?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ext4: record the checksum algorithm in use in the superblock
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ext4 / super.c
blobf489ffb3605da387129da7d3d71debca3fa397e8
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
43
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
46
47 #include "ext4.h"
48 #include "ext4_extents.h"
49 #include "ext4_jbd2.h"
50 #include "xattr.h"
51 #include "acl.h"
52 #include "mballoc.h"
53
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
56
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
62
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static const char *ext4_decode_error(struct super_block *sb, int errno,
73 char nbuf[16]);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static void ext4_write_super(struct super_block *sb);
78 static int ext4_freeze(struct super_block *sb);
79 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 const char *dev_name, void *data);
81 static inline int ext2_feature_set_ok(struct super_block *sb);
82 static inline int ext3_feature_set_ok(struct super_block *sb);
83 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
84 static void ext4_destroy_lazyinit_thread(void);
85 static void ext4_unregister_li_request(struct super_block *sb);
86 static void ext4_clear_request_list(void);
87
88 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89 static struct file_system_type ext2_fs_type = {
90 .owner = THIS_MODULE,
91 .name = "ext2",
92 .mount = ext4_mount,
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
95 };
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #else
98 #define IS_EXT2_SB(sb) (0)
99 #endif
100
101
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
106 .mount = ext4_mount,
107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109 };
110 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111 #else
112 #define IS_EXT3_SB(sb) (0)
113 #endif
114
115 static int ext4_verify_csum_type(struct super_block *sb,
116 struct ext4_super_block *es)
117 {
118 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
119 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
120 return 1;
121
122 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
123 }
124
125 void *ext4_kvmalloc(size_t size, gfp_t flags)
126 {
127 void *ret;
128
129 ret = kmalloc(size, flags);
130 if (!ret)
131 ret = __vmalloc(size, flags, PAGE_KERNEL);
132 return ret;
133 }
134
135 void *ext4_kvzalloc(size_t size, gfp_t flags)
136 {
137 void *ret;
138
139 ret = kzalloc(size, flags);
140 if (!ret)
141 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
142 return ret;
143 }
144
145 void ext4_kvfree(void *ptr)
146 {
147 if (is_vmalloc_addr(ptr))
148 vfree(ptr);
149 else
150 kfree(ptr);
151
152 }
153
154 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
155 struct ext4_group_desc *bg)
156 {
157 return le32_to_cpu(bg->bg_block_bitmap_lo) |
158 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
159 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
160 }
161
162 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
163 struct ext4_group_desc *bg)
164 {
165 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
166 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
167 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
168 }
169
170 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
171 struct ext4_group_desc *bg)
172 {
173 return le32_to_cpu(bg->bg_inode_table_lo) |
174 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
175 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
176 }
177
178 __u32 ext4_free_group_clusters(struct super_block *sb,
179 struct ext4_group_desc *bg)
180 {
181 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
182 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
183 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
184 }
185
186 __u32 ext4_free_inodes_count(struct super_block *sb,
187 struct ext4_group_desc *bg)
188 {
189 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
192 }
193
194 __u32 ext4_used_dirs_count(struct super_block *sb,
195 struct ext4_group_desc *bg)
196 {
197 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
200 }
201
202 __u32 ext4_itable_unused_count(struct super_block *sb,
203 struct ext4_group_desc *bg)
204 {
205 return le16_to_cpu(bg->bg_itable_unused_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
208 }
209
210 void ext4_block_bitmap_set(struct super_block *sb,
211 struct ext4_group_desc *bg, ext4_fsblk_t blk)
212 {
213 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
214 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
215 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
216 }
217
218 void ext4_inode_bitmap_set(struct super_block *sb,
219 struct ext4_group_desc *bg, ext4_fsblk_t blk)
220 {
221 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
222 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
223 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
224 }
225
226 void ext4_inode_table_set(struct super_block *sb,
227 struct ext4_group_desc *bg, ext4_fsblk_t blk)
228 {
229 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
230 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
231 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
232 }
233
234 void ext4_free_group_clusters_set(struct super_block *sb,
235 struct ext4_group_desc *bg, __u32 count)
236 {
237 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
238 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
240 }
241
242 void ext4_free_inodes_set(struct super_block *sb,
243 struct ext4_group_desc *bg, __u32 count)
244 {
245 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
248 }
249
250 void ext4_used_dirs_set(struct super_block *sb,
251 struct ext4_group_desc *bg, __u32 count)
252 {
253 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
256 }
257
258 void ext4_itable_unused_set(struct super_block *sb,
259 struct ext4_group_desc *bg, __u32 count)
260 {
261 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
264 }
265
266
267 /* Just increment the non-pointer handle value */
268 static handle_t *ext4_get_nojournal(void)
269 {
270 handle_t *handle = current->journal_info;
271 unsigned long ref_cnt = (unsigned long)handle;
272
273 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
274
275 ref_cnt++;
276 handle = (handle_t *)ref_cnt;
277
278 current->journal_info = handle;
279 return handle;
280 }
281
282
283 /* Decrement the non-pointer handle value */
284 static void ext4_put_nojournal(handle_t *handle)
285 {
286 unsigned long ref_cnt = (unsigned long)handle;
287
288 BUG_ON(ref_cnt == 0);
289
290 ref_cnt--;
291 handle = (handle_t *)ref_cnt;
292
293 current->journal_info = handle;
294 }
295
296 /*
297 * Wrappers for jbd2_journal_start/end.
298 *
299 * The only special thing we need to do here is to make sure that all
300 * journal_end calls result in the superblock being marked dirty, so
301 * that sync() will call the filesystem's write_super callback if
302 * appropriate.
303 *
304 * To avoid j_barrier hold in userspace when a user calls freeze(),
305 * ext4 prevents a new handle from being started by s_frozen, which
306 * is in an upper layer.
307 */
308 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
309 {
310 journal_t *journal;
311 handle_t *handle;
312
313 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
314 if (sb->s_flags & MS_RDONLY)
315 return ERR_PTR(-EROFS);
316
317 journal = EXT4_SB(sb)->s_journal;
318 handle = ext4_journal_current_handle();
319
320 /*
321 * If a handle has been started, it should be allowed to
322 * finish, otherwise deadlock could happen between freeze
323 * and others(e.g. truncate) due to the restart of the
324 * journal handle if the filesystem is forzen and active
325 * handles are not stopped.
326 */
327 if (!handle)
328 vfs_check_frozen(sb, SB_FREEZE_TRANS);
329
330 if (!journal)
331 return ext4_get_nojournal();
332 /*
333 * Special case here: if the journal has aborted behind our
334 * backs (eg. EIO in the commit thread), then we still need to
335 * take the FS itself readonly cleanly.
336 */
337 if (is_journal_aborted(journal)) {
338 ext4_abort(sb, "Detected aborted journal");
339 return ERR_PTR(-EROFS);
340 }
341 return jbd2_journal_start(journal, nblocks);
342 }
343
344 /*
345 * The only special thing we need to do here is to make sure that all
346 * jbd2_journal_stop calls result in the superblock being marked dirty, so
347 * that sync() will call the filesystem's write_super callback if
348 * appropriate.
349 */
350 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
351 {
352 struct super_block *sb;
353 int err;
354 int rc;
355
356 if (!ext4_handle_valid(handle)) {
357 ext4_put_nojournal(handle);
358 return 0;
359 }
360 sb = handle->h_transaction->t_journal->j_private;
361 err = handle->h_err;
362 rc = jbd2_journal_stop(handle);
363
364 if (!err)
365 err = rc;
366 if (err)
367 __ext4_std_error(sb, where, line, err);
368 return err;
369 }
370
371 void ext4_journal_abort_handle(const char *caller, unsigned int line,
372 const char *err_fn, struct buffer_head *bh,
373 handle_t *handle, int err)
374 {
375 char nbuf[16];
376 const char *errstr = ext4_decode_error(NULL, err, nbuf);
377
378 BUG_ON(!ext4_handle_valid(handle));
379
380 if (bh)
381 BUFFER_TRACE(bh, "abort");
382
383 if (!handle->h_err)
384 handle->h_err = err;
385
386 if (is_handle_aborted(handle))
387 return;
388
389 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
390 caller, line, errstr, err_fn);
391
392 jbd2_journal_abort_handle(handle);
393 }
394
395 static void __save_error_info(struct super_block *sb, const char *func,
396 unsigned int line)
397 {
398 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
399
400 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
401 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
402 es->s_last_error_time = cpu_to_le32(get_seconds());
403 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
404 es->s_last_error_line = cpu_to_le32(line);
405 if (!es->s_first_error_time) {
406 es->s_first_error_time = es->s_last_error_time;
407 strncpy(es->s_first_error_func, func,
408 sizeof(es->s_first_error_func));
409 es->s_first_error_line = cpu_to_le32(line);
410 es->s_first_error_ino = es->s_last_error_ino;
411 es->s_first_error_block = es->s_last_error_block;
412 }
413 /*
414 * Start the daily error reporting function if it hasn't been
415 * started already
416 */
417 if (!es->s_error_count)
418 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
419 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
420 }
421
422 static void save_error_info(struct super_block *sb, const char *func,
423 unsigned int line)
424 {
425 __save_error_info(sb, func, line);
426 ext4_commit_super(sb, 1);
427 }
428
429 /*
430 * The del_gendisk() function uninitializes the disk-specific data
431 * structures, including the bdi structure, without telling anyone
432 * else. Once this happens, any attempt to call mark_buffer_dirty()
433 * (for example, by ext4_commit_super), will cause a kernel OOPS.
434 * This is a kludge to prevent these oops until we can put in a proper
435 * hook in del_gendisk() to inform the VFS and file system layers.
436 */
437 static int block_device_ejected(struct super_block *sb)
438 {
439 struct inode *bd_inode = sb->s_bdev->bd_inode;
440 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
441
442 return bdi->dev == NULL;
443 }
444
445 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
446 {
447 struct super_block *sb = journal->j_private;
448 struct ext4_sb_info *sbi = EXT4_SB(sb);
449 int error = is_journal_aborted(journal);
450 struct ext4_journal_cb_entry *jce, *tmp;
451
452 spin_lock(&sbi->s_md_lock);
453 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
454 list_del_init(&jce->jce_list);
455 spin_unlock(&sbi->s_md_lock);
456 jce->jce_func(sb, jce, error);
457 spin_lock(&sbi->s_md_lock);
458 }
459 spin_unlock(&sbi->s_md_lock);
460 }
461
462 /* Deal with the reporting of failure conditions on a filesystem such as
463 * inconsistencies detected or read IO failures.
464 *
465 * On ext2, we can store the error state of the filesystem in the
466 * superblock. That is not possible on ext4, because we may have other
467 * write ordering constraints on the superblock which prevent us from
468 * writing it out straight away; and given that the journal is about to
469 * be aborted, we can't rely on the current, or future, transactions to
470 * write out the superblock safely.
471 *
472 * We'll just use the jbd2_journal_abort() error code to record an error in
473 * the journal instead. On recovery, the journal will complain about
474 * that error until we've noted it down and cleared it.
475 */
476
477 static void ext4_handle_error(struct super_block *sb)
478 {
479 if (sb->s_flags & MS_RDONLY)
480 return;
481
482 if (!test_opt(sb, ERRORS_CONT)) {
483 journal_t *journal = EXT4_SB(sb)->s_journal;
484
485 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
486 if (journal)
487 jbd2_journal_abort(journal, -EIO);
488 }
489 if (test_opt(sb, ERRORS_RO)) {
490 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
491 sb->s_flags |= MS_RDONLY;
492 }
493 if (test_opt(sb, ERRORS_PANIC))
494 panic("EXT4-fs (device %s): panic forced after error\n",
495 sb->s_id);
496 }
497
498 void __ext4_error(struct super_block *sb, const char *function,
499 unsigned int line, const char *fmt, ...)
500 {
501 struct va_format vaf;
502 va_list args;
503
504 va_start(args, fmt);
505 vaf.fmt = fmt;
506 vaf.va = &args;
507 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
508 sb->s_id, function, line, current->comm, &vaf);
509 va_end(args);
510
511 ext4_handle_error(sb);
512 }
513
514 void ext4_error_inode(struct inode *inode, const char *function,
515 unsigned int line, ext4_fsblk_t block,
516 const char *fmt, ...)
517 {
518 va_list args;
519 struct va_format vaf;
520 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
521
522 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
523 es->s_last_error_block = cpu_to_le64(block);
524 save_error_info(inode->i_sb, function, line);
525 va_start(args, fmt);
526 vaf.fmt = fmt;
527 vaf.va = &args;
528 if (block)
529 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
530 "inode #%lu: block %llu: comm %s: %pV\n",
531 inode->i_sb->s_id, function, line, inode->i_ino,
532 block, current->comm, &vaf);
533 else
534 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
535 "inode #%lu: comm %s: %pV\n",
536 inode->i_sb->s_id, function, line, inode->i_ino,
537 current->comm, &vaf);
538 va_end(args);
539
540 ext4_handle_error(inode->i_sb);
541 }
542
543 void ext4_error_file(struct file *file, const char *function,
544 unsigned int line, ext4_fsblk_t block,
545 const char *fmt, ...)
546 {
547 va_list args;
548 struct va_format vaf;
549 struct ext4_super_block *es;
550 struct inode *inode = file->f_dentry->d_inode;
551 char pathname[80], *path;
552
553 es = EXT4_SB(inode->i_sb)->s_es;
554 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
555 save_error_info(inode->i_sb, function, line);
556 path = d_path(&(file->f_path), pathname, sizeof(pathname));
557 if (IS_ERR(path))
558 path = "(unknown)";
559 va_start(args, fmt);
560 vaf.fmt = fmt;
561 vaf.va = &args;
562 if (block)
563 printk(KERN_CRIT
564 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
565 "block %llu: comm %s: path %s: %pV\n",
566 inode->i_sb->s_id, function, line, inode->i_ino,
567 block, current->comm, path, &vaf);
568 else
569 printk(KERN_CRIT
570 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
571 "comm %s: path %s: %pV\n",
572 inode->i_sb->s_id, function, line, inode->i_ino,
573 current->comm, path, &vaf);
574 va_end(args);
575
576 ext4_handle_error(inode->i_sb);
577 }
578
579 static const char *ext4_decode_error(struct super_block *sb, int errno,
580 char nbuf[16])
581 {
582 char *errstr = NULL;
583
584 switch (errno) {
585 case -EIO:
586 errstr = "IO failure";
587 break;
588 case -ENOMEM:
589 errstr = "Out of memory";
590 break;
591 case -EROFS:
592 if (!sb || (EXT4_SB(sb)->s_journal &&
593 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
594 errstr = "Journal has aborted";
595 else
596 errstr = "Readonly filesystem";
597 break;
598 default:
599 /* If the caller passed in an extra buffer for unknown
600 * errors, textualise them now. Else we just return
601 * NULL. */
602 if (nbuf) {
603 /* Check for truncated error codes... */
604 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
605 errstr = nbuf;
606 }
607 break;
608 }
609
610 return errstr;
611 }
612
613 /* __ext4_std_error decodes expected errors from journaling functions
614 * automatically and invokes the appropriate error response. */
615
616 void __ext4_std_error(struct super_block *sb, const char *function,
617 unsigned int line, int errno)
618 {
619 char nbuf[16];
620 const char *errstr;
621
622 /* Special case: if the error is EROFS, and we're not already
623 * inside a transaction, then there's really no point in logging
624 * an error. */
625 if (errno == -EROFS && journal_current_handle() == NULL &&
626 (sb->s_flags & MS_RDONLY))
627 return;
628
629 errstr = ext4_decode_error(sb, errno, nbuf);
630 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
631 sb->s_id, function, line, errstr);
632 save_error_info(sb, function, line);
633
634 ext4_handle_error(sb);
635 }
636
637 /*
638 * ext4_abort is a much stronger failure handler than ext4_error. The
639 * abort function may be used to deal with unrecoverable failures such
640 * as journal IO errors or ENOMEM at a critical moment in log management.
641 *
642 * We unconditionally force the filesystem into an ABORT|READONLY state,
643 * unless the error response on the fs has been set to panic in which
644 * case we take the easy way out and panic immediately.
645 */
646
647 void __ext4_abort(struct super_block *sb, const char *function,
648 unsigned int line, const char *fmt, ...)
649 {
650 va_list args;
651
652 save_error_info(sb, function, line);
653 va_start(args, fmt);
654 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
655 function, line);
656 vprintk(fmt, args);
657 printk("\n");
658 va_end(args);
659
660 if ((sb->s_flags & MS_RDONLY) == 0) {
661 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
662 sb->s_flags |= MS_RDONLY;
663 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
664 if (EXT4_SB(sb)->s_journal)
665 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
666 save_error_info(sb, function, line);
667 }
668 if (test_opt(sb, ERRORS_PANIC))
669 panic("EXT4-fs panic from previous error\n");
670 }
671
672 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
673 {
674 struct va_format vaf;
675 va_list args;
676
677 va_start(args, fmt);
678 vaf.fmt = fmt;
679 vaf.va = &args;
680 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
681 va_end(args);
682 }
683
684 void __ext4_warning(struct super_block *sb, const char *function,
685 unsigned int line, const char *fmt, ...)
686 {
687 struct va_format vaf;
688 va_list args;
689
690 va_start(args, fmt);
691 vaf.fmt = fmt;
692 vaf.va = &args;
693 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
694 sb->s_id, function, line, &vaf);
695 va_end(args);
696 }
697
698 void __ext4_grp_locked_error(const char *function, unsigned int line,
699 struct super_block *sb, ext4_group_t grp,
700 unsigned long ino, ext4_fsblk_t block,
701 const char *fmt, ...)
702 __releases(bitlock)
703 __acquires(bitlock)
704 {
705 struct va_format vaf;
706 va_list args;
707 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
708
709 es->s_last_error_ino = cpu_to_le32(ino);
710 es->s_last_error_block = cpu_to_le64(block);
711 __save_error_info(sb, function, line);
712
713 va_start(args, fmt);
714
715 vaf.fmt = fmt;
716 vaf.va = &args;
717 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
718 sb->s_id, function, line, grp);
719 if (ino)
720 printk(KERN_CONT "inode %lu: ", ino);
721 if (block)
722 printk(KERN_CONT "block %llu:", (unsigned long long) block);
723 printk(KERN_CONT "%pV\n", &vaf);
724 va_end(args);
725
726 if (test_opt(sb, ERRORS_CONT)) {
727 ext4_commit_super(sb, 0);
728 return;
729 }
730
731 ext4_unlock_group(sb, grp);
732 ext4_handle_error(sb);
733 /*
734 * We only get here in the ERRORS_RO case; relocking the group
735 * may be dangerous, but nothing bad will happen since the
736 * filesystem will have already been marked read/only and the
737 * journal has been aborted. We return 1 as a hint to callers
738 * who might what to use the return value from
739 * ext4_grp_locked_error() to distinguish between the
740 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
741 * aggressively from the ext4 function in question, with a
742 * more appropriate error code.
