?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ARM: 7220/1: mmc: mmci: Fixup error handling for dma
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blob065ff37aa3b0724996865f04a8b6ec9395307f5d
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_itable");
1144 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
1145 seq_printf(seq, ",init_itable=%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, Opt_init_itable, Opt_noinit_itable,
1322 };
1323
1324 static const match_table_t tokens = {
1325 {Opt_bsd_df, "bsddf"},
1326 {Opt_minix_df, "minixdf"},
1327 {Opt_grpid, "grpid"},
1328 {Opt_grpid, "bsdgroups"},
1329 {Opt_nogrpid, "nogrpid"},
1330 {Opt_nogrpid, "sysvgroups"},
1331 {Opt_resgid, "resgid=%u"},
1332 {Opt_resuid, "resuid=%u"},
1333 {Opt_sb, "sb=%u"},
1334 {Opt_err_cont, "errors=continue"},
1335 {Opt_err_panic, "errors=panic"},
1336 {Opt_err_ro, "errors=remount-ro"},
1337 {Opt_nouid32, "nouid32"},
1338 {Opt_debug, "debug"},
1339 {Opt_oldalloc, "oldalloc"},
1340 {Opt_orlov, "orlov"},
1341 {Opt_user_xattr, "user_xattr"},
1342 {Opt_nouser_xattr, "nouser_xattr"},
1343 {Opt_acl, "acl"},
1344 {Opt_noacl, "noacl"},
1345 {Opt_noload, "noload"},
1346 {Opt_noload, "norecovery"},
1347 {Opt_nobh, "nobh"},
1348 {Opt_bh, "bh"},
1349 {Opt_commit, "commit=%u"},
1350 {Opt_min_batch_time, "min_batch_time=%u"},
1351 {Opt_max_batch_time, "max_batch_time=%u"},
1352 {Opt_journal_update, "journal=update"},
1353 {Opt_journal_dev, "journal_dev=%u"},
1354 {Opt_journal_checksum, "journal_checksum"},
1355 {Opt_journal_async_commit, "journal_async_commit"},
1356 {Opt_abort, "abort"},
1357 {Opt_data_journal, "data=journal"},
1358 {Opt_data_ordered, "data=ordered"},
1359 {Opt_data_writeback, "data=writeback"},
1360 {Opt_data_err_abort, "data_err=abort"},
1361 {Opt_data_err_ignore, "data_err=ignore"},
1362 {Opt_offusrjquota, "usrjquota="},
1363 {Opt_usrjquota, "usrjquota=%s"},
1364 {Opt_offgrpjquota, "grpjquota="},
1365 {Opt_grpjquota, "grpjquota=%s"},
1366 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1367 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1368 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1369 {Opt_grpquota, "grpquota"},
1370 {Opt_noquota, "noquota"},
1371 {Opt_quota, "quota"},
1372 {Opt_usrquota, "usrquota"},
1373 {Opt_barrier, "barrier=%u"},
1374 {Opt_barrier, "barrier"},
1375 {Opt_nobarrier, "nobarrier"},
1376 {Opt_i_version, "i_version"},
1377 {Opt_stripe, "stripe=%u"},
1378 {Opt_resize, "resize"},
1379 {Opt_delalloc, "delalloc"},
1380 {Opt_nodelalloc, "nodelalloc"},
1381 {Opt_mblk_io_submit, "mblk_io_submit"},
1382 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1383 {Opt_block_validity, "block_validity"},
1384 {Opt_noblock_validity, "noblock_validity"},
1385 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1386 {Opt_journal_ioprio, "journal_ioprio=%u"},
1387 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1388 {Opt_auto_da_alloc, "auto_da_alloc"},
1389 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1390 {Opt_dioread_nolock, "dioread_nolock"},
1391 {Opt_dioread_lock, "dioread_lock"},
1392 {Opt_discard, "discard"},
1393 {Opt_nodiscard, "nodiscard"},
1394 {Opt_init_itable, "init_itable=%u"},
1395 {Opt_init_itable, "init_itable"},
1396 {Opt_noinit_itable, "noinit_itable"},
1397 {Opt_err, NULL},
1398 };
1399
1400 static ext4_fsblk_t get_sb_block(void **data)
1401 {
1402 ext4_fsblk_t sb_block;
1403 char *options = (char *) *data;
1404
1405 if (!options || strncmp(options, "sb=", 3) != 0)
1406 return 1; /* Default location */
1407
1408 options += 3;
1409 /* TODO: use simple_strtoll with >32bit ext4 */
1410 sb_block = simple_strtoul(options, &options, 0);
1411 if (*options && *options != ',') {
1412 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1413 (char *) *data);
1414 return 1;
1415 }
1416 if (*options == ',')
1417 options++;
1418 *data = (void *) options;
1419
1420 return sb_block;
1421 }
1422
1423 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1424 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1425 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1426
1427 #ifdef CONFIG_QUOTA
1428 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1429 {
1430 struct ext4_sb_info *sbi = EXT4_SB(sb);
1431 char *qname;
1432
1433 if (sb_any_quota_loaded(sb) &&
1434 !sbi->s_qf_names[qtype]) {
1435 ext4_msg(sb, KERN_ERR,
1436 "Cannot change journaled "
1437 "quota options when quota turned on");
1438 return 0;
1439 }
1440 qname = match_strdup(args);
1441 if (!qname) {
1442 ext4_msg(sb, KERN_ERR,
1443 "Not enough memory for storing quotafile name");
1444 return 0;
1445 }
1446 if (sbi->s_qf_names[qtype] &&
1447 strcmp(sbi->s_qf_names[qtype], qname)) {
1448 ext4_msg(sb, KERN_ERR,
1449 "%s quota file already specified", QTYPE2NAME(qtype));
1450 kfree(qname);
1451 return 0;
1452 }
1453 sbi->s_qf_names[qtype] = qname;
1454 if (strchr(sbi->s_qf_names[qtype], '/')) {
1455 ext4_msg(sb, KERN_ERR,
1456 "quotafile must be on filesystem root");
1457 kfree(sbi->s_qf_names[qtype]);
1458 sbi->s_qf_names[qtype] = NULL;
1459 return 0;
1460 }
1461 set_opt(sb, QUOTA);
1462 return 1;
1463 }
1464
1465 static int clear_qf_name(struct super_block *sb, int qtype)
1466 {
1467
1468 struct ext4_sb_info *sbi = EXT4_SB(sb);
1469
1470 if (sb_any_quota_loaded(sb) &&
1471 sbi->s_qf_names[qtype]) {
1472 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1473 " when quota turned on");
1474 return 0;
1475 }
1476 /*
1477 * The space will be released later when all options are confirmed
1478 * to be correct
1479 */
1480 sbi->s_qf_names[qtype] = NULL;
1481 return 1;
1482 }
1483 #endif
1484
1485 static int parse_options(char *options, struct super_block *sb,
1486 unsigned long *journal_devnum,
1487 unsigned int *journal_ioprio,
1488 ext4_fsblk_t *n_blocks_count, int is_remount)
1489 {
1490 struct ext4_sb_info *sbi = EXT4_SB(sb);
1491 char *p;
1492 substring_t args[MAX_OPT_ARGS];
1493 int data_opt = 0;
1494 int option;
1495 #ifdef CONFIG_QUOTA
1496 int qfmt;
1497 #endif
1498
1499 if (!options)
1500 return 1;
1501
1502 while ((p = strsep(&options, ",")) != NULL) {
1503 int token;
1504 if (!*p)
1505 continue;
1506
1507 /*
1508 * Initialize args struct so we know whether arg was
1509 * found; some options take optional arguments.
1510 */
1511 args[0].to = args[0].from = NULL;
1512 token = match_token(p, tokens, args);
1513 switch (token) {
1514 case Opt_bsd_df:
1515 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1516 clear_opt(sb, MINIX_DF);
1517 break;
1518 case Opt_minix_df:
1519 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1520 set_opt(sb, MINIX_DF);
1521
1522 break;
1523 case Opt_grpid:
1524 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1525 set_opt(sb, GRPID);
1526
1527 break;
1528 case Opt_nogrpid:
1529 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
1530 clear_opt(sb, GRPID);
1531
1532 break;
1533 case Opt_resuid:
1534 if (match_int(&args[0], &option))
1535 return 0;
1536 sbi->s_resuid = option;
1537 break;
1538 case Opt_resgid:
1539 if (match_int(&args[0], &option))
1540 return 0;
1541 sbi->s_resgid = option;
1542 break;
1543 case Opt_sb:
1544 /* handled by get_sb_block() instead of here */
1545 /* *sb_block = match_int(&args[0]); */
1546 break;
1547 case Opt_err_panic:
1548 clear_opt(sb, ERRORS_CONT);
1549 clear_opt(sb, ERRORS_RO);
1550 set_opt(sb, ERRORS_PANIC);
1551 break;
1552 case Opt_err_ro:
1553 clear_opt(sb, ERRORS_CONT);
1554 clear_opt(sb, ERRORS_PANIC);
1555 set_opt(sb, ERRORS_RO);
1556 break;
1557 case Opt_err_cont:
1558 clear_opt(sb, ERRORS_RO);
1559 clear_opt(sb, ERRORS_PANIC);
1560 set_opt(sb, ERRORS_CONT);
1561 break;
1562 case Opt_nouid32:
1563 set_opt(sb, NO_UID32);
1564 break;
1565 case Opt_debug:
1566 set_opt(sb, DEBUG);
1567 break;
1568 case Opt_oldalloc:
1569 set_opt(sb, OLDALLOC);
1570 break;
1571 case Opt_orlov:
1572 clear_opt(sb, OLDALLOC);
1573 break;
1574 #ifdef CONFIG_EXT4_FS_XATTR
1575 case Opt_user_xattr:
1576 set_opt(sb, XATTR_USER);
1577 break;
1578 case Opt_nouser_xattr:
1579 clear_opt(sb, XATTR_USER);
1580 break;
1581 #else
1582 case Opt_user_xattr:
1583 case Opt_nouser_xattr:
1584 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
1585 break;
1586 #endif
1587 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1588 case Opt_acl:
1589 set_opt(sb, POSIX_ACL);
1590 break;
1591 case Opt_noacl:
1592 clear_opt(sb, POSIX_ACL);
1593 break;
1594 #else
1595 case Opt_acl:
1596 case Opt_noacl:
1597 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
1598 break;
1599 #endif
1600 case Opt_journal_update:
1601 /* @@@ FIXME */
1602 /* Eventually we will want to be able to create
1603 a journal file here. For now, only allow the
1604 user to specify an existing inode to be the
1605 journal file. */
1606 if (is_remount) {
1607 ext4_msg(sb, KERN_ERR,
1608 "Cannot specify journal on remount");
1609 return 0;
1610 }
1611 set_opt(sb, UPDATE_JOURNAL);
1612 break;
1613 case Opt_journal_dev:
1614 if (is_remount) {
1615 ext4_msg(sb, KERN_ERR,
1616 "Cannot specify journal on remount");
1617 return 0;
1618 }
1619 if (match_int(&args[0], &option))
1620 return 0;
1621 *journal_devnum = option;
1622 break;
1623 case Opt_journal_checksum:
1624 set_opt(sb, JOURNAL_CHECKSUM);
1625 break;
1626 case Opt_journal_async_commit:
1627 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1628 set_opt(sb, JOURNAL_CHECKSUM);
1629 break;
1630 case Opt_noload:
1631 set_opt(sb, NOLOAD);
1632 break;
1633 case Opt_commit:
1634 if (match_int(&args[0], &option))
1635 return 0;
1636 if (option < 0)
1637 return 0;
1638 if (option == 0)
1639 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1640 sbi->s_commit_interval = HZ * option;
1641 break;
1642 case Opt_max_batch_time:
1643 if (match_int(&args[0], &option))
1644 return 0;
1645 if (option < 0)
1646 return 0;
1647 if (option == 0)
1648 option = EXT4_DEF_MAX_BATCH_TIME;
1649 sbi->s_max_batch_time = option;
1650 break;
1651 case Opt_min_batch_time:
1652 if (match_int(&args[0], &option))
1653 return 0;
1654 if (option < 0)
1655 return 0;
1656 sbi->s_min_batch_time = option;
1657 break;
1658 case Opt_data_journal:
1659 data_opt = EXT4_MOUNT_JOURNAL_DATA;
1660 goto datacheck;
1661 case Opt_data_ordered:
1662 data_opt = EXT4_MOUNT_ORDERED_DATA;
1663 goto datacheck;
1664 case Opt_data_writeback:
1665 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
1666 datacheck:
1667 if (is_remount) {
1668 if (test_opt(sb, DATA_FLAGS) != data_opt) {
1669 ext4_msg(sb, KERN_ERR,
1670 "Cannot change data mode on remount");
1671 return 0;
1672 }
1673 } else {
1674 clear_opt(sb, DATA_FLAGS);
1675 sbi->s_mount_opt |= data_opt;
1676 }
1677 break;
1678 case Opt_data_err_abort:
1679 set_opt(sb, DATA_ERR_ABORT);
1680 break;
1681 case Opt_data_err_ignore:
1682 clear_opt(sb, DATA_ERR_ABORT);
1683 break;
1684 #ifdef CONFIG_QUOTA
1685 case Opt_usrjquota:
1686 if (!set_qf_name(sb, USRQUOTA, &args[0]))
1687 return 0;
1688 break;
1689 case Opt_grpjquota:
1690 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
1691 return 0;
1692 break;
1693 case Opt_offusrjquota:
1694 if (!clear_qf_name(sb, USRQUOTA))
1695 return 0;
1696 break;
1697 case Opt_offgrpjquota:
1698 if (!clear_qf_name(sb, GRPQUOTA))
1699 return 0;
1700 break;
1701
1702 case Opt_jqfmt_vfsold:
1703 qfmt = QFMT_VFS_OLD;
1704 goto set_qf_format;
1705 case Opt_jqfmt_vfsv0:
1706 qfmt = QFMT_VFS_V0;
1707 goto set_qf_format;
1708 case Opt_jqfmt_vfsv1:
1709 qfmt = QFMT_VFS_V1;
1710 set_qf_format:
1711 if (sb_any_quota_loaded(sb) &&
1712 sbi->s_jquota_fmt != qfmt) {
1713 ext4_msg(sb, KERN_ERR, "Cannot change "
1714 "journaled quota options when "
1715 "quota turned on");
1716 return 0;
1717 }
1718 sbi->s_jquota_fmt = qfmt;
1719 break;
1720 case Opt_quota:
1721 case Opt_usrquota:
1722 set_opt(sb, QUOTA);
1723 set_opt(sb, USRQUOTA);
1724 break;
1725 case Opt_grpquota:
1726 set_opt(sb, QUOTA);
1727 set_opt(sb, GRPQUOTA);
1728 break;
1729 case Opt_noquota:
1730 if (sb_any_quota_loaded(sb)) {
1731 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1732 "options when quota turned on");
1733 return 0;
1734 }
1735 clear_opt(sb, QUOTA);
1736 clear_opt(sb, USRQUOTA);
1737 clear_opt(sb, GRPQUOTA);
1738 break;
1739 #else
1740 case Opt_quota:
1741 case Opt_usrquota:
1742 case Opt_grpquota:
1743 ext4_msg(sb, KERN_ERR,
1744 "quota options not supported");
1745 break;
1746 case Opt_usrjquota:
1747 case Opt_grpjquota:
1748 case Opt_offusrjquota:
1749 case Opt_offgrpjquota:
1750 case Opt_jqfmt_vfsold:
1751 case Opt_jqfmt_vfsv0:
1752 case Opt_jqfmt_vfsv1:
1753 ext4_msg(sb, KERN_ERR,
1754 "journaled quota options not supported");
1755 break;
1756 case Opt_noquota:
1757 break;
