?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge remote-tracking branch 'moduleh/module.h-split'
[linux-2.6/next.git] / fs / ext4 / super.c
blob8c9bb73f952d6afcb789882742a5161f9b6bf7ce
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
43
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
46
47 #include "ext4.h"
48 #include "ext4_jbd2.h"
49 #include "xattr.h"
50 #include "acl.h"
51 #include "mballoc.h"
52
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
55
56 static struct proc_dir_entry *ext4_proc_root;
57 static struct kset *ext4_kset;
58 static struct ext4_lazy_init *ext4_li_info;
59 static struct mutex ext4_li_mtx;
60 static struct ext4_features *ext4_feat;
61
62 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
63 unsigned long journal_devnum);
64 static int ext4_commit_super(struct super_block *sb, int sync);
65 static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67 static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
69 static int ext4_sync_fs(struct super_block *sb, int wait);
70 static const char *ext4_decode_error(struct super_block *sb, int errno,
71 char nbuf[16]);
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static void ext4_write_super(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type = {
88 .owner = THIS_MODULE,
89 .name = "ext2",
90 .mount = ext4_mount,
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
93 };
94 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
95 #else
96 #define IS_EXT2_SB(sb) (0)
97 #endif
98
99
100 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
101 static struct file_system_type ext3_fs_type = {
102 .owner = THIS_MODULE,
103 .name = "ext3",
104 .mount = ext4_mount,
105 .kill_sb = kill_block_super,
106 .fs_flags = FS_REQUIRES_DEV,
107 };
108 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
109 #else
110 #define IS_EXT3_SB(sb) (0)
111 #endif
112
113 void *ext4_kvmalloc(size_t size, gfp_t flags)
114 {
115 void *ret;
116
117 ret = kmalloc(size, flags);
118 if (!ret)
119 ret = __vmalloc(size, flags, PAGE_KERNEL);
120 return ret;
121 }
122
123 void *ext4_kvzalloc(size_t size, gfp_t flags)
124 {
125 void *ret;
126
127 ret = kzalloc(size, flags);
128 if (!ret)
129 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
130 return ret;
131 }
132
133 void ext4_kvfree(void *ptr)
134 {
135 if (is_vmalloc_addr(ptr))
136 vfree(ptr);
137 else
138 kfree(ptr);
139
140 }
141
142 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
143 struct ext4_group_desc *bg)
144 {
145 return le32_to_cpu(bg->bg_block_bitmap_lo) |
146 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
147 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
148 }
149
150 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
151 struct ext4_group_desc *bg)
152 {
153 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
154 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
155 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
156 }
157
158 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
159 struct ext4_group_desc *bg)
160 {
161 return le32_to_cpu(bg->bg_inode_table_lo) |
162 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
163 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
164 }
165
166 __u32 ext4_free_blks_count(struct super_block *sb,
167 struct ext4_group_desc *bg)
168 {
169 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
170 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
171 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
172 }
173
174 __u32 ext4_free_inodes_count(struct super_block *sb,
175 struct ext4_group_desc *bg)
176 {
177 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
178 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
179 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
180 }
181
182 __u32 ext4_used_dirs_count(struct super_block *sb,
183 struct ext4_group_desc *bg)
184 {
185 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
186 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
187 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
188 }
189
190 __u32 ext4_itable_unused_count(struct super_block *sb,
191 struct ext4_group_desc *bg)
192 {
193 return le16_to_cpu(bg->bg_itable_unused_lo) |
194 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
195 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
196 }
197
198 void ext4_block_bitmap_set(struct super_block *sb,
199 struct ext4_group_desc *bg, ext4_fsblk_t blk)
200 {
201 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
202 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
203 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
204 }
205
206 void ext4_inode_bitmap_set(struct super_block *sb,
207 struct ext4_group_desc *bg, ext4_fsblk_t blk)
208 {
209 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
210 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
211 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
212 }
213
214 void ext4_inode_table_set(struct super_block *sb,
215 struct ext4_group_desc *bg, ext4_fsblk_t blk)
216 {
217 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
218 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
219 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
220 }
221
222 void ext4_free_blks_set(struct super_block *sb,
223 struct ext4_group_desc *bg, __u32 count)
224 {
225 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
226 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
227 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
228 }
229
230 void ext4_free_inodes_set(struct super_block *sb,
231 struct ext4_group_desc *bg, __u32 count)
232 {
233 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
234 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
235 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
236 }
237
238 void ext4_used_dirs_set(struct super_block *sb,
239 struct ext4_group_desc *bg, __u32 count)
240 {
241 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
242 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
243 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
244 }
245
246 void ext4_itable_unused_set(struct super_block *sb,
247 struct ext4_group_desc *bg, __u32 count)
248 {
249 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
250 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
251 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
252 }
253
254
255 /* Just increment the non-pointer handle value */
256 static handle_t *ext4_get_nojournal(void)
257 {
258 handle_t *handle = current->journal_info;
259 unsigned long ref_cnt = (unsigned long)handle;
260
261 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
262
263 ref_cnt++;
264 handle = (handle_t *)ref_cnt;
265
266 current->journal_info = handle;
267 return handle;
268 }
269
270
271 /* Decrement the non-pointer handle value */
272 static void ext4_put_nojournal(handle_t *handle)
273 {
274 unsigned long ref_cnt = (unsigned long)handle;
275
276 BUG_ON(ref_cnt == 0);
277
278 ref_cnt--;
279 handle = (handle_t *)ref_cnt;
280
281 current->journal_info = handle;
282 }
283
284 /*
285 * Wrappers for jbd2_journal_start/end.
286 *
287 * The only special thing we need to do here is to make sure that all
288 * journal_end calls result in the superblock being marked dirty, so
289 * that sync() will call the filesystem's write_super callback if
290 * appropriate.
291 *
292 * To avoid j_barrier hold in userspace when a user calls freeze(),
293 * ext4 prevents a new handle from being started by s_frozen, which
294 * is in an upper layer.
295 */
296 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
297 {
298 journal_t *journal;
299 handle_t *handle;
300
301 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
302 if (sb->s_flags & MS_RDONLY)
303 return ERR_PTR(-EROFS);
304
305 journal = EXT4_SB(sb)->s_journal;
306 handle = ext4_journal_current_handle();
307
308 /*
309 * If a handle has been started, it should be allowed to
310 * finish, otherwise deadlock could happen between freeze
311 * and others(e.g. truncate) due to the restart of the
312 * journal handle if the filesystem is forzen and active
313 * handles are not stopped.
314 */
315 if (!handle)
316 vfs_check_frozen(sb, SB_FREEZE_TRANS);
317
318 if (!journal)
319 return ext4_get_nojournal();
320 /*
321 * Special case here: if the journal has aborted behind our
322 * backs (eg. EIO in the commit thread), then we still need to
323 * take the FS itself readonly cleanly.
324 */
325 if (is_journal_aborted(journal)) {
326 ext4_abort(sb, "Detected aborted journal");
327 return ERR_PTR(-EROFS);
328 }
329 return jbd2_journal_start(journal, nblocks);
330 }
331
332 /*
333 * The only special thing we need to do here is to make sure that all
334 * jbd2_journal_stop calls result in the superblock being marked dirty, so
335 * that sync() will call the filesystem's write_super callback if
336 * appropriate.
337 */
338 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
339 {
340 struct super_block *sb;
341 int err;
342 int rc;
343
344 if (!ext4_handle_valid(handle)) {
345 ext4_put_nojournal(handle);
346 return 0;
347 }
348 sb = handle->h_transaction->t_journal->j_private;
349 err = handle->h_err;
350 rc = jbd2_journal_stop(handle);
351
352 if (!err)
353 err = rc;
354 if (err)
355 __ext4_std_error(sb, where, line, err);
356 return err;
357 }
358
359 void ext4_journal_abort_handle(const char *caller, unsigned int line,
360 const char *err_fn, struct buffer_head *bh,
361 handle_t *handle, int err)
362 {
363 char nbuf[16];
364 const char *errstr = ext4_decode_error(NULL, err, nbuf);
365
366 BUG_ON(!ext4_handle_valid(handle));
367
368 if (bh)
369 BUFFER_TRACE(bh, "abort");
370
371 if (!handle->h_err)
372 handle->h_err = err;
373
374 if (is_handle_aborted(handle))
375 return;
376
377 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
378 caller, line, errstr, err_fn);
379
380 jbd2_journal_abort_handle(handle);
381 }
382
383 static void __save_error_info(struct super_block *sb, const char *func,
384 unsigned int line)
385 {
386 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
387
388 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
389 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
390 es->s_last_error_time = cpu_to_le32(get_seconds());
391 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
392 es->s_last_error_line = cpu_to_le32(line);
393 if (!es->s_first_error_time) {
394 es->s_first_error_time = es->s_last_error_time;
395 strncpy(es->s_first_error_func, func,
396 sizeof(es->s_first_error_func));
397 es->s_first_error_line = cpu_to_le32(line);
398 es->s_first_error_ino = es->s_last_error_ino;
399 es->s_first_error_block = es->s_last_error_block;
400 }
401 /*
402 * Start the daily error reporting function if it hasn't been
403 * started already
404 */
405 if (!es->s_error_count)
406 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
407 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
408 }
409
410 static void save_error_info(struct super_block *sb, const char *func,
411 unsigned int line)
412 {
413 __save_error_info(sb, func, line);
414 ext4_commit_super(sb, 1);
415 }
416
417
418 /* Deal with the reporting of failure conditions on a filesystem such as
419 * inconsistencies detected or read IO failures.
420 *
421 * On ext2, we can store the error state of the filesystem in the
422 * superblock. That is not possible on ext4, because we may have other
423 * write ordering constraints on the superblock which prevent us from
424 * writing it out straight away; and given that the journal is about to
425 * be aborted, we can't rely on the current, or future, transactions to
426 * write out the superblock safely.
427 *
428 * We'll just use the jbd2_journal_abort() error code to record an error in
429 * the journal instead. On recovery, the journal will complain about
430 * that error until we've noted it down and cleared it.
431 */
432
433 static void ext4_handle_error(struct super_block *sb)
434 {
435 if (sb->s_flags & MS_RDONLY)
436 return;
437
438 if (!test_opt(sb, ERRORS_CONT)) {
439 journal_t *journal = EXT4_SB(sb)->s_journal;
440
441 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
442 if (journal)
443 jbd2_journal_abort(journal, -EIO);
444 }
445 if (test_opt(sb, ERRORS_RO)) {
446 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
447 sb->s_flags |= MS_RDONLY;
448 }
449 if (test_opt(sb, ERRORS_PANIC))
450 panic("EXT4-fs (device %s): panic forced after error\n",
451 sb->s_id);
452 }
453
454 void __ext4_error(struct super_block *sb, const char *function,
455 unsigned int line, const char *fmt, ...)
456 {
457 struct va_format vaf;
458 va_list args;
459
460 va_start(args, fmt);
461 vaf.fmt = fmt;
462 vaf.va = &args;
463 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
464 sb->s_id, function, line, current->comm, &vaf);
465 va_end(args);
466
467 ext4_handle_error(sb);
468 }
469
470 void ext4_error_inode(struct inode *inode, const char *function,
471 unsigned int line, ext4_fsblk_t block,
472 const char *fmt, ...)
473 {
474 va_list args;
475 struct va_format vaf;
476 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
477
478 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
479 es->s_last_error_block = cpu_to_le64(block);
480 save_error_info(inode->i_sb, function, line);
481 va_start(args, fmt);
482 vaf.fmt = fmt;
483 vaf.va = &args;
484 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
485 inode->i_sb->s_id, function, line, inode->i_ino);
486 if (block)
487 printk(KERN_CONT "block %llu: ", block);
488 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
489 va_end(args);
490
491 ext4_handle_error(inode->i_sb);
492 }
493
494 void ext4_error_file(struct file *file, const char *function,
495 unsigned int line, ext4_fsblk_t block,
496 const char *fmt, ...)
497 {
498 va_list args;
499 struct va_format vaf;
500 struct ext4_super_block *es;
501 struct inode *inode = file->f_dentry->d_inode;
502 char pathname[80], *path;
503
504 es = EXT4_SB(inode->i_sb)->s_es;
505 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
506 save_error_info(inode->i_sb, function, line);
507 path = d_path(&(file->f_path), pathname, sizeof(pathname));
508 if (IS_ERR(path))
509 path = "(unknown)";
510 printk(KERN_CRIT
511 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
512 inode->i_sb->s_id, function, line, inode->i_ino);
513 if (block)
514 printk(KERN_CONT "block %llu: ", block);
515 va_start(args, fmt);
516 vaf.fmt = fmt;
517 vaf.va = &args;
518 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
519 va_end(args);
520
521 ext4_handle_error(inode->i_sb);
522 }
523
524 static const char *ext4_decode_error(struct super_block *sb, int errno,
525 char nbuf[16])
526 {
527 char *errstr = NULL;
528
529 switch (errno) {
530 case -EIO:
531 errstr = "IO failure";
532 break;
533 case -ENOMEM:
534 errstr = "Out of memory";
535 break;
536 case -EROFS:
537 if (!sb || (EXT4_SB(sb)->s_journal &&
538 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
539 errstr = "Journal has aborted";
540 else
541 errstr = "Readonly filesystem";
542 break;
543 default:
544 /* If the caller passed in an extra buffer for unknown
545 * errors, textualise them now. Else we just return
546 * NULL. */
547 if (nbuf) {
548 /* Check for truncated error codes... */
549 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
550 errstr = nbuf;
551 }
552 break;
553 }
554
555 return errstr;
556 }
557
558 /* __ext4_std_error decodes expected errors from journaling functions
559 * automatically and invokes the appropriate error response. */
560
561 void __ext4_std_error(struct super_block *sb, const char *function,
562 unsigned int line, int errno)
563 {
564 char nbuf[16];
565 const char *errstr;
566
567 /* Special case: if the error is EROFS, and we're not already
568 * inside a transaction, then there's really no point in logging
569 * an error. */
570 if (errno == -EROFS && journal_current_handle() == NULL &&
571 (sb->s_flags & MS_RDONLY))
572 return;
573
574 errstr = ext4_decode_error(sb, errno, nbuf);
575 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
576 sb->s_id, function, line, errstr);
577 save_error_info(sb, function, line);
578
579 ext4_handle_error(sb);
580 }
581
582 /*
583 * ext4_abort is a much stronger failure handler than ext4_error. The
584 * abort function may be used to deal with unrecoverable failures such
585 * as journal IO errors or ENOMEM at a critical moment in log management.
586 *
587 * We unconditionally force the filesystem into an ABORT|READONLY state,
588 * unless the error response on the fs has been set to panic in which
589 * case we take the easy way out and panic immediately.
590 */
591
592 void __ext4_abort(struct super_block *sb, const char *function,
593 unsigned int line, const char *fmt, ...)
594 {
595 va_list args;
596
597 save_error_info(sb, function, line);
598 va_start(args, fmt);
599 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
600 function, line);
601 vprintk(fmt, args);
602 printk("\n");
603 va_end(args);
604
605 if ((sb->s_flags & MS_RDONLY) == 0) {
606 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
607 sb->s_flags |= MS_RDONLY;
608 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
609 if (EXT4_SB(sb)->s_journal)
610 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
611 save_error_info(sb, function, line);
612 }
613 if (test_opt(sb, ERRORS_PANIC))
614 panic("EXT4-fs panic from previous error\n");
615 }
616
617 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
618 {
619 struct va_format vaf;
620 va_list args;
621
622 va_start(args, fmt);
623 vaf.fmt = fmt;
624 vaf.va = &args;
625 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
626 va_end(args);
627 }
628
629 void __ext4_warning(struct super_block *sb, const char *function,
630 unsigned int line, const char *fmt, ...)
631 {
632 struct va_format vaf;
633 va_list args;
634
635 va_start(args, fmt);
636 vaf.fmt = fmt;
637 vaf.va = &args;
638 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
639 sb->s_id, function, line, &vaf);
640 va_end(args);
641 }
642
643 void __ext4_grp_locked_error(const char *function, unsigned int line,
644 struct super_block *sb, ext4_group_t grp,
645 unsigned long ino, ext4_fsblk_t block,
646 const char *fmt, ...)
647 __releases(bitlock)
648 __acquires(bitlock)
649 {
650 struct va_format vaf;
651 va_list args;
652 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
653
654 es->s_last_error_ino = cpu_to_le32(ino);
655 es->s_last_error_block = cpu_to_le64(block);
656 __save_error_info(sb, function, line);
657
658 va_start(args, fmt);
659
660 vaf.fmt = fmt;
661 vaf.va = &args;
662 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
663 sb->s_id, function, line, grp);
664 if (ino)
665 printk(KERN_CONT "inode %lu: ", ino);
666 if (block)
667 printk(KERN_CONT "block %llu:", (unsigned long long) block);
668 printk(KERN_CONT "%pV\n", &vaf);
669 va_end(args);
670
671 if (test_opt(sb, ERRORS_CONT)) {
672 ext4_commit_super(sb, 0);
673 return;
674 }
675
676 ext4_unlock_group(sb, grp);
677 ext4_handle_error(sb);
678 /*
679 * We only get here in the ERRORS_RO case; relocking the group
680 * may be dangerous, but nothing bad will happen since the
681 * filesystem will have already been marked read/only and the
682 * journal has been aborted. We return 1 as a hint to callers
683 * who might what to use the return value from
684 * ext4_grp_locked_error() to distinguish between the
685 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
686 * aggressively from the ext4 function in question, with a
687 * more appropriate error code.
688 */
689 ext4_lock_group(sb, grp);
690 return;
691 }
692
693 void ext4_update_dynamic_rev(struct super_block *sb)
694 {
695 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
696
697 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
698 return;
699
700 ext4_warning(sb,
701 "updating to rev %d because of new feature flag, "
702 "running e2fsck is recommended",
703 EXT4_DYNAMIC_REV);
704
705 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
706 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
707 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
708 /* leave es->s_feature_*compat flags alone */
709 /* es->s_uuid will be set by e2fsck if empty */
710
711 /*
712 * The rest of the superblock fields should be zero, and if not it
713 * means they are likely already in use, so leave them alone. We
714 * can leave it up to e2fsck to clean up any inconsistencies there.
715 */
716 }
717
718 /*
719 * Open the external journal device
720 */
721 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
722 {
723 struct block_device *bdev;
724 char b[BDEVNAME_SIZE];
725
726 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
727 if (IS_ERR(bdev))
728 goto fail;
729 return bdev;
730
731 fail:
732 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
733 __bdevname(dev, b), PTR_ERR(bdev));
734 return NULL;
735 }
736
737 /*
738 * Release the journal device
739 */
740 static int ext4_blkdev_put(struct block_device *bdev)
741 {
742 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
743 }
744
745 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
746 {
747 struct block_device *bdev;
748 int ret = -ENODEV;
749
750 bdev = sbi->journal_bdev;
751 if (bdev) {
752 ret = ext4_blkdev_put(bdev);
753 sbi->journal_bdev = NULL;
754 }
755 return ret;
756 }
757
758 static inline struct inode *orphan_list_entry(struct list_head *l)
759 {
760 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
761 }
762
763 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
764 {
765 struct list_head *l;
766
767 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
768 le32_to_cpu(sbi->s_es->s_last_orphan));
769
770 printk(KERN_ERR "sb_info orphan list:\n");
771 list_for_each(l, &sbi->s_orphan) {
772 struct inode *inode = orphan_list_entry(l);
773 printk(KERN_ERR " "
774 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
775 inode->i_sb->s_id, inode->i_ino, inode,
776 inode->i_mode, inode->i_nlink,
777 NEXT_ORPHAN(inode));
778 }
779 }
780
781 static void ext4_put_super(struct super_block *sb)
782 {
783 struct ext4_sb_info *sbi = EXT4_SB(sb);
784 struct ext4_super_block *es = sbi->s_es;
785 int i, err;
786
787 ext4_unregister_li_request(sb);
788 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
789
790 flush_workqueue(sbi->dio_unwritten_wq);
791 destroy_workqueue(sbi->dio_unwritten_wq);
792
793 lock_super(sb);
794 if (sb->s_dirt)
795 ext4_commit_super(sb, 1);
796
797 if (sbi->s_journal) {
798 err = jbd2_journal_destroy(sbi->s_journal);
799 sbi->s_journal = NULL;
800 if (err < 0)
801 ext4_abort(sb, "Couldn't clean up the journal");
802 }
803
804 del_timer(&sbi->s_err_report);
805 ext4_release_system_zone(sb);
806 ext4_mb_release(sb);
807 ext4_ext_release(sb);
808 ext4_xattr_put_super(sb);
809
810 if (!(sb->s_flags & MS_RDONLY)) {
811 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
812 es->s_state = cpu_to_le16(sbi->s_mount_state);
813 ext4_commit_super(sb, 1);
814 }
815 if (sbi->s_proc) {
816 remove_proc_entry(sb->s_id, ext4_proc_root);
817 }
818 kobject_del(&sbi->s_kobj);
819
820 for (i = 0; i < sbi->s_gdb_count; i++)
821 brelse(sbi->s_group_desc[i]);
822 ext4_kvfree(sbi->s_group_desc);
823 ext4_kvfree(sbi->s_flex_groups);
824 percpu_counter_destroy(&sbi->s_freeblocks_counter);
825 percpu_counter_destroy(&sbi->s_freeinodes_counter);
826 percpu_counter_destroy(&sbi->s_dirs_counter);
827 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
828 brelse(sbi->s_sbh);
829 #ifdef CONFIG_QUOTA
830 for (i = 0; i < MAXQUOTAS; i++)
831 kfree(sbi->s_qf_names[i]);
832 #endif
833
834 /* Debugging code just in case the in-memory inode orphan list
835 * isn't empty. The on-disk one can be non-empty if we've
836 * detected an error and taken the fs readonly, but the
837 * in-memory list had better be clean by this point. */
838 if (!list_empty(&sbi->s_orphan))
839 dump_orphan_list(sb, sbi);
840 J_ASSERT(list_empty(&sbi->s_orphan));
841
842 invalidate_bdev(sb->s_bdev);
843 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
844 /*
845 * Invalidate the journal device's buffers. We don't want them
846 * floating about in memory - the physical journal device may
847 * hotswapped, and it breaks the `ro-after' testing code.
848 */
849 sync_blockdev(sbi->journal_bdev);
850 invalidate_bdev(sbi->journal_bdev);
851 ext4_blkdev_remove(sbi);
852 }
853 if (sbi->s_mmp_tsk)
854 kthread_stop(sbi->s_mmp_tsk);
855 sb->s_fs_info = NULL;
856 /*
857 * Now that we are completely done shutting down the
858 * superblock, we need to actually destroy the kobject.
859 */
860 unlock_super(sb);
861 kobject_put(&sbi->s_kobj);
862 wait_for_completion(&sbi->s_kobj_unregister);
863 kfree(sbi->s_blockgroup_lock);
864 kfree(sbi);
865 }
866
867 static struct kmem_cache *ext4_inode_cachep;
868
869 /*
870 * Called inside transaction, so use GFP_NOFS
871 */
872 static struct inode *ext4_alloc_inode(struct super_block *sb)
873 {
874 struct ext4_inode_info *ei;
875
876 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
877 if (!ei)
878 return NULL;
879
880 ei->vfs_inode.i_version = 1;
881 ei->vfs_inode.i_data.writeback_index = 0;
882 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
883 INIT_LIST_HEAD(&ei->i_prealloc_list);
884 spin_lock_init(&ei->i_prealloc_lock);
885 ei->i_reserved_data_blocks = 0;
886 ei->i_reserved_meta_blocks = 0;
887 ei->i_allocated_meta_blocks = 0;
888 ei->i_da_metadata_calc_len = 0;
889 spin_lock_init(&(ei->i_block_reservation_lock));
890 #ifdef CONFIG_QUOTA
891 ei->i_reserved_quota = 0;
892 #endif
893 ei->jinode = NULL;
894 INIT_LIST_HEAD(&ei->i_completed_io_list);
895 spin_lock_init(&ei->i_completed_io_lock);
896 ei->cur_aio_dio = NULL;
897 ei->i_sync_tid = 0;
898 ei->i_datasync_tid = 0;
899 atomic_set(&ei->i_ioend_count, 0);
900 atomic_set(&ei->i_aiodio_unwritten, 0);
901
902 return &ei->vfs_inode;
903 }
904
905 static int ext4_drop_inode(struct inode *inode)
906 {
907 int drop = generic_drop_inode(inode);
908
909 trace_ext4_drop_inode(inode, drop);
910 return drop;
911 }
912
913 static void ext4_i_callback(struct rcu_head *head)
914 {
915 struct inode *inode = container_of(head, struct inode, i_rcu);
916 INIT_LIST_HEAD(&inode->i_dentry);
917 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
918 }
919
920 static void ext4_destroy_inode(struct inode *inode)
921 {
922 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
923 ext4_msg(inode->i_sb, KERN_ERR,
924 "Inode %lu (%p): orphan list check failed!",
925 inode->i_ino, EXT4_I(inode));
926 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
927 EXT4_I(inode), sizeof(struct ext4_inode_info),
928 true);
929 dump_stack();
930 }
931 call_rcu(&inode->i_rcu, ext4_i_callback);
932 }
933
934 static void init_once(void *foo)
935 {
936 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
937
938 INIT_LIST_HEAD(&ei->i_orphan);
939 #ifdef CONFIG_EXT4_FS_XATTR
940 init_rwsem(&ei->xattr_sem);
941 #endif
942 init_rwsem(&ei->i_data_sem);
943 inode_init_once(&ei->vfs_inode);
944 }
945
946 static int init_inodecache(void)
947 {
948 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
949 sizeof(struct ext4_inode_info),
950 0, (SLAB_RECLAIM_ACCOUNT|
951 SLAB_MEM_SPREAD),
952 init_once);
953 if (ext4_inode_cachep == NULL)
954 return -ENOMEM;
955 return 0;
956 }
957
958 static void destroy_inodecache(void)
959 {
960 kmem_cache_destroy(ext4_inode_cachep);
961 }
962
963 void ext4_clear_inode(struct inode *inode)
964 {
965 invalidate_inode_buffers(inode);
966 end_writeback(inode);
967 dquot_drop(inode);
968 ext4_discard_preallocations(inode);
969 if (EXT4_I(inode)->jinode) {
970 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
971 EXT4_I(inode)->jinode);
972 jbd2_free_inode(EXT4_I(inode)->jinode);
973 EXT4_I(inode)->jinode = NULL;
974 }
975 }
976
977 static inline void ext4_show_quota_options(struct seq_file *seq,
978 struct super_block *sb)
979 {
980 #if defined(CONFIG_QUOTA)
981 struct ext4_sb_info *sbi = EXT4_SB(sb);
982
983 if (sbi->s_jquota_fmt) {
984 char *fmtname = "";
985
986 switch (sbi->s_jquota_fmt) {
987 case QFMT_VFS_OLD:
988 fmtname = "vfsold";
989 break;
990 case QFMT_VFS_V0:
991 fmtname = "vfsv0";
992 break;
993 case QFMT_VFS_V1:
994 fmtname = "vfsv1";
995 break;
996 }
997 seq_printf(seq, ",jqfmt=%s", fmtname);
998 }
999
1000 if (sbi->s_qf_names[USRQUOTA])
1001 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1002
1003 if (sbi->s_qf_names[GRPQUOTA])
1004 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1005
1006 if (test_opt(sb, USRQUOTA))
1007 seq_puts(seq, ",usrquota");
1008
1009 if (test_opt(sb, GRPQUOTA))
1010 seq_puts(seq, ",grpquota");
1011 #endif
1012 }
1013
1014 /*
1015 * Show an option if
1016 * - it's set to a non-default value OR
1017 * - if the per-sb default is different from the global default
1018 */
1019 static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
1020 {
1021 int def_errors;
1022 unsigned long def_mount_opts;
1023 struct super_block *sb = vfs->mnt_sb;
1024 struct ext4_sb_info *sbi = EXT4_SB(sb);
1025 struct ext4_super_block *es = sbi->s_es;
1026
1027 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
1028 def_errors = le16_to_cpu(es->s_errors);
1029
1030 if (sbi->s_sb_block != 1)
1031 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1032 if (test_opt(sb, MINIX_DF))
1033 seq_puts(seq, ",minixdf");
1034 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
1035 seq_puts(seq, ",grpid");
1036 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1037 seq_puts(seq, ",nogrpid");
1038 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1039 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1040 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1041 }
1042 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1043 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1044 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1045 }
1046 if (test_opt(sb, ERRORS_RO)) {
1047 if (def_errors == EXT4_ERRORS_PANIC ||
1048 def_errors == EXT4_ERRORS_CONTINUE) {
1049 seq_puts(seq, ",errors=remount-ro");
1050 }
1051 }
1052 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1053 seq_puts(seq, ",errors=continue");
1054 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1055 seq_puts(seq, ",errors=panic");
1056 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
1057 seq_puts(seq, ",nouid32");
1058 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
1059 seq_puts(seq, ",debug");
1060 if (test_opt(sb, OLDALLOC))
1061 seq_puts(seq, ",oldalloc");
1062 #ifdef CONFIG_EXT4_FS_XATTR
1063 if (test_opt(sb, XATTR_USER))
1064 seq_puts(seq, ",user_xattr");
1065 if (!test_opt(sb, XATTR_USER))
1066 seq_puts(seq, ",nouser_xattr");
1067 #endif
1068 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1069 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
1070 seq_puts(seq, ",acl");
1071 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1072 seq_puts(seq, ",noacl");
1073 #endif
1074 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
1075 seq_printf(seq, ",commit=%u",
1076 (unsigned) (sbi->s_commit_interval / HZ));
1077 }
1078 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1079 seq_printf(seq, ",min_batch_time=%u",
1080 (unsigned) sbi->s_min_batch_time);
1081 }
1082 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1083 seq_printf(seq, ",max_batch_time=%u",
1084 (unsigned) sbi->s_min_batch_time);
1085 }
1086
1087 /*
1088 * We're changing the default of barrier mount option, so
1089 * let's always display its mount state so it's clear what its
1090 * status is.
1091 */
1092 seq_puts(seq, ",barrier=");
1093 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
1094 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1095 seq_puts(seq, ",journal_async_commit");
1096 else if (test_opt(sb, JOURNAL_CHECKSUM))
1097 seq_puts(seq, ",journal_checksum");
1098 if (test_opt(sb, I_VERSION))
1099 seq_puts(seq, ",i_version");
1100 if (!test_opt(sb, DELALLOC) &&
1101 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
1102 seq_puts(seq, ",nodelalloc");
1103
1104 if (!test_opt(sb, MBLK_IO_SUBMIT))
1105 seq_puts(seq, ",nomblk_io_submit");
1106 if (sbi->s_stripe)
1107 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
1108 /*
1109 * journal mode get enabled in different ways
1110 * So just print the value even if we didn't specify it
1111 */
1112 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1113 seq_puts(seq, ",data=journal");
1114 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1115 seq_puts(seq, ",data=ordered");
1116 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1117 seq_puts(seq, ",data=writeback");
1118
1119 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1120 seq_printf(seq, ",inode_readahead_blks=%u",
1121 sbi->s_inode_readahead_blks);
1122
1123 if (test_opt(sb, DATA_ERR_ABORT))
1124 seq_puts(seq, ",data_err=abort");
1125
1126 if (test_opt(sb, NO_AUTO_DA_ALLOC))
1127 seq_puts(seq, ",noauto_da_alloc");
1128
1129 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
1130 seq_puts(seq, ",discard");
1131
1132 if (test_opt(sb, NOLOAD))
1133 seq_puts(seq, ",norecovery");
1134
1135 if (test_opt(sb, DIOREAD_NOLOCK))
1136 seq_puts(seq, ",dioread_nolock");
1137
1138 if (test_opt(sb, BLOCK_VALIDITY) &&
1139 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1140 seq_puts(seq, ",block_validity");
1141
1142 if (!test_opt(sb, INIT_INODE_TABLE))
1143 seq_puts(seq, ",noinit_inode_table");
1144 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1145 seq_printf(seq, ",init_inode_table=%u",
1146 (unsigned) sbi->s_li_wait_mult);
1147
1148 ext4_show_quota_options(seq, sb);
1149
1150 return 0;
1151 }
1152
1153 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1154 u64 ino, u32 generation)
1155 {
1156 struct inode *inode;
1157
1158 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1159 return ERR_PTR(-ESTALE);
1160 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1161 return ERR_PTR(-ESTALE);
1162
1163 /* iget isn't really right if the inode is currently unallocated!!
1164 *
1165 * ext4_read_inode will return a bad_inode if the inode had been
1166 * deleted, so we should be safe.
1167 *
1168 * Currently we don't know the generation for parent directory, so
1169 * a generation of 0 means "accept any"
1170 */
1171 inode = ext4_iget(sb, ino);
1172 if (IS_ERR(inode))
1173 return ERR_CAST(inode);
1174 if (generation && inode->i_generation != generation) {
1175 iput(inode);
1176 return ERR_PTR(-ESTALE);
1177 }
1178
1179 return inode;
1180 }
1181
1182 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1183 int fh_len, int fh_type)
1184 {
1185 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1186 ext4_nfs_get_inode);
1187 }
1188
1189 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1190 int fh_len, int fh_type)
1191 {
1192 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1193 ext4_nfs_get_inode);
1194 }
1195
1196 /*
1197 * Try to release metadata pages (indirect blocks, directories) which are
1198 * mapped via the block device. Since these pages could have journal heads
1199 * which would prevent try_to_free_buffers() from freeing them, we must use
1200 * jbd2 layer's try_to_free_buffers() function to release them.
1201 */
1202 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1203 gfp_t wait)
1204 {
1205 journal_t *journal = EXT4_SB(sb)->s_journal;
1206
1207 WARN_ON(PageChecked(page));
1208 if (!page_has_buffers(page))
1209 return 0;
1210 if (journal)
1211 return jbd2_journal_try_to_free_buffers(journal, page,
1212 wait & ~__GFP_WAIT);
1213 return try_to_free_buffers(page);
1214 }
1215
1216 #ifdef CONFIG_QUOTA
1217 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1218 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1219
1220 static int ext4_write_dquot(struct dquot *dquot);
1221 static int ext4_acquire_dquot(struct dquot *dquot);
1222 static int ext4_release_dquot(struct dquot *dquot);
1223 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1224 static int ext4_write_info(struct super_block *sb, int type);
1225 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1226 struct path *path);
1227 static int ext4_quota_off(struct super_block *sb, int type);
1228 static int ext4_quota_on_mount(struct super_block *sb, int type);
1229 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1230 size_t len, loff_t off);
1231 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1232 const char *data, size_t len, loff_t off);
1233
1234 static const struct dquot_operations ext4_quota_operations = {
1235 .get_reserved_space = ext4_get_reserved_space,
1236 .write_dquot = ext4_write_dquot,
1237 .acquire_dquot = ext4_acquire_dquot,
1238 .release_dquot = ext4_release_dquot,
1239 .mark_dirty = ext4_mark_dquot_dirty,
1240 .write_info = ext4_write_info,
1241 .alloc_dquot = dquot_alloc,
1242 .destroy_dquot = dquot_destroy,
1243 };
1244
1245 static const struct quotactl_ops ext4_qctl_operations = {
1246 .quota_on = ext4_quota_on,
1247 .quota_off = ext4_quota_off,
1248 .quota_sync = dquot_quota_sync,
1249 .get_info = dquot_get_dqinfo,
1250 .set_info = dquot_set_dqinfo,
1251 .get_dqblk = dquot_get_dqblk,
1252 .set_dqblk = dquot_set_dqblk
1253 };
1254 #endif
1255
1256 static const struct super_operations ext4_sops = {
1257 .alloc_inode = ext4_alloc_inode,
1258 .destroy_inode = ext4_destroy_inode,
1259 .write_inode = ext4_write_inode,
1260 .dirty_inode = ext4_dirty_inode,
1261 .drop_inode = ext4_drop_inode,
1262 .evict_inode = ext4_evict_inode,
1263 .put_super = ext4_put_super,
1264 .sync_fs = ext4_sync_fs,
1265 .freeze_fs = ext4_freeze,
1266 .unfreeze_fs = ext4_unfreeze,
1267 .statfs = ext4_statfs,
1268 .remount_fs = ext4_remount,
1269 .show_options = ext4_show_options,
1270 #ifdef CONFIG_QUOTA
1271 .quota_read = ext4_quota_read,
1272 .quota_write = ext4_quota_write,
1273 #endif
1274 .bdev_try_to_free_page = bdev_try_to_free_page,
1275 };
1276
1277 static const struct super_operations ext4_nojournal_sops = {
1278 .alloc_inode = ext4_alloc_inode,
1279 .destroy_inode = ext4_destroy_inode,
1280 .write_inode = ext4_write_inode,
1281 .dirty_inode = ext4_dirty_inode,
1282 .drop_inode = ext4_drop_inode,
1283 .evict_inode = ext4_evict_inode,
1284 .write_super = ext4_write_super,
1285 .put_super = ext4_put_super,
1286 .statfs = ext4_statfs,
1287 .remount_fs = ext4_remount,
1288 .show_options = ext4_show_options,
1289 #ifdef CONFIG_QUOTA
1290 .quota_read = ext4_quota_read,
1291 .quota_write = ext4_quota_write,
1292 #endif
1293 .bdev_try_to_free_page = bdev_try_to_free_page,
1294 };
1295
1296 static const struct export_operations ext4_export_ops = {
1297 .fh_to_dentry = ext4_fh_to_dentry,
1298 .fh_to_parent = ext4_fh_to_parent,
1299 .get_parent = ext4_get_parent,
1300 };
1301
1302 enum {
1303 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1304 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1305 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
1306 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1307 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
1308 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1309 Opt_journal_update, Opt_journal_dev,
1310 Opt_journal_checksum, Opt_journal_async_commit,
1311 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1312 Opt_data_err_abort, Opt_data_err_ignore,
1313 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1314 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1315 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1316 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1317 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1318 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1319 Opt_inode_readahead_blks, Opt_journal_ioprio,
1320 Opt_dioread_nolock, Opt_dioread_lock,
1321 Opt_discard, Opt_nodiscard,
1322 Opt_init_inode_table, Opt_noinit_inode_table,
1323 };
1324
1325 static const match_table_t tokens = {
1326 {Opt_bsd_df, "bsddf"},
1327 {Opt_minix_df, "minixdf"},
1328 {Opt_grpid, "grpid"},
1329 {Opt_grpid, "bsdgroups"},
1330 {Opt_nogrpid, "nogrpid"},
1331 {Opt_nogrpid, "sysvgroups"},
1332 {Opt_resgid, "resgid=%u"},
1333 {Opt_resuid, "resuid=%u"},
1334 {Opt_sb, "sb=%u"},
1335 {Opt_err_cont, "errors=continue"},
1336 {Opt_err_panic, "errors=panic"},
1337 {Opt_err_ro, "errors=remount-ro"},
1338 {Opt_nouid32, "nouid32"},
1339 {Opt_debug, "debug"},
1340 {Opt_oldalloc, "oldalloc"},
1341 {Opt_orlov, "orlov"},
1342 {Opt_user_xattr, "user_xattr"},
1343 {Opt_nouser_xattr, "nouser_xattr"},
1344 {Opt_acl, "acl"},
1345 {Opt_noacl, "noacl"},
1346 {Opt_noload, "noload"},
1347 {Opt_noload, "norecovery"},
1348 {Opt_nobh, "nobh"},
1349 {Opt_bh, "bh"},
1350 {Opt_commit, "commit=%u"},
1351 {Opt_min_batch_time, "min_batch_time=%u"},
1352 {Opt_max_batch_time, "max_batch_time=%u"},
1353 {Opt_journal_update, "journal=update"},
1354 {Opt_journal_dev, "journal_dev=%u"},
1355 {Opt_journal_checksum, "journal_checksum"},
1356 {Opt_journal_async_commit, "journal_async_commit"},
1357 {Opt_abort, "abort"},
1358 {Opt_data_journal, "data=journal"},
1359 {Opt_data_ordered, "data=ordered"},
1360 {Opt_data_writeback, "data=writeback"},
1361 {Opt_data_err_abort, "data_err=abort"},
1362 {Opt_data_err_ignore, "data_err=ignore"},
1363 {Opt_offusrjquota, "usrjquota="},
1364 {Opt_usrjquota, "usrjquota=%s"},
1365 {Opt_offgrpjquota, "grpjquota="},
1366 {Opt_grpjquota, "grpjquota=%s"},
1367 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1368 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1369 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1370 {Opt_grpquota, "grpquota"},
1371 {Opt_noquota, "noquota"},
1372 {Opt_quota, "quota"},
1373 {Opt_usrquota, "usrquota"},
1374 {Opt_barrier, "barrier=%u"},
1375 {Opt_barrier, "barrier"},
1376 {Opt_nobarrier, "nobarrier"},
1377 {Opt_i_version, "i_version"},
1378 {Opt_stripe, "stripe=%u"},
1379 {Opt_resize, "resize"},
1380 {Opt_delalloc, "delalloc"},
1381 {Opt_nodelalloc, "nodelalloc"},
1382 {Opt_mblk_io_submit, "mblk_io_submit"},
1383 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1384 {Opt_block_validity, "block_validity"},
1385 {Opt_noblock_validity, "noblock_validity"},
1386 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1387 {Opt_journal_ioprio, "journal_ioprio=%u"},
1388 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1389 {Opt_auto_da_alloc, "auto_da_alloc"},
1390 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1391 {Opt_dioread_nolock, "dioread_nolock"},
1392 {Opt_dioread_lock, "dioread_lock"},
1393 {Opt_discard, "discard"},
1394 {Opt_nodiscard, "nodiscard"},
1395 {Opt_init_inode_table, "init_itable=%u"},
1396 {Opt_init_inode_table, "init_itable"},
1397 {Opt_noinit_inode_table, "noinit_itable"},
1398 {Opt_err, NULL},
1399 };
1400
1401 static ext4_fsblk_t get_sb_block(void **data)
1402 {
1403 ext4_fsblk_t sb_block;
1404 char *options = (char *) *data;
1405
1406 if (!options || strncmp(options, "sb=", 3) != 0)
1407 return 1; /* Default location */
1408
1409 options += 3;
1410 /* TODO: use simple_strtoll with >32bit ext4 */
1411 sb_block = simple_strtoul(options, &options, 0);
1412 if (*options && *options != ',') {
1413 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1414 (char *) *data);
1415 return 1;
1416 }
1417 if (*options == ',')
1418 options++;
1419 *data = (void *) options;
1420
1421 return sb_block;
1422 }
1423
1424 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1425 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1426 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1427
1428 #ifdef CONFIG_QUOTA
1429 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1430 {
1431 struct ext4_sb_info *sbi = EXT4_SB(sb);
1432 char *qname;
1433
1434 if (sb_any_quota_loaded(sb) &&
1435 !sbi->s_qf_names[qtype]) {
1436 ext4_msg(sb, KERN_ERR,
1437 "Cannot change journaled "
1438 "quota options when quota turned on");
1439 return 0;
1440 }
1441 qname = match_strdup(args);
1442 if (!qname) {
1443 ext4_msg(sb, KERN_ERR,
1444 "Not enough memory for storing quotafile name");
1445 return 0;
1446 }
1447 if (sbi->s_qf_names[qtype] &&
1448 strcmp(sbi->s_qf_names[qtype], qname)) {
1449 ext4_msg(sb, KERN_ERR,
1450 "%s quota file already specified", QTYPE2NAME(qtype));
1451 kfree(qname);
1452 return 0;
1453 }
1454 sbi->s_qf_names[qtype] = qname;
1455 if (strchr(sbi->s_qf_names[qtype], '/')) {
1456 ext4_msg(sb, KERN_ERR,
1457 "quotafile must be on filesystem root");
1458 kfree(sbi->s_qf_names[qtype]);
1459 sbi->s_qf_names[qtype] = NULL;
1460 return 0;
1461 }
1462 set_opt(sb, QUOTA);
1463 return 1;
1464 }
1465
1466 static int clear_qf_name(struct super_block *sb, int qtype)
1467 {
1468
1469 struct ext4_sb_info *sbi = EXT4_SB(sb);
1470
1471 if (sb_any_quota_loaded(sb) &&
1472 sbi->s_qf_names[qtype]) {
1473 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1474 " when quota turned on");
1475 return 0;
1476 }
1477 /*
1478 * The space will be released later when all options are confirmed
1479 * to be correct
1480 */
1481 sbi->s_qf_names[qtype] = NULL;
1482 return 1;
1483 }
1484 #endif
1485
1486 static int parse_options(char *options, struct super_block *sb,
1487 unsigned long *journal_devnum,
1488 unsigned int *journal_ioprio,
1489 ext4_fsblk_t *n_blocks_count, int is_remount)
1490 {
1491 struct ext4_sb_info *sbi = EXT4_SB(sb);
1492 char *p;
1493 substring_t args[MAX_OPT_ARGS];
1494 int data_opt = 0;
1495 int option;
1496 #ifdef CONFIG_QUOTA
1497 int qfmt;
1498 #endif
1499
1500 if (!options)
1501 return 1;
1502
1503 while ((p = strsep(&options, ",")) != NULL) {
1504 int token;
1505 if (!*p)
1506 continue;
1507
1508 /*
1509 * Initialize args struct so we know whether arg was
1510 * found; some options take optional arguments.
1511 */
1512 args[0].to = args[0].from = NULL;
1513 token = match_token(p, tokens, args);
1514 switch (token) {
1515 case Opt_bsd_df:
1516 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1517 clear_opt(sb, MINIX_DF);
1518 break;
1519 case Opt_minix_df:
1520 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1521 set_opt(sb, MINIX_DF);
1522
1523 break;
1524 case Opt_grpid:
1525 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1526 set_opt(sb, GRPID);
1527
1528 break;
1529 case Opt_nogrpid:
1530 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1531 clear_opt(sb, GRPID);
1532
1533 break;
1534 case Opt_resuid:
1535 if (match_int(&args[0], &option))
1536 return 0;
1537 sbi->s_resuid = option;
1538 break;
1539 case Opt_resgid:
1540 if (match_int(&args[0], &option))
1541 return 0;
1542 sbi->s_resgid = option;
1543 break;
1544 case Opt_sb:
1545 /* handled by get_sb_block() instead of here */
1546 /* *sb_block = match_int(&args[0]); */
1547 break;
1548 case Opt_err_panic:
1549 clear_opt(sb, ERRORS_CONT);
1550 clear_opt(sb, ERRORS_RO);
1551 set_opt(sb, ERRORS_PANIC);
1552 break;
1553 case Opt_err_ro:
1554 clear_opt(sb, ERRORS_CONT);
1555 clear_opt(sb, ERRORS_PANIC);
1556 set_opt(sb, ERRORS_RO);
1557 break;
1558 case Opt_err_cont:
1559 clear_opt(sb, ERRORS_RO);
1560 clear_opt(sb, ERRORS_PANIC);
1561 set_opt(sb, ERRORS_CONT);
1562 break;
1563 case Opt_nouid32:
1564 set_opt(sb, NO_UID32);
1565 break;
1566 case Opt_debug:
1567 set_opt(sb, DEBUG);
1568 break;
1569 case Opt_oldalloc:
1570 set_opt(sb, OLDALLOC);
1571 break;
1572 case Opt_orlov:
1573 clear_opt(sb, OLDALLOC);
1574 break;
1575 #ifdef CONFIG_EXT4_FS_XATTR
1576 case Opt_user_xattr:
1577 set_opt(sb, XATTR_USER);
1578 break;
1579 case Opt_nouser_xattr:
1580 clear_opt(sb, XATTR_USER);
1581 break;
1582 #else
1583 case Opt_user_xattr:
1584 case Opt_nouser_xattr:
1585 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1586 break;
1587 #endif
1588 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1589 case Opt_acl:
1590 set_opt(sb, POSIX_ACL);
1591 break;
1592 case Opt_noacl:
1593 clear_opt(sb, POSIX_ACL);
1594 break;
1595 #else
1596 case Opt_acl:
1597 case Opt_noacl:
1598 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1599 break;
1600 #endif
1601 case Opt_journal_update:
1602 /* @@@ FIXME */
1603 /* Eventually we will want to be able to create
1604 a journal file here. For now, only allow the
1605 user to specify an existing inode to be the
1606 journal file. */
1607 if (is_remount) {
1608 ext4_msg(sb, KERN_ERR,
1609 "Cannot specify journal on remount");
1610 return 0;
1611 }
1612 set_opt(sb, UPDATE_JOURNAL);
1613 break;
1614 case Opt_journal_dev:
1615 if (is_remount) {
1616 ext4_msg(sb, KERN_ERR,
1617 "Cannot specify journal on remount");
1618 return 0;
1619 }
1620 if (match_int(&args[0], &option))
1621 return 0;
1622 *journal_devnum = option;
1623 break;
1624 case Opt_journal_checksum:
1625 set_opt(sb, JOURNAL_CHECKSUM);
1626 break;
1627 case Opt_journal_async_commit:
1628 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1629 set_opt(sb, JOURNAL_CHECKSUM);
1630 break;
1631 case Opt_noload:
1632 set_opt(sb, NOLOAD);
1633 break;
1634 case Opt_commit:
1635 if (match_int(&args[0], &option))
1636 return 0;
1637 if (option < 0)
1638 return 0;
1639 if (option == 0)
1640 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1641 sbi->s_commit_interval = HZ * option;
1642 break;
1643 case Opt_max_batch_time:
1644 if (match_int(&args[0], &option))
1645 return 0;
1646 if (option < 0)
1647 return 0;
1648 if (option == 0)
1649 option = EXT4_DEF_MAX_BATCH_TIME;
1650 sbi->s_max_batch_time = option;
1651 break;
1652 case Opt_min_batch_time:
1653 if (match_int(&args[0], &option))
1654 return 0;
1655 if (option < 0)
1656 return 0;
1657 sbi->s_min_batch_time = option;
1658 break;
1659 case Opt_data_journal:
1660 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1661 goto datacheck;
1662 case Opt_data_ordered:
1663 data_opt = EXT4_MOUNT_ORDERED_DATA;
1664 goto datacheck;
1665 case Opt_data_writeback:
1666 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1667 datacheck:
1668 if (is_remount) {
1669 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1670 ext4_msg(sb, KERN_ERR,
1671 "Cannot change data mode on remount");
1672 return 0;
1673 }
1674 } else {
1675 clear_opt(sb, DATA_FLAGS);
1676 sbi->s_mount_opt |= data_opt;
1677 }
1678 break;
1679 case Opt_data_err_abort:
1680 set_opt(sb, DATA_ERR_ABORT);
1681 break;
1682 case Opt_data_err_ignore:
1683 clear_opt(sb, DATA_ERR_ABORT);
1684 break;
1685 #ifdef CONFIG_QUOTA
1686 case Opt_usrjquota:
1687 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1688 return 0;
1689 break;
1690 case Opt_grpjquota:
1691 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1692 return 0;
1693 break;
1694 case Opt_offusrjquota:
1695 if (!clear_qf_name(sb, USRQUOTA))
1696 return 0;
1697 break;
1698 case Opt_offgrpjquota:
1699 if (!clear_qf_name(sb, GRPQUOTA))
1700 return 0;
1701 break;
1702
1703 case Opt_jqfmt_vfsold:
1704 qfmt = QFMT_VFS_OLD;
1705 goto set_qf_format;
1706 case Opt_jqfmt_vfsv0:
1707 qfmt = QFMT_VFS_V0;
1708 goto set_qf_format;
1709 case Opt_jqfmt_vfsv1:
1710 qfmt = QFMT_VFS_V1;
1711 set_qf_format:
1712 if (sb_any_quota_loaded(sb) &&
1713 sbi->s_jquota_fmt != qfmt) {
1714 ext4_msg(sb, KERN_ERR, "Cannot change "
1715 "journaled quota options when "
1716 "quota turned on");
1717 return 0;
1718 }
1719 sbi->s_jquota_fmt = qfmt;
1720 break;
1721 case Opt_quota:
1722 case Opt_usrquota:
1723 set_opt(sb, QUOTA);
1724 set_opt(sb, USRQUOTA);
1725 break;
1726 case Opt_grpquota:
1727 set_opt(sb, QUOTA);
1728 set_opt(sb, GRPQUOTA);
1729 break;
1730 case Opt_noquota:
1731 if (sb_any_quota_loaded(sb)) {
1732 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1733 "options when quota turned on");
1734 return 0;
1735 }
1736 clear_opt(sb, QUOTA);
1737 clear_opt(sb, USRQUOTA);
1738 clear_opt(sb, GRPQUOTA);
1739 break;
1740 #else
1741 case Opt_quota:
1742 case Opt_usrquota:
1743 case Opt_grpquota:
1744 ext4_msg(sb, KERN_ERR,
1745 "quota options not supported");
1746 break;
1747 case Opt_usrjquota:
1748 case Opt_grpjquota:
1749 case Opt_offusrjquota:
1750 case Opt_offgrpjquota:
1751 case Opt_jqfmt_vfsold:
1752 case Opt_jqfmt_vfsv0:
1753 case Opt_jqfmt_vfsv1:
1754 ext4_msg(sb, KERN_ERR,
1755 "journaled quota options not supported");
1756 break;
1757 case Opt_noquota:
1758 break;
1759 #endif
1760 case Opt_abort:
1761 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1762 break;
1763 case Opt_nobarrier:
1764 clear_opt(sb, BARRIER);
1765 break;
1766 case Opt_barrier:
1767 if (args[0].from) {
1768 if (match_int(&args[0], &option))
1769 return 0;
1770 } else
1771 option = 1; /* No argument, default to 1 */
1772 if (option)
1773 set_opt(sb, BARRIER);
1774 else
1775 clear_opt(sb, BARRIER);
1776 break;
1777 case Opt_ignore:
1778 break;
1779 case Opt_resize:
1780 if (!is_remount) {
1781 ext4_msg(sb, KERN_ERR,
1782 "resize option only available "
1783 "for remount");
1784 return 0;
1785 }
1786 if (match_int(&args[0], &option) != 0)
1787 return 0;
1788 *n_blocks_count = option;
1789 break;
1790 case Opt_nobh:
1791 ext4_msg(sb, KERN_WARNING,
1792 "Ignoring deprecated nobh option");
1793 break;
1794 case Opt_bh:
1795 ext4_msg(sb, KERN_WARNING,
1796 "Ignoring deprecated bh option");
1797 break;
1798 case Opt_i_version:
1799 set_opt(sb, I_VERSION);
1800 sb->s_flags |= MS_I_VERSION;
1801 break;
1802 case Opt_nodelalloc:
1803 clear_opt(sb, DELALLOC);
1804 clear_opt2(sb, EXPLICIT_DELALLOC);
1805 break;
1806 case Opt_mblk_io_submit:
1807 set_opt(sb, MBLK_IO_SUBMIT);
1808 break;
1809 case Opt_nomblk_io_submit:
1810 clear_opt(sb, MBLK_IO_SUBMIT);
1811 break;
1812 case Opt_stripe:
1813 if (match_int(&args[0], &option))
1814 return 0;
1815 if (option < 0)
1816 return 0;
1817 sbi->s_stripe = option;
1818 break;
1819 case Opt_delalloc:
1820 set_opt(sb, DELALLOC);
1821 set_opt2(sb, EXPLICIT_DELALLOC);
1822 break;
1823 case Opt_block_validity:
1824 set_opt(sb, BLOCK_VALIDITY);
1825 break;
1826 case Opt_noblock_validity:
1827 clear_opt(sb, BLOCK_VALIDITY);
1828 break;
1829 case Opt_inode_readahead_blks:
1830 if (match_int(&args[0], &option))
1831 return 0;
1832 if (option < 0 || option > (1 << 30))
1833 return 0;
1834 if (option && !is_power_of_2(option)) {
1835 ext4_msg(sb, KERN_ERR,
1836 "EXT4-fs: inode_readahead_blks"
1837 " must be a power of 2");
1838 return 0;
1839 }
1840 sbi->s_inode_readahead_blks = option;
1841 break;
1842 case Opt_journal_ioprio:
1843 if (match_int(&args[0], &option))
1844 return 0;
1845 if (option < 0 || option > 7)
1846 break;
1847 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1848 option);
1849 break;
1850 case Opt_noauto_da_alloc:
1851 set_opt(sb, NO_AUTO_DA_ALLOC);
1852 break;
1853 case Opt_auto_da_alloc:
1854 if (args[0].from) {
1855 if (match_int(&args[0], &option))
1856 return 0;
1857 } else
1858 option = 1; /* No argument, default to 1 */
1859 if (option)
1860 clear_opt(sb, NO_AUTO_DA_ALLOC);
1861 else
1862 set_opt(sb,NO_AUTO_DA_ALLOC);
1863 break;
1864 case Opt_discard:
1865 set_opt(sb, DISCARD);
1866 break;
1867 case Opt_nodiscard:
1868 clear_opt(sb, DISCARD);
1869 break;
1870 case Opt_dioread_nolock:
1871 set_opt(sb, DIOREAD_NOLOCK);
1872 break;
1873 case Opt_dioread_lock:
1874 clear_opt(sb, DIOREAD_NOLOCK);
1875 break;
1876 case Opt_init_inode_table:
1877 set_opt(sb, INIT_INODE_TABLE);
1878 if (args[0].from) {
1879 if (match_int(&args[0], &option))
1880 return 0;
1881 } else
1882 option = EXT4_DEF_LI_WAIT_MULT;
1883 if (option < 0)
1884 return 0;
1885 sbi->s_li_wait_mult = option;
1886 break;
1887 case Opt_noinit_inode_table:
1888 clear_opt(sb, INIT_INODE_TABLE);
1889 break;
1890 default:
1891 ext4_msg(sb, KERN_ERR,
1892 "Unrecognized mount option \"%s\" "
1893 "or missing value", p);
1894 return 0;
1895 }
1896 }
1897 #ifdef CONFIG_QUOTA
1898 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1899 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1900 clear_opt(sb, USRQUOTA);
1901
1902 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1903 clear_opt(sb, GRPQUOTA);
1904
1905 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1906 ext4_msg(sb, KERN_ERR, "old and new quota "
1907 "format mixing");
1908 return 0;
1909 }
1910
1911 if (!sbi->s_jquota_fmt) {
1912 ext4_msg(sb, KERN_ERR, "journaled quota format "
1913 "not specified");
1914 return 0;
1915 }
1916 } else {
1917 if (sbi->s_jquota_fmt) {
1918 ext4_msg(sb, KERN_ERR, "journaled quota format "
1919 "specified with no journaling "
1920 "enabled");
1921 return 0;
1922 }
1923 }
1924 #endif
1925 return 1;
1926 }
1927
1928 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1929 int read_only)
1930 {
1931 struct ext4_sb_info *sbi = EXT4_SB(sb);
1932 int res = 0;
1933
1934 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1935 ext4_msg(sb, KERN_ERR, "revision level too high, "
1936 "forcing read-only mode");
1937 res = MS_RDONLY;
1938 }
1939 if (read_only)
1940 return res;
1941 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1942 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1943 "running e2fsck is recommended");
1944 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1945 ext4_msg(sb, KERN_WARNING,
1946 "warning: mounting fs with errors, "
1947 "running e2fsck is recommended");
1948 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1949 le16_to_cpu(es->s_mnt_count) >=
1950 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1951 ext4_msg(sb, KERN_WARNING,
1952 "warning: maximal mount count reached, "
1953 "running e2fsck is recommended");
1954 else if (le32_to_cpu(es->s_checkinterval) &&
1955 (le32_to_cpu(es->s_lastcheck) +
1956 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1957 ext4_msg(sb, KERN_WARNING,
1958 "warning: checktime reached, "
1959 "running e2fsck is recommended");
1960 if (!sbi->s_journal)
1961 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1962 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1963 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1964 le16_add_cpu(&es->s_mnt_count, 1);
1965 es->s_mtime = cpu_to_le32(get_seconds());
1966 ext4_update_dynamic_rev(sb);
1967 if (sbi->s_journal)
1968 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1969
1970 ext4_commit_super(sb, 1);
1971 if (test_opt(sb, DEBUG))
1972 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1973 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1974 sb->s_blocksize,
1975 sbi->s_groups_count,
1976 EXT4_BLOCKS_PER_GROUP(sb),
1977 EXT4_INODES_PER_GROUP(sb),
1978 sbi->s_mount_opt, sbi->s_mount_opt2);
1979
1980 cleancache_init_fs(sb);
1981 return res;
1982 }
1983
1984 static int ext4_fill_flex_info(struct super_block *sb)
1985 {
1986 struct ext4_sb_info *sbi = EXT4_SB(sb);
1987 struct ext4_group_desc *gdp = NULL;
1988 ext4_group_t flex_group_count;
1989 ext4_group_t flex_group;
1990 int groups_per_flex = 0;
1991 size_t size;
1992 int i;
1993
1994 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1995 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1996
1997 if (groups_per_flex < 2) {
1998 sbi->s_log_groups_per_flex = 0;
1999 return 1;
2000 }
2001
2002 /* We allocate both existing and potentially added groups */
2003 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2004 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2005 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2006 size = flex_group_count * sizeof(struct flex_groups);
2007 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2008 if (sbi->s_flex_groups == NULL) {
2009 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2010 flex_group_count);
2011 goto failed;
2012 }
2013
2014 for (i = 0; i < sbi->s_groups_count; i++) {
2015 gdp = ext4_get_group_desc(sb, i, NULL);
2016
2017 flex_group = ext4_flex_group(sbi, i);
2018 atomic_add(ext4_free_inodes_count(sb, gdp),
2019 &sbi->s_flex_groups[flex_group].free_inodes);
2020 atomic_add(ext4_free_blks_count(sb, gdp),
2021 &sbi->s_flex_groups[flex_group].free_blocks);
2022 atomic_add(ext4_used_dirs_count(sb, gdp),
2023 &sbi->s_flex_groups[flex_group].used_dirs);
2024 }
2025
2026 return 1;
2027 failed:
2028 return 0;
2029 }
2030
2031 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2032 struct ext4_group_desc *gdp)
2033 {
2034 __u16 crc = 0;
2035
2036 if (sbi->s_es->s_feature_ro_compat &
2037 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2038 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2039 __le32 le_group = cpu_to_le32(block_group);
2040
2041 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2042 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2043 crc = crc16(crc, (__u8 *)gdp, offset);
2044 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2045 /* for checksum of struct ext4_group_desc do the rest...*/
2046 if ((sbi->s_es->s_feature_incompat &
2047 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2048 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2049 crc = crc16(crc, (__u8 *)gdp + offset,
2050 le16_to_cpu(sbi->s_es->s_desc_size) -
2051 offset);
2052 }
2053
2054 return cpu_to_le16(crc);
2055 }
2056
2057 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2058 struct ext4_group_desc *gdp)
2059 {
2060 if ((sbi->s_es->s_feature_ro_compat &
2061 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2062 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2063 return 0;
2064
2065 return 1;
2066 }
2067
2068 /* Called at mount-time, super-block is locked */
2069 static int ext4_check_descriptors(struct super_block *sb,
2070 ext4_group_t *first_not_zeroed)
2071 {
2072 struct ext4_sb_info *sbi = EXT4_SB(sb);
2073 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2074 ext4_fsblk_t last_block;
2075 ext4_fsblk_t block_bitmap;
2076 ext4_fsblk_t inode_bitmap;
2077 ext4_fsblk_t inode_table;
2078 int flexbg_flag = 0;
2079 ext4_group_t i, grp = sbi->s_groups_count;
2080
2081 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2082 flexbg_flag = 1;
2083
2084 ext4_debug("Checking group descriptors");
2085
2086 for (i = 0; i < sbi->s_groups_count; i++) {
2087 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2088
2089 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2090 last_block = ext4_blocks_count(sbi->s_es) - 1;
2091 else
2092 last_block = first_block +
2093 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2094
2095 if ((grp == sbi->s_groups_count) &&
2096 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2097 grp = i;
2098
2099 block_bitmap = ext4_block_bitmap(sb, gdp);
2100 if (block_bitmap < first_block || block_bitmap > last_block) {
2101 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2102 "Block bitmap for group %u not in group "
2103 "(block %llu)!", i, block_bitmap);
2104 return 0;
2105 }
2106 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2107 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2108 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2109 "Inode bitmap for group %u not in group "
2110 "(block %llu)!", i, inode_bitmap);
2111 return 0;
2112 }
2113 inode_table = ext4_inode_table(sb, gdp);
2114 if (inode_table < first_block ||
2115 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2116 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2117 "Inode table for group %u not in group "
2118 "(block %llu)!", i, inode_table);
2119 return 0;
2120 }
2121 ext4_lock_group(sb, i);
2122 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2123 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2124 "Checksum for group %u failed (%u!=%u)",
2125 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2126 gdp)), le16_to_cpu(gdp->bg_checksum));
2127 if (!(sb->s_flags & MS_RDONLY)) {
2128 ext4_unlock_group(sb, i);
2129 return 0;
2130 }
2131 }
2132 ext4_unlock_group(sb, i);
2133 if (!flexbg_flag)
2134 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2135 }
2136 if (NULL != first_not_zeroed)
2137 *first_not_zeroed = grp;
2138
2139 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2140 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2141 return 1;
2142 }
2143
2144 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2145 * the superblock) which were deleted from all directories, but held open by
2146 * a process at the time of a crash. We walk the list and try to delete these
2147 * inodes at recovery time (only with a read-write filesystem).
2148 *
2149 * In order to keep the orphan inode chain consistent during traversal (in
2150 * case of crash during recovery), we link each inode into the superblock
2151 * orphan list_head and handle it the same way as an inode deletion during
2152 * normal operation (which journals the operations for us).
2153 *
2154 * We only do an iget() and an iput() on each inode, which is very safe if we
2155 * accidentally point at an in-use or already deleted inode. The worst that
2156 * can happen in this case is that we get a "bit already cleared" message from
2157 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2158 * e2fsck was run on this filesystem, and it must have already done the orphan
2159 * inode cleanup for us, so we can safely abort without any further action.
2160 */
2161 static void ext4_orphan_cleanup(struct super_block *sb,
2162 struct ext4_super_block *es)
2163 {
2164 unsigned int s_flags = sb->s_flags;
2165 int nr_orphans = 0, nr_truncates = 0;
2166 #ifdef CONFIG_QUOTA
2167 int i;
2168 #endif
2169 if (!es->s_last_orphan) {
2170 jbd_debug(4, "no orphan inodes to clean up\n");
2171 return;
2172 }
2173
2174 if (bdev_read_only(sb->s_bdev)) {
2175 ext4_msg(sb, KERN_ERR, "write access "
2176 "unavailable, skipping orphan cleanup");
2177 return;
2178 }
2179
2180 /* Check if feature set would not allow a r/w mount */
2181 if (!ext4_feature_set_ok(sb, 0)) {
2182 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2183 "unknown ROCOMPAT features");
2184 return;
2185 }
2186
2187 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2188 if (es->s_last_orphan)
2189 jbd_debug(1, "Errors on filesystem, "
2190 "clearing orphan list.\n");
2191 es->s_last_orphan = 0;
2192 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2193 return;
2194 }
2195
2196 if (s_flags & MS_RDONLY) {
2197 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2198 sb->s_flags &= ~MS_RDONLY;
2199 }
2200 #ifdef CONFIG_QUOTA
2201 /* Needed for iput() to work correctly and not trash data */
2202 sb->s_flags |= MS_ACTIVE;
2203 /* Turn on quotas so that they are updated correctly */
2204 for (i = 0; i < MAXQUOTAS; i++) {
2205 if (EXT4_SB(sb)->s_qf_names[i]) {
2206 int ret = ext4_quota_on_mount(sb, i);
2207 if (ret < 0)
2208 ext4_msg(sb, KERN_ERR,
2209 "Cannot turn on journaled "
2210 "quota: error %d", ret);
2211 }
2212 }
2213 #endif
2214
2215 while (es->s_last_orphan) {
2216 struct inode *inode;
2217
2218 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2219 if (IS_ERR(inode)) {
2220 es->s_last_orphan = 0;
2221 break;
2222 }
2223
2224 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2225 dquot_initialize(inode);
2226 if (inode->i_nlink) {
2227 ext4_msg(sb, KERN_DEBUG,
2228 "%s: truncating inode %lu to %lld bytes",
2229 __func__, inode->i_ino, inode->i_size);
2230 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2231 inode->i_ino, inode->i_size);
2232 ext4_truncate(inode);
2233 nr_truncates++;
2234 } else {
2235 ext4_msg(sb, KERN_DEBUG,
2236 "%s: deleting unreferenced inode %lu",
2237 __func__, inode->i_ino);
2238 jbd_debug(2, "deleting unreferenced inode %lu\n",
2239 inode->i_ino);
2240 nr_orphans++;
2241 }
2242 iput(inode); /* The delete magic happens here! */
2243 }
2244
2245 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2246
2247 if (nr_orphans)
2248 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2249 PLURAL(nr_orphans));
2250 if (nr_truncates)
2251 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2252 PLURAL(nr_truncates));
2253 #ifdef CONFIG_QUOTA
2254 /* Turn quotas off */
2255 for (i = 0; i < MAXQUOTAS; i++) {
2256 if (sb_dqopt(sb)->files[i])
2257 dquot_quota_off(sb, i);
2258 }
2259 #endif
2260 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2261 }
2262
2263 /*
2264 * Maximal extent format file size.
2265 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2266 * extent format containers, within a sector_t, and within i_blocks
2267 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2268 * so that won't be a limiting factor.
2269 *
2270 * However there is other limiting factor. We do store extents in the form
2271 * of starting block and length, hence the resulting length of the extent
2272 * covering maximum file size must fit into on-disk format containers as
2273 * well. Given that length is always by 1 unit bigger than max unit (because
2274 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2275 *
2276 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2277 */
2278 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2279 {
2280 loff_t res;
2281 loff_t upper_limit = MAX_LFS_FILESIZE;
2282
2283 /* small i_blocks in vfs inode? */
2284 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2285 /*
2286 * CONFIG_LBDAF is not enabled implies the inode
2287 * i_block represent total blocks in 512 bytes
2288 * 32 == size of vfs inode i_blocks * 8
2289 */
2290 upper_limit = (1LL << 32) - 1;
2291
2292 /* total blocks in file system block size */
2293 upper_limit >>= (blkbits - 9);
2294 upper_limit <<= blkbits;
2295 }
2296
2297 /*
2298 * 32-bit extent-start container, ee_block. We lower the maxbytes
2299 * by one fs block, so ee_len can cover the extent of maximum file
2300 * size
2301 */
2302 res = (1LL << 32) - 1;
2303 res <<= blkbits;
2304
2305 /* Sanity check against vm- & vfs- imposed limits */
2306 if (res > upper_limit)
2307 res = upper_limit;
2308
2309 return res;
2310 }
2311
2312 /*
2313 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2314 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2315 * We need to be 1 filesystem block less than the 2^48 sector limit.
2316 */
2317 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2318 {
2319 loff_t res = EXT4_NDIR_BLOCKS;
2320 int meta_blocks;
2321 loff_t upper_limit;
2322 /* This is calculated to be the largest file size for a dense, block
2323 * mapped file such that the file's total number of 512-byte sectors,
2324 * including data and all indirect blocks, does not exceed (2^48 - 1).
2325 *
2326 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2327 * number of 512-byte sectors of the file.
2328 */
2329
2330 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2331 /*
2332 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2333 * the inode i_block field represents total file blocks in
2334 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2335 */
2336 upper_limit = (1LL << 32) - 1;
2337
2338 /* total blocks in file system block size */
2339 upper_limit >>= (bits - 9);
2340
2341 } else {
2342 /*
2343 * We use 48 bit ext4_inode i_blocks
2344 * With EXT4_HUGE_FILE_FL set the i_blocks
2345 * represent total number of blocks in
2346 * file system block size
2347 */
2348 upper_limit = (1LL << 48) - 1;
2349
2350 }
2351
2352 /* indirect blocks */
2353 meta_blocks = 1;
2354 /* double indirect blocks */
2355 meta_blocks += 1 + (1LL << (bits-2));
2356 /* tripple indirect blocks */
2357 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2358
2359 upper_limit -= meta_blocks;
2360 upper_limit <<= bits;
2361
2362 res += 1LL << (bits-2);
2363 res += 1LL << (2*(bits-2));
2364 res += 1LL << (3*(bits-2));
2365 res <<= bits;
2366 if (res > upper_limit)
2367 res = upper_limit;
2368
2369 if (res > MAX_LFS_FILESIZE)
2370 res = MAX_LFS_FILESIZE;
2371
2372 return res;
2373 }
2374
2375 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2376 ext4_fsblk_t logical_sb_block, int nr)
2377 {
2378 struct ext4_sb_info *sbi = EXT4_SB(sb);
2379 ext4_group_t bg, first_meta_bg;
2380 int has_super = 0;
2381
2382 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2383
2384 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2385 nr < first_meta_bg)
2386 return logical_sb_block + nr + 1;
2387 bg = sbi->s_desc_per_block * nr;
2388 if (ext4_bg_has_super(sb, bg))
2389 has_super = 1;
2390
2391 return (has_super + ext4_group_first_block_no(sb, bg));
2392 }
2393
2394 /**
2395 * ext4_get_stripe_size: Get the stripe size.
2396 * @sbi: In memory super block info
2397 *
2398 * If we have specified it via mount option, then
2399 * use the mount option value. If the value specified at mount time is
2400 * greater than the blocks per group use the super block value.
2401 * If the super block value is greater than blocks per group return 0.
2402 * Allocator needs it be less than blocks per group.
2403 *
2404 */
2405 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2406 {
2407 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2408 unsigned long stripe_width =
2409 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2410 int ret;
2411
2412 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2413 ret = sbi->s_stripe;
2414 else if (stripe_width <= sbi->s_blocks_per_group)
2415 ret = stripe_width;
2416 else if (stride <= sbi->s_blocks_per_group)
2417 ret = stride;
2418 else
2419 ret = 0;
2420
2421 /*
2422 * If the stripe width is 1, this makes no sense and
2423 * we set it to 0 to turn off stripe handling code.
2424 */
2425 if (ret <= 1)
2426 ret = 0;
2427
2428 return ret;
2429 }
2430
2431 /* sysfs supprt */
2432
2433 struct ext4_attr {
2434 struct attribute attr;
2435 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2436 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2437 const char *, size_t);
2438 int offset;
2439 };
2440
2441 static int parse_strtoul(const char *buf,
2442 unsigned long max, unsigned long *value)
2443 {
2444 char *endp;
2445
2446 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2447 endp = skip_spaces(endp);
2448 if (*endp || *value > max)
2449 return -EINVAL;
2450
2451 return 0;
2452 }
2453
2454 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2455 struct ext4_sb_info *sbi,
2456 char *buf)
2457 {
2458 return snprintf(buf, PAGE_SIZE, "%llu\n",
2459 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2460 }
2461
2462 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2463 struct ext4_sb_info *sbi, char *buf)
2464 {
2465 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2466
2467 if (!sb->s_bdev->bd_part)
2468 return snprintf(buf, PAGE_SIZE, "0\n");
2469 return snprintf(buf, PAGE_SIZE, "%lu\n",
2470 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2471 sbi->s_sectors_written_start) >> 1);
2472 }
2473
2474 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2475 struct ext4_sb_info *sbi, char *buf)
2476 {
2477 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2478
2479 if (!sb->s_bdev->bd_part)
2480 return snprintf(buf, PAGE_SIZE, "0\n");
2481 return snprintf(buf, PAGE_SIZE, "%llu\n",
2482 (unsigned long long)(sbi->s_kbytes_written +
2483 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2484 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2485 }
2486
2487 static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2488 struct ext4_sb_info *sbi, char *buf)
2489 {
2490 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2491 }
2492
2493 static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2494 struct ext4_sb_info *sbi, char *buf)
2495 {
2496 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2497 }
2498
2499 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2500 struct ext4_sb_info *sbi,
2501 const char *buf, size_t count)
2502 {
2503 unsigned long t;
2504
2505 if (parse_strtoul(buf, 0x40000000, &t))
2506 return -EINVAL;
2507
2508 if (t && !is_power_of_2(t))
2509 return -EINVAL;
2510
2511 sbi->s_inode_readahead_blks = t;
2512 return count;
2513 }
2514
2515 static ssize_t sbi_ui_show(struct ext4_attr *a,
2516 struct ext4_sb_info *sbi, char *buf)
2517 {
2518 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2519
2520 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2521 }
2522
2523 static ssize_t sbi_ui_store(struct ext4_attr *a,
2524 struct ext4_sb_info *sbi,
2525 const char *buf, size_t count)
2526 {
2527 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2528 unsigned long t;
2529
2530 if (parse_strtoul(buf, 0xffffffff, &t))
2531 return -EINVAL;
2532 *ui = t;
2533 return count;
2534 }
2535
2536 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2537 static struct ext4_attr ext4_attr_##_name = { \
2538 .attr = {.name = __stringify(_name), .mode = _mode }, \
2539 .show = _show, \
2540 .store = _store, \
2541 .offset = offsetof(struct ext4_sb_info, _elname), \
2542 }
2543 #define EXT4_ATTR(name, mode, show, store) \
2544 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2545
2546 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2547 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2548 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2549 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2550 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2551 #define ATTR_LIST(name) &ext4_attr_##name.attr
2552
2553 EXT4_RO_ATTR(delayed_allocation_blocks);
2554 EXT4_RO_ATTR(session_write_kbytes);
2555 EXT4_RO_ATTR(lifetime_write_kbytes);
2556 EXT4_RO_ATTR(extent_cache_hits);
2557 EXT4_RO_ATTR(extent_cache_misses);
2558 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2559 inode_readahead_blks_store, s_inode_readahead_blks);
2560 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2561 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2562 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2563 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2564 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2565 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2566 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2567 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2568
2569 static struct attribute *ext4_attrs[] = {
2570 ATTR_LIST(delayed_allocation_blocks),
2571 ATTR_LIST(session_write_kbytes),
2572 ATTR_LIST(lifetime_write_kbytes),
2573 ATTR_LIST(extent_cache_hits),
2574 ATTR_LIST(extent_cache_misses),
2575 ATTR_LIST(inode_readahead_blks),
2576 ATTR_LIST(inode_goal),
2577 ATTR_LIST(mb_stats),
2578 ATTR_LIST(mb_max_to_scan),
2579 ATTR_LIST(mb_min_to_scan),
2580 ATTR_LIST(mb_order2_req),
2581 ATTR_LIST(mb_stream_req),
2582 ATTR_LIST(mb_group_prealloc),
2583 ATTR_LIST(max_writeback_mb_bump),
2584 NULL,
2585 };
2586
2587 /* Features this copy of ext4 supports */
2588 EXT4_INFO_ATTR(lazy_itable_init);
2589 EXT4_INFO_ATTR(batched_discard);
2590
2591 static struct attribute *ext4_feat_attrs[] = {
2592 ATTR_LIST(lazy_itable_init),
2593 ATTR_LIST(batched_discard),
2594 NULL,
2595 };
2596
2597 static ssize_t ext4_attr_show(struct kobject *kobj,
2598 struct attribute *attr, char *buf)
2599 {
2600 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2601 s_kobj);
2602 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2603
2604 return a->show ? a->show(a, sbi, buf) : 0;
2605 }
2606
2607 static ssize_t ext4_attr_store(struct kobject *kobj,
2608 struct attribute *attr,
2609 const char *buf, size_t len)
2610 {
2611 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2612 s_kobj);
2613 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2614
2615 return a->store ? a->store(a, sbi, buf, len) : 0;
2616 }
2617
2618 static void ext4_sb_release(struct kobject *kobj)
2619 {
2620 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2621 s_kobj);
2622 complete(&sbi->s_kobj_unregister);
2623 }
2624
2625 static const struct sysfs_ops ext4_attr_ops = {
2626 .show = ext4_attr_show,
2627 .store = ext4_attr_store,
2628 };
2629
2630 static struct kobj_type ext4_ktype = {
2631 .default_attrs = ext4_attrs,
2632 .sysfs_ops = &ext4_attr_ops,
2633 .release = ext4_sb_release,
2634 };
2635
2636 static void ext4_feat_release(struct kobject *kobj)
2637 {
2638 complete(&ext4_feat->f_kobj_unregister);
2639 }
2640
2641 static struct kobj_type ext4_feat_ktype = {
2642 .default_attrs = ext4_feat_attrs,
2643 .sysfs_ops = &ext4_attr_ops,
2644 .release = ext4_feat_release,
2645 };
2646
2647 /*
2648 * Check whether this filesystem can be mounted based on
2649 * the features present and the RDONLY/RDWR mount requested.
2650 * Returns 1 if this filesystem can be mounted as requested,
2651 * 0 if it cannot be.
2652 */
2653 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2654 {
2655 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2656 ext4_msg(sb, KERN_ERR,
2657 "Couldn't mount because of "
2658 "unsupported optional features (%x)",
2659 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2660 ~EXT4_FEATURE_INCOMPAT_SUPP));
2661 return 0;
2662 }
2663
2664 if (readonly)
2665 return 1;
2666
2667 /* Check that feature set is OK for a read-write mount */
2668 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2669 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2670 "unsupported optional features (%x)",
2671 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2672 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2673 return 0;
2674 }
2675 /*
2676 * Large file size enabled file system can only be mounted
2677 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2678 */
2679 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2680 if (sizeof(blkcnt_t) < sizeof(u64)) {
2681 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2682 "cannot be mounted RDWR without "
2683 "CONFIG_LBDAF");
2684 return 0;
2685 }
2686 }
2687 return 1;
2688 }
2689
2690 /*
2691 * This function is called once a day if we have errors logged
2692 * on the file system
2693 */
2694 static void print_daily_error_info(unsigned long arg)
2695 {
2696 struct super_block *sb = (struct super_block *) arg;
2697 struct ext4_sb_info *sbi;
2698 struct ext4_super_block *es;
2699
2700 sbi = EXT4_SB(sb);
2701 es = sbi->s_es;
2702
2703 if (es->s_error_count)
2704 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2705 le32_to_cpu(es->s_error_count));
2706 if (es->s_first_error_time) {
2707 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2708 sb->s_id, le32_to_cpu(es->s_first_error_time),
2709 (int) sizeof(es->s_first_error_func),
2710 es->s_first_error_func,
2711 le32_to_cpu(es->s_first_error_line));
2712 if (es->s_first_error_ino)
2713 printk(": inode %u",
2714 le32_to_cpu(es->s_first_error_ino));
2715 if (es->s_first_error_block)
2716 printk(": block %llu", (unsigned long long)
2717 le64_to_cpu(es->s_first_error_block));
2718 printk("\n");
2719 }
2720 if (es->s_last_error_time) {
2721 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2722 sb->s_id, le32_to_cpu(es->s_last_error_time),
2723 (int) sizeof(es->s_last_error_func),
2724 es->s_last_error_func,
2725 le32_to_cpu(es->s_last_error_line));
2726 if (es->s_last_error_ino)
2727 printk(": inode %u",
2728 le32_to_cpu(es->s_last_error_ino));
2729 if (es->s_last_error_block)
2730 printk(": block %llu", (unsigned long long)
2731 le64_to_cpu(es->s_last_error_block));
2732 printk("\n");
2733 }
2734 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2735 }
2736
2737 /* Find next suitable group and run ext4_init_inode_table */
2738 static int ext4_run_li_request(struct ext4_li_request *elr)
2739 {
2740 struct ext4_group_desc *gdp = NULL;
2741 ext4_group_t group, ngroups;
2742 struct super_block *sb;
2743 unsigned long timeout = 0;
2744 int ret = 0;
2745
2746 sb = elr->lr_super;
2747 ngroups = EXT4_SB(sb)->s_groups_count;
2748
2749 for (group = elr->lr_next_group; group < ngroups; group++) {
2750 gdp = ext4_get_group_desc(sb, group, NULL);
2751 if (!gdp) {
2752 ret = 1;
2753 break;
2754 }
2755
2756 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2757 break;
2758 }
2759
2760 if (group == ngroups)
2761 ret = 1;
2762
2763 if (!ret) {
2764 timeout = jiffies;
2765 ret = ext4_init_inode_table(sb, group,
2766 elr->lr_timeout ? 0 : 1);
2767 if (elr->lr_timeout == 0) {
2768 timeout = (jiffies - timeout) *
2769 elr->lr_sbi->s_li_wait_mult;
2770 elr->lr_timeout = timeout;
2771 }
2772 elr->lr_next_sched = jiffies + elr->lr_timeout;
2773 elr->lr_next_group = group + 1;
2774 }
2775
2776 return ret;
2777 }
2778
2779 /*
2780 * Remove lr_request from the list_request and free the
2781 * request structure. Should be called with li_list_mtx held
2782 */
2783 static void ext4_remove_li_request(struct ext4_li_request *elr)
2784 {
2785 struct ext4_sb_info *sbi;
2786
2787 if (!elr)
2788 return;
2789
2790 sbi = elr->lr_sbi;
2791
2792 list_del(&elr->lr_request);
2793 sbi->s_li_request = NULL;
2794 kfree(elr);
2795 }
2796
2797 static void ext4_unregister_li_request(struct super_block *sb)
2798 {
2799 mutex_lock(&ext4_li_mtx);
2800 if (!ext4_li_info) {
2801 mutex_unlock(&ext4_li_mtx);
2802 return;
2803 }
2804
2805 mutex_lock(&ext4_li_info->li_list_mtx);
2806 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2807 mutex_unlock(&ext4_li_info->li_list_mtx);
2808 mutex_unlock(&ext4_li_mtx);
2809 }
2810
2811 static struct task_struct *ext4_lazyinit_task;
2812
2813 /*
2814 * This is the function where ext4lazyinit thread lives. It walks
2815 * through the request list searching for next scheduled filesystem.
2816 * When such a fs is found, run the lazy initialization request
2817 * (ext4_rn_li_request) and keep track of the time spend in this
2818 * function. Based on that time we compute next schedule time of
2819 * the request. When walking through the list is complete, compute
2820 * next waking time and put itself into sleep.
2821 */
2822 static int ext4_lazyinit_thread(void *arg)
2823 {
2824 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2825 struct list_head *pos, *n;
2826 struct ext4_li_request *elr;
2827 unsigned long next_wakeup, cur;
2828
2829 BUG_ON(NULL == eli);
2830
2831 cont_thread:
2832 while (true) {
2833 next_wakeup = MAX_JIFFY_OFFSET;
2834
2835 mutex_lock(&eli->li_list_mtx);
2836 if (list_empty(&eli->li_request_list)) {
2837 mutex_unlock(&eli->li_list_mtx);
2838 goto exit_thread;
2839 }
2840
2841 list_for_each_safe(pos, n, &eli->li_request_list) {
2842 elr = list_entry(pos, struct ext4_li_request,
2843 lr_request);
2844
2845 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2846 if (ext4_run_li_request(elr) != 0) {
2847 /* error, remove the lazy_init job */
2848 ext4_remove_li_request(elr);
2849 continue;
2850 }
2851 }
2852
2853 if (time_before(elr->lr_next_sched, next_wakeup))
2854 next_wakeup = elr->lr_next_sched;
2855 }
2856 mutex_unlock(&eli->li_list_mtx);
2857
2858 try_to_freeze();
2859
2860 cur = jiffies;
2861 if ((time_after_eq(cur, next_wakeup)) ||
2862 (MAX_JIFFY_OFFSET == next_wakeup)) {
2863 cond_resched();
2864 continue;
2865 }
2866
2867 schedule_timeout_interruptible(next_wakeup - cur);
2868
2869 if (kthread_should_stop()) {
2870 ext4_clear_request_list();
2871 goto exit_thread;
2872 }
2873 }
2874
2875 exit_thread:
2876 /*
2877 * It looks like the request list is empty, but we need
2878 * to check it under the li_list_mtx lock, to prevent any
2879 * additions into it, and of course we should lock ext4_li_mtx
2880 * to atomically free the list and ext4_li_info, because at
2881 * this point another ext4 filesystem could be registering
2882 * new one.
2883 */
2884 mutex_lock(&ext4_li_mtx);
2885 mutex_lock(&eli->li_list_mtx);
2886 if (!list_empty(&eli->li_request_list)) {
2887 mutex_unlock(&eli->li_list_mtx);
2888 mutex_unlock(&ext4_li_mtx);
2889 goto cont_thread;
2890 }
2891 mutex_unlock(&eli->li_list_mtx);
2892 kfree(ext4_li_info);
2893 ext4_li_info = NULL;
2894 mutex_unlock(&ext4_li_mtx);
2895
2896 return 0;
2897 }
2898
2899 static void ext4_clear_request_list(void)
2900 {
2901 struct list_head *pos, *n;
2902 struct ext4_li_request *elr;
2903
2904 mutex_lock(&ext4_li_info->li_list_mtx);
2905 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2906 elr = list_entry(pos, struct ext4_li_request,
2907 lr_request);
2908 ext4_remove_li_request(elr);
2909 }
2910 mutex_unlock(&ext4_li_info->li_list_mtx);
2911 }
2912
2913 static int ext4_run_lazyinit_thread(void)
2914 {
2915 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2916 ext4_li_info, "ext4lazyinit");
2917 if (IS_ERR(ext4_lazyinit_task)) {
2918 int err = PTR_ERR(ext4_lazyinit_task);
2919 ext4_clear_request_list();
2920 kfree(ext4_li_info);
2921 ext4_li_info = NULL;
2922 printk(KERN_CRIT "EXT4: error %d creating inode table "
2923 "initialization thread\n",
2924 err);
2925 return err;
2926 }
2927 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2928 return 0;
2929 }
2930
2931 /*
2932 * Check whether it make sense to run itable init. thread or not.
2933 * If there is at least one uninitialized inode table, return
2934 * corresponding group number, else the loop goes through all
2935 * groups and return total number of groups.
2936 */
2937 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2938 {
2939 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2940 struct ext4_group_desc *gdp = NULL;
2941
2942 for (group = 0; group < ngroups; group++) {
2943 gdp = ext4_get_group_desc(sb, group, NULL);
2944 if (!gdp)
2945 continue;
2946
2947 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2948 break;
2949 }
2950
2951 return group;
2952 }
2953
2954 static int ext4_li_info_new(void)
2955 {
2956 struct ext4_lazy_init *eli = NULL;
2957
2958 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2959 if (!eli)
2960 return -ENOMEM;
2961
2962 INIT_LIST_HEAD(&eli->li_request_list);
2963 mutex_init(&eli->li_list_mtx);
2964
2965 eli->li_state |= EXT4_LAZYINIT_QUIT;
2966
2967 ext4_li_info = eli;
2968
2969 return 0;
2970 }
2971
2972 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2973 ext4_group_t start)
2974 {
2975 struct ext4_sb_info *sbi = EXT4_SB(sb);
2976 struct ext4_li_request *elr;
2977 unsigned long rnd;
2978
2979 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2980 if (!elr)
2981 return NULL;
2982
2983 elr->lr_super = sb;
2984 elr->lr_sbi = sbi;
2985 elr->lr_next_group = start;
2986
2987 /*
2988 * Randomize first schedule time of the request to
2989 * spread the inode table initialization requests
2990 * better.
2991 */
2992 get_random_bytes(&rnd, sizeof(rnd));
2993 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2994 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2995
2996 return elr;
2997 }
2998
2999 static int ext4_register_li_request(struct super_block *sb,
3000 ext4_group_t first_not_zeroed)
3001 {
3002 struct ext4_sb_info *sbi = EXT4_SB(sb);
3003 struct ext4_li_request *elr;
3004 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3005 int ret = 0;
3006
3007 if (sbi->s_li_request != NULL) {
3008 /*
3009 * Reset timeout so it can be computed again, because
3010 * s_li_wait_mult might have changed.
3011 */
3012 sbi->s_li_request->lr_timeout = 0;
3013 return 0;
3014 }
3015
3016 if (first_not_zeroed == ngroups ||
3017 (sb->s_flags & MS_RDONLY) ||
3018 !test_opt(sb, INIT_INODE_TABLE))
3019 return 0;
3020
3021 elr = ext4_li_request_new(sb, first_not_zeroed);
3022 if (!elr)
3023 return -ENOMEM;
3024
3025 mutex_lock(&ext4_li_mtx);
3026
3027 if (NULL == ext4_li_info) {
3028 ret = ext4_li_info_new();
3029 if (ret)
3030 goto out;
3031 }
3032
3033 mutex_lock(&ext4_li_info->li_list_mtx);
3034 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3035 mutex_unlock(&ext4_li_info->li_list_mtx);
3036
3037 sbi->s_li_request = elr;
3038 /*
3039 * set elr to NULL here since it has been inserted to
3040 * the request_list and the removal and free of it is
3041 * handled by ext4_clear_request_list from now on.
3042 */
3043 elr = NULL;
3044
3045 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3046 ret = ext4_run_lazyinit_thread();
3047 if (ret)
3048 goto out;
3049 }
3050 out:
3051 mutex_unlock(&ext4_li_mtx);
3052 if (ret)
3053 kfree(elr);
3054 return ret;
3055 }
3056
3057 /*
3058 * We do not need to lock anything since this is called on
3059 * module unload.
3060 */
3061 static void ext4_destroy_lazyinit_thread(void)
3062 {
3063 /*
3064 * If thread exited earlier
3065 * there's nothing to be done.
3066 */
3067 if (!ext4_li_info || !ext4_lazyinit_task)
3068 return;
3069
3070 kthread_stop(ext4_lazyinit_task);
3071 }
3072
3073 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3074 __releases(kernel_lock)
3075 __acquires(kernel_lock)
3076 {
3077 char *orig_data = kstrdup(data, GFP_KERNEL);
3078 struct buffer_head *bh;
3079 struct ext4_super_block *es = NULL;
3080 struct ext4_sb_info *sbi;
3081 ext4_fsblk_t block;
3082 ext4_fsblk_t sb_block = get_sb_block(&data);
3083 ext4_fsblk_t logical_sb_block;
3084 unsigned long offset = 0;
3085 unsigned long journal_devnum = 0;
3086 unsigned long def_mount_opts;
3087 struct inode *root;
3088 char *cp;
3089 const char *descr;
3090 int ret = -ENOMEM;
3091 int blocksize;
3092 unsigned int db_count;
3093 unsigned int i;
3094 int needs_recovery, has_huge_files;
3095 __u64 blocks_count;
3096 int err;
3097 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3098 ext4_group_t first_not_zeroed;
3099
3100 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3101 if (!sbi)
3102 goto out_free_orig;
3103
3104 sbi->s_blockgroup_lock =
3105 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3106 if (!sbi->s_blockgroup_lock) {
3107 kfree(sbi);
3108 goto out_free_orig;
3109 }
3110 sb->s_fs_info = sbi;
3111 sbi->s_mount_opt = 0;
3112 sbi->s_resuid = EXT4_DEF_RESUID;
3113 sbi->s_resgid = EXT4_DEF_RESGID;
3114 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3115 sbi->s_sb_block = sb_block;
3116 if (sb->s_bdev->bd_part)
3117 sbi->s_sectors_written_start =
3118 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3119
3120 /* Cleanup superblock name */
3121 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3122 *cp = '!';
3123
3124 ret = -EINVAL;
3125 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3126 if (!blocksize) {
3127 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3128 goto out_fail;
3129 }
3130
3131 /*
3132 * The ext4 superblock will not be buffer aligned for other than 1kB
3133 * block sizes. We need to calculate the offset from buffer start.
3134 */
3135 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3136 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3137 offset = do_div(logical_sb_block, blocksize);
3138 } else {
3139 logical_sb_block = sb_block;
3140 }
3141
3142 if (!(bh = sb_bread(sb, logical_sb_block))) {
3143 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3144 goto out_fail;
3145 }
3146 /*
3147 * Note: s_es must be initialized as soon as possible because
3148 * some ext4 macro-instructions depend on its value
3149 */
3150 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3151 sbi->s_es = es;
3152 sb->s_magic = le16_to_cpu(es->s_magic);
3153 if (sb->s_magic != EXT4_SUPER_MAGIC)
3154 goto cantfind_ext4;
3155 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3156
3157 /* Set defaults before we parse the mount options */
3158 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3159 set_opt(sb, INIT_INODE_TABLE);
3160 if (def_mount_opts & EXT4_DEFM_DEBUG)
3161 set_opt(sb, DEBUG);
3162 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3163 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3164 "2.6.38");
3165 set_opt(sb, GRPID);
3166 }
3167 if (def_mount_opts & EXT4_DEFM_UID16)
3168 set_opt(sb, NO_UID32);
3169 /* xattr user namespace & acls are now defaulted on */
3170 #ifdef CONFIG_EXT4_FS_XATTR
3171 set_opt(sb, XATTR_USER);
3172 #endif
3173 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3174 set_opt(sb, POSIX_ACL);
3175 #endif
3176 set_opt(sb, MBLK_IO_SUBMIT);
3177 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3178 set_opt(sb, JOURNAL_DATA);
3179 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3180 set_opt(sb, ORDERED_DATA);
3181 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3182 set_opt(sb, WRITEBACK_DATA);
3183
3184 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3185 set_opt(sb, ERRORS_PANIC);
3186 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3187 set_opt(sb, ERRORS_CONT);
3188 else
3189 set_opt(sb, ERRORS_RO);
3190 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3191 set_opt(sb, BLOCK_VALIDITY);
3192 if (def_mount_opts & EXT4_DEFM_DISCARD)
3193 set_opt(sb, DISCARD);
3194
3195 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3196 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3197 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3198 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3199 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3200
3201 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3202 set_opt(sb, BARRIER);
3203
3204 /*
3205 * enable delayed allocation by default
3206 * Use -o nodelalloc to turn it off
3207 */
3208 if (!IS_EXT3_SB(sb) &&
3209 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3210 set_opt(sb, DELALLOC);
3211
3212 /*
3213 * set default s_li_wait_mult for lazyinit, for the case there is
3214 * no mount option specified.
3215 */
3216 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3217
3218 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3219 &journal_devnum, &journal_ioprio, NULL, 0)) {
3220 ext4_msg(sb, KERN_WARNING,
3221 "failed to parse options in superblock: %s",
3222 sbi->s_es->s_mount_opts);
3223 }
3224 if (!parse_options((char *) data, sb, &journal_devnum,
3225 &journal_ioprio, NULL, 0))
3226 goto failed_mount;
3227
3228 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3229 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3230 "with data=journal disables delayed "
3231 "allocation and O_DIRECT support!\n");
3232 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3233 ext4_msg(sb, KERN_ERR, "can't mount with "
3234 "both data=journal and delalloc");
3235 goto failed_mount;
3236 }
3237 if (test_opt(sb, DIOREAD_NOLOCK)) {
3238 ext4_msg(sb, KERN_ERR, "can't mount with "
3239 "both data=journal and delalloc");
3240 goto failed_mount;
3241 }
3242 if (test_opt(sb, DELALLOC))
3243 clear_opt(sb, DELALLOC);
3244 }
3245 if (test_opt(sb, DIOREAD_NOLOCK)) {
3246 if (sb->s_blocksize < PAGE_SIZE) {
3247 ext4_msg(sb, KERN_ERR, "can't mount with "
3248 "dioread_nolock if block size != PAGE_SIZE");
3249 goto failed_mount;
3250 }
3251 }
3252
3253 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3254 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3255
3256 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3257 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3258 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3259 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3260 ext4_msg(sb, KERN_WARNING,
3261 "feature flags set on rev 0 fs, "
3262 "running e2fsck is recommended");
3263
3264 if (IS_EXT2_SB(sb)) {
3265 if (ext2_feature_set_ok(sb))
3266 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3267 "using the ext4 subsystem");
3268 else {
3269 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3270 "to feature incompatibilities");
3271 goto failed_mount;
3272 }
3273 }
3274
3275 if (IS_EXT3_SB(sb)) {
3276 if (ext3_feature_set_ok(sb))
3277 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3278 "using the ext4 subsystem");
3279 else {
3280 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3281 "to feature incompatibilities");
3282 goto failed_mount;
3283 }
3284 }
3285
3286 /*
3287 * Check feature flags regardless of the revision level, since we
3288 * previously didn't change the revision level when setting the flags,
3289 * so there is a chance incompat flags are set on a rev 0 filesystem.
3290 */
3291 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3292 goto failed_mount;
3293
3294 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3295
3296 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3297 blocksize > EXT4_MAX_BLOCK_SIZE) {
3298 ext4_msg(sb, KERN_ERR,
3299 "Unsupported filesystem blocksize %d", blocksize);
3300 goto failed_mount;
3301 }
3302
3303 if (sb->s_blocksize != blocksize) {
3304 /* Validate the filesystem blocksize */
3305 if (!sb_set_blocksize(sb, blocksize)) {
3306 ext4_msg(sb, KERN_ERR, "bad block size %d",
3307 blocksize);
3308 goto failed_mount;
3309 }
3310
3311 brelse(bh);
3312 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3313 offset = do_div(logical_sb_block, blocksize);
3314 bh = sb_bread(sb, logical_sb_block);
3315 if (!bh) {
3316 ext4_msg(sb, KERN_ERR,
3317 "Can't read superblock on 2nd try");
3318 goto failed_mount;
3319 }
3320 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3321 sbi->s_es = es;
3322 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3323 ext4_msg(sb, KERN_ERR,
3324 "Magic mismatch, very weird!");
3325 goto failed_mount;
3326 }
3327 }
3328
3329 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3330 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3331 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3332 has_huge_files);
3333 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3334
3335 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3336 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3337 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3338 } else {
3339 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3340 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3341 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3342 (!is_power_of_2(sbi->s_inode_size)) ||
3343 (sbi->s_inode_size > blocksize)) {
3344 ext4_msg(sb, KERN_ERR,
3345 "unsupported inode size: %d",
3346 sbi->s_inode_size);
3347 goto failed_mount;
3348 }
3349 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3350 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3351 }
3352
3353 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3354 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3355 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3356 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3357 !is_power_of_2(sbi->s_desc_size)) {
3358 ext4_msg(sb, KERN_ERR,
3359 "unsupported descriptor size %lu",
3360 sbi->s_desc_size);
3361 goto failed_mount;
3362 }
3363 } else
3364 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3365
3366 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3367 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3368 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3369 goto cantfind_ext4;
3370
3371 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3372 if (sbi->s_inodes_per_block == 0)
3373 goto cantfind_ext4;
3374 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3375 sbi->s_inodes_per_block;
3376 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3377 sbi->s_sbh = bh;
3378 sbi->s_mount_state = le16_to_cpu(es->s_state);
3379 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3380 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3381
3382 for (i = 0; i < 4; i++)
3383 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3384 sbi->s_def_hash_version = es->s_def_hash_version;
3385 i = le32_to_cpu(es->s_flags);
3386 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3387 sbi->s_hash_unsigned = 3;
3388 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3389 #ifdef __CHAR_UNSIGNED__
3390 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3391 sbi->s_hash_unsigned = 3;
3392 #else
3393 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3394 #endif
3395 sb->s_dirt = 1;
3396 }
3397
3398 if (sbi->s_blocks_per_group > blocksize * 8) {
3399 ext4_msg(sb, KERN_ERR,
3400 "#blocks per group too big: %lu",
3401 sbi->s_blocks_per_group);
3402 goto failed_mount;
3403 }
3404 if (sbi->s_inodes_per_group > blocksize * 8) {
3405 ext4_msg(sb, KERN_ERR,
3406 "#inodes per group too big: %lu",
3407 sbi->s_inodes_per_group);
3408 goto failed_mount;
3409 }
3410
3411 /*
3412 * Test whether we have more sectors than will fit in sector_t,
3413 * and whether the max offset is addressable by the page cache.
3414 */
3415 err = generic_check_addressable(sb->s_blocksize_bits,
3416 ext4_blocks_count(es));
3417 if (err) {
3418 ext4_msg(sb, KERN_ERR, "filesystem"
3419 " too large to mount safely on this system");
3420 if (sizeof(sector_t) < 8)
3421 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3422 ret = err;
3423 goto failed_mount;
3424 }
3425
3426 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3427 goto cantfind_ext4;
3428
3429 /* check blocks count against device size */
3430 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3431 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3432 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3433 "exceeds size of device (%llu blocks)",
3434 ext4_blocks_count(es), blocks_count);
3435 goto failed_mount;
3436 }
3437
3438 /*
3439 * It makes no sense for the first data block to be beyond the end
3440 * of the filesystem.
3441 */
3442 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3443 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3444 "block %u is beyond end of filesystem (%llu)",
3445 le32_to_cpu(es->s_first_data_block),
3446 ext4_blocks_count(es));
3447 goto failed_mount;
3448 }
3449 blocks_count = (ext4_blocks_count(es) -
3450 le32_to_cpu(es->s_first_data_block) +
3451 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3452 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3453 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3454 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3455 "(block count %llu, first data block %u, "
3456 "blocks per group %lu)", sbi->s_groups_count,
3457 ext4_blocks_count(es),
3458 le32_to_cpu(es->s_first_data_block),
3459 EXT4_BLOCKS_PER_GROUP(sb));
3460 goto failed_mount;
3461 }
3462 sbi->s_groups_count = blocks_count;
3463 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3464 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3465 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3466 EXT4_DESC_PER_BLOCK(sb);
3467 sbi->s_group_desc = ext4_kvmalloc(db_count *
3468 sizeof(struct buffer_head *),
3469 GFP_KERNEL);
3470 if (sbi->s_group_desc == NULL) {
3471 ext4_msg(sb, KERN_ERR, "not enough memory");
3472 goto failed_mount;
3473 }
3474
3475 #ifdef CONFIG_PROC_FS
3476 if (ext4_proc_root)
3477 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3478 #endif
3479
3480 bgl_lock_init(sbi->s_blockgroup_lock);
3481
3482 for (i = 0; i < db_count; i++) {
3483 block = descriptor_loc(sb, logical_sb_block, i);
3484 sbi->s_group_desc[i] = sb_bread(sb, block);
3485 if (!sbi->s_group_desc[i]) {
3486 ext4_msg(sb, KERN_ERR,
3487 "can't read group descriptor %d", i);
3488 db_count = i;
3489 goto failed_mount2;
3490 }
3491 }
3492 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3493 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3494 goto failed_mount2;
3495 }
3496 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3497 if (!ext4_fill_flex_info(sb)) {
3498 ext4_msg(sb, KERN_ERR,
3499 "unable to initialize "
3500 "flex_bg meta info!");
3501 goto failed_mount2;
3502 }
3503
3504 sbi->s_gdb_count = db_count;
3505 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3506 spin_lock_init(&sbi->s_next_gen_lock);
3507
3508 init_timer(&sbi->s_err_report);
3509 sbi->s_err_report.function = print_daily_error_info;
3510 sbi->s_err_report.data = (unsigned long) sb;
3511
3512 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3513 ext4_count_free_blocks(sb));
3514 if (!err) {
3515 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3516 ext4_count_free_inodes(sb));
3517 }
3518 if (!err) {
3519 err = percpu_counter_init(&sbi->s_dirs_counter,
3520 ext4_count_dirs(sb));
3521 }
3522 if (!err) {
3523 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3524 }
3525 if (err) {
3526 ext4_msg(sb, KERN_ERR, "insufficient memory");
3527 goto failed_mount3;
3528 }
3529
3530 sbi->s_stripe = ext4_get_stripe_size(sbi);
3531 sbi->s_max_writeback_mb_bump = 128;
3532
3533 /*
3534 * set up enough so that it can read an inode
3535 */
3536 if (!test_opt(sb, NOLOAD) &&
3537 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3538 sb->s_op = &ext4_sops;
3539 else
3540 sb->s_op = &ext4_nojournal_sops;
3541 sb->s_export_op = &ext4_export_ops;
3542 sb->s_xattr = ext4_xattr_handlers;
3543 #ifdef CONFIG_QUOTA
3544 sb->s_qcop = &ext4_qctl_operations;
3545 sb->dq_op = &ext4_quota_operations;
3546 #endif
3547 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3548
3549 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3550 mutex_init(&sbi->s_orphan_lock);
3551 sbi->s_resize_flags = 0;
3552
3553 sb->s_root = NULL;
3554
3555 needs_recovery = (es->s_last_orphan != 0 ||
3556 EXT4_HAS_INCOMPAT_FEATURE(sb,
3557 EXT4_FEATURE_INCOMPAT_RECOVER));
3558
3559 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3560 !(sb->s_flags & MS_RDONLY))
3561 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3562 goto failed_mount3;
3563
3564 /*
3565 * The first inode we look at is the journal inode. Don't try
3566 * root first: it may be modified in the journal!
3567 */
3568 if (!test_opt(sb, NOLOAD) &&
3569 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3570 if (ext4_load_journal(sb, es, journal_devnum))
3571 goto failed_mount3;
3572 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3573 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3574 ext4_msg(sb, KERN_ERR, "required journal recovery "
3575 "suppressed and not mounted read-only");
3576 goto failed_mount_wq;
3577 } else {
3578 clear_opt(sb, DATA_FLAGS);
3579 sbi->s_journal = NULL;
3580 needs_recovery = 0;
3581 goto no_journal;
3582 }
3583
3584 if (ext4_blocks_count(es) > 0xffffffffULL &&
3585 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3586 JBD2_FEATURE_INCOMPAT_64BIT)) {
3587 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3588 goto failed_mount_wq;
3589 }
3590
3591 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3592 jbd2_journal_set_features(sbi->s_journal,
3593 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3594 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3595 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3596 jbd2_journal_set_features(sbi->s_journal,
3597 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3598 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3599 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3600 } else {
3601 jbd2_journal_clear_features(sbi->s_journal,
3602 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3603 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3604 }
3605
3606 /* We have now updated the journal if required, so we can
3607 * validate the data journaling mode. */
3608 switch (test_opt(sb, DATA_FLAGS)) {
3609 case 0:
3610 /* No mode set, assume a default based on the journal
3611 * capabilities: ORDERED_DATA if the journal can
3612 * cope, else JOURNAL_DATA
3613 */
3614 if (jbd2_journal_check_available_features
3615 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3616 set_opt(sb, ORDERED_DATA);
3617 else
3618 set_opt(sb, JOURNAL_DATA);
3619 break;
3620
3621 case EXT4_MOUNT_ORDERED_DATA:
3622 case EXT4_MOUNT_WRITEBACK_DATA:
3623 if (!jbd2_journal_check_available_features
3624 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3625 ext4_msg(sb, KERN_ERR, "Journal does not support "
3626 "requested data journaling mode");
3627 goto failed_mount_wq;
3628 }
3629 default:
3630 break;
3631 }
3632 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3633
3634 /*
3635 * The journal may have updated the bg summary counts, so we
3636 * need to update the global counters.
3637 */
3638 percpu_counter_set(&sbi->s_freeblocks_counter,
3639 ext4_count_free_blocks(sb));
3640 percpu_counter_set(&sbi->s_freeinodes_counter,
3641 ext4_count_free_inodes(sb));
3642 percpu_counter_set(&sbi->s_dirs_counter,
3643 ext4_count_dirs(sb));
3644 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
3645
3646 no_journal:
3647 /*
3648 * The maximum number of concurrent works can be high and
3649 * concurrency isn't really necessary. Limit it to 1.
3650 */
3651 EXT4_SB(sb)->dio_unwritten_wq =
3652 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3653 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3654 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3655 goto failed_mount_wq;
3656 }
3657
3658 /*
3659 * The jbd2_journal_load will have done any necessary log recovery,
3660 * so we can safely mount the rest of the filesystem now.
3661 */
3662
3663 root = ext4_iget(sb, EXT4_ROOT_INO);
3664 if (IS_ERR(root)) {
3665 ext4_msg(sb, KERN_ERR, "get root inode failed");
3666 ret = PTR_ERR(root);
3667 root = NULL;
3668 goto failed_mount4;
3669 }
3670 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3671 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3672 goto failed_mount4;
3673 }
3674 sb->s_root = d_alloc_root(root);
3675 if (!sb->s_root) {
3676 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3677 ret = -ENOMEM;
3678 goto failed_mount4;
3679 }
3680
3681 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3682
3683 /* determine the minimum size of new large inodes, if present */
3684 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3685 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3686 EXT4_GOOD_OLD_INODE_SIZE;
3687 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3688 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3689 if (sbi->s_want_extra_isize <
3690 le16_to_cpu(es->s_want_extra_isize))
3691 sbi->s_want_extra_isize =
3692 le16_to_cpu(es->s_want_extra_isize);
3693 if (sbi->s_want_extra_isize <
3694 le16_to_cpu(es->s_min_extra_isize))
3695 sbi->s_want_extra_isize =
3696 le16_to_cpu(es->s_min_extra_isize);
3697 }
3698 }
3699 /* Check if enough inode space is available */
3700 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3701 sbi->s_inode_size) {
3702 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3703 EXT4_GOOD_OLD_INODE_SIZE;
3704 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3705 "available");
3706 }
3707
3708 err = ext4_setup_system_zone(sb);
3709 if (err) {
3710 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3711 "zone (%d)", err);
3712 goto failed_mount4;
3713 }
3714
3715 ext4_ext_init(sb);
3716 err = ext4_mb_init(sb, needs_recovery);
3717 if (err) {
3718 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3719 err);
3720 goto failed_mount4;
3721 }
3722
3723 err = ext4_register_li_request(sb, first_not_zeroed);
3724 if (err)
3725 goto failed_mount4;
3726
3727 sbi->s_kobj.kset = ext4_kset;
3728 init_completion(&sbi->s_kobj_unregister);
3729 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3730 "%s", sb->s_id);
3731 if (err) {
3732 ext4_mb_release(sb);
3733 ext4_ext_release(sb);
3734 goto failed_mount4;
3735 };
3736
3737 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3738 ext4_orphan_cleanup(sb, es);
3739 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3740 if (needs_recovery) {
3741 ext4_msg(sb, KERN_INFO, "recovery complete");
3742 ext4_mark_recovery_complete(sb, es);
3743 }
3744 if (EXT4_SB(sb)->s_journal) {
3745 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3746 descr = " journalled data mode";
3747 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3748 descr = " ordered data mode";
3749 else
3750 descr = " writeback data mode";
3751 } else
3752 descr = "out journal";
3753
3754 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3755 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3756 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3757
3758 if (es->s_error_count)
3759 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3760
3761 kfree(orig_data);
3762 return 0;
3763
3764 cantfind_ext4:
3765 if (!silent)
3766 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3767 goto failed_mount;
3768
3769 failed_mount4:
3770 iput(root);
3771 sb->s_root = NULL;
3772 ext4_msg(sb, KERN_ERR, "mount failed");
3773 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3774 failed_mount_wq:
3775 ext4_release_system_zone(sb);
3776 if (sbi->s_journal) {
3777 jbd2_journal_destroy(sbi->s_journal);
3778 sbi->s_journal = NULL;
3779 }
3780 failed_mount3:
3781 del_timer(&sbi->s_err_report);
3782 if (sbi->s_flex_groups)
3783 ext4_kvfree(sbi->s_flex_groups);
3784 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3785 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3786 percpu_counter_destroy(&sbi->s_dirs_counter);
3787 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3788 if (sbi->s_mmp_tsk)
3789 kthread_stop(sbi->s_mmp_tsk);
3790 failed_mount2:
3791 for (i = 0; i < db_count; i++)
3792 brelse(sbi->s_group_desc[i]);
3793 ext4_kvfree(sbi->s_group_desc);
3794 failed_mount:
3795 if (sbi->s_proc) {
3796 remove_proc_entry(sb->s_id, ext4_proc_root);
3797 }
3798 #ifdef CONFIG_QUOTA
3799 for (i = 0; i < MAXQUOTAS; i++)
3800 kfree(sbi->s_qf_names[i]);
3801 #endif
3802 ext4_blkdev_remove(sbi);
3803 brelse(bh);
3804 out_fail:
3805 sb->s_fs_info = NULL;
3806 kfree(sbi->s_blockgroup_lock);
3807 kfree(sbi);
3808 out_free_orig:
3809 kfree(orig_data);
3810 return ret;
3811 }
3812
3813 /*
3814 * Setup any per-fs journal parameters now. We'll do this both on
3815 * initial mount, once the journal has been initialised but before we've
3816 * done any recovery; and again on any subsequent remount.
3817 */
3818 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3819 {
3820 struct ext4_sb_info *sbi = EXT4_SB(sb);
3821
3822 journal->j_commit_interval = sbi->s_commit_interval;
3823 journal->j_min_batch_time = sbi->s_min_batch_time;
3824 journal->j_max_batch_time = sbi->s_max_batch_time;
3825
3826 write_lock(&journal->j_state_lock);
3827 if (test_opt(sb, BARRIER))
3828 journal->j_flags |= JBD2_BARRIER;
3829 else
3830 journal->j_flags &= ~JBD2_BARRIER;
3831 if (test_opt(sb, DATA_ERR_ABORT))
3832 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3833 else
3834 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3835 write_unlock(&journal->j_state_lock);
3836 }
3837
3838 static journal_t *ext4_get_journal(struct super_block *sb,
3839 unsigned int journal_inum)
3840 {
3841 struct inode *journal_inode;
3842 journal_t *journal;
3843
3844 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3845
3846 /* First, test for the existence of a valid inode on disk. Bad
3847 * things happen if we iget() an unused inode, as the subsequent
3848 * iput() will try to delete it. */
3849
3850 journal_inode = ext4_iget(sb, journal_inum);
3851 if (IS_ERR(journal_inode)) {
3852 ext4_msg(sb, KERN_ERR, "no journal found");
3853 return NULL;
3854 }
3855 if (!journal_inode->i_nlink) {
3856 make_bad_inode(journal_inode);
3857 iput(journal_inode);
3858 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3859 return NULL;
3860 }
3861
3862 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3863 journal_inode, journal_inode->i_size);
3864 if (!S_ISREG(journal_inode->i_mode)) {
3865 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3866 iput(journal_inode);
3867 return NULL;
3868 }
3869
3870 journal = jbd2_journal_init_inode(journal_inode);
3871 if (!journal) {
3872 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3873 iput(journal_inode);
3874 return NULL;
3875 }
3876 journal->j_private = sb;
3877 ext4_init_journal_params(sb, journal);
3878 return journal;
3879 }
3880
3881 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3882 dev_t j_dev)
3883 {
3884 struct buffer_head *bh;
3885 journal_t *journal;
3886 ext4_fsblk_t start;
3887 ext4_fsblk_t len;
3888 int hblock, blocksize;
3889 ext4_fsblk_t sb_block;
3890 unsigned long offset;
3891 struct ext4_super_block *es;
3892 struct block_device *bdev;
3893
3894 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3895
3896 bdev = ext4_blkdev_get(j_dev, sb);
3897 if (bdev == NULL)
3898 return NULL;
3899
3900 blocksize = sb->s_blocksize;
3901 hblock = bdev_logical_block_size(bdev);
3902 if (blocksize < hblock) {
3903 ext4_msg(sb, KERN_ERR,
3904 "blocksize too small for journal device");
3905 goto out_bdev;
3906 }
3907
3908 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3909 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3910 set_blocksize(bdev, blocksize);
3911 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3912 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3913 "external journal");
3914 goto out_bdev;
3915 }
3916
3917 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3918 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3919 !(le32_to_cpu(es->s_feature_incompat) &
3920 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3921 ext4_msg(sb, KERN_ERR, "external journal has "
3922 "bad superblock");
3923 brelse(bh);
3924 goto out_bdev;
3925 }
3926
3927 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3928 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3929 brelse(bh);
3930 goto out_bdev;
3931 }
3932
3933 len = ext4_blocks_count(es);
3934 start = sb_block + 1;
3935 brelse(bh); /* we're done with the superblock */
3936
3937 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3938 start, len, blocksize);
3939 if (!journal) {
3940 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3941 goto out_bdev;
3942 }
3943 journal->j_private = sb;
3944 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3945 wait_on_buffer(journal->j_sb_buffer);
3946 if (!buffer_uptodate(journal->j_sb_buffer)) {
3947 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3948 goto out_journal;
3949 }
3950 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3951 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3952 "user (unsupported) - %d",
3953 be32_to_cpu(journal->j_superblock->s_nr_users));
3954 goto out_journal;
3955 }
3956 EXT4_SB(sb)->journal_bdev = bdev;
3957 ext4_init_journal_params(sb, journal);
3958 return journal;
3959
3960 out_journal:
3961 jbd2_journal_destroy(journal);
3962 out_bdev:
3963 ext4_blkdev_put(bdev);
3964 return NULL;
3965 }
3966
3967 static int ext4_load_journal(struct super_block *sb,
3968 struct ext4_super_block *es,
3969 unsigned long journal_devnum)
3970 {
3971 journal_t *journal;
3972 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3973 dev_t journal_dev;
3974 int err = 0;
3975 int really_read_only;
3976
3977 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3978
3979 if (journal_devnum &&
3980 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3981 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3982 "numbers have changed");
3983 journal_dev = new_decode_dev(journal_devnum);
3984 } else
3985 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3986
3987 really_read_only = bdev_read_only(sb->s_bdev);
3988
3989 /*
3990 * Are we loading a blank journal or performing recovery after a
3991 * crash? For recovery, we need to check in advance whether we
3992 * can get read-write access to the device.
3993 */
3994 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3995 if (sb->s_flags & MS_RDONLY) {
3996 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3997 "required on readonly filesystem");
3998 if (really_read_only) {
3999 ext4_msg(sb, KERN_ERR, "write access "
4000 "unavailable, cannot proceed");
4001 return -EROFS;
4002 }
4003 ext4_msg(sb, KERN_INFO, "write access will "
4004 "be enabled during recovery");
4005 }
4006 }
4007
4008 if (journal_inum && journal_dev) {
4009 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4010 "and inode journals!");
4011 return -EINVAL;
4012 }
4013
4014 if (journal_inum) {
4015 if (!(journal = ext4_get_journal(sb, journal_inum)))
4016 return -EINVAL;
4017 } else {
4018 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4019 return -EINVAL;
4020 }
4021
4022 if (!(journal->j_flags & JBD2_BARRIER))
4023 ext4_msg(sb, KERN_INFO, "barriers disabled");
4024
4025 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
4026 err = jbd2_journal_update_format(journal);
4027 if (err) {
4028 ext4_msg(sb, KERN_ERR, "error updating journal");
4029 jbd2_journal_destroy(journal);
4030 return err;
4031 }
4032 }
4033
4034 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4035 err = jbd2_journal_wipe(journal, !really_read_only);
4036 if (!err) {
4037 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4038 if (save)
4039 memcpy(save, ((char *) es) +
4040 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4041 err = jbd2_journal_load(journal);
4042 if (save)
4043 memcpy(((char *) es) + EXT4_S_ERR_START,
4044 save, EXT4_S_ERR_LEN);
4045 kfree(save);
4046 }
4047
4048 if (err) {
4049 ext4_msg(sb, KERN_ERR, "error loading journal");
4050 jbd2_journal_destroy(journal);
4051 return err;
4052 }
4053
4054 EXT4_SB(sb)->s_journal = journal;
4055 ext4_clear_journal_err(sb, es);
4056
4057 if (!really_read_only && journal_devnum &&
4058 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4059 es->s_journal_dev = cpu_to_le32(journal_devnum);
4060
4061 /* Make sure we flush the recovery flag to disk. */
4062 ext4_commit_super(sb, 1);
4063 }
4064
4065 return 0;
4066 }
4067
4068 static int ext4_commit_super(struct super_block *sb, int sync)
4069 {
4070 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4071 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4072 int error = 0;
4073
4074 if (!sbh)
4075 return error;
4076 if (buffer_write_io_error(sbh)) {
4077 /*
4078 * Oh, dear. A previous attempt to write the
4079 * superblock failed. This could happen because the
4080 * USB device was yanked out. Or it could happen to
4081 * be a transient write error and maybe the block will
4082 * be remapped. Nothing we can do but to retry the
4083 * write and hope for the best.
4084 */
4085 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4086 "superblock detected");
4087 clear_buffer_write_io_error(sbh);
4088 set_buffer_uptodate(sbh);
4089 }
4090 /*
4091 * If the file system is mounted read-only, don't update the
4092 * superblock write time. This avoids updating the superblock
4093 * write time when we are mounting the root file system
4094 * read/only but we need to replay the journal; at that point,
4095 * for people who are east of GMT and who make their clock
4096 * tick in localtime for Windows bug-for-bug compatibility,
4097 * the clock is set in the future, and this will cause e2fsck
4098 * to complain and force a full file system check.
4099 */
4100 if (!(sb->s_flags & MS_RDONLY))
4101 es->s_wtime = cpu_to_le32(get_seconds());
4102 if (sb->s_bdev->bd_part)
4103 es->s_kbytes_written =
4104 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4105 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4106 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4107 else
4108 es->s_kbytes_written =
4109 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4110 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
4111 &EXT4_SB(sb)->s_freeblocks_counter));
4112 es->s_free_inodes_count =
4113 cpu_to_le32(percpu_counter_sum_positive(
4114 &EXT4_SB(sb)->s_freeinodes_counter));
4115 sb->s_dirt = 0;
4116 BUFFER_TRACE(sbh, "marking dirty");
4117 mark_buffer_dirty(sbh);
4118 if (sync) {
4119 error = sync_dirty_buffer(sbh);
4120 if (error)
4121 return error;
4122
4123 error = buffer_write_io_error(sbh);
4124 if (error) {
4125 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4126 "superblock");
4127 clear_buffer_write_io_error(sbh);
4128 set_buffer_uptodate(sbh);
4129 }
4130 }
4131 return error;
4132 }
4133
4134 /*
4135 * Have we just finished recovery? If so, and if we are mounting (or
4136 * remounting) the filesystem readonly, then we will end up with a
4137 * consistent fs on disk. Record that fact.
4138 */
4139 static void ext4_mark_recovery_complete(struct super_block *sb,
4140 struct ext4_super_block *es)
4141 {
4142 journal_t *journal = EXT4_SB(sb)->s_journal;
4143
4144 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4145 BUG_ON(journal != NULL);
4146 return;
4147 }
4148 jbd2_journal_lock_updates(journal);
4149 if (jbd2_journal_flush(journal) < 0)
4150 goto out;
4151
4152 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4153 sb->s_flags & MS_RDONLY) {
4154 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4155 ext4_commit_super(sb, 1);
4156 }
4157
4158 out:
4159 jbd2_journal_unlock_updates(journal);
4160 }
4161
4162 /*
4163 * If we are mounting (or read-write remounting) a filesystem whose journal
4164 * has recorded an error from a previous lifetime, move that error to the
4165 * main filesystem now.
4166 */
4167 static void ext4_clear_journal_err(struct super_block *sb,
4168 struct ext4_super_block *es)
4169 {
4170 journal_t *journal;
4171 int j_errno;
4172 const char *errstr;
4173
4174 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4175
4176 journal = EXT4_SB(sb)->s_journal;
4177
4178 /*
4179 * Now check for any error status which may have been recorded in the
4180 * journal by a prior ext4_error() or ext4_abort()
4181 */
4182
4183 j_errno = jbd2_journal_errno(journal);
4184 if (j_errno) {
4185 char nbuf[16];
4186
4187 errstr = ext4_decode_error(sb, j_errno, nbuf);
4188 ext4_warning(sb, "Filesystem error recorded "
4189 "from previous mount: %s", errstr);
4190 ext4_warning(sb, "Marking fs in need of filesystem check.");
4191
4192 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4193 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4194 ext4_commit_super(sb, 1);
4195
4196 jbd2_journal_clear_err(journal);
4197 }
4198 }
4199
4200 /*
4201 * Force the running and committing transactions to commit,
4202 * and wait on the commit.
4203 */
4204 int ext4_force_commit(struct super_block *sb)
4205 {
4206 journal_t *journal;
4207 int ret = 0;
4208
4209 if (sb->s_flags & MS_RDONLY)
4210 return 0;
4211
4212 journal = EXT4_SB(sb)->s_journal;
4213 if (journal) {
4214 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4215 ret = ext4_journal_force_commit(journal);
4216 }
4217
4218 return ret;
4219 }
4220
4221 static void ext4_write_super(struct super_block *sb)
4222 {
4223 lock_super(sb);
4224 ext4_commit_super(sb, 1);
4225 unlock_super(sb);
4226 }
4227
4228 static int ext4_sync_fs(struct super_block *sb, int wait)
4229 {
4230 int ret = 0;
4231 tid_t target;
4232 struct ext4_sb_info *sbi = EXT4_SB(sb);
4233
4234 trace_ext4_sync_fs(sb, wait);
4235 flush_workqueue(sbi->dio_unwritten_wq);
4236 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4237 if (wait)
4238 jbd2_log_wait_commit(sbi->s_journal, target);
4239 }
4240 return ret;
4241 }
4242
4243 /*
4244 * LVM calls this function before a (read-only) snapshot is created. This
4245 * gives us a chance to flush the journal completely and mark the fs clean.
4246 *
4247 * Note that only this function cannot bring a filesystem to be in a clean
4248 * state independently, because ext4 prevents a new handle from being started
4249 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4250 * the upper layer.
4251 */
4252 static int ext4_freeze(struct super_block *sb)
4253 {
4254 int error = 0;
4255 journal_t *journal;
4256
4257 if (sb->s_flags & MS_RDONLY)
4258 return 0;
4259
4260 journal = EXT4_SB(sb)->s_journal;
4261
4262 /* Now we set up the journal barrier. */
4263 jbd2_journal_lock_updates(journal);
4264
4265 /*
4266 * Don't clear the needs_recovery flag if we failed to flush
4267 * the journal.
4268 */
4269 error = jbd2_journal_flush(journal);
4270 if (error < 0)
4271 goto out;
4272
4273 /* Journal blocked and flushed, clear needs_recovery flag. */
4274 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4275 error = ext4_commit_super(sb, 1);
4276 out:
4277 /* we rely on s_frozen to stop further updates */
4278 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4279 return error;
4280 }
4281
4282 /*
4283 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4284 * flag here, even though the filesystem is not technically dirty yet.
4285 */
4286 static int ext4_unfreeze(struct super_block *sb)
4287 {
4288 if (sb->s_flags & MS_RDONLY)
4289 return 0;
4290
4291 lock_super(sb);
4292 /* Reset the needs_recovery flag before the fs is unlocked. */
4293 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4294 ext4_commit_super(sb, 1);
4295 unlock_super(sb);
4296 return 0;
4297 }
4298
4299 /*
4300 * Structure to save mount options for ext4_remount's benefit
4301 */
4302 struct ext4_mount_options {
4303 unsigned long s_mount_opt;
4304 unsigned long s_mount_opt2;
4305 uid_t s_resuid;
4306 gid_t s_resgid;
4307 unsigned long s_commit_interval;
4308 u32 s_min_batch_time, s_max_batch_time;
4309 #ifdef CONFIG_QUOTA
4310 int s_jquota_fmt;
4311 char *s_qf_names[MAXQUOTAS];
4312 #endif
4313 };
4314
4315 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4316 {
4317 struct ext4_super_block *es;
4318 struct ext4_sb_info *sbi = EXT4_SB(sb);
4319 ext4_fsblk_t n_blocks_count = 0;
4320 unsigned long old_sb_flags;
4321 struct ext4_mount_options old_opts;
4322 int enable_quota = 0;
4323 ext4_group_t g;
4324 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4325 int err = 0;
4326 #ifdef CONFIG_QUOTA
4327 int i;
4328 #endif
4329 char *orig_data = kstrdup(data, GFP_KERNEL);
4330
4331 /* Store the original options */
4332 lock_super(sb);
4333 old_sb_flags = sb->s_flags;
4334 old_opts.s_mount_opt = sbi->s_mount_opt;
4335 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4336 old_opts.s_resuid = sbi->s_resuid;
4337 old_opts.s_resgid = sbi->s_resgid;
4338 old_opts.s_commit_interval = sbi->s_commit_interval;
4339 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4340 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4341 #ifdef CONFIG_QUOTA
4342 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4343 for (i = 0; i < MAXQUOTAS; i++)
4344 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4345 #endif
4346 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4347 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4348
4349 /*
4350 * Allow the "check" option to be passed as a remount option.
4351 */
4352 if (!parse_options(data, sb, NULL, &journal_ioprio,
4353 &n_blocks_count, 1)) {
4354 err = -EINVAL;
4355 goto restore_opts;
4356 }
4357
4358 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4359 ext4_abort(sb, "Abort forced by user");
4360
4361 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4362 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4363
4364 es = sbi->s_es;
4365
4366 if (sbi->s_journal) {
4367 ext4_init_journal_params(sb, sbi->s_journal);
4368 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4369 }
4370
4371 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4372 n_blocks_count > ext4_blocks_count(es)) {
4373 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4374 err = -EROFS;
4375 goto restore_opts;
4376 }
4377
4378 if (*flags & MS_RDONLY) {
4379 err = dquot_suspend(sb, -1);
4380 if (err < 0)
4381 goto restore_opts;
4382
4383 /*
4384 * First of all, the unconditional stuff we have to do
4385 * to disable replay of the journal when we next remount
4386 */
4387 sb->s_flags |= MS_RDONLY;
4388
4389 /*
4390 * OK, test if we are remounting a valid rw partition
4391 * readonly, and if so set the rdonly flag and then
4392 * mark the partition as valid again.
4393 */
4394 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4395 (sbi->s_mount_state & EXT4_VALID_FS))
4396 es->s_state = cpu_to_le16(sbi->s_mount_state);
4397
4398 if (sbi->s_journal)
4399 ext4_mark_recovery_complete(sb, es);
4400 } else {
4401 /* Make sure we can mount this feature set readwrite */
4402 if (!ext4_feature_set_ok(sb, 0)) {
4403 err = -EROFS;
4404 goto restore_opts;
4405 }
4406 /*
4407 * Make sure the group descriptor checksums
4408 * are sane. If they aren't, refuse to remount r/w.
4409 */
4410 for (g = 0; g < sbi->s_groups_count; g++) {
4411 struct ext4_group_desc *gdp =
4412 ext4_get_group_desc(sb, g, NULL);
4413
4414 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4415 ext4_msg(sb, KERN_ERR,
4416 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4417 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4418 le16_to_cpu(gdp->bg_checksum));
4419 err = -EINVAL;
4420 goto restore_opts;
4421 }
4422 }
4423
4424 /*
4425 * If we have an unprocessed orphan list hanging
4426 * around from a previously readonly bdev mount,
4427 * require a full umount/remount for now.
4428 */
4429 if (es->s_last_orphan) {
4430 ext4_msg(sb, KERN_WARNING, "Couldn't "
4431 "remount RDWR because of unprocessed "
4432 "orphan inode list. Please "
4433 "umount/remount instead");
4434 err = -EINVAL;
4435 goto restore_opts;
4436 }
4437
4438 /*
4439 * Mounting a RDONLY partition read-write, so reread
4440 * and store the current valid flag. (It may have
4441 * been changed by e2fsck since we originally mounted
4442 * the partition.)
4443 */
4444 if (sbi->s_journal)
4445 ext4_clear_journal_err(sb, es);
4446 sbi->s_mount_state = le16_to_cpu(es->s_state);
4447 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4448 goto restore_opts;
4449 if (!ext4_setup_super(sb, es, 0))
4450 sb->s_flags &= ~MS_RDONLY;
4451 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4452 EXT4_FEATURE_INCOMPAT_MMP))
4453 if (ext4_multi_mount_protect(sb,
4454 le64_to_cpu(es->s_mmp_block))) {
4455 err = -EROFS;
4456 goto restore_opts;
4457 }
4458 enable_quota = 1;
4459 }
4460 }
4461
4462 /*
4463 * Reinitialize lazy itable initialization thread based on
4464 * current settings
4465 */
4466 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4467 ext4_unregister_li_request(sb);
4468 else {
4469 ext4_group_t first_not_zeroed;
4470 first_not_zeroed = ext4_has_uninit_itable(sb);
4471 ext4_register_li_request(sb, first_not_zeroed);
4472 }
4473
4474 ext4_setup_system_zone(sb);
4475 if (sbi->s_journal == NULL)
4476 ext4_commit_super(sb, 1);
4477
4478 #ifdef CONFIG_QUOTA
4479 /* Release old quota file names */
4480 for (i = 0; i < MAXQUOTAS; i++)
4481 if (old_opts.s_qf_names[i] &&
4482 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4483 kfree(old_opts.s_qf_names[i]);
4484 #endif
4485 unlock_super(sb);
4486 if (enable_quota)
4487 dquot_resume(sb, -1);
4488
4489 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4490 kfree(orig_data);
4491 return 0;
4492
4493 restore_opts:
4494 sb->s_flags = old_sb_flags;
4495 sbi->s_mount_opt = old_opts.s_mount_opt;
4496 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4497 sbi->s_resuid = old_opts.s_resuid;
4498 sbi->s_resgid = old_opts.s_resgid;
4499 sbi->s_commit_interval = old_opts.s_commit_interval;
4500 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4501 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4502 #ifdef CONFIG_QUOTA
4503 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4504 for (i = 0; i < MAXQUOTAS; i++) {
4505 if (sbi->s_qf_names[i] &&
4506 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4507 kfree(sbi->s_qf_names[i]);
4508 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4509 }
4510 #endif
4511 unlock_super(sb);
4512 kfree(orig_data);
4513 return err;
4514 }
4515
4516 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4517 {
4518 struct super_block *sb = dentry->d_sb;
4519 struct ext4_sb_info *sbi = EXT4_SB(sb);
4520 struct ext4_super_block *es = sbi->s_es;
4521 u64 fsid;
4522 s64 bfree;
4523
4524 if (test_opt(sb, MINIX_DF)) {
4525 sbi->s_overhead_last = 0;
4526 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4527 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4528 ext4_fsblk_t overhead = 0;
4529
4530 /*
4531 * Compute the overhead (FS structures). This is constant
4532 * for a given filesystem unless the number of block groups
4533 * changes so we cache the previous value until it does.
4534 */
4535
4536 /*
4537 * All of the blocks before first_data_block are
4538 * overhead
4539 */
4540 overhead = le32_to_cpu(es->s_first_data_block);
4541
4542 /*
4543 * Add the overhead attributed to the superblock and
4544 * block group descriptors. If the sparse superblocks
4545 * feature is turned on, then not all groups have this.
4546 */
4547 for (i = 0; i < ngroups; i++) {
4548 overhead += ext4_bg_has_super(sb, i) +
4549 ext4_bg_num_gdb(sb, i);
4550 cond_resched();
4551 }
4552
4553 /*
4554 * Every block group has an inode bitmap, a block
4555 * bitmap, and an inode table.
4556 */
4557 overhead += ngroups * (2 + sbi->s_itb_per_group);
4558 sbi->s_overhead_last = overhead;
4559 smp_wmb();
4560 sbi->s_blocks_last = ext4_blocks_count(es);
4561 }
4562
4563 buf->f_type = EXT4_SUPER_MAGIC;
4564 buf->f_bsize = sb->s_blocksize;
4565 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
4566 bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4567 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
4568 /* prevent underflow in case that few free space is available */
4569 buf->f_bfree = max_t(s64, bfree, 0);
4570 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4571 if (buf->f_bfree < ext4_r_blocks_count(es))
4572 buf->f_bavail = 0;
4573 buf->f_files = le32_to_cpu(es->s_inodes_count);
4574 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4575 buf->f_namelen = EXT4_NAME_LEN;
4576 fsid = le64_to_cpup((void *)es->s_uuid) ^
4577 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4578 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4579 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4580
4581 return 0;
4582 }
4583
4584 /* Helper function for writing quotas on sync - we need to start transaction
4585 * before quota file is locked for write. Otherwise the are possible deadlocks:
4586 * Process 1 Process 2
4587 * ext4_create() quota_sync()
4588 * jbd2_journal_start() write_dquot()
4589 * dquot_initialize() down(dqio_mutex)
4590 * down(dqio_mutex) jbd2_journal_start()
4591 *
4592 */
4593
4594 #ifdef CONFIG_QUOTA
4595
4596 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4597 {
4598 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4599 }
4600
4601 static int ext4_write_dquot(struct dquot *dquot)
4602 {
4603 int ret, err;
4604 handle_t *handle;
4605 struct inode *inode;
4606
4607 inode = dquot_to_inode(dquot);
4608 handle = ext4_journal_start(inode,
4609 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4610 if (IS_ERR(handle))
4611 return PTR_ERR(handle);
4612 ret = dquot_commit(dquot);
4613 err = ext4_journal_stop(handle);
4614 if (!ret)
4615 ret = err;
4616 return ret;
4617 }
4618
4619 static int ext4_acquire_dquot(struct dquot *dquot)
4620 {
4621 int ret, err;
4622 handle_t *handle;
4623
4624 handle = ext4_journal_start(dquot_to_inode(dquot),
4625 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4626 if (IS_ERR(handle))
4627 return PTR_ERR(handle);
4628 ret = dquot_acquire(dquot);
4629 err = ext4_journal_stop(handle);
4630 if (!ret)
4631 ret = err;
4632 return ret;
4633 }
4634
4635 static int ext4_release_dquot(struct dquot *dquot)
4636 {
4637 int ret, err;
4638 handle_t *handle;
4639
4640 handle = ext4_journal_start(dquot_to_inode(dquot),
4641 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4642 if (IS_ERR(handle)) {
4643 /* Release dquot anyway to avoid endless cycle in dqput() */
4644 dquot_release(dquot);
4645 return PTR_ERR(handle);
4646 }
4647 ret = dquot_release(dquot);
4648 err = ext4_journal_stop(handle);
4649 if (!ret)
4650 ret = err;
4651 return ret;
4652 }
4653
4654 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4655 {
4656 /* Are we journaling quotas? */
4657 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4658 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4659 dquot_mark_dquot_dirty(dquot);
4660 return ext4_write_dquot(dquot);
4661 } else {
4662 return dquot_mark_dquot_dirty(dquot);
4663 }
4664 }
4665
4666 static int ext4_write_info(struct super_block *sb, int type)
4667 {
4668 int ret, err;
4669 handle_t *handle;
4670
4671 /* Data block + inode block */
4672 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4673 if (IS_ERR(handle))
4674 return PTR_ERR(handle);
4675 ret = dquot_commit_info(sb, type);
4676 err = ext4_journal_stop(handle);
4677 if (!ret)
4678 ret = err;
4679 return ret;
4680 }
4681
4682 /*
4683 * Turn on quotas during mount time - we need to find
4684 * the quota file and such...
4685 */
4686 static int ext4_quota_on_mount(struct super_block *sb, int type)
4687 {
4688 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4689 EXT4_SB(sb)->s_jquota_fmt, type);
4690 }
4691
4692 /*
4693 * Standard function to be called on quota_on
4694 */
4695 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4696 struct path *path)
4697 {
4698 int err;
4699
4700 if (!test_opt(sb, QUOTA))
4701 return -EINVAL;
4702
4703 /* Quotafile not on the same filesystem? */
4704 if (path->mnt->mnt_sb != sb)
4705 return -EXDEV;
4706 /* Journaling quota? */
4707 if (EXT4_SB(sb)->s_qf_names[type]) {
4708 /* Quotafile not in fs root? */
4709 if (path->dentry->d_parent != sb->s_root)
4710 ext4_msg(sb, KERN_WARNING,
4711 "Quota file not on filesystem root. "
4712 "Journaled quota will not work");
4713 }
4714
4715 /*
4716 * When we journal data on quota file, we have to flush journal to see
4717 * all updates to the file when we bypass pagecache...
4718 */
4719 if (EXT4_SB(sb)->s_journal &&
4720 ext4_should_journal_data(path->dentry->d_inode)) {
4721 /*
4722 * We don't need to lock updates but journal_flush() could
4723 * otherwise be livelocked...
4724 */
4725 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4726 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4727 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4728 if (err)
4729 return err;
4730 }
4731
4732 return dquot_quota_on(sb, type, format_id, path);
4733 }
4734
4735 static int ext4_quota_off(struct super_block *sb, int type)
4736 {
4737 struct inode *inode = sb_dqopt(sb)->files[type];
4738 handle_t *handle;
4739
4740 /* Force all delayed allocation blocks to be allocated.
4741 * Caller already holds s_umount sem */
4742 if (test_opt(sb, DELALLOC))
4743 sync_filesystem(sb);
4744
4745 if (!inode)
4746 goto out;
4747
4748 /* Update modification times of quota files when userspace can
4749 * start looking at them */
4750 handle = ext4_journal_start(inode, 1);
4751 if (IS_ERR(handle))
4752 goto out;
4753 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4754 ext4_mark_inode_dirty(handle, inode);
4755 ext4_journal_stop(handle);
4756
4757 out:
4758 return dquot_quota_off(sb, type);
4759 }
4760
4761 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4762 * acquiring the locks... As quota files are never truncated and quota code
4763 * itself serializes the operations (and no one else should touch the files)
4764 * we don't have to be afraid of races */
4765 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4766 size_t len, loff_t off)
4767 {
4768 struct inode *inode = sb_dqopt(sb)->files[type];
4769 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4770 int err = 0;
4771 int offset = off & (sb->s_blocksize - 1);
4772 int tocopy;
4773 size_t toread;
4774 struct buffer_head *bh;
4775 loff_t i_size = i_size_read(inode);
4776
4777 if (off > i_size)
4778 return 0;
4779 if (off+len > i_size)
4780 len = i_size-off;
4781 toread = len;
4782 while (toread > 0) {
4783 tocopy = sb->s_blocksize - offset < toread ?
4784 sb->s_blocksize - offset : toread;
4785 bh = ext4_bread(NULL, inode, blk, 0, &err);
4786 if (err)
4787 return err;
4788 if (!bh) /* A hole? */
4789 memset(data, 0, tocopy);
4790 else
4791 memcpy(data, bh->b_data+offset, tocopy);
4792 brelse(bh);
4793 offset = 0;
4794 toread -= tocopy;
4795 data += tocopy;
4796 blk++;
4797 }
4798 return len;
4799 }
4800
4801 /* Write to quotafile (we know the transaction is already started and has
4802 * enough credits) */
4803 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4804 const char *data, size_t len, loff_t off)
4805 {
4806 struct inode *inode = sb_dqopt(sb)->files[type];
4807 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4808 int err = 0;
4809 int offset = off & (sb->s_blocksize - 1);
4810 struct buffer_head *bh;
4811 handle_t *handle = journal_current_handle();
4812
4813 if (EXT4_SB(sb)->s_journal && !handle) {
4814 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4815 " cancelled because transaction is not started",
4816 (unsigned long long)off, (unsigned long long)len);
4817 return -EIO;
4818 }
4819 /*
4820 * Since we account only one data block in transaction credits,
4821 * then it is impossible to cross a block boundary.
4822 */
4823 if (sb->s_blocksize - offset < len) {
4824 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4825 " cancelled because not block aligned",
4826 (unsigned long long)off, (unsigned long long)len);
4827 return -EIO;
4828 }
4829
4830 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4831 bh = ext4_bread(handle, inode, blk, 1, &err);
4832 if (!bh)
4833 goto out;
4834 err = ext4_journal_get_write_access(handle, bh);
4835 if (err) {
4836 brelse(bh);
4837 goto out;
4838 }
4839 lock_buffer(bh);
4840 memcpy(bh->b_data+offset, data, len);
4841 flush_dcache_page(bh->b_page);
4842 unlock_buffer(bh);
4843 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4844 brelse(bh);
4845 out:
4846 if (err) {
4847 mutex_unlock(&inode->i_mutex);
4848 return err;
4849 }
4850 if (inode->i_size < off + len) {
4851 i_size_write(inode, off + len);
4852 EXT4_I(inode)->i_disksize = inode->i_size;
4853 ext4_mark_inode_dirty(handle, inode);
4854 }
4855 mutex_unlock(&inode->i_mutex);
4856 return len;
4857 }
4858
4859 #endif
4860
4861 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4862 const char *dev_name, void *data)
4863 {
4864 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4865 }
4866
4867 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4868 static inline void register_as_ext2(void)
4869 {
4870 int err = register_filesystem(&ext2_fs_type);
4871 if (err)
4872 printk(KERN_WARNING
4873 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4874 }
4875
4876 static inline void unregister_as_ext2(void)
4877 {
4878 unregister_filesystem(&ext2_fs_type);
4879 }
4880
4881 static inline int ext2_feature_set_ok(struct super_block *sb)
4882 {
4883 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4884 return 0;
4885 if (sb->s_flags & MS_RDONLY)
4886 return 1;
4887 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4888 return 0;
4889 return 1;
4890 }
4891 MODULE_ALIAS("ext2");
4892 #else
4893 static inline void register_as_ext2(void) { }
4894 static inline void unregister_as_ext2(void) { }
4895 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4896 #endif
4897
4898 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4899 static inline void register_as_ext3(void)
4900 {
4901 int err = register_filesystem(&ext3_fs_type);
4902 if (err)
4903 printk(KERN_WARNING
4904 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4905 }
4906
4907 static inline void unregister_as_ext3(void)
4908 {
4909 unregister_filesystem(&ext3_fs_type);
4910 }
4911
4912 static inline int ext3_feature_set_ok(struct super_block *sb)
4913 {
4914 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4915 return 0;
4916 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4917 return 0;
4918 if (sb->s_flags & MS_RDONLY)
4919 return 1;
4920 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4921 return 0;
4922 return 1;
4923 }
4924 MODULE_ALIAS("ext3");
4925 #else
4926 static inline void register_as_ext3(void) { }
4927 static inline void unregister_as_ext3(void) { }
4928 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
4929 #endif
4930
4931 static struct file_system_type ext4_fs_type = {
4932 .owner = THIS_MODULE,
4933 .name = "ext4",
4934 .mount = ext4_mount,
4935 .kill_sb = kill_block_super,
4936 .fs_flags = FS_REQUIRES_DEV,
4937 };
4938
4939 static int __init ext4_init_feat_adverts(void)
4940 {
4941 struct ext4_features *ef;
4942 int ret = -ENOMEM;
4943
4944 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4945 if (!ef)
4946 goto out;
4947
4948 ef->f_kobj.kset = ext4_kset;
4949 init_completion(&ef->f_kobj_unregister);
4950 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4951 "features");
4952 if (ret) {
4953 kfree(ef);
4954 goto out;
4955 }
4956
4957 ext4_feat = ef;
4958 ret = 0;
4959 out:
4960 return ret;
4961 }
4962
4963 static void ext4_exit_feat_adverts(void)
4964 {
4965 kobject_put(&ext4_feat->f_kobj);
4966 wait_for_completion(&ext4_feat->f_kobj_unregister);
4967 kfree(ext4_feat);
4968 }
4969
4970 /* Shared across all ext4 file systems */
4971 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4972 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4973
4974 static int __init ext4_init_fs(void)
4975 {
4976 int i, err;
4977
4978 ext4_check_flag_values();
4979
4980 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4981 mutex_init(&ext4__aio_mutex[i]);
4982 init_waitqueue_head(&ext4__ioend_wq[i]);
4983 }
4984
4985 err = ext4_init_pageio();
4986 if (err)
4987 return err;
4988 err = ext4_init_system_zone();
4989 if (err)
4990 goto out7;
4991 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4992 if (!ext4_kset)
4993 goto out6;
4994 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4995 if (!ext4_proc_root)
4996 goto out5;
4997
4998 err = ext4_init_feat_adverts();
4999 if (err)
5000 goto out4;
5001
5002 err = ext4_init_mballoc();
5003 if (err)
5004 goto out3;
5005
5006 err = ext4_init_xattr();
5007 if (err)
5008 goto out2;
5009 err = init_inodecache();
5010 if (err)
5011 goto out1;
5012 register_as_ext3();
5013 register_as_ext2();
5014 err = register_filesystem(&ext4_fs_type);
5015 if (err)
5016 goto out;
5017
5018 ext4_li_info = NULL;
5019 mutex_init(&ext4_li_mtx);
5020 return 0;
5021 out:
5022 unregister_as_ext2();
5023 unregister_as_ext3();
5024 destroy_inodecache();
5025 out1:
5026 ext4_exit_xattr();
5027 out2:
5028 ext4_exit_mballoc();
5029 out3:
5030 ext4_exit_feat_adverts();
5031 out4:
5032 remove_proc_entry("fs/ext4", NULL);
5033 out5:
5034 kset_unregister(ext4_kset);
5035 out6:
5036 ext4_exit_system_zone();
5037 out7:
5038 ext4_exit_pageio();
5039 return err;
5040 }
5041
5042 static void __exit ext4_exit_fs(void)
5043 {
5044 ext4_destroy_lazyinit_thread();
5045 unregister_as_ext2();
5046 unregister_as_ext3();
5047 unregister_filesystem(&ext4_fs_type);
5048 destroy_inodecache();
5049 ext4_exit_xattr();
5050 ext4_exit_mballoc();
5051 ext4_exit_feat_adverts();
5052 remove_proc_entry("fs/ext4", NULL);
5053 kset_unregister(ext4_kset);
5054 ext4_exit_system_zone();
5055 ext4_exit_pageio();
5056 }
5057
5058 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5059 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5060 MODULE_LICENSE("GPL");
5061 module_init(ext4_init_fs)
5062 module_exit(ext4_exit_fs)
linux-next