743 */
744 ext4_lock_group(sb, grp);
745 return;
746 }
747
748 void ext4_update_dynamic_rev(struct super_block *sb)
749 {
750 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
751
752 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
753 return;
754
755 ext4_warning(sb,
756 "updating to rev %d because of new feature flag, "
757 "running e2fsck is recommended",
758 EXT4_DYNAMIC_REV);
759
760 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
761 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
762 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
763 /* leave es->s_feature_*compat flags alone */
764 /* es->s_uuid will be set by e2fsck if empty */
765
766 /*
767 * The rest of the superblock fields should be zero, and if not it
768 * means they are likely already in use, so leave them alone. We
769 * can leave it up to e2fsck to clean up any inconsistencies there.
770 */
771 }
772
773 /*
774 * Open the external journal device
775 */
776 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
777 {
778 struct block_device *bdev;
779 char b[BDEVNAME_SIZE];
780
781 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
782 if (IS_ERR(bdev))
783 goto fail;
784 return bdev;
785
786 fail:
787 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
788 __bdevname(dev, b), PTR_ERR(bdev));
789 return NULL;
790 }
791
792 /*
793 * Release the journal device
794 */
795 static int ext4_blkdev_put(struct block_device *bdev)
796 {
797 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
798 }
799
800 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
801 {
802 struct block_device *bdev;
803 int ret = -ENODEV;
804
805 bdev = sbi->journal_bdev;
806 if (bdev) {
807 ret = ext4_blkdev_put(bdev);
808 sbi->journal_bdev = NULL;
809 }
810 return ret;
811 }
812
813 static inline struct inode *orphan_list_entry(struct list_head *l)
814 {
815 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
816 }
817
818 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
819 {
820 struct list_head *l;
821
822 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
823 le32_to_cpu(sbi->s_es->s_last_orphan));
824
825 printk(KERN_ERR "sb_info orphan list:\n");
826 list_for_each(l, &sbi->s_orphan) {
827 struct inode *inode = orphan_list_entry(l);
828 printk(KERN_ERR " "
829 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
830 inode->i_sb->s_id, inode->i_ino, inode,
831 inode->i_mode, inode->i_nlink,
832 NEXT_ORPHAN(inode));
833 }
834 }
835
836 static void ext4_put_super(struct super_block *sb)
837 {
838 struct ext4_sb_info *sbi = EXT4_SB(sb);
839 struct ext4_super_block *es = sbi->s_es;
840 int i, err;
841
842 ext4_unregister_li_request(sb);
843 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
844
845 flush_workqueue(sbi->dio_unwritten_wq);
846 destroy_workqueue(sbi->dio_unwritten_wq);
847
848 lock_super(sb);
849 if (sbi->s_journal) {
850 err = jbd2_journal_destroy(sbi->s_journal);
851 sbi->s_journal = NULL;
852 if (err < 0)
853 ext4_abort(sb, "Couldn't clean up the journal");
854 }
855
856 del_timer(&sbi->s_err_report);
857 ext4_release_system_zone(sb);
858 ext4_mb_release(sb);
859 ext4_ext_release(sb);
860 ext4_xattr_put_super(sb);
861
862 if (!(sb->s_flags & MS_RDONLY)) {
863 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
864 es->s_state = cpu_to_le16(sbi->s_mount_state);
865 }
866 if (sb->s_dirt || !(sb->s_flags & MS_RDONLY))
867 ext4_commit_super(sb, 1);
868
869 if (sbi->s_proc) {
870 remove_proc_entry("options", sbi->s_proc);
871 remove_proc_entry(sb->s_id, ext4_proc_root);
872 }
873 kobject_del(&sbi->s_kobj);
874
875 for (i = 0; i < sbi->s_gdb_count; i++)
876 brelse(sbi->s_group_desc[i]);
877 ext4_kvfree(sbi->s_group_desc);
878 ext4_kvfree(sbi->s_flex_groups);
879 percpu_counter_destroy(&sbi->s_freeclusters_counter);
880 percpu_counter_destroy(&sbi->s_freeinodes_counter);
881 percpu_counter_destroy(&sbi->s_dirs_counter);
882 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
883 brelse(sbi->s_sbh);
884 #ifdef CONFIG_QUOTA
885 for (i = 0; i < MAXQUOTAS; i++)
886 kfree(sbi->s_qf_names[i]);
887 #endif
888
889 /* Debugging code just in case the in-memory inode orphan list
890 * isn't empty. The on-disk one can be non-empty if we've
891 * detected an error and taken the fs readonly, but the
892 * in-memory list had better be clean by this point. */
893 if (!list_empty(&sbi->s_orphan))
894 dump_orphan_list(sb, sbi);
895 J_ASSERT(list_empty(&sbi->s_orphan));
896
897 invalidate_bdev(sb->s_bdev);
898 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
899 /*
900 * Invalidate the journal device's buffers. We don't want them
901 * floating about in memory - the physical journal device may
902 * hotswapped, and it breaks the `ro-after' testing code.
903 */
904 sync_blockdev(sbi->journal_bdev);
905 invalidate_bdev(sbi->journal_bdev);
906 ext4_blkdev_remove(sbi);
907 }
908 if (sbi->s_mmp_tsk)
909 kthread_stop(sbi->s_mmp_tsk);
910 sb->s_fs_info = NULL;
911 /*
912 * Now that we are completely done shutting down the
913 * superblock, we need to actually destroy the kobject.
914 */
915 unlock_super(sb);
916 kobject_put(&sbi->s_kobj);
917 wait_for_completion(&sbi->s_kobj_unregister);
918 kfree(sbi->s_blockgroup_lock);
919 kfree(sbi);
920 }
921
922 static struct kmem_cache *ext4_inode_cachep;
923
924 /*
925 * Called inside transaction, so use GFP_NOFS
926 */
927 static struct inode *ext4_alloc_inode(struct super_block *sb)
928 {
929 struct ext4_inode_info *ei;
930
931 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
932 if (!ei)
933 return NULL;
934
935 ei->vfs_inode.i_version = 1;
936 ei->vfs_inode.i_data.writeback_index = 0;
937 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
938 INIT_LIST_HEAD(&ei->i_prealloc_list);
939 spin_lock_init(&ei->i_prealloc_lock);
940 ei->i_reserved_data_blocks = 0;
941 ei->i_reserved_meta_blocks = 0;
942 ei->i_allocated_meta_blocks = 0;
943 ei->i_da_metadata_calc_len = 0;
944 spin_lock_init(&(ei->i_block_reservation_lock));
945 #ifdef CONFIG_QUOTA
946 ei->i_reserved_quota = 0;
947 #endif
948 ei->jinode = NULL;
949 INIT_LIST_HEAD(&ei->i_completed_io_list);
950 spin_lock_init(&ei->i_completed_io_lock);
951 ei->cur_aio_dio = NULL;
952 ei->i_sync_tid = 0;
953 ei->i_datasync_tid = 0;
954 atomic_set(&ei->i_ioend_count, 0);
955 atomic_set(&ei->i_aiodio_unwritten, 0);
956
957 return &ei->vfs_inode;
958 }
959
960 static int ext4_drop_inode(struct inode *inode)
961 {
962 int drop = generic_drop_inode(inode);
963
964 trace_ext4_drop_inode(inode, drop);
965 return drop;
966 }
967
968 static void ext4_i_callback(struct rcu_head *head)
969 {
970 struct inode *inode = container_of(head, struct inode, i_rcu);
971 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
972 }
973
974 static void ext4_destroy_inode(struct inode *inode)
975 {
976 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
977 ext4_msg(inode->i_sb, KERN_ERR,
978 "Inode %lu (%p): orphan list check failed!",
979 inode->i_ino, EXT4_I(inode));
980 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
981 EXT4_I(inode), sizeof(struct ext4_inode_info),
982 true);
983 dump_stack();
984 }
985 call_rcu(&inode->i_rcu, ext4_i_callback);
986 }
987
988 static void init_once(void *foo)
989 {
990 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
991
992 INIT_LIST_HEAD(&ei->i_orphan);
993 #ifdef CONFIG_EXT4_FS_XATTR
994 init_rwsem(&ei->xattr_sem);
995 #endif
996 init_rwsem(&ei->i_data_sem);
997 inode_init_once(&ei->vfs_inode);
998 }
999
1000 static int init_inodecache(void)
1001 {
1002 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1003 sizeof(struct ext4_inode_info),
1004 0, (SLAB_RECLAIM_ACCOUNT|
1005 SLAB_MEM_SPREAD),
1006 init_once);
1007 if (ext4_inode_cachep == NULL)
1008 return -ENOMEM;
1009 return 0;
1010 }
1011
1012 static void destroy_inodecache(void)
1013 {
1014 kmem_cache_destroy(ext4_inode_cachep);
1015 }
1016
1017 void ext4_clear_inode(struct inode *inode)
1018 {
1019 invalidate_inode_buffers(inode);
1020 end_writeback(inode);
1021 dquot_drop(inode);
1022 ext4_discard_preallocations(inode);
1023 if (EXT4_I(inode)->jinode) {
1024 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1025 EXT4_I(inode)->jinode);
1026 jbd2_free_inode(EXT4_I(inode)->jinode);
1027 EXT4_I(inode)->jinode = NULL;
1028 }
1029 }
1030
1031 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1032 u64 ino, u32 generation)
1033 {
1034 struct inode *inode;
1035
1036 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1037 return ERR_PTR(-ESTALE);
1038 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1039 return ERR_PTR(-ESTALE);
1040
1041 /* iget isn't really right if the inode is currently unallocated!!
1042 *
1043 * ext4_read_inode will return a bad_inode if the inode had been
1044 * deleted, so we should be safe.
1045 *
1046 * Currently we don't know the generation for parent directory, so
1047 * a generation of 0 means "accept any"
1048 */
1049 inode = ext4_iget(sb, ino);
1050 if (IS_ERR(inode))
1051 return ERR_CAST(inode);
1052 if (generation && inode->i_generation != generation) {
1053 iput(inode);
1054 return ERR_PTR(-ESTALE);
1055 }
1056
1057 return inode;
1058 }
1059
1060 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1061 int fh_len, int fh_type)
1062 {
1063 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1064 ext4_nfs_get_inode);
1065 }
1066
1067 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1068 int fh_len, int fh_type)
1069 {
1070 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1071 ext4_nfs_get_inode);
1072 }
1073
1074 /*
1075 * Try to release metadata pages (indirect blocks, directories) which are
1076 * mapped via the block device. Since these pages could have journal heads
1077 * which would prevent try_to_free_buffers() from freeing them, we must use
1078 * jbd2 layer's try_to_free_buffers() function to release them.
1079 */
1080 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1081 gfp_t wait)
1082 {
1083 journal_t *journal = EXT4_SB(sb)->s_journal;
1084
1085 WARN_ON(PageChecked(page));
1086 if (!page_has_buffers(page))
1087 return 0;
1088 if (journal)
1089 return jbd2_journal_try_to_free_buffers(journal, page,
1090 wait & ~__GFP_WAIT);
1091 return try_to_free_buffers(page);
1092 }
1093
1094 #ifdef CONFIG_QUOTA
1095 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1096 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1097
1098 static int ext4_write_dquot(struct dquot *dquot);
1099 static int ext4_acquire_dquot(struct dquot *dquot);
1100 static int ext4_release_dquot(struct dquot *dquot);
1101 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1102 static int ext4_write_info(struct super_block *sb, int type);
1103 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1104 struct path *path);
1105 static int ext4_quota_off(struct super_block *sb, int type);
1106 static int ext4_quota_on_mount(struct super_block *sb, int type);
1107 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1108 size_t len, loff_t off);
1109 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1110 const char *data, size_t len, loff_t off);
1111
1112 static const struct dquot_operations ext4_quota_operations = {
1113 .get_reserved_space = ext4_get_reserved_space,
1114 .write_dquot = ext4_write_dquot,
1115 .acquire_dquot = ext4_acquire_dquot,
1116 .release_dquot = ext4_release_dquot,
1117 .mark_dirty = ext4_mark_dquot_dirty,
1118 .write_info = ext4_write_info,
1119 .alloc_dquot = dquot_alloc,
1120 .destroy_dquot = dquot_destroy,
1121 };
1122
1123 static const struct quotactl_ops ext4_qctl_operations = {
1124 .quota_on = ext4_quota_on,
1125 .quota_off = ext4_quota_off,
1126 .quota_sync = dquot_quota_sync,
1127 .get_info = dquot_get_dqinfo,
1128 .set_info = dquot_set_dqinfo,
1129 .get_dqblk = dquot_get_dqblk,
1130 .set_dqblk = dquot_set_dqblk
1131 };
1132 #endif
1133
1134 static const struct super_operations ext4_sops = {
1135 .alloc_inode = ext4_alloc_inode,
1136 .destroy_inode = ext4_destroy_inode,
1137 .write_inode = ext4_write_inode,
1138 .dirty_inode = ext4_dirty_inode,
1139 .drop_inode = ext4_drop_inode,
1140 .evict_inode = ext4_evict_inode,
1141 .put_super = ext4_put_super,
1142 .sync_fs = ext4_sync_fs,
1143 .freeze_fs = ext4_freeze,
1144 .unfreeze_fs = ext4_unfreeze,
1145 .statfs = ext4_statfs,
1146 .remount_fs = ext4_remount,
1147 .show_options = ext4_show_options,
1148 #ifdef CONFIG_QUOTA
1149 .quota_read = ext4_quota_read,
1150 .quota_write = ext4_quota_write,
1151 #endif
1152 .bdev_try_to_free_page = bdev_try_to_free_page,
1153 };
1154
1155 static const struct super_operations ext4_nojournal_sops = {
1156 .alloc_inode = ext4_alloc_inode,
1157 .destroy_inode = ext4_destroy_inode,
1158 .write_inode = ext4_write_inode,
1159 .dirty_inode = ext4_dirty_inode,
1160 .drop_inode = ext4_drop_inode,
1161 .evict_inode = ext4_evict_inode,
1162 .write_super = ext4_write_super,
1163 .put_super = ext4_put_super,
1164 .statfs = ext4_statfs,
1165 .remount_fs = ext4_remount,
1166 .show_options = ext4_show_options,
1167 #ifdef CONFIG_QUOTA
1168 .quota_read = ext4_quota_read,
1169 .quota_write = ext4_quota_write,
1170 #endif
1171 .bdev_try_to_free_page = bdev_try_to_free_page,
1172 };
1173
1174 static const struct export_operations ext4_export_ops = {
1175 .fh_to_dentry = ext4_fh_to_dentry,
1176 .fh_to_parent = ext4_fh_to_parent,
1177 .get_parent = ext4_get_parent,
1178 };
1179
1180 enum {
1181 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1182 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1183 Opt_nouid32, Opt_debug, Opt_removed,
1184 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1185 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1186 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1187 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1188 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1189 Opt_data_err_abort, Opt_data_err_ignore,
1190 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1191 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1192 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1193 Opt_usrquota, Opt_grpquota, Opt_i_version,
1194 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1195 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1196 Opt_inode_readahead_blks, Opt_journal_ioprio,
1197 Opt_dioread_nolock, Opt_dioread_lock,
1198 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1199 };
1200
1201 static const match_table_t tokens = {
1202 {Opt_bsd_df, "bsddf"},
1203 {Opt_minix_df, "minixdf"},
1204 {Opt_grpid, "grpid"},
1205 {Opt_grpid, "bsdgroups"},
1206 {Opt_nogrpid, "nogrpid"},
1207 {Opt_nogrpid, "sysvgroups"},
1208 {Opt_resgid, "resgid=%u"},
1209 {Opt_resuid, "resuid=%u"},
1210 {Opt_sb, "sb=%u"},
1211 {Opt_err_cont, "errors=continue"},
1212 {Opt_err_panic, "errors=panic"},
1213 {Opt_err_ro, "errors=remount-ro"},
1214 {Opt_nouid32, "nouid32"},
1215 {Opt_debug, "debug"},
1216 {Opt_removed, "oldalloc"},
1217 {Opt_removed, "orlov"},
1218 {Opt_user_xattr, "user_xattr"},
1219 {Opt_nouser_xattr, "nouser_xattr"},
1220 {Opt_acl, "acl"},
1221 {Opt_noacl, "noacl"},
1222 {Opt_noload, "norecovery"},
1223 {Opt_noload, "noload"},
1224 {Opt_removed, "nobh"},
1225 {Opt_removed, "bh"},
1226 {Opt_commit, "commit=%u"},
1227 {Opt_min_batch_time, "min_batch_time=%u"},
1228 {Opt_max_batch_time, "max_batch_time=%u"},
1229 {Opt_journal_dev, "journal_dev=%u"},
1230 {Opt_journal_checksum, "journal_checksum"},
1231 {Opt_journal_async_commit, "journal_async_commit"},
1232 {Opt_abort, "abort"},
1233 {Opt_data_journal, "data=journal"},
1234 {Opt_data_ordered, "data=ordered"},
1235 {Opt_data_writeback, "data=writeback"},
1236 {Opt_data_err_abort, "data_err=abort"},
1237 {Opt_data_err_ignore, "data_err=ignore"},
1238 {Opt_offusrjquota, "usrjquota="},
1239 {Opt_usrjquota, "usrjquota=%s"},
1240 {Opt_offgrpjquota, "grpjquota="},
1241 {Opt_grpjquota, "grpjquota=%s"},
1242 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1243 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1244 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1245 {Opt_grpquota, "grpquota"},
1246 {Opt_noquota, "noquota"},
1247 {Opt_quota, "quota"},
1248 {Opt_usrquota, "usrquota"},
1249 {Opt_barrier, "barrier=%u"},
1250 {Opt_barrier, "barrier"},
1251 {Opt_nobarrier, "nobarrier"},
1252 {Opt_i_version, "i_version"},
1253 {Opt_stripe, "stripe=%u"},
1254 {Opt_delalloc, "delalloc"},
1255 {Opt_nodelalloc, "nodelalloc"},
1256 {Opt_mblk_io_submit, "mblk_io_submit"},
1257 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1258 {Opt_block_validity, "block_validity"},
1259 {Opt_noblock_validity, "noblock_validity"},
1260 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1261 {Opt_journal_ioprio, "journal_ioprio=%u"},
1262 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1263 {Opt_auto_da_alloc, "auto_da_alloc"},
1264 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1265 {Opt_dioread_nolock, "dioread_nolock"},
1266 {Opt_dioread_lock, "dioread_lock"},
1267 {Opt_discard, "discard"},
1268 {Opt_nodiscard, "nodiscard"},
1269 {Opt_init_itable, "init_itable=%u"},
1270 {Opt_init_itable, "init_itable"},
1271 {Opt_noinit_itable, "noinit_itable"},
1272 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1273 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1274 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1275 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1276 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1277 {Opt_err, NULL},
1278 };
1279
1280 static ext4_fsblk_t get_sb_block(void **data)
1281 {
1282 ext4_fsblk_t sb_block;
1283 char *options = (char *) *data;
1284
1285 if (!options || strncmp(options, "sb=", 3) != 0)
1286 return 1; /* Default location */
1287
1288 options += 3;
1289 /* TODO: use simple_strtoll with >32bit ext4 */
1290 sb_block = simple_strtoul(options, &options, 0);
1291 if (*options && *options != ',') {
1292 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1293 (char *) *data);
1294 return 1;
1295 }
1296 if (*options == ',')
1297 options++;
1298 *data = (void *) options;
1299
1300 return sb_block;
1301 }
1302
1303 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1304 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1305 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1306
1307 #ifdef CONFIG_QUOTA
1308 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1309 {
1310 struct ext4_sb_info *sbi = EXT4_SB(sb);
1311 char *qname;
1312
1313 if (sb_any_quota_loaded(sb) &&
1314 !sbi->s_qf_names[qtype]) {
1315 ext4_msg(sb, KERN_ERR,
1316 "Cannot change journaled "
1317 "quota options when quota turned on");
1318 return -1;
1319 }
1320 qname = match_strdup(args);
1321 if (!qname) {
1322 ext4_msg(sb, KERN_ERR,
1323 "Not enough memory for storing quotafile name");
1324 return -1;
1325 }
1326 if (sbi->s_qf_names[qtype] &&
1327 strcmp(sbi->s_qf_names[qtype], qname)) {
1328 ext4_msg(sb, KERN_ERR,
1329 "%s quota file already specified", QTYPE2NAME(qtype));
1330 kfree(qname);
1331 return -1;
1332 }
1333 sbi->s_qf_names[qtype] = qname;
1334 if (strchr(sbi->s_qf_names[qtype], '/')) {
1335 ext4_msg(sb, KERN_ERR,
1336 "quotafile must be on filesystem root");
1337 kfree(sbi->s_qf_names[qtype]);
1338 sbi->s_qf_names[qtype] = NULL;
1339 return -1;
1340 }
1341 set_opt(sb, QUOTA);
1342 return 1;
1343 }
1344
1345 static int clear_qf_name(struct super_block *sb, int qtype)
1346 {
1347
1348 struct ext4_sb_info *sbi = EXT4_SB(sb);
1349
1350 if (sb_any_quota_loaded(sb) &&
1351 sbi->s_qf_names[qtype]) {
1352 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1353 " when quota turned on");
1354 return -1;
1355 }
1356 /*
1357 * The space will be released later when all options are confirmed
1358 * to be correct
1359 */
1360 sbi->s_qf_names[qtype] = NULL;
1361 return 1;
1362 }
1363 #endif
1364
1365 #define MOPT_SET 0x0001
1366 #define MOPT_CLEAR 0x0002
1367 #define MOPT_NOSUPPORT 0x0004
1368 #define MOPT_EXPLICIT 0x0008
1369 #define MOPT_CLEAR_ERR 0x0010
1370 #define MOPT_GTE0 0x0020
1371 #ifdef CONFIG_QUOTA
1372 #define MOPT_Q 0
1373 #define MOPT_QFMT 0x0040
1374 #else
1375 #define MOPT_Q MOPT_NOSUPPORT
1376 #define MOPT_QFMT MOPT_NOSUPPORT
1377 #endif
1378 #define MOPT_DATAJ 0x0080
1379
1380 static const struct mount_opts {
1381 int token;
1382 int mount_opt;
1383 int flags;
1384 } ext4_mount_opts[] = {
1385 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1386 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1387 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1388 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1389 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1390 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1391 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1392 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1393 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1394 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1395 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1396 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1397 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1398 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1399 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1400 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1401 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1402 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1403 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1404 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1405 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1406 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1407 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1408 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1409 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1410 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1411 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1412 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1413 {Opt_commit, 0, MOPT_GTE0},