1758 #endif
1759 case Opt_abort:
1760 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1761 break;
1762 case Opt_nobarrier:
1763 clear_opt(sb, BARRIER);
1764 break;
1765 case Opt_barrier:
1766 if (args[0].from) {
1767 if (match_int(&args[0], &option))
1768 return 0;
1769 } else
1770 option = 1; /* No argument, default to 1 */
1771 if (option)
1772 set_opt(sb, BARRIER);
1773 else
1774 clear_opt(sb, BARRIER);
1775 break;
1776 case Opt_ignore:
1777 break;
1778 case Opt_resize:
1779 if (!is_remount) {
1780 ext4_msg(sb, KERN_ERR,
1781 "resize option only available "
1782 "for remount");
1783 return 0;
1784 }
1785 if (match_int(&args[0], &option) != 0)
1786 return 0;
1787 *n_blocks_count = option;
1788 break;
1789 case Opt_nobh:
1790 ext4_msg(sb, KERN_WARNING,
1791 "Ignoring deprecated nobh option");
1792 break;
1793 case Opt_bh:
1794 ext4_msg(sb, KERN_WARNING,
1795 "Ignoring deprecated bh option");
1796 break;
1797 case Opt_i_version:
1798 set_opt(sb, I_VERSION);
1799 sb->s_flags |= MS_I_VERSION;
1800 break;
1801 case Opt_nodelalloc:
1802 clear_opt(sb, DELALLOC);
1803 break;
1804 case Opt_mblk_io_submit:
1805 set_opt(sb, MBLK_IO_SUBMIT);
1806 break;
1807 case Opt_nomblk_io_submit:
1808 clear_opt(sb, MBLK_IO_SUBMIT);
1809 break;
1810 case Opt_stripe:
1811 if (match_int(&args[0], &option))
1812 return 0;
1813 if (option < 0)
1814 return 0;
1815 sbi->s_stripe = option;
1816 break;
1817 case Opt_delalloc:
1818 set_opt(sb, DELALLOC);
1819 break;
1820 case Opt_block_validity:
1821 set_opt(sb, BLOCK_VALIDITY);
1822 break;
1823 case Opt_noblock_validity:
1824 clear_opt(sb, BLOCK_VALIDITY);
1825 break;
1826 case Opt_inode_readahead_blks:
1827 if (match_int(&args[0], &option))
1828 return 0;
1829 if (option < 0 || option > (1 << 30))
1830 return 0;
1831 if (option && !is_power_of_2(option)) {
1832 ext4_msg(sb, KERN_ERR,
1833 "EXT4-fs: inode_readahead_blks"
1834 " must be a power of 2");
1835 return 0;
1836 }
1837 sbi->s_inode_readahead_blks = option;
1838 break;
1839 case Opt_journal_ioprio:
1840 if (match_int(&args[0], &option))
1841 return 0;
1842 if (option < 0 || option > 7)
1843 break;
1844 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1845 option);
1846 break;
1847 case Opt_noauto_da_alloc:
1848 set_opt(sb, NO_AUTO_DA_ALLOC);
1849 break;
1850 case Opt_auto_da_alloc:
1851 if (args[0].from) {
1852 if (match_int(&args[0], &option))
1853 return 0;
1854 } else
1855 option = 1; /* No argument, default to 1 */
1856 if (option)
1857 clear_opt(sb, NO_AUTO_DA_ALLOC);
1858 else
1859 set_opt(sb,NO_AUTO_DA_ALLOC);
1860 break;
1861 case Opt_discard:
1862 set_opt(sb, DISCARD);
1863 break;
1864 case Opt_nodiscard:
1865 clear_opt(sb, DISCARD);
1866 break;
1867 case Opt_dioread_nolock:
1868 set_opt(sb, DIOREAD_NOLOCK);
1869 break;
1870 case Opt_dioread_lock:
1871 clear_opt(sb, DIOREAD_NOLOCK);
1872 break;
1873 case Opt_init_itable:
1874 set_opt(sb, INIT_INODE_TABLE);
1875 if (args[0].from) {
1876 if (match_int(&args[0], &option))
1877 return 0;
1878 } else
1879 option = EXT4_DEF_LI_WAIT_MULT;
1880 if (option < 0)
1881 return 0;
1882 sbi->s_li_wait_mult = option;
1883 break;
1884 case Opt_noinit_itable:
1885 clear_opt(sb, INIT_INODE_TABLE);
1886 break;
1887 default:
1888 ext4_msg(sb, KERN_ERR,
1889 "Unrecognized mount option \"%s\" "
1890 "or missing value", p);
1891 return 0;
1892 }
1893 }
1894 #ifdef CONFIG_QUOTA
1895 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1896 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1897 clear_opt(sb, USRQUOTA);
1898
1899 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1900 clear_opt(sb, GRPQUOTA);
1901
1902 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1903 ext4_msg(sb, KERN_ERR, "old and new quota "
1904 "format mixing");
1905 return 0;
1906 }
1907
1908 if (!sbi->s_jquota_fmt) {
1909 ext4_msg(sb, KERN_ERR, "journaled quota format "
1910 "not specified");
1911 return 0;
1912 }
1913 } else {
1914 if (sbi->s_jquota_fmt) {
1915 ext4_msg(sb, KERN_ERR, "journaled quota format "
1916 "specified with no journaling "
1917 "enabled");
1918 return 0;
1919 }
1920 }
1921 #endif
1922 return 1;
1923 }
1924
1925 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1926 int read_only)
1927 {
1928 struct ext4_sb_info *sbi = EXT4_SB(sb);
1929 int res = 0;
1930
1931 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1932 ext4_msg(sb, KERN_ERR, "revision level too high, "
1933 "forcing read-only mode");
1934 res = MS_RDONLY;
1935 }
1936 if (read_only)
1937 return res;
1938 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1939 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1940 "running e2fsck is recommended");
1941 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1942 ext4_msg(sb, KERN_WARNING,
1943 "warning: mounting fs with errors, "
1944 "running e2fsck is recommended");
1945 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1946 le16_to_cpu(es->s_mnt_count) >=
1947 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1948 ext4_msg(sb, KERN_WARNING,
1949 "warning: maximal mount count reached, "
1950 "running e2fsck is recommended");
1951 else if (le32_to_cpu(es->s_checkinterval) &&
1952 (le32_to_cpu(es->s_lastcheck) +
1953 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1954 ext4_msg(sb, KERN_WARNING,
1955 "warning: checktime reached, "
1956 "running e2fsck is recommended");
1957 if (!sbi->s_journal)
1958 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1959 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1960 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1961 le16_add_cpu(&es->s_mnt_count, 1);
1962 es->s_mtime = cpu_to_le32(get_seconds());
1963 ext4_update_dynamic_rev(sb);
1964 if (sbi->s_journal)
1965 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1966
1967 ext4_commit_super(sb, 1);
1968 if (test_opt(sb, DEBUG))
1969 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1970 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1971 sb->s_blocksize,
1972 sbi->s_groups_count,
1973 EXT4_BLOCKS_PER_GROUP(sb),
1974 EXT4_INODES_PER_GROUP(sb),
1975 sbi->s_mount_opt, sbi->s_mount_opt2);
1976
1977 cleancache_init_fs(sb);
1978 return res;
1979 }
1980
1981 static int ext4_fill_flex_info(struct super_block *sb)
1982 {
1983 struct ext4_sb_info *sbi = EXT4_SB(sb);
1984 struct ext4_group_desc *gdp = NULL;
1985 ext4_group_t flex_group_count;
1986 ext4_group_t flex_group;
1987 int groups_per_flex = 0;
1988 size_t size;
1989 int i;
1990
1991 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1992 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1993
1994 if (groups_per_flex < 2) {
1995 sbi->s_log_groups_per_flex = 0;
1996 return 1;
1997 }
1998
1999 /* We allocate both existing and potentially added groups */
2000 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
2001 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2002 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
2003 size = flex_group_count * sizeof(struct flex_groups);
2004 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
2005 if (sbi->s_flex_groups == NULL) {
2006 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2007 flex_group_count);
2008 goto failed;
2009 }
2010
2011 for (i = 0; i < sbi->s_groups_count; i++) {
2012 gdp = ext4_get_group_desc(sb, i, NULL);
2013
2014 flex_group = ext4_flex_group(sbi, i);
2015 atomic_add(ext4_free_inodes_count(sb, gdp),
2016 &sbi->s_flex_groups[flex_group].free_inodes);
2017 atomic_add(ext4_free_blks_count(sb, gdp),
2018 &sbi->s_flex_groups[flex_group].free_blocks);
2019 atomic_add(ext4_used_dirs_count(sb, gdp),
2020 &sbi->s_flex_groups[flex_group].used_dirs);
2021 }
2022
2023 return 1;
2024 failed:
2025 return 0;
2026 }
2027
2028 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2029 struct ext4_group_desc *gdp)
2030 {
2031 __u16 crc = 0;
2032
2033 if (sbi->s_es->s_feature_ro_compat &
2034 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2035 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2036 __le32 le_group = cpu_to_le32(block_group);
2037
2038 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2039 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2040 crc = crc16(crc, (__u8 *)gdp, offset);
2041 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2042 /* for checksum of struct ext4_group_desc do the rest...*/
2043 if ((sbi->s_es->s_feature_incompat &
2044 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2045 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2046 crc = crc16(crc, (__u8 *)gdp + offset,
2047 le16_to_cpu(sbi->s_es->s_desc_size) -
2048 offset);
2049 }
2050
2051 return cpu_to_le16(crc);
2052 }
2053
2054 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2055 struct ext4_group_desc *gdp)
2056 {
2057 if ((sbi->s_es->s_feature_ro_compat &
2058 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2059 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2060 return 0;
2061
2062 return 1;
2063 }
2064
2065 /* Called at mount-time, super-block is locked */
2066 static int ext4_check_descriptors(struct super_block *sb,
2067 ext4_group_t *first_not_zeroed)
2068 {
2069 struct ext4_sb_info *sbi = EXT4_SB(sb);
2070 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2071 ext4_fsblk_t last_block;
2072 ext4_fsblk_t block_bitmap;
2073 ext4_fsblk_t inode_bitmap;
2074 ext4_fsblk_t inode_table;
2075 int flexbg_flag = 0;
2076 ext4_group_t i, grp = sbi->s_groups_count;
2077
2078 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2079 flexbg_flag = 1;
2080
2081 ext4_debug("Checking group descriptors");
2082
2083 for (i = 0; i < sbi->s_groups_count; i++) {
2084 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2085
2086 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2087 last_block = ext4_blocks_count(sbi->s_es) - 1;
2088 else
2089 last_block = first_block +
2090 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2091
2092 if ((grp == sbi->s_groups_count) &&
2093 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2094 grp = i;
2095
2096 block_bitmap = ext4_block_bitmap(sb, gdp);
2097 if (block_bitmap < first_block || block_bitmap > last_block) {
2098 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2099 "Block bitmap for group %u not in group "
2100 "(block %llu)!", i, block_bitmap);
2101 return 0;
2102 }
2103 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2104 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2105 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2106 "Inode bitmap for group %u not in group "
2107 "(block %llu)!", i, inode_bitmap);
2108 return 0;
2109 }
2110 inode_table = ext4_inode_table(sb, gdp);
2111 if (inode_table < first_block ||
2112 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2113 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2114 "Inode table for group %u not in group "
2115 "(block %llu)!", i, inode_table);
2116 return 0;
2117 }
2118 ext4_lock_group(sb, i);
2119 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2120 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2121 "Checksum for group %u failed (%u!=%u)",
2122 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2123 gdp)), le16_to_cpu(gdp->bg_checksum));
2124 if (!(sb->s_flags & MS_RDONLY)) {
2125 ext4_unlock_group(sb, i);
2126 return 0;
2127 }
2128 }
2129 ext4_unlock_group(sb, i);
2130 if (!flexbg_flag)
2131 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2132 }
2133 if (NULL != first_not_zeroed)
2134 *first_not_zeroed = grp;
2135
2136 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2137 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2138 return 1;
2139 }
2140
2141 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2142 * the superblock) which were deleted from all directories, but held open by
2143 * a process at the time of a crash. We walk the list and try to delete these
2144 * inodes at recovery time (only with a read-write filesystem).
2145 *
2146 * In order to keep the orphan inode chain consistent during traversal (in
2147 * case of crash during recovery), we link each inode into the superblock
2148 * orphan list_head and handle it the same way as an inode deletion during
2149 * normal operation (which journals the operations for us).
2150 *
2151 * We only do an iget() and an iput() on each inode, which is very safe if we
2152 * accidentally point at an in-use or already deleted inode. The worst that
2153 * can happen in this case is that we get a "bit already cleared" message from
2154 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2155 * e2fsck was run on this filesystem, and it must have already done the orphan
2156 * inode cleanup for us, so we can safely abort without any further action.
2157 */
2158 static void ext4_orphan_cleanup(struct super_block *sb,
2159 struct ext4_super_block *es)
2160 {
2161 unsigned int s_flags = sb->s_flags;
2162 int nr_orphans = 0, nr_truncates = 0;
2163 #ifdef CONFIG_QUOTA
2164 int i;
2165 #endif
2166 if (!es->s_last_orphan) {
2167 jbd_debug(4, "no orphan inodes to clean up\n");
2168 return;
2169 }
2170
2171 if (bdev_read_only(sb->s_bdev)) {
2172 ext4_msg(sb, KERN_ERR, "write access "
2173 "unavailable, skipping orphan cleanup");
2174 return;
2175 }
2176
2177 /* Check if feature set would not allow a r/w mount */
2178 if (!ext4_feature_set_ok(sb, 0)) {
2179 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2180 "unknown ROCOMPAT features");
2181 return;
2182 }
2183
2184 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2185 if (es->s_last_orphan)
2186 jbd_debug(1, "Errors on filesystem, "
2187 "clearing orphan list.\n");
2188 es->s_last_orphan = 0;
2189 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2190 return;
2191 }
2192
2193 if (s_flags & MS_RDONLY) {
2194 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2195 sb->s_flags &= ~MS_RDONLY;
2196 }
2197 #ifdef CONFIG_QUOTA
2198 /* Needed for iput() to work correctly and not trash data */
2199 sb->s_flags |= MS_ACTIVE;
2200 /* Turn on quotas so that they are updated correctly */
2201 for (i = 0; i < MAXQUOTAS; i++) {
2202 if (EXT4_SB(sb)->s_qf_names[i]) {
2203 int ret = ext4_quota_on_mount(sb, i);
2204 if (ret < 0)
2205 ext4_msg(sb, KERN_ERR,
2206 "Cannot turn on journaled "
2207 "quota: error %d", ret);
2208 }
2209 }
2210 #endif
2211
2212 while (es->s_last_orphan) {
2213 struct inode *inode;
2214
2215 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2216 if (IS_ERR(inode)) {
2217 es->s_last_orphan = 0;
2218 break;
2219 }
2220
2221 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2222 dquot_initialize(inode);
2223 if (inode->i_nlink) {
2224 ext4_msg(sb, KERN_DEBUG,
2225 "%s: truncating inode %lu to %lld bytes",
2226 __func__, inode->i_ino, inode->i_size);
2227 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2228 inode->i_ino, inode->i_size);
2229 ext4_truncate(inode);
2230 nr_truncates++;
2231 } else {
2232 ext4_msg(sb, KERN_DEBUG,
2233 "%s: deleting unreferenced inode %lu",
2234 __func__, inode->i_ino);
2235 jbd_debug(2, "deleting unreferenced inode %lu\n",
2236 inode->i_ino);
2237 nr_orphans++;
2238 }
2239 iput(inode); /* The delete magic happens here! */
2240 }
2241
2242 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2243
2244 if (nr_orphans)
2245 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2246 PLURAL(nr_orphans));
2247 if (nr_truncates)
2248 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2249 PLURAL(nr_truncates));
2250 #ifdef CONFIG_QUOTA
2251 /* Turn quotas off */
2252 for (i = 0; i < MAXQUOTAS; i++) {
2253 if (sb_dqopt(sb)->files[i])
2254 dquot_quota_off(sb, i);
2255 }
2256 #endif
2257 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2258 }
2259
2260 /*
2261 * Maximal extent format file size.
2262 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2263 * extent format containers, within a sector_t, and within i_blocks
2264 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2265 * so that won't be a limiting factor.
2266 *
2267 * However there is other limiting factor. We do store extents in the form
2268 * of starting block and length, hence the resulting length of the extent
2269 * covering maximum file size must fit into on-disk format containers as
2270 * well. Given that length is always by 1 unit bigger than max unit (because
2271 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2272 *
2273 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2274 */
2275 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2276 {
2277 loff_t res;
2278 loff_t upper_limit = MAX_LFS_FILESIZE;
2279
2280 /* small i_blocks in vfs inode? */
2281 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2282 /*
2283 * CONFIG_LBDAF is not enabled implies the inode
2284 * i_block represent total blocks in 512 bytes
2285 * 32 == size of vfs inode i_blocks * 8
2286 */
2287 upper_limit = (1LL << 32) - 1;
2288
2289 /* total blocks in file system block size */
2290 upper_limit >>= (blkbits - 9);
2291 upper_limit <<= blkbits;
2292 }
2293
2294 /*
2295 * 32-bit extent-start container, ee_block. We lower the maxbytes
2296 * by one fs block, so ee_len can cover the extent of maximum file
2297 * size
2298 */
2299 res = (1LL << 32) - 1;
2300 res <<= blkbits;
2301
2302 /* Sanity check against vm- & vfs- imposed limits */
2303 if (res > upper_limit)
2304 res = upper_limit;
2305
2306 return res;
2307 }
2308
2309 /*
2310 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2311 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2312 * We need to be 1 filesystem block less than the 2^48 sector limit.
2313 */
2314 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2315 {
2316 loff_t res = EXT4_NDIR_BLOCKS;
2317 int meta_blocks;
2318 loff_t upper_limit;
2319 /* This is calculated to be the largest file size for a dense, block
2320 * mapped file such that the file's total number of 512-byte sectors,
2321 * including data and all indirect blocks, does not exceed (2^48 - 1).
2322 *
2323 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2324 * number of 512-byte sectors of the file.
2325 */
2326
2327 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2328 /*
2329 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2330 * the inode i_block field represents total file blocks in
2331 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2332 */
2333 upper_limit = (1LL << 32) - 1;
2334
2335 /* total blocks in file system block size */
2336 upper_limit >>= (bits - 9);
2337
2338 } else {
2339 /*
2340 * We use 48 bit ext4_inode i_blocks
2341 * With EXT4_HUGE_FILE_FL set the i_blocks
2342 * represent total number of blocks in
2343 * file system block size
2344 */
2345 upper_limit = (1LL << 48) - 1;
2346
2347 }
2348
2349 /* indirect blocks */
2350 meta_blocks = 1;
2351 /* double indirect blocks */
2352 meta_blocks += 1 + (1LL << (bits-2));
2353 /* tripple indirect blocks */
2354 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2355
2356 upper_limit -= meta_blocks;
2357 upper_limit <<= bits;
2358
2359 res += 1LL << (bits-2);
2360 res += 1LL << (2*(bits-2));
2361 res += 1LL << (3*(bits-2));
2362 res <<= bits;
2363 if (res > upper_limit)
2364 res = upper_limit;
2365
2366 if (res > MAX_LFS_FILESIZE)
2367 res = MAX_LFS_FILESIZE;
2368
2369 return res;
2370 }
2371
2372 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2373 ext4_fsblk_t logical_sb_block, int nr)
2374 {
2375 struct ext4_sb_info *sbi = EXT4_SB(sb);
2376 ext4_group_t bg, first_meta_bg;
2377 int has_super = 0;
2378
2379 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2380
2381 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2382 nr < first_meta_bg)
2383 return logical_sb_block + nr + 1;
2384 bg = sbi->s_desc_per_block * nr;
2385 if (ext4_bg_has_super(sb, bg))
2386 has_super = 1;
2387
2388 return (has_super + ext4_group_first_block_no(sb, bg));
2389 }
2390
2391 /**
2392 * ext4_get_stripe_size: Get the stripe size.
2393 * @sbi: In memory super block info
2394 *
2395 * If we have specified it via mount option, then
2396 * use the mount option value. If the value specified at mount time is
2397 * greater than the blocks per group use the super block value.
2398 * If the super block value is greater than blocks per group return 0.
2399 * Allocator needs it be less than blocks per group.
2400 *
2401 */
2402 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2403 {
2404 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2405 unsigned long stripe_width =
2406 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2407 int ret;
2408
2409 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2410 ret = sbi->s_stripe;
2411 else if (stripe_width <= sbi->s_blocks_per_group)
2412 ret = stripe_width;
2413 else if (stride <= sbi->s_blocks_per_group)
2414 ret = stride;
2415 else
2416 ret = 0;
2417
2418 /*
2419 * If the stripe width is 1, this makes no sense and
2420 * we set it to 0 to turn off stripe handling code.
2421 */
2422 if (ret <= 1)
2423 ret = 0;
2424
2425 return ret;
2426 }
2427
2428 /* sysfs supprt */
2429
2430 struct ext4_attr {
2431 struct attribute attr;
2432 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2433 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2434 const char *, size_t);
2435 int offset;
2436 };
2437
2438 static int parse_strtoul(const char *buf,
2439 unsigned long max, unsigned long *value)
2440 {
2441 char *endp;
2442
2443 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2444 endp = skip_spaces(endp);
2445 if (*endp || *value > max)
2446 return -EINVAL;
2447
2448 return 0;
2449 }
2450
2451 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2452 struct ext4_sb_info *sbi,
2453 char *buf)
2454 {
2455 return snprintf(buf, PAGE_SIZE, "%llu\n",
2456 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2457 }
2458
2459 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2460 struct ext4_sb_info *sbi, char *buf)
2461 {
2462 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2463
2464 if (!sb->s_bdev->bd_part)
2465 return snprintf(buf, PAGE_SIZE, "0\n");
2466 return snprintf(buf, PAGE_SIZE, "%lu\n",
2467 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2468 sbi->s_sectors_written_start) >> 1);
2469 }
2470
2471 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2472 struct ext4_sb_info *sbi, char *buf)
2473 {
2474 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2475
2476 if (!sb->s_bdev->bd_part)
2477 return snprintf(buf, PAGE_SIZE, "0\n");
2478 return snprintf(buf, PAGE_SIZE, "%llu\n",
2479 (unsigned long long)(sbi->s_kbytes_written +
2480 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2481 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2482 }
2483
2484 static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2485 struct ext4_sb_info *sbi, char *buf)
2486 {
2487 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2488 }
2489
2490 static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2491 struct ext4_sb_info *sbi, char *buf)
2492 {
2493 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2494 }
2495
2496 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2497 struct ext4_sb_info *sbi,
2498 const char *buf, size_t count)
2499 {
2500 unsigned long t;
2501
2502 if (parse_strtoul(buf, 0x40000000, &t))
2503 return -EINVAL;
2504
2505 if (t && !is_power_of_2(t))
2506 return -EINVAL;
2507
2508 sbi->s_inode_readahead_blks = t;
2509 return count;
2510 }
2511
2512 static ssize_t sbi_ui_show(struct ext4_attr *a,
2513 struct ext4_sb_info *sbi, char *buf)
2514 {
2515 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2516
2517 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2518 }
2519
2520 static ssize_t sbi_ui_store(struct ext4_attr *a,
2521 struct ext4_sb_info *sbi,
2522 const char *buf, size_t count)
2523 {
2524 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2525 unsigned long t;
2526
2527 if (parse_strtoul(buf, 0xffffffff, &t))
2528 return -EINVAL;
2529 *ui = t;
2530 return count;
2531 }
2532
2533 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2534 static struct ext4_attr ext4_attr_##_name = { \
2535 .attr = {.name = __stringify(_name), .mode = _mode }, \
2536 .show = _show, \
2537 .store = _store, \
2538 .offset = offsetof(struct ext4_sb_info, _elname), \
2539 }
2540 #define EXT4_ATTR(name, mode, show, store) \
2541 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2542
2543 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2544 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2545 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2546 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2547 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2548 #define ATTR_LIST(name) &ext4_attr_##name.attr
2549
2550 EXT4_RO_ATTR(delayed_allocation_blocks);
2551 EXT4_RO_ATTR(session_write_kbytes);
2552 EXT4_RO_ATTR(lifetime_write_kbytes);
2553 EXT4_RO_ATTR(extent_cache_hits);
2554 EXT4_RO_ATTR(extent_cache_misses);
2555 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2556 inode_readahead_blks_store, s_inode_readahead_blks);
2557 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2558 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2559 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2560 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2561 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2562 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2563 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2564 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2565
2566 static struct attribute *ext4_attrs[] = {
2567 ATTR_LIST(delayed_allocation_blocks),
2568 ATTR_LIST(session_write_kbytes),
2569 ATTR_LIST(lifetime_write_kbytes),
2570 ATTR_LIST(extent_cache_hits),
2571 ATTR_LIST(extent_cache_misses),
2572 ATTR_LIST(inode_readahead_blks),
2573 ATTR_LIST(inode_goal),
2574 ATTR_LIST(mb_stats),
2575 ATTR_LIST(mb_max_to_scan),
2576 ATTR_LIST(mb_min_to_scan),
2577 ATTR_LIST(mb_order2_req),
2578 ATTR_LIST(mb_stream_req),
2579 ATTR_LIST(mb_group_prealloc),
2580 ATTR_LIST(max_writeback_mb_bump),
2581 NULL,
2582 };
2583
2584 /* Features this copy of ext4 supports */
2585 EXT4_INFO_ATTR(lazy_itable_init);
2586 EXT4_INFO_ATTR(batched_discard);
2587
2588 static struct attribute *ext4_feat_attrs[] = {
2589 ATTR_LIST(lazy_itable_init),
2590 ATTR_LIST(batched_discard),
2591 NULL,
2592 };
2593
2594 static ssize_t ext4_attr_show(struct kobject *kobj,
2595 struct attribute *attr, char *buf)
2596 {
2597 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2598 s_kobj);
2599 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2600
2601 return a->show ? a->show(a, sbi, buf) : 0;
2602 }
2603
2604 static ssize_t ext4_attr_store(struct kobject *kobj,
2605 struct attribute *attr,
2606 const char *buf, size_t len)
2607 {
2608 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2609 s_kobj);
2610 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2611
2612 return a->store ? a->store(a, sbi, buf, len) : 0;
2613 }
2614
2615 static void ext4_sb_release(struct kobject *kobj)
2616 {
2617 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2618 s_kobj);
2619 complete(&sbi->s_kobj_unregister);
2620 }
2621
2622 static const struct sysfs_ops ext4_attr_ops = {
2623 .show = ext4_attr_show,
2624 .store = ext4_attr_store,
2625 };
2626
2627 static struct kobj_type ext4_ktype = {
2628 .default_attrs = ext4_attrs,
2629 .sysfs_ops = &ext4_attr_ops,
2630 .release = ext4_sb_release,
2631 };
2632
2633 static void ext4_feat_release(struct kobject *kobj)
2634 {
2635 complete(&ext4_feat->f_kobj_unregister);
2636 }
2637
2638 static struct kobj_type ext4_feat_ktype = {
2639 .default_attrs = ext4_feat_attrs,
2640 .sysfs_ops = &ext4_attr_ops,
2641 .release = ext4_feat_release,
2642 };
2643
2644 /*
2645 * Check whether this filesystem can be mounted based on
2646 * the features present and the RDONLY/RDWR mount requested.
2647 * Returns 1 if this filesystem can be mounted as requested,
2648 * 0 if it cannot be.
2649 */
2650 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2651 {
2652 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2653 ext4_msg(sb, KERN_ERR,
2654 "Couldn't mount because of "
2655 "unsupported optional features (%x)",
2656 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2657 ~EXT4_FEATURE_INCOMPAT_SUPP));
2658 return 0;
2659 }
2660
2661 if (readonly)
2662 return 1;
2663
2664 /* Check that feature set is OK for a read-write mount */
2665 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2666 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2667 "unsupported optional features (%x)",
2668 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2669 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2670 return 0;
2671 }
2672 /*
2673 * Large file size enabled file system can only be mounted
2674 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2675 */
2676 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2677 if (sizeof(blkcnt_t) < sizeof(u64)) {
2678 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2679 "cannot be mounted RDWR without "
2680 "CONFIG_LBDAF");
2681 return 0;
2682 }
2683 }
2684 return 1;
2685 }
2686
2687 /*
2688 * This function is called once a day if we have errors logged
2689 * on the file system
2690 */
2691 static void print_daily_error_info(unsigned long arg)
2692 {
2693 struct super_block *sb = (struct super_block *) arg;
2694 struct ext4_sb_info *sbi;
2695 struct ext4_super_block *es;
2696
2697 sbi = EXT4_SB(sb);
2698 es = sbi->s_es;
2699
2700 if (es->s_error_count)
2701 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2702 le32_to_cpu(es->s_error_count));
2703 if (es->s_first_error_time) {
2704 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2705 sb->s_id, le32_to_cpu(es->s_first_error_time),
2706 (int) sizeof(es->s_first_error_func),
2707 es->s_first_error_func,
2708 le32_to_cpu(es->s_first_error_line));
2709 if (es->s_first_error_ino)
2710 printk(": inode %u",
2711 le32_to_cpu(es->s_first_error_ino));
2712 if (es->s_first_error_block)
2713 printk(": block %llu", (unsigned long long)
2714 le64_to_cpu(es->s_first_error_block));
2715 printk("\n");
2716 }
2717 if (es->s_last_error_time) {
2718 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2719 sb->s_id, le32_to_cpu(es->s_last_error_time),
2720 (int) sizeof(es->s_last_error_func),
2721 es->s_last_error_func,
2722 le32_to_cpu(es->s_last_error_line));
2723 if (es->s_last_error_ino)
2724 printk(": inode %u",
2725 le32_to_cpu(es->s_last_error_ino));
2726 if (es->s_last_error_block)
2727 printk(": block %llu", (unsigned long long)
2728 le64_to_cpu(es->s_last_error_block));
2729 printk("\n");
2730 }
2731 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2732 }
2733
2734 /* Find next suitable group and run ext4_init_inode_table */
2735 static int ext4_run_li_request(struct ext4_li_request *elr)
2736 {
2737 struct ext4_group_desc *gdp = NULL;
2738 ext4_group_t group, ngroups;
2739 struct super_block *sb;
2740 unsigned long timeout = 0;
2741 int ret = 0;
2742
2743 sb = elr->lr_super;
2744 ngroups = EXT4_SB(sb)->s_groups_count;
2745
2746 for (group = elr->lr_next_group; group < ngroups; group++) {
2747 gdp = ext4_get_group_desc(sb, group, NULL);
2748 if (!gdp) {
2749 ret = 1;
2750 break;
2751 }
2752
2753 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2754 break;
2755 }
2756
2757 if (group == ngroups)
2758 ret = 1;
2759
2760 if (!ret) {
2761 timeout = jiffies;
2762 ret = ext4_init_inode_table(sb, group,
2763 elr->lr_timeout ? 0 : 1);
2764 if (elr->lr_timeout == 0) {
2765 timeout = (jiffies - timeout) *
2766 elr->lr_sbi->s_li_wait_mult;
2767 elr->lr_timeout = timeout;
2768 }
2769 elr->lr_next_sched = jiffies + elr->lr_timeout;
2770 elr->lr_next_group = group + 1;
2771 }
2772
2773 return ret;
2774 }
2775
2776 /*
2777 * Remove lr_request from the list_request and free the
2778 * request structure. Should be called with li_list_mtx held
2779 */
2780 static void ext4_remove_li_request(struct ext4_li_request *elr)
2781 {
2782 struct ext4_sb_info *sbi;
2783
2784 if (!elr)
2785 return;
2786
2787 sbi = elr->lr_sbi;
2788
2789 list_del(&elr->lr_request);
2790 sbi->s_li_request = NULL;
2791 kfree(elr);
2792 }
2793
2794 static void ext4_unregister_li_request(struct super_block *sb)
2795 {
2796 mutex_lock(&ext4_li_mtx);
2797 if (!ext4_li_info) {
2798 mutex_unlock(&ext4_li_mtx);
2799 return;
2800 }
2801
2802 mutex_lock(&ext4_li_info->li_list_mtx);
2803 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2804 mutex_unlock(&ext4_li_info->li_list_mtx);
2805 mutex_unlock(&ext4_li_mtx);
2806 }
2807
2808 static struct task_struct *ext4_lazyinit_task;
2809
2810 /*
2811 * This is the function where ext4lazyinit thread lives. It walks
2812 * through the request list searching for next scheduled filesystem.
2813 * When such a fs is found, run the lazy initialization request
2814 * (ext4_rn_li_request) and keep track of the time spend in this
2815 * function. Based on that time we compute next schedule time of
2816 * the request. When walking through the list is complete, compute
2817 * next waking time and put itself into sleep.
2818 */
2819 static int ext4_lazyinit_thread(void *arg)
2820 {
2821 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2822 struct list_head *pos, *n;
2823 struct ext4_li_request *elr;
2824 unsigned long next_wakeup, cur;
2825
2826 BUG_ON(NULL == eli);
2827
2828 cont_thread:
2829 while (true) {
2830 next_wakeup = MAX_JIFFY_OFFSET;
2831
2832 mutex_lock(&eli->li_list_mtx);
2833 if (list_empty(&eli->li_request_list)) {
2834 mutex_unlock(&eli->li_list_mtx);
2835 goto exit_thread;
2836 }
2837
2838 list_for_each_safe(pos, n, &eli->li_request_list) {
2839 elr = list_entry(pos, struct ext4_li_request,
2840 lr_request);
2841
2842 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2843 if (ext4_run_li_request(elr) != 0) {
2844 /* error, remove the lazy_init job */
2845 ext4_remove_li_request(elr);
2846 continue;
2847 }
2848 }
2849
2850 if (time_before(elr->lr_next_sched, next_wakeup))
2851 next_wakeup = elr->lr_next_sched;
2852 }
2853 mutex_unlock(&eli->li_list_mtx);
2854
2855 if (freezing(current))
2856 refrigerator();
2857
2858 cur = jiffies;
2859 if ((time_after_eq(cur, next_wakeup)) ||
2860 (MAX_JIFFY_OFFSET == next_wakeup)) {
2861 cond_resched();
2862 continue;
2863 }
2864
2865 schedule_timeout_interruptible(next_wakeup - cur);
2866
2867 if (kthread_should_stop()) {
2868 ext4_clear_request_list();
2869 goto exit_thread;
2870 }
2871 }
2872
2873 exit_thread:
2874 /*
2875 * It looks like the request list is empty, but we need
2876 * to check it under the li_list_mtx lock, to prevent any
2877 * additions into it, and of course we should lock ext4_li_mtx
2878 * to atomically free the list and ext4_li_info, because at
2879 * this point another ext4 filesystem could be registering
2880 * new one.
2881 */
2882 mutex_lock(&ext4_li_mtx);
2883 mutex_lock(&eli->li_list_mtx);
2884 if (!list_empty(&eli->li_request_list)) {
2885 mutex_unlock(&eli->li_list_mtx);
2886 mutex_unlock(&ext4_li_mtx);
2887 goto cont_thread;
2888 }
2889 mutex_unlock(&eli->li_list_mtx);
2890 kfree(ext4_li_info);
2891 ext4_li_info = NULL;
2892 mutex_unlock(&ext4_li_mtx);
2893
2894 return 0;
2895 }
2896
2897 static void ext4_clear_request_list(void)
2898 {
2899 struct list_head *pos, *n;
2900 struct ext4_li_request *elr;
2901
2902 mutex_lock(&ext4_li_info->li_list_mtx);
2903 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2904 elr = list_entry(pos, struct ext4_li_request,
2905 lr_request);
2906 ext4_remove_li_request(elr);
2907 }
2908 mutex_unlock(&ext4_li_info->li_list_mtx);
2909 }
2910
2911 static int ext4_run_lazyinit_thread(void)
2912 {
2913 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2914 ext4_li_info, "ext4lazyinit");
2915 if (IS_ERR(ext4_lazyinit_task)) {
2916 int err = PTR_ERR(ext4_lazyinit_task);
2917 ext4_clear_request_list();
2918 kfree(ext4_li_info);
2919 ext4_li_info = NULL;
2920 printk(KERN_CRIT "EXT4: error %d creating inode table "
2921 "initialization thread\n",
2922 err);
2923 return err;
2924 }
2925 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2926 return 0;
2927 }
2928
2929 /*
2930 * Check whether it make sense to run itable init. thread or not.
2931 * If there is at least one uninitialized inode table, return
2932 * corresponding group number, else the loop goes through all
2933 * groups and return total number of groups.
2934 */
2935 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2936 {
2937 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2938 struct ext4_group_desc *gdp = NULL;
2939
2940 for (group = 0; group < ngroups; group++) {
2941 gdp = ext4_get_group_desc(sb, group, NULL);
2942 if (!gdp)
2943 continue;
2944
2945 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2946 break;
2947 }
2948
2949 return group;
2950 }
2951
2952 static int ext4_li_info_new(void)
2953 {
2954 struct ext4_lazy_init *eli = NULL;
2955
2956 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2957 if (!eli)
2958 return -ENOMEM;
2959
2960 INIT_LIST_HEAD(&eli->li_request_list);
2961 mutex_init(&eli->li_list_mtx);
2962
2963 eli->li_state |= EXT4_LAZYINIT_QUIT;
2964
2965 ext4_li_info = eli;
2966
2967 return 0;
2968 }
2969
2970 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2971 ext4_group_t start)
2972 {
2973 struct ext4_sb_info *sbi = EXT4_SB(sb);
2974 struct ext4_li_request *elr;
2975 unsigned long rnd;
2976
2977 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2978 if (!elr)
2979 return NULL;
2980
2981 elr->lr_super = sb;
2982 elr->lr_sbi = sbi;
2983 elr->lr_next_group = start;
2984
2985 /*
2986 * Randomize first schedule time of the request to
2987 * spread the inode table initialization requests
2988 * better.
2989 */
2990 get_random_bytes(&rnd, sizeof(rnd));
2991 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2992 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2993
2994 return elr;
2995 }
2996
2997 static int ext4_register_li_request(struct super_block *sb,
2998 ext4_group_t first_not_zeroed)
2999 {
3000 struct ext4_sb_info *sbi = EXT4_SB(sb);
3001 struct ext4_li_request *elr;
3002 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3003 int ret = 0;
3004
3005 if (sbi->s_li_request != NULL) {
3006 /*
3007 * Reset timeout so it can be computed again, because
3008 * s_li_wait_mult might have changed.
3009 */
3010 sbi->s_li_request->lr_timeout = 0;
3011 return 0;
3012 }
3013
3014 if (first_not_zeroed == ngroups ||
3015 (sb->s_flags & MS_RDONLY) ||
3016 !test_opt(sb, INIT_INODE_TABLE))
3017 return 0;
3018
3019 elr = ext4_li_request_new(sb, first_not_zeroed);
3020 if (!elr)
3021 return -ENOMEM;
3022
3023 mutex_lock(&ext4_li_mtx);
3024
3025 if (NULL == ext4_li_info) {
3026 ret = ext4_li_info_new();
3027 if (ret)
3028 goto out;
3029 }
3030
3031 mutex_lock(&ext4_li_info->li_list_mtx);
3032 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3033 mutex_unlock(&ext4_li_info->li_list_mtx);
3034
3035 sbi->s_li_request = elr;
3036 /*
3037 * set elr to NULL here since it has been inserted to
3038 * the request_list and the removal and free of it is
3039 * handled by ext4_clear_request_list from now on.
3040 */
3041 elr = NULL;
3042
3043 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3044 ret = ext4_run_lazyinit_thread();
3045 if (ret)
3046 goto out;
3047 }
3048 out:
3049 mutex_unlock(&ext4_li_mtx);
3050 if (ret)
3051 kfree(elr);
3052 return ret;
3053 }
3054
3055 /*
3056 * We do not need to lock anything since this is called on
3057 * module unload.
3058 */
3059 static void ext4_destroy_lazyinit_thread(void)
3060 {
3061 /*
3062 * If thread exited earlier
3063 * there's nothing to be done.
3064 */
3065 if (!ext4_li_info || !ext4_lazyinit_task)
3066 return;
3067
3068 kthread_stop(ext4_lazyinit_task);
3069 }
3070
3071 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3072 __releases(kernel_lock)
3073 __acquires(kernel_lock)
3074 {
3075 char *orig_data = kstrdup(data, GFP_KERNEL);
3076 struct buffer_head *bh;
3077 struct ext4_super_block *es = NULL;
3078 struct ext4_sb_info *sbi;
3079 ext4_fsblk_t block;
3080 ext4_fsblk_t sb_block = get_sb_block(&data);
3081 ext4_fsblk_t logical_sb_block;
3082 unsigned long offset = 0;
3083 unsigned long journal_devnum = 0;
3084 unsigned long def_mount_opts;
3085 struct inode *root;
3086 char *cp;
3087 const char *descr;
3088 int ret = -ENOMEM;
3089 int blocksize;
3090 unsigned int db_count;
3091 unsigned int i;
3092 int needs_recovery, has_huge_files;
3093 __u64 blocks_count;
3094 int err;
3095 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3096 ext4_group_t first_not_zeroed;
3097
3098 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3099 if (!sbi)
3100 goto out_free_orig;
3101
3102 sbi->s_blockgroup_lock =
3103 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3104 if (!sbi->s_blockgroup_lock) {
3105 kfree(sbi);
3106 goto out_free_orig;
3107 }
3108 sb->s_fs_info = sbi;
3109 sbi->s_mount_opt = 0;
3110 sbi->s_resuid = EXT4_DEF_RESUID;
3111 sbi->s_resgid = EXT4_DEF_RESGID;
3112 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3113 sbi->s_sb_block = sb_block;
3114 if (sb->s_bdev->bd_part)
3115 sbi->s_sectors_written_start =
3116 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3117
3118 /* Cleanup superblock name */
3119 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3120 *cp = '!';
3121
3122 ret = -EINVAL;
3123 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3124 if (!blocksize) {
3125 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3126 goto out_fail;
3127 }
3128
3129 /*
3130 * The ext4 superblock will not be buffer aligned for other than 1kB
3131 * block sizes. We need to calculate the offset from buffer start.
3132 */
3133 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3134 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3135 offset = do_div(logical_sb_block, blocksize);
3136 } else {
3137 logical_sb_block = sb_block;
3138 }
3139
3140 if (!(bh = sb_bread(sb, logical_sb_block))) {
3141 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3142 goto out_fail;
3143 }
3144 /*
3145 * Note: s_es must be initialized as soon as possible because
3146 * some ext4 macro-instructions depend on its value
3147 */
3148 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3149 sbi->s_es = es;
3150 sb->s_magic = le16_to_cpu(es->s_magic);
3151 if (sb->s_magic != EXT4_SUPER_MAGIC)
3152 goto cantfind_ext4;
3153 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3154
3155 /* Set defaults before we parse the mount options */
3156 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3157 set_opt(sb, INIT_INODE_TABLE);
3158 if (def_mount_opts & EXT4_DEFM_DEBUG)
3159 set_opt(sb, DEBUG);
3160 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3161 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3162 "2.6.38");
3163 set_opt(sb, GRPID);
3164 }
3165 if (def_mount_opts & EXT4_DEFM_UID16)
3166 set_opt(sb, NO_UID32);
3167 /* xattr user namespace & acls are now defaulted on */
3168 #ifdef CONFIG_EXT4_FS_XATTR
3169 set_opt(sb, XATTR_USER);
3170 #endif
3171 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3172 set_opt(sb, POSIX_ACL);
3173 #endif
3174 set_opt(sb, MBLK_IO_SUBMIT);
3175 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3176 set_opt(sb, JOURNAL_DATA);
3177 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3178 set_opt(sb, ORDERED_DATA);
3179 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3180 set_opt(sb, WRITEBACK_DATA);
3181
3182 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3183 set_opt(sb, ERRORS_PANIC);
3184 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3185 set_opt(sb, ERRORS_CONT);
3186 else
3187 set_opt(sb, ERRORS_RO);
3188 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3189 set_opt(sb, BLOCK_VALIDITY);
3190 if (def_mount_opts & EXT4_DEFM_DISCARD)
3191 set_opt(sb, DISCARD);
3192
3193 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3194 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3195 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3196 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3197 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3198
3199 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3200 set_opt(sb, BARRIER);
3201
3202 /*
3203 * enable delayed allocation by default
3204 * Use -o nodelalloc to turn it off
3205 */
3206 if (!IS_EXT3_SB(sb) &&
3207 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3208 set_opt(sb, DELALLOC);
3209
3210 /*
3211 * set default s_li_wait_mult for lazyinit, for the case there is
3212 * no mount option specified.
3213 */
3214 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3215
3216 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3217 &journal_devnum, &journal_ioprio, NULL, 0)) {
3218 ext4_msg(sb, KERN_WARNING,
3219 "failed to parse options in superblock: %s",
3220 sbi->s_es->s_mount_opts);
3221 }
3222 if (!parse_options((char *) data, sb, &journal_devnum,
3223 &journal_ioprio, NULL, 0))
3224 goto failed_mount;
3225
3226 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3227 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3228
3229 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3230 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3231 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3232 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3233 ext4_msg(sb, KERN_WARNING,
3234 "feature flags set on rev 0 fs, "
3235 "running e2fsck is recommended");
3236
3237 if (IS_EXT2_SB(sb)) {
3238 if (ext2_feature_set_ok(sb))
3239 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3240 "using the ext4 subsystem");
3241 else {
3242 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3243 "to feature incompatibilities");
3244 goto failed_mount;
3245 }
3246 }
3247
3248 if (IS_EXT3_SB(sb)) {
3249 if (ext3_feature_set_ok(sb))
3250 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3251 "using the ext4 subsystem");
3252 else {
3253 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3254 "to feature incompatibilities");
3255 goto failed_mount;
3256 }
3257 }
3258
3259 /*
3260 * Check feature flags regardless of the revision level, since we
3261 * previously didn't change the revision level when setting the flags,
3262 * so there is a chance incompat flags are set on a rev 0 filesystem.
3263 */
3264 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3265 goto failed_mount;
3266
3267 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3268
3269 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3270 blocksize > EXT4_MAX_BLOCK_SIZE) {
3271 ext4_msg(sb, KERN_ERR,
3272 "Unsupported filesystem blocksize %d", blocksize);
3273 goto failed_mount;
3274 }
3275
3276 if (sb->s_blocksize != blocksize) {
3277 /* Validate the filesystem blocksize */
3278 if (!sb_set_blocksize(sb, blocksize)) {
3279 ext4_msg(sb, KERN_ERR, "bad block size %d",
3280 blocksize);
3281 goto failed_mount;
3282 }
3283
3284 brelse(bh);
3285 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3286 offset = do_div(logical_sb_block, blocksize);
3287 bh = sb_bread(sb, logical_sb_block);
3288 if (!bh) {
3289 ext4_msg(sb, KERN_ERR,
3290 "Can't read superblock on 2nd try");
3291 goto failed_mount;
3292 }
3293 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3294 sbi->s_es = es;
3295 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3296 ext4_msg(sb, KERN_ERR,
3297 "Magic mismatch, very weird!");
3298 goto failed_mount;
3299 }
3300 }
3301
3302 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3303 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3304 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3305 has_huge_files);
3306 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3307
3308 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3309 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3310 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3311 } else {
3312 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3313 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3314 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3315 (!is_power_of_2(sbi->s_inode_size)) ||
3316 (sbi->s_inode_size > blocksize)) {
3317 ext4_msg(sb, KERN_ERR,
3318 "unsupported inode size: %d",
3319 sbi->s_inode_size);
3320 goto failed_mount;
3321 }
3322 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3323 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3324 }
3325
3326 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3327 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3328 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3329 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3330 !is_power_of_2(sbi->s_desc_size)) {
3331 ext4_msg(sb, KERN_ERR,
3332 "unsupported descriptor size %lu",
3333 sbi->s_desc_size);
3334 goto failed_mount;
3335 }
3336 } else
3337 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3338
3339 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3340 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3341 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3342 goto cantfind_ext4;
3343
3344 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3345 if (sbi->s_inodes_per_block == 0)
3346 goto cantfind_ext4;
3347 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3348 sbi->s_inodes_per_block;
3349 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3350 sbi->s_sbh = bh;
3351 sbi->s_mount_state = le16_to_cpu(es->s_state);
3352 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3353 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3354
3355 for (i = 0; i < 4; i++)
3356 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3357 sbi->s_def_hash_version = es->s_def_hash_version;
3358 i = le32_to_cpu(es->s_flags);
3359 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3360 sbi->s_hash_unsigned = 3;
3361 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3362 #ifdef __CHAR_UNSIGNED__
3363 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3364 sbi->s_hash_unsigned = 3;
3365 #else
3366 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3367 #endif
3368 sb->s_dirt = 1;
3369 }
3370
3371 if (sbi->s_blocks_per_group > blocksize * 8) {
3372 ext4_msg(sb, KERN_ERR,
3373 "#blocks per group too big: %lu",
3374 sbi->s_blocks_per_group);
3375 goto failed_mount;
3376 }
3377 if (sbi->s_inodes_per_group > blocksize * 8) {
3378 ext4_msg(sb, KERN_ERR,
3379 "#inodes per group too big: %lu",
3380 sbi->s_inodes_per_group);
3381 goto failed_mount;
3382 }
3383
3384 /*
3385 * Test whether we have more sectors than will fit in sector_t,
3386 * and whether the max offset is addressable by the page cache.
3387 */
3388 err = generic_check_addressable(sb->s_blocksize_bits,
3389 ext4_blocks_count(es));
3390 if (err) {
3391 ext4_msg(sb, KERN_ERR, "filesystem"
3392 " too large to mount safely on this system");
3393 if (sizeof(sector_t) < 8)
3394 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3395 ret = err;
3396 goto failed_mount;
3397 }
3398
3399 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3400 goto cantfind_ext4;
3401
3402 /* check blocks count against device size */
3403 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3404 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3405 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3406 "exceeds size of device (%llu blocks)",
3407 ext4_blocks_count(es), blocks_count);
3408 goto failed_mount;
3409 }
3410
3411 /*
3412 * It makes no sense for the first data block to be beyond the end
3413 * of the filesystem.
3414 */
3415 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3416 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3417 "block %u is beyond end of filesystem (%llu)",
3418 le32_to_cpu(es->s_first_data_block),
3419 ext4_blocks_count(es));
3420 goto failed_mount;
3421 }
3422 blocks_count = (ext4_blocks_count(es) -
3423 le32_to_cpu(es->s_first_data_block) +
3424 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3425 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3426 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3427 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3428 "(block count %llu, first data block %u, "
3429 "blocks per group %lu)", sbi->s_groups_count,
3430 ext4_blocks_count(es),
3431 le32_to_cpu(es->s_first_data_block),
3432 EXT4_BLOCKS_PER_GROUP(sb));
3433 goto failed_mount;
3434 }
3435 sbi->s_groups_count = blocks_count;
3436 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3437 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3438 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3439 EXT4_DESC_PER_BLOCK(sb);
3440 sbi->s_group_desc = ext4_kvmalloc(db_count *
3441 sizeof(struct buffer_head *),
3442 GFP_KERNEL);
3443 if (sbi->s_group_desc == NULL) {
3444 ext4_msg(sb, KERN_ERR, "not enough memory");
3445 goto failed_mount;
3446 }
3447
3448 #ifdef CONFIG_PROC_FS
3449 if (ext4_proc_root)
3450 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3451 #endif
3452
3453 bgl_lock_init(sbi->s_blockgroup_lock);
3454
3455 for (i = 0; i < db_count; i++) {
3456 block = descriptor_loc(sb, logical_sb_block, i);
3457 sbi->s_group_desc[i] = sb_bread(sb, block);
3458 if (!sbi->s_group_desc[i]) {
3459 ext4_msg(sb, KERN_ERR,
3460 "can't read group descriptor %d", i);
3461 db_count = i;
3462 goto failed_mount2;
3463 }
3464 }
3465 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3466 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3467 goto failed_mount2;
3468 }
3469 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3470 if (!ext4_fill_flex_info(sb)) {
3471 ext4_msg(sb, KERN_ERR,
3472 "unable to initialize "
3473 "flex_bg meta info!");
3474 goto failed_mount2;
3475 }
3476
3477 sbi->s_gdb_count = db_count;
3478 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3479 spin_lock_init(&sbi->s_next_gen_lock);
3480
3481 init_timer(&sbi->s_err_report);
3482 sbi->s_err_report.function = print_daily_error_info;
3483 sbi->s_err_report.data = (unsigned long) sb;
3484
3485 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3486 ext4_count_free_blocks(sb));
3487 if (!err) {
3488 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3489 ext4_count_free_inodes(sb));
3490 }
3491 if (!err) {
3492 err = percpu_counter_init(&sbi->s_dirs_counter,
3493 ext4_count_dirs(sb));
3494 }
3495 if (!err) {
3496 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3497 }
3498 if (err) {
3499 ext4_msg(sb, KERN_ERR, "insufficient memory");
3500 goto failed_mount3;
3501 }
3502
3503 sbi->s_stripe = ext4_get_stripe_size(sbi);
3504 sbi->s_max_writeback_mb_bump = 128;
3505
3506 /*
3507 * set up enough so that it can read an inode
3508 */
3509 if (!test_opt(sb, NOLOAD) &&
3510 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3511 sb->s_op = &ext4_sops;
3512 else
3513 sb->s_op = &ext4_nojournal_sops;
3514 sb->s_export_op = &ext4_export_ops;
3515 sb->s_xattr = ext4_xattr_handlers;
3516 #ifdef CONFIG_QUOTA
3517 sb->s_qcop = &ext4_qctl_operations;
3518 sb->dq_op = &ext4_quota_operations;
3519 #endif
3520 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3521
3522 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3523 mutex_init(&sbi->s_orphan_lock);
3524 sbi->s_resize_flags = 0;
3525
3526 sb->s_root = NULL;
3527
3528 needs_recovery = (es->s_last_orphan != 0 ||
3529 EXT4_HAS_INCOMPAT_FEATURE(sb,
3530 EXT4_FEATURE_INCOMPAT_RECOVER));
3531
3532 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3533 !(sb->s_flags & MS_RDONLY))
3534 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3535 goto failed_mount3;
3536
3537 /*
3538 * The first inode we look at is the journal inode. Don't try
3539 * root first: it may be modified in the journal!
3540 */
3541 if (!test_opt(sb, NOLOAD) &&
3542 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3543 if (ext4_load_journal(sb, es, journal_devnum))
3544 goto failed_mount3;
3545 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3546 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3547 ext4_msg(sb, KERN_ERR, "required journal recovery "
3548 "suppressed and not mounted read-only");
3549 goto failed_mount_wq;
3550 } else {
3551 clear_opt(sb, DATA_FLAGS);
3552 sbi->s_journal = NULL;
3553 needs_recovery = 0;
3554 goto no_journal;
3555 }
3556
3557 if (ext4_blocks_count(es) > 0xffffffffULL &&
3558 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3559 JBD2_FEATURE_INCOMPAT_64BIT)) {
3560 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3561 goto failed_mount_wq;
3562 }
3563
3564 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3565 jbd2_journal_set_features(sbi->s_journal,
3566 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3567 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3568 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3569 jbd2_journal_set_features(sbi->s_journal,
3570 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3571 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3572 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3573 } else {
3574 jbd2_journal_clear_features(sbi->s_journal,
3575 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3576 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3577 }
3578
3579 /* We have now updated the journal if required, so we can
3580 * validate the data journaling mode. */
3581 switch (test_opt(sb, DATA_FLAGS)) {
3582 case 0:
3583 /* No mode set, assume a default based on the journal
3584 * capabilities: ORDERED_DATA if the journal can
3585 * cope, else JOURNAL_DATA
3586 */
3587 if (jbd2_journal_check_available_features
3588 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3589 set_opt(sb, ORDERED_DATA);
3590 else
3591 set_opt(sb, JOURNAL_DATA);
3592 break;
3593
3594 case EXT4_MOUNT_ORDERED_DATA:
3595 case EXT4_MOUNT_WRITEBACK_DATA:
3596 if (!jbd2_journal_check_available_features
3597 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3598 ext4_msg(sb, KERN_ERR, "Journal does not support "
3599 "requested data journaling mode");
3600 goto failed_mount_wq;
3601 }
3602 default:
3603 break;
3604 }
3605 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3606
3607 /*
3608 * The journal may have updated the bg summary counts, so we
3609 * need to update the global counters.
3610 */
3611 percpu_counter_set(&sbi->s_freeblocks_counter,
3612 ext4_count_free_blocks(sb));
3613 percpu_counter_set(&sbi->s_freeinodes_counter,
3614 ext4_count_free_inodes(sb));
3615 percpu_counter_set(&sbi->s_dirs_counter,
3616 ext4_count_dirs(sb));
3617 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
3618
3619 no_journal:
3620 /*
3621 * The maximum number of concurrent works can be high and
3622 * concurrency isn't really necessary. Limit it to 1.
3623 */
3624 EXT4_SB(sb)->dio_unwritten_wq =
3625 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3626 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3627 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3628 goto failed_mount_wq;
3629 }
3630
3631 /*
3632 * The jbd2_journal_load will have done any necessary log recovery,
3633 * so we can safely mount the rest of the filesystem now.
3634 */
3635
3636 root = ext4_iget(sb, EXT4_ROOT_INO);
3637 if (IS_ERR(root)) {
3638 ext4_msg(sb, KERN_ERR, "get root inode failed");
3639 ret = PTR_ERR(root);
3640 root = NULL;
3641 goto failed_mount4;
3642 }
3643 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3644 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3645 goto failed_mount4;
3646 }
3647 sb->s_root = d_alloc_root(root);
3648 if (!sb->s_root) {
3649 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3650 ret = -ENOMEM;
3651 goto failed_mount4;
3652 }
3653
3654 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3655
3656 /* determine the minimum size of new large inodes, if present */
3657 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3658 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3659 EXT4_GOOD_OLD_INODE_SIZE;
3660 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3661 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3662 if (sbi->s_want_extra_isize <
3663 le16_to_cpu(es->s_want_extra_isize))
3664 sbi->s_want_extra_isize =
3665 le16_to_cpu(es->s_want_extra_isize);
3666 if (sbi->s_want_extra_isize <
3667 le16_to_cpu(es->s_min_extra_isize))
3668 sbi->s_want_extra_isize =
3669 le16_to_cpu(es->s_min_extra_isize);
3670 }
3671 }
3672 /* Check if enough inode space is available */
3673 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3674 sbi->s_inode_size) {
3675 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3676 EXT4_GOOD_OLD_INODE_SIZE;
3677 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3678 "available");
3679 }
3680
3681 if (test_opt(sb, DELALLOC) &&
3682 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
3683 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3684 "requested data journaling mode");
3685 clear_opt(sb, DELALLOC);
3686 }
3687 if (test_opt(sb, DIOREAD_NOLOCK)) {
3688 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3689 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3690 "option - requested data journaling mode");
3691 clear_opt(sb, DIOREAD_NOLOCK);
3692 }
3693 if (sb->s_blocksize < PAGE_SIZE) {
3694 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3695 "option - block size is too small");
3696 clear_opt(sb, DIOREAD_NOLOCK);
3697 }
3698 }
3699
3700 err = ext4_setup_system_zone(sb);
3701 if (err) {
3702 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3703 "zone (%d)", err);
3704 goto failed_mount4;
3705 }
3706
3707 ext4_ext_init(sb);
3708 err = ext4_mb_init(sb, needs_recovery);
3709 if (err) {
3710 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3711 err);
3712 goto failed_mount4;
3713 }
3714
3715 err = ext4_register_li_request(sb, first_not_zeroed);
3716 if (err)
3717 goto failed_mount4;
3718
3719 sbi->s_kobj.kset = ext4_kset;
3720 init_completion(&sbi->s_kobj_unregister);
3721 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3722 "%s", sb->s_id);
3723 if (err) {
3724 ext4_mb_release(sb);
3725 ext4_ext_release(sb);
3726 goto failed_mount4;
3727 };
3728
3729 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3730 ext4_orphan_cleanup(sb, es);
3731 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3732 if (needs_recovery) {
3733 ext4_msg(sb, KERN_INFO, "recovery complete");
3734 ext4_mark_recovery_complete(sb, es);
3735 }
3736 if (EXT4_SB(sb)->s_journal) {
3737 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3738 descr = " journalled data mode";
3739 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3740 descr = " ordered data mode";
3741 else
3742 descr = " writeback data mode";
3743 } else
3744 descr = "out journal";
3745
3746 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3747 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3748 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3749
3750 if (es->s_error_count)
3751 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3752
3753 kfree(orig_data);
3754 return 0;
3755
3756 cantfind_ext4:
3757 if (!silent)
3758 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3759 goto failed_mount;
3760
3761 failed_mount4:
3762 iput(root);
3763 sb->s_root = NULL;
3764 ext4_msg(sb, KERN_ERR, "mount failed");
3765 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3766 failed_mount_wq:
3767 ext4_release_system_zone(sb);
3768 if (sbi->s_journal) {
3769 jbd2_journal_destroy(sbi->s_journal);
3770 sbi->s_journal = NULL;
3771 }
3772 failed_mount3:
3773 del_timer(&sbi->s_err_report);
3774 if (sbi->s_flex_groups)
3775 ext4_kvfree(sbi->s_flex_groups);
3776 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3777 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3778 percpu_counter_destroy(&sbi->s_dirs_counter);
3779 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
3780 if (sbi->s_mmp_tsk)
3781 kthread_stop(sbi->s_mmp_tsk);
3782 failed_mount2:
3783 for (i = 0; i < db_count; i++)
3784 brelse(sbi->s_group_desc[i]);
3785 ext4_kvfree(sbi->s_group_desc);
3786 failed_mount:
3787 if (sbi->s_proc) {
3788 remove_proc_entry(sb->s_id, ext4_proc_root);
3789 }
3790 #ifdef CONFIG_QUOTA
3791 for (i = 0; i < MAXQUOTAS; i++)
3792 kfree(sbi->s_qf_names[i]);
3793 #endif
3794 ext4_blkdev_remove(sbi);
3795 brelse(bh);
3796 out_fail:
3797 sb->s_fs_info = NULL;
3798 kfree(sbi->s_blockgroup_lock);
3799 kfree(sbi);
3800 out_free_orig:
3801 kfree(orig_data);
3802 return ret;
3803 }
3804
3805 /*
3806 * Setup any per-fs journal parameters now. We'll do this both on
3807 * initial mount, once the journal has been initialised but before we've
3808 * done any recovery; and again on any subsequent remount.
3809 */
3810 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3811 {
3812 struct ext4_sb_info *sbi = EXT4_SB(sb);
3813
3814 journal->j_commit_interval = sbi->s_commit_interval;
3815 journal->j_min_batch_time = sbi->s_min_batch_time;
3816 journal->j_max_batch_time = sbi->s_max_batch_time;
3817
3818 write_lock(&journal->j_state_lock);
3819 if (test_opt(sb, BARRIER))
3820 journal->j_flags |= JBD2_BARRIER;
3821 else
3822 journal->j_flags &= ~JBD2_BARRIER;
3823 if (test_opt(sb, DATA_ERR_ABORT))
3824 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3825 else
3826 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3827 write_unlock(&journal->j_state_lock);
3828 }
3829
3830 static journal_t *ext4_get_journal(struct super_block *sb,
3831 unsigned int journal_inum)
3832 {
3833 struct inode *journal_inode;
3834 journal_t *journal;
3835
3836 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3837
3838 /* First, test for the existence of a valid inode on disk. Bad
3839 * things happen if we iget() an unused inode, as the subsequent
3840 * iput() will try to delete it. */
3841
3842 journal_inode = ext4_iget(sb, journal_inum);
3843 if (IS_ERR(journal_inode)) {
3844 ext4_msg(sb, KERN_ERR, "no journal found");
3845 return NULL;
3846 }
3847 if (!journal_inode->i_nlink) {
3848 make_bad_inode(journal_inode);
3849 iput(journal_inode);
3850 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3851 return NULL;
3852 }
3853
3854 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3855 journal_inode, journal_inode->i_size);
3856 if (!S_ISREG(journal_inode->i_mode)) {
3857 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3858 iput(journal_inode);
3859 return NULL;
3860 }
3861
3862 journal = jbd2_journal_init_inode(journal_inode);
3863 if (!journal) {
3864 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3865 iput(journal_inode);
3866 return NULL;
3867 }
3868 journal->j_private = sb;
3869 ext4_init_journal_params(sb, journal);
3870 return journal;
3871 }
3872
3873 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3874 dev_t j_dev)
3875 {
3876 struct buffer_head *bh;
3877 journal_t *journal;
3878 ext4_fsblk_t start;
3879 ext4_fsblk_t len;
3880 int hblock, blocksize;
3881 ext4_fsblk_t sb_block;
3882 unsigned long offset;
3883 struct ext4_super_block *es;
3884 struct block_device *bdev;
3885
3886 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3887
3888 bdev = ext4_blkdev_get(j_dev, sb);
3889 if (bdev == NULL)
3890 return NULL;
3891
3892 blocksize = sb->s_blocksize;
3893 hblock = bdev_logical_block_size(bdev);
3894 if (blocksize < hblock) {
3895 ext4_msg(sb, KERN_ERR,
3896 "blocksize too small for journal device");
3897 goto out_bdev;
3898 }
3899
3900 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3901 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3902 set_blocksize(bdev, blocksize);
3903 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3904 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3905 "external journal");
3906 goto out_bdev;
3907 }
3908
3909 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3910 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3911 !(le32_to_cpu(es->s_feature_incompat) &
3912 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3913 ext4_msg(sb, KERN_ERR, "external journal has "
3914 "bad superblock");
3915 brelse(bh);
3916 goto out_bdev;
3917 }
3918
3919 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3920 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3921 brelse(bh);
3922 goto out_bdev;
3923 }
3924
3925 len = ext4_blocks_count(es);
3926 start = sb_block + 1;
3927 brelse(bh); /* we're done with the superblock */
3928
3929 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3930 start, len, blocksize);
3931 if (!journal) {
3932 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3933 goto out_bdev;
3934 }
3935 journal->j_private = sb;
3936 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3937 wait_on_buffer(journal->j_sb_buffer);
3938 if (!buffer_uptodate(journal->j_sb_buffer)) {
3939 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3940 goto out_journal;
3941 }
3942 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3943 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3944 "user (unsupported) - %d",
3945 be32_to_cpu(journal->j_superblock->s_nr_users));
3946 goto out_journal;
3947 }
3948 EXT4_SB(sb)->journal_bdev = bdev;
3949 ext4_init_journal_params(sb, journal);
3950 return journal;
3951
3952 out_journal:
3953 jbd2_journal_destroy(journal);
3954 out_bdev:
3955 ext4_blkdev_put(bdev);
3956 return NULL;
3957 }
3958
3959 static int ext4_load_journal(struct super_block *sb,
3960 struct ext4_super_block *es,
3961 unsigned long journal_devnum)
3962 {
3963 journal_t *journal;
3964 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3965 dev_t journal_dev;
3966 int err = 0;
3967 int really_read_only;
3968
3969 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3970
3971 if (journal_devnum &&
3972 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3973 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3974 "numbers have changed");
3975 journal_dev = new_decode_dev(journal_devnum);
3976 } else
3977 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3978
3979 really_read_only = bdev_read_only(sb->s_bdev);
3980
3981 /*
3982 * Are we loading a blank journal or performing recovery after a
3983 * crash? For recovery, we need to check in advance whether we
3984 * can get read-write access to the device.
3985 */
3986 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3987 if (sb->s_flags & MS_RDONLY) {
3988 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3989 "required on readonly filesystem");
3990 if (really_read_only) {
3991 ext4_msg(sb, KERN_ERR, "write access "
3992 "unavailable, cannot proceed");
3993 return -EROFS;
3994 }
3995 ext4_msg(sb, KERN_INFO, "write access will "
3996 "be enabled during recovery");
3997 }
3998 }
3999
4000 if (journal_inum && journal_dev) {
4001 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4002 "and inode journals!");
4003 return -EINVAL;
4004 }
4005
4006 if (journal_inum) {
4007 if (!(journal = ext4_get_journal(sb, journal_inum)))
4008 return -EINVAL;
4009 } else {
4010 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4011 return -EINVAL;
4012 }
4013
4014 if (!(journal->j_flags & JBD2_BARRIER))
4015 ext4_msg(sb, KERN_INFO, "barriers disabled");
4016
4017 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
4018 err = jbd2_journal_update_format(journal);
4019 if (err) {
4020 ext4_msg(sb, KERN_ERR, "error updating journal");
4021 jbd2_journal_destroy(journal);
4022 return err;
4023 }
4024 }
4025
4026 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4027 err = jbd2_journal_wipe(journal, !really_read_only);
4028 if (!err) {
4029 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4030 if (save)
4031 memcpy(save, ((char *) es) +
4032 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4033 err = jbd2_journal_load(journal);
4034 if (save)
4035 memcpy(((char *) es) + EXT4_S_ERR_START,
4036 save, EXT4_S_ERR_LEN);
4037 kfree(save);
4038 }
4039
4040 if (err) {
4041 ext4_msg(sb, KERN_ERR, "error loading journal");
4042 jbd2_journal_destroy(journal);
4043 return err;
4044 }
4045
4046 EXT4_SB(sb)->s_journal = journal;
4047 ext4_clear_journal_err(sb, es);
4048
4049 if (!really_read_only && journal_devnum &&
4050 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4051 es->s_journal_dev = cpu_to_le32(journal_devnum);
4052
4053 /* Make sure we flush the recovery flag to disk. */
4054 ext4_commit_super(sb, 1);
4055 }
4056
4057 return 0;
4058 }
4059
4060 static int ext4_commit_super(struct super_block *sb, int sync)
4061 {
4062 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4063 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4064 int error = 0;
4065
4066 if (!sbh)
4067 return error;
4068 if (buffer_write_io_error(sbh)) {
4069 /*
4070 * Oh, dear. A previous attempt to write the
4071 * superblock failed. This could happen because the
4072 * USB device was yanked out. Or it could happen to
4073 * be a transient write error and maybe the block will
4074 * be remapped. Nothing we can do but to retry the
4075 * write and hope for the best.
4076 */
4077 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4078 "superblock detected");
4079 clear_buffer_write_io_error(sbh);
4080 set_buffer_uptodate(sbh);
4081 }
4082 /*
4083 * If the file system is mounted read-only, don't update the
4084 * superblock write time. This avoids updating the superblock
4085 * write time when we are mounting the root file system
4086 * read/only but we need to replay the journal; at that point,
4087 * for people who are east of GMT and who make their clock
4088 * tick in localtime for Windows bug-for-bug compatibility,
4089 * the clock is set in the future, and this will cause e2fsck
4090 * to complain and force a full file system check.
4091 */
4092 if (!(sb->s_flags & MS_RDONLY))
4093 es->s_wtime = cpu_to_le32(get_seconds());
4094 if (sb->s_bdev->bd_part)
4095 es->s_kbytes_written =
4096 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4097 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4098 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4099 else
4100 es->s_kbytes_written =
4101 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4102 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
4103 &EXT4_SB(sb)->s_freeblocks_counter));
4104 es->s_free_inodes_count =
4105 cpu_to_le32(percpu_counter_sum_positive(
4106 &EXT4_SB(sb)->s_freeinodes_counter));
4107 sb->s_dirt = 0;
4108 BUFFER_TRACE(sbh, "marking dirty");
4109 mark_buffer_dirty(sbh);
4110 if (sync) {
4111 error = sync_dirty_buffer(sbh);
4112 if (error)
4113 return error;
4114
4115 error = buffer_write_io_error(sbh);
4116 if (error) {
4117 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4118 "superblock");
4119 clear_buffer_write_io_error(sbh);
4120 set_buffer_uptodate(sbh);
4121 }
4122 }
4123 return error;
4124 }
4125
4126 /*
4127 * Have we just finished recovery? If so, and if we are mounting (or
4128 * remounting) the filesystem readonly, then we will end up with a
4129 * consistent fs on disk. Record that fact.
4130 */
4131 static void ext4_mark_recovery_complete(struct super_block *sb,
4132 struct ext4_super_block *es)
4133 {
4134 journal_t *journal = EXT4_SB(sb)->s_journal;
4135
4136 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4137 BUG_ON(journal != NULL);
4138 return;
4139 }
4140 jbd2_journal_lock_updates(journal);
4141 if (jbd2_journal_flush(journal) < 0)
4142 goto out;
4143
4144 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4145 sb->s_flags & MS_RDONLY) {
4146 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4147 ext4_commit_super(sb, 1);
4148 }
4149
4150 out:
4151 jbd2_journal_unlock_updates(journal);
4152 }
4153
4154 /*
4155 * If we are mounting (or read-write remounting) a filesystem whose journal
4156 * has recorded an error from a previous lifetime, move that error to the
4157 * main filesystem now.
4158 */
4159 static void ext4_clear_journal_err(struct super_block *sb,
4160 struct ext4_super_block *es)
4161 {
4162 journal_t *journal;
4163 int j_errno;
4164 const char *errstr;
4165
4166 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4167
4168 journal = EXT4_SB(sb)->s_journal;
4169
4170 /*
4171 * Now check for any error status which may have been recorded in the
4172 * journal by a prior ext4_error() or ext4_abort()
4173 */
4174
4175 j_errno = jbd2_journal_errno(journal);
4176 if (j_errno) {
4177 char nbuf[16];
4178
4179 errstr = ext4_decode_error(sb, j_errno, nbuf);
4180 ext4_warning(sb, "Filesystem error recorded "
4181 "from previous mount: %s", errstr);
4182 ext4_warning(sb, "Marking fs in need of filesystem check.");
4183
4184 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4185 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4186 ext4_commit_super(sb, 1);
4187
4188 jbd2_journal_clear_err(journal);
4189 }
4190 }
4191
4192 /*
4193 * Force the running and committing transactions to commit,
4194 * and wait on the commit.
4195 */
4196 int ext4_force_commit(struct super_block *sb)
4197 {
4198 journal_t *journal;
4199 int ret = 0;
4200
4201 if (sb->s_flags & MS_RDONLY)
4202 return 0;
4203
4204 journal = EXT4_SB(sb)->s_journal;
4205 if (journal) {
4206 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4207 ret = ext4_journal_force_commit(journal);
4208 }
4209
4210 return ret;
4211 }
4212
4213 static void ext4_write_super(struct super_block *sb)
4214 {
4215 lock_super(sb);
4216 ext4_commit_super(sb, 1);
4217 unlock_super(sb);
4218 }
4219
4220 static int ext4_sync_fs(struct super_block *sb, int wait)
4221 {
4222 int ret = 0;
4223 tid_t target;
4224 struct ext4_sb_info *sbi = EXT4_SB(sb);
4225
4226 trace_ext4_sync_fs(sb, wait);
4227 flush_workqueue(sbi->dio_unwritten_wq);
4228 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4229 if (wait)
4230 jbd2_log_wait_commit(sbi->s_journal, target);
4231 }
4232 return ret;
4233 }
4234
4235 /*
4236 * LVM calls this function before a (read-only) snapshot is created. This
4237 * gives us a chance to flush the journal completely and mark the fs clean.
4238 *
4239 * Note that only this function cannot bring a filesystem to be in a clean
4240 * state independently, because ext4 prevents a new handle from being started
4241 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4242 * the upper layer.
4243 */
4244 static int ext4_freeze(struct super_block *sb)
4245 {
4246 int error = 0;
4247 journal_t *journal;
4248
4249 if (sb->s_flags & MS_RDONLY)
4250 return 0;
4251
4252 journal = EXT4_SB(sb)->s_journal;
4253
4254 /* Now we set up the journal barrier. */
4255 jbd2_journal_lock_updates(journal);
4256
4257 /*
4258 * Don't clear the needs_recovery flag if we failed to flush
4259 * the journal.
4260 */
4261 error = jbd2_journal_flush(journal);
4262 if (error < 0)
4263 goto out;
4264
4265 /* Journal blocked and flushed, clear needs_recovery flag. */
4266 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4267 error = ext4_commit_super(sb, 1);
4268 out:
4269 /* we rely on s_frozen to stop further updates */
4270 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4271 return error;
4272 }
4273
4274 /*
4275 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4276 * flag here, even though the filesystem is not technically dirty yet.
4277 */
4278 static int ext4_unfreeze(struct super_block *sb)
4279 {
4280 if (sb->s_flags & MS_RDONLY)
4281 return 0;
4282
4283 lock_super(sb);
4284 /* Reset the needs_recovery flag before the fs is unlocked. */
4285 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4286 ext4_commit_super(sb, 1);
4287 unlock_super(sb);
4288 return 0;
4289 }
4290
4291 /*
4292 * Structure to save mount options for ext4_remount's benefit
4293 */
4294 struct ext4_mount_options {
4295 unsigned long s_mount_opt;
4296 unsigned long s_mount_opt2;
4297 uid_t s_resuid;
4298 gid_t s_resgid;
4299 unsigned long s_commit_interval;
4300 u32 s_min_batch_time, s_max_batch_time;
4301 #ifdef CONFIG_QUOTA
4302 int s_jquota_fmt;
4303 char *s_qf_names[MAXQUOTAS];
4304 #endif
4305 };
4306
4307 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4308 {
4309 struct ext4_super_block *es;
4310 struct ext4_sb_info *sbi = EXT4_SB(sb);
4311 ext4_fsblk_t n_blocks_count = 0;
4312 unsigned long old_sb_flags;
4313 struct ext4_mount_options old_opts;
4314 int enable_quota = 0;
4315 ext4_group_t g;
4316 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4317 int err = 0;
4318 #ifdef CONFIG_QUOTA
4319 int i;
4320 #endif
4321 char *orig_data = kstrdup(data, GFP_KERNEL);
4322
4323 /* Store the original options */
4324 lock_super(sb);
4325 old_sb_flags = sb->s_flags;
4326 old_opts.s_mount_opt = sbi->s_mount_opt;
4327 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4328 old_opts.s_resuid = sbi->s_resuid;
4329 old_opts.s_resgid = sbi->s_resgid;
4330 old_opts.s_commit_interval = sbi->s_commit_interval;
4331 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4332 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4333 #ifdef CONFIG_QUOTA
4334 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4335 for (i = 0; i < MAXQUOTAS; i++)
4336 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4337 #endif
4338 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4339 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4340
4341 /*
4342 * Allow the "check" option to be passed as a remount option.
4343 */
4344 if (!parse_options(data, sb, NULL, &journal_ioprio,
4345 &n_blocks_count, 1)) {
4346 err = -EINVAL;
4347 goto restore_opts;
4348 }
4349
4350 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4351 ext4_abort(sb, "Abort forced by user");
4352
4353 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4354 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4355
4356 es = sbi->s_es;
4357
4358 if (sbi->s_journal) {
4359 ext4_init_journal_params(sb, sbi->s_journal);
4360 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4361 }
4362
4363 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
4364 n_blocks_count > ext4_blocks_count(es)) {
4365 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4366 err = -EROFS;
4367 goto restore_opts;
4368 }
4369
4370 if (*flags & MS_RDONLY) {
4371 err = dquot_suspend(sb, -1);
4372 if (err < 0)
4373 goto restore_opts;
4374
4375 /*
4376 * First of all, the unconditional stuff we have to do
4377 * to disable replay of the journal when we next remount
4378 */
4379 sb->s_flags |= MS_RDONLY;
4380
4381 /*
4382 * OK, test if we are remounting a valid rw partition
4383 * readonly, and if so set the rdonly flag and then
4384 * mark the partition as valid again.
4385 */
4386 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4387 (sbi->s_mount_state & EXT4_VALID_FS))
4388 es->s_state = cpu_to_le16(sbi->s_mount_state);
4389
4390 if (sbi->s_journal)
4391 ext4_mark_recovery_complete(sb, es);
4392 } else {
4393 /* Make sure we can mount this feature set readwrite */
4394 if (!ext4_feature_set_ok(sb, 0)) {
4395 err = -EROFS;
4396 goto restore_opts;
4397 }
4398 /*
4399 * Make sure the group descriptor checksums
4400 * are sane. If they aren't, refuse to remount r/w.
4401 */
4402 for (g = 0; g < sbi->s_groups_count; g++) {
4403 struct ext4_group_desc *gdp =
4404 ext4_get_group_desc(sb, g, NULL);
4405
4406 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4407 ext4_msg(sb, KERN_ERR,
4408 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4409 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4410 le16_to_cpu(gdp->bg_checksum));
4411 err = -EINVAL;
4412 goto restore_opts;
4413 }
4414 }
4415
4416 /*
4417 * If we have an unprocessed orphan list hanging
4418 * around from a previously readonly bdev mount,
4419 * require a full umount/remount for now.
4420 */
4421 if (es->s_last_orphan) {
4422 ext4_msg(sb, KERN_WARNING, "Couldn't "
4423 "remount RDWR because of unprocessed "
4424 "orphan inode list. Please "
4425 "umount/remount instead");
4426 err = -EINVAL;
4427 goto restore_opts;
4428 }
4429
4430 /*
4431 * Mounting a RDONLY partition read-write, so reread
4432 * and store the current valid flag. (It may have
4433 * been changed by e2fsck since we originally mounted
4434 * the partition.)
4435 */
4436 if (sbi->s_journal)
4437 ext4_clear_journal_err(sb, es);
4438 sbi->s_mount_state = le16_to_cpu(es->s_state);
4439 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
4440 goto restore_opts;
4441 if (!ext4_setup_super(sb, es, 0))
4442 sb->s_flags &= ~MS_RDONLY;
4443 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4444 EXT4_FEATURE_INCOMPAT_MMP))
4445 if (ext4_multi_mount_protect(sb,
4446 le64_to_cpu(es->s_mmp_block))) {
4447 err = -EROFS;
4448 goto restore_opts;
4449 }
4450 enable_quota = 1;
4451 }
4452 }
4453
4454 /*
4455 * Reinitialize lazy itable initialization thread based on
4456 * current settings
4457 */
4458 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4459 ext4_unregister_li_request(sb);
4460 else {
4461 ext4_group_t first_not_zeroed;
4462 first_not_zeroed = ext4_has_uninit_itable(sb);
4463 ext4_register_li_request(sb, first_not_zeroed);
4464 }
4465
4466 ext4_setup_system_zone(sb);
4467 if (sbi->s_journal == NULL)
4468 ext4_commit_super(sb, 1);
4469
4470 #ifdef CONFIG_QUOTA
4471 /* Release old quota file names */
4472 for (i = 0; i < MAXQUOTAS; i++)
4473 if (old_opts.s_qf_names[i] &&
4474 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4475 kfree(old_opts.s_qf_names[i]);
4476 #endif
4477 unlock_super(sb);
4478 if (enable_quota)
4479 dquot_resume(sb, -1);
4480
4481 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4482 kfree(orig_data);
4483 return 0;
4484
4485 restore_opts:
4486 sb->s_flags = old_sb_flags;
4487 sbi->s_mount_opt = old_opts.s_mount_opt;
4488 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4489 sbi->s_resuid = old_opts.s_resuid;
4490 sbi->s_resgid = old_opts.s_resgid;
4491 sbi->s_commit_interval = old_opts.s_commit_interval;
4492 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4493 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4494 #ifdef CONFIG_QUOTA
4495 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4496 for (i = 0; i < MAXQUOTAS; i++) {
4497 if (sbi->s_qf_names[i] &&
4498 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4499 kfree(sbi->s_qf_names[i]);
4500 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4501 }
4502 #endif
4503 unlock_super(sb);
4504 kfree(orig_data);
4505 return err;
4506 }
4507
4508 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4509 {
4510 struct super_block *sb = dentry->d_sb;
4511 struct ext4_sb_info *sbi = EXT4_SB(sb);
4512 struct ext4_super_block *es = sbi->s_es;
4513 u64 fsid;
4514 s64 bfree;
4515
4516 if (test_opt(sb, MINIX_DF)) {
4517 sbi->s_overhead_last = 0;
4518 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4519 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4520 ext4_fsblk_t overhead = 0;
4521
4522 /*
4523 * Compute the overhead (FS structures). This is constant
4524 * for a given filesystem unless the number of block groups
4525 * changes so we cache the previous value until it does.
4526 */
4527
4528 /*
4529 * All of the blocks before first_data_block are
4530 * overhead
4531 */
4532 overhead = le32_to_cpu(es->s_first_data_block);
4533
4534 /*
4535 * Add the overhead attributed to the superblock and
4536 * block group descriptors. If the sparse superblocks
4537 * feature is turned on, then not all groups have this.
4538 */
4539 for (i = 0; i < ngroups; i++) {
4540 overhead += ext4_bg_has_super(sb, i) +
4541 ext4_bg_num_gdb(sb, i);
4542 cond_resched();
4543 }
4544
4545 /*
4546 * Every block group has an inode bitmap, a block
4547 * bitmap, and an inode table.
4548 */
4549 overhead += ngroups * (2 + sbi->s_itb_per_group);
4550 sbi->s_overhead_last = overhead;
4551 smp_wmb();
4552 sbi->s_blocks_last = ext4_blocks_count(es);
4553 }
4554
4555 buf->f_type = EXT4_SUPER_MAGIC;
4556 buf->f_bsize = sb->s_blocksize;
4557 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
4558 bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4559 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
4560 /* prevent underflow in case that few free space is available */
4561 buf->f_bfree = max_t(s64, bfree, 0);
4562 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4563 if (buf->f_bfree < ext4_r_blocks_count(es))
4564 buf->f_bavail = 0;
4565 buf->f_files = le32_to_cpu(es->s_inodes_count);
4566 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4567 buf->f_namelen = EXT4_NAME_LEN;
4568 fsid = le64_to_cpup((void *)es->s_uuid) ^
4569 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4570 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4571 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4572
4573 return 0;
4574 }
4575
4576 /* Helper function for writing quotas on sync - we need to start transaction
4577 * before quota file is locked for write. Otherwise the are possible deadlocks:
4578 * Process 1 Process 2
4579 * ext4_create() quota_sync()
4580 * jbd2_journal_start() write_dquot()
4581 * dquot_initialize() down(dqio_mutex)
4582 * down(dqio_mutex) jbd2_journal_start()
4583 *
4584 */
4585
4586 #ifdef CONFIG_QUOTA
4587
4588 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4589 {
4590 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4591 }
4592
4593 static int ext4_write_dquot(struct dquot *dquot)
4594 {
4595 int ret, err;
4596 handle_t *handle;
4597 struct inode *inode;
4598
4599 inode = dquot_to_inode(dquot);
4600 handle = ext4_journal_start(inode,
4601 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4602 if (IS_ERR(handle))
4603 return PTR_ERR(handle);
4604 ret = dquot_commit(dquot);
4605 err = ext4_journal_stop(handle);
4606 if (!ret)
4607 ret = err;
4608 return ret;
4609 }
4610
4611 static int ext4_acquire_dquot(struct dquot *dquot)
4612 {
4613 int ret, err;
4614 handle_t *handle;
4615
4616 handle = ext4_journal_start(dquot_to_inode(dquot),
4617 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4618 if (IS_ERR(handle))
4619 return PTR_ERR(handle);
4620 ret = dquot_acquire(dquot);
4621 err = ext4_journal_stop(handle);
4622 if (!ret)
4623 ret = err;
4624 return ret;
4625 }
4626
4627 static int ext4_release_dquot(struct dquot *dquot)
4628 {
4629 int ret, err;
4630 handle_t *handle;
4631
4632 handle = ext4_journal_start(dquot_to_inode(dquot),
4633 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4634 if (IS_ERR(handle)) {
4635 /* Release dquot anyway to avoid endless cycle in dqput() */
4636 dquot_release(dquot);
4637 return PTR_ERR(handle);
4638 }
4639 ret = dquot_release(dquot);
4640 err = ext4_journal_stop(handle);
4641 if (!ret)
4642 ret = err;
4643 return ret;
4644 }
4645
4646 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4647 {
4648 /* Are we journaling quotas? */
4649 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4650 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4651 dquot_mark_dquot_dirty(dquot);
4652 return ext4_write_dquot(dquot);
4653 } else {
4654 return dquot_mark_dquot_dirty(dquot);
4655 }
4656 }
4657
4658 static int ext4_write_info(struct super_block *sb, int type)
4659 {
4660 int ret, err;
4661 handle_t *handle;
4662
4663 /* Data block + inode block */
4664 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4665 if (IS_ERR(handle))
4666 return PTR_ERR(handle);
4667 ret = dquot_commit_info(sb, type);
4668 err = ext4_journal_stop(handle);
4669 if (!ret)
4670 ret = err;
4671 return ret;
4672 }
4673
4674 /*
4675 * Turn on quotas during mount time - we need to find
4676 * the quota file and such...
4677 */
4678 static int ext4_quota_on_mount(struct super_block *sb, int type)
4679 {
4680 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4681 EXT4_SB(sb)->s_jquota_fmt, type);
4682 }
4683
4684 /*
4685 * Standard function to be called on quota_on
4686 */
4687 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4688 struct path *path)
4689 {
4690 int err;
4691
4692 if (!test_opt(sb, QUOTA))
4693 return -EINVAL;
4694
4695 /* Quotafile not on the same filesystem? */
4696 if (path->mnt->mnt_sb != sb)
4697 return -EXDEV;
4698 /* Journaling quota? */
4699 if (EXT4_SB(sb)->s_qf_names[type]) {
4700 /* Quotafile not in fs root? */
4701 if (path->dentry->d_parent != sb->s_root)
4702 ext4_msg(sb, KERN_WARNING,
4703 "Quota file not on filesystem root. "
4704 "Journaled quota will not work");
4705 }
4706
4707 /*
4708 * When we journal data on quota file, we have to flush journal to see
4709 * all updates to the file when we bypass pagecache...
4710 */
4711 if (EXT4_SB(sb)->s_journal &&
4712 ext4_should_journal_data(path->dentry->d_inode)) {
4713 /*
4714 * We don't need to lock updates but journal_flush() could
4715 * otherwise be livelocked...
4716 */
4717 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4718 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4719 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4720 if (err)
4721 return err;
4722 }
4723
4724 return dquot_quota_on(sb, type, format_id, path);
4725 }
4726
4727 static int ext4_quota_off(struct super_block *sb, int type)
4728 {
4729 struct inode *inode = sb_dqopt(sb)->files[type];
4730 handle_t *handle;
4731
4732 /* Force all delayed allocation blocks to be allocated.
4733 * Caller already holds s_umount sem */
4734 if (test_opt(sb, DELALLOC))
4735 sync_filesystem(sb);
4736
4737 if (!inode)
4738 goto out;
4739
4740 /* Update modification times of quota files when userspace can
4741 * start looking at them */
4742 handle = ext4_journal_start(inode, 1);
4743 if (IS_ERR(handle))
4744 goto out;
4745 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4746 ext4_mark_inode_dirty(handle, inode);
4747 ext4_journal_stop(handle);
4748
4749 out:
4750 return dquot_quota_off(sb, type);
4751 }
4752
4753 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4754 * acquiring the locks... As quota files are never truncated and quota code
4755 * itself serializes the operations (and no one else should touch the files)
4756 * we don't have to be afraid of races */
4757 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4758 size_t len, loff_t off)
4759 {
4760 struct inode *inode = sb_dqopt(sb)->files[type];
4761 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4762 int err = 0;
4763 int offset = off & (sb->s_blocksize - 1);
4764 int tocopy;
4765 size_t toread;
4766 struct buffer_head *bh;
4767 loff_t i_size = i_size_read(inode);
4768
4769 if (off > i_size)
4770 return 0;
4771 if (off+len > i_size)
4772 len = i_size-off;
4773 toread = len;
4774 while (toread > 0) {
4775 tocopy = sb->s_blocksize - offset < toread ?
4776 sb->s_blocksize - offset : toread;
4777 bh = ext4_bread(NULL, inode, blk, 0, &err);
4778 if (err)
4779 return err;
4780 if (!bh) /* A hole? */
4781 memset(data, 0, tocopy);
4782 else
4783 memcpy(data, bh->b_data+offset, tocopy);
4784 brelse(bh);
4785 offset = 0;
4786 toread -= tocopy;
4787 data += tocopy;
4788 blk++;
4789 }
4790 return len;
4791 }
4792
4793 /* Write to quotafile (we know the transaction is already started and has
4794 * enough credits) */
4795 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4796 const char *data, size_t len, loff_t off)
4797 {
4798 struct inode *inode = sb_dqopt(sb)->files[type];
4799 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4800 int err = 0;
4801 int offset = off & (sb->s_blocksize - 1);
4802 struct buffer_head *bh;
4803 handle_t *handle = journal_current_handle();
4804
4805 if (EXT4_SB(sb)->s_journal && !handle) {
4806 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4807 " cancelled because transaction is not started",
4808 (unsigned long long)off, (unsigned long long)len);
4809 return -EIO;
4810 }
4811 /*
4812 * Since we account only one data block in transaction credits,
4813 * then it is impossible to cross a block boundary.
4814 */
4815 if (sb->s_blocksize - offset < len) {
4816 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4817 " cancelled because not block aligned",
4818 (unsigned long long)off, (unsigned long long)len);
4819 return -EIO;
4820 }
4821
4822 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4823 bh = ext4_bread(handle, inode, blk, 1, &err);
4824 if (!bh)
4825 goto out;
4826 err = ext4_journal_get_write_access(handle, bh);
4827 if (err) {
4828 brelse(bh);
4829 goto out;
4830 }
4831 lock_buffer(bh);
4832 memcpy(bh->b_data+offset, data, len);
4833 flush_dcache_page(bh->b_page);
4834 unlock_buffer(bh);
4835 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4836 brelse(bh);
4837 out:
4838 if (err) {
4839 mutex_unlock(&inode->i_mutex);
4840 return err;
4841 }
4842 if (inode->i_size < off + len) {
4843 i_size_write(inode, off + len);
4844 EXT4_I(inode)->i_disksize = inode->i_size;
4845 ext4_mark_inode_dirty(handle, inode);
4846 }
4847 mutex_unlock(&inode->i_mutex);
4848 return len;
4849 }
4850
4851 #endif
4852
4853 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4854 const char *dev_name, void *data)
4855 {
4856 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4857 }
4858
4859 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4860 static inline void register_as_ext2(void)
4861 {
4862 int err = register_filesystem(&ext2_fs_type);
4863 if (err)
4864 printk(KERN_WARNING
4865 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4866 }
4867
4868 static inline void unregister_as_ext2(void)
4869 {
4870 unregister_filesystem(&ext2_fs_type);
4871 }
4872
4873 static inline int ext2_feature_set_ok(struct super_block *sb)
4874 {
4875 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4876 return 0;
4877 if (sb->s_flags & MS_RDONLY)
4878 return 1;
4879 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4880 return 0;
4881 return 1;
4882 }
4883 MODULE_ALIAS("ext2");
4884 #else
4885 static inline void register_as_ext2(void) { }
4886 static inline void unregister_as_ext2(void) { }
4887 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4888 #endif
4889
4890 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4891 static inline void register_as_ext3(void)
4892 {
4893 int err = register_filesystem(&ext3_fs_type);
4894 if (err)
4895 printk(KERN_WARNING
4896 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4897 }
4898
4899 static inline void unregister_as_ext3(void)
4900 {
4901 unregister_filesystem(&ext3_fs_type);
4902 }
4903
4904 static inline int ext3_feature_set_ok(struct super_block *sb)
4905 {
4906 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4907 return 0;
4908 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4909 return 0;
4910 if (sb->s_flags & MS_RDONLY)
4911 return 1;
4912 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4913 return 0;
4914 return 1;
4915 }
4916 MODULE_ALIAS("ext3");
4917 #else
4918 static inline void register_as_ext3(void) { }
4919 static inline void unregister_as_ext3(void) { }
4920 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
4921 #endif
4922
4923 static struct file_system_type ext4_fs_type = {
4924 .owner = THIS_MODULE,
4925 .name = "ext4",
4926 .mount = ext4_mount,
4927 .kill_sb = kill_block_super,
4928 .fs_flags = FS_REQUIRES_DEV,
4929 };
4930
4931 static int __init ext4_init_feat_adverts(void)
4932 {
4933 struct ext4_features *ef;
4934 int ret = -ENOMEM;
4935
4936 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4937 if (!ef)
4938 goto out;
4939
4940 ef->f_kobj.kset = ext4_kset;
4941 init_completion(&ef->f_kobj_unregister);
4942 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4943 "features");
4944 if (ret) {
4945 kfree(ef);
4946 goto out;
4947 }
4948
4949 ext4_feat = ef;
4950 ret = 0;
4951 out:
4952 return ret;
4953 }
4954
4955 static void ext4_exit_feat_adverts(void)
4956 {
4957 kobject_put(&ext4_feat->f_kobj);
4958 wait_for_completion(&ext4_feat->f_kobj_unregister);
4959 kfree(ext4_feat);
4960 }
4961
4962 /* Shared across all ext4 file systems */
4963 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4964 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4965
4966 static int __init ext4_init_fs(void)
4967 {
4968 int i, err;
4969
4970 ext4_check_flag_values();
4971
4972 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4973 mutex_init(&ext4__aio_mutex[i]);
4974 init_waitqueue_head(&ext4__ioend_wq[i]);
4975 }
4976
4977 err = ext4_init_pageio();
4978 if (err)
4979 return err;
4980 err = ext4_init_system_zone();
4981 if (err)
4982 goto out7;
4983 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4984 if (!ext4_kset)
4985 goto out6;
4986 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4987 if (!ext4_proc_root)
4988 goto out5;
4989
4990 err = ext4_init_feat_adverts();
4991 if (err)
4992 goto out4;
4993
4994 err = ext4_init_mballoc();
4995 if (err)
4996 goto out3;
4997
4998 err = ext4_init_xattr();
4999 if (err)
5000 goto out2;
5001 err = init_inodecache();
5002 if (err)
5003 goto out1;
5004 register_as_ext3();
5005 register_as_ext2();
5006 err = register_filesystem(&ext4_fs_type);
5007 if (err)
5008 goto out;
5009
5010 ext4_li_info = NULL;
5011 mutex_init(&ext4_li_mtx);
5012 return 0;
5013 out:
5014 unregister_as_ext2();
5015 unregister_as_ext3();
5016 destroy_inodecache();
5017 out1:
5018 ext4_exit_xattr();
5019 out2:
5020 ext4_exit_mballoc();
5021 out3:
5022 ext4_exit_feat_adverts();
5023 out4:
5024 remove_proc_entry("fs/ext4", NULL);
5025 out5:
5026 kset_unregister(ext4_kset);
5027 out6:
5028 ext4_exit_system_zone();
5029 out7:
5030 ext4_exit_pageio();
5031 return err;
5032 }
5033
5034 static void __exit ext4_exit_fs(void)
5035 {
5036 ext4_destroy_lazyinit_thread();
5037 unregister_as_ext2();
5038 unregister_as_ext3();
5039 unregister_filesystem(&ext4_fs_type);
5040 destroy_inodecache();
5041 ext4_exit_xattr();
5042 ext4_exit_mballoc();
5043 ext4_exit_feat_adverts();
5044 remove_proc_entry("fs/ext4", NULL);
5045 kset_unregister(ext4_kset);
5046 ext4_exit_system_zone();
5047 ext4_exit_pageio();
5048 }
5049
5050 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5051 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5052 MODULE_LICENSE("GPL");
5053 module_init(ext4_init_fs)
5054 module_exit(ext4_exit_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree