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
1571 sbi->s_mount_opt |= data_opt;
1572 }
1573 break;
1574 case Opt_data_err_abort:
1575 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1576 break;
1577 case Opt_data_err_ignore:
1578 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1579 break;
1580 #ifdef CONFIG_QUOTA
1581 case Opt_usrjquota:
1582 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1583 return 0;
1584 break;
1585 case Opt_grpjquota:
1586 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1587 return 0;
1588 break;
1589 case Opt_offusrjquota:
1590 if (!clear_qf_name(sb, USRQUOTA))
1591 return 0;
1592 break;
1593 case Opt_offgrpjquota:
1594 if (!clear_qf_name(sb, GRPQUOTA))
1595 return 0;
1596 break;
1597
1598 case Opt_jqfmt_vfsold:
1599 qfmt = QFMT_VFS_OLD;
1600 goto set_qf_format;
1601 case Opt_jqfmt_vfsv0:
1602 qfmt = QFMT_VFS_V0;
1603 goto set_qf_format;
1604 case Opt_jqfmt_vfsv1:
1605 qfmt = QFMT_VFS_V1;
1606 set_qf_format:
1607 if (sb_any_quota_loaded(sb) &&
1608 sbi->s_jquota_fmt != qfmt) {
1609 ext4_msg(sb, KERN_ERR, "Cannot change "
1610 "journaled quota options when "
1611 "quota turned on");
1612 return 0;
1613 }
1614 sbi->s_jquota_fmt = qfmt;
1615 break;
1616 case Opt_quota:
1617 case Opt_usrquota:
1618 set_opt(sbi->s_mount_opt, QUOTA);
1619 set_opt(sbi->s_mount_opt, USRQUOTA);
1620 break;
1621 case Opt_grpquota:
1622 set_opt(sbi->s_mount_opt, QUOTA);
1623 set_opt(sbi->s_mount_opt, GRPQUOTA);
1624 break;
1625 case Opt_noquota:
1626 if (sb_any_quota_loaded(sb)) {
1627 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1628 "options when quota turned on");
1629 return 0;
1630 }
1631 clear_opt(sbi->s_mount_opt, QUOTA);
1632 clear_opt(sbi->s_mount_opt, USRQUOTA);
1633 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1634 break;
1635 #else
1636 case Opt_quota:
1637 case Opt_usrquota:
1638 case Opt_grpquota:
1639 ext4_msg(sb, KERN_ERR,
1640 "quota options not supported");
1641 break;
1642 case Opt_usrjquota:
1643 case Opt_grpjquota:
1644 case Opt_offusrjquota:
1645 case Opt_offgrpjquota:
1646 case Opt_jqfmt_vfsold:
1647 case Opt_jqfmt_vfsv0:
1648 case Opt_jqfmt_vfsv1:
1649 ext4_msg(sb, KERN_ERR,
1650 "journaled quota options not supported");
1651 break;
1652 case Opt_noquota:
1653 break;
1654 #endif
1655 case Opt_abort:
1656 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1657 break;
1658 case Opt_nobarrier:
1659 clear_opt(sbi->s_mount_opt, BARRIER);
1660 break;
1661 case Opt_barrier:
1662 if (args[0].from) {
1663 if (match_int(&args[0], &option))
1664 return 0;
1665 } else
1666 option = 1; /* No argument, default to 1 */
1667 if (option)
1668 set_opt(sbi->s_mount_opt, BARRIER);
1669 else
1670 clear_opt(sbi->s_mount_opt, BARRIER);
1671 break;
1672 case Opt_ignore:
1673 break;
1674 case Opt_resize:
1675 if (!is_remount) {
1676 ext4_msg(sb, KERN_ERR,
1677 "resize option only available "
1678 "for remount");
1679 return 0;
1680 }
1681 if (match_int(&args[0], &option) != 0)
1682 return 0;
1683 *n_blocks_count = option;
1684 break;
1685 case Opt_nobh:
1686 ext4_msg(sb, KERN_WARNING,
1687 "Ignoring deprecated nobh option");
1688 break;
1689 case Opt_bh:
1690 ext4_msg(sb, KERN_WARNING,
1691 "Ignoring deprecated bh option");
1692 break;
1693 case Opt_i_version:
1694 set_opt(sbi->s_mount_opt, I_VERSION);
1695 sb->s_flags |= MS_I_VERSION;
1696 break;
1697 case Opt_nodelalloc:
1698 clear_opt(sbi->s_mount_opt, DELALLOC);
1699 break;
1700 case Opt_stripe:
1701 if (match_int(&args[0], &option))
1702 return 0;
1703 if (option < 0)
1704 return 0;
1705 sbi->s_stripe = option;
1706 break;
1707 case Opt_delalloc:
1708 set_opt(sbi->s_mount_opt, DELALLOC);
1709 break;
1710 case Opt_block_validity:
1711 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1712 break;
1713 case Opt_noblock_validity:
1714 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1715 break;
1716 case Opt_inode_readahead_blks:
1717 if (match_int(&args[0], &option))
1718 return 0;
1719 if (option < 0 || option > (1 << 30))
1720 return 0;
1721 if (!is_power_of_2(option)) {
1722 ext4_msg(sb, KERN_ERR,
1723 "EXT4-fs: inode_readahead_blks"
1724 " must be a power of 2");
1725 return 0;
1726 }
1727 sbi->s_inode_readahead_blks = option;
1728 break;
1729 case Opt_journal_ioprio:
1730 if (match_int(&args[0], &option))
1731 return 0;
1732 if (option < 0 || option > 7)
1733 break;
1734 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1735 option);
1736 break;
1737 case Opt_noauto_da_alloc:
1738 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1739 break;
1740 case Opt_auto_da_alloc:
1741 if (args[0].from) {
1742 if (match_int(&args[0], &option))
1743 return 0;
1744 } else
1745 option = 1; /* No argument, default to 1 */
1746 if (option)
1747 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1748 else
1749 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1750 break;
1751 case Opt_discard:
1752 set_opt(sbi->s_mount_opt, DISCARD);
1753 break;
1754 case Opt_nodiscard:
1755 clear_opt(sbi->s_mount_opt, DISCARD);
1756 break;
1757 case Opt_dioread_nolock:
1758 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1759 break;
1760 case Opt_dioread_lock:
1761 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1762 break;
1763 default:
1764 ext4_msg(sb, KERN_ERR,
1765 "Unrecognized mount option \"%s\" "
1766 "or missing value", p);
1767 return 0;
1768 }
1769 }
1770 #ifdef CONFIG_QUOTA
1771 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1772 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1773 clear_opt(sbi->s_mount_opt, USRQUOTA);
1774
1775 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1776 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1777
1778 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1779 ext4_msg(sb, KERN_ERR, "old and new quota "
1780 "format mixing");
1781 return 0;
1782 }
1783
1784 if (!sbi->s_jquota_fmt) {
1785 ext4_msg(sb, KERN_ERR, "journaled quota format "
1786 "not specified");
1787 return 0;
1788 }
1789 } else {
1790 if (sbi->s_jquota_fmt) {
1791 ext4_msg(sb, KERN_ERR, "journaled quota format "
1792 "specified with no journaling "
1793 "enabled");
1794 return 0;
1795 }
1796 }
1797 #endif
1798 return 1;
1799 }
1800
1801 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1802 int read_only)
1803 {
1804 struct ext4_sb_info *sbi = EXT4_SB(sb);
1805 int res = 0;
1806
1807 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1808 ext4_msg(sb, KERN_ERR, "revision level too high, "
1809 "forcing read-only mode");
1810 res = MS_RDONLY;
1811 }
1812 if (read_only)
1813 return res;
1814 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1815 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1816 "running e2fsck is recommended");
1817 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1818 ext4_msg(sb, KERN_WARNING,
1819 "warning: mounting fs with errors, "
1820 "running e2fsck is recommended");
1821 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1822 le16_to_cpu(es->s_mnt_count) >=
1823 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1824 ext4_msg(sb, KERN_WARNING,
1825 "warning: maximal mount count reached, "
1826 "running e2fsck is recommended");
1827 else if (le32_to_cpu(es->s_checkinterval) &&
1828 (le32_to_cpu(es->s_lastcheck) +
1829 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1830 ext4_msg(sb, KERN_WARNING,
1831 "warning: checktime reached, "
1832 "running e2fsck is recommended");
1833 if (!sbi->s_journal)
1834 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1835 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1836 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1837 le16_add_cpu(&es->s_mnt_count, 1);
1838 es->s_mtime = cpu_to_le32(get_seconds());
1839 ext4_update_dynamic_rev(sb);
1840 if (sbi->s_journal)
1841 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1842
1843 ext4_commit_super(sb, 1);
1844 if (test_opt(sb, DEBUG))
1845 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1846 "bpg=%lu, ipg=%lu, mo=%04x]\n",
1847 sb->s_blocksize,
1848 sbi->s_groups_count,
1849 EXT4_BLOCKS_PER_GROUP(sb),
1850 EXT4_INODES_PER_GROUP(sb),
1851 sbi->s_mount_opt);
1852
1853 return res;
1854 }
1855
1856 static int ext4_fill_flex_info(struct super_block *sb)
1857 {
1858 struct ext4_sb_info *sbi = EXT4_SB(sb);
1859 struct ext4_group_desc *gdp = NULL;
1860 ext4_group_t flex_group_count;
1861 ext4_group_t flex_group;
1862 int groups_per_flex = 0;
1863 size_t size;
1864 int i;
1865
1866 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1867 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1868
1869 if (groups_per_flex < 2) {
1870 sbi->s_log_groups_per_flex = 0;
1871 return 1;
1872 }
1873
1874 /* We allocate both existing and potentially added groups */
1875 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1876 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1877 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1878 size = flex_group_count * sizeof(struct flex_groups);
1879 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1880 if (sbi->s_flex_groups == NULL) {
1881 sbi->s_flex_groups = vmalloc(size);
1882 if (sbi->s_flex_groups)
1883 memset(sbi->s_flex_groups, 0, size);
1884 }
1885 if (sbi->s_flex_groups == NULL) {
1886 ext4_msg(sb, KERN_ERR, "not enough memory for "
1887 "%u flex groups", flex_group_count);
1888 goto failed;
1889 }
1890
1891 for (i = 0; i < sbi->s_groups_count; i++) {
1892 gdp = ext4_get_group_desc(sb, i, NULL);
1893
1894 flex_group = ext4_flex_group(sbi, i);
1895 atomic_add(ext4_free_inodes_count(sb, gdp),
1896 &sbi->s_flex_groups[flex_group].free_inodes);
1897 atomic_add(ext4_free_blks_count(sb, gdp),
1898 &sbi->s_flex_groups[flex_group].free_blocks);
1899 atomic_add(ext4_used_dirs_count(sb, gdp),
1900 &sbi->s_flex_groups[flex_group].used_dirs);
1901 }
1902
1903 return 1;
1904 failed:
1905 return 0;
1906 }
1907
1908 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1909 struct ext4_group_desc *gdp)
1910 {
1911 __u16 crc = 0;
1912
1913 if (sbi->s_es->s_feature_ro_compat &
1914 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1915 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1916 __le32 le_group = cpu_to_le32(block_group);
1917
1918 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1919 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1920 crc = crc16(crc, (__u8 *)gdp, offset);
1921 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1922 /* for checksum of struct ext4_group_desc do the rest...*/
1923 if ((sbi->s_es->s_feature_incompat &
1924 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1925 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1926 crc = crc16(crc, (__u8 *)gdp + offset,
1927 le16_to_cpu(sbi->s_es->s_desc_size) -
1928 offset);
1929 }
1930
1931 return cpu_to_le16(crc);
1932 }
1933
1934 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1935 struct ext4_group_desc *gdp)
1936 {
1937 if ((sbi->s_es->s_feature_ro_compat &
1938 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1939 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1940 return 0;
1941
1942 return 1;
1943 }
1944
1945 /* Called at mount-time, super-block is locked */
1946 static int ext4_check_descriptors(struct super_block *sb)
1947 {
1948 struct ext4_sb_info *sbi = EXT4_SB(sb);
1949 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1950 ext4_fsblk_t last_block;
1951 ext4_fsblk_t block_bitmap;
1952 ext4_fsblk_t inode_bitmap;
1953 ext4_fsblk_t inode_table;
1954 int flexbg_flag = 0;
1955 ext4_group_t i;
1956
1957 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1958 flexbg_flag = 1;
1959
1960 ext4_debug("Checking group descriptors");
1961
1962 for (i = 0; i < sbi->s_groups_count; i++) {
1963 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1964
1965 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1966 last_block = ext4_blocks_count(sbi->s_es) - 1;
1967 else
1968 last_block = first_block +
1969 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1970
1971 block_bitmap = ext4_block_bitmap(sb, gdp);
1972 if (block_bitmap < first_block || block_bitmap > last_block) {
1973 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1974 "Block bitmap for group %u not in group "
1975 "(block %llu)!", i, block_bitmap);
1976 return 0;
1977 }
1978 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1979 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1980 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1981 "Inode bitmap for group %u not in group "
1982 "(block %llu)!", i, inode_bitmap);
1983 return 0;
1984 }
1985 inode_table = ext4_inode_table(sb, gdp);
1986 if (inode_table < first_block ||
1987 inode_table + sbi->s_itb_per_group - 1 > last_block) {
1988 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1989 "Inode table for group %u not in group "
1990 "(block %llu)!", i, inode_table);
1991 return 0;
1992 }
1993 ext4_lock_group(sb, i);
1994 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1995 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1996 "Checksum for group %u failed (%u!=%u)",
1997 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1998 gdp)), le16_to_cpu(gdp->bg_checksum));
1999 if (!(sb->s_flags & MS_RDONLY)) {
2000 ext4_unlock_group(sb, i);
2001 return 0;
2002 }
2003 }
2004 ext4_unlock_group(sb, i);
2005 if (!flexbg_flag)
2006 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2007 }
2008
2009 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2010 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2011 return 1;
2012 }
2013
2014 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2015 * the superblock) which were deleted from all directories, but held open by
2016 * a process at the time of a crash. We walk the list and try to delete these
2017 * inodes at recovery time (only with a read-write filesystem).
2018 *
2019 * In order to keep the orphan inode chain consistent during traversal (in
2020 * case of crash during recovery), we link each inode into the superblock
2021 * orphan list_head and handle it the same way as an inode deletion during
2022 * normal operation (which journals the operations for us).
2023 *
2024 * We only do an iget() and an iput() on each inode, which is very safe if we
2025 * accidentally point at an in-use or already deleted inode. The worst that
2026 * can happen in this case is that we get a "bit already cleared" message from
2027 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2028 * e2fsck was run on this filesystem, and it must have already done the orphan
2029 * inode cleanup for us, so we can safely abort without any further action.
2030 */
2031 static void ext4_orphan_cleanup(struct super_block *sb,
2032 struct ext4_super_block *es)
2033 {
2034 unsigned int s_flags = sb->s_flags;
2035 int nr_orphans = 0, nr_truncates = 0;
2036 #ifdef CONFIG_QUOTA
2037 int i;
2038 #endif
2039 if (!es->s_last_orphan) {
2040 jbd_debug(4, "no orphan inodes to clean up\n");
2041 return;
2042 }
2043
2044 if (bdev_read_only(sb->s_bdev)) {
2045 ext4_msg(sb, KERN_ERR, "write access "
2046 "unavailable, skipping orphan cleanup");
2047 return;
2048 }
2049
2050 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2051 if (es->s_last_orphan)
2052 jbd_debug(1, "Errors on filesystem, "
2053 "clearing orphan list.\n");
2054 es->s_last_orphan = 0;
2055 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2056 return;
2057 }
2058
2059 if (s_flags & MS_RDONLY) {
2060 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2061 sb->s_flags &= ~MS_RDONLY;
2062 }
2063 #ifdef CONFIG_QUOTA
2064 /* Needed for iput() to work correctly and not trash data */
2065 sb->s_flags |= MS_ACTIVE;
2066 /* Turn on quotas so that they are updated correctly */
2067 for (i = 0; i < MAXQUOTAS; i++) {
2068 if (EXT4_SB(sb)->s_qf_names[i]) {
2069 int ret = ext4_quota_on_mount(sb, i);
2070 if (ret < 0)
2071 ext4_msg(sb, KERN_ERR,
2072 "Cannot turn on journaled "
2073 "quota: error %d", ret);
2074 }
2075 }
2076 #endif
2077
2078 while (es->s_last_orphan) {
2079 struct inode *inode;
2080
2081 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2082 if (IS_ERR(inode)) {
2083 es->s_last_orphan = 0;
2084 break;
2085 }
2086
2087 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2088 dquot_initialize(inode);
2089 if (inode->i_nlink) {
2090 ext4_msg(sb, KERN_DEBUG,
2091 "%s: truncating inode %lu to %lld bytes",
2092 __func__, inode->i_ino, inode->i_size);
2093 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2094 inode->i_ino, inode->i_size);
2095 ext4_truncate(inode);
2096 nr_truncates++;
2097 } else {
2098 ext4_msg(sb, KERN_DEBUG,
2099 "%s: deleting unreferenced inode %lu",
2100 __func__, inode->i_ino);
2101 jbd_debug(2, "deleting unreferenced inode %lu\n",
2102 inode->i_ino);
2103 nr_orphans++;
2104 }
2105 iput(inode); /* The delete magic happens here! */
2106 }
2107
2108 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2109
2110 if (nr_orphans)
2111 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2112 PLURAL(nr_orphans));
2113 if (nr_truncates)
2114 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2115 PLURAL(nr_truncates));
2116 #ifdef CONFIG_QUOTA
2117 /* Turn quotas off */
2118 for (i = 0; i < MAXQUOTAS; i++) {
2119 if (sb_dqopt(sb)->files[i])
2120 dquot_quota_off(sb, i);
2121 }
2122 #endif
2123 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2124 }
2125
2126 /*
2127 * Maximal extent format file size.
2128 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2129 * extent format containers, within a sector_t, and within i_blocks
2130 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2131 * so that won't be a limiting factor.
2132 *
2133 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2134 */
2135 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2136 {
2137 loff_t res;
2138 loff_t upper_limit = MAX_LFS_FILESIZE;
2139
2140 /* small i_blocks in vfs inode? */
2141 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2142 /*
2143 * CONFIG_LBDAF is not enabled implies the inode
2144 * i_block represent total blocks in 512 bytes
2145 * 32 == size of vfs inode i_blocks * 8
2146 */
2147 upper_limit = (1LL << 32) - 1;
2148
2149 /* total blocks in file system block size */
2150 upper_limit >>= (blkbits - 9);
2151 upper_limit <<= blkbits;
2152 }
2153
2154 /* 32-bit extent-start container, ee_block */
2155 res = 1LL << 32;
2156 res <<= blkbits;
2157 res -= 1;
2158
2159 /* Sanity check against vm- & vfs- imposed limits */
2160 if (res > upper_limit)
2161 res = upper_limit;
2162
2163 return res;
2164 }
2165
2166 /*
2167 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2168 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2169 * We need to be 1 filesystem block less than the 2^48 sector limit.
2170 */
2171 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2172 {
2173 loff_t res = EXT4_NDIR_BLOCKS;
2174 int meta_blocks;
2175 loff_t upper_limit;
2176 /* This is calculated to be the largest file size for a dense, block
2177 * mapped file such that the file's total number of 512-byte sectors,
2178 * including data and all indirect blocks, does not exceed (2^48 - 1).
2179 *
2180 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2181 * number of 512-byte sectors of the file.
2182 */
2183
2184 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2185 /*
2186 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2187 * the inode i_block field represents total file blocks in
2188 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2189 */
2190 upper_limit = (1LL << 32) - 1;
2191
2192 /* total blocks in file system block size */
2193 upper_limit >>= (bits - 9);
2194
2195 } else {
2196 /*
2197 * We use 48 bit ext4_inode i_blocks
2198 * With EXT4_HUGE_FILE_FL set the i_blocks
2199 * represent total number of blocks in
2200 * file system block size
2201 */
2202 upper_limit = (1LL << 48) - 1;
2203
2204 }
2205
2206 /* indirect blocks */
2207 meta_blocks = 1;
2208 /* double indirect blocks */
2209 meta_blocks += 1 + (1LL << (bits-2));
2210 /* tripple indirect blocks */
2211 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2212
2213 upper_limit -= meta_blocks;
2214 upper_limit <<= bits;
2215
2216 res += 1LL << (bits-2);
2217 res += 1LL << (2*(bits-2));
2218 res += 1LL << (3*(bits-2));
2219 res <<= bits;
2220 if (res > upper_limit)
2221 res = upper_limit;
2222
2223 if (res > MAX_LFS_FILESIZE)
2224 res = MAX_LFS_FILESIZE;
2225
2226 return res;
2227 }
2228
2229 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2230 ext4_fsblk_t logical_sb_block, int nr)
2231 {
2232 struct ext4_sb_info *sbi = EXT4_SB(sb);
2233 ext4_group_t bg, first_meta_bg;
2234 int has_super = 0;
2235
2236 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2237
2238 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2239 nr < first_meta_bg)
2240 return logical_sb_block + nr + 1;
2241 bg = sbi->s_desc_per_block * nr;
2242 if (ext4_bg_has_super(sb, bg))
2243 has_super = 1;
2244
2245 return (has_super + ext4_group_first_block_no(sb, bg));
2246 }
2247
2248 /**
2249 * ext4_get_stripe_size: Get the stripe size.
2250 * @sbi: In memory super block info
2251 *
2252 * If we have specified it via mount option, then
2253 * use the mount option value. If the value specified at mount time is
2254 * greater than the blocks per group use the super block value.
2255 * If the super block value is greater than blocks per group return 0.
2256 * Allocator needs it be less than blocks per group.
2257 *
2258 */
2259 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2260 {
2261 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2262 unsigned long stripe_width =
2263 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2264
2265 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2266 return sbi->s_stripe;
2267
2268 if (stripe_width <= sbi->s_blocks_per_group)
2269 return stripe_width;
2270
2271 if (stride <= sbi->s_blocks_per_group)
2272 return stride;
2273
2274 return 0;
2275 }
2276
2277 /* sysfs supprt */
2278
2279 struct ext4_attr {
2280 struct attribute attr;
2281 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2282 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2283 const char *, size_t);
2284 int offset;
2285 };
2286
2287 static int parse_strtoul(const char *buf,
2288 unsigned long max, unsigned long *value)
2289 {
2290 char *endp;
2291
2292 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2293 endp = skip_spaces(endp);
2294 if (*endp || *value > max)
2295 return -EINVAL;
2296
2297 return 0;
2298 }
2299
2300 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2301 struct ext4_sb_info *sbi,
2302 char *buf)
2303 {
2304 return snprintf(buf, PAGE_SIZE, "%llu\n",
2305 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2306 }
2307
2308 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2309 struct ext4_sb_info *sbi, char *buf)
2310 {
2311 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2312
2313 if (!sb->s_bdev->bd_part)
2314 return snprintf(buf, PAGE_SIZE, "0\n");
2315 return snprintf(buf, PAGE_SIZE, "%lu\n",
2316 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2317 sbi->s_sectors_written_start) >> 1);
2318 }
2319
2320 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2321 struct ext4_sb_info *sbi, char *buf)
2322 {
2323 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2324
2325 if (!sb->s_bdev->bd_part)
2326 return snprintf(buf, PAGE_SIZE, "0\n");
2327 return snprintf(buf, PAGE_SIZE, "%llu\n",
2328 (unsigned long long)(sbi->s_kbytes_written +
2329 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2330 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2331 }
2332
2333 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2334 struct ext4_sb_info *sbi,
2335 const char *buf, size_t count)
2336 {
2337 unsigned long t;
2338
2339 if (parse_strtoul(buf, 0x40000000, &t))
2340 return -EINVAL;
2341
2342 if (!is_power_of_2(t))
2343 return -EINVAL;
2344
2345 sbi->s_inode_readahead_blks = t;
2346 return count;
2347 }
2348
2349 static ssize_t sbi_ui_show(struct ext4_attr *a,
2350 struct ext4_sb_info *sbi, char *buf)
2351 {
2352 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2353
2354 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2355 }
2356
2357 static ssize_t sbi_ui_store(struct ext4_attr *a,
2358 struct ext4_sb_info *sbi,
2359 const char *buf, size_t count)
2360 {
2361 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2362 unsigned long t;
2363
2364 if (parse_strtoul(buf, 0xffffffff, &t))
2365 return -EINVAL;
2366 *ui = t;
2367 return count;
2368 }
2369
2370 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2371 static struct ext4_attr ext4_attr_##_name = { \
2372 .attr = {.name = __stringify(_name), .mode = _mode }, \
2373 .show = _show, \
2374 .store = _store, \
2375 .offset = offsetof(struct ext4_sb_info, _elname), \
2376 }
2377 #define EXT4_ATTR(name, mode, show, store) \
2378 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2379
2380 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2381 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2382 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2383 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2384 #define ATTR_LIST(name) &ext4_attr_##name.attr
2385
2386 EXT4_RO_ATTR(delayed_allocation_blocks);
2387 EXT4_RO_ATTR(session_write_kbytes);
2388 EXT4_RO_ATTR(lifetime_write_kbytes);
2389 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2390 inode_readahead_blks_store, s_inode_readahead_blks);
2391 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2392 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2393 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2394 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2395 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2396 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2397 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2398 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2399
2400 static struct attribute *ext4_attrs[] = {
2401 ATTR_LIST(delayed_allocation_blocks),
2402 ATTR_LIST(session_write_kbytes),
2403 ATTR_LIST(lifetime_write_kbytes),
2404 ATTR_LIST(inode_readahead_blks),
2405 ATTR_LIST(inode_goal),
2406 ATTR_LIST(mb_stats),
2407 ATTR_LIST(mb_max_to_scan),
2408 ATTR_LIST(mb_min_to_scan),
2409 ATTR_LIST(mb_order2_req),
2410 ATTR_LIST(mb_stream_req),
2411 ATTR_LIST(mb_group_prealloc),
2412 ATTR_LIST(max_writeback_mb_bump),
2413 NULL,
2414 };
2415
2416 static ssize_t ext4_attr_show(struct kobject *kobj,
2417 struct attribute *attr, char *buf)
2418 {
2419 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2420 s_kobj);
2421 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2422
2423 return a->show ? a->show(a, sbi, buf) : 0;
2424 }
2425
2426 static ssize_t ext4_attr_store(struct kobject *kobj,
2427 struct attribute *attr,
2428 const char *buf, size_t len)
2429 {
2430 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2431 s_kobj);
2432 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2433
2434 return a->store ? a->store(a, sbi, buf, len) : 0;
2435 }
2436
2437 static void ext4_sb_release(struct kobject *kobj)
2438 {
2439 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2440 s_kobj);
2441 complete(&sbi->s_kobj_unregister);
2442 }
2443
2444
2445 static const struct sysfs_ops ext4_attr_ops = {
2446 .show = ext4_attr_show,
2447 .store = ext4_attr_store,
2448 };
2449
2450 static struct kobj_type ext4_ktype = {
2451 .default_attrs = ext4_attrs,
2452 .sysfs_ops = &ext4_attr_ops,
2453 .release = ext4_sb_release,
2454 };
2455
2456 /*
2457 * Check whether this filesystem can be mounted based on
2458 * the features present and the RDONLY/RDWR mount requested.
2459 * Returns 1 if this filesystem can be mounted as requested,
2460 * 0 if it cannot be.
2461 */
2462 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2463 {
2464 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2465 ext4_msg(sb, KERN_ERR,
2466 "Couldn't mount because of "
2467 "unsupported optional features (%x)",
2468 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2469 ~EXT4_FEATURE_INCOMPAT_SUPP));
2470 return 0;
2471 }
2472
2473 if (readonly)
2474 return 1;
2475
2476 /* Check that feature set is OK for a read-write mount */
2477 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2478 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2479 "unsupported optional features (%x)",
2480 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2481 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2482 return 0;
2483 }
2484 /*
2485 * Large file size enabled file system can only be mounted
2486 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2487 */
2488 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2489 if (sizeof(blkcnt_t) < sizeof(u64)) {
2490 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2491 "cannot be mounted RDWR without "
2492 "CONFIG_LBDAF");
2493 return 0;
2494 }
2495 }
2496 return 1;
2497 }
2498
2499 /*
2500 * This function is called once a day if we have errors logged
2501 * on the file system
2502 */
2503 static void print_daily_error_info(unsigned long arg)
2504 {
2505 struct super_block *sb = (struct super_block *) arg;
2506 struct ext4_sb_info *sbi;
2507 struct ext4_super_block *es;
2508
2509 sbi = EXT4_SB(sb);
2510 es = sbi->s_es;
2511
2512 if (es->s_error_count)
2513 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2514 le32_to_cpu(es->s_error_count));
2515 if (es->s_first_error_time) {
2516 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2517 sb->s_id, le32_to_cpu(es->s_first_error_time),
2518 (int) sizeof(es->s_first_error_func),
2519 es->s_first_error_func,
2520 le32_to_cpu(es->s_first_error_line));
2521 if (es->s_first_error_ino)
2522 printk(": inode %u",
2523 le32_to_cpu(es->s_first_error_ino));
2524 if (es->s_first_error_block)
2525 printk(": block %llu", (unsigned long long)
2526 le64_to_cpu(es->s_first_error_block));
2527 printk("\n");
2528 }
2529 if (es->s_last_error_time) {
2530 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2531 sb->s_id, le32_to_cpu(es->s_last_error_time),
2532 (int) sizeof(es->s_last_error_func),
2533 es->s_last_error_func,
2534 le32_to_cpu(es->s_last_error_line));
2535 if (es->s_last_error_ino)
2536 printk(": inode %u",
2537 le32_to_cpu(es->s_last_error_ino));
2538 if (es->s_last_error_block)
2539 printk(": block %llu", (unsigned long long)
2540 le64_to_cpu(es->s_last_error_block));
2541 printk("\n");
2542 }
2543 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2544 }
2545
2546 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2547 __releases(kernel_lock)
2548 __acquires(kernel_lock)
2549 {
2550 char *orig_data = kstrdup(data, GFP_KERNEL);
2551 struct buffer_head *bh;
2552 struct ext4_super_block *es = NULL;
2553 struct ext4_sb_info *sbi;
2554 ext4_fsblk_t block;
2555 ext4_fsblk_t sb_block = get_sb_block(&data);
2556 ext4_fsblk_t logical_sb_block;
2557 unsigned long offset = 0;
2558 unsigned long journal_devnum = 0;
2559 unsigned long def_mount_opts;
2560 struct inode *root;
2561 char *cp;
2562 const char *descr;
2563 int ret = -ENOMEM;
2564 int blocksize;
2565 unsigned int db_count;
2566 unsigned int i;
2567 int needs_recovery, has_huge_files;
2568 __u64 blocks_count;
2569 int err;
2570 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2571
2572 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2573 if (!sbi)
2574 goto out_free_orig;
2575
2576 sbi->s_blockgroup_lock =
2577 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2578 if (!sbi->s_blockgroup_lock) {
2579 kfree(sbi);
2580 goto out_free_orig;
2581 }
2582 sb->s_fs_info = sbi;
2583 sbi->s_mount_opt = 0;
2584 sbi->s_resuid = EXT4_DEF_RESUID;
2585 sbi->s_resgid = EXT4_DEF_RESGID;
2586 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2587 sbi->s_sb_block = sb_block;
2588 if (sb->s_bdev->bd_part)
2589 sbi->s_sectors_written_start =
2590 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2591
2592 unlock_kernel();
2593
2594 /* Cleanup superblock name */
2595 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2596 *cp = '!';
2597
2598 ret = -EINVAL;
2599 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2600 if (!blocksize) {
2601 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
2602 goto out_fail;
2603 }
2604
2605 /*
2606 * The ext4 superblock will not be buffer aligned for other than 1kB
2607 * block sizes. We need to calculate the offset from buffer start.
2608 */
2609 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
2610 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2611 offset = do_div(logical_sb_block, blocksize);
2612 } else {
2613 logical_sb_block = sb_block;
2614 }
2615
2616 if (!(bh = sb_bread(sb, logical_sb_block))) {
2617 ext4_msg(sb, KERN_ERR, "unable to read superblock");
2618 goto out_fail;
2619 }
2620 /*
2621 * Note: s_es must be initialized as soon as possible because
2622 * some ext4 macro-instructions depend on its value
2623 */
2624 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2625 sbi->s_es = es;
2626 sb->s_magic = le16_to_cpu(es->s_magic);
2627 if (sb->s_magic != EXT4_SUPER_MAGIC)
2628 goto cantfind_ext4;
2629 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
2630
2631 /* Set defaults before we parse the mount options */
2632 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
2633 if (def_mount_opts & EXT4_DEFM_DEBUG)
2634 set_opt(sbi->s_mount_opt, DEBUG);
2635 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2636 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2637 "2.6.38");
2638 set_opt(sbi->s_mount_opt, GRPID);
2639 }
2640 if (def_mount_opts & EXT4_DEFM_UID16)
2641 set_opt(sbi->s_mount_opt, NO_UID32);
2642 #ifdef CONFIG_EXT4_FS_XATTR
2643 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
2644 set_opt(sbi->s_mount_opt, XATTR_USER);
2645 #endif
2646 #ifdef CONFIG_EXT4_FS_POSIX_ACL
2647 if (def_mount_opts & EXT4_DEFM_ACL)
2648 set_opt(sbi->s_mount_opt, POSIX_ACL);
2649 #endif
2650 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2651 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2652 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2653 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2654 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2655 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2656
2657 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
2658 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
2659 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
2660 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2661 else
2662 set_opt(sbi->s_mount_opt, ERRORS_RO);
2663 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
2664 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
2665 if (def_mount_opts & EXT4_DEFM_DISCARD)
2666 set_opt(sbi->s_mount_opt, DISCARD);
2667
2668 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2669 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2670 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2671 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2672 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2673
2674 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
2675 set_opt(sbi->s_mount_opt, BARRIER);
2676
2677 /*
2678 * enable delayed allocation by default
2679 * Use -o nodelalloc to turn it off
2680 */
2681 if (!IS_EXT3_SB(sb) &&
2682 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
2683 set_opt(sbi->s_mount_opt, DELALLOC);
2684
2685 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
2686 &journal_devnum, &journal_ioprio, NULL, 0)) {
2687 ext4_msg(sb, KERN_WARNING,
2688 "failed to parse options in superblock: %s",
2689 sbi->s_es->s_mount_opts);
2690 }
2691 if (!parse_options((char *) data, sb, &journal_devnum,
2692 &journal_ioprio, NULL, 0))
2693 goto failed_mount;
2694
2695 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2696 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
2697
2698 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2699 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2700 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2701 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
2702 ext4_msg(sb, KERN_WARNING,
2703 "feature flags set on rev 0 fs, "
2704 "running e2fsck is recommended");
2705
2706 /*
2707 * Check feature flags regardless of the revision level, since we
2708 * previously didn't change the revision level when setting the flags,
2709 * so there is a chance incompat flags are set on a rev 0 filesystem.
2710 */
2711 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
2712 goto failed_mount;
2713
2714 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2715
2716 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2717 blocksize > EXT4_MAX_BLOCK_SIZE) {
2718 ext4_msg(sb, KERN_ERR,
2719 "Unsupported filesystem blocksize %d", blocksize);
2720 goto failed_mount;
2721 }
2722
2723 if (sb->s_blocksize != blocksize) {
2724 /* Validate the filesystem blocksize */
2725 if (!sb_set_blocksize(sb, blocksize)) {
2726 ext4_msg(sb, KERN_ERR, "bad block size %d",
2727 blocksize);
2728 goto failed_mount;
2729 }
2730
2731 brelse(bh);
2732 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2733 offset = do_div(logical_sb_block, blocksize);
2734 bh = sb_bread(sb, logical_sb_block);
2735 if (!bh) {
2736 ext4_msg(sb, KERN_ERR,
2737 "Can't read superblock on 2nd try");
2738 goto failed_mount;
2739 }
2740 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
2741 sbi->s_es = es;
2742 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2743 ext4_msg(sb, KERN_ERR,
2744 "Magic mismatch, very weird!");
2745 goto failed_mount;
2746 }
2747 }
2748
2749 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2750 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2751 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2752 has_huge_files);
2753 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
2754
2755 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2756 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2757 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
2758 } else {
2759 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2760 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
2761 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
2762 (!is_power_of_2(sbi->s_inode_size)) ||
2763 (sbi->s_inode_size > blocksize)) {
2764 ext4_msg(sb, KERN_ERR,
2765 "unsupported inode size: %d",
2766 sbi->s_inode_size);
2767 goto failed_mount;
2768 }
2769 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2770 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
2771 }
2772
2773 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2774 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
2775 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
2776 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
2777 !is_power_of_2(sbi->s_desc_size)) {
2778 ext4_msg(sb, KERN_ERR,
2779 "unsupported descriptor size %lu",
2780 sbi->s_desc_size);
2781 goto failed_mount;
2782 }
2783 } else
2784 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
2785
2786 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
2787 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
2788 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
2789 goto cantfind_ext4;
2790
2791 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
2792 if (sbi->s_inodes_per_block == 0)
2793 goto cantfind_ext4;
2794 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2795 sbi->s_inodes_per_block;
2796 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
2797 sbi->s_sbh = bh;
2798 sbi->s_mount_state = le16_to_cpu(es->s_state);
2799 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2800 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2801
2802 for (i = 0; i < 4; i++)
2803 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2804 sbi->s_def_hash_version = es->s_def_hash_version;
2805 i = le32_to_cpu(es->s_flags);
2806 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2807 sbi->s_hash_unsigned = 3;
2808 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2809 #ifdef __CHAR_UNSIGNED__
2810 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2811 sbi->s_hash_unsigned = 3;
2812 #else
2813 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2814 #endif
2815 sb->s_dirt = 1;
2816 }
2817
2818 if (sbi->s_blocks_per_group > blocksize * 8) {
2819 ext4_msg(sb, KERN_ERR,
2820 "#blocks per group too big: %lu",
2821 sbi->s_blocks_per_group);
2822 goto failed_mount;
2823 }
2824 if (sbi->s_inodes_per_group > blocksize * 8) {
2825 ext4_msg(sb, KERN_ERR,
2826 "#inodes per group too big: %lu",
2827 sbi->s_inodes_per_group);
2828 goto failed_mount;
2829 }
2830
2831 /*
2832 * Test whether we have more sectors than will fit in sector_t,
2833 * and whether the max offset is addressable by the page cache.
2834 */
2835 if ((ext4_blocks_count(es) >
2836 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2837 (ext4_blocks_count(es) >
2838 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2839 ext4_msg(sb, KERN_ERR, "filesystem"
2840 " too large to mount safely on this system");
2841 if (sizeof(sector_t) < 8)
2842 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2843 ret = -EFBIG;
2844 goto failed_mount;
2845 }
2846
2847 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2848 goto cantfind_ext4;
2849
2850 /* check blocks count against device size */
2851 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2852 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
2853 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2854 "exceeds size of device (%llu blocks)",
2855 ext4_blocks_count(es), blocks_count);
2856 goto failed_mount;
2857 }
2858
2859 /*
2860 * It makes no sense for the first data block to be beyond the end
2861 * of the filesystem.
2862 */
2863 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2864 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2865 "block %u is beyond end of filesystem (%llu)",
2866 le32_to_cpu(es->s_first_data_block),
2867 ext4_blocks_count(es));
2868 goto failed_mount;
2869 }
2870 blocks_count = (ext4_blocks_count(es) -
2871 le32_to_cpu(es->s_first_data_block) +
2872 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2873 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2874 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2875 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
2876 "(block count %llu, first data block %u, "
2877 "blocks per group %lu)", sbi->s_groups_count,
2878 ext4_blocks_count(es),
2879 le32_to_cpu(es->s_first_data_block),
2880 EXT4_BLOCKS_PER_GROUP(sb));
2881 goto failed_mount;
2882 }
2883 sbi->s_groups_count = blocks_count;
2884 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2885 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
2886 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2887 EXT4_DESC_PER_BLOCK(sb);
2888 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
2889 GFP_KERNEL);
2890 if (sbi->s_group_desc == NULL) {
2891 ext4_msg(sb, KERN_ERR, "not enough memory");
2892 goto failed_mount;
2893 }
2894
2895 #ifdef CONFIG_PROC_FS
2896 if (ext4_proc_root)
2897 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2898 #endif
2899
2900 bgl_lock_init(sbi->s_blockgroup_lock);
2901
2902 for (i = 0; i < db_count; i++) {
2903 block = descriptor_loc(sb, logical_sb_block, i);
2904 sbi->s_group_desc[i] = sb_bread(sb, block);
2905 if (!sbi->s_group_desc[i]) {
2906 ext4_msg(sb, KERN_ERR,
2907 "can't read group descriptor %d", i);
2908 db_count = i;
2909 goto failed_mount2;
2910 }
2911 }
2912 if (!ext4_check_descriptors(sb)) {
2913 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2914 goto failed_mount2;
2915 }
2916 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2917 if (!ext4_fill_flex_info(sb)) {
2918 ext4_msg(sb, KERN_ERR,
2919 "unable to initialize "
2920 "flex_bg meta info!");
2921 goto failed_mount2;
2922 }
2923
2924 sbi->s_gdb_count = db_count;
2925 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2926 spin_lock_init(&sbi->s_next_gen_lock);
2927
2928 sbi->s_stripe = ext4_get_stripe_size(sbi);
2929 sbi->s_max_writeback_mb_bump = 128;
2930
2931 /*
2932 * set up enough so that it can read an inode
2933 */
2934 if (!test_opt(sb, NOLOAD) &&
2935 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2936 sb->s_op = &ext4_sops;
2937 else
2938 sb->s_op = &ext4_nojournal_sops;
2939 sb->s_export_op = &ext4_export_ops;
2940 sb->s_xattr = ext4_xattr_handlers;
2941 #ifdef CONFIG_QUOTA
2942 sb->s_qcop = &ext4_qctl_operations;
2943 sb->dq_op = &ext4_quota_operations;
2944 #endif
2945 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2946 mutex_init(&sbi->s_orphan_lock);
2947 mutex_init(&sbi->s_resize_lock);
2948
2949 sb->s_root = NULL;
2950
2951 needs_recovery = (es->s_last_orphan != 0 ||
2952 EXT4_HAS_INCOMPAT_FEATURE(sb,
2953 EXT4_FEATURE_INCOMPAT_RECOVER));
2954
2955 /*
2956 * The first inode we look at is the journal inode. Don't try
2957 * root first: it may be modified in the journal!
2958 */
2959 if (!test_opt(sb, NOLOAD) &&
2960 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2961 if (ext4_load_journal(sb, es, journal_devnum))
2962 goto failed_mount3;
2963 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2964 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2965 ext4_msg(sb, KERN_ERR, "required journal recovery "
2966 "suppressed and not mounted read-only");
2967 goto failed_mount_wq;
2968 } else {
2969 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2970 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2971 sbi->s_journal = NULL;
2972 needs_recovery = 0;
2973 goto no_journal;
2974 }
2975
2976 if (ext4_blocks_count(es) > 0xffffffffULL &&
2977 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2978 JBD2_FEATURE_INCOMPAT_64BIT)) {
2979 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
2980 goto failed_mount_wq;
2981 }
2982
2983 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2984 jbd2_journal_set_features(sbi->s_journal,
2985 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2986 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2987 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2988 jbd2_journal_set_features(sbi->s_journal,
2989 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2990 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2991 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2992 } else {
2993 jbd2_journal_clear_features(sbi->s_journal,
2994 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2995 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2996 }
2997
2998 /* We have now updated the journal if required, so we can
2999 * validate the data journaling mode. */
3000 switch (test_opt(sb, DATA_FLAGS)) {
3001 case 0:
3002 /* No mode set, assume a default based on the journal
3003 * capabilities: ORDERED_DATA if the journal can
3004 * cope, else JOURNAL_DATA
3005 */
3006 if (jbd2_journal_check_available_features
3007 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3008 set_opt(sbi->s_mount_opt, ORDERED_DATA);
3009 else
3010 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
3011 break;
3012
3013 case EXT4_MOUNT_ORDERED_DATA:
3014 case EXT4_MOUNT_WRITEBACK_DATA:
3015 if (!jbd2_journal_check_available_features
3016 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3017 ext4_msg(sb, KERN_ERR, "Journal does not support "
3018 "requested data journaling mode");
3019 goto failed_mount_wq;
3020 }
3021 default:
3022 break;
3023 }
3024 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3025
3026 no_journal:
3027 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3028 ext4_count_free_blocks(sb));
3029 if (!err)
3030 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3031 ext4_count_free_inodes(sb));
3032 if (!err)
3033 err = percpu_counter_init(&sbi->s_dirs_counter,
3034 ext4_count_dirs(sb));
3035 if (!err)
3036 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3037 if (err) {
3038 ext4_msg(sb, KERN_ERR, "insufficient memory");
3039 goto failed_mount_wq;
3040 }
3041
3042 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3043 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3044 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3045 goto failed_mount_wq;
3046 }
3047
3048 /*
3049 * The jbd2_journal_load will have done any necessary log recovery,
3050 * so we can safely mount the rest of the filesystem now.
3051 */
3052
3053 root = ext4_iget(sb, EXT4_ROOT_INO);
3054 if (IS_ERR(root)) {
3055 ext4_msg(sb, KERN_ERR, "get root inode failed");
3056 ret = PTR_ERR(root);
3057 goto failed_mount4;
3058 }
3059 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3060 iput(root);
3061 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3062 goto failed_mount4;
3063 }
3064 sb->s_root = d_alloc_root(root);
3065 if (!sb->s_root) {
3066 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3067 iput(root);
3068 ret = -ENOMEM;
3069 goto failed_mount4;
3070 }
3071
3072 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3073
3074 /* determine the minimum size of new large inodes, if present */
3075 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3076 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3077 EXT4_GOOD_OLD_INODE_SIZE;
3078 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3079 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3080 if (sbi->s_want_extra_isize <
3081 le16_to_cpu(es->s_want_extra_isize))
3082 sbi->s_want_extra_isize =
3083 le16_to_cpu(es->s_want_extra_isize);
3084 if (sbi->s_want_extra_isize <
3085 le16_to_cpu(es->s_min_extra_isize))
3086 sbi->s_want_extra_isize =
3087 le16_to_cpu(es->s_min_extra_isize);
3088 }
3089 }
3090 /* Check if enough inode space is available */
3091 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3092 sbi->s_inode_size) {
3093 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3094 EXT4_GOOD_OLD_INODE_SIZE;
3095 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3096 "available");
3097 }
3098
3099 if (test_opt(sb, DELALLOC) &&
3100 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3101 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3102 "requested data journaling mode");
3103 clear_opt(sbi->s_mount_opt, DELALLOC);
3104 }
3105 if (test_opt(sb, DIOREAD_NOLOCK)) {
3106 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3107 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3108 "option - requested data journaling mode");
3109 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3110 }
3111 if (sb->s_blocksize < PAGE_SIZE) {
3112 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3113 "option - block size is too small");
3114 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3115 }
3116 }
3117
3118 err = ext4_setup_system_zone(sb);
3119 if (err) {
3120 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3121 "zone (%d)", err);
3122 goto failed_mount4;
3123 }
3124
3125 ext4_ext_init(sb);
3126 err = ext4_mb_init(sb, needs_recovery);
3127 if (err) {
3128 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3129 err);
3130 goto failed_mount4;
3131 }
3132
3133 sbi->s_kobj.kset = ext4_kset;
3134 init_completion(&sbi->s_kobj_unregister);
3135 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3136 "%s", sb->s_id);
3137 if (err) {
3138 ext4_mb_release(sb);
3139 ext4_ext_release(sb);
3140 goto failed_mount4;
3141 };
3142
3143 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3144 ext4_orphan_cleanup(sb, es);
3145 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3146 if (needs_recovery) {
3147 ext4_msg(sb, KERN_INFO, "recovery complete");
3148 ext4_mark_recovery_complete(sb, es);
3149 }
3150 if (EXT4_SB(sb)->s_journal) {
3151 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3152 descr = " journalled data mode";
3153 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3154 descr = " ordered data mode";
3155 else
3156 descr = " writeback data mode";
3157 } else
3158 descr = "out journal";
3159
3160 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3161 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3162 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3163
3164 init_timer(&sbi->s_err_report);
3165 sbi->s_err_report.function = print_daily_error_info;
3166 sbi->s_err_report.data = (unsigned long) sb;
3167 if (es->s_error_count)
3168 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3169
3170 lock_kernel();
3171 kfree(orig_data);
3172 return 0;
3173
3174 cantfind_ext4:
3175 if (!silent)
3176 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3177 goto failed_mount;
3178
3179 failed_mount4:
3180 ext4_msg(sb, KERN_ERR, "mount failed");
3181 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3182 failed_mount_wq:
3183 ext4_release_system_zone(sb);
3184 if (sbi->s_journal) {
3185 jbd2_journal_destroy(sbi->s_journal);
3186 sbi->s_journal = NULL;
3187 }
3188 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3189 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3190 percpu_counter_destroy(&sbi->s_dirs_counter);
3191 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3192 failed_mount3:
3193 if (sbi->s_flex_groups) {
3194 if (is_vmalloc_addr(sbi->s_flex_groups))
3195 vfree(sbi->s_flex_groups);
3196 else
3197 kfree(sbi->s_flex_groups);
3198 }
3199 failed_mount2:
3200 for (i = 0; i < db_count; i++)
3201 brelse(sbi->s_group_desc[i]);
3202 kfree(sbi->s_group_desc);
3203 failed_mount:
3204 if (sbi->s_proc) {
3205 remove_proc_entry(sb->s_id, ext4_proc_root);
3206 }
3207 #ifdef CONFIG_QUOTA
3208 for (i = 0; i < MAXQUOTAS; i++)
3209 kfree(sbi->s_qf_names[i]);
3210 #endif
3211 ext4_blkdev_remove(sbi);
3212 brelse(bh);
3213 out_fail:
3214 sb->s_fs_info = NULL;
3215 kfree(sbi->s_blockgroup_lock);
3216 kfree(sbi);
3217 lock_kernel();
3218 out_free_orig:
3219 kfree(orig_data);
3220 return ret;
3221 }
3222
3223 /*
3224 * Setup any per-fs journal parameters now. We'll do this both on
3225 * initial mount, once the journal has been initialised but before we've
3226 * done any recovery; and again on any subsequent remount.
3227 */
3228 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3229 {
3230 struct ext4_sb_info *sbi = EXT4_SB(sb);
3231
3232 journal->j_commit_interval = sbi->s_commit_interval;
3233 journal->j_min_batch_time = sbi->s_min_batch_time;
3234 journal->j_max_batch_time = sbi->s_max_batch_time;
3235
3236 write_lock(&journal->j_state_lock);
3237 if (test_opt(sb, BARRIER))
3238 journal->j_flags |= JBD2_BARRIER;
3239 else
3240 journal->j_flags &= ~JBD2_BARRIER;
3241 if (test_opt(sb, DATA_ERR_ABORT))
3242 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3243 else
3244 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3245 write_unlock(&journal->j_state_lock);
3246 }
3247
3248 static journal_t *ext4_get_journal(struct super_block *sb,
3249 unsigned int journal_inum)
3250 {
3251 struct inode *journal_inode;
3252 journal_t *journal;
3253
3254 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3255
3256 /* First, test for the existence of a valid inode on disk. Bad
3257 * things happen if we iget() an unused inode, as the subsequent
3258 * iput() will try to delete it. */
3259
3260 journal_inode = ext4_iget(sb, journal_inum);
3261 if (IS_ERR(journal_inode)) {
3262 ext4_msg(sb, KERN_ERR, "no journal found");
3263 return NULL;
3264 }
3265 if (!journal_inode->i_nlink) {
3266 make_bad_inode(journal_inode);
3267 iput(journal_inode);
3268 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3269 return NULL;
3270 }
3271
3272 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3273 journal_inode, journal_inode->i_size);
3274 if (!S_ISREG(journal_inode->i_mode)) {
3275 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3276 iput(journal_inode);
3277 return NULL;
3278 }
3279
3280 journal = jbd2_journal_init_inode(journal_inode);
3281 if (!journal) {
3282 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3283 iput(journal_inode);
3284 return NULL;
3285 }
3286 journal->j_private = sb;
3287 ext4_init_journal_params(sb, journal);
3288 return journal;
3289 }
3290
3291 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3292 dev_t j_dev)
3293 {
3294 struct buffer_head *bh;
3295 journal_t *journal;
3296 ext4_fsblk_t start;
3297 ext4_fsblk_t len;
3298 int hblock, blocksize;
3299 ext4_fsblk_t sb_block;
3300 unsigned long offset;
3301 struct ext4_super_block *es;
3302 struct block_device *bdev;
3303
3304 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3305
3306 bdev = ext4_blkdev_get(j_dev, sb);
3307 if (bdev == NULL)
3308 return NULL;
3309
3310 if (bd_claim(bdev, sb)) {
3311 ext4_msg(sb, KERN_ERR,
3312 "failed to claim external journal device");
3313 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
3314 return NULL;
3315 }
3316
3317 blocksize = sb->s_blocksize;
3318 hblock = bdev_logical_block_size(bdev);
3319 if (blocksize < hblock) {
3320 ext4_msg(sb, KERN_ERR,
3321 "blocksize too small for journal device");
3322 goto out_bdev;
3323 }
3324
3325 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3326 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3327 set_blocksize(bdev, blocksize);
3328 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3329 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3330 "external journal");
3331 goto out_bdev;
3332 }
3333
3334 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3335 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3336 !(le32_to_cpu(es->s_feature_incompat) &
3337 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3338 ext4_msg(sb, KERN_ERR, "external journal has "
3339 "bad superblock");
3340 brelse(bh);
3341 goto out_bdev;
3342 }
3343
3344 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3345 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3346 brelse(bh);
3347 goto out_bdev;
3348 }
3349
3350 len = ext4_blocks_count(es);
3351 start = sb_block + 1;
3352 brelse(bh); /* we're done with the superblock */
3353
3354 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3355 start, len, blocksize);
3356 if (!journal) {
3357 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3358 goto out_bdev;
3359 }
3360 journal->j_private = sb;
3361 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3362 wait_on_buffer(journal->j_sb_buffer);
3363 if (!buffer_uptodate(journal->j_sb_buffer)) {
3364 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3365 goto out_journal;
3366 }
3367 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3368 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3369 "user (unsupported) - %d",
3370 be32_to_cpu(journal->j_superblock->s_nr_users));
3371 goto out_journal;
3372 }
3373 EXT4_SB(sb)->journal_bdev = bdev;
3374 ext4_init_journal_params(sb, journal);
3375 return journal;
3376
3377 out_journal:
3378 jbd2_journal_destroy(journal);
3379 out_bdev:
3380 ext4_blkdev_put(bdev);
3381 return NULL;
3382 }
3383
3384 static int ext4_load_journal(struct super_block *sb,
3385 struct ext4_super_block *es,
3386 unsigned long journal_devnum)
3387 {
3388 journal_t *journal;
3389 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3390 dev_t journal_dev;
3391 int err = 0;
3392 int really_read_only;
3393
3394 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3395
3396 if (journal_devnum &&
3397 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3398 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3399 "numbers have changed");
3400 journal_dev = new_decode_dev(journal_devnum);
3401 } else
3402 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3403
3404 really_read_only = bdev_read_only(sb->s_bdev);
3405
3406 /*
3407 * Are we loading a blank journal or performing recovery after a
3408 * crash? For recovery, we need to check in advance whether we
3409 * can get read-write access to the device.
3410 */
3411 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3412 if (sb->s_flags & MS_RDONLY) {
3413 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3414 "required on readonly filesystem");
3415 if (really_read_only) {
3416 ext4_msg(sb, KERN_ERR, "write access "
3417 "unavailable, cannot proceed");
3418 return -EROFS;
3419 }
3420 ext4_msg(sb, KERN_INFO, "write access will "
3421 "be enabled during recovery");
3422 }
3423 }
3424
3425 if (journal_inum && journal_dev) {
3426 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3427 "and inode journals!");
3428 return -EINVAL;
3429 }
3430
3431 if (journal_inum) {
3432 if (!(journal = ext4_get_journal(sb, journal_inum)))
3433 return -EINVAL;
3434 } else {
3435 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3436 return -EINVAL;
3437 }
3438
3439 if (!(journal->j_flags & JBD2_BARRIER))
3440 ext4_msg(sb, KERN_INFO, "barriers disabled");
3441
3442 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
3443 err = jbd2_journal_update_format(journal);
3444 if (err) {
3445 ext4_msg(sb, KERN_ERR, "error updating journal");
3446 jbd2_journal_destroy(journal);
3447 return err;
3448 }
3449 }
3450
3451 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3452 err = jbd2_journal_wipe(journal, !really_read_only);
3453 if (!err) {
3454 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3455 if (save)
3456 memcpy(save, ((char *) es) +
3457 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3458 err = jbd2_journal_load(journal);
3459 if (save)
3460 memcpy(((char *) es) + EXT4_S_ERR_START,
3461 save, EXT4_S_ERR_LEN);
3462 kfree(save);
3463 }
3464
3465 if (err) {
3466 ext4_msg(sb, KERN_ERR, "error loading journal");
3467 jbd2_journal_destroy(journal);
3468 return err;
3469 }
3470
3471 EXT4_SB(sb)->s_journal = journal;
3472 ext4_clear_journal_err(sb, es);
3473
3474 if (journal_devnum &&
3475 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3476 es->s_journal_dev = cpu_to_le32(journal_devnum);
3477
3478 /* Make sure we flush the recovery flag to disk. */
3479 ext4_commit_super(sb, 1);
3480 }
3481
3482 return 0;
3483 }
3484
3485 static int ext4_commit_super(struct super_block *sb, int sync)
3486 {
3487 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3488 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
3489 int error = 0;
3490
3491 if (!sbh)
3492 return error;
3493 if (buffer_write_io_error(sbh)) {
3494 /*
3495 * Oh, dear. A previous attempt to write the
3496 * superblock failed. This could happen because the
3497 * USB device was yanked out. Or it could happen to
3498 * be a transient write error and maybe the block will
3499 * be remapped. Nothing we can do but to retry the
3500 * write and hope for the best.
3501 */
3502 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3503 "superblock detected");
3504 clear_buffer_write_io_error(sbh);
3505 set_buffer_uptodate(sbh);
3506 }
3507 /*
3508 * If the file system is mounted read-only, don't update the
3509 * superblock write time. This avoids updating the superblock
3510 * write time when we are mounting the root file system
3511 * read/only but we need to replay the journal; at that point,
3512 * for people who are east of GMT and who make their clock
3513 * tick in localtime for Windows bug-for-bug compatibility,
3514 * the clock is set in the future, and this will cause e2fsck
3515 * to complain and force a full file system check.
3516 */
3517 if (!(sb->s_flags & MS_RDONLY))
3518 es->s_wtime = cpu_to_le32(get_seconds());
3519 if (sb->s_bdev->bd_part)
3520 es->s_kbytes_written =
3521 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3522 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3523 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3524 else
3525 es->s_kbytes_written =
3526 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
3527 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3528 &EXT4_SB(sb)->s_freeblocks_counter));
3529 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3530 &EXT4_SB(sb)->s_freeinodes_counter));
3531 sb->s_dirt = 0;
3532 BUFFER_TRACE(sbh, "marking dirty");
3533 mark_buffer_dirty(sbh);
3534 if (sync) {
3535 error = sync_dirty_buffer(sbh);
3536 if (error)
3537 return error;
3538
3539 error = buffer_write_io_error(sbh);
3540 if (error) {
3541 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3542 "superblock");
3543 clear_buffer_write_io_error(sbh);
3544 set_buffer_uptodate(sbh);
3545 }
3546 }
3547 return error;
3548 }
3549
3550 /*
3551 * Have we just finished recovery? If so, and if we are mounting (or
3552 * remounting) the filesystem readonly, then we will end up with a
3553 * consistent fs on disk. Record that fact.
3554 */
3555 static void ext4_mark_recovery_complete(struct super_block *sb,
3556 struct ext4_super_block *es)
3557 {
3558 journal_t *journal = EXT4_SB(sb)->s_journal;
3559
3560 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3561 BUG_ON(journal != NULL);
3562 return;
3563 }
3564 jbd2_journal_lock_updates(journal);
3565 if (jbd2_journal_flush(journal) < 0)
3566 goto out;
3567
3568 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
3569 sb->s_flags & MS_RDONLY) {
3570 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3571 ext4_commit_super(sb, 1);
3572 }
3573
3574 out:
3575 jbd2_journal_unlock_updates(journal);
3576 }
3577
3578 /*
3579 * If we are mounting (or read-write remounting) a filesystem whose journal
3580 * has recorded an error from a previous lifetime, move that error to the
3581 * main filesystem now.
3582 */
3583 static void ext4_clear_journal_err(struct super_block *sb,
3584 struct ext4_super_block *es)
3585 {
3586 journal_t *journal;
3587 int j_errno;
3588 const char *errstr;
3589
3590 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3591
3592 journal = EXT4_SB(sb)->s_journal;
3593
3594 /*
3595 * Now check for any error status which may have been recorded in the
3596 * journal by a prior ext4_error() or ext4_abort()
3597 */
3598
3599 j_errno = jbd2_journal_errno(journal);
3600 if (j_errno) {
3601 char nbuf[16];
3602
3603 errstr = ext4_decode_error(sb, j_errno, nbuf);
3604 ext4_warning(sb, "Filesystem error recorded "
3605 "from previous mount: %s", errstr);
3606 ext4_warning(sb, "Marking fs in need of filesystem check.");
3607
3608 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3609 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
3610 ext4_commit_super(sb, 1);
3611
3612 jbd2_journal_clear_err(journal);
3613 }
3614 }
3615
3616 /*
3617 * Force the running and committing transactions to commit,
3618 * and wait on the commit.
3619 */
3620 int ext4_force_commit(struct super_block *sb)
3621 {
3622 journal_t *journal;
3623 int ret = 0;
3624
3625 if (sb->s_flags & MS_RDONLY)
3626 return 0;
3627
3628 journal = EXT4_SB(sb)->s_journal;
3629 if (journal) {
3630 vfs_check_frozen(sb, SB_FREEZE_TRANS);
3631 ret = ext4_journal_force_commit(journal);
3632 }
3633
3634 return ret;
3635 }
3636
3637 static void ext4_write_super(struct super_block *sb)
3638 {
3639 lock_super(sb);
3640 ext4_commit_super(sb, 1);
3641 unlock_super(sb);
3642 }
3643
3644 static int ext4_sync_fs(struct super_block *sb, int wait)
3645 {
3646 int ret = 0;
3647 tid_t target;
3648 struct ext4_sb_info *sbi = EXT4_SB(sb);
3649
3650 trace_ext4_sync_fs(sb, wait);
3651 flush_workqueue(sbi->dio_unwritten_wq);
3652 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
3653 if (wait)
3654 jbd2_log_wait_commit(sbi->s_journal, target);
3655 }
3656 return ret;
3657 }
3658
3659 /*
3660 * LVM calls this function before a (read-only) snapshot is created. This
3661 * gives us a chance to flush the journal completely and mark the fs clean.
3662 */
3663 static int ext4_freeze(struct super_block *sb)
3664 {
3665 int error = 0;
3666 journal_t *journal;
3667
3668 if (sb->s_flags & MS_RDONLY)
3669 return 0;
3670
3671 journal = EXT4_SB(sb)->s_journal;
3672
3673 /* Now we set up the journal barrier. */
3674 jbd2_journal_lock_updates(journal);
3675
3676 /*
3677 * Don't clear the needs_recovery flag if we failed to flush
3678 * the journal.
3679 */
3680 error = jbd2_journal_flush(journal);
3681 if (error < 0)
3682 goto out;
3683
3684 /* Journal blocked and flushed, clear needs_recovery flag. */
3685 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3686 error = ext4_commit_super(sb, 1);
3687 out:
3688 /* we rely on s_frozen to stop further updates */
3689 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3690 return error;
3691 }
3692
3693 /*
3694 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3695 * flag here, even though the filesystem is not technically dirty yet.
3696 */
3697 static int ext4_unfreeze(struct super_block *sb)
3698 {
3699 if (sb->s_flags & MS_RDONLY)
3700 return 0;
3701
3702 lock_super(sb);
3703 /* Reset the needs_recovery flag before the fs is unlocked. */
3704 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3705 ext4_commit_super(sb, 1);
3706 unlock_super(sb);
3707 return 0;
3708 }
3709
3710 static int ext4_remount(struct super_block *sb, int *flags, char *data)
3711 {
3712 struct ext4_super_block *es;
3713 struct ext4_sb_info *sbi = EXT4_SB(sb);
3714 ext4_fsblk_t n_blocks_count = 0;
3715 unsigned long old_sb_flags;
3716 struct ext4_mount_options old_opts;
3717 int enable_quota = 0;
3718 ext4_group_t g;
3719 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3720 int err;
3721 #ifdef CONFIG_QUOTA
3722 int i;
3723 #endif
3724 char *orig_data = kstrdup(data, GFP_KERNEL);
3725
3726 lock_kernel();
3727
3728 /* Store the original options */
3729 lock_super(sb);
3730 old_sb_flags = sb->s_flags;
3731 old_opts.s_mount_opt = sbi->s_mount_opt;
3732 old_opts.s_resuid = sbi->s_resuid;
3733 old_opts.s_resgid = sbi->s_resgid;
3734 old_opts.s_commit_interval = sbi->s_commit_interval;
3735 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3736 old_opts.s_max_batch_time = sbi->s_max_batch_time;
3737 #ifdef CONFIG_QUOTA
3738 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3739 for (i = 0; i < MAXQUOTAS; i++)
3740 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3741 #endif
3742 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3743 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
3744
3745 /*
3746 * Allow the "check" option to be passed as a remount option.
3747 */
3748 if (!parse_options(data, sb, NULL, &journal_ioprio,
3749 &n_blocks_count, 1)) {
3750 err = -EINVAL;
3751 goto restore_opts;
3752 }
3753
3754 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3755 ext4_abort(sb, "Abort forced by user");
3756
3757 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3758 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3759
3760 es = sbi->s_es;
3761
3762 if (sbi->s_journal) {
3763 ext4_init_journal_params(sb, sbi->s_journal);
3764 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3765 }
3766
3767 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
3768 n_blocks_count > ext4_blocks_count(es)) {
3769 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
3770 err = -EROFS;
3771 goto restore_opts;
3772 }
3773
3774 if (*flags & MS_RDONLY) {
3775 err = dquot_suspend(sb, -1);
3776 if (err < 0)
3777 goto restore_opts;
3778
3779 /*
3780 * First of all, the unconditional stuff we have to do
3781 * to disable replay of the journal when we next remount
3782 */
3783 sb->s_flags |= MS_RDONLY;
3784
3785 /*
3786 * OK, test if we are remounting a valid rw partition
3787 * readonly, and if so set the rdonly flag and then
3788 * mark the partition as valid again.
3789 */
3790 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3791 (sbi->s_mount_state & EXT4_VALID_FS))
3792 es->s_state = cpu_to_le16(sbi->s_mount_state);
3793
3794 if (sbi->s_journal)
3795 ext4_mark_recovery_complete(sb, es);
3796 } else {
3797 /* Make sure we can mount this feature set readwrite */
3798 if (!ext4_feature_set_ok(sb, 0)) {
3799 err = -EROFS;
3800 goto restore_opts;
3801 }
3802 /*
3803 * Make sure the group descriptor checksums
3804 * are sane. If they aren't, refuse to remount r/w.
3805 */
3806 for (g = 0; g < sbi->s_groups_count; g++) {
3807 struct ext4_group_desc *gdp =
3808 ext4_get_group_desc(sb, g, NULL);
3809
3810 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3811 ext4_msg(sb, KERN_ERR,
3812 "ext4_remount: Checksum for group %u failed (%u!=%u)",
3813 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3814 le16_to_cpu(gdp->bg_checksum));
3815 err = -EINVAL;
3816 goto restore_opts;
3817 }
3818 }
3819
3820 /*
3821 * If we have an unprocessed orphan list hanging
3822 * around from a previously readonly bdev mount,
3823 * require a full umount/remount for now.
3824 */
3825 if (es->s_last_orphan) {
3826 ext4_msg(sb, KERN_WARNING, "Couldn't "
3827 "remount RDWR because of unprocessed "
3828 "orphan inode list. Please "
3829 "umount/remount instead");
3830 err = -EINVAL;
3831 goto restore_opts;
3832 }
3833
3834 /*
3835 * Mounting a RDONLY partition read-write, so reread
3836 * and store the current valid flag. (It may have
3837 * been changed by e2fsck since we originally mounted
3838 * the partition.)
3839 */
3840 if (sbi->s_journal)
3841 ext4_clear_journal_err(sb, es);
3842 sbi->s_mount_state = le16_to_cpu(es->s_state);
3843 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
3844 goto restore_opts;
3845 if (!ext4_setup_super(sb, es, 0))
3846 sb->s_flags &= ~MS_RDONLY;
3847 enable_quota = 1;
3848 }
3849 }
3850 ext4_setup_system_zone(sb);
3851 if (sbi->s_journal == NULL)
3852 ext4_commit_super(sb, 1);
3853
3854 #ifdef CONFIG_QUOTA
3855 /* Release old quota file names */
3856 for (i = 0; i < MAXQUOTAS; i++)
3857 if (old_opts.s_qf_names[i] &&
3858 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3859 kfree(old_opts.s_qf_names[i]);
3860 #endif
3861 unlock_super(sb);
3862 unlock_kernel();
3863 if (enable_quota)
3864 dquot_resume(sb, -1);
3865
3866 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
3867 kfree(orig_data);
3868 return 0;
3869
3870 restore_opts:
3871 sb->s_flags = old_sb_flags;
3872 sbi->s_mount_opt = old_opts.s_mount_opt;
3873 sbi->s_resuid = old_opts.s_resuid;
3874 sbi->s_resgid = old_opts.s_resgid;
3875 sbi->s_commit_interval = old_opts.s_commit_interval;
3876 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3877 sbi->s_max_batch_time = old_opts.s_max_batch_time;
3878 #ifdef CONFIG_QUOTA
3879 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3880 for (i = 0; i < MAXQUOTAS; i++) {
3881 if (sbi->s_qf_names[i] &&
3882 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3883 kfree(sbi->s_qf_names[i]);
3884 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3885 }
3886 #endif
3887 unlock_super(sb);
3888 unlock_kernel();
3889 kfree(orig_data);
3890 return err;
3891 }
3892
3893 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
3894 {
3895 struct super_block *sb = dentry->d_sb;
3896 struct ext4_sb_info *sbi = EXT4_SB(sb);
3897 struct ext4_super_block *es = sbi->s_es;
3898 u64 fsid;
3899
3900 if (test_opt(sb, MINIX_DF)) {
3901 sbi->s_overhead_last = 0;
3902 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
3903 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3904 ext4_fsblk_t overhead = 0;
3905
3906 /*
3907 * Compute the overhead (FS structures). This is constant
3908 * for a given filesystem unless the number of block groups
3909 * changes so we cache the previous value until it does.
3910 */
3911
3912 /*
3913 * All of the blocks before first_data_block are
3914 * overhead
3915 */
3916 overhead = le32_to_cpu(es->s_first_data_block);
3917
3918 /*
3919 * Add the overhead attributed to the superblock and
3920 * block group descriptors. If the sparse superblocks
3921 * feature is turned on, then not all groups have this.
3922 */
3923 for (i = 0; i < ngroups; i++) {
3924 overhead += ext4_bg_has_super(sb, i) +
3925 ext4_bg_num_gdb(sb, i);
3926 cond_resched();
3927 }
3928
3929 /*
3930 * Every block group has an inode bitmap, a block
3931 * bitmap, and an inode table.
3932 */
3933 overhead += ngroups * (2 + sbi->s_itb_per_group);
3934 sbi->s_overhead_last = overhead;
3935 smp_wmb();
3936 sbi->s_blocks_last = ext4_blocks_count(es);
3937 }
3938
3939 buf->f_type = EXT4_SUPER_MAGIC;
3940 buf->f_bsize = sb->s_blocksize;
3941 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
3942 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3943 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
3944 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3945 if (buf->f_bfree < ext4_r_blocks_count(es))
3946 buf->f_bavail = 0;
3947 buf->f_files = le32_to_cpu(es->s_inodes_count);
3948 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
3949 buf->f_namelen = EXT4_NAME_LEN;
3950 fsid = le64_to_cpup((void *)es->s_uuid) ^
3951 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3952 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3953 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
3954
3955 return 0;
3956 }
3957
3958 /* Helper function for writing quotas on sync - we need to start transaction
3959 * before quota file is locked for write. Otherwise the are possible deadlocks:
3960 * Process 1 Process 2
3961 * ext4_create() quota_sync()
3962 * jbd2_journal_start() write_dquot()
3963 * dquot_initialize() down(dqio_mutex)
3964 * down(dqio_mutex) jbd2_journal_start()
3965 *
3966 */
3967
3968 #ifdef CONFIG_QUOTA
3969
3970 static inline struct inode *dquot_to_inode(struct dquot *dquot)
3971 {
3972 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3973 }
3974
3975 static int ext4_write_dquot(struct dquot *dquot)
3976 {
3977 int ret, err;
3978 handle_t *handle;
3979 struct inode *inode;
3980
3981 inode = dquot_to_inode(dquot);
3982 handle = ext4_journal_start(inode,
3983 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
3984 if (IS_ERR(handle))
3985 return PTR_ERR(handle);
3986 ret = dquot_commit(dquot);
3987 err = ext4_journal_stop(handle);
3988 if (!ret)
3989 ret = err;
3990 return ret;
3991 }
3992
3993 static int ext4_acquire_dquot(struct dquot *dquot)
3994 {
3995 int ret, err;
3996 handle_t *handle;
3997
3998 handle = ext4_journal_start(dquot_to_inode(dquot),
3999 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4000 if (IS_ERR(handle))
4001 return PTR_ERR(handle);
4002 ret = dquot_acquire(dquot);
4003 err = ext4_journal_stop(handle);
4004 if (!ret)
4005 ret = err;
4006 return ret;
4007 }
4008
4009 static int ext4_release_dquot(struct dquot *dquot)
4010 {
4011 int ret, err;
4012 handle_t *handle;
4013
4014 handle = ext4_journal_start(dquot_to_inode(dquot),
4015 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4016 if (IS_ERR(handle)) {
4017 /* Release dquot anyway to avoid endless cycle in dqput() */
4018 dquot_release(dquot);
4019 return PTR_ERR(handle);
4020 }
4021 ret = dquot_release(dquot);
4022 err = ext4_journal_stop(handle);
4023 if (!ret)
4024 ret = err;
4025 return ret;
4026 }
4027
4028 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4029 {
4030 /* Are we journaling quotas? */
4031 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4032 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4033 dquot_mark_dquot_dirty(dquot);
4034 return ext4_write_dquot(dquot);
4035 } else {
4036 return dquot_mark_dquot_dirty(dquot);
4037 }
4038 }
4039
4040 static int ext4_write_info(struct super_block *sb, int type)
4041 {
4042 int ret, err;
4043 handle_t *handle;
4044
4045 /* Data block + inode block */
4046 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4047 if (IS_ERR(handle))
4048 return PTR_ERR(handle);
4049 ret = dquot_commit_info(sb, type);
4050 err = ext4_journal_stop(handle);
4051 if (!ret)
4052 ret = err;
4053 return ret;
4054 }
4055
4056 /*
4057 * Turn on quotas during mount time - we need to find
4058 * the quota file and such...
4059 */
4060 static int ext4_quota_on_mount(struct super_block *sb, int type)
4061 {
4062 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4063 EXT4_SB(sb)->s_jquota_fmt, type);
4064 }
4065
4066 /*
4067 * Standard function to be called on quota_on
4068 */
4069 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4070 char *name)
4071 {
4072 int err;
4073 struct path path;
4074
4075 if (!test_opt(sb, QUOTA))
4076 return -EINVAL;
4077
4078 err = kern_path(name, LOOKUP_FOLLOW, &path);
4079 if (err)
4080 return err;
4081
4082 /* Quotafile not on the same filesystem? */
4083 if (path.mnt->mnt_sb != sb) {
4084 path_put(&path);
4085 return -EXDEV;
4086 }
4087 /* Journaling quota? */
4088 if (EXT4_SB(sb)->s_qf_names[type]) {
4089 /* Quotafile not in fs root? */
4090 if (path.dentry->d_parent != sb->s_root)
4091 ext4_msg(sb, KERN_WARNING,
4092 "Quota file not on filesystem root. "
4093 "Journaled quota will not work");
4094 }
4095
4096 /*
4097 * When we journal data on quota file, we have to flush journal to see
4098 * all updates to the file when we bypass pagecache...
4099 */
4100 if (EXT4_SB(sb)->s_journal &&
4101 ext4_should_journal_data(path.dentry->d_inode)) {
4102 /*
4103 * We don't need to lock updates but journal_flush() could
4104 * otherwise be livelocked...
4105 */
4106 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4107 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4108 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4109 if (err) {
4110 path_put(&path);
4111 return err;
4112 }
4113 }
4114
4115 err = dquot_quota_on_path(sb, type, format_id, &path);
4116 path_put(&path);
4117 return err;
4118 }
4119
4120 static int ext4_quota_off(struct super_block *sb, int type)
4121 {
4122 /* Force all delayed allocation blocks to be allocated */
4123 if (test_opt(sb, DELALLOC)) {
4124 down_read(&sb->s_umount);
4125 sync_filesystem(sb);
4126 up_read(&sb->s_umount);
4127 }
4128
4129 return dquot_quota_off(sb, type);
4130 }
4131
4132 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4133 * acquiring the locks... As quota files are never truncated and quota code
4134 * itself serializes the operations (and noone else should touch the files)
4135 * we don't have to be afraid of races */
4136 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4137 size_t len, loff_t off)
4138 {
4139 struct inode *inode = sb_dqopt(sb)->files[type];
4140 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4141 int err = 0;
4142 int offset = off & (sb->s_blocksize - 1);
4143 int tocopy;
4144 size_t toread;
4145 struct buffer_head *bh;
4146 loff_t i_size = i_size_read(inode);
4147
4148 if (off > i_size)
4149 return 0;
4150 if (off+len > i_size)
4151 len = i_size-off;
4152 toread = len;
4153 while (toread > 0) {
4154 tocopy = sb->s_blocksize - offset < toread ?
4155 sb->s_blocksize - offset : toread;
4156 bh = ext4_bread(NULL, inode, blk, 0, &err);
4157 if (err)
4158 return err;
4159 if (!bh) /* A hole? */
4160 memset(data, 0, tocopy);
4161 else
4162 memcpy(data, bh->b_data+offset, tocopy);
4163 brelse(bh);
4164 offset = 0;
4165 toread -= tocopy;
4166 data += tocopy;
4167 blk++;
4168 }
4169 return len;
4170 }
4171
4172 /* Write to quotafile (we know the transaction is already started and has
4173 * enough credits) */
4174 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4175 const char *data, size_t len, loff_t off)
4176 {
4177 struct inode *inode = sb_dqopt(sb)->files[type];
4178 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4179 int err = 0;
4180 int offset = off & (sb->s_blocksize - 1);
4181 struct buffer_head *bh;
4182 handle_t *handle = journal_current_handle();
4183
4184 if (EXT4_SB(sb)->s_journal && !handle) {
4185 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4186 " cancelled because transaction is not started",
4187 (unsigned long long)off, (unsigned long long)len);
4188 return -EIO;
4189 }
4190 /*
4191 * Since we account only one data block in transaction credits,
4192 * then it is impossible to cross a block boundary.
4193 */
4194 if (sb->s_blocksize - offset < len) {
4195 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4196 " cancelled because not block aligned",
4197 (unsigned long long)off, (unsigned long long)len);
4198 return -EIO;
4199 }
4200
4201 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4202 bh = ext4_bread(handle, inode, blk, 1, &err);
4203 if (!bh)
4204 goto out;
4205 err = ext4_journal_get_write_access(handle, bh);
4206 if (err) {
4207 brelse(bh);
4208 goto out;
4209 }
4210 lock_buffer(bh);
4211 memcpy(bh->b_data+offset, data, len);
4212 flush_dcache_page(bh->b_page);
4213 unlock_buffer(bh);
4214 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4215 brelse(bh);
4216 out:
4217 if (err) {
4218 mutex_unlock(&inode->i_mutex);
4219 return err;
4220 }
4221 if (inode->i_size < off + len) {
4222 i_size_write(inode, off + len);
4223 EXT4_I(inode)->i_disksize = inode->i_size;
4224 }
4225 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4226 ext4_mark_inode_dirty(handle, inode);
4227 mutex_unlock(&inode->i_mutex);
4228 return len;
4229 }
4230
4231 #endif
4232
4233 static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4234 const char *dev_name, void *data, struct vfsmount *mnt)
4235 {
4236 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
4237 }
4238
4239 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && \
4240 defined(CONFIG_EXT4_USE_FOR_EXT23)
4241 static struct file_system_type ext2_fs_type = {
4242 .owner = THIS_MODULE,
4243 .name = "ext2",
4244 .get_sb = ext4_get_sb,
4245 .kill_sb = kill_block_super,
4246 .fs_flags = FS_REQUIRES_DEV,
4247 };
4248
4249 static inline void register_as_ext2(void)
4250 {
4251 int err = register_filesystem(&ext2_fs_type);
4252 if (err)
4253 printk(KERN_WARNING
4254 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4255 }
4256
4257 static inline void unregister_as_ext2(void)
4258 {
4259 unregister_filesystem(&ext2_fs_type);
4260 }
4261 MODULE_ALIAS("ext2");
4262 #else
4263 static inline void register_as_ext2(void) { }
4264 static inline void unregister_as_ext2(void) { }
4265 #endif
4266
4267 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && \
4268 defined(CONFIG_EXT4_USE_FOR_EXT23)
4269 static inline void register_as_ext3(void)
4270 {
4271 int err = register_filesystem(&ext3_fs_type);
4272 if (err)
4273 printk(KERN_WARNING
4274 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4275 }
4276
4277 static inline void unregister_as_ext3(void)
4278 {
4279 unregister_filesystem(&ext3_fs_type);
4280 }
4281 MODULE_ALIAS("ext3");
4282 #else
4283 static inline void register_as_ext3(void) { }
4284 static inline void unregister_as_ext3(void) { }
4285 #endif
4286
4287 static struct file_system_type ext4_fs_type = {
4288 .owner = THIS_MODULE,
4289 .name = "ext4",
4290 .get_sb = ext4_get_sb,
4291 .kill_sb = kill_block_super,
4292 .fs_flags = FS_REQUIRES_DEV,
4293 };
4294
4295 static int __init init_ext4_fs(void)
4296 {
4297 int err;
4298
4299 ext4_check_flag_values();
4300 err = init_ext4_system_zone();
4301 if (err)
4302 return err;
4303 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4304 if (!ext4_kset)
4305 goto out4;
4306 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4307 err = init_ext4_mballoc();
4308 if (err)
4309 goto out3;
4310
4311 err = init_ext4_xattr();
4312 if (err)
4313 goto out2;
4314 err = init_inodecache();
4315 if (err)
4316 goto out1;
4317 register_as_ext2();
4318 register_as_ext3();
4319 err = register_filesystem(&ext4_fs_type);
4320 if (err)
4321 goto out;
4322 return 0;
4323 out:
4324 unregister_as_ext2();
4325 unregister_as_ext3();
4326 destroy_inodecache();
4327 out1:
4328 exit_ext4_xattr();
4329 out2:
4330 exit_ext4_mballoc();
4331 out3:
4332 remove_proc_entry("fs/ext4", NULL);
4333 kset_unregister(ext4_kset);
4334 out4:
4335 exit_ext4_system_zone();
4336 return err;
4337 }
4338
4339 static void __exit exit_ext4_fs(void)
4340 {
4341 unregister_as_ext2();
4342 unregister_as_ext3();
4343 unregister_filesystem(&ext4_fs_type);
4344 destroy_inodecache();
4345 exit_ext4_xattr();
4346 exit_ext4_mballoc();
4347 remove_proc_entry("fs/ext4", NULL);
4348 kset_unregister(ext4_kset);
4349 exit_ext4_system_zone();
4350 }
4351
4352 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4353 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4354 MODULE_LICENSE("GPL");
4355 module_init(init_ext4_fs)
4356 module_exit(exit_ext4_fs)
Tomato firmware and modifications.