1414 {Opt_max_batch_time, 0, MOPT_GTE0},
1415 {Opt_min_batch_time, 0, MOPT_GTE0},
1416 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1417 {Opt_init_itable, 0, MOPT_GTE0},
1418 {Opt_stripe, 0, MOPT_GTE0},
1419 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1420 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1421 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1422 #ifdef CONFIG_EXT4_FS_XATTR
1423 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1424 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1425 #else
1426 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1427 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
1428 #endif
1429 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1430 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1431 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1432 #else
1433 {Opt_acl, 0, MOPT_NOSUPPORT},
1434 {Opt_noacl, 0, MOPT_NOSUPPORT},
1435 #endif
1436 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1437 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1438 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1439 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1440 MOPT_SET | MOPT_Q},
1441 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1442 MOPT_SET | MOPT_Q},
1443 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1444 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1445 {Opt_usrjquota, 0, MOPT_Q},
1446 {Opt_grpjquota, 0, MOPT_Q},
1447 {Opt_offusrjquota, 0, MOPT_Q},
1448 {Opt_offgrpjquota, 0, MOPT_Q},
1449 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1450 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1451 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1452 {Opt_err, 0, 0}
1453 };
1454
1455 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1456 substring_t *args, unsigned long *journal_devnum,
1457 unsigned int *journal_ioprio, int is_remount)
1458 {
1459 struct ext4_sb_info *sbi = EXT4_SB(sb);
1460 const struct mount_opts *m;
1461 int arg = 0;
1462
1463 #ifdef CONFIG_QUOTA
1464 if (token == Opt_usrjquota)
1465 return set_qf_name(sb, USRQUOTA, &args[0]);
1466 else if (token == Opt_grpjquota)
1467 return set_qf_name(sb, GRPQUOTA, &args[0]);
1468 else if (token == Opt_offusrjquota)
1469 return clear_qf_name(sb, USRQUOTA);
1470 else if (token == Opt_offgrpjquota)
1471 return clear_qf_name(sb, GRPQUOTA);
1472 #endif
1473 if (args->from && match_int(args, &arg))
1474 return -1;
1475 switch (token) {
1476 case Opt_noacl:
1477 case Opt_nouser_xattr:
1478 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1479 break;
1480 case Opt_sb:
1481 return 1; /* handled by get_sb_block() */
1482 case Opt_removed:
1483 ext4_msg(sb, KERN_WARNING,
1484 "Ignoring removed %s option", opt);
1485 return 1;
1486 case Opt_resuid:
1487 sbi->s_resuid = arg;
1488 return 1;
1489 case Opt_resgid:
1490 sbi->s_resgid = arg;
1491 return 1;
1492 case Opt_abort:
1493 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1494 return 1;
1495 case Opt_i_version:
1496 sb->s_flags |= MS_I_VERSION;
1497 return 1;
1498 case Opt_journal_dev:
1499 if (is_remount) {
1500 ext4_msg(sb, KERN_ERR,
1501 "Cannot specify journal on remount");
1502 return -1;
1503 }
1504 *journal_devnum = arg;
1505 return 1;
1506 case Opt_journal_ioprio:
1507 if (arg < 0 || arg > 7)
1508 return -1;
1509 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1510 return 1;
1511 }
1512
1513 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1514 if (token != m->token)
1515 continue;
1516 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1517 return -1;
1518 if (m->flags & MOPT_EXPLICIT)
1519 set_opt2(sb, EXPLICIT_DELALLOC);
1520 if (m->flags & MOPT_CLEAR_ERR)
1521 clear_opt(sb, ERRORS_MASK);
1522 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1523 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1524 "options when quota turned on");
1525 return -1;
1526 }
1527
1528 if (m->flags & MOPT_NOSUPPORT) {
1529 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1530 } else if (token == Opt_commit) {
1531 if (arg == 0)
1532 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1533 sbi->s_commit_interval = HZ * arg;
1534 } else if (token == Opt_max_batch_time) {
1535 if (arg == 0)
1536 arg = EXT4_DEF_MAX_BATCH_TIME;
1537 sbi->s_max_batch_time = arg;
1538 } else if (token == Opt_min_batch_time) {
1539 sbi->s_min_batch_time = arg;
1540 } else if (token == Opt_inode_readahead_blks) {
1541 if (arg > (1 << 30))
1542 return -1;
1543 if (arg && !is_power_of_2(arg)) {
1544 ext4_msg(sb, KERN_ERR,
1545 "EXT4-fs: inode_readahead_blks"
1546 " must be a power of 2");
1547 return -1;
1548 }
1549 sbi->s_inode_readahead_blks = arg;
1550 } else if (token == Opt_init_itable) {
1551 set_opt(sb, INIT_INODE_TABLE);
1552 if (!args->from)
1553 arg = EXT4_DEF_LI_WAIT_MULT;
1554 sbi->s_li_wait_mult = arg;
1555 } else if (token == Opt_stripe) {
1556 sbi->s_stripe = arg;
1557 } else if (m->flags & MOPT_DATAJ) {
1558 if (is_remount) {
1559 if (!sbi->s_journal)
1560 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1561 else if (test_opt(sb, DATA_FLAGS) !=
1562 m->mount_opt) {
1563 ext4_msg(sb, KERN_ERR,
1564 "Cannot change data mode on remount");
1565 return -1;
1566 }
1567 } else {
1568 clear_opt(sb, DATA_FLAGS);
1569 sbi->s_mount_opt |= m->mount_opt;
1570 }
1571 #ifdef CONFIG_QUOTA
1572 } else if (m->flags & MOPT_QFMT) {
1573 if (sb_any_quota_loaded(sb) &&
1574 sbi->s_jquota_fmt != m->mount_opt) {
1575 ext4_msg(sb, KERN_ERR, "Cannot "
1576 "change journaled quota options "
1577 "when quota turned on");
1578 return -1;
1579 }
1580 sbi->s_jquota_fmt = m->mount_opt;
1581 #endif
1582 } else {
1583 if (!args->from)
1584 arg = 1;
1585 if (m->flags & MOPT_CLEAR)
1586 arg = !arg;
1587 else if (unlikely(!(m->flags & MOPT_SET))) {
1588 ext4_msg(sb, KERN_WARNING,
1589 "buggy handling of option %s", opt);
1590 WARN_ON(1);
1591 return -1;
1592 }
1593 if (arg != 0)
1594 sbi->s_mount_opt |= m->mount_opt;
1595 else
1596 sbi->s_mount_opt &= ~m->mount_opt;
1597 }
1598 return 1;
1599 }
1600 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1601 "or missing value", opt);
1602 return -1;
1603 }
1604
1605 static int parse_options(char *options, struct super_block *sb,
1606 unsigned long *journal_devnum,
1607 unsigned int *journal_ioprio,
1608 int is_remount)
1609 {
1610 #ifdef CONFIG_QUOTA
1611 struct ext4_sb_info *sbi = EXT4_SB(sb);
1612 #endif
1613 char *p;
1614 substring_t args[MAX_OPT_ARGS];
1615 int token;
1616
1617 if (!options)
1618 return 1;
1619
1620 while ((p = strsep(&options, ",")) != NULL) {
1621 if (!*p)
1622 continue;
1623 /*
1624 * Initialize args struct so we know whether arg was
1625 * found; some options take optional arguments.
1626 */
1627 args[0].to = args[0].from = 0;
1628 token = match_token(p, tokens, args);
1629 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1630 journal_ioprio, is_remount) < 0)
1631 return 0;
1632 }
1633 #ifdef CONFIG_QUOTA
1634 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1635 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1636 clear_opt(sb, USRQUOTA);
1637
1638 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1639 clear_opt(sb, GRPQUOTA);
1640
1641 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1642 ext4_msg(sb, KERN_ERR, "old and new quota "
1643 "format mixing");
1644 return 0;
1645 }
1646
1647 if (!sbi->s_jquota_fmt) {
1648 ext4_msg(sb, KERN_ERR, "journaled quota format "
1649 "not specified");
1650 return 0;
1651 }
1652 } else {
1653 if (sbi->s_jquota_fmt) {
1654 ext4_msg(sb, KERN_ERR, "journaled quota format "
1655 "specified with no journaling "
1656 "enabled");
1657 return 0;
1658 }
1659 }
1660 #endif
1661 return 1;
1662 }
1663
1664 static inline void ext4_show_quota_options(struct seq_file *seq,
1665 struct super_block *sb)
1666 {
1667 #if defined(CONFIG_QUOTA)
1668 struct ext4_sb_info *sbi = EXT4_SB(sb);
1669
1670 if (sbi->s_jquota_fmt) {
1671 char *fmtname = "";
1672
1673 switch (sbi->s_jquota_fmt) {
1674 case QFMT_VFS_OLD:
1675 fmtname = "vfsold";
1676 break;
1677 case QFMT_VFS_V0:
1678 fmtname = "vfsv0";
1679 break;
1680 case QFMT_VFS_V1:
1681 fmtname = "vfsv1";
1682 break;
1683 }
1684 seq_printf(seq, ",jqfmt=%s", fmtname);
1685 }
1686
1687 if (sbi->s_qf_names[USRQUOTA])
1688 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1689
1690 if (sbi->s_qf_names[GRPQUOTA])
1691 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1692
1693 if (test_opt(sb, USRQUOTA))
1694 seq_puts(seq, ",usrquota");
1695
1696 if (test_opt(sb, GRPQUOTA))
1697 seq_puts(seq, ",grpquota");
1698 #endif
1699 }
1700
1701 static const char *token2str(int token)
1702 {
1703 static const struct match_token *t;
1704
1705 for (t = tokens; t->token != Opt_err; t++)
1706 if (t->token == token && !strchr(t->pattern, '='))
1707 break;
1708 return t->pattern;
1709 }
1710
1711 /*
1712 * Show an option if
1713 * - it's set to a non-default value OR
1714 * - if the per-sb default is different from the global default
1715 */
1716 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1717 int nodefs)
1718 {
1719 struct ext4_sb_info *sbi = EXT4_SB(sb);
1720 struct ext4_super_block *es = sbi->s_es;
1721 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1722 const struct mount_opts *m;
1723 char sep = nodefs ? '\n' : ',';
1724
1725 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1726 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1727
1728 if (sbi->s_sb_block != 1)
1729 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1730
1731 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1732 int want_set = m->flags & MOPT_SET;
1733 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1734 (m->flags & MOPT_CLEAR_ERR))
1735 continue;
1736 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1737 continue; /* skip if same as the default */
1738 if ((want_set &&
1739 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1740 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1741 continue; /* select Opt_noFoo vs Opt_Foo */
1742 SEQ_OPTS_PRINT("%s", token2str(m->token));
1743 }
1744
1745 if (nodefs || sbi->s_resuid != EXT4_DEF_RESUID ||
1746 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1747 SEQ_OPTS_PRINT("resuid=%u", sbi->s_resuid);
1748 if (nodefs || sbi->s_resgid != EXT4_DEF_RESGID ||
1749 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1750 SEQ_OPTS_PRINT("resgid=%u", sbi->s_resgid);
1751 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1752 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1753 SEQ_OPTS_PUTS("errors=remount-ro");
1754 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1755 SEQ_OPTS_PUTS("errors=continue");
1756 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1757 SEQ_OPTS_PUTS("errors=panic");
1758 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1759 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1760 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1761 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1762 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1763 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1764 if (sb->s_flags & MS_I_VERSION)
1765 SEQ_OPTS_PUTS("i_version");
1766 if (nodefs || sbi->s_stripe)
1767 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1768 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1769 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1770 SEQ_OPTS_PUTS("data=journal");
1771 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1772 SEQ_OPTS_PUTS("data=ordered");
1773 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1774 SEQ_OPTS_PUTS("data=writeback");
1775 }
1776 if (nodefs ||
1777 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1778 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1779 sbi->s_inode_readahead_blks);
1780
1781 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1782 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1783 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1784
1785 ext4_show_quota_options(seq, sb);
1786 return 0;
1787 }
1788
1789 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1790 {
1791 return _ext4_show_options(seq, root->d_sb, 0);
1792 }
1793
1794 static int options_seq_show(struct seq_file *seq, void *offset)
1795 {
1796 struct super_block *sb = seq->private;
1797 int rc;
1798
1799 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1800 rc = _ext4_show_options(seq, sb, 1);
1801 seq_puts(seq, "\n");
1802 return rc;
1803 }
1804
1805 static int options_open_fs(struct inode *inode, struct file *file)
1806 {
1807 return single_open(file, options_seq_show, PDE(inode)->data);
1808 }
1809
1810 static const struct file_operations ext4_seq_options_fops = {
1811 .owner = THIS_MODULE,
1812 .open = options_open_fs,
1813 .read = seq_read,
1814 .llseek = seq_lseek,
1815 .release = single_release,
1816 };
1817
1818 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1819 int read_only)
1820 {
1821 struct ext4_sb_info *sbi = EXT4_SB(sb);
1822 int res = 0;
1823
1824 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1825 ext4_msg(sb, KERN_ERR, "revision level too high, "
1826 "forcing read-only mode");
1827 res = MS_RDONLY;
1828 }
1829 if (read_only)
1830 goto done;
1831 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1832 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1833 "running e2fsck is recommended");
1834 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1835 ext4_msg(sb, KERN_WARNING,
1836 "warning: mounting fs with errors, "
1837 "running e2fsck is recommended");
1838 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1839 le16_to_cpu(es->s_mnt_count) >=
1840 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1841 ext4_msg(sb, KERN_WARNING,
1842 "warning: maximal mount count reached, "
1843 "running e2fsck is recommended");
1844 else if (le32_to_cpu(es->s_checkinterval) &&
1845 (le32_to_cpu(es->s_lastcheck) +
1846 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1847 ext4_msg(sb, KERN_WARNING,
1848 "warning: checktime reached, "
1849 "running e2fsck is recommended");
1850 if (!sbi->s_journal)
1851 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1852 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1853 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1854 le16_add_cpu(&es->s_mnt_count, 1);
1855 es->s_mtime = cpu_to_le32(get_seconds());
1856 ext4_update_dynamic_rev(sb);
1857 if (sbi->s_journal)
1858 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1859
1860 ext4_commit_super(sb, 1);
1861 done:
1862 if (test_opt(sb, DEBUG))
1863 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1864 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1865 sb->s_blocksize,
1866 sbi->s_groups_count,
1867 EXT4_BLOCKS_PER_GROUP(sb),
1868 EXT4_INODES_PER_GROUP(sb),
1869 sbi->s_mount_opt, sbi->s_mount_opt2);
1870
1871 cleancache_init_fs(sb);
1872 return res;
1873 }
1874
1875 static int ext4_fill_flex_info(struct super_block *sb)
1876 {
1877 struct ext4_sb_info *sbi = EXT4_SB(sb);
1878 struct ext4_group_desc *gdp = NULL;
1879 ext4_group_t flex_group_count;
1880 ext4_group_t flex_group;
1881 unsigned int groups_per_flex = 0;
1882 size_t size;
1883 int i;
1884
1885 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1886 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1887 sbi->s_log_groups_per_flex = 0;
1888 return 1;
1889 }
1890 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1891
1892 /* We allocate both existing and potentially added groups */
1893 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1894 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1895 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1896 size = flex_group_count * sizeof(struct flex_groups);
1897 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
1898 if (sbi->s_flex_groups == NULL) {
1899 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1900 flex_group_count);
1901 goto failed;
1902 }
1903
1904 for (i = 0; i < sbi->s_groups_count; i++) {
1905 gdp = ext4_get_group_desc(sb, i, NULL);
1906
1907 flex_group = ext4_flex_group(sbi, i);
1908 atomic_add(ext4_free_inodes_count(sb, gdp),
1909 &sbi->s_flex_groups[flex_group].free_inodes);
1910 atomic_add(ext4_free_group_clusters(sb, gdp),
1911 &sbi->s_flex_groups[flex_group].free_clusters);
1912 atomic_add(ext4_used_dirs_count(sb, gdp),
1913 &sbi->s_flex_groups[flex_group].used_dirs);
1914 }
1915
1916 return 1;
1917 failed:
1918 return 0;
1919 }
1920
1921 __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1922 struct ext4_group_desc *gdp)
1923 {
1924 __u16 crc = 0;
1925
1926 if (sbi->s_es->s_feature_ro_compat &
1927 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1928 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1929 __le32 le_group = cpu_to_le32(block_group);
1930
1931 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1932 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1933 crc = crc16(crc, (__u8 *)gdp, offset);
1934 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1935 /* for checksum of struct ext4_group_desc do the rest...*/
1936 if ((sbi->s_es->s_feature_incompat &
1937 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1938 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1939 crc = crc16(crc, (__u8 *)gdp + offset,
1940 le16_to_cpu(sbi->s_es->s_desc_size) -
1941 offset);
1942 }
1943
1944 return cpu_to_le16(crc);
1945 }
1946
1947 int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1948 struct ext4_group_desc *gdp)
1949 {
1950 if ((sbi->s_es->s_feature_ro_compat &
1951 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1952 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1953 return 0;
1954
1955 return 1;
1956 }
1957
1958 /* Called at mount-time, super-block is locked */
1959 static int ext4_check_descriptors(struct super_block *sb,
1960 ext4_group_t *first_not_zeroed)
1961 {
1962 struct ext4_sb_info *sbi = EXT4_SB(sb);
1963 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1964 ext4_fsblk_t last_block;
1965 ext4_fsblk_t block_bitmap;
1966 ext4_fsblk_t inode_bitmap;
1967 ext4_fsblk_t inode_table;
1968 int flexbg_flag = 0;
1969 ext4_group_t i, grp = sbi->s_groups_count;
1970
1971 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1972 flexbg_flag = 1;
1973
1974 ext4_debug("Checking group descriptors");
1975
1976 for (i = 0; i < sbi->s_groups_count; i++) {
1977 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1978
1979 if (i == sbi->s_groups_count - 1 || flexbg_flag)
1980 last_block = ext4_blocks_count(sbi->s_es) - 1;
1981 else
1982 last_block = first_block +
1983 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1984
1985 if ((grp == sbi->s_groups_count) &&
1986 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
1987 grp = i;
1988
1989 block_bitmap = ext4_block_bitmap(sb, gdp);
1990 if (block_bitmap < first_block || block_bitmap > last_block) {
1991 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1992 "Block bitmap for group %u not in group "
1993 "(block %llu)!", i, block_bitmap);
1994 return 0;
1995 }
1996 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1997 if (inode_bitmap < first_block || inode_bitmap > last_block) {
1998 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1999 "Inode bitmap for group %u not in group "
2000 "(block %llu)!", i, inode_bitmap);
2001 return 0;
2002 }
2003 inode_table = ext4_inode_table(sb, gdp);
2004 if (inode_table < first_block ||
2005 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2006 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2007 "Inode table for group %u not in group "
2008 "(block %llu)!", i, inode_table);
2009 return 0;
2010 }
2011 ext4_lock_group(sb, i);
2012 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
2013 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2014 "Checksum for group %u failed (%u!=%u)",
2015 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2016 gdp)), le16_to_cpu(gdp->bg_checksum));
2017 if (!(sb->s_flags & MS_RDONLY)) {
2018 ext4_unlock_group(sb, i);
2019 return 0;
2020 }
2021 }
2022 ext4_unlock_group(sb, i);
2023 if (!flexbg_flag)
2024 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2025 }
2026 if (NULL != first_not_zeroed)
2027 *first_not_zeroed = grp;
2028
2029 ext4_free_blocks_count_set(sbi->s_es,
2030 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2031 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2032 return 1;
2033 }
2034
2035 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2036 * the superblock) which were deleted from all directories, but held open by
2037 * a process at the time of a crash. We walk the list and try to delete these
2038 * inodes at recovery time (only with a read-write filesystem).
2039 *
2040 * In order to keep the orphan inode chain consistent during traversal (in
2041 * case of crash during recovery), we link each inode into the superblock
2042 * orphan list_head and handle it the same way as an inode deletion during
2043 * normal operation (which journals the operations for us).
2044 *
2045 * We only do an iget() and an iput() on each inode, which is very safe if we
2046 * accidentally point at an in-use or already deleted inode. The worst that
2047 * can happen in this case is that we get a "bit already cleared" message from
2048 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2049 * e2fsck was run on this filesystem, and it must have already done the orphan
2050 * inode cleanup for us, so we can safely abort without any further action.
2051 */
2052 static void ext4_orphan_cleanup(struct super_block *sb,
2053 struct ext4_super_block *es)
2054 {
2055 unsigned int s_flags = sb->s_flags;
2056 int nr_orphans = 0, nr_truncates = 0;
2057 #ifdef CONFIG_QUOTA
2058 int i;
2059 #endif
2060 if (!es->s_last_orphan) {
2061 jbd_debug(4, "no orphan inodes to clean up\n");
2062 return;
2063 }
2064
2065 if (bdev_read_only(sb->s_bdev)) {
2066 ext4_msg(sb, KERN_ERR, "write access "
2067 "unavailable, skipping orphan cleanup");
2068 return;
2069 }
2070
2071 /* Check if feature set would not allow a r/w mount */
2072 if (!ext4_feature_set_ok(sb, 0)) {
2073 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2074 "unknown ROCOMPAT features");
2075 return;
2076 }
2077
2078 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2079 if (es->s_last_orphan)
2080 jbd_debug(1, "Errors on filesystem, "
2081 "clearing orphan list.\n");
2082 es->s_last_orphan = 0;
2083 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2084 return;
2085 }
2086
2087 if (s_flags & MS_RDONLY) {
2088 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2089 sb->s_flags &= ~MS_RDONLY;
2090 }
2091 #ifdef CONFIG_QUOTA
2092 /* Needed for iput() to work correctly and not trash data */
2093 sb->s_flags |= MS_ACTIVE;
2094 /* Turn on quotas so that they are updated correctly */
2095 for (i = 0; i < MAXQUOTAS; i++) {
2096 if (EXT4_SB(sb)->s_qf_names[i]) {
2097 int ret = ext4_quota_on_mount(sb, i);
2098 if (ret < 0)
2099 ext4_msg(sb, KERN_ERR,
2100 "Cannot turn on journaled "
2101 "quota: error %d", ret);
2102 }
2103 }
2104 #endif
2105
2106 while (es->s_last_orphan) {
2107 struct inode *inode;
2108
2109 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2110 if (IS_ERR(inode)) {
2111 es->s_last_orphan = 0;
2112 break;
2113 }
2114
2115 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2116 dquot_initialize(inode);
2117 if (inode->i_nlink) {
2118 ext4_msg(sb, KERN_DEBUG,
2119 "%s: truncating inode %lu to %lld bytes",
2120 __func__, inode->i_ino, inode->i_size);
2121 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2122 inode->i_ino, inode->i_size);
2123 ext4_truncate(inode);
2124 nr_truncates++;
2125 } else {
2126 ext4_msg(sb, KERN_DEBUG,
2127 "%s: deleting unreferenced inode %lu",
2128 __func__, inode->i_ino);
2129 jbd_debug(2, "deleting unreferenced inode %lu\n",
2130 inode->i_ino);
2131 nr_orphans++;
2132 }
2133 iput(inode); /* The delete magic happens here! */
2134 }
2135
2136 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2137
2138 if (nr_orphans)
2139 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2140 PLURAL(nr_orphans));
2141 if (nr_truncates)
2142 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2143 PLURAL(nr_truncates));
2144 #ifdef CONFIG_QUOTA
2145 /* Turn quotas off */
2146 for (i = 0; i < MAXQUOTAS; i++) {
2147 if (sb_dqopt(sb)->files[i])
2148 dquot_quota_off(sb, i);
2149 }
2150 #endif
2151 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2152 }
2153
2154 /*
2155 * Maximal extent format file size.
2156 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2157 * extent format containers, within a sector_t, and within i_blocks
2158 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2159 * so that won't be a limiting factor.
2160 *
2161 * However there is other limiting factor. We do store extents in the form
2162 * of starting block and length, hence the resulting length of the extent
2163 * covering maximum file size must fit into on-disk format containers as
2164 * well. Given that length is always by 1 unit bigger than max unit (because
2165 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2166 *
2167 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2168 */
2169 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2170 {
2171 loff_t res;
2172 loff_t upper_limit = MAX_LFS_FILESIZE;
2173
2174 /* small i_blocks in vfs inode? */
2175 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2176 /*
2177 * CONFIG_LBDAF is not enabled implies the inode
2178 * i_block represent total blocks in 512 bytes
2179 * 32 == size of vfs inode i_blocks * 8
2180 */
2181 upper_limit = (1LL << 32) - 1;
2182
2183 /* total blocks in file system block size */
2184 upper_limit >>= (blkbits - 9);
2185 upper_limit <<= blkbits;
2186 }
2187
2188 /*
2189 * 32-bit extent-start container, ee_block. We lower the maxbytes
2190 * by one fs block, so ee_len can cover the extent of maximum file
2191 * size
2192 */
2193 res = (1LL << 32) - 1;
2194 res <<= blkbits;
2195
2196 /* Sanity check against vm- & vfs- imposed limits */
2197 if (res > upper_limit)
2198 res = upper_limit;
2199
2200 return res;
2201 }
2202
2203 /*
2204 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2205 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2206 * We need to be 1 filesystem block less than the 2^48 sector limit.
2207 */
2208 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2209 {
2210 loff_t res = EXT4_NDIR_BLOCKS;
2211 int meta_blocks;
2212 loff_t upper_limit;
2213 /* This is calculated to be the largest file size for a dense, block
2214 * mapped file such that the file's total number of 512-byte sectors,
2215 * including data and all indirect blocks, does not exceed (2^48 - 1).
2216 *
2217 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2218 * number of 512-byte sectors of the file.
2219 */
2220
2221 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2222 /*
2223 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2224 * the inode i_block field represents total file blocks in
2225 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2226 */
2227 upper_limit = (1LL << 32) - 1;
2228
2229 /* total blocks in file system block size */
2230 upper_limit >>= (bits - 9);
2231
2232 } else {
2233 /*
2234 * We use 48 bit ext4_inode i_blocks
2235 * With EXT4_HUGE_FILE_FL set the i_blocks
2236 * represent total number of blocks in
2237 * file system block size
2238 */
2239 upper_limit = (1LL << 48) - 1;
2240
2241 }
2242
2243 /* indirect blocks */
2244 meta_blocks = 1;
2245 /* double indirect blocks */
2246 meta_blocks += 1 + (1LL << (bits-2));
2247 /* tripple indirect blocks */
2248 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2249
2250 upper_limit -= meta_blocks;
2251 upper_limit <<= bits;
2252
2253 res += 1LL << (bits-2);
2254 res += 1LL << (2*(bits-2));
2255 res += 1LL << (3*(bits-2));
2256 res <<= bits;
2257 if (res > upper_limit)
2258 res = upper_limit;
2259
2260 if (res > MAX_LFS_FILESIZE)
2261 res = MAX_LFS_FILESIZE;
2262
2263 return res;
2264 }
2265
2266 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2267 ext4_fsblk_t logical_sb_block, int nr)
2268 {
2269 struct ext4_sb_info *sbi = EXT4_SB(sb);
2270 ext4_group_t bg, first_meta_bg;
2271 int has_super = 0;
2272
2273 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2274
2275 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2276 nr < first_meta_bg)
2277 return logical_sb_block + nr + 1;
2278 bg = sbi->s_desc_per_block * nr;
2279 if (ext4_bg_has_super(sb, bg))
2280 has_super = 1;
2281
2282 return (has_super + ext4_group_first_block_no(sb, bg));
2283 }
2284
2285 /**
2286 * ext4_get_stripe_size: Get the stripe size.
2287 * @sbi: In memory super block info
2288 *
2289 * If we have specified it via mount option, then
2290 * use the mount option value. If the value specified at mount time is
2291 * greater than the blocks per group use the super block value.
2292 * If the super block value is greater than blocks per group return 0.
2293 * Allocator needs it be less than blocks per group.
2294 *
2295 */
2296 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2297 {
2298 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2299 unsigned long stripe_width =
2300 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2301 int ret;
2302
2303 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2304 ret = sbi->s_stripe;
2305 else if (stripe_width <= sbi->s_blocks_per_group)
2306 ret = stripe_width;
2307 else if (stride <= sbi->s_blocks_per_group)
2308 ret = stride;
2309 else
2310 ret = 0;
2311
2312 /*
2313 * If the stripe width is 1, this makes no sense and
2314 * we set it to 0 to turn off stripe handling code.
2315 */
2316 if (ret <= 1)
2317 ret = 0;
2318
2319 return ret;
2320 }
2321
2322 /* sysfs supprt */
2323
2324 struct ext4_attr {
2325 struct attribute attr;
2326 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2327 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2328 const char *, size_t);
2329 int offset;
2330 };
2331
2332 static int parse_strtoul(const char *buf,
2333 unsigned long max, unsigned long *value)
2334 {
2335 char *endp;
2336
2337 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2338 endp = skip_spaces(endp);
2339 if (*endp || *value > max)
2340 return -EINVAL;
2341
2342 return 0;
2343 }
2344
2345 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2346 struct ext4_sb_info *sbi,
2347 char *buf)
2348 {
2349 return snprintf(buf, PAGE_SIZE, "%llu\n",
2350 (s64) EXT4_C2B(sbi,
2351 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2352 }
2353
2354 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2355 struct ext4_sb_info *sbi, char *buf)
2356 {
2357 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2358
2359 if (!sb->s_bdev->bd_part)
2360 return snprintf(buf, PAGE_SIZE, "0\n");
2361 return snprintf(buf, PAGE_SIZE, "%lu\n",
2362 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2363 sbi->s_sectors_written_start) >> 1);
2364 }
2365
2366 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2367 struct ext4_sb_info *sbi, char *buf)
2368 {
2369 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2370
2371 if (!sb->s_bdev->bd_part)
2372 return snprintf(buf, PAGE_SIZE, "0\n");
2373 return snprintf(buf, PAGE_SIZE, "%llu\n",
2374 (unsigned long long)(sbi->s_kbytes_written +
2375 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2376 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2377 }
2378
2379 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2380 struct ext4_sb_info *sbi,
2381 const char *buf, size_t count)
2382 {
2383 unsigned long t;
2384
2385 if (parse_strtoul(buf, 0x40000000, &t))
2386 return -EINVAL;
2387
2388 if (t && !is_power_of_2(t))
2389 return -EINVAL;
2390
2391 sbi->s_inode_readahead_blks = t;
2392 return count;
2393 }
2394
2395 static ssize_t sbi_ui_show(struct ext4_attr *a,
2396 struct ext4_sb_info *sbi, char *buf)
2397 {
2398 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2399
2400 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2401 }
2402
2403 static ssize_t sbi_ui_store(struct ext4_attr *a,
2404 struct ext4_sb_info *sbi,
2405 const char *buf, size_t count)
2406 {
2407 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2408 unsigned long t;
2409
2410 if (parse_strtoul(buf, 0xffffffff, &t))
2411 return -EINVAL;
2412 *ui = t;
2413 return count;
2414 }
2415
2416 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2417 static struct ext4_attr ext4_attr_##_name = { \
2418 .attr = {.name = __stringify(_name), .mode = _mode }, \
2419 .show = _show, \
2420 .store = _store, \
2421 .offset = offsetof(struct ext4_sb_info, _elname), \
2422 }
2423 #define EXT4_ATTR(name, mode, show, store) \
2424 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2425
2426 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2427 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2428 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2429 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2430 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2431 #define ATTR_LIST(name) &ext4_attr_##name.attr
2432
2433 EXT4_RO_ATTR(delayed_allocation_blocks);
2434 EXT4_RO_ATTR(session_write_kbytes);
2435 EXT4_RO_ATTR(lifetime_write_kbytes);
2436 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2437 inode_readahead_blks_store, s_inode_readahead_blks);
2438 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2439 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2440 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2441 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2442 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2443 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2444 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2445 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2446
2447 static struct attribute *ext4_attrs[] = {
2448 ATTR_LIST(delayed_allocation_blocks),
2449 ATTR_LIST(session_write_kbytes),
2450 ATTR_LIST(lifetime_write_kbytes),
2451 ATTR_LIST(inode_readahead_blks),
2452 ATTR_LIST(inode_goal),
2453 ATTR_LIST(mb_stats),
2454 ATTR_LIST(mb_max_to_scan),
2455 ATTR_LIST(mb_min_to_scan),
2456 ATTR_LIST(mb_order2_req),
2457 ATTR_LIST(mb_stream_req),
2458 ATTR_LIST(mb_group_prealloc),
2459 ATTR_LIST(max_writeback_mb_bump),
2460 NULL,
2461 };
2462
2463 /* Features this copy of ext4 supports */
2464 EXT4_INFO_ATTR(lazy_itable_init);
2465 EXT4_INFO_ATTR(batched_discard);
2466
2467 static struct attribute *ext4_feat_attrs[] = {
2468 ATTR_LIST(lazy_itable_init),
2469 ATTR_LIST(batched_discard),
2470 NULL,
2471 };
2472
2473 static ssize_t ext4_attr_show(struct kobject *kobj,
2474 struct attribute *attr, char *buf)
2475 {
2476 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2477 s_kobj);
2478 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2479
2480 return a->show ? a->show(a, sbi, buf) : 0;
2481 }
2482
2483 static ssize_t ext4_attr_store(struct kobject *kobj,
2484 struct attribute *attr,
2485 const char *buf, size_t len)
2486 {
2487 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2488 s_kobj);
2489 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2490
2491 return a->store ? a->store(a, sbi, buf, len) : 0;
2492 }
2493
2494 static void ext4_sb_release(struct kobject *kobj)
2495 {
2496 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2497 s_kobj);
2498 complete(&sbi->s_kobj_unregister);
2499 }
2500
2501 static const struct sysfs_ops ext4_attr_ops = {
2502 .show = ext4_attr_show,
2503 .store = ext4_attr_store,
2504 };
2505
2506 static struct kobj_type ext4_ktype = {
2507 .default_attrs = ext4_attrs,
2508 .sysfs_ops = &ext4_attr_ops,
2509 .release = ext4_sb_release,
2510 };
2511
2512 static void ext4_feat_release(struct kobject *kobj)
2513 {
2514 complete(&ext4_feat->f_kobj_unregister);
2515 }
2516
2517 static struct kobj_type ext4_feat_ktype = {
2518 .default_attrs = ext4_feat_attrs,
2519 .sysfs_ops = &ext4_attr_ops,
2520 .release = ext4_feat_release,
2521 };
2522
2523 /*
2524 * Check whether this filesystem can be mounted based on
2525 * the features present and the RDONLY/RDWR mount requested.
2526 * Returns 1 if this filesystem can be mounted as requested,
2527 * 0 if it cannot be.
2528 */
2529 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2530 {
2531 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2532 ext4_msg(sb, KERN_ERR,
2533 "Couldn't mount because of "
2534 "unsupported optional features (%x)",
2535 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2536 ~EXT4_FEATURE_INCOMPAT_SUPP));
2537 return 0;
2538 }
2539
2540 if (readonly)
2541 return 1;
2542
2543 /* Check that feature set is OK for a read-write mount */
2544 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2545 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2546 "unsupported optional features (%x)",
2547 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2548 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2549 return 0;
2550 }
2551 /*
2552 * Large file size enabled file system can only be mounted
2553 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2554 */
2555 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2556 if (sizeof(blkcnt_t) < sizeof(u64)) {
2557 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2558 "cannot be mounted RDWR without "
2559 "CONFIG_LBDAF");
2560 return 0;
2561 }
2562 }
2563 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2564 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2565 ext4_msg(sb, KERN_ERR,
2566 "Can't support bigalloc feature without "
2567 "extents feature\n");
2568 return 0;
2569 }
2570 return 1;
2571 }
2572
2573 /*
2574 * This function is called once a day if we have errors logged
2575 * on the file system
2576 */
2577 static void print_daily_error_info(unsigned long arg)
2578 {
2579 struct super_block *sb = (struct super_block *) arg;
2580 struct ext4_sb_info *sbi;
2581 struct ext4_super_block *es;
2582
2583 sbi = EXT4_SB(sb);
2584 es = sbi->s_es;
2585
2586 if (es->s_error_count)
2587 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2588 le32_to_cpu(es->s_error_count));
2589 if (es->s_first_error_time) {
2590 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2591 sb->s_id, le32_to_cpu(es->s_first_error_time),
2592 (int) sizeof(es->s_first_error_func),
2593 es->s_first_error_func,
2594 le32_to_cpu(es->s_first_error_line));
2595 if (es->s_first_error_ino)
2596 printk(": inode %u",
2597 le32_to_cpu(es->s_first_error_ino));
2598 if (es->s_first_error_block)
2599 printk(": block %llu", (unsigned long long)
2600 le64_to_cpu(es->s_first_error_block));
2601 printk("\n");
2602 }
2603 if (es->s_last_error_time) {
2604 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2605 sb->s_id, le32_to_cpu(es->s_last_error_time),
2606 (int) sizeof(es->s_last_error_func),
2607 es->s_last_error_func,
2608 le32_to_cpu(es->s_last_error_line));
2609 if (es->s_last_error_ino)
2610 printk(": inode %u",
2611 le32_to_cpu(es->s_last_error_ino));
2612 if (es->s_last_error_block)
2613 printk(": block %llu", (unsigned long long)
2614 le64_to_cpu(es->s_last_error_block));
2615 printk("\n");
2616 }
2617 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2618 }
2619
2620 /* Find next suitable group and run ext4_init_inode_table */
2621 static int ext4_run_li_request(struct ext4_li_request *elr)
2622 {
2623 struct ext4_group_desc *gdp = NULL;
2624 ext4_group_t group, ngroups;
2625 struct super_block *sb;
2626 unsigned long timeout = 0;
2627 int ret = 0;
2628
2629 sb = elr->lr_super;
2630 ngroups = EXT4_SB(sb)->s_groups_count;
2631
2632 for (group = elr->lr_next_group; group < ngroups; group++) {
2633 gdp = ext4_get_group_desc(sb, group, NULL);
2634 if (!gdp) {
2635 ret = 1;
2636 break;
2637 }
2638
2639 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2640 break;
2641 }
2642
2643 if (group == ngroups)
2644 ret = 1;
2645
2646 if (!ret) {
2647 timeout = jiffies;
2648 ret = ext4_init_inode_table(sb, group,
2649 elr->lr_timeout ? 0 : 1);
2650 if (elr->lr_timeout == 0) {
2651 timeout = (jiffies - timeout) *
2652 elr->lr_sbi->s_li_wait_mult;
2653 elr->lr_timeout = timeout;
2654 }
2655 elr->lr_next_sched = jiffies + elr->lr_timeout;
2656 elr->lr_next_group = group + 1;
2657 }
2658
2659 return ret;
2660 }
2661
2662 /*
2663 * Remove lr_request from the list_request and free the
2664 * request structure. Should be called with li_list_mtx held
2665 */
2666 static void ext4_remove_li_request(struct ext4_li_request *elr)
2667 {
2668 struct ext4_sb_info *sbi;
2669
2670 if (!elr)
2671 return;
2672
2673 sbi = elr->lr_sbi;
2674
2675 list_del(&elr->lr_request);
2676 sbi->s_li_request = NULL;
2677 kfree(elr);
2678 }
2679
2680 static void ext4_unregister_li_request(struct super_block *sb)
2681 {
2682 mutex_lock(&ext4_li_mtx);
2683 if (!ext4_li_info) {
2684 mutex_unlock(&ext4_li_mtx);
2685 return;
2686 }
2687
2688 mutex_lock(&ext4_li_info->li_list_mtx);
2689 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2690 mutex_unlock(&ext4_li_info->li_list_mtx);
2691 mutex_unlock(&ext4_li_mtx);
2692 }
2693
2694 static struct task_struct *ext4_lazyinit_task;
2695
2696 /*
2697 * This is the function where ext4lazyinit thread lives. It walks
2698 * through the request list searching for next scheduled filesystem.
2699 * When such a fs is found, run the lazy initialization request
2700 * (ext4_rn_li_request) and keep track of the time spend in this
2701 * function. Based on that time we compute next schedule time of
2702 * the request. When walking through the list is complete, compute
2703 * next waking time and put itself into sleep.
2704 */
2705 static int ext4_lazyinit_thread(void *arg)
2706 {
2707 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2708 struct list_head *pos, *n;
2709 struct ext4_li_request *elr;
2710 unsigned long next_wakeup, cur;
2711
2712 BUG_ON(NULL == eli);
2713
2714 cont_thread:
2715 while (true) {
2716 next_wakeup = MAX_JIFFY_OFFSET;
2717
2718 mutex_lock(&eli->li_list_mtx);
2719 if (list_empty(&eli->li_request_list)) {
2720 mutex_unlock(&eli->li_list_mtx);
2721 goto exit_thread;
2722 }
2723
2724 list_for_each_safe(pos, n, &eli->li_request_list) {
2725 elr = list_entry(pos, struct ext4_li_request,
2726 lr_request);
2727
2728 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2729 if (ext4_run_li_request(elr) != 0) {
2730 /* error, remove the lazy_init job */
2731 ext4_remove_li_request(elr);
2732 continue;
2733 }
2734 }
2735
2736 if (time_before(elr->lr_next_sched, next_wakeup))
2737 next_wakeup = elr->lr_next_sched;
2738 }
2739 mutex_unlock(&eli->li_list_mtx);
2740
2741 try_to_freeze();
2742
2743 cur = jiffies;
2744 if ((time_after_eq(cur, next_wakeup)) ||
2745 (MAX_JIFFY_OFFSET == next_wakeup)) {
2746 cond_resched();
2747 continue;
2748 }
2749
2750 schedule_timeout_interruptible(next_wakeup - cur);
2751
2752 if (kthread_should_stop()) {
2753 ext4_clear_request_list();
2754 goto exit_thread;
2755 }
2756 }
2757
2758 exit_thread:
2759 /*
2760 * It looks like the request list is empty, but we need
2761 * to check it under the li_list_mtx lock, to prevent any
2762 * additions into it, and of course we should lock ext4_li_mtx
2763 * to atomically free the list and ext4_li_info, because at
2764 * this point another ext4 filesystem could be registering
2765 * new one.
2766 */
2767 mutex_lock(&ext4_li_mtx);
2768 mutex_lock(&eli->li_list_mtx);
2769 if (!list_empty(&eli->li_request_list)) {
2770 mutex_unlock(&eli->li_list_mtx);
2771 mutex_unlock(&ext4_li_mtx);
2772 goto cont_thread;
2773 }
2774 mutex_unlock(&eli->li_list_mtx);
2775 kfree(ext4_li_info);
2776 ext4_li_info = NULL;
2777 mutex_unlock(&ext4_li_mtx);
2778
2779 return 0;
2780 }
2781
2782 static void ext4_clear_request_list(void)
2783 {
2784 struct list_head *pos, *n;
2785 struct ext4_li_request *elr;
2786
2787 mutex_lock(&ext4_li_info->li_list_mtx);
2788 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2789 elr = list_entry(pos, struct ext4_li_request,
2790 lr_request);
2791 ext4_remove_li_request(elr);
2792 }
2793 mutex_unlock(&ext4_li_info->li_list_mtx);
2794 }
2795
2796 static int ext4_run_lazyinit_thread(void)
2797 {
2798 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2799 ext4_li_info, "ext4lazyinit");
2800 if (IS_ERR(ext4_lazyinit_task)) {
2801 int err = PTR_ERR(ext4_lazyinit_task);
2802 ext4_clear_request_list();
2803 kfree(ext4_li_info);
2804 ext4_li_info = NULL;
2805 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2806 "initialization thread\n",
2807 err);
2808 return err;
2809 }
2810 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2811 return 0;
2812 }
2813
2814 /*
2815 * Check whether it make sense to run itable init. thread or not.
2816 * If there is at least one uninitialized inode table, return
2817 * corresponding group number, else the loop goes through all
2818 * groups and return total number of groups.
2819 */
2820 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2821 {
2822 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2823 struct ext4_group_desc *gdp = NULL;
2824
2825 for (group = 0; group < ngroups; group++) {
2826 gdp = ext4_get_group_desc(sb, group, NULL);
2827 if (!gdp)
2828 continue;
2829
2830 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2831 break;
2832 }
2833
2834 return group;
2835 }
2836
2837 static int ext4_li_info_new(void)
2838 {
2839 struct ext4_lazy_init *eli = NULL;
2840
2841 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2842 if (!eli)
2843 return -ENOMEM;
2844
2845 INIT_LIST_HEAD(&eli->li_request_list);
2846 mutex_init(&eli->li_list_mtx);
2847
2848 eli->li_state |= EXT4_LAZYINIT_QUIT;
2849
2850 ext4_li_info = eli;
2851
2852 return 0;
2853 }
2854
2855 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2856 ext4_group_t start)
2857 {
2858 struct ext4_sb_info *sbi = EXT4_SB(sb);
2859 struct ext4_li_request *elr;
2860 unsigned long rnd;
2861
2862 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2863 if (!elr)
2864 return NULL;
2865
2866 elr->lr_super = sb;
2867 elr->lr_sbi = sbi;
2868 elr->lr_next_group = start;
2869
2870 /*
2871 * Randomize first schedule time of the request to
2872 * spread the inode table initialization requests
2873 * better.
2874 */
2875 get_random_bytes(&rnd, sizeof(rnd));
2876 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2877 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2878
2879 return elr;
2880 }
2881
2882 static int ext4_register_li_request(struct super_block *sb,
2883 ext4_group_t first_not_zeroed)
2884 {
2885 struct ext4_sb_info *sbi = EXT4_SB(sb);
2886 struct ext4_li_request *elr;
2887 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2888 int ret = 0;
2889
2890 if (sbi->s_li_request != NULL) {
2891 /*
2892 * Reset timeout so it can be computed again, because
2893 * s_li_wait_mult might have changed.
2894 */
2895 sbi->s_li_request->lr_timeout = 0;
2896 return 0;
2897 }
2898
2899 if (first_not_zeroed == ngroups ||
2900 (sb->s_flags & MS_RDONLY) ||
2901 !test_opt(sb, INIT_INODE_TABLE))
2902 return 0;
2903
2904 elr = ext4_li_request_new(sb, first_not_zeroed);
2905 if (!elr)
2906 return -ENOMEM;
2907
2908 mutex_lock(&ext4_li_mtx);
2909
2910 if (NULL == ext4_li_info) {
2911 ret = ext4_li_info_new();
2912 if (ret)
2913 goto out;
2914 }
2915
2916 mutex_lock(&ext4_li_info->li_list_mtx);
2917 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2918 mutex_unlock(&ext4_li_info->li_list_mtx);
2919
2920 sbi->s_li_request = elr;
2921 /*
2922 * set elr to NULL here since it has been inserted to
2923 * the request_list and the removal and free of it is
2924 * handled by ext4_clear_request_list from now on.
2925 */
2926 elr = NULL;
2927
2928 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2929 ret = ext4_run_lazyinit_thread();
2930 if (ret)
2931 goto out;
2932 }
2933 out:
2934 mutex_unlock(&ext4_li_mtx);
2935 if (ret)
2936 kfree(elr);
2937 return ret;
2938 }
2939
2940 /*
2941 * We do not need to lock anything since this is called on
2942 * module unload.
2943 */
2944 static void ext4_destroy_lazyinit_thread(void)
2945 {
2946 /*
2947 * If thread exited earlier
2948 * there's nothing to be done.
2949 */
2950 if (!ext4_li_info || !ext4_lazyinit_task)
2951 return;
2952
2953 kthread_stop(ext4_lazyinit_task);
2954 }
2955
2956 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2957 {
2958 char *orig_data = kstrdup(data, GFP_KERNEL);
2959 struct buffer_head *bh;
2960 struct ext4_super_block *es = NULL;
2961 struct ext4_sb_info *sbi;
2962 ext4_fsblk_t block;
2963 ext4_fsblk_t sb_block = get_sb_block(&data);
2964 ext4_fsblk_t logical_sb_block;
2965 unsigned long offset = 0;
2966 unsigned long journal_devnum = 0;
2967 unsigned long def_mount_opts;
2968 struct inode *root;
2969 char *cp;
2970 const char *descr;
2971 int ret = -ENOMEM;
2972 int blocksize, clustersize;
2973 unsigned int db_count;
2974 unsigned int i;
2975 int needs_recovery, has_huge_files, has_bigalloc;
2976 __u64 blocks_count;
2977 int err;
2978 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2979 ext4_group_t first_not_zeroed;
2980
2981 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2982 if (!sbi)
2983 goto out_free_orig;
2984
2985 sbi->s_blockgroup_lock =
2986 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2987 if (!sbi->s_blockgroup_lock) {
2988 kfree(sbi);
2989 goto out_free_orig;
2990 }
2991 sb->s_fs_info = sbi;
2992 sbi->s_mount_opt = 0;
2993 sbi->s_resuid = EXT4_DEF_RESUID;
2994 sbi->s_resgid = EXT4_DEF_RESGID;
2995 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2996 sbi->s_sb_block = sb_block;
2997 if (sb->s_bdev->bd_part)
2998 sbi->s_sectors_written_start =
2999 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3000
3001 /* Cleanup superblock name */
3002 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3003 *cp = '!';
3004
3005 ret = -EINVAL;
3006 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3007 if (!blocksize) {
3008 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3009 goto out_fail;
3010 }
3011
3012 /*
3013 * The ext4 superblock will not be buffer aligned for other than 1kB
3014 * block sizes. We need to calculate the offset from buffer start.
3015 */
3016 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3017 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3018 offset = do_div(logical_sb_block, blocksize);
3019 } else {
3020 logical_sb_block = sb_block;
3021 }
3022
3023 if (!(bh = sb_bread(sb, logical_sb_block))) {
3024 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3025 goto out_fail;
3026 }
3027 /*
3028 * Note: s_es must be initialized as soon as possible because
3029 * some ext4 macro-instructions depend on its value
3030 */
3031 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3032 sbi->s_es = es;
3033 sb->s_magic = le16_to_cpu(es->s_magic);
3034 if (sb->s_magic != EXT4_SUPER_MAGIC)
3035 goto cantfind_ext4;
3036 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3037
3038 /* Check for a known checksum algorithm */
3039 if (!ext4_verify_csum_type(sb, es)) {
3040 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3041 "unknown checksum algorithm.");
3042 silent = 1;
3043 goto cantfind_ext4;
3044 }
3045
3046 /* Set defaults before we parse the mount options */
3047 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3048 set_opt(sb, INIT_INODE_TABLE);
3049 if (def_mount_opts & EXT4_DEFM_DEBUG)
3050 set_opt(sb, DEBUG);
3051 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3052 set_opt(sb, GRPID);
3053 if (def_mount_opts & EXT4_DEFM_UID16)
3054 set_opt(sb, NO_UID32);
3055 /* xattr user namespace & acls are now defaulted on */
3056 #ifdef CONFIG_EXT4_FS_XATTR
3057 set_opt(sb, XATTR_USER);
3058 #endif
3059 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3060 set_opt(sb, POSIX_ACL);
3061 #endif
3062 set_opt(sb, MBLK_IO_SUBMIT);
3063 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3064 set_opt(sb, JOURNAL_DATA);
3065 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3066 set_opt(sb, ORDERED_DATA);
3067 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3068 set_opt(sb, WRITEBACK_DATA);
3069
3070 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3071 set_opt(sb, ERRORS_PANIC);
3072 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3073 set_opt(sb, ERRORS_CONT);
3074 else
3075 set_opt(sb, ERRORS_RO);
3076 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3077 set_opt(sb, BLOCK_VALIDITY);
3078 if (def_mount_opts & EXT4_DEFM_DISCARD)
3079 set_opt(sb, DISCARD);
3080
3081 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3082 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
3083 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3084 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3085 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3086
3087 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3088 set_opt(sb, BARRIER);
3089
3090 /*
3091 * enable delayed allocation by default
3092 * Use -o nodelalloc to turn it off
3093 */
3094 if (!IS_EXT3_SB(sb) &&
3095 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3096 set_opt(sb, DELALLOC);
3097
3098 /*
3099 * set default s_li_wait_mult for lazyinit, for the case there is
3100 * no mount option specified.
3101 */
3102 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3103
3104 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3105 &journal_devnum, &journal_ioprio, 0)) {
3106 ext4_msg(sb, KERN_WARNING,
3107 "failed to parse options in superblock: %s",
3108 sbi->s_es->s_mount_opts);
3109 }
3110 sbi->s_def_mount_opt = sbi->s_mount_opt;
3111 if (!parse_options((char *) data, sb, &journal_devnum,
3112 &journal_ioprio, 0))
3113 goto failed_mount;
3114
3115 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3116 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3117 "with data=journal disables delayed "
3118 "allocation and O_DIRECT support!\n");
3119 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3120 ext4_msg(sb, KERN_ERR, "can't mount with "
3121 "both data=journal and delalloc");
3122 goto failed_mount;
3123 }
3124 if (test_opt(sb, DIOREAD_NOLOCK)) {
3125 ext4_msg(sb, KERN_ERR, "can't mount with "
3126 "both data=journal and delalloc");
3127 goto failed_mount;
3128 }
3129 if (test_opt(sb, DELALLOC))
3130 clear_opt(sb, DELALLOC);
3131 }
3132
3133 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3134 if (test_opt(sb, DIOREAD_NOLOCK)) {
3135 if (blocksize < PAGE_SIZE) {
3136 ext4_msg(sb, KERN_ERR, "can't mount with "
3137 "dioread_nolock if block size != PAGE_SIZE");
3138 goto failed_mount;
3139 }
3140 }
3141
3142 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3143 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3144
3145 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3146 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3147 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3148 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3149 ext4_msg(sb, KERN_WARNING,
3150 "feature flags set on rev 0 fs, "
3151 "running e2fsck is recommended");
3152
3153 if (IS_EXT2_SB(sb)) {
3154 if (ext2_feature_set_ok(sb))
3155 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3156 "using the ext4 subsystem");
3157 else {
3158 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3159 "to feature incompatibilities");
3160 goto failed_mount;
3161 }
3162 }
3163
3164 if (IS_EXT3_SB(sb)) {
3165 if (ext3_feature_set_ok(sb))
3166 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3167 "using the ext4 subsystem");
3168 else {
3169 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3170 "to feature incompatibilities");
3171 goto failed_mount;
3172 }
3173 }
3174
3175 /*
3176 * Check feature flags regardless of the revision level, since we
3177 * previously didn't change the revision level when setting the flags,
3178 * so there is a chance incompat flags are set on a rev 0 filesystem.
3179 */
3180 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3181 goto failed_mount;
3182
3183 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3184 blocksize > EXT4_MAX_BLOCK_SIZE) {
3185 ext4_msg(sb, KERN_ERR,
3186 "Unsupported filesystem blocksize %d", blocksize);
3187 goto failed_mount;
3188 }
3189
3190 if (sb->s_blocksize != blocksize) {
3191 /* Validate the filesystem blocksize */
3192 if (!sb_set_blocksize(sb, blocksize)) {
3193 ext4_msg(sb, KERN_ERR, "bad block size %d",
3194 blocksize);
3195 goto failed_mount;
3196 }
3197
3198 brelse(bh);
3199 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3200 offset = do_div(logical_sb_block, blocksize);
3201 bh = sb_bread(sb, logical_sb_block);
3202 if (!bh) {
3203 ext4_msg(sb, KERN_ERR,
3204 "Can't read superblock on 2nd try");
3205 goto failed_mount;
3206 }
3207 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
3208 sbi->s_es = es;
3209 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3210 ext4_msg(sb, KERN_ERR,
3211 "Magic mismatch, very weird!");
3212 goto failed_mount;
3213 }
3214 }
3215
3216 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3217 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3218 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3219 has_huge_files);
3220 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3221
3222 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3223 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3224 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3225 } else {
3226 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3227 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3228 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3229 (!is_power_of_2(sbi->s_inode_size)) ||
3230 (sbi->s_inode_size > blocksize)) {
3231 ext4_msg(sb, KERN_ERR,
3232 "unsupported inode size: %d",
3233 sbi->s_inode_size);
3234 goto failed_mount;
3235 }
3236 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3237 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3238 }
3239
3240 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3241 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3242 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3243 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3244 !is_power_of_2(sbi->s_desc_size)) {
3245 ext4_msg(sb, KERN_ERR,
3246 "unsupported descriptor size %lu",
3247 sbi->s_desc_size);
3248 goto failed_mount;
3249 }
3250 } else
3251 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3252
3253 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3254 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3255 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3256 goto cantfind_ext4;
3257
3258 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3259 if (sbi->s_inodes_per_block == 0)
3260 goto cantfind_ext4;
3261 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3262 sbi->s_inodes_per_block;
3263 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3264 sbi->s_sbh = bh;
3265 sbi->s_mount_state = le16_to_cpu(es->s_state);
3266 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3267 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3268
3269 for (i = 0; i < 4; i++)
3270 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3271 sbi->s_def_hash_version = es->s_def_hash_version;
3272 i = le32_to_cpu(es->s_flags);
3273 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3274 sbi->s_hash_unsigned = 3;
3275 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3276 #ifdef __CHAR_UNSIGNED__
3277 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3278 sbi->s_hash_unsigned = 3;
3279 #else
3280 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3281 #endif
3282 }
3283
3284 /* Handle clustersize */
3285 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3286 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3287 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3288 if (has_bigalloc) {
3289 if (clustersize < blocksize) {
3290 ext4_msg(sb, KERN_ERR,
3291 "cluster size (%d) smaller than "
3292 "block size (%d)", clustersize, blocksize);
3293 goto failed_mount;
3294 }
3295 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3296 le32_to_cpu(es->s_log_block_size);
3297 sbi->s_clusters_per_group =
3298 le32_to_cpu(es->s_clusters_per_group);
3299 if (sbi->s_clusters_per_group > blocksize * 8) {
3300 ext4_msg(sb, KERN_ERR,
3301 "#clusters per group too big: %lu",
3302 sbi->s_clusters_per_group);
3303 goto failed_mount;
3304 }
3305 if (sbi->s_blocks_per_group !=
3306 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3307 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3308 "clusters per group (%lu) inconsistent",
3309 sbi->s_blocks_per_group,
3310 sbi->s_clusters_per_group);
3311 goto failed_mount;
3312 }
3313 } else {
3314 if (clustersize != blocksize) {
3315 ext4_warning(sb, "fragment/cluster size (%d) != "
3316 "block size (%d)", clustersize,
3317 blocksize);
3318 clustersize = blocksize;
3319 }
3320 if (sbi->s_blocks_per_group > blocksize * 8) {
3321 ext4_msg(sb, KERN_ERR,
3322 "#blocks per group too big: %lu",
3323 sbi->s_blocks_per_group);
3324 goto failed_mount;
3325 }
3326 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3327 sbi->s_cluster_bits = 0;
3328 }
3329 sbi->s_cluster_ratio = clustersize / blocksize;
3330
3331 if (sbi->s_inodes_per_group > blocksize * 8) {
3332 ext4_msg(sb, KERN_ERR,
3333 "#inodes per group too big: %lu",
3334 sbi->s_inodes_per_group);
3335 goto failed_mount;
3336 }
3337
3338 /*
3339 * Test whether we have more sectors than will fit in sector_t,
3340 * and whether the max offset is addressable by the page cache.
3341 */
3342 err = generic_check_addressable(sb->s_blocksize_bits,
3343 ext4_blocks_count(es));
3344 if (err) {
3345 ext4_msg(sb, KERN_ERR, "filesystem"
3346 " too large to mount safely on this system");
3347 if (sizeof(sector_t) < 8)
3348 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3349 ret = err;
3350 goto failed_mount;
3351 }
3352
3353 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3354 goto cantfind_ext4;
3355
3356 /* check blocks count against device size */
3357 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3358 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3359 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3360 "exceeds size of device (%llu blocks)",
3361 ext4_blocks_count(es), blocks_count);
3362 goto failed_mount;
3363 }
3364
3365 /*
3366 * It makes no sense for the first data block to be beyond the end
3367 * of the filesystem.
3368 */
3369 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3370 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3371 "block %u is beyond end of filesystem (%llu)",
3372 le32_to_cpu(es->s_first_data_block),
3373 ext4_blocks_count(es));
3374 goto failed_mount;
3375 }
3376 blocks_count = (ext4_blocks_count(es) -
3377 le32_to_cpu(es->s_first_data_block) +
3378 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3379 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3380 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3381 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3382 "(block count %llu, first data block %u, "
3383 "blocks per group %lu)", sbi->s_groups_count,
3384 ext4_blocks_count(es),
3385 le32_to_cpu(es->s_first_data_block),
3386 EXT4_BLOCKS_PER_GROUP(sb));
3387 goto failed_mount;
3388 }
3389 sbi->s_groups_count = blocks_count;
3390 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3391 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3392 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3393 EXT4_DESC_PER_BLOCK(sb);
3394 sbi->s_group_desc = ext4_kvmalloc(db_count *
3395 sizeof(struct buffer_head *),
3396 GFP_KERNEL);
3397 if (sbi->s_group_desc == NULL) {
3398 ext4_msg(sb, KERN_ERR, "not enough memory");
3399 goto failed_mount;
3400 }
3401
3402 if (ext4_proc_root)
3403 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3404
3405 if (sbi->s_proc)
3406 proc_create_data("options", S_IRUGO, sbi->s_proc,
3407 &ext4_seq_options_fops, sb);
3408
3409 bgl_lock_init(sbi->s_blockgroup_lock);
3410
3411 for (i = 0; i < db_count; i++) {
3412 block = descriptor_loc(sb, logical_sb_block, i);
3413 sbi->s_group_desc[i] = sb_bread(sb, block);
3414 if (!sbi->s_group_desc[i]) {
3415 ext4_msg(sb, KERN_ERR,
3416 "can't read group descriptor %d", i);
3417 db_count = i;
3418 goto failed_mount2;
3419 }
3420 }
3421 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3422 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3423 goto failed_mount2;
3424 }
3425 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3426 if (!ext4_fill_flex_info(sb)) {
3427 ext4_msg(sb, KERN_ERR,
3428 "unable to initialize "
3429 "flex_bg meta info!");
3430 goto failed_mount2;
3431 }
3432
3433 sbi->s_gdb_count = db_count;
3434 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3435 spin_lock_init(&sbi->s_next_gen_lock);
3436
3437 init_timer(&sbi->s_err_report);
3438 sbi->s_err_report.function = print_daily_error_info;
3439 sbi->s_err_report.data = (unsigned long) sb;
3440
3441 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3442 ext4_count_free_clusters(sb));
3443 if (!err) {
3444 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3445 ext4_count_free_inodes(sb));
3446 }
3447 if (!err) {
3448 err = percpu_counter_init(&sbi->s_dirs_counter,
3449 ext4_count_dirs(sb));
3450 }
3451 if (!err) {
3452 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3453 }
3454 if (err) {
3455 ext4_msg(sb, KERN_ERR, "insufficient memory");
3456 goto failed_mount3;
3457 }
3458
3459 sbi->s_stripe = ext4_get_stripe_size(sbi);
3460 sbi->s_max_writeback_mb_bump = 128;
3461
3462 /*
3463 * set up enough so that it can read an inode
3464 */
3465 if (!test_opt(sb, NOLOAD) &&
3466 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3467 sb->s_op = &ext4_sops;
3468 else
3469 sb->s_op = &ext4_nojournal_sops;
3470 sb->s_export_op = &ext4_export_ops;
3471 sb->s_xattr = ext4_xattr_handlers;
3472 #ifdef CONFIG_QUOTA
3473 sb->s_qcop = &ext4_qctl_operations;
3474 sb->dq_op = &ext4_quota_operations;
3475 #endif
3476 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3477
3478 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3479 mutex_init(&sbi->s_orphan_lock);
3480 sbi->s_resize_flags = 0;
3481
3482 sb->s_root = NULL;
3483
3484 needs_recovery = (es->s_last_orphan != 0 ||
3485 EXT4_HAS_INCOMPAT_FEATURE(sb,
3486 EXT4_FEATURE_INCOMPAT_RECOVER));
3487
3488 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3489 !(sb->s_flags & MS_RDONLY))
3490 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3491 goto failed_mount3;
3492
3493 /*
3494 * The first inode we look at is the journal inode. Don't try
3495 * root first: it may be modified in the journal!
3496 */
3497 if (!test_opt(sb, NOLOAD) &&
3498 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3499 if (ext4_load_journal(sb, es, journal_devnum))
3500 goto failed_mount3;
3501 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3502 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3503 ext4_msg(sb, KERN_ERR, "required journal recovery "
3504 "suppressed and not mounted read-only");
3505 goto failed_mount_wq;
3506 } else {
3507 clear_opt(sb, DATA_FLAGS);
3508 sbi->s_journal = NULL;
3509 needs_recovery = 0;
3510 goto no_journal;
3511 }
3512
3513 if (ext4_blocks_count(es) > 0xffffffffULL &&
3514 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3515 JBD2_FEATURE_INCOMPAT_64BIT)) {
3516 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3517 goto failed_mount_wq;
3518 }
3519
3520 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3521 jbd2_journal_set_features(sbi->s_journal,
3522 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3523 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3524 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3525 jbd2_journal_set_features(sbi->s_journal,
3526 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
3527 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3528 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3529 } else {
3530 jbd2_journal_clear_features(sbi->s_journal,
3531 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3532 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3533 }
3534
3535 /* We have now updated the journal if required, so we can
3536 * validate the data journaling mode. */
3537 switch (test_opt(sb, DATA_FLAGS)) {
3538 case 0:
3539 /* No mode set, assume a default based on the journal
3540 * capabilities: ORDERED_DATA if the journal can
3541 * cope, else JOURNAL_DATA
3542 */
3543 if (jbd2_journal_check_available_features
3544 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3545 set_opt(sb, ORDERED_DATA);
3546 else
3547 set_opt(sb, JOURNAL_DATA);
3548 break;
3549
3550 case EXT4_MOUNT_ORDERED_DATA:
3551 case EXT4_MOUNT_WRITEBACK_DATA:
3552 if (!jbd2_journal_check_available_features
3553 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3554 ext4_msg(sb, KERN_ERR, "Journal does not support "
3555 "requested data journaling mode");
3556 goto failed_mount_wq;
3557 }
3558 default:
3559 break;
3560 }
3561 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3562
3563 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3564
3565 /*
3566 * The journal may have updated the bg summary counts, so we
3567 * need to update the global counters.
3568 */
3569 percpu_counter_set(&sbi->s_freeclusters_counter,
3570 ext4_count_free_clusters(sb));
3571 percpu_counter_set(&sbi->s_freeinodes_counter,
3572 ext4_count_free_inodes(sb));
3573 percpu_counter_set(&sbi->s_dirs_counter,
3574 ext4_count_dirs(sb));
3575 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3576
3577 no_journal:
3578 /*
3579 * The maximum number of concurrent works can be high and
3580 * concurrency isn't really necessary. Limit it to 1.
3581 */
3582 EXT4_SB(sb)->dio_unwritten_wq =
3583 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3584 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3585 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3586 goto failed_mount_wq;
3587 }
3588
3589 /*
3590 * The jbd2_journal_load will have done any necessary log recovery,
3591 * so we can safely mount the rest of the filesystem now.
3592 */
3593
3594 root = ext4_iget(sb, EXT4_ROOT_INO);
3595 if (IS_ERR(root)) {
3596 ext4_msg(sb, KERN_ERR, "get root inode failed");
3597 ret = PTR_ERR(root);
3598 root = NULL;
3599 goto failed_mount4;
3600 }
3601 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3602 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3603 iput(root);
3604 goto failed_mount4;
3605 }
3606 sb->s_root = d_make_root(root);
3607 if (!sb->s_root) {
3608 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3609 ret = -ENOMEM;
3610 goto failed_mount4;
3611 }
3612
3613 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
3614
3615 /* determine the minimum size of new large inodes, if present */
3616 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3617 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3618 EXT4_GOOD_OLD_INODE_SIZE;
3619 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3620 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3621 if (sbi->s_want_extra_isize <
3622 le16_to_cpu(es->s_want_extra_isize))
3623 sbi->s_want_extra_isize =
3624 le16_to_cpu(es->s_want_extra_isize);
3625 if (sbi->s_want_extra_isize <
3626 le16_to_cpu(es->s_min_extra_isize))
3627 sbi->s_want_extra_isize =
3628 le16_to_cpu(es->s_min_extra_isize);
3629 }
3630 }
3631 /* Check if enough inode space is available */
3632 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3633 sbi->s_inode_size) {
3634 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3635 EXT4_GOOD_OLD_INODE_SIZE;
3636 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3637 "available");
3638 }
3639
3640 err = ext4_setup_system_zone(sb);
3641 if (err) {
3642 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3643 "zone (%d)", err);
3644 goto failed_mount4a;
3645 }
3646
3647 ext4_ext_init(sb);
3648 err = ext4_mb_init(sb, needs_recovery);
3649 if (err) {
3650 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3651 err);
3652 goto failed_mount5;
3653 }
3654
3655 err = ext4_register_li_request(sb, first_not_zeroed);
3656 if (err)
3657 goto failed_mount6;
3658
3659 sbi->s_kobj.kset = ext4_kset;
3660 init_completion(&sbi->s_kobj_unregister);
3661 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3662 "%s", sb->s_id);
3663 if (err)
3664 goto failed_mount7;
3665
3666 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3667 ext4_orphan_cleanup(sb, es);
3668 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3669 if (needs_recovery) {
3670 ext4_msg(sb, KERN_INFO, "recovery complete");
3671 ext4_mark_recovery_complete(sb, es);
3672 }
3673 if (EXT4_SB(sb)->s_journal) {
3674 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3675 descr = " journalled data mode";
3676 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3677 descr = " ordered data mode";
3678 else
3679 descr = " writeback data mode";
3680 } else
3681 descr = "out journal";
3682
3683 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3684 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3685 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3686
3687 if (es->s_error_count)
3688 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3689
3690 kfree(orig_data);
3691 return 0;
3692
3693 cantfind_ext4:
3694 if (!silent)
3695 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3696 goto failed_mount;
3697
3698 failed_mount7:
3699 ext4_unregister_li_request(sb);
3700 failed_mount6:
3701 ext4_mb_release(sb);
3702 failed_mount5:
3703 ext4_ext_release(sb);
3704 ext4_release_system_zone(sb);
3705 failed_mount4a:
3706 dput(sb->s_root);
3707 sb->s_root = NULL;
3708 failed_mount4:
3709 ext4_msg(sb, KERN_ERR, "mount failed");
3710 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3711 failed_mount_wq:
3712 if (sbi->s_journal) {
3713 jbd2_journal_destroy(sbi->s_journal);
3714 sbi->s_journal = NULL;
3715 }
3716 failed_mount3:
3717 del_timer(&sbi->s_err_report);
3718 if (sbi->s_flex_groups)
3719 ext4_kvfree(sbi->s_flex_groups);
3720 percpu_counter_destroy(&sbi->s_freeclusters_counter);
3721 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3722 percpu_counter_destroy(&sbi->s_dirs_counter);
3723 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
3724 if (sbi->s_mmp_tsk)
3725 kthread_stop(sbi->s_mmp_tsk);
3726 failed_mount2:
3727 for (i = 0; i < db_count; i++)
3728 brelse(sbi->s_group_desc[i]);
3729 ext4_kvfree(sbi->s_group_desc);
3730 failed_mount:
3731 if (sbi->s_proc) {
3732 remove_proc_entry("options", sbi->s_proc);
3733 remove_proc_entry(sb->s_id, ext4_proc_root);
3734 }
3735 #ifdef CONFIG_QUOTA
3736 for (i = 0; i < MAXQUOTAS; i++)
3737 kfree(sbi->s_qf_names[i]);
3738 #endif
3739 ext4_blkdev_remove(sbi);
3740 brelse(bh);
3741 out_fail:
3742 sb->s_fs_info = NULL;
3743 kfree(sbi->s_blockgroup_lock);
3744 kfree(sbi);
3745 out_free_orig:
3746 kfree(orig_data);
3747 return ret;
3748 }
3749
3750 /*
3751 * Setup any per-fs journal parameters now. We'll do this both on
3752 * initial mount, once the journal has been initialised but before we've
3753 * done any recovery; and again on any subsequent remount.
3754 */
3755 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3756 {
3757 struct ext4_sb_info *sbi = EXT4_SB(sb);
3758
3759 journal->j_commit_interval = sbi->s_commit_interval;
3760 journal->j_min_batch_time = sbi->s_min_batch_time;
3761 journal->j_max_batch_time = sbi->s_max_batch_time;
3762
3763 write_lock(&journal->j_state_lock);
3764 if (test_opt(sb, BARRIER))
3765 journal->j_flags |= JBD2_BARRIER;
3766 else
3767 journal->j_flags &= ~JBD2_BARRIER;
3768 if (test_opt(sb, DATA_ERR_ABORT))
3769 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3770 else
3771 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3772 write_unlock(&journal->j_state_lock);
3773 }
3774
3775 static journal_t *ext4_get_journal(struct super_block *sb,
3776 unsigned int journal_inum)
3777 {
3778 struct inode *journal_inode;
3779 journal_t *journal;
3780
3781 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3782
3783 /* First, test for the existence of a valid inode on disk. Bad
3784 * things happen if we iget() an unused inode, as the subsequent
3785 * iput() will try to delete it. */
3786
3787 journal_inode = ext4_iget(sb, journal_inum);
3788 if (IS_ERR(journal_inode)) {
3789 ext4_msg(sb, KERN_ERR, "no journal found");
3790 return NULL;
3791 }
3792 if (!journal_inode->i_nlink) {
3793 make_bad_inode(journal_inode);
3794 iput(journal_inode);
3795 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
3796 return NULL;
3797 }
3798
3799 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
3800 journal_inode, journal_inode->i_size);
3801 if (!S_ISREG(journal_inode->i_mode)) {
3802 ext4_msg(sb, KERN_ERR, "invalid journal inode");
3803 iput(journal_inode);
3804 return NULL;
3805 }
3806
3807 journal = jbd2_journal_init_inode(journal_inode);
3808 if (!journal) {
3809 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
3810 iput(journal_inode);
3811 return NULL;
3812 }
3813 journal->j_private = sb;
3814 ext4_init_journal_params(sb, journal);
3815 return journal;
3816 }
3817
3818 static journal_t *ext4_get_dev_journal(struct super_block *sb,
3819 dev_t j_dev)
3820 {
3821 struct buffer_head *bh;
3822 journal_t *journal;
3823 ext4_fsblk_t start;
3824 ext4_fsblk_t len;
3825 int hblock, blocksize;
3826 ext4_fsblk_t sb_block;
3827 unsigned long offset;
3828 struct ext4_super_block *es;
3829 struct block_device *bdev;
3830
3831 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3832
3833 bdev = ext4_blkdev_get(j_dev, sb);
3834 if (bdev == NULL)
3835 return NULL;
3836
3837 blocksize = sb->s_blocksize;
3838 hblock = bdev_logical_block_size(bdev);
3839 if (blocksize < hblock) {
3840 ext4_msg(sb, KERN_ERR,
3841 "blocksize too small for journal device");
3842 goto out_bdev;
3843 }
3844
3845 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3846 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
3847 set_blocksize(bdev, blocksize);
3848 if (!(bh = __bread(bdev, sb_block, blocksize))) {
3849 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3850 "external journal");
3851 goto out_bdev;
3852 }
3853
3854 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3855 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
3856 !(le32_to_cpu(es->s_feature_incompat) &
3857 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
3858 ext4_msg(sb, KERN_ERR, "external journal has "
3859 "bad superblock");
3860 brelse(bh);
3861 goto out_bdev;
3862 }
3863
3864 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
3865 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
3866 brelse(bh);
3867 goto out_bdev;
3868 }
3869
3870 len = ext4_blocks_count(es);
3871 start = sb_block + 1;
3872 brelse(bh); /* we're done with the superblock */
3873
3874 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
3875 start, len, blocksize);
3876 if (!journal) {
3877 ext4_msg(sb, KERN_ERR, "failed to create device journal");
3878 goto out_bdev;
3879 }
3880 journal->j_private = sb;
3881 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3882 wait_on_buffer(journal->j_sb_buffer);
3883 if (!buffer_uptodate(journal->j_sb_buffer)) {
3884 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
3885 goto out_journal;
3886 }
3887 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
3888 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3889 "user (unsupported) - %d",
3890 be32_to_cpu(journal->j_superblock->s_nr_users));
3891 goto out_journal;
3892 }
3893 EXT4_SB(sb)->journal_bdev = bdev;
3894 ext4_init_journal_params(sb, journal);
3895 return journal;
3896
3897 out_journal:
3898 jbd2_journal_destroy(journal);
3899 out_bdev:
3900 ext4_blkdev_put(bdev);
3901 return NULL;
3902 }
3903
3904 static int ext4_load_journal(struct super_block *sb,
3905 struct ext4_super_block *es,
3906 unsigned long journal_devnum)
3907 {
3908 journal_t *journal;
3909 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3910 dev_t journal_dev;
3911 int err = 0;
3912 int really_read_only;
3913
3914 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3915
3916 if (journal_devnum &&
3917 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3918 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3919 "numbers have changed");
3920 journal_dev = new_decode_dev(journal_devnum);
3921 } else
3922 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3923
3924 really_read_only = bdev_read_only(sb->s_bdev);
3925
3926 /*
3927 * Are we loading a blank journal or performing recovery after a
3928 * crash? For recovery, we need to check in advance whether we
3929 * can get read-write access to the device.
3930 */
3931 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3932 if (sb->s_flags & MS_RDONLY) {
3933 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3934 "required on readonly filesystem");
3935 if (really_read_only) {
3936 ext4_msg(sb, KERN_ERR, "write access "
3937 "unavailable, cannot proceed");
3938 return -EROFS;
3939 }
3940 ext4_msg(sb, KERN_INFO, "write access will "
3941 "be enabled during recovery");
3942 }
3943 }
3944
3945 if (journal_inum && journal_dev) {
3946 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3947 "and inode journals!");
3948 return -EINVAL;
3949 }
3950
3951 if (journal_inum) {
3952 if (!(journal = ext4_get_journal(sb, journal_inum)))
3953 return -EINVAL;
3954 } else {
3955 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
3956 return -EINVAL;
3957 }
3958
3959 if (!(journal->j_flags & JBD2_BARRIER))
3960 ext4_msg(sb, KERN_INFO, "barriers disabled");
3961
3962 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3963 err = jbd2_journal_wipe(journal, !really_read_only);
3964 if (!err) {
3965 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3966 if (save)
3967 memcpy(save, ((char *) es) +
3968 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3969 err = jbd2_journal_load(journal);
3970 if (save)
3971 memcpy(((char *) es) + EXT4_S_ERR_START,
3972 save, EXT4_S_ERR_LEN);
3973 kfree(save);
3974 }
3975
3976 if (err) {
3977 ext4_msg(sb, KERN_ERR, "error loading journal");
3978 jbd2_journal_destroy(journal);
3979 return err;
3980 }
3981
3982 EXT4_SB(sb)->s_journal = journal;
3983 ext4_clear_journal_err(sb, es);
3984
3985 if (!really_read_only && journal_devnum &&
3986 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3987 es->s_journal_dev = cpu_to_le32(journal_devnum);
3988
3989 /* Make sure we flush the recovery flag to disk. */
3990 ext4_commit_super(sb, 1);
3991 }
3992
3993 return 0;
3994 }
3995
3996 static int ext4_commit_super(struct super_block *sb, int sync)
3997 {
3998 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
3999 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4000 int error = 0;
4001
4002 if (!sbh || block_device_ejected(sb))
4003 return error;
4004 if (buffer_write_io_error(sbh)) {
4005 /*
4006 * Oh, dear. A previous attempt to write the
4007 * superblock failed. This could happen because the
4008 * USB device was yanked out. Or it could happen to
4009 * be a transient write error and maybe the block will
4010 * be remapped. Nothing we can do but to retry the
4011 * write and hope for the best.
4012 */
4013 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4014 "superblock detected");
4015 clear_buffer_write_io_error(sbh);
4016 set_buffer_uptodate(sbh);
4017 }
4018 /*
4019 * If the file system is mounted read-only, don't update the
4020 * superblock write time. This avoids updating the superblock
4021 * write time when we are mounting the root file system
4022 * read/only but we need to replay the journal; at that point,
4023 * for people who are east of GMT and who make their clock
4024 * tick in localtime for Windows bug-for-bug compatibility,
4025 * the clock is set in the future, and this will cause e2fsck
4026 * to complain and force a full file system check.
4027 */
4028 if (!(sb->s_flags & MS_RDONLY))
4029 es->s_wtime = cpu_to_le32(get_seconds());
4030 if (sb->s_bdev->bd_part)
4031 es->s_kbytes_written =
4032 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4033 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4034 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4035 else
4036 es->s_kbytes_written =
4037 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4038 ext4_free_blocks_count_set(es,
4039 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4040 &EXT4_SB(sb)->s_freeclusters_counter)));
4041 es->s_free_inodes_count =
4042 cpu_to_le32(percpu_counter_sum_positive(
4043 &EXT4_SB(sb)->s_freeinodes_counter));
4044 sb->s_dirt = 0;
4045 BUFFER_TRACE(sbh, "marking dirty");
4046 mark_buffer_dirty(sbh);
4047 if (sync) {
4048 error = sync_dirty_buffer(sbh);
4049 if (error)
4050 return error;
4051
4052 error = buffer_write_io_error(sbh);
4053 if (error) {
4054 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4055 "superblock");
4056 clear_buffer_write_io_error(sbh);
4057 set_buffer_uptodate(sbh);
4058 }
4059 }
4060 return error;
4061 }
4062
4063 /*
4064 * Have we just finished recovery? If so, and if we are mounting (or
4065 * remounting) the filesystem readonly, then we will end up with a
4066 * consistent fs on disk. Record that fact.
4067 */
4068 static void ext4_mark_recovery_complete(struct super_block *sb,
4069 struct ext4_super_block *es)
4070 {
4071 journal_t *journal = EXT4_SB(sb)->s_journal;
4072
4073 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4074 BUG_ON(journal != NULL);
4075 return;
4076 }
4077 jbd2_journal_lock_updates(journal);
4078 if (jbd2_journal_flush(journal) < 0)
4079 goto out;
4080
4081 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4082 sb->s_flags & MS_RDONLY) {
4083 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4084 ext4_commit_super(sb, 1);
4085 }
4086
4087 out:
4088 jbd2_journal_unlock_updates(journal);
4089 }
4090
4091 /*
4092 * If we are mounting (or read-write remounting) a filesystem whose journal
4093 * has recorded an error from a previous lifetime, move that error to the
4094 * main filesystem now.
4095 */
4096 static void ext4_clear_journal_err(struct super_block *sb,
4097 struct ext4_super_block *es)
4098 {
4099 journal_t *journal;
4100 int j_errno;
4101 const char *errstr;
4102
4103 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4104
4105 journal = EXT4_SB(sb)->s_journal;
4106
4107 /*
4108 * Now check for any error status which may have been recorded in the
4109 * journal by a prior ext4_error() or ext4_abort()
4110 */
4111
4112 j_errno = jbd2_journal_errno(journal);
4113 if (j_errno) {
4114 char nbuf[16];
4115
4116 errstr = ext4_decode_error(sb, j_errno, nbuf);
4117 ext4_warning(sb, "Filesystem error recorded "
4118 "from previous mount: %s", errstr);
4119 ext4_warning(sb, "Marking fs in need of filesystem check.");
4120
4121 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4122 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4123 ext4_commit_super(sb, 1);
4124
4125 jbd2_journal_clear_err(journal);
4126 }
4127 }
4128
4129 /*
4130 * Force the running and committing transactions to commit,
4131 * and wait on the commit.
4132 */
4133 int ext4_force_commit(struct super_block *sb)
4134 {
4135 journal_t *journal;
4136 int ret = 0;
4137
4138 if (sb->s_flags & MS_RDONLY)
4139 return 0;
4140
4141 journal = EXT4_SB(sb)->s_journal;
4142 if (journal) {
4143 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4144 ret = ext4_journal_force_commit(journal);
4145 }
4146
4147 return ret;
4148 }
4149
4150 static void ext4_write_super(struct super_block *sb)
4151 {
4152 lock_super(sb);
4153 ext4_commit_super(sb, 1);
4154 unlock_super(sb);
4155 }
4156
4157 static int ext4_sync_fs(struct super_block *sb, int wait)
4158 {
4159 int ret = 0;
4160 tid_t target;
4161 struct ext4_sb_info *sbi = EXT4_SB(sb);
4162
4163 trace_ext4_sync_fs(sb, wait);
4164 flush_workqueue(sbi->dio_unwritten_wq);
4165 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4166 if (wait)
4167 jbd2_log_wait_commit(sbi->s_journal, target);
4168 }
4169 return ret;
4170 }
4171
4172 /*
4173 * LVM calls this function before a (read-only) snapshot is created. This
4174 * gives us a chance to flush the journal completely and mark the fs clean.
4175 *
4176 * Note that only this function cannot bring a filesystem to be in a clean
4177 * state independently, because ext4 prevents a new handle from being started
4178 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4179 * the upper layer.
4180 */
4181 static int ext4_freeze(struct super_block *sb)
4182 {
4183 int error = 0;
4184 journal_t *journal;
4185
4186 if (sb->s_flags & MS_RDONLY)
4187 return 0;
4188
4189 journal = EXT4_SB(sb)->s_journal;
4190
4191 /* Now we set up the journal barrier. */
4192 jbd2_journal_lock_updates(journal);
4193
4194 /*
4195 * Don't clear the needs_recovery flag if we failed to flush
4196 * the journal.
4197 */
4198 error = jbd2_journal_flush(journal);
4199 if (error < 0)
4200 goto out;
4201
4202 /* Journal blocked and flushed, clear needs_recovery flag. */
4203 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4204 error = ext4_commit_super(sb, 1);
4205 out:
4206 /* we rely on s_frozen to stop further updates */
4207 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4208 return error;
4209 }
4210
4211 /*
4212 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4213 * flag here, even though the filesystem is not technically dirty yet.
4214 */
4215 static int ext4_unfreeze(struct super_block *sb)
4216 {
4217 if (sb->s_flags & MS_RDONLY)
4218 return 0;
4219
4220 lock_super(sb);
4221 /* Reset the needs_recovery flag before the fs is unlocked. */
4222 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4223 ext4_commit_super(sb, 1);
4224 unlock_super(sb);
4225 return 0;
4226 }
4227
4228 /*
4229 * Structure to save mount options for ext4_remount's benefit
4230 */
4231 struct ext4_mount_options {
4232 unsigned long s_mount_opt;
4233 unsigned long s_mount_opt2;
4234 uid_t s_resuid;
4235 gid_t s_resgid;
4236 unsigned long s_commit_interval;
4237 u32 s_min_batch_time, s_max_batch_time;
4238 #ifdef CONFIG_QUOTA
4239 int s_jquota_fmt;
4240 char *s_qf_names[MAXQUOTAS];
4241 #endif
4242 };
4243
4244 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4245 {
4246 struct ext4_super_block *es;
4247 struct ext4_sb_info *sbi = EXT4_SB(sb);
4248 unsigned long old_sb_flags;
4249 struct ext4_mount_options old_opts;
4250 int enable_quota = 0;
4251 ext4_group_t g;
4252 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4253 int err = 0;
4254 #ifdef CONFIG_QUOTA
4255 int i;
4256 #endif
4257 char *orig_data = kstrdup(data, GFP_KERNEL);
4258
4259 /* Store the original options */
4260 lock_super(sb);
4261 old_sb_flags = sb->s_flags;
4262 old_opts.s_mount_opt = sbi->s_mount_opt;
4263 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4264 old_opts.s_resuid = sbi->s_resuid;
4265 old_opts.s_resgid = sbi->s_resgid;
4266 old_opts.s_commit_interval = sbi->s_commit_interval;
4267 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4268 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4269 #ifdef CONFIG_QUOTA
4270 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4271 for (i = 0; i < MAXQUOTAS; i++)
4272 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4273 #endif
4274 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4275 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4276
4277 /*
4278 * Allow the "check" option to be passed as a remount option.
4279 */
4280 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4281 err = -EINVAL;
4282 goto restore_opts;
4283 }
4284
4285 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4286 ext4_abort(sb, "Abort forced by user");
4287
4288 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4289 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4290
4291 es = sbi->s_es;
4292
4293 if (sbi->s_journal) {
4294 ext4_init_journal_params(sb, sbi->s_journal);
4295 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4296 }
4297
4298 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4299 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4300 err = -EROFS;
4301 goto restore_opts;
4302 }
4303
4304 if (*flags & MS_RDONLY) {
4305 err = dquot_suspend(sb, -1);
4306 if (err < 0)
4307 goto restore_opts;
4308
4309 /*
4310 * First of all, the unconditional stuff we have to do
4311 * to disable replay of the journal when we next remount
4312 */
4313 sb->s_flags |= MS_RDONLY;
4314
4315 /*
4316 * OK, test if we are remounting a valid rw partition
4317 * readonly, and if so set the rdonly flag and then
4318 * mark the partition as valid again.
4319 */
4320 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4321 (sbi->s_mount_state & EXT4_VALID_FS))
4322 es->s_state = cpu_to_le16(sbi->s_mount_state);
4323
4324 if (sbi->s_journal)
4325 ext4_mark_recovery_complete(sb, es);
4326 } else {
4327 /* Make sure we can mount this feature set readwrite */
4328 if (!ext4_feature_set_ok(sb, 0)) {
4329 err = -EROFS;
4330 goto restore_opts;
4331 }
4332 /*
4333 * Make sure the group descriptor checksums
4334 * are sane. If they aren't, refuse to remount r/w.
4335 */
4336 for (g = 0; g < sbi->s_groups_count; g++) {
4337 struct ext4_group_desc *gdp =
4338 ext4_get_group_desc(sb, g, NULL);
4339
4340 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
4341 ext4_msg(sb, KERN_ERR,
4342 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4343 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4344 le16_to_cpu(gdp->bg_checksum));
4345 err = -EINVAL;
4346 goto restore_opts;
4347 }
4348 }
4349
4350 /*
4351 * If we have an unprocessed orphan list hanging
4352 * around from a previously readonly bdev mount,
4353 * require a full umount/remount for now.
4354 */
4355 if (es->s_last_orphan) {
4356 ext4_msg(sb, KERN_WARNING, "Couldn't "
4357 "remount RDWR because of unprocessed "
4358 "orphan inode list. Please "
4359 "umount/remount instead");
4360 err = -EINVAL;
4361 goto restore_opts;
4362 }
4363
4364 /*
4365 * Mounting a RDONLY partition read-write, so reread
4366 * and store the current valid flag. (It may have
4367 * been changed by e2fsck since we originally mounted
4368 * the partition.)
4369 */
4370 if (sbi->s_journal)
4371 ext4_clear_journal_err(sb, es);
4372 sbi->s_mount_state = le16_to_cpu(es->s_state);
4373 if (!ext4_setup_super(sb, es, 0))
4374 sb->s_flags &= ~MS_RDONLY;
4375 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4376 EXT4_FEATURE_INCOMPAT_MMP))
4377 if (ext4_multi_mount_protect(sb,
4378 le64_to_cpu(es->s_mmp_block))) {
4379 err = -EROFS;
4380 goto restore_opts;
4381 }
4382 enable_quota = 1;
4383 }
4384 }
4385
4386 /*
4387 * Reinitialize lazy itable initialization thread based on
4388 * current settings
4389 */
4390 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4391 ext4_unregister_li_request(sb);
4392 else {
4393 ext4_group_t first_not_zeroed;
4394 first_not_zeroed = ext4_has_uninit_itable(sb);
4395 ext4_register_li_request(sb, first_not_zeroed);
4396 }
4397
4398 ext4_setup_system_zone(sb);
4399 if (sbi->s_journal == NULL)
4400 ext4_commit_super(sb, 1);
4401
4402 #ifdef CONFIG_QUOTA
4403 /* Release old quota file names */
4404 for (i = 0; i < MAXQUOTAS; i++)
4405 if (old_opts.s_qf_names[i] &&
4406 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4407 kfree(old_opts.s_qf_names[i]);
4408 #endif
4409 unlock_super(sb);
4410 if (enable_quota)
4411 dquot_resume(sb, -1);
4412
4413 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4414 kfree(orig_data);
4415 return 0;
4416
4417 restore_opts:
4418 sb->s_flags = old_sb_flags;
4419 sbi->s_mount_opt = old_opts.s_mount_opt;
4420 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4421 sbi->s_resuid = old_opts.s_resuid;
4422 sbi->s_resgid = old_opts.s_resgid;
4423 sbi->s_commit_interval = old_opts.s_commit_interval;
4424 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4425 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4426 #ifdef CONFIG_QUOTA
4427 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4428 for (i = 0; i < MAXQUOTAS; i++) {
4429 if (sbi->s_qf_names[i] &&
4430 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4431 kfree(sbi->s_qf_names[i]);
4432 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4433 }
4434 #endif
4435 unlock_super(sb);
4436 kfree(orig_data);
4437 return err;
4438 }
4439
4440 /*
4441 * Note: calculating the overhead so we can be compatible with
4442 * historical BSD practice is quite difficult in the face of
4443 * clusters/bigalloc. This is because multiple metadata blocks from
4444 * different block group can end up in the same allocation cluster.
4445 * Calculating the exact overhead in the face of clustered allocation
4446 * requires either O(all block bitmaps) in memory or O(number of block
4447 * groups**2) in time. We will still calculate the superblock for
4448 * older file systems --- and if we come across with a bigalloc file
4449 * system with zero in s_overhead_clusters the estimate will be close to
4450 * correct especially for very large cluster sizes --- but for newer
4451 * file systems, it's better to calculate this figure once at mkfs
4452 * time, and store it in the superblock. If the superblock value is
4453 * present (even for non-bigalloc file systems), we will use it.
4454 */
4455 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4456 {
4457 struct super_block *sb = dentry->d_sb;
4458 struct ext4_sb_info *sbi = EXT4_SB(sb);
4459 struct ext4_super_block *es = sbi->s_es;
4460 struct ext4_group_desc *gdp;
4461 u64 fsid;
4462 s64 bfree;
4463
4464 if (test_opt(sb, MINIX_DF)) {
4465 sbi->s_overhead_last = 0;
4466 } else if (es->s_overhead_clusters) {
4467 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
4468 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
4469 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4470 ext4_fsblk_t overhead = 0;
4471
4472 /*
4473 * Compute the overhead (FS structures). This is constant
4474 * for a given filesystem unless the number of block groups
4475 * changes so we cache the previous value until it does.
4476 */
4477
4478 /*
4479 * All of the blocks before first_data_block are
4480 * overhead
4481 */
4482 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4483
4484 /*
4485 * Add the overhead found in each block group
4486 */
4487 for (i = 0; i < ngroups; i++) {
4488 gdp = ext4_get_group_desc(sb, i, NULL);
4489 overhead += ext4_num_overhead_clusters(sb, i, gdp);
4490 cond_resched();
4491 }
4492 sbi->s_overhead_last = overhead;
4493 smp_wmb();
4494 sbi->s_blocks_last = ext4_blocks_count(es);
4495 }
4496
4497 buf->f_type = EXT4_SUPER_MAGIC;
4498 buf->f_bsize = sb->s_blocksize;
4499 buf->f_blocks = (ext4_blocks_count(es) -
4500 EXT4_C2B(sbi, sbi->s_overhead_last));
4501 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4502 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4503 /* prevent underflow in case that few free space is available */
4504 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4505 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4506 if (buf->f_bfree < ext4_r_blocks_count(es))
4507 buf->f_bavail = 0;
4508 buf->f_files = le32_to_cpu(es->s_inodes_count);
4509 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4510 buf->f_namelen = EXT4_NAME_LEN;
4511 fsid = le64_to_cpup((void *)es->s_uuid) ^
4512 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4513 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4514 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4515
4516 return 0;
4517 }
4518
4519 /* Helper function for writing quotas on sync - we need to start transaction
4520 * before quota file is locked for write. Otherwise the are possible deadlocks:
4521 * Process 1 Process 2
4522 * ext4_create() quota_sync()
4523 * jbd2_journal_start() write_dquot()
4524 * dquot_initialize() down(dqio_mutex)
4525 * down(dqio_mutex) jbd2_journal_start()
4526 *
4527 */
4528
4529 #ifdef CONFIG_QUOTA
4530
4531 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4532 {
4533 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4534 }
4535
4536 static int ext4_write_dquot(struct dquot *dquot)
4537 {
4538 int ret, err;
4539 handle_t *handle;
4540 struct inode *inode;
4541
4542 inode = dquot_to_inode(dquot);
4543 handle = ext4_journal_start(inode,
4544 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4545 if (IS_ERR(handle))
4546 return PTR_ERR(handle);
4547 ret = dquot_commit(dquot);
4548 err = ext4_journal_stop(handle);
4549 if (!ret)
4550 ret = err;
4551 return ret;
4552 }
4553
4554 static int ext4_acquire_dquot(struct dquot *dquot)
4555 {
4556 int ret, err;
4557 handle_t *handle;
4558
4559 handle = ext4_journal_start(dquot_to_inode(dquot),
4560 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4561 if (IS_ERR(handle))
4562 return PTR_ERR(handle);
4563 ret = dquot_acquire(dquot);
4564 err = ext4_journal_stop(handle);
4565 if (!ret)
4566 ret = err;
4567 return ret;
4568 }
4569
4570 static int ext4_release_dquot(struct dquot *dquot)
4571 {
4572 int ret, err;
4573 handle_t *handle;
4574
4575 handle = ext4_journal_start(dquot_to_inode(dquot),
4576 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4577 if (IS_ERR(handle)) {
4578 /* Release dquot anyway to avoid endless cycle in dqput() */
4579 dquot_release(dquot);
4580 return PTR_ERR(handle);
4581 }
4582 ret = dquot_release(dquot);
4583 err = ext4_journal_stop(handle);
4584 if (!ret)
4585 ret = err;
4586 return ret;
4587 }
4588
4589 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4590 {
4591 /* Are we journaling quotas? */
4592 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4593 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4594 dquot_mark_dquot_dirty(dquot);
4595 return ext4_write_dquot(dquot);
4596 } else {
4597 return dquot_mark_dquot_dirty(dquot);
4598 }
4599 }
4600
4601 static int ext4_write_info(struct super_block *sb, int type)
4602 {
4603 int ret, err;
4604 handle_t *handle;
4605
4606 /* Data block + inode block */
4607 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4608 if (IS_ERR(handle))
4609 return PTR_ERR(handle);
4610 ret = dquot_commit_info(sb, type);
4611 err = ext4_journal_stop(handle);
4612 if (!ret)
4613 ret = err;
4614 return ret;
4615 }
4616
4617 /*
4618 * Turn on quotas during mount time - we need to find
4619 * the quota file and such...
4620 */
4621 static int ext4_quota_on_mount(struct super_block *sb, int type)
4622 {
4623 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4624 EXT4_SB(sb)->s_jquota_fmt, type);
4625 }
4626
4627 /*
4628 * Standard function to be called on quota_on
4629 */
4630 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4631 struct path *path)
4632 {
4633 int err;
4634
4635 if (!test_opt(sb, QUOTA))
4636 return -EINVAL;
4637
4638 /* Quotafile not on the same filesystem? */
4639 if (path->dentry->d_sb != sb)
4640 return -EXDEV;
4641 /* Journaling quota? */
4642 if (EXT4_SB(sb)->s_qf_names[type]) {
4643 /* Quotafile not in fs root? */
4644 if (path->dentry->d_parent != sb->s_root)
4645 ext4_msg(sb, KERN_WARNING,
4646 "Quota file not on filesystem root. "
4647 "Journaled quota will not work");
4648 }
4649
4650 /*
4651 * When we journal data on quota file, we have to flush journal to see
4652 * all updates to the file when we bypass pagecache...
4653 */
4654 if (EXT4_SB(sb)->s_journal &&
4655 ext4_should_journal_data(path->dentry->d_inode)) {
4656 /*
4657 * We don't need to lock updates but journal_flush() could
4658 * otherwise be livelocked...
4659 */
4660 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4661 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4662 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4663 if (err)
4664 return err;
4665 }
4666
4667 return dquot_quota_on(sb, type, format_id, path);
4668 }
4669
4670 static int ext4_quota_off(struct super_block *sb, int type)
4671 {
4672 struct inode *inode = sb_dqopt(sb)->files[type];
4673 handle_t *handle;
4674
4675 /* Force all delayed allocation blocks to be allocated.
4676 * Caller already holds s_umount sem */
4677 if (test_opt(sb, DELALLOC))
4678 sync_filesystem(sb);
4679
4680 if (!inode)
4681 goto out;
4682
4683 /* Update modification times of quota files when userspace can
4684 * start looking at them */
4685 handle = ext4_journal_start(inode, 1);
4686 if (IS_ERR(handle))
4687 goto out;
4688 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4689 ext4_mark_inode_dirty(handle, inode);
4690 ext4_journal_stop(handle);
4691
4692 out:
4693 return dquot_quota_off(sb, type);
4694 }
4695
4696 /* Read data from quotafile - avoid pagecache and such because we cannot afford
4697 * acquiring the locks... As quota files are never truncated and quota code
4698 * itself serializes the operations (and no one else should touch the files)
4699 * we don't have to be afraid of races */
4700 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4701 size_t len, loff_t off)
4702 {
4703 struct inode *inode = sb_dqopt(sb)->files[type];
4704 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4705 int err = 0;
4706 int offset = off & (sb->s_blocksize - 1);
4707 int tocopy;
4708 size_t toread;
4709 struct buffer_head *bh;
4710 loff_t i_size = i_size_read(inode);
4711
4712 if (off > i_size)
4713 return 0;
4714 if (off+len > i_size)
4715 len = i_size-off;
4716 toread = len;
4717 while (toread > 0) {
4718 tocopy = sb->s_blocksize - offset < toread ?
4719 sb->s_blocksize - offset : toread;
4720 bh = ext4_bread(NULL, inode, blk, 0, &err);
4721 if (err)
4722 return err;
4723 if (!bh) /* A hole? */
4724 memset(data, 0, tocopy);
4725 else
4726 memcpy(data, bh->b_data+offset, tocopy);
4727 brelse(bh);
4728 offset = 0;
4729 toread -= tocopy;
4730 data += tocopy;
4731 blk++;
4732 }
4733 return len;
4734 }
4735
4736 /* Write to quotafile (we know the transaction is already started and has
4737 * enough credits) */
4738 static ssize_t ext4_quota_write(struct super_block *sb, int type,
4739 const char *data, size_t len, loff_t off)
4740 {
4741 struct inode *inode = sb_dqopt(sb)->files[type];
4742 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4743 int err = 0;
4744 int offset = off & (sb->s_blocksize - 1);
4745 struct buffer_head *bh;
4746 handle_t *handle = journal_current_handle();
4747
4748 if (EXT4_SB(sb)->s_journal && !handle) {
4749 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4750 " cancelled because transaction is not started",
4751 (unsigned long long)off, (unsigned long long)len);
4752 return -EIO;
4753 }
4754 /*
4755 * Since we account only one data block in transaction credits,
4756 * then it is impossible to cross a block boundary.
4757 */
4758 if (sb->s_blocksize - offset < len) {
4759 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4760 " cancelled because not block aligned",
4761 (unsigned long long)off, (unsigned long long)len);
4762 return -EIO;
4763 }
4764
4765 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4766 bh = ext4_bread(handle, inode, blk, 1, &err);
4767 if (!bh)
4768 goto out;
4769 err = ext4_journal_get_write_access(handle, bh);
4770 if (err) {
4771 brelse(bh);
4772 goto out;
4773 }
4774 lock_buffer(bh);
4775 memcpy(bh->b_data+offset, data, len);
4776 flush_dcache_page(bh->b_page);
4777 unlock_buffer(bh);
4778 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4779 brelse(bh);
4780 out:
4781 if (err) {
4782 mutex_unlock(&inode->i_mutex);
4783 return err;
4784 }
4785 if (inode->i_size < off + len) {
4786 i_size_write(inode, off + len);
4787 EXT4_I(inode)->i_disksize = inode->i_size;
4788 ext4_mark_inode_dirty(handle, inode);
4789 }
4790 mutex_unlock(&inode->i_mutex);
4791 return len;
4792 }
4793
4794 #endif
4795
4796 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4797 const char *dev_name, void *data)
4798 {
4799 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
4800 }
4801
4802 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4803 static inline void register_as_ext2(void)
4804 {
4805 int err = register_filesystem(&ext2_fs_type);
4806 if (err)
4807 printk(KERN_WARNING
4808 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4809 }
4810
4811 static inline void unregister_as_ext2(void)
4812 {
4813 unregister_filesystem(&ext2_fs_type);
4814 }
4815
4816 static inline int ext2_feature_set_ok(struct super_block *sb)
4817 {
4818 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4819 return 0;
4820 if (sb->s_flags & MS_RDONLY)
4821 return 1;
4822 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4823 return 0;
4824 return 1;
4825 }
4826 MODULE_ALIAS("ext2");
4827 #else
4828 static inline void register_as_ext2(void) { }
4829 static inline void unregister_as_ext2(void) { }
4830 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
4831 #endif
4832
4833 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
4834 static inline void register_as_ext3(void)
4835 {
4836 int err = register_filesystem(&ext3_fs_type);
4837 if (err)
4838 printk(KERN_WARNING
4839 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4840 }
4841
4842 static inline void unregister_as_ext3(void)
4843 {
4844 unregister_filesystem(&ext3_fs_type);
4845 }
4846
4847 static inline int ext3_feature_set_ok(struct super_block *sb)
4848 {
4849 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4850 return 0;
4851 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4852 return 0;
4853 if (sb->s_flags & MS_RDONLY)
4854 return 1;
4855 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
4856 return 0;
4857 return 1;
4858 }
4859 MODULE_ALIAS("ext3");
4860 #else
4861 static inline void register_as_ext3(void) { }
4862 static inline void unregister_as_ext3(void) { }
4863 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
4864 #endif
4865
4866 static struct file_system_type ext4_fs_type = {
4867 .owner = THIS_MODULE,
4868 .name = "ext4",
4869 .mount = ext4_mount,
4870 .kill_sb = kill_block_super,
4871 .fs_flags = FS_REQUIRES_DEV,
4872 };
4873
4874 static int __init ext4_init_feat_adverts(void)
4875 {
4876 struct ext4_features *ef;
4877 int ret = -ENOMEM;
4878
4879 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4880 if (!ef)
4881 goto out;
4882
4883 ef->f_kobj.kset = ext4_kset;
4884 init_completion(&ef->f_kobj_unregister);
4885 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4886 "features");
4887 if (ret) {
4888 kfree(ef);
4889 goto out;
4890 }
4891
4892 ext4_feat = ef;
4893 ret = 0;
4894 out:
4895 return ret;
4896 }
4897
4898 static void ext4_exit_feat_adverts(void)
4899 {
4900 kobject_put(&ext4_feat->f_kobj);
4901 wait_for_completion(&ext4_feat->f_kobj_unregister);
4902 kfree(ext4_feat);
4903 }
4904
4905 /* Shared across all ext4 file systems */
4906 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4907 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4908
4909 static int __init ext4_init_fs(void)
4910 {
4911 int i, err;
4912
4913 ext4_li_info = NULL;
4914 mutex_init(&ext4_li_mtx);
4915
4916 ext4_check_flag_values();
4917
4918 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4919 mutex_init(&ext4__aio_mutex[i]);
4920 init_waitqueue_head(&ext4__ioend_wq[i]);
4921 }
4922
4923 err = ext4_init_pageio();
4924 if (err)
4925 return err;
4926 err = ext4_init_system_zone();
4927 if (err)
4928 goto out6;
4929 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4930 if (!ext4_kset)
4931 goto out5;
4932 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4933
4934 err = ext4_init_feat_adverts();
4935 if (err)
4936 goto out4;
4937
4938 err = ext4_init_mballoc();
4939 if (err)
4940 goto out3;
4941
4942 err = ext4_init_xattr();
4943 if (err)
4944 goto out2;
4945 err = init_inodecache();
4946 if (err)
4947 goto out1;
4948 register_as_ext3();
4949 register_as_ext2();
4950 err = register_filesystem(&ext4_fs_type);
4951 if (err)
4952 goto out;
4953
4954 return 0;
4955 out:
4956 unregister_as_ext2();
4957 unregister_as_ext3();
4958 destroy_inodecache();
4959 out1:
4960 ext4_exit_xattr();
4961 out2:
4962 ext4_exit_mballoc();
4963 out3:
4964 ext4_exit_feat_adverts();
4965 out4:
4966 if (ext4_proc_root)
4967 remove_proc_entry("fs/ext4", NULL);
4968 kset_unregister(ext4_kset);
4969 out5:
4970 ext4_exit_system_zone();
4971 out6:
4972 ext4_exit_pageio();
4973 return err;
4974 }
4975
4976 static void __exit ext4_exit_fs(void)
4977 {
4978 ext4_destroy_lazyinit_thread();
4979 unregister_as_ext2();
4980 unregister_as_ext3();
4981 unregister_filesystem(&ext4_fs_type);
4982 destroy_inodecache();
4983 ext4_exit_xattr();
4984 ext4_exit_mballoc();
4985 ext4_exit_feat_adverts();
4986 remove_proc_entry("fs/ext4", NULL);
4987 kset_unregister(ext4_kset);
4988 ext4_exit_system_zone();
4989 ext4_exit_pageio();
4990 }
4991
4992 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4993 MODULE_DESCRIPTION("Fourth Extended Filesystem");
4994 MODULE_LICENSE("GPL");
4995 module_init(ext4_init_fs)
4996 module_exit(ext4_exit_fs)
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree