projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge tag 'v3.10.107' into update
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / ext4 / super.c
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" /* Needed for trace points definition */
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 int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
78 static inline int ext2_feature_set_ok(struct super_block *sb);
79 static inline int ext3_feature_set_ok(struct super_block *sb);
80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
84 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
85
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type = {
88 .owner = THIS_MODULE,
89 .name = "ext2",
90 .mount = ext4_mount,
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
93 };
94 MODULE_ALIAS_FS("ext2");
95 MODULE_ALIAS("ext2");
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 MODULE_ALIAS_FS("ext3");
111 MODULE_ALIAS("ext3");
112 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
113 #else
114 #define IS_EXT3_SB(sb) (0)
115 #endif
116
117 static int ext4_verify_csum_type(struct super_block *sb,
118 struct ext4_super_block *es)
119 {
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
122 return 1;
123
124 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
125 }
126
127 static __le32 ext4_superblock_csum(struct super_block *sb,
128 struct ext4_super_block *es)
129 {
130 struct ext4_sb_info *sbi = EXT4_SB(sb);
131 int offset = offsetof(struct ext4_super_block, s_checksum);
132 __u32 csum;
133
134 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
135
136 return cpu_to_le32(csum);
137 }
138
139 int ext4_superblock_csum_verify(struct super_block *sb,
140 struct ext4_super_block *es)
141 {
142 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
143 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
144 return 1;
145
146 return es->s_checksum == ext4_superblock_csum(sb, es);
147 }
148
149 void ext4_superblock_csum_set(struct super_block *sb)
150 {
151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152
153 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
154 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
155 return;
156
157 es->s_checksum = ext4_superblock_csum(sb, es);
158 }
159
160 void *ext4_kvmalloc(size_t size, gfp_t flags)
161 {
162 void *ret;
163
164 ret = kmalloc(size, flags);
165 if (!ret)
166 ret = __vmalloc(size, flags, PAGE_KERNEL);
167 return ret;
168 }
169
170 void *ext4_kvzalloc(size_t size, gfp_t flags)
171 {
172 void *ret;
173
174 ret = kzalloc(size, flags);
175 if (!ret)
176 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
177 return ret;
178 }
179
180 void ext4_kvfree(void *ptr)
181 {
182 if (is_vmalloc_addr(ptr))
183 vfree(ptr);
184 else
185 kfree(ptr);
186
187 }
188
189 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
190 struct ext4_group_desc *bg)
191 {
192 return le32_to_cpu(bg->bg_block_bitmap_lo) |
193 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
194 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
195 }
196
197 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
198 struct ext4_group_desc *bg)
199 {
200 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
201 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
202 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
203 }
204
205 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
206 struct ext4_group_desc *bg)
207 {
208 return le32_to_cpu(bg->bg_inode_table_lo) |
209 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
210 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
211 }
212
213 __u32 ext4_free_group_clusters(struct super_block *sb,
214 struct ext4_group_desc *bg)
215 {
216 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
217 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
218 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
219 }
220
221 __u32 ext4_free_inodes_count(struct super_block *sb,
222 struct ext4_group_desc *bg)
223 {
224 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
225 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
226 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
227 }
228
229 __u32 ext4_used_dirs_count(struct super_block *sb,
230 struct ext4_group_desc *bg)
231 {
232 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
233 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
234 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
235 }
236
237 __u32 ext4_itable_unused_count(struct super_block *sb,
238 struct ext4_group_desc *bg)
239 {
240 return le16_to_cpu(bg->bg_itable_unused_lo) |
241 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
242 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
243 }
244
245 void ext4_block_bitmap_set(struct super_block *sb,
246 struct ext4_group_desc *bg, ext4_fsblk_t blk)
247 {
248 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
249 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
250 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
251 }
252
253 void ext4_inode_bitmap_set(struct super_block *sb,
254 struct ext4_group_desc *bg, ext4_fsblk_t blk)
255 {
256 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
257 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
258 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
259 }
260
261 void ext4_inode_table_set(struct super_block *sb,
262 struct ext4_group_desc *bg, ext4_fsblk_t blk)
263 {
264 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
265 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
266 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
267 }
268
269 void ext4_free_group_clusters_set(struct super_block *sb,
270 struct ext4_group_desc *bg, __u32 count)
271 {
272 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
273 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
274 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
275 }
276
277 void ext4_free_inodes_set(struct super_block *sb,
278 struct ext4_group_desc *bg, __u32 count)
279 {
280 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
281 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
282 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
283 }
284
285 void ext4_used_dirs_set(struct super_block *sb,
286 struct ext4_group_desc *bg, __u32 count)
287 {
288 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
289 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
290 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
291 }
292
293 void ext4_itable_unused_set(struct super_block *sb,
294 struct ext4_group_desc *bg, __u32 count)
295 {
296 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
297 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
298 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
299 }
300
301
302 static void __save_error_info(struct super_block *sb, const char *func,
303 unsigned int line)
304 {
305 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
306
307 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
308 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
309 es->s_last_error_time = cpu_to_le32(get_seconds());
310 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
311 es->s_last_error_line = cpu_to_le32(line);
312 if (!es->s_first_error_time) {
313 es->s_first_error_time = es->s_last_error_time;
314 strncpy(es->s_first_error_func, func,
315 sizeof(es->s_first_error_func));
316 es->s_first_error_line = cpu_to_le32(line);
317 es->s_first_error_ino = es->s_last_error_ino;
318 es->s_first_error_block = es->s_last_error_block;
319 }
320 /*
321 * Start the daily error reporting function if it hasn't been
322 * started already
323 */
324 if (!es->s_error_count)
325 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
326 le32_add_cpu(&es->s_error_count, 1);
327 }
328
329 static void save_error_info(struct super_block *sb, const char *func,
330 unsigned int line)
331 {
332 __save_error_info(sb, func, line);
333 ext4_commit_super(sb, 1);
334 }
335
336 /*
337 * The del_gendisk() function uninitializes the disk-specific data
338 * structures, including the bdi structure, without telling anyone
339 * else. Once this happens, any attempt to call mark_buffer_dirty()
340 * (for example, by ext4_commit_super), will cause a kernel OOPS.
341 * This is a kludge to prevent these oops until we can put in a proper
342 * hook in del_gendisk() to inform the VFS and file system layers.
343 */
344 static int block_device_ejected(struct super_block *sb)
345 {
346 struct inode *bd_inode = sb->s_bdev->bd_inode;
347 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
348
349 return bdi->dev == NULL;
350 }
351
352 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
353 {
354 struct super_block *sb = journal->j_private;
355 struct ext4_sb_info *sbi = EXT4_SB(sb);
356 int error = is_journal_aborted(journal);
357 struct ext4_journal_cb_entry *jce;
358
359 BUG_ON(txn->t_state == T_FINISHED);
360 spin_lock(&sbi->s_md_lock);
361 while (!list_empty(&txn->t_private_list)) {
362 jce = list_entry(txn->t_private_list.next,
363 struct ext4_journal_cb_entry, jce_list);
364 list_del_init(&jce->jce_list);
365 spin_unlock(&sbi->s_md_lock);
366 jce->jce_func(sb, jce, error);
367 spin_lock(&sbi->s_md_lock);
368 }
369 spin_unlock(&sbi->s_md_lock);
370 }
371
372 /* Deal with the reporting of failure conditions on a filesystem such as
373 * inconsistencies detected or read IO failures.
374 *
375 * On ext2, we can store the error state of the filesystem in the
376 * superblock. That is not possible on ext4, because we may have other
377 * write ordering constraints on the superblock which prevent us from
378 * writing it out straight away; and given that the journal is about to
379 * be aborted, we can't rely on the current, or future, transactions to
380 * write out the superblock safely.
381 *
382 * We'll just use the jbd2_journal_abort() error code to record an error in
383 * the journal instead. On recovery, the journal will complain about
384 * that error until we've noted it down and cleared it.
385 */
386
387 static void ext4_handle_error(struct super_block *sb)
388 {
389 if (sb->s_flags & MS_RDONLY)
390 return;
391
392 if (!test_opt(sb, ERRORS_CONT)) {
393 journal_t *journal = EXT4_SB(sb)->s_journal;
394
395 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
396 if (journal)
397 jbd2_journal_abort(journal, -EIO);
398 }
399 if (test_opt(sb, ERRORS_RO)) {
400 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
401 sb->s_flags |= MS_RDONLY;
402 }
403 if (test_opt(sb, ERRORS_PANIC)) {
404 if (EXT4_SB(sb)->s_journal &&
405 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
406 return;
407 panic("EXT4-fs (device %s): panic forced after error\n",
408 sb->s_id);
409 }
410 }
411
412 void __ext4_error(struct super_block *sb, const char *function,
413 unsigned int line, const char *fmt, ...)
414 {
415 struct va_format vaf;
416 va_list args;
417
418 va_start(args, fmt);
419 vaf.fmt = fmt;
420 vaf.va = &args;
421 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
422 sb->s_id, function, line, current->comm, &vaf);
423 va_end(args);
424 save_error_info(sb, function, line);
425
426 ext4_handle_error(sb);
427 }
428
429 void ext4_error_inode(struct inode *inode, const char *function,
430 unsigned int line, ext4_fsblk_t block,
431 const char *fmt, ...)
432 {
433 va_list args;
434 struct va_format vaf;
435 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
436
437 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
438 es->s_last_error_block = cpu_to_le64(block);
439 save_error_info(inode->i_sb, function, line);
440 va_start(args, fmt);
441 vaf.fmt = fmt;
442 vaf.va = &args;
443 if (block)
444 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
445 "inode #%lu: block %llu: comm %s: %pV\n",
446 inode->i_sb->s_id, function, line, inode->i_ino,
447 block, current->comm, &vaf);
448 else
449 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
450 "inode #%lu: comm %s: %pV\n",
451 inode->i_sb->s_id, function, line, inode->i_ino,
452 current->comm, &vaf);
453 va_end(args);
454
455 ext4_handle_error(inode->i_sb);
456 }
457
458 void ext4_error_file(struct file *file, const char *function,
459 unsigned int line, ext4_fsblk_t block,
460 const char *fmt, ...)
461 {
462 va_list args;
463 struct va_format vaf;
464 struct ext4_super_block *es;
465 struct inode *inode = file_inode(file);
466 char pathname[80], *path;
467
468 es = EXT4_SB(inode->i_sb)->s_es;
469 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
470 save_error_info(inode->i_sb, function, line);
471 path = d_path(&(file->f_path), pathname, sizeof(pathname));
472 if (IS_ERR(path))
473 path = "(unknown)";
474 va_start(args, fmt);
475 vaf.fmt = fmt;
476 vaf.va = &args;
477 if (block)
478 printk(KERN_CRIT
479 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
480 "block %llu: comm %s: path %s: %pV\n",
481 inode->i_sb->s_id, function, line, inode->i_ino,
482 block, current->comm, path, &vaf);
483 else
484 printk(KERN_CRIT
485 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
486 "comm %s: path %s: %pV\n",
487 inode->i_sb->s_id, function, line, inode->i_ino,
488 current->comm, path, &vaf);
489 va_end(args);
490
491 ext4_handle_error(inode->i_sb);
492 }
493
494 const char *ext4_decode_error(struct super_block *sb, int errno,
495 char nbuf[16])
496 {
497 char *errstr = NULL;
498
499 switch (errno) {
500 case -EIO:
501 errstr = "IO failure";
502 break;
503 case -ENOMEM:
504 errstr = "Out of memory";
505 break;
506 case -EROFS:
507 if (!sb || (EXT4_SB(sb)->s_journal &&
508 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
509 errstr = "Journal has aborted";
510 else
511 errstr = "Readonly filesystem";
512 break;
513 default:
514 /* If the caller passed in an extra buffer for unknown
515 * errors, textualise them now. Else we just return
516 * NULL. */
517 if (nbuf) {
518 /* Check for truncated error codes... */
519 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
520 errstr = nbuf;
521 }
522 break;
523 }
524
525 return errstr;
526 }
527
528 /* __ext4_std_error decodes expected errors from journaling functions
529 * automatically and invokes the appropriate error response. */
530
531 void __ext4_std_error(struct super_block *sb, const char *function,
532 unsigned int line, int errno)
533 {
534 char nbuf[16];
535 const char *errstr;
536
537 /* Special case: if the error is EROFS, and we're not already
538 * inside a transaction, then there's really no point in logging
539 * an error. */
540 if (errno == -EROFS && journal_current_handle() == NULL &&
541 (sb->s_flags & MS_RDONLY))
542 return;
543
544 errstr = ext4_decode_error(sb, errno, nbuf);
545 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
546 sb->s_id, function, line, errstr);
547 save_error_info(sb, function, line);
548
549 ext4_handle_error(sb);
550 }
551
552 /*
553 * ext4_abort is a much stronger failure handler than ext4_error. The
554 * abort function may be used to deal with unrecoverable failures such
555 * as journal IO errors or ENOMEM at a critical moment in log management.
556 *
557 * We unconditionally force the filesystem into an ABORT|READONLY state,
558 * unless the error response on the fs has been set to panic in which
559 * case we take the easy way out and panic immediately.
560 */
561
562 void __ext4_abort(struct super_block *sb, const char *function,
563 unsigned int line, const char *fmt, ...)
564 {
565 va_list args;
566
567 save_error_info(sb, function, line);
568 va_start(args, fmt);
569 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
570 function, line);
571 vprintk(fmt, args);
572 printk("\n");
573 va_end(args);
574
575 if ((sb->s_flags & MS_RDONLY) == 0) {
576 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
577 sb->s_flags |= MS_RDONLY;
578 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
579 if (EXT4_SB(sb)->s_journal)
580 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
581 save_error_info(sb, function, line);
582 }
583 if (test_opt(sb, ERRORS_PANIC)) {
584 if (EXT4_SB(sb)->s_journal &&
585 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
586 return;
587 panic("EXT4-fs panic from previous error\n");
588 }
589 }
590
591 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
592 {
593 struct va_format vaf;
594 va_list args;
595
596 va_start(args, fmt);
597 vaf.fmt = fmt;
598 vaf.va = &args;
599 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
600 va_end(args);
601 }
602
603 void __ext4_warning(struct super_block *sb, const char *function,
604 unsigned int line, const char *fmt, ...)
605 {
606 struct va_format vaf;
607 va_list args;
608
609 va_start(args, fmt);
610 vaf.fmt = fmt;
611 vaf.va = &args;
612 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
613 sb->s_id, function, line, &vaf);
614 va_end(args);
615 }
616
617 void __ext4_grp_locked_error(const char *function, unsigned int line,
618 struct super_block *sb, ext4_group_t grp,
619 unsigned long ino, ext4_fsblk_t block,
620 const char *fmt, ...)
621 __releases(bitlock)
622 __acquires(bitlock)
623 {
624 struct va_format vaf;
625 va_list args;
626 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
627
628 es->s_last_error_ino = cpu_to_le32(ino);
629 es->s_last_error_block = cpu_to_le64(block);
630 __save_error_info(sb, function, line);
631
632 va_start(args, fmt);
633
634 vaf.fmt = fmt;
635 vaf.va = &args;
636 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
637 sb->s_id, function, line, grp);
638 if (ino)
639 printk(KERN_CONT "inode %lu: ", ino);
640 if (block)
641 printk(KERN_CONT "block %llu:", (unsigned long long) block);
642 printk(KERN_CONT "%pV\n", &vaf);
643 va_end(args);
644
645 if (test_opt(sb, ERRORS_CONT)) {
646 ext4_commit_super(sb, 0);
647 return;
648 }
649
650 ext4_unlock_group(sb, grp);
651 ext4_handle_error(sb);
652 /*
653 * We only get here in the ERRORS_RO case; relocking the group
654 * may be dangerous, but nothing bad will happen since the
655 * filesystem will have already been marked read/only and the
656 * journal has been aborted. We return 1 as a hint to callers
657 * who might what to use the return value from
658 * ext4_grp_locked_error() to distinguish between the
659 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
660 * aggressively from the ext4 function in question, with a
661 * more appropriate error code.
662 */
663 ext4_lock_group(sb, grp);
664 return;
665 }
666
667 void ext4_update_dynamic_rev(struct super_block *sb)
668 {
669 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
670
671 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
672 return;
673
674 ext4_warning(sb,
675 "updating to rev %d because of new feature flag, "
676 "running e2fsck is recommended",
677 EXT4_DYNAMIC_REV);
678
679 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
680 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
681 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
682 /* leave es->s_feature_*compat flags alone */
683 /* es->s_uuid will be set by e2fsck if empty */
684
685 /*
686 * The rest of the superblock fields should be zero, and if not it
687 * means they are likely already in use, so leave them alone. We
688 * can leave it up to e2fsck to clean up any inconsistencies there.
689 */
690 }
691
692 /*
693 * Open the external journal device
694 */
695 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
696 {
697 struct block_device *bdev;
698 char b[BDEVNAME_SIZE];
699
700 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
701 if (IS_ERR(bdev))
702 goto fail;
703 return bdev;
704
705 fail:
706 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
707 __bdevname(dev, b), PTR_ERR(bdev));
708 return NULL;
709 }
710
711 /*
712 * Release the journal device
713 */
714 static void ext4_blkdev_put(struct block_device *bdev)
715 {
716 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
717 }
718
719 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
720 {
721 struct block_device *bdev;
722 bdev = sbi->journal_bdev;
723 if (bdev) {
724 ext4_blkdev_put(bdev);
725 sbi->journal_bdev = NULL;
726 }
727 }
728
729 static inline struct inode *orphan_list_entry(struct list_head *l)
730 {
731 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
732 }
733
734 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
735 {
736 struct list_head *l;
737
738 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
739 le32_to_cpu(sbi->s_es->s_last_orphan));
740
741 printk(KERN_ERR "sb_info orphan list:\n");
742 list_for_each(l, &sbi->s_orphan) {
743 struct inode *inode = orphan_list_entry(l);
744 printk(KERN_ERR " "
745 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
746 inode->i_sb->s_id, inode->i_ino, inode,
747 inode->i_mode, inode->i_nlink,
748 NEXT_ORPHAN(inode));
749 }
750 }
751
752 static void ext4_put_super(struct super_block *sb)
753 {
754 struct ext4_sb_info *sbi = EXT4_SB(sb);
755 struct ext4_super_block *es = sbi->s_es;
756 int aborted = 0;
757 int i, err;
758
759 ext4_unregister_li_request(sb);
760 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
761
762 flush_workqueue(sbi->dio_unwritten_wq);
763 destroy_workqueue(sbi->dio_unwritten_wq);
764
765 if (sbi->s_journal) {
766 aborted = is_journal_aborted(sbi->s_journal);
767 err = jbd2_journal_destroy(sbi->s_journal);
768 sbi->s_journal = NULL;
769 if ((err < 0) && !aborted)
770 ext4_abort(sb, "Couldn't clean up the journal");
771 }
772
773 ext4_es_unregister_shrinker(sb);
774 del_timer(&sbi->s_err_report);
775 ext4_release_system_zone(sb);
776 ext4_mb_release(sb);
777 ext4_ext_release(sb);
778 ext4_xattr_put_super(sb);
779
780 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
781 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
782 es->s_state = cpu_to_le16(sbi->s_mount_state);
783 }
784 if (!(sb->s_flags & MS_RDONLY))
785 ext4_commit_super(sb, 1);
786
787 if (sbi->s_proc) {
788 remove_proc_entry("options", sbi->s_proc);
789 remove_proc_entry(sb->s_id, ext4_proc_root);
790 }
791 kobject_del(&sbi->s_kobj);
792
793 for (i = 0; i < sbi->s_gdb_count; i++)
794 brelse(sbi->s_group_desc[i]);
795 ext4_kvfree(sbi->s_group_desc);
796 ext4_kvfree(sbi->s_flex_groups);
797 percpu_counter_destroy(&sbi->s_freeclusters_counter);
798 percpu_counter_destroy(&sbi->s_freeinodes_counter);
799 percpu_counter_destroy(&sbi->s_dirs_counter);
800 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
801 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
802 brelse(sbi->s_sbh);
803 #ifdef CONFIG_QUOTA
804 for (i = 0; i < MAXQUOTAS; i++)
805 kfree(sbi->s_qf_names[i]);
806 #endif
807
808 /* Debugging code just in case the in-memory inode orphan list
809 * isn't empty. The on-disk one can be non-empty if we've
810 * detected an error and taken the fs readonly, but the
811 * in-memory list had better be clean by this point. */
812 if (!list_empty(&sbi->s_orphan))
813 dump_orphan_list(sb, sbi);
814 J_ASSERT(list_empty(&sbi->s_orphan));
815
816 sync_blockdev(sb->s_bdev);
817 invalidate_bdev(sb->s_bdev);
818 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
819 /*
820 * Invalidate the journal device's buffers. We don't want them
821 * floating about in memory - the physical journal device may
822 * hotswapped, and it breaks the `ro-after' testing code.
823 */
824 sync_blockdev(sbi->journal_bdev);
825 invalidate_bdev(sbi->journal_bdev);
826 ext4_blkdev_remove(sbi);
827 }
828 if (sbi->s_mmp_tsk)
829 kthread_stop(sbi->s_mmp_tsk);
830 sb->s_fs_info = NULL;
831 /*
832 * Now that we are completely done shutting down the
833 * superblock, we need to actually destroy the kobject.
834 */
835 kobject_put(&sbi->s_kobj);
836 wait_for_completion(&sbi->s_kobj_unregister);
837 if (sbi->s_chksum_driver)
838 crypto_free_shash(sbi->s_chksum_driver);
839 kfree(sbi->s_blockgroup_lock);
840 kfree(sbi);
841 }
842
843 static struct kmem_cache *ext4_inode_cachep;
844
845 /*
846 * Called inside transaction, so use GFP_NOFS
847 */
848 static struct inode *ext4_alloc_inode(struct super_block *sb)
849 {
850 struct ext4_inode_info *ei;
851
852 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
853 if (!ei)
854 return NULL;
855
856 ei->vfs_inode.i_version = 1;
857 INIT_LIST_HEAD(&ei->i_prealloc_list);
858 spin_lock_init(&ei->i_prealloc_lock);
859 ext4_es_init_tree(&ei->i_es_tree);
860 rwlock_init(&ei->i_es_lock);
861 INIT_LIST_HEAD(&ei->i_es_lru);
862 ei->i_es_lru_nr = 0;
863 ei->i_reserved_data_blocks = 0;
864 ei->i_reserved_meta_blocks = 0;
865 ei->i_allocated_meta_blocks = 0;
866 ei->i_da_metadata_calc_len = 0;
867 ei->i_da_metadata_calc_last_lblock = 0;
868 spin_lock_init(&(ei->i_block_reservation_lock));
869 #ifdef CONFIG_QUOTA
870 ei->i_reserved_quota = 0;
871 #endif
872 ei->jinode = NULL;
873 INIT_LIST_HEAD(&ei->i_completed_io_list);
874 spin_lock_init(&ei->i_completed_io_lock);
875 ei->i_sync_tid = 0;
876 ei->i_datasync_tid = 0;
877 atomic_set(&ei->i_ioend_count, 0);
878 atomic_set(&ei->i_unwritten, 0);
879 INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
880
881 return &ei->vfs_inode;
882 }
883
884 static int ext4_drop_inode(struct inode *inode)
885 {
886 int drop = generic_drop_inode(inode);
887
888 trace_ext4_drop_inode(inode, drop);
889 return drop;
890 }
891
892 static void ext4_i_callback(struct rcu_head *head)
893 {
894 struct inode *inode = container_of(head, struct inode, i_rcu);
895 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
896 }
897
898 static void ext4_destroy_inode(struct inode *inode)
899 {
900 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
901 ext4_msg(inode->i_sb, KERN_ERR,
902 "Inode %lu (%p): orphan list check failed!",
903 inode->i_ino, EXT4_I(inode));
904 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
905 EXT4_I(inode), sizeof(struct ext4_inode_info),
906 true);
907 dump_stack();
908 }
909 call_rcu(&inode->i_rcu, ext4_i_callback);
910 }
911
912 static void init_once(void *foo)
913 {
914 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
915
916 INIT_LIST_HEAD(&ei->i_orphan);
917 init_rwsem(&ei->xattr_sem);
918 init_rwsem(&ei->i_data_sem);
919 inode_init_once(&ei->vfs_inode);
920 }
921
922 static int init_inodecache(void)
923 {
924 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
925 sizeof(struct ext4_inode_info),
926 0, (SLAB_RECLAIM_ACCOUNT|
927 SLAB_MEM_SPREAD),
928 init_once);
929 if (ext4_inode_cachep == NULL)
930 return -ENOMEM;
931 return 0;
932 }
933
934 static void destroy_inodecache(void)
935 {
936 /*
937 * Make sure all delayed rcu free inodes are flushed before we
938 * destroy cache.
939 */
940 rcu_barrier();
941 kmem_cache_destroy(ext4_inode_cachep);
942 }
943
944 void ext4_clear_inode(struct inode *inode)
945 {
946 invalidate_inode_buffers(inode);
947 clear_inode(inode);
948 dquot_drop(inode);
949 ext4_discard_preallocations(inode);
950 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
951 ext4_es_lru_del(inode);
952 if (EXT4_I(inode)->jinode) {
953 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
954 EXT4_I(inode)->jinode);
955 jbd2_free_inode(EXT4_I(inode)->jinode);
956 EXT4_I(inode)->jinode = NULL;
957 }
958 }
959
960 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
961 u64 ino, u32 generation)
962 {
963 struct inode *inode;
964
965 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
966 return ERR_PTR(-ESTALE);
967 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
968 return ERR_PTR(-ESTALE);
969
970 /* iget isn't really right if the inode is currently unallocated!!
971 *
972 * ext4_read_inode will return a bad_inode if the inode had been
973 * deleted, so we should be safe.
974 *
975 * Currently we don't know the generation for parent directory, so
976 * a generation of 0 means "accept any"
977 */
978 inode = ext4_iget_normal(sb, ino);
979 if (IS_ERR(inode))
980 return ERR_CAST(inode);
981 if (generation && inode->i_generation != generation) {
982 iput(inode);
983 return ERR_PTR(-ESTALE);
984 }
985
986 return inode;
987 }
988
989 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
990 int fh_len, int fh_type)
991 {
992 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
993 ext4_nfs_get_inode);
994 }
995
996 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
997 int fh_len, int fh_type)
998 {
999 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1000 ext4_nfs_get_inode);
1001 }
1002
1003 /*
1004 * Try to release metadata pages (indirect blocks, directories) which are
1005 * mapped via the block device. Since these pages could have journal heads
1006 * which would prevent try_to_free_buffers() from freeing them, we must use
1007 * jbd2 layer's try_to_free_buffers() function to release them.
1008 */
1009 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1010 gfp_t wait)
1011 {
1012 journal_t *journal = EXT4_SB(sb)->s_journal;
1013
1014 WARN_ON(PageChecked(page));
1015 if (!page_has_buffers(page))
1016 return 0;
1017 if (journal)
1018 return jbd2_journal_try_to_free_buffers(journal, page,
1019 wait & ~__GFP_WAIT);
1020 return try_to_free_buffers(page);
1021 }
1022
1023 #ifdef CONFIG_QUOTA
1024 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1025 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1026
1027 static int ext4_write_dquot(struct dquot *dquot);
1028 static int ext4_acquire_dquot(struct dquot *dquot);
1029 static int ext4_release_dquot(struct dquot *dquot);
1030 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1031 static int ext4_write_info(struct super_block *sb, int type);
1032 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1033 struct path *path);
1034 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1035 int format_id);
1036 static int ext4_quota_off(struct super_block *sb, int type);
1037 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1038 static int ext4_quota_on_mount(struct super_block *sb, int type);
1039 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1040 size_t len, loff_t off);
1041 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1042 const char *data, size_t len, loff_t off);
1043 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1044 unsigned int flags);
1045 static int ext4_enable_quotas(struct super_block *sb);
1046
1047 static const struct dquot_operations ext4_quota_operations = {
1048 .get_reserved_space = ext4_get_reserved_space,
1049 .write_dquot = ext4_write_dquot,
1050 .acquire_dquot = ext4_acquire_dquot,
1051 .release_dquot = ext4_release_dquot,
1052 .mark_dirty = ext4_mark_dquot_dirty,
1053 .write_info = ext4_write_info,
1054 .alloc_dquot = dquot_alloc,
1055 .destroy_dquot = dquot_destroy,
1056 };
1057
1058 static const struct quotactl_ops ext4_qctl_operations = {
1059 .quota_on = ext4_quota_on,
1060 .quota_off = ext4_quota_off,
1061 .quota_sync = dquot_quota_sync,
1062 .get_info = dquot_get_dqinfo,
1063 .set_info = dquot_set_dqinfo,
1064 .get_dqblk = dquot_get_dqblk,
1065 .set_dqblk = dquot_set_dqblk
1066 };
1067
1068 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1069 .quota_on_meta = ext4_quota_on_sysfile,
1070 .quota_off = ext4_quota_off_sysfile,
1071 .quota_sync = dquot_quota_sync,
1072 .get_info = dquot_get_dqinfo,
1073 .set_info = dquot_set_dqinfo,
1074 .get_dqblk = dquot_get_dqblk,
1075 .set_dqblk = dquot_set_dqblk
1076 };
1077 #endif
1078
1079 static const struct super_operations ext4_sops = {
1080 .alloc_inode = ext4_alloc_inode,
1081 .destroy_inode = ext4_destroy_inode,
1082 .write_inode = ext4_write_inode,
1083 .dirty_inode = ext4_dirty_inode,
1084 .drop_inode = ext4_drop_inode,
1085 .evict_inode = ext4_evict_inode,
1086 .put_super = ext4_put_super,
1087 .sync_fs = ext4_sync_fs,
1088 .freeze_fs = ext4_freeze,
1089 .unfreeze_fs = ext4_unfreeze,
1090 .statfs = ext4_statfs,
1091 .remount_fs = ext4_remount,
1092 .show_options = ext4_show_options,
1093 #ifdef CONFIG_QUOTA
1094 .quota_read = ext4_quota_read,
1095 .quota_write = ext4_quota_write,
1096 #endif
1097 .bdev_try_to_free_page = bdev_try_to_free_page,
1098 };
1099
1100 static const struct super_operations ext4_nojournal_sops = {
1101 .alloc_inode = ext4_alloc_inode,
1102 .destroy_inode = ext4_destroy_inode,
1103 .write_inode = ext4_write_inode,
1104 .dirty_inode = ext4_dirty_inode,
1105 .drop_inode = ext4_drop_inode,
1106 .evict_inode = ext4_evict_inode,
1107 .put_super = ext4_put_super,
1108 .statfs = ext4_statfs,
1109 .remount_fs = ext4_remount,
1110 .show_options = ext4_show_options,
1111 #ifdef CONFIG_QUOTA
1112 .quota_read = ext4_quota_read,
1113 .quota_write = ext4_quota_write,
1114 #endif
1115 .bdev_try_to_free_page = bdev_try_to_free_page,
1116 };
1117
1118 static const struct export_operations ext4_export_ops = {
1119 .fh_to_dentry = ext4_fh_to_dentry,
1120 .fh_to_parent = ext4_fh_to_parent,
1121 .get_parent = ext4_get_parent,
1122 };
1123
1124 enum {
1125 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1126 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1127 Opt_nouid32, Opt_debug, Opt_removed,
1128 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1129 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1130 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1131 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1132 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1133 Opt_data_err_abort, Opt_data_err_ignore,
1134 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1135 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1136 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1137 Opt_usrquota, Opt_grpquota, Opt_i_version,
1138 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1139 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1140 Opt_inode_readahead_blks, Opt_journal_ioprio,
1141 Opt_dioread_nolock, Opt_dioread_lock,
1142 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1143 Opt_max_dir_size_kb,
1144 };
1145
1146 static const match_table_t tokens = {
1147 {Opt_bsd_df, "bsddf"},
1148 {Opt_minix_df, "minixdf"},
1149 {Opt_grpid, "grpid"},
1150 {Opt_grpid, "bsdgroups"},
1151 {Opt_nogrpid, "nogrpid"},
1152 {Opt_nogrpid, "sysvgroups"},
1153 {Opt_resgid, "resgid=%u"},
1154 {Opt_resuid, "resuid=%u"},
1155 {Opt_sb, "sb=%u"},
1156 {Opt_err_cont, "errors=continue"},
1157 {Opt_err_panic, "errors=panic"},
1158 {Opt_err_ro, "errors=remount-ro"},
1159 {Opt_nouid32, "nouid32"},
1160 {Opt_debug, "debug"},
1161 {Opt_removed, "oldalloc"},
1162 {Opt_removed, "orlov"},
1163 {Opt_user_xattr, "user_xattr"},
1164 {Opt_nouser_xattr, "nouser_xattr"},
1165 {Opt_acl, "acl"},
1166 {Opt_noacl, "noacl"},
1167 {Opt_noload, "norecovery"},
1168 {Opt_noload, "noload"},
1169 {Opt_removed, "nobh"},
1170 {Opt_removed, "bh"},
1171 {Opt_commit, "commit=%u"},
1172 {Opt_min_batch_time, "min_batch_time=%u"},
1173 {Opt_max_batch_time, "max_batch_time=%u"},
1174 {Opt_journal_dev, "journal_dev=%u"},
1175 {Opt_journal_checksum, "journal_checksum"},
1176 {Opt_journal_async_commit, "journal_async_commit"},
1177 {Opt_abort, "abort"},
1178 {Opt_data_journal, "data=journal"},
1179 {Opt_data_ordered, "data=ordered"},
1180 {Opt_data_writeback, "data=writeback"},
1181 {Opt_data_err_abort, "data_err=abort"},
1182 {Opt_data_err_ignore, "data_err=ignore"},
1183 {Opt_offusrjquota, "usrjquota="},
1184 {Opt_usrjquota, "usrjquota=%s"},
1185 {Opt_offgrpjquota, "grpjquota="},
1186 {Opt_grpjquota, "grpjquota=%s"},
1187 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1188 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1189 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1190 {Opt_grpquota, "grpquota"},
1191 {Opt_noquota, "noquota"},
1192 {Opt_quota, "quota"},
1193 {Opt_usrquota, "usrquota"},
1194 {Opt_barrier, "barrier=%u"},
1195 {Opt_barrier, "barrier"},
1196 {Opt_nobarrier, "nobarrier"},
1197 {Opt_i_version, "i_version"},
1198 {Opt_stripe, "stripe=%u"},
1199 {Opt_delalloc, "delalloc"},
1200 {Opt_nodelalloc, "nodelalloc"},
1201 {Opt_removed, "mblk_io_submit"},
1202 {Opt_removed, "nomblk_io_submit"},
1203 {Opt_block_validity, "block_validity"},
1204 {Opt_noblock_validity, "noblock_validity"},
1205 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1206 {Opt_journal_ioprio, "journal_ioprio=%u"},
1207 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1208 {Opt_auto_da_alloc, "auto_da_alloc"},
1209 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1210 {Opt_dioread_nolock, "dioread_nolock"},
1211 {Opt_dioread_lock, "dioread_lock"},
1212 {Opt_discard, "discard"},
1213 {Opt_nodiscard, "nodiscard"},
1214 {Opt_init_itable, "init_itable=%u"},
1215 {Opt_init_itable, "init_itable"},
1216 {Opt_noinit_itable, "noinit_itable"},
1217 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1218 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1219 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1220 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1221 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1222 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1223 {Opt_err, NULL},
1224 };
1225
1226 static ext4_fsblk_t get_sb_block(void **data)
1227 {
1228 ext4_fsblk_t sb_block;
1229 char *options = (char *) *data;
1230
1231 if (!options || strncmp(options, "sb=", 3) != 0)
1232 return 1; /* Default location */
1233
1234 options += 3;
1235 /* TODO: use simple_strtoll with >32bit ext4 */
1236 sb_block = simple_strtoul(options, &options, 0);
1237 if (*options && *options != ',') {
1238 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1239 (char *) *data);
1240 return 1;
1241 }
1242 if (*options == ',')
1243 options++;
1244 *data = (void *) options;
1245
1246 return sb_block;
1247 }
1248
1249 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1250 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1251 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1252
1253 #ifdef CONFIG_QUOTA
1254 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1255 {
1256 struct ext4_sb_info *sbi = EXT4_SB(sb);
1257 char *qname;
1258 int ret = -1;
1259
1260 if (sb_any_quota_loaded(sb) &&
1261 !sbi->s_qf_names[qtype]) {
1262 ext4_msg(sb, KERN_ERR,
1263 "Cannot change journaled "
1264 "quota options when quota turned on");
1265 return -1;
1266 }
1267 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1268 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1269 "when QUOTA feature is enabled");
1270 return -1;
1271 }
1272 qname = match_strdup(args);
1273 if (!qname) {
1274 ext4_msg(sb, KERN_ERR,
1275 "Not enough memory for storing quotafile name");
1276 return -1;
1277 }
1278 if (sbi->s_qf_names[qtype]) {
1279 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1280 ret = 1;
1281 else
1282 ext4_msg(sb, KERN_ERR,
1283 "%s quota file already specified",
1284 QTYPE2NAME(qtype));
1285 goto errout;
1286 }
1287 if (strchr(qname, '/')) {
1288 ext4_msg(sb, KERN_ERR,
1289 "quotafile must be on filesystem root");
1290 goto errout;
1291 }
1292 sbi->s_qf_names[qtype] = qname;
1293 set_opt(sb, QUOTA);
1294 return 1;
1295 errout:
1296 kfree(qname);
1297 return ret;
1298 }
1299
1300 static int clear_qf_name(struct super_block *sb, int qtype)
1301 {
1302
1303 struct ext4_sb_info *sbi = EXT4_SB(sb);
1304
1305 if (sb_any_quota_loaded(sb) &&
1306 sbi->s_qf_names[qtype]) {
1307 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1308 " when quota turned on");
1309 return -1;
1310 }
1311 kfree(sbi->s_qf_names[qtype]);
1312 sbi->s_qf_names[qtype] = NULL;
1313 return 1;
1314 }
1315 #endif
1316
1317 #define MOPT_SET 0x0001
1318 #define MOPT_CLEAR 0x0002
1319 #define MOPT_NOSUPPORT 0x0004
1320 #define MOPT_EXPLICIT 0x0008
1321 #define MOPT_CLEAR_ERR 0x0010
1322 #define MOPT_GTE0 0x0020
1323 #ifdef CONFIG_QUOTA
1324 #define MOPT_Q 0
1325 #define MOPT_QFMT 0x0040
1326 #else
1327 #define MOPT_Q MOPT_NOSUPPORT
1328 #define MOPT_QFMT MOPT_NOSUPPORT
1329 #endif
1330 #define MOPT_DATAJ 0x0080
1331 #define MOPT_NO_EXT2 0x0100
1332 #define MOPT_NO_EXT3 0x0200
1333 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1334
1335 static const struct mount_opts {
1336 int token;
1337 int mount_opt;
1338 int flags;
1339 } ext4_mount_opts[] = {
1340 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1341 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1342 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1343 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1344 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1345 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1346 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1347 MOPT_EXT4_ONLY | MOPT_SET},
1348 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1349 MOPT_EXT4_ONLY | MOPT_CLEAR},
1350 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1351 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1352 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1353 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1354 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1355 MOPT_EXT4_ONLY | MOPT_CLEAR},
1356 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1357 MOPT_EXT4_ONLY | MOPT_SET},
1358 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1359 EXT4_MOUNT_JOURNAL_CHECKSUM),
1360 MOPT_EXT4_ONLY | MOPT_SET},
1361 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1362 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1363 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1364 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1365 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1366 MOPT_NO_EXT2 | MOPT_SET},
1367 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1368 MOPT_NO_EXT2 | MOPT_CLEAR},
1369 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1370 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1371 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1372 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1373 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1374 {Opt_commit, 0, MOPT_GTE0},
1375 {Opt_max_batch_time, 0, MOPT_GTE0},
1376 {Opt_min_batch_time, 0, MOPT_GTE0},
1377 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1378 {Opt_init_itable, 0, MOPT_GTE0},
1379 {Opt_stripe, 0, MOPT_GTE0},
1380 {Opt_resuid, 0, MOPT_GTE0},
1381 {Opt_resgid, 0, MOPT_GTE0},
1382 {Opt_journal_dev, 0, MOPT_GTE0},
1383 {Opt_journal_ioprio, 0, MOPT_GTE0},
1384 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1385 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1386 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1387 MOPT_NO_EXT2 | MOPT_DATAJ},
1388 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1389 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1390 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1391 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1392 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1393 #else
1394 {Opt_acl, 0, MOPT_NOSUPPORT},
1395 {Opt_noacl, 0, MOPT_NOSUPPORT},
1396 #endif
1397 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1398 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1399 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1400 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1401 MOPT_SET | MOPT_Q},
1402 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1403 MOPT_SET | MOPT_Q},
1404 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1405 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1406 {Opt_usrjquota, 0, MOPT_Q},
1407 {Opt_grpjquota, 0, MOPT_Q},
1408 {Opt_offusrjquota, 0, MOPT_Q},
1409 {Opt_offgrpjquota, 0, MOPT_Q},
1410 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1411 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1412 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1413 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1414 {Opt_err, 0, 0}
1415 };
1416
1417 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1418 substring_t *args, unsigned long *journal_devnum,
1419 unsigned int *journal_ioprio, int is_remount)
1420 {
1421 struct ext4_sb_info *sbi = EXT4_SB(sb);
1422 const struct mount_opts *m;
1423 kuid_t uid;
1424 kgid_t gid;
1425 int arg = 0;
1426
1427 #ifdef CONFIG_QUOTA
1428 if (token == Opt_usrjquota)
1429 return set_qf_name(sb, USRQUOTA, &args[0]);
1430 else if (token == Opt_grpjquota)
1431 return set_qf_name(sb, GRPQUOTA, &args[0]);
1432 else if (token == Opt_offusrjquota)
1433 return clear_qf_name(sb, USRQUOTA);
1434 else if (token == Opt_offgrpjquota)
1435 return clear_qf_name(sb, GRPQUOTA);
1436 #endif
1437 switch (token) {
1438 case Opt_noacl:
1439 case Opt_nouser_xattr:
1440 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1441 break;
1442 case Opt_sb:
1443 return 1; /* handled by get_sb_block() */
1444 case Opt_removed:
1445 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1446 return 1;
1447 case Opt_abort:
1448 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1449 return 1;
1450 case Opt_i_version:
1451 sb->s_flags |= MS_I_VERSION;
1452 return 1;
1453 }
1454
1455 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1456 if (token == m->token)
1457 break;
1458
1459 if (m->token == Opt_err) {
1460 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1461 "or missing value", opt);
1462 return -1;
1463 }
1464
1465 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1466 ext4_msg(sb, KERN_ERR,
1467 "Mount option \"%s\" incompatible with ext2", opt);
1468 return -1;
1469 }
1470 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1471 ext4_msg(sb, KERN_ERR,
1472 "Mount option \"%s\" incompatible with ext3", opt);
1473 return -1;
1474 }
1475
1476 if (args->from && match_int(args, &arg))
1477 return -1;
1478 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1479 return -1;
1480 if (m->flags & MOPT_EXPLICIT)
1481 set_opt2(sb, EXPLICIT_DELALLOC);
1482 if (m->flags & MOPT_CLEAR_ERR)
1483 clear_opt(sb, ERRORS_MASK);
1484 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1485 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1486 "options when quota turned on");
1487 return -1;
1488 }
1489
1490 if (m->flags & MOPT_NOSUPPORT) {
1491 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1492 } else if (token == Opt_commit) {
1493 if (arg == 0)
1494 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1495 sbi->s_commit_interval = HZ * arg;
1496 } else if (token == Opt_max_batch_time) {
1497 sbi->s_max_batch_time = arg;
1498 } else if (token == Opt_min_batch_time) {
1499 sbi->s_min_batch_time = arg;
1500 } else if (token == Opt_inode_readahead_blks) {
1501 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1502 ext4_msg(sb, KERN_ERR,
1503 "EXT4-fs: inode_readahead_blks must be "
1504 "0 or a power of 2 smaller than 2^31");
1505 return -1;
1506 }
1507 sbi->s_inode_readahead_blks = arg;
1508 } else if (token == Opt_init_itable) {
1509 set_opt(sb, INIT_INODE_TABLE);
1510 if (!args->from)
1511 arg = EXT4_DEF_LI_WAIT_MULT;
1512 sbi->s_li_wait_mult = arg;
1513 } else if (token == Opt_max_dir_size_kb) {
1514 sbi->s_max_dir_size_kb = arg;
1515 } else if (token == Opt_stripe) {
1516 sbi->s_stripe = arg;
1517 } else if (token == Opt_resuid) {
1518 uid = make_kuid(current_user_ns(), arg);
1519 if (!uid_valid(uid)) {
1520 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1521 return -1;
1522 }
1523 sbi->s_resuid = uid;
1524 } else if (token == Opt_resgid) {
1525 gid = make_kgid(current_user_ns(), arg);
1526 if (!gid_valid(gid)) {
1527 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1528 return -1;
1529 }
1530 sbi->s_resgid = gid;
1531 } else if (token == Opt_journal_dev) {
1532 if (is_remount) {
1533 ext4_msg(sb, KERN_ERR,
1534 "Cannot specify journal on remount");
1535 return -1;
1536 }
1537 *journal_devnum = arg;
1538 } else if (token == Opt_journal_ioprio) {
1539 if (arg > 7) {
1540 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1541 " (must be 0-7)");
1542 return -1;
1543 }
1544 *journal_ioprio =
1545 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1546 } else if (m->flags & MOPT_DATAJ) {
1547 if (is_remount) {
1548 if (!sbi->s_journal)
1549 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1550 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1551 ext4_msg(sb, KERN_ERR,
1552 "Cannot change data mode on remount");
1553 return -1;
1554 }
1555 } else {
1556 clear_opt(sb, DATA_FLAGS);
1557 sbi->s_mount_opt |= m->mount_opt;
1558 }
1559 #ifdef CONFIG_QUOTA
1560 } else if (m->flags & MOPT_QFMT) {
1561 if (sb_any_quota_loaded(sb) &&
1562 sbi->s_jquota_fmt != m->mount_opt) {
1563 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1564 "quota options when quota turned on");
1565 return -1;
1566 }
1567 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1568 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1569 ext4_msg(sb, KERN_ERR,
1570 "Cannot set journaled quota options "
1571 "when QUOTA feature is enabled");
1572 return -1;
1573 }
1574 sbi->s_jquota_fmt = m->mount_opt;
1575 #endif
1576 } else {
1577 if (!args->from)
1578 arg = 1;
1579 if (m->flags & MOPT_CLEAR)
1580 arg = !arg;
1581 else if (unlikely(!(m->flags & MOPT_SET))) {
1582 ext4_msg(sb, KERN_WARNING,
1583 "buggy handling of option %s", opt);
1584 WARN_ON(1);
1585 return -1;
1586 }
1587 if (arg != 0)
1588 sbi->s_mount_opt |= m->mount_opt;
1589 else
1590 sbi->s_mount_opt &= ~m->mount_opt;
1591 }
1592 return 1;
1593 }
1594
1595 static int parse_options(char *options, struct super_block *sb,
1596 unsigned long *journal_devnum,
1597 unsigned int *journal_ioprio,
1598 int is_remount)
1599 {
1600 struct ext4_sb_info *sbi = EXT4_SB(sb);
1601 char *p;
1602 substring_t args[MAX_OPT_ARGS];
1603 int token;
1604
1605 if (!options)
1606 return 1;
1607
1608 while ((p = strsep(&options, ",")) != NULL) {
1609 if (!*p)
1610 continue;
1611 /*
1612 * Initialize args struct so we know whether arg was
1613 * found; some options take optional arguments.
1614 */
1615 args[0].to = args[0].from = NULL;
1616 token = match_token(p, tokens, args);
1617 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1618 journal_ioprio, is_remount) < 0)
1619 return 0;
1620 }
1621 #ifdef CONFIG_QUOTA
1622 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1623 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1624 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1625 "feature is enabled");
1626 return 0;
1627 }
1628 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1629 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1630 clear_opt(sb, USRQUOTA);
1631
1632 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1633 clear_opt(sb, GRPQUOTA);
1634
1635 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1636 ext4_msg(sb, KERN_ERR, "old and new quota "
1637 "format mixing");
1638 return 0;
1639 }
1640
1641 if (!sbi->s_jquota_fmt) {
1642 ext4_msg(sb, KERN_ERR, "journaled quota format "
1643 "not specified");
1644 return 0;
1645 }
1646 }
1647 #endif
1648 if (test_opt(sb, DIOREAD_NOLOCK)) {
1649 int blocksize =
1650 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1651
1652 if (blocksize < PAGE_CACHE_SIZE) {
1653 ext4_msg(sb, KERN_ERR, "can't mount with "
1654 "dioread_nolock if block size != PAGE_SIZE");
1655 return 0;
1656 }
1657 }
1658 return 1;
1659 }
1660
1661 static inline void ext4_show_quota_options(struct seq_file *seq,
1662 struct super_block *sb)
1663 {
1664 #if defined(CONFIG_QUOTA)
1665 struct ext4_sb_info *sbi = EXT4_SB(sb);
1666
1667 if (sbi->s_jquota_fmt) {
1668 char *fmtname = "";
1669
1670 switch (sbi->s_jquota_fmt) {
1671 case QFMT_VFS_OLD:
1672 fmtname = "vfsold";
1673 break;
1674 case QFMT_VFS_V0:
1675 fmtname = "vfsv0";
1676 break;
1677 case QFMT_VFS_V1:
1678 fmtname = "vfsv1";
1679 break;
1680 }
1681 seq_printf(seq, ",jqfmt=%s", fmtname);
1682 }
1683
1684 if (sbi->s_qf_names[USRQUOTA])
1685 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1686
1687 if (sbi->s_qf_names[GRPQUOTA])
1688 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1689 #endif
1690 }
1691
1692 static const char *token2str(int token)
1693 {
1694 const struct match_token *t;
1695
1696 for (t = tokens; t->token != Opt_err; t++)
1697 if (t->token == token && !strchr(t->pattern, '='))
1698 break;
1699 return t->pattern;
1700 }
1701
1702 /*
1703 * Show an option if
1704 * - it's set to a non-default value OR
1705 * - if the per-sb default is different from the global default
1706 */
1707 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1708 int nodefs)
1709 {
1710 struct ext4_sb_info *sbi = EXT4_SB(sb);
1711 struct ext4_super_block *es = sbi->s_es;
1712 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1713 const struct mount_opts *m;
1714 char sep = nodefs ? '\n' : ',';
1715
1716 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1717 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1718
1719 if (sbi->s_sb_block != 1)
1720 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1721
1722 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1723 int want_set = m->flags & MOPT_SET;
1724 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1725 (m->flags & MOPT_CLEAR_ERR))
1726 continue;
1727 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1728 continue; /* skip if same as the default */
1729 if ((want_set &&
1730 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1731 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1732 continue; /* select Opt_noFoo vs Opt_Foo */
1733 SEQ_OPTS_PRINT("%s", token2str(m->token));
1734 }
1735
1736 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1737 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1738 SEQ_OPTS_PRINT("resuid=%u",
1739 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1740 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1741 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1742 SEQ_OPTS_PRINT("resgid=%u",
1743 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1744 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1745 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1746 SEQ_OPTS_PUTS("errors=remount-ro");
1747 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1748 SEQ_OPTS_PUTS("errors=continue");
1749 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1750 SEQ_OPTS_PUTS("errors=panic");
1751 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1752 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1753 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1754 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1755 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1756 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1757 if (sb->s_flags & MS_I_VERSION)
1758 SEQ_OPTS_PUTS("i_version");
1759 if (nodefs || sbi->s_stripe)
1760 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1761 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1762 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1763 SEQ_OPTS_PUTS("data=journal");
1764 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1765 SEQ_OPTS_PUTS("data=ordered");
1766 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1767 SEQ_OPTS_PUTS("data=writeback");
1768 }
1769 if (nodefs ||
1770 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1771 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1772 sbi->s_inode_readahead_blks);
1773
1774 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1775 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1776 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1777 if (nodefs || sbi->s_max_dir_size_kb)
1778 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1779
1780 ext4_show_quota_options(seq, sb);
1781 return 0;
1782 }
1783
1784 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1785 {
1786 return _ext4_show_options(seq, root->d_sb, 0);
1787 }
1788
1789 static int options_seq_show(struct seq_file *seq, void *offset)
1790 {
1791 struct super_block *sb = seq->private;
1792 int rc;
1793
1794 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1795 rc = _ext4_show_options(seq, sb, 1);
1796 seq_puts(seq, "\n");
1797 return rc;
1798 }
1799
1800 static int options_open_fs(struct inode *inode, struct file *file)
1801 {
1802 return single_open(file, options_seq_show, PDE_DATA(inode));
1803 }
1804
1805 static const struct file_operations ext4_seq_options_fops = {
1806 .owner = THIS_MODULE,
1807 .open = options_open_fs,
1808 .read = seq_read,
1809 .llseek = seq_lseek,
1810 .release = single_release,
1811 };
1812
1813 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1814 int read_only)
1815 {
1816 struct ext4_sb_info *sbi = EXT4_SB(sb);
1817 int res = 0;
1818
1819 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1820 ext4_msg(sb, KERN_ERR, "revision level too high, "
1821 "forcing read-only mode");
1822 res = MS_RDONLY;
1823 }
1824 if (read_only)
1825 goto done;
1826 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1827 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1828 "running e2fsck is recommended");
1829 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1830 ext4_msg(sb, KERN_WARNING,
1831 "warning: mounting fs with errors, "
1832 "running e2fsck is recommended");
1833 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1834 le16_to_cpu(es->s_mnt_count) >=
1835 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1836 ext4_msg(sb, KERN_WARNING,
1837 "warning: maximal mount count reached, "
1838 "running e2fsck is recommended");
1839 else if (le32_to_cpu(es->s_checkinterval) &&
1840 (le32_to_cpu(es->s_lastcheck) +
1841 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1842 ext4_msg(sb, KERN_WARNING,
1843 "warning: checktime reached, "
1844 "running e2fsck is recommended");
1845 if (!sbi->s_journal)
1846 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1847 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1848 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1849 le16_add_cpu(&es->s_mnt_count, 1);
1850 es->s_mtime = cpu_to_le32(get_seconds());
1851 ext4_update_dynamic_rev(sb);
1852 if (sbi->s_journal)
1853 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1854
1855 ext4_commit_super(sb, 1);
1856 done:
1857 if (test_opt(sb, DEBUG))
1858 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1859 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1860 sb->s_blocksize,
1861 sbi->s_groups_count,
1862 EXT4_BLOCKS_PER_GROUP(sb),
1863 EXT4_INODES_PER_GROUP(sb),
1864 sbi->s_mount_opt, sbi->s_mount_opt2);
1865
1866 cleancache_init_fs(sb);
1867 return res;
1868 }
1869
1870 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1871 {
1872 struct ext4_sb_info *sbi = EXT4_SB(sb);
1873 struct flex_groups *new_groups;
1874 int size;
1875
1876 if (!sbi->s_log_groups_per_flex)
1877 return 0;
1878
1879 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1880 if (size <= sbi->s_flex_groups_allocated)
1881 return 0;
1882
1883 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1884 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1885 if (!new_groups) {
1886 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1887 size / (int) sizeof(struct flex_groups));
1888 return -ENOMEM;
1889 }
1890
1891 if (sbi->s_flex_groups) {
1892 memcpy(new_groups, sbi->s_flex_groups,
1893 (sbi->s_flex_groups_allocated *
1894 sizeof(struct flex_groups)));
1895 ext4_kvfree(sbi->s_flex_groups);
1896 }
1897 sbi->s_flex_groups = new_groups;
1898 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1899 return 0;
1900 }
1901
1902 static int ext4_fill_flex_info(struct super_block *sb)
1903 {
1904 struct ext4_sb_info *sbi = EXT4_SB(sb);
1905 struct ext4_group_desc *gdp = NULL;
1906 ext4_group_t flex_group;
1907 unsigned int groups_per_flex = 0;
1908 int i, err;
1909
1910 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1911 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1912 sbi->s_log_groups_per_flex = 0;
1913 return 1;
1914 }
1915 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1916
1917 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1918 if (err)
1919 goto failed;
1920
1921 for (i = 0; i < sbi->s_groups_count; i++) {
1922 gdp = ext4_get_group_desc(sb, i, NULL);
1923
1924 flex_group = ext4_flex_group(sbi, i);
1925 atomic_add(ext4_free_inodes_count(sb, gdp),
1926 &sbi->s_flex_groups[flex_group].free_inodes);
1927 atomic64_add(ext4_free_group_clusters(sb, gdp),
1928 &sbi->s_flex_groups[flex_group].free_clusters);
1929 atomic_add(ext4_used_dirs_count(sb, gdp),
1930 &sbi->s_flex_groups[flex_group].used_dirs);
1931 }
1932
1933 return 1;
1934 failed:
1935 return 0;
1936 }
1937
1938 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1939 struct ext4_group_desc *gdp)
1940 {
1941 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1942 __u16 crc = 0;
1943 __le32 le_group = cpu_to_le32(block_group);
1944
1945 if ((sbi->s_es->s_feature_ro_compat &
1946 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1947 /* Use new metadata_csum algorithm */
1948 __u32 csum32;
1949 __u16 dummy_csum = 0;
1950
1951 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1952 sizeof(le_group));
1953 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
1954 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
1955 sizeof(dummy_csum));
1956 offset += sizeof(dummy_csum);
1957 if (offset < sbi->s_desc_size)
1958 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
1959 sbi->s_desc_size - offset);
1960
1961 crc = csum32 & 0xFFFF;
1962 goto out;
1963 }
1964
1965 /* old crc16 code */
1966 if (!(sbi->s_es->s_feature_ro_compat &
1967 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
1968 return 0;
1969
1970 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1971 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1972 crc = crc16(crc, (__u8 *)gdp, offset);
1973 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1974 /* for checksum of struct ext4_group_desc do the rest...*/
1975 if ((sbi->s_es->s_feature_incompat &
1976 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1977 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1978 crc = crc16(crc, (__u8 *)gdp + offset,
1979 le16_to_cpu(sbi->s_es->s_desc_size) -
1980 offset);
1981
1982 out:
1983 return cpu_to_le16(crc);
1984 }
1985
1986 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
1987 struct ext4_group_desc *gdp)
1988 {
1989 if (ext4_has_group_desc_csum(sb) &&
1990 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
1991 block_group, gdp)))
1992 return 0;
1993
1994 return 1;
1995 }
1996
1997 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
1998 struct ext4_group_desc *gdp)
1999 {
2000 if (!ext4_has_group_desc_csum(sb))
2001 return;
2002 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2003 }
2004
2005 /* Called at mount-time, super-block is locked */
2006 static int ext4_check_descriptors(struct super_block *sb,
2007 ext4_fsblk_t sb_block,
2008 ext4_group_t *first_not_zeroed)
2009 {
2010 struct ext4_sb_info *sbi = EXT4_SB(sb);
2011 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2012 ext4_fsblk_t last_block;
2013 ext4_fsblk_t block_bitmap;
2014 ext4_fsblk_t inode_bitmap;
2015 ext4_fsblk_t inode_table;
2016 int flexbg_flag = 0;
2017 ext4_group_t i, grp = sbi->s_groups_count;
2018
2019 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2020 flexbg_flag = 1;
2021
2022 ext4_debug("Checking group descriptors");
2023
2024 for (i = 0; i < sbi->s_groups_count; i++) {
2025 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2026
2027 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2028 last_block = ext4_blocks_count(sbi->s_es) - 1;
2029 else
2030 last_block = first_block +
2031 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2032
2033 if ((grp == sbi->s_groups_count) &&
2034 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2035 grp = i;
2036
2037 block_bitmap = ext4_block_bitmap(sb, gdp);
2038 if (block_bitmap == sb_block) {
2039 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2040 "Block bitmap for group %u overlaps "
2041 "superblock", i);
2042 }
2043 if (block_bitmap < first_block || block_bitmap > last_block) {
2044 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2045 "Block bitmap for group %u not in group "
2046 "(block %llu)!", i, block_bitmap);
2047 return 0;
2048 }
2049 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2050 if (inode_bitmap == sb_block) {
2051 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2052 "Inode bitmap for group %u overlaps "
2053 "superblock", i);
2054 }
2055 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2056 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2057 "Inode bitmap for group %u not in group "
2058 "(block %llu)!", i, inode_bitmap);
2059 return 0;
2060 }
2061 inode_table = ext4_inode_table(sb, gdp);
2062 if (inode_table == sb_block) {
2063 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2064 "Inode table for group %u overlaps "
2065 "superblock", i);
2066 }
2067 if (inode_table < first_block ||
2068 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2069 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2070 "Inode table for group %u not in group "
2071 "(block %llu)!", i, inode_table);
2072 return 0;
2073 }
2074 ext4_lock_group(sb, i);
2075 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2076 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2077 "Checksum for group %u failed (%u!=%u)",
2078 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2079 gdp)), le16_to_cpu(gdp->bg_checksum));
2080 if (!(sb->s_flags & MS_RDONLY)) {
2081 ext4_unlock_group(sb, i);
2082 return 0;
2083 }
2084 }
2085 ext4_unlock_group(sb, i);
2086 if (!flexbg_flag)
2087 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2088 }
2089 if (NULL != first_not_zeroed)
2090 *first_not_zeroed = grp;
2091
2092 ext4_free_blocks_count_set(sbi->s_es,
2093 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2094 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2095 return 1;
2096 }
2097
2098 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2099 * the superblock) which were deleted from all directories, but held open by
2100 * a process at the time of a crash. We walk the list and try to delete these
2101 * inodes at recovery time (only with a read-write filesystem).
2102 *
2103 * In order to keep the orphan inode chain consistent during traversal (in
2104 * case of crash during recovery), we link each inode into the superblock
2105 * orphan list_head and handle it the same way as an inode deletion during
2106 * normal operation (which journals the operations for us).
2107 *
2108 * We only do an iget() and an iput() on each inode, which is very safe if we
2109 * accidentally point at an in-use or already deleted inode. The worst that
2110 * can happen in this case is that we get a "bit already cleared" message from
2111 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2112 * e2fsck was run on this filesystem, and it must have already done the orphan
2113 * inode cleanup for us, so we can safely abort without any further action.
2114 */
2115 static void ext4_orphan_cleanup(struct super_block *sb,
2116 struct ext4_super_block *es)
2117 {
2118 unsigned int s_flags = sb->s_flags;
2119 int nr_orphans = 0, nr_truncates = 0;
2120 #ifdef CONFIG_QUOTA
2121 int i;
2122 #endif
2123 if (!es->s_last_orphan) {
2124 jbd_debug(4, "no orphan inodes to clean up\n");
2125 return;
2126 }
2127
2128 if (bdev_read_only(sb->s_bdev)) {
2129 ext4_msg(sb, KERN_ERR, "write access "
2130 "unavailable, skipping orphan cleanup");
2131 return;
2132 }
2133
2134 /* Check if feature set would not allow a r/w mount */
2135 if (!ext4_feature_set_ok(sb, 0)) {
2136 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2137 "unknown ROCOMPAT features");
2138 return;
2139 }
2140
2141 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2142 /* don't clear list on RO mount w/ errors */
2143 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2144 jbd_debug(1, "Errors on filesystem, "
2145 "clearing orphan list.\n");
2146 es->s_last_orphan = 0;
2147 }
2148 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2149 return;
2150 }
2151
2152 if (s_flags & MS_RDONLY) {
2153 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2154 sb->s_flags &= ~MS_RDONLY;
2155 }
2156 #ifdef CONFIG_QUOTA
2157 /* Needed for iput() to work correctly and not trash data */
2158 sb->s_flags |= MS_ACTIVE;
2159 /* Turn on quotas so that they are updated correctly */
2160 for (i = 0; i < MAXQUOTAS; i++) {
2161 if (EXT4_SB(sb)->s_qf_names[i]) {
2162 int ret = ext4_quota_on_mount(sb, i);
2163 if (ret < 0)
2164 ext4_msg(sb, KERN_ERR,
2165 "Cannot turn on journaled "
2166 "quota: error %d", ret);
2167 }
2168 }
2169 #endif
2170
2171 while (es->s_last_orphan) {
2172 struct inode *inode;
2173
2174 /*
2175 * We may have encountered an error during cleanup; if
2176 * so, skip the rest.
2177 */
2178 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2179 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2180 es->s_last_orphan = 0;
2181 break;
2182 }
2183
2184 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2185 if (IS_ERR(inode)) {
2186 es->s_last_orphan = 0;
2187 break;
2188 }
2189
2190 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2191 dquot_initialize(inode);
2192 if (inode->i_nlink) {
2193 ext4_msg(sb, KERN_DEBUG,
2194 "%s: truncating inode %lu to %lld bytes",
2195 __func__, inode->i_ino, inode->i_size);
2196 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2197 inode->i_ino, inode->i_size);
2198 mutex_lock(&inode->i_mutex);
2199 ext4_truncate(inode);
2200 mutex_unlock(&inode->i_mutex);
2201 nr_truncates++;
2202 } else {
2203 ext4_msg(sb, KERN_DEBUG,
2204 "%s: deleting unreferenced inode %lu",
2205 __func__, inode->i_ino);
2206 jbd_debug(2, "deleting unreferenced inode %lu\n",
2207 inode->i_ino);
2208 nr_orphans++;
2209 }
2210 iput(inode); /* The delete magic happens here! */
2211 }
2212
2213 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2214
2215 if (nr_orphans)
2216 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2217 PLURAL(nr_orphans));
2218 if (nr_truncates)
2219 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2220 PLURAL(nr_truncates));
2221 #ifdef CONFIG_QUOTA
2222 /* Turn quotas off */
2223 for (i = 0; i < MAXQUOTAS; i++) {
2224 if (sb_dqopt(sb)->files[i])
2225 dquot_quota_off(sb, i);
2226 }
2227 #endif
2228 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2229 }
2230
2231 /*
2232 * Maximal extent format file size.
2233 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2234 * extent format containers, within a sector_t, and within i_blocks
2235 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2236 * so that won't be a limiting factor.
2237 *
2238 * However there is other limiting factor. We do store extents in the form
2239 * of starting block and length, hence the resulting length of the extent
2240 * covering maximum file size must fit into on-disk format containers as
2241 * well. Given that length is always by 1 unit bigger than max unit (because
2242 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2243 *
2244 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2245 */
2246 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2247 {
2248 loff_t res;
2249 loff_t upper_limit = MAX_LFS_FILESIZE;
2250
2251 /* small i_blocks in vfs inode? */
2252 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2253 /*
2254 * CONFIG_LBDAF is not enabled implies the inode
2255 * i_block represent total blocks in 512 bytes
2256 * 32 == size of vfs inode i_blocks * 8
2257 */
2258 upper_limit = (1LL << 32) - 1;
2259
2260 /* total blocks in file system block size */
2261 upper_limit >>= (blkbits - 9);
2262 upper_limit <<= blkbits;
2263 }
2264
2265 /*
2266 * 32-bit extent-start container, ee_block. We lower the maxbytes
2267 * by one fs block, so ee_len can cover the extent of maximum file
2268 * size
2269 */
2270 res = (1LL << 32) - 1;
2271 res <<= blkbits;
2272
2273 /* Sanity check against vm- & vfs- imposed limits */
2274 if (res > upper_limit)
2275 res = upper_limit;
2276
2277 return res;
2278 }
2279
2280 /*
2281 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2282 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2283 * We need to be 1 filesystem block less than the 2^48 sector limit.
2284 */
2285 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2286 {
2287 loff_t res = EXT4_NDIR_BLOCKS;
2288 int meta_blocks;
2289 loff_t upper_limit;
2290 /* This is calculated to be the largest file size for a dense, block
2291 * mapped file such that the file's total number of 512-byte sectors,
2292 * including data and all indirect blocks, does not exceed (2^48 - 1).
2293 *
2294 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2295 * number of 512-byte sectors of the file.
2296 */
2297
2298 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2299 /*
2300 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2301 * the inode i_block field represents total file blocks in
2302 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2303 */
2304 upper_limit = (1LL << 32) - 1;
2305
2306 /* total blocks in file system block size */
2307 upper_limit >>= (bits - 9);
2308
2309 } else {
2310 /*
2311 * We use 48 bit ext4_inode i_blocks
2312 * With EXT4_HUGE_FILE_FL set the i_blocks
2313 * represent total number of blocks in
2314 * file system block size
2315 */
2316 upper_limit = (1LL << 48) - 1;
2317
2318 }
2319
2320 /* indirect blocks */
2321 meta_blocks = 1;
2322 /* double indirect blocks */
2323 meta_blocks += 1 + (1LL << (bits-2));
2324 /* tripple indirect blocks */
2325 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2326
2327 upper_limit -= meta_blocks;
2328 upper_limit <<= bits;
2329
2330 res += 1LL << (bits-2);
2331 res += 1LL << (2*(bits-2));
2332 res += 1LL << (3*(bits-2));
2333 res <<= bits;
2334 if (res > upper_limit)
2335 res = upper_limit;
2336
2337 if (res > MAX_LFS_FILESIZE)
2338 res = MAX_LFS_FILESIZE;
2339
2340 return res;
2341 }
2342
2343 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2344 ext4_fsblk_t logical_sb_block, int nr)
2345 {
2346 struct ext4_sb_info *sbi = EXT4_SB(sb);
2347 ext4_group_t bg, first_meta_bg;
2348 int has_super = 0;
2349
2350 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2351
2352 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2353 nr < first_meta_bg)
2354 return logical_sb_block + nr + 1;
2355 bg = sbi->s_desc_per_block * nr;
2356 if (ext4_bg_has_super(sb, bg))
2357 has_super = 1;
2358
2359 return (has_super + ext4_group_first_block_no(sb, bg));
2360 }
2361
2362 /**
2363 * ext4_get_stripe_size: Get the stripe size.
2364 * @sbi: In memory super block info
2365 *
2366 * If we have specified it via mount option, then
2367 * use the mount option value. If the value specified at mount time is
2368 * greater than the blocks per group use the super block value.
2369 * If the super block value is greater than blocks per group return 0.
2370 * Allocator needs it be less than blocks per group.
2371 *
2372 */
2373 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2374 {
2375 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2376 unsigned long stripe_width =
2377 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2378 int ret;
2379
2380 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2381 ret = sbi->s_stripe;
2382 else if (stripe_width <= sbi->s_blocks_per_group)
2383 ret = stripe_width;
2384 else if (stride <= sbi->s_blocks_per_group)
2385 ret = stride;
2386 else
2387 ret = 0;
2388
2389 /*
2390 * If the stripe width is 1, this makes no sense and
2391 * we set it to 0 to turn off stripe handling code.
2392 */
2393 if (ret <= 1)
2394 ret = 0;
2395
2396 return ret;
2397 }
2398
2399 /* sysfs supprt */
2400
2401 struct ext4_attr {
2402 struct attribute attr;
2403 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2404 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2405 const char *, size_t);
2406 int offset;
2407 };
2408
2409 static int parse_strtoull(const char *buf,
2410 unsigned long long max, unsigned long long *value)
2411 {
2412 int ret;
2413
2414 ret = kstrtoull(skip_spaces(buf), 0, value);
2415 if (!ret && *value > max)
2416 ret = -EINVAL;
2417 return ret;
2418 }
2419
2420 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2421 struct ext4_sb_info *sbi,
2422 char *buf)
2423 {
2424 return snprintf(buf, PAGE_SIZE, "%llu\n",
2425 (s64) EXT4_C2B(sbi,
2426 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2427 }
2428
2429 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2430 struct ext4_sb_info *sbi, char *buf)
2431 {
2432 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2433
2434 if (!sb->s_bdev->bd_part)
2435 return snprintf(buf, PAGE_SIZE, "0\n");
2436 return snprintf(buf, PAGE_SIZE, "%lu\n",
2437 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2438 sbi->s_sectors_written_start) >> 1);
2439 }
2440
2441 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2442 struct ext4_sb_info *sbi, char *buf)
2443 {
2444 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2445
2446 if (!sb->s_bdev->bd_part)
2447 return snprintf(buf, PAGE_SIZE, "0\n");
2448 return snprintf(buf, PAGE_SIZE, "%llu\n",
2449 (unsigned long long)(sbi->s_kbytes_written +
2450 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2451 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2452 }
2453
2454 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2455 struct ext4_sb_info *sbi,
2456 const char *buf, size_t count)
2457 {
2458 unsigned long t;
2459 int ret;
2460
2461 ret = kstrtoul(skip_spaces(buf), 0, &t);
2462 if (ret)
2463 return ret;
2464
2465 if (t && (!is_power_of_2(t) || t > 0x40000000))
2466 return -EINVAL;
2467
2468 sbi->s_inode_readahead_blks = t;
2469 return count;
2470 }
2471
2472 static ssize_t sbi_ui_show(struct ext4_attr *a,
2473 struct ext4_sb_info *sbi, char *buf)
2474 {
2475 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2476
2477 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2478 }
2479
2480 static ssize_t sbi_ui_store(struct ext4_attr *a,
2481 struct ext4_sb_info *sbi,
2482 const char *buf, size_t count)
2483 {
2484 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2485 unsigned long t;
2486 int ret;
2487
2488 ret = kstrtoul(skip_spaces(buf), 0, &t);
2489 if (ret)
2490 return ret;
2491 *ui = t;
2492 return count;
2493 }
2494
2495 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2496 struct ext4_sb_info *sbi, char *buf)
2497 {
2498 return snprintf(buf, PAGE_SIZE, "%llu\n",
2499 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2500 }
2501
2502 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2503 struct ext4_sb_info *sbi,
2504 const char *buf, size_t count)
2505 {
2506 unsigned long long val;
2507 int ret;
2508
2509 if (parse_strtoull(buf, -1ULL, &val))
2510 return -EINVAL;
2511 ret = ext4_reserve_clusters(sbi, val);
2512
2513 return ret ? ret : count;
2514 }
2515
2516 static ssize_t trigger_test_error(struct ext4_attr *a,
2517 struct ext4_sb_info *sbi,
2518 const char *buf, size_t count)
2519 {
2520 int len = count;
2521
2522 if (!capable(CAP_SYS_ADMIN))
2523 return -EPERM;
2524
2525 if (len && buf[len-1] == '\n')
2526 len--;
2527
2528 if (len)
2529 ext4_error(sbi->s_sb, "%.*s", len, buf);
2530 return count;
2531 }
2532
2533 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2534 static struct ext4_attr ext4_attr_##_name = { \
2535 .attr = {.name = __stringify(_name), .mode = _mode }, \
2536 .show = _show, \
2537 .store = _store, \
2538 .offset = offsetof(struct ext4_sb_info, _elname), \
2539 }
2540 #define EXT4_ATTR(name, mode, show, store) \
2541 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2542
2543 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2544 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2545 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2546 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2547 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2548 #define ATTR_LIST(name) &ext4_attr_##name.attr
2549
2550 EXT4_RO_ATTR(delayed_allocation_blocks);
2551 EXT4_RO_ATTR(session_write_kbytes);
2552 EXT4_RO_ATTR(lifetime_write_kbytes);
2553 EXT4_RW_ATTR(reserved_clusters);
2554 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2555 inode_readahead_blks_store, s_inode_readahead_blks);
2556 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2557 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2558 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2559 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2560 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2561 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2562 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2563 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2564 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2565 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2566
2567 static struct attribute *ext4_attrs[] = {
2568 ATTR_LIST(delayed_allocation_blocks),
2569 ATTR_LIST(session_write_kbytes),
2570 ATTR_LIST(lifetime_write_kbytes),
2571 ATTR_LIST(reserved_clusters),
2572 ATTR_LIST(inode_readahead_blks),
2573 ATTR_LIST(inode_goal),
2574 ATTR_LIST(mb_stats),
2575 ATTR_LIST(mb_max_to_scan),
2576 ATTR_LIST(mb_min_to_scan),
2577 ATTR_LIST(mb_order2_req),
2578 ATTR_LIST(mb_stream_req),
2579 ATTR_LIST(mb_group_prealloc),
2580 ATTR_LIST(max_writeback_mb_bump),
2581 ATTR_LIST(extent_max_zeroout_kb),
2582 ATTR_LIST(trigger_fs_error),
2583 NULL,
2584 };
2585
2586 /* Features this copy of ext4 supports */
2587 EXT4_INFO_ATTR(lazy_itable_init);
2588 EXT4_INFO_ATTR(batched_discard);
2589 EXT4_INFO_ATTR(meta_bg_resize);
2590
2591 static struct attribute *ext4_feat_attrs[] = {
2592 ATTR_LIST(lazy_itable_init),
2593 ATTR_LIST(batched_discard),
2594 ATTR_LIST(meta_bg_resize),
2595 NULL,
2596 };
2597
2598 static ssize_t ext4_attr_show(struct kobject *kobj,
2599 struct attribute *attr, char *buf)
2600 {
2601 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2602 s_kobj);
2603 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2604
2605 return a->show ? a->show(a, sbi, buf) : 0;
2606 }
2607
2608 static ssize_t ext4_attr_store(struct kobject *kobj,
2609 struct attribute *attr,
2610 const char *buf, size_t len)
2611 {
2612 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2613 s_kobj);
2614 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2615
2616 return a->store ? a->store(a, sbi, buf, len) : 0;
2617 }
2618
2619 static void ext4_sb_release(struct kobject *kobj)
2620 {
2621 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2622 s_kobj);
2623 complete(&sbi->s_kobj_unregister);
2624 }
2625
2626 static const struct sysfs_ops ext4_attr_ops = {
2627 .show = ext4_attr_show,
2628 .store = ext4_attr_store,
2629 };
2630
2631 static struct kobj_type ext4_ktype = {
2632 .default_attrs = ext4_attrs,
2633 .sysfs_ops = &ext4_attr_ops,
2634 .release = ext4_sb_release,
2635 };
2636
2637 static void ext4_feat_release(struct kobject *kobj)
2638 {
2639 complete(&ext4_feat->f_kobj_unregister);
2640 }
2641
2642 static struct kobj_type ext4_feat_ktype = {
2643 .default_attrs = ext4_feat_attrs,
2644 .sysfs_ops = &ext4_attr_ops,
2645 .release = ext4_feat_release,
2646 };
2647
2648 /*
2649 * Check whether this filesystem can be mounted based on
2650 * the features present and the RDONLY/RDWR mount requested.
2651 * Returns 1 if this filesystem can be mounted as requested,
2652 * 0 if it cannot be.
2653 */
2654 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2655 {
2656 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2657 ext4_msg(sb, KERN_ERR,
2658 "Couldn't mount because of "
2659 "unsupported optional features (%x)",
2660 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2661 ~EXT4_FEATURE_INCOMPAT_SUPP));
2662 return 0;
2663 }
2664
2665 if (readonly)
2666 return 1;
2667
2668 /* Check that feature set is OK for a read-write mount */
2669 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2670 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2671 "unsupported optional features (%x)",
2672 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2673 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2674 return 0;
2675 }
2676 /*
2677 * Large file size enabled file system can only be mounted
2678 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2679 */
2680 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2681 if (sizeof(blkcnt_t) < sizeof(u64)) {
2682 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2683 "cannot be mounted RDWR without "
2684 "CONFIG_LBDAF");
2685 return 0;
2686 }
2687 }
2688 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2689 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2690 ext4_msg(sb, KERN_ERR,
2691 "Can't support bigalloc feature without "
2692 "extents feature\n");
2693 return 0;
2694 }
2695
2696 #ifndef CONFIG_QUOTA
2697 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2698 !readonly) {
2699 ext4_msg(sb, KERN_ERR,
2700 "Filesystem with quota feature cannot be mounted RDWR "
2701 "without CONFIG_QUOTA");
2702 return 0;
2703 }
2704 #endif /* CONFIG_QUOTA */
2705 return 1;
2706 }
2707
2708 /*
2709 * This function is called once a day if we have errors logged
2710 * on the file system
2711 */
2712 static void print_daily_error_info(unsigned long arg)
2713 {
2714 struct super_block *sb = (struct super_block *) arg;
2715 struct ext4_sb_info *sbi;
2716 struct ext4_super_block *es;
2717
2718 sbi = EXT4_SB(sb);
2719 es = sbi->s_es;
2720
2721 if (es->s_error_count)
2722 /* fsck newer than v1.41.13 is needed to clean this condition. */
2723 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2724 le32_to_cpu(es->s_error_count));
2725 if (es->s_first_error_time) {
2726 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2727 sb->s_id, le32_to_cpu(es->s_first_error_time),
2728 (int) sizeof(es->s_first_error_func),
2729 es->s_first_error_func,
2730 le32_to_cpu(es->s_first_error_line));
2731 if (es->s_first_error_ino)
2732 printk(": inode %u",
2733 le32_to_cpu(es->s_first_error_ino));
2734 if (es->s_first_error_block)
2735 printk(": block %llu", (unsigned long long)
2736 le64_to_cpu(es->s_first_error_block));
2737 printk("\n");
2738 }
2739 if (es->s_last_error_time) {
2740 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2741 sb->s_id, le32_to_cpu(es->s_last_error_time),
2742 (int) sizeof(es->s_last_error_func),
2743 es->s_last_error_func,
2744 le32_to_cpu(es->s_last_error_line));
2745 if (es->s_last_error_ino)
2746 printk(": inode %u",
2747 le32_to_cpu(es->s_last_error_ino));
2748 if (es->s_last_error_block)
2749 printk(": block %llu", (unsigned long long)
2750 le64_to_cpu(es->s_last_error_block));
2751 printk("\n");
2752 }
2753 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2754 }
2755
2756 /* Find next suitable group and run ext4_init_inode_table */
2757 static int ext4_run_li_request(struct ext4_li_request *elr)
2758 {
2759 struct ext4_group_desc *gdp = NULL;
2760 ext4_group_t group, ngroups;
2761 struct super_block *sb;
2762 unsigned long timeout = 0;
2763 int ret = 0;
2764
2765 sb = elr->lr_super;
2766 ngroups = EXT4_SB(sb)->s_groups_count;
2767
2768 sb_start_write(sb);
2769 for (group = elr->lr_next_group; group < ngroups; group++) {
2770 gdp = ext4_get_group_desc(sb, group, NULL);
2771 if (!gdp) {
2772 ret = 1;
2773 break;
2774 }
2775
2776 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2777 break;
2778 }
2779
2780 if (group >= ngroups)
2781 ret = 1;
2782
2783 if (!ret) {
2784 timeout = jiffies;
2785 ret = ext4_init_inode_table(sb, group,
2786 elr->lr_timeout ? 0 : 1);
2787 if (elr->lr_timeout == 0) {
2788 timeout = (jiffies - timeout) *
2789 elr->lr_sbi->s_li_wait_mult;
2790 elr->lr_timeout = timeout;
2791 }
2792 elr->lr_next_sched = jiffies + elr->lr_timeout;
2793 elr->lr_next_group = group + 1;
2794 }
2795 sb_end_write(sb);
2796
2797 return ret;
2798 }
2799
2800 /*
2801 * Remove lr_request from the list_request and free the
2802 * request structure. Should be called with li_list_mtx held
2803 */
2804 static void ext4_remove_li_request(struct ext4_li_request *elr)
2805 {
2806 struct ext4_sb_info *sbi;
2807
2808 if (!elr)
2809 return;
2810
2811 sbi = elr->lr_sbi;
2812
2813 list_del(&elr->lr_request);
2814 sbi->s_li_request = NULL;
2815 kfree(elr);
2816 }
2817
2818 static void ext4_unregister_li_request(struct super_block *sb)
2819 {
2820 mutex_lock(&ext4_li_mtx);
2821 if (!ext4_li_info) {
2822 mutex_unlock(&ext4_li_mtx);
2823 return;
2824 }
2825
2826 mutex_lock(&ext4_li_info->li_list_mtx);
2827 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2828 mutex_unlock(&ext4_li_info->li_list_mtx);
2829 mutex_unlock(&ext4_li_mtx);
2830 }
2831
2832 static struct task_struct *ext4_lazyinit_task;
2833
2834 /*
2835 * This is the function where ext4lazyinit thread lives. It walks
2836 * through the request list searching for next scheduled filesystem.
2837 * When such a fs is found, run the lazy initialization request
2838 * (ext4_rn_li_request) and keep track of the time spend in this
2839 * function. Based on that time we compute next schedule time of
2840 * the request. When walking through the list is complete, compute
2841 * next waking time and put itself into sleep.
2842 */
2843 static int ext4_lazyinit_thread(void *arg)
2844 {
2845 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2846 struct list_head *pos, *n;
2847 struct ext4_li_request *elr;
2848 unsigned long next_wakeup, cur;
2849
2850 BUG_ON(NULL == eli);
2851 set_freezable();
2852
2853 cont_thread:
2854 while (true) {
2855 next_wakeup = MAX_JIFFY_OFFSET;
2856
2857 mutex_lock(&eli->li_list_mtx);
2858 if (list_empty(&eli->li_request_list)) {
2859 mutex_unlock(&eli->li_list_mtx);
2860 goto exit_thread;
2861 }
2862
2863 list_for_each_safe(pos, n, &eli->li_request_list) {
2864 elr = list_entry(pos, struct ext4_li_request,
2865 lr_request);
2866
2867 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2868 if (ext4_run_li_request(elr) != 0) {
2869 /* error, remove the lazy_init job */
2870 ext4_remove_li_request(elr);
2871 continue;
2872 }
2873 }
2874
2875 if (time_before(elr->lr_next_sched, next_wakeup))
2876 next_wakeup = elr->lr_next_sched;
2877 }
2878 mutex_unlock(&eli->li_list_mtx);
2879
2880 try_to_freeze();
2881
2882 cur = jiffies;
2883 if ((time_after_eq(cur, next_wakeup)) ||
2884 (MAX_JIFFY_OFFSET == next_wakeup)) {
2885 cond_resched();
2886 continue;
2887 }
2888
2889 schedule_timeout_interruptible(next_wakeup - cur);
2890
2891 if (kthread_freezable_should_stop(NULL)) {
2892 ext4_clear_request_list();
2893 goto exit_thread;
2894 }
2895 }
2896
2897 exit_thread:
2898 /*
2899 * It looks like the request list is empty, but we need
2900 * to check it under the li_list_mtx lock, to prevent any
2901 * additions into it, and of course we should lock ext4_li_mtx
2902 * to atomically free the list and ext4_li_info, because at
2903 * this point another ext4 filesystem could be registering
2904 * new one.
2905 */
2906 mutex_lock(&ext4_li_mtx);
2907 mutex_lock(&eli->li_list_mtx);
2908 if (!list_empty(&eli->li_request_list)) {
2909 mutex_unlock(&eli->li_list_mtx);
2910 mutex_unlock(&ext4_li_mtx);
2911 goto cont_thread;
2912 }
2913 mutex_unlock(&eli->li_list_mtx);
2914 kfree(ext4_li_info);
2915 ext4_li_info = NULL;
2916 mutex_unlock(&ext4_li_mtx);
2917
2918 return 0;
2919 }
2920
2921 static void ext4_clear_request_list(void)
2922 {
2923 struct list_head *pos, *n;
2924 struct ext4_li_request *elr;
2925
2926 mutex_lock(&ext4_li_info->li_list_mtx);
2927 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2928 elr = list_entry(pos, struct ext4_li_request,
2929 lr_request);
2930 ext4_remove_li_request(elr);
2931 }
2932 mutex_unlock(&ext4_li_info->li_list_mtx);
2933 }
2934
2935 static int ext4_run_lazyinit_thread(void)
2936 {
2937 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2938 ext4_li_info, "ext4lazyinit");
2939 if (IS_ERR(ext4_lazyinit_task)) {
2940 int err = PTR_ERR(ext4_lazyinit_task);
2941 ext4_clear_request_list();
2942 kfree(ext4_li_info);
2943 ext4_li_info = NULL;
2944 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2945 "initialization thread\n",
2946 err);
2947 return err;
2948 }
2949 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2950 return 0;
2951 }
2952
2953 /*
2954 * Check whether it make sense to run itable init. thread or not.
2955 * If there is at least one uninitialized inode table, return
2956 * corresponding group number, else the loop goes through all
2957 * groups and return total number of groups.
2958 */
2959 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2960 {
2961 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2962 struct ext4_group_desc *gdp = NULL;
2963
2964 for (group = 0; group < ngroups; group++) {
2965 gdp = ext4_get_group_desc(sb, group, NULL);
2966 if (!gdp)
2967 continue;
2968
2969 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2970 break;
2971 }
2972
2973 return group;
2974 }
2975
2976 static int ext4_li_info_new(void)
2977 {
2978 struct ext4_lazy_init *eli = NULL;
2979
2980 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2981 if (!eli)
2982 return -ENOMEM;
2983
2984 INIT_LIST_HEAD(&eli->li_request_list);
2985 mutex_init(&eli->li_list_mtx);
2986
2987 eli->li_state |= EXT4_LAZYINIT_QUIT;
2988
2989 ext4_li_info = eli;
2990
2991 return 0;
2992 }
2993
2994 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2995 ext4_group_t start)
2996 {
2997 struct ext4_sb_info *sbi = EXT4_SB(sb);
2998 struct ext4_li_request *elr;
2999 unsigned long rnd;
3000
3001 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3002 if (!elr)
3003 return NULL;
3004
3005 elr->lr_super = sb;
3006 elr->lr_sbi = sbi;
3007 elr->lr_next_group = start;
3008
3009 /*
3010 * Randomize first schedule time of the request to
3011 * spread the inode table initialization requests
3012 * better.
3013 */
3014 get_random_bytes(&rnd, sizeof(rnd));
3015 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3016 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3017
3018 return elr;
3019 }
3020
3021 int ext4_register_li_request(struct super_block *sb,
3022 ext4_group_t first_not_zeroed)
3023 {
3024 struct ext4_sb_info *sbi = EXT4_SB(sb);
3025 struct ext4_li_request *elr = NULL;
3026 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3027 int ret = 0;
3028
3029 mutex_lock(&ext4_li_mtx);
3030 if (sbi->s_li_request != NULL) {
3031 /*
3032 * Reset timeout so it can be computed again, because
3033 * s_li_wait_mult might have changed.
3034 */
3035 sbi->s_li_request->lr_timeout = 0;
3036 goto out;
3037 }
3038
3039 if (first_not_zeroed == ngroups ||
3040 (sb->s_flags & MS_RDONLY) ||
3041 !test_opt(sb, INIT_INODE_TABLE))
3042 goto out;
3043
3044 elr = ext4_li_request_new(sb, first_not_zeroed);
3045 if (!elr) {
3046 ret = -ENOMEM;
3047 goto out;
3048 }
3049
3050 if (NULL == ext4_li_info) {
3051 ret = ext4_li_info_new();
3052 if (ret)
3053 goto out;
3054 }
3055
3056 mutex_lock(&ext4_li_info->li_list_mtx);
3057 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3058 mutex_unlock(&ext4_li_info->li_list_mtx);
3059
3060 sbi->s_li_request = elr;
3061 /*
3062 * set elr to NULL here since it has been inserted to
3063 * the request_list and the removal and free of it is
3064 * handled by ext4_clear_request_list from now on.
3065 */
3066 elr = NULL;
3067
3068 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3069 ret = ext4_run_lazyinit_thread();
3070 if (ret)
3071 goto out;
3072 }
3073 out:
3074 mutex_unlock(&ext4_li_mtx);
3075 if (ret)
3076 kfree(elr);
3077 return ret;
3078 }
3079
3080 /*
3081 * We do not need to lock anything since this is called on
3082 * module unload.
3083 */
3084 static void ext4_destroy_lazyinit_thread(void)
3085 {
3086 /*
3087 * If thread exited earlier
3088 * there's nothing to be done.
3089 */
3090 if (!ext4_li_info || !ext4_lazyinit_task)
3091 return;
3092
3093 kthread_stop(ext4_lazyinit_task);
3094 }
3095
3096 static int set_journal_csum_feature_set(struct super_block *sb)
3097 {
3098 int ret = 1;
3099 int compat, incompat;
3100 struct ext4_sb_info *sbi = EXT4_SB(sb);
3101
3102 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3103 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3104 /* journal checksum v2 */
3105 compat = 0;
3106 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3107 } else {
3108 /* journal checksum v1 */
3109 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3110 incompat = 0;
3111 }
3112
3113 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3114 ret = jbd2_journal_set_features(sbi->s_journal,
3115 compat, 0,
3116 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3117 incompat);
3118 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3119 ret = jbd2_journal_set_features(sbi->s_journal,
3120 compat, 0,
3121 incompat);
3122 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3123 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3124 } else {
3125 jbd2_journal_clear_features(sbi->s_journal,
3126 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3127 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3128 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3129 }
3130
3131 return ret;
3132 }
3133
3134 /*
3135 * Note: calculating the overhead so we can be compatible with
3136 * historical BSD practice is quite difficult in the face of
3137 * clusters/bigalloc. This is because multiple metadata blocks from
3138 * different block group can end up in the same allocation cluster.
3139 * Calculating the exact overhead in the face of clustered allocation
3140 * requires either O(all block bitmaps) in memory or O(number of block
3141 * groups**2) in time. We will still calculate the superblock for
3142 * older file systems --- and if we come across with a bigalloc file
3143 * system with zero in s_overhead_clusters the estimate will be close to
3144 * correct especially for very large cluster sizes --- but for newer
3145 * file systems, it's better to calculate this figure once at mkfs
3146 * time, and store it in the superblock. If the superblock value is
3147 * present (even for non-bigalloc file systems), we will use it.
3148 */
3149 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3150 char *buf)
3151 {
3152 struct ext4_sb_info *sbi = EXT4_SB(sb);
3153 struct ext4_group_desc *gdp;
3154 ext4_fsblk_t first_block, last_block, b;
3155 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3156 int s, j, count = 0;
3157
3158 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3159 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3160 sbi->s_itb_per_group + 2);
3161
3162 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3163 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3164 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3165 for (i = 0; i < ngroups; i++) {
3166 gdp = ext4_get_group_desc(sb, i, NULL);
3167 b = ext4_block_bitmap(sb, gdp);
3168 if (b >= first_block && b <= last_block) {
3169 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3170 count++;
3171 }
3172 b = ext4_inode_bitmap(sb, gdp);
3173 if (b >= first_block && b <= last_block) {
3174 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3175 count++;
3176 }
3177 b = ext4_inode_table(sb, gdp);
3178 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3179 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3180 int c = EXT4_B2C(sbi, b - first_block);
3181 ext4_set_bit(c, buf);
3182 count++;
3183 }
3184 if (i != grp)
3185 continue;
3186 s = 0;
3187 if (ext4_bg_has_super(sb, grp)) {
3188 ext4_set_bit(s++, buf);
3189 count++;
3190 }
3191 j = ext4_bg_num_gdb(sb, grp);
3192 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3193 ext4_error(sb, "Invalid number of block group "
3194 "descriptor blocks: %d", j);
3195 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3196 }
3197 count += j;
3198 for (; j > 0; j--)
3199 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3200 }
3201 if (!count)
3202 return 0;
3203 return EXT4_CLUSTERS_PER_GROUP(sb) -
3204 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3205 }
3206
3207 /*
3208 * Compute the overhead and stash it in sbi->s_overhead
3209 */
3210 int ext4_calculate_overhead(struct super_block *sb)
3211 {
3212 struct ext4_sb_info *sbi = EXT4_SB(sb);
3213 struct ext4_super_block *es = sbi->s_es;
3214 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3215 ext4_fsblk_t overhead = 0;
3216 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3217
3218 if (!buf)
3219 return -ENOMEM;
3220
3221 /*
3222 * Compute the overhead (FS structures). This is constant
3223 * for a given filesystem unless the number of block groups
3224 * changes so we cache the previous value until it does.
3225 */
3226
3227 /*
3228 * All of the blocks before first_data_block are overhead
3229 */
3230 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3231
3232 /*
3233 * Add the overhead found in each block group
3234 */
3235 for (i = 0; i < ngroups; i++) {
3236 int blks;
3237
3238 blks = count_overhead(sb, i, buf);
3239 overhead += blks;
3240 if (blks)
3241 memset(buf, 0, PAGE_SIZE);
3242 cond_resched();
3243 }
3244 /* Add the journal blocks as well */
3245 if (sbi->s_journal)
3246 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3247
3248 sbi->s_overhead = overhead;
3249 smp_wmb();
3250 free_page((unsigned long) buf);
3251 return 0;
3252 }
3253
3254
3255 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3256 {
3257 ext4_fsblk_t resv_clusters;
3258
3259 /*
3260 * There's no need to reserve anything when we aren't using extents.
3261 * The space estimates are exact, there are no unwritten extents,
3262 * hole punching doesn't need new metadata... This is needed especially
3263 * to keep ext2/3 backward compatibility.
3264 */
3265 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3266 return 0;
3267 /*
3268 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3269 * This should cover the situations where we can not afford to run
3270 * out of space like for example punch hole, or converting
3271 * uninitialized extents in delalloc path. In most cases such
3272 * allocation would require 1, or 2 blocks, higher numbers are
3273 * very rare.
3274 */
3275 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3276 EXT4_SB(sb)->s_cluster_bits;
3277
3278 do_div(resv_clusters, 50);
3279 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3280
3281 return resv_clusters;
3282 }
3283
3284
3285 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3286 {
3287 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3288 sbi->s_cluster_bits;
3289
3290 if (count >= clusters)
3291 return -EINVAL;
3292
3293 atomic64_set(&sbi->s_resv_clusters, count);
3294 return 0;
3295 }
3296
3297 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3298 {
3299 char *orig_data = kstrdup(data, GFP_KERNEL);
3300 struct buffer_head *bh;
3301 struct ext4_super_block *es = NULL;
3302 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3303 ext4_fsblk_t block;
3304 ext4_fsblk_t sb_block = get_sb_block(&data);
3305 ext4_fsblk_t logical_sb_block;
3306 unsigned long offset = 0;
3307 unsigned long journal_devnum = 0;
3308 unsigned long def_mount_opts;
3309 struct inode *root;
3310 char *cp;
3311 const char *descr;
3312 int ret = -ENOMEM;
3313 int blocksize, clustersize;
3314 unsigned int db_count;
3315 unsigned int i;
3316 int needs_recovery, has_huge_files, has_bigalloc;
3317 __u64 blocks_count;
3318 int err = 0;
3319 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3320 ext4_group_t first_not_zeroed;
3321
3322 if ((data && !orig_data) || !sbi)
3323 goto out_free_base;
3324
3325 sbi->s_blockgroup_lock =
3326 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3327 if (!sbi->s_blockgroup_lock)
3328 goto out_free_base;
3329
3330 sb->s_fs_info = sbi;
3331 sbi->s_sb = sb;
3332 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3333 sbi->s_sb_block = sb_block;
3334 if (sb->s_bdev->bd_part)
3335 sbi->s_sectors_written_start =
3336 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3337
3338 /* Cleanup superblock name */
3339 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3340 *cp = '!';
3341
3342 /* -EINVAL is default */
3343 ret = -EINVAL;
3344 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3345 if (!blocksize) {
3346 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3347 goto out_fail;
3348 }
3349
3350 /*
3351 * The ext4 superblock will not be buffer aligned for other than 1kB
3352 * block sizes. We need to calculate the offset from buffer start.
3353 */
3354 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3355 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3356 offset = do_div(logical_sb_block, blocksize);
3357 } else {
3358 logical_sb_block = sb_block;
3359 }
3360
3361 if (!(bh = sb_bread(sb, logical_sb_block))) {
3362 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3363 goto out_fail;
3364 }
3365 /*
3366 * Note: s_es must be initialized as soon as possible because
3367 * some ext4 macro-instructions depend on its value
3368 */
3369 es = (struct ext4_super_block *) (bh->b_data + offset);
3370 sbi->s_es = es;
3371 sb->s_magic = le16_to_cpu(es->s_magic);
3372 if (sb->s_magic != EXT4_SUPER_MAGIC)
3373 goto cantfind_ext4;
3374 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3375
3376 /* Warn if metadata_csum and gdt_csum are both set. */
3377 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3378 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3379 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3380 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3381 "redundant flags; please run fsck.");
3382
3383 /* Check for a known checksum algorithm */
3384 if (!ext4_verify_csum_type(sb, es)) {
3385 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3386 "unknown checksum algorithm.");
3387 silent = 1;
3388 goto cantfind_ext4;
3389 }
3390
3391 /* Load the checksum driver */
3392 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3393 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3394 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3395 if (IS_ERR(sbi->s_chksum_driver)) {
3396 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3397 ret = PTR_ERR(sbi->s_chksum_driver);
3398 sbi->s_chksum_driver = NULL;
3399 goto failed_mount;
3400 }
3401 }
3402
3403 /* Check superblock checksum */
3404 if (!ext4_superblock_csum_verify(sb, es)) {
3405 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3406 "invalid superblock checksum. Run e2fsck?");
3407 silent = 1;
3408 goto cantfind_ext4;
3409 }
3410
3411 /* Precompute checksum seed for all metadata */
3412 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3413 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3414 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3415 sizeof(es->s_uuid));
3416
3417 /* Set defaults before we parse the mount options */
3418 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3419 set_opt(sb, INIT_INODE_TABLE);
3420 if (def_mount_opts & EXT4_DEFM_DEBUG)
3421 set_opt(sb, DEBUG);
3422 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3423 set_opt(sb, GRPID);
3424 if (def_mount_opts & EXT4_DEFM_UID16)
3425 set_opt(sb, NO_UID32);
3426 /* xattr user namespace & acls are now defaulted on */
3427 set_opt(sb, XATTR_USER);
3428 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3429 set_opt(sb, POSIX_ACL);
3430 #endif
3431 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3432 set_opt(sb, JOURNAL_DATA);
3433 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3434 set_opt(sb, ORDERED_DATA);
3435 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3436 set_opt(sb, WRITEBACK_DATA);
3437
3438 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3439 set_opt(sb, ERRORS_PANIC);
3440 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3441 set_opt(sb, ERRORS_CONT);
3442 else
3443 set_opt(sb, ERRORS_RO);
3444 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3445 set_opt(sb, BLOCK_VALIDITY);
3446 if (def_mount_opts & EXT4_DEFM_DISCARD)
3447 set_opt(sb, DISCARD);
3448
3449 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3450 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3451 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3452 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3453 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3454
3455 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3456 set_opt(sb, BARRIER);
3457
3458 /*
3459 * enable delayed allocation by default
3460 * Use -o nodelalloc to turn it off
3461 */
3462 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3463 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3464 set_opt(sb, DELALLOC);
3465
3466 /*
3467 * set default s_li_wait_mult for lazyinit, for the case there is
3468 * no mount option specified.
3469 */
3470 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3471
3472 if (sbi->s_es->s_mount_opts[0]) {
3473 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3474 sizeof(sbi->s_es->s_mount_opts),
3475 GFP_KERNEL);
3476 if (!s_mount_opts)
3477 goto failed_mount;
3478 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3479 &journal_ioprio, 0)) {
3480 ext4_msg(sb, KERN_WARNING,
3481 "failed to parse options in superblock: %s",
3482 s_mount_opts);
3483 }
3484 kfree(s_mount_opts);
3485 }
3486 sbi->s_def_mount_opt = sbi->s_mount_opt;
3487 if (!parse_options((char *) data, sb, &journal_devnum,
3488 &journal_ioprio, 0))
3489 goto failed_mount;
3490
3491 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3492 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3493 "with data=journal disables delayed "
3494 "allocation and O_DIRECT support!\n");
3495 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3496 ext4_msg(sb, KERN_ERR, "can't mount with "
3497 "both data=journal and delalloc");
3498 goto failed_mount;
3499 }
3500 if (test_opt(sb, DIOREAD_NOLOCK)) {
3501 ext4_msg(sb, KERN_ERR, "can't mount with "
3502 "both data=journal and dioread_nolock");
3503 goto failed_mount;
3504 }
3505 if (test_opt(sb, DELALLOC))
3506 clear_opt(sb, DELALLOC);
3507 }
3508
3509 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3510 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3511
3512 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3513 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3514 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3515 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3516 ext4_msg(sb, KERN_WARNING,
3517 "feature flags set on rev 0 fs, "
3518 "running e2fsck is recommended");
3519
3520 if (IS_EXT2_SB(sb)) {
3521 if (ext2_feature_set_ok(sb))
3522 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3523 "using the ext4 subsystem");
3524 else {
3525 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3526 "to feature incompatibilities");
3527 goto failed_mount;
3528 }
3529 }
3530
3531 if (IS_EXT3_SB(sb)) {
3532 if (ext3_feature_set_ok(sb))
3533 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3534 "using the ext4 subsystem");
3535 else {
3536 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3537 "to feature incompatibilities");
3538 goto failed_mount;
3539 }
3540 }
3541
3542 /*
3543 * Check feature flags regardless of the revision level, since we
3544 * previously didn't change the revision level when setting the flags,
3545 * so there is a chance incompat flags are set on a rev 0 filesystem.
3546 */
3547 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3548 goto failed_mount;
3549
3550 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3551 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3552 blocksize > EXT4_MAX_BLOCK_SIZE) {
3553 ext4_msg(sb, KERN_ERR,
3554 "Unsupported filesystem blocksize %d (%d log_block_size)",
3555 blocksize, le32_to_cpu(es->s_log_block_size));
3556 goto failed_mount;
3557 }
3558 if (le32_to_cpu(es->s_log_block_size) >
3559 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3560 ext4_msg(sb, KERN_ERR,
3561 "Invalid log block size: %u",
3562 le32_to_cpu(es->s_log_block_size));
3563 goto failed_mount;
3564 }
3565
3566 if (sb->s_blocksize != blocksize) {
3567 /* Validate the filesystem blocksize */
3568 if (!sb_set_blocksize(sb, blocksize)) {
3569 ext4_msg(sb, KERN_ERR, "bad block size %d",
3570 blocksize);
3571 goto failed_mount;
3572 }
3573
3574 brelse(bh);
3575 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3576 offset = do_div(logical_sb_block, blocksize);
3577 bh = sb_bread(sb, logical_sb_block);
3578 if (!bh) {
3579 ext4_msg(sb, KERN_ERR,
3580 "Can't read superblock on 2nd try");
3581 goto failed_mount;
3582 }
3583 es = (struct ext4_super_block *)(bh->b_data + offset);
3584 sbi->s_es = es;
3585 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3586 ext4_msg(sb, KERN_ERR,
3587 "Magic mismatch, very weird!");
3588 goto failed_mount;
3589 }
3590 }
3591
3592 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3593 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3594 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3595 has_huge_files);
3596 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3597
3598 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3599 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3600 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3601 } else {
3602 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3603 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3604 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3605 (!is_power_of_2(sbi->s_inode_size)) ||
3606 (sbi->s_inode_size > blocksize)) {
3607 ext4_msg(sb, KERN_ERR,
3608 "unsupported inode size: %d",
3609 sbi->s_inode_size);
3610 goto failed_mount;
3611 }
3612 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3613 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3614 }
3615
3616 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3617 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3618 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3619 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3620 !is_power_of_2(sbi->s_desc_size)) {
3621 ext4_msg(sb, KERN_ERR,
3622 "unsupported descriptor size %lu",
3623 sbi->s_desc_size);
3624 goto failed_mount;
3625 }
3626 } else
3627 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3628
3629 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3630 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3631
3632 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3633 if (sbi->s_inodes_per_block == 0)
3634 goto cantfind_ext4;
3635 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3636 sbi->s_inodes_per_group > blocksize * 8) {
3637 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3638 sbi->s_blocks_per_group);
3639 goto failed_mount;
3640 }
3641 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3642 sbi->s_inodes_per_block;
3643 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3644 sbi->s_sbh = bh;
3645 sbi->s_mount_state = le16_to_cpu(es->s_state);
3646 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3647 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3648
3649 for (i = 0; i < 4; i++)
3650 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3651 sbi->s_def_hash_version = es->s_def_hash_version;
3652 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3653 i = le32_to_cpu(es->s_flags);
3654 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3655 sbi->s_hash_unsigned = 3;
3656 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3657 #ifdef __CHAR_UNSIGNED__
3658 if (!(sb->s_flags & MS_RDONLY))
3659 es->s_flags |=
3660 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3661 sbi->s_hash_unsigned = 3;
3662 #else
3663 if (!(sb->s_flags & MS_RDONLY))
3664 es->s_flags |=
3665 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3666 #endif
3667 }
3668 }
3669
3670 /* Handle clustersize */
3671 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3672 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3673 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3674 if (has_bigalloc) {
3675 if (clustersize < blocksize) {
3676 ext4_msg(sb, KERN_ERR,
3677 "cluster size (%d) smaller than "
3678 "block size (%d)", clustersize, blocksize);
3679 goto failed_mount;
3680 }
3681 if (le32_to_cpu(es->s_log_cluster_size) >
3682 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3683 ext4_msg(sb, KERN_ERR,
3684 "Invalid log cluster size: %u",
3685 le32_to_cpu(es->s_log_cluster_size));
3686 goto failed_mount;
3687 }
3688 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3689 le32_to_cpu(es->s_log_block_size);
3690 sbi->s_clusters_per_group =
3691 le32_to_cpu(es->s_clusters_per_group);
3692 if (sbi->s_clusters_per_group > blocksize * 8) {
3693 ext4_msg(sb, KERN_ERR,
3694 "#clusters per group too big: %lu",
3695 sbi->s_clusters_per_group);
3696 goto failed_mount;
3697 }
3698 if (sbi->s_blocks_per_group !=
3699 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3700 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3701 "clusters per group (%lu) inconsistent",
3702 sbi->s_blocks_per_group,
3703 sbi->s_clusters_per_group);
3704 goto failed_mount;
3705 }
3706 } else {
3707 if (clustersize != blocksize) {
3708 ext4_warning(sb, "fragment/cluster size (%d) != "
3709 "block size (%d)", clustersize,
3710 blocksize);
3711 clustersize = blocksize;
3712 }
3713 if (sbi->s_blocks_per_group > blocksize * 8) {
3714 ext4_msg(sb, KERN_ERR,
3715 "#blocks per group too big: %lu",
3716 sbi->s_blocks_per_group);
3717 goto failed_mount;
3718 }
3719 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3720 sbi->s_cluster_bits = 0;
3721 }
3722 sbi->s_cluster_ratio = clustersize / blocksize;
3723
3724 /* Do we have standard group size of clustersize * 8 blocks ? */
3725 if (sbi->s_blocks_per_group == clustersize << 3)
3726 set_opt2(sb, STD_GROUP_SIZE);
3727
3728 /*
3729 * Test whether we have more sectors than will fit in sector_t,
3730 * and whether the max offset is addressable by the page cache.
3731 */
3732 err = generic_check_addressable(sb->s_blocksize_bits,
3733 ext4_blocks_count(es));
3734 if (err) {
3735 ext4_msg(sb, KERN_ERR, "filesystem"
3736 " too large to mount safely on this system");
3737 if (sizeof(sector_t) < 8)
3738 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3739 goto failed_mount;
3740 }
3741
3742 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3743 goto cantfind_ext4;
3744
3745 /* check blocks count against device size */
3746 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3747 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3748 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3749 "exceeds size of device (%llu blocks)",
3750 ext4_blocks_count(es), blocks_count);
3751 goto failed_mount;
3752 }
3753
3754 /*
3755 * It makes no sense for the first data block to be beyond the end
3756 * of the filesystem.
3757 */
3758 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3759 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3760 "block %u is beyond end of filesystem (%llu)",
3761 le32_to_cpu(es->s_first_data_block),
3762 ext4_blocks_count(es));
3763 goto failed_mount;
3764 }
3765 blocks_count = (ext4_blocks_count(es) -
3766 le32_to_cpu(es->s_first_data_block) +
3767 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3768 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3769 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3770 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3771 "(block count %llu, first data block %u, "
3772 "blocks per group %lu)", sbi->s_groups_count,
3773 ext4_blocks_count(es),
3774 le32_to_cpu(es->s_first_data_block),
3775 EXT4_BLOCKS_PER_GROUP(sb));
3776 goto failed_mount;
3777 }
3778 sbi->s_groups_count = blocks_count;
3779 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3780 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3781 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3782 EXT4_DESC_PER_BLOCK(sb);
3783 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG)) {
3784 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3785 ext4_msg(sb, KERN_WARNING,
3786 "first meta block group too large: %u "
3787 "(group descriptor block count %u)",
3788 le32_to_cpu(es->s_first_meta_bg), db_count);
3789 goto failed_mount;
3790 }
3791 }
3792 sbi->s_group_desc = ext4_kvmalloc(db_count *
3793 sizeof(struct buffer_head *),
3794 GFP_KERNEL);
3795 if (sbi->s_group_desc == NULL) {
3796 ext4_msg(sb, KERN_ERR, "not enough memory");
3797 ret = -ENOMEM;
3798 goto failed_mount;
3799 }
3800
3801 if (ext4_proc_root)
3802 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3803
3804 if (sbi->s_proc)
3805 proc_create_data("options", S_IRUGO, sbi->s_proc,
3806 &ext4_seq_options_fops, sb);
3807
3808 bgl_lock_init(sbi->s_blockgroup_lock);
3809
3810 for (i = 0; i < db_count; i++) {
3811 block = descriptor_loc(sb, logical_sb_block, i);
3812 sbi->s_group_desc[i] = sb_bread(sb, block);
3813 if (!sbi->s_group_desc[i]) {
3814 ext4_msg(sb, KERN_ERR,
3815 "can't read group descriptor %d", i);
3816 db_count = i;
3817 goto failed_mount2;
3818 }
3819 }
3820 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3821 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3822 goto failed_mount2;
3823 }
3824 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3825 if (!ext4_fill_flex_info(sb)) {
3826 ext4_msg(sb, KERN_ERR,
3827 "unable to initialize "
3828 "flex_bg meta info!");
3829 goto failed_mount2;
3830 }
3831
3832 sbi->s_gdb_count = db_count;
3833 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3834 spin_lock_init(&sbi->s_next_gen_lock);
3835
3836 init_timer(&sbi->s_err_report);
3837 sbi->s_err_report.function = print_daily_error_info;
3838 sbi->s_err_report.data = (unsigned long) sb;
3839
3840 /* Register extent status tree shrinker */
3841 ext4_es_register_shrinker(sb);
3842
3843 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3844 ext4_count_free_clusters(sb));
3845 if (!err) {
3846 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3847 ext4_count_free_inodes(sb));
3848 }
3849 if (!err) {
3850 err = percpu_counter_init(&sbi->s_dirs_counter,
3851 ext4_count_dirs(sb));
3852 }
3853 if (!err) {
3854 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3855 }
3856 if (!err) {
3857 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3858 }
3859 if (err) {
3860 ext4_msg(sb, KERN_ERR, "insufficient memory");
3861 goto failed_mount3;
3862 }
3863
3864 sbi->s_stripe = ext4_get_stripe_size(sbi);
3865 sbi->s_max_writeback_mb_bump = 128;
3866 sbi->s_extent_max_zeroout_kb = 32;
3867
3868 /*
3869 * set up enough so that it can read an inode
3870 */
3871 if (!test_opt(sb, NOLOAD) &&
3872 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3873 sb->s_op = &ext4_sops;
3874 else
3875 sb->s_op = &ext4_nojournal_sops;
3876 sb->s_export_op = &ext4_export_ops;
3877 sb->s_xattr = ext4_xattr_handlers;
3878 #ifdef CONFIG_QUOTA
3879 sb->dq_op = &ext4_quota_operations;
3880 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3881 sb->s_qcop = &ext4_qctl_sysfile_operations;
3882 else
3883 sb->s_qcop = &ext4_qctl_operations;
3884 #endif
3885 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3886
3887 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3888 mutex_init(&sbi->s_orphan_lock);
3889
3890 sb->s_root = NULL;
3891
3892 needs_recovery = (es->s_last_orphan != 0 ||
3893 EXT4_HAS_INCOMPAT_FEATURE(sb,
3894 EXT4_FEATURE_INCOMPAT_RECOVER));
3895
3896 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3897 !(sb->s_flags & MS_RDONLY))
3898 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3899 goto failed_mount3;
3900
3901 /*
3902 * The first inode we look at is the journal inode. Don't try
3903 * root first: it may be modified in the journal!
3904 */
3905 if (!test_opt(sb, NOLOAD) &&
3906 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3907 err = ext4_load_journal(sb, es, journal_devnum);
3908 if (err)
3909 goto failed_mount3;
3910 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3911 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3912 ext4_msg(sb, KERN_ERR, "required journal recovery "
3913 "suppressed and not mounted read-only");
3914 goto failed_mount_wq;
3915 } else {
3916 clear_opt(sb, DATA_FLAGS);
3917 sbi->s_journal = NULL;
3918 needs_recovery = 0;
3919 goto no_journal;
3920 }
3921
3922 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3923 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3924 JBD2_FEATURE_INCOMPAT_64BIT)) {
3925 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3926 goto failed_mount_wq;
3927 }
3928
3929 if (!set_journal_csum_feature_set(sb)) {
3930 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3931 "feature set");
3932 goto failed_mount_wq;
3933 }
3934
3935 /* We have now updated the journal if required, so we can
3936 * validate the data journaling mode. */
3937 switch (test_opt(sb, DATA_FLAGS)) {
3938 case 0:
3939 /* No mode set, assume a default based on the journal
3940 * capabilities: ORDERED_DATA if the journal can
3941 * cope, else JOURNAL_DATA
3942 */
3943 if (jbd2_journal_check_available_features
3944 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3945 set_opt(sb, ORDERED_DATA);
3946 else
3947 set_opt(sb, JOURNAL_DATA);
3948 break;
3949
3950 case EXT4_MOUNT_ORDERED_DATA:
3951 case EXT4_MOUNT_WRITEBACK_DATA:
3952 if (!jbd2_journal_check_available_features
3953 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3954 ext4_msg(sb, KERN_ERR, "Journal does not support "
3955 "requested data journaling mode");
3956 goto failed_mount_wq;
3957 }
3958 default:
3959 break;
3960 }
3961 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3962
3963 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3964
3965 /*
3966 * The journal may have updated the bg summary counts, so we
3967 * need to update the global counters.
3968 */
3969 percpu_counter_set(&sbi->s_freeclusters_counter,
3970 ext4_count_free_clusters(sb));
3971 percpu_counter_set(&sbi->s_freeinodes_counter,
3972 ext4_count_free_inodes(sb));
3973 percpu_counter_set(&sbi->s_dirs_counter,
3974 ext4_count_dirs(sb));
3975 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3976
3977 no_journal:
3978 /*
3979 * Get the # of file system overhead blocks from the
3980 * superblock if present.
3981 */
3982 if (es->s_overhead_clusters)
3983 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3984 else {
3985 err = ext4_calculate_overhead(sb);
3986 if (err)
3987 goto failed_mount_wq;
3988 }
3989
3990 /*
3991 * The maximum number of concurrent works can be high and
3992 * concurrency isn't really necessary. Limit it to 1.
3993 */
3994 EXT4_SB(sb)->dio_unwritten_wq =
3995 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3996 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3997 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3998 ret = -ENOMEM;
3999 goto failed_mount_wq;
4000 }
4001
4002 /*
4003 * The jbd2_journal_load will have done any necessary log recovery,
4004 * so we can safely mount the rest of the filesystem now.
4005 */
4006
4007 root = ext4_iget(sb, EXT4_ROOT_INO);
4008 if (IS_ERR(root)) {
4009 ext4_msg(sb, KERN_ERR, "get root inode failed");
4010 ret = PTR_ERR(root);
4011 root = NULL;
4012 goto failed_mount4;
4013 }
4014 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4015 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4016 iput(root);
4017 goto failed_mount4;
4018 }
4019 sb->s_root = d_make_root(root);
4020 if (!sb->s_root) {
4021 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4022 ret = -ENOMEM;
4023 goto failed_mount4;
4024 }
4025
4026 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4027 sb->s_flags |= MS_RDONLY;
4028
4029 /* determine the minimum size of new large inodes, if present */
4030 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4031 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4032 EXT4_GOOD_OLD_INODE_SIZE;
4033 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4034 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4035 if (sbi->s_want_extra_isize <
4036 le16_to_cpu(es->s_want_extra_isize))
4037 sbi->s_want_extra_isize =
4038 le16_to_cpu(es->s_want_extra_isize);
4039 if (sbi->s_want_extra_isize <
4040 le16_to_cpu(es->s_min_extra_isize))
4041 sbi->s_want_extra_isize =
4042 le16_to_cpu(es->s_min_extra_isize);
4043 }
4044 }
4045 /* Check if enough inode space is available */
4046 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4047 sbi->s_inode_size) {
4048 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4049 EXT4_GOOD_OLD_INODE_SIZE;
4050 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4051 "available");
4052 }
4053
4054 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4055 if (err) {
4056 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4057 "reserved pool", ext4_calculate_resv_clusters(sb));
4058 goto failed_mount4a;
4059 }
4060
4061 err = ext4_setup_system_zone(sb);
4062 if (err) {
4063 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4064 "zone (%d)", err);
4065 goto failed_mount4a;
4066 }
4067
4068 ext4_ext_init(sb);
4069 err = ext4_mb_init(sb);
4070 if (err) {
4071 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4072 err);
4073 goto failed_mount5;
4074 }
4075
4076 err = ext4_register_li_request(sb, first_not_zeroed);
4077 if (err)
4078 goto failed_mount6;
4079
4080 sbi->s_kobj.kset = ext4_kset;
4081 init_completion(&sbi->s_kobj_unregister);
4082 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4083 "%s", sb->s_id);
4084 if (err)
4085 goto failed_mount7;
4086
4087 #ifdef CONFIG_QUOTA
4088 /* Enable quota usage during mount. */
4089 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4090 !(sb->s_flags & MS_RDONLY)) {
4091 err = ext4_enable_quotas(sb);
4092 if (err)
4093 goto failed_mount8;
4094 }
4095 #endif /* CONFIG_QUOTA */
4096
4097 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4098 ext4_orphan_cleanup(sb, es);
4099 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4100 if (needs_recovery) {
4101 ext4_msg(sb, KERN_INFO, "recovery complete");
4102 ext4_mark_recovery_complete(sb, es);
4103 }
4104 if (EXT4_SB(sb)->s_journal) {
4105 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4106 descr = " journalled data mode";
4107 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4108 descr = " ordered data mode";
4109 else
4110 descr = " writeback data mode";
4111 } else
4112 descr = "out journal";
4113
4114 if (test_opt(sb, DISCARD)) {
4115 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4116 if (!blk_queue_discard(q))
4117 ext4_msg(sb, KERN_WARNING,
4118 "mounting with \"discard\" option, but "
4119 "the device does not support discard");
4120 }
4121
4122 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4123 "Opts: %.*s%s%s", descr,
4124 (int) sizeof(sbi->s_es->s_mount_opts),
4125 sbi->s_es->s_mount_opts,
4126 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4127
4128 if (es->s_error_count)
4129 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4130
4131 kfree(orig_data);
4132 return 0;
4133
4134 cantfind_ext4:
4135 if (!silent)
4136 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4137 goto failed_mount;
4138
4139 #ifdef CONFIG_QUOTA
4140 failed_mount8:
4141 kobject_del(&sbi->s_kobj);
4142 #endif
4143 failed_mount7:
4144 ext4_unregister_li_request(sb);
4145 failed_mount6:
4146 ext4_mb_release(sb);
4147 failed_mount5:
4148 ext4_ext_release(sb);
4149 ext4_release_system_zone(sb);
4150 failed_mount4a:
4151 dput(sb->s_root);
4152 sb->s_root = NULL;
4153 failed_mount4:
4154 ext4_msg(sb, KERN_ERR, "mount failed");
4155 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4156 failed_mount_wq:
4157 if (sbi->s_journal) {
4158 jbd2_journal_destroy(sbi->s_journal);
4159 sbi->s_journal = NULL;
4160 }
4161 failed_mount3:
4162 ext4_es_unregister_shrinker(sb);
4163 del_timer(&sbi->s_err_report);
4164 if (sbi->s_flex_groups)
4165 ext4_kvfree(sbi->s_flex_groups);
4166 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4167 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4168 percpu_counter_destroy(&sbi->s_dirs_counter);
4169 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4170 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4171 if (sbi->s_mmp_tsk)
4172 kthread_stop(sbi->s_mmp_tsk);
4173 failed_mount2:
4174 for (i = 0; i < db_count; i++)
4175 brelse(sbi->s_group_desc[i]);
4176 ext4_kvfree(sbi->s_group_desc);
4177 failed_mount:
4178 if (sbi->s_chksum_driver)
4179 crypto_free_shash(sbi->s_chksum_driver);
4180 if (sbi->s_proc) {
4181 remove_proc_entry("options", sbi->s_proc);
4182 remove_proc_entry(sb->s_id, ext4_proc_root);
4183 }
4184 #ifdef CONFIG_QUOTA
4185 for (i = 0; i < MAXQUOTAS; i++)
4186 kfree(sbi->s_qf_names[i]);
4187 #endif
4188 ext4_blkdev_remove(sbi);
4189 brelse(bh);
4190 out_fail:
4191 sb->s_fs_info = NULL;
4192 kfree(sbi->s_blockgroup_lock);
4193 out_free_base:
4194 kfree(sbi);
4195 kfree(orig_data);
4196 return err ? err : ret;
4197 }
4198
4199 /*
4200 * Setup any per-fs journal parameters now. We'll do this both on
4201 * initial mount, once the journal has been initialised but before we've
4202 * done any recovery; and again on any subsequent remount.
4203 */
4204 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4205 {
4206 struct ext4_sb_info *sbi = EXT4_SB(sb);
4207
4208 journal->j_commit_interval = sbi->s_commit_interval;
4209 journal->j_min_batch_time = sbi->s_min_batch_time;
4210 journal->j_max_batch_time = sbi->s_max_batch_time;
4211
4212 write_lock(&journal->j_state_lock);
4213 if (test_opt(sb, BARRIER))
4214 journal->j_flags |= JBD2_BARRIER;
4215 else
4216 journal->j_flags &= ~JBD2_BARRIER;
4217 if (test_opt(sb, DATA_ERR_ABORT))
4218 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4219 else
4220 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4221 write_unlock(&journal->j_state_lock);
4222 }
4223
4224 static journal_t *ext4_get_journal(struct super_block *sb,
4225 unsigned int journal_inum)
4226 {
4227 struct inode *journal_inode;
4228 journal_t *journal;
4229
4230 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4231
4232 /* First, test for the existence of a valid inode on disk. Bad
4233 * things happen if we iget() an unused inode, as the subsequent
4234 * iput() will try to delete it. */
4235
4236 journal_inode = ext4_iget(sb, journal_inum);
4237 if (IS_ERR(journal_inode)) {
4238 ext4_msg(sb, KERN_ERR, "no journal found");
4239 return NULL;
4240 }
4241 if (!journal_inode->i_nlink) {
4242 make_bad_inode(journal_inode);
4243 iput(journal_inode);
4244 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4245 return NULL;
4246 }
4247
4248 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4249 journal_inode, journal_inode->i_size);
4250 if (!S_ISREG(journal_inode->i_mode)) {
4251 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4252 iput(journal_inode);
4253 return NULL;
4254 }
4255
4256 journal = jbd2_journal_init_inode(journal_inode);
4257 if (!journal) {
4258 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4259 iput(journal_inode);
4260 return NULL;
4261 }
4262 journal->j_private = sb;
4263 ext4_init_journal_params(sb, journal);
4264 return journal;
4265 }
4266
4267 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4268 dev_t j_dev)
4269 {
4270 struct buffer_head *bh;
4271 journal_t *journal;
4272 ext4_fsblk_t start;
4273 ext4_fsblk_t len;
4274 int hblock, blocksize;
4275 ext4_fsblk_t sb_block;
4276 unsigned long offset;
4277 struct ext4_super_block *es;
4278 struct block_device *bdev;
4279
4280 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4281
4282 bdev = ext4_blkdev_get(j_dev, sb);
4283 if (bdev == NULL)
4284 return NULL;
4285
4286 blocksize = sb->s_blocksize;
4287 hblock = bdev_logical_block_size(bdev);
4288 if (blocksize < hblock) {
4289 ext4_msg(sb, KERN_ERR,
4290 "blocksize too small for journal device");
4291 goto out_bdev;
4292 }
4293
4294 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4295 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4296 set_blocksize(bdev, blocksize);
4297 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4298 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4299 "external journal");
4300 goto out_bdev;
4301 }
4302
4303 es = (struct ext4_super_block *) (bh->b_data + offset);
4304 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4305 !(le32_to_cpu(es->s_feature_incompat) &
4306 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4307 ext4_msg(sb, KERN_ERR, "external journal has "
4308 "bad superblock");
4309 brelse(bh);
4310 goto out_bdev;
4311 }
4312
4313 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4314 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4315 brelse(bh);
4316 goto out_bdev;
4317 }
4318
4319 len = ext4_blocks_count(es);
4320 start = sb_block + 1;
4321 brelse(bh); /* we're done with the superblock */
4322
4323 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4324 start, len, blocksize);
4325 if (!journal) {
4326 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4327 goto out_bdev;
4328 }
4329 journal->j_private = sb;
4330 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4331 wait_on_buffer(journal->j_sb_buffer);
4332 if (!buffer_uptodate(journal->j_sb_buffer)) {
4333 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4334 goto out_journal;
4335 }
4336 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4337 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4338 "user (unsupported) - %d",
4339 be32_to_cpu(journal->j_superblock->s_nr_users));
4340 goto out_journal;
4341 }
4342 EXT4_SB(sb)->journal_bdev = bdev;
4343 ext4_init_journal_params(sb, journal);
4344 return journal;
4345
4346 out_journal:
4347 jbd2_journal_destroy(journal);
4348 out_bdev:
4349 ext4_blkdev_put(bdev);
4350 return NULL;
4351 }
4352
4353 static int ext4_load_journal(struct super_block *sb,
4354 struct ext4_super_block *es,
4355 unsigned long journal_devnum)
4356 {
4357 journal_t *journal;
4358 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4359 dev_t journal_dev;
4360 int err = 0;
4361 int really_read_only;
4362
4363 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4364
4365 if (journal_devnum &&
4366 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4367 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4368 "numbers have changed");
4369 journal_dev = new_decode_dev(journal_devnum);
4370 } else
4371 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4372
4373 really_read_only = bdev_read_only(sb->s_bdev);
4374
4375 /*
4376 * Are we loading a blank journal or performing recovery after a
4377 * crash? For recovery, we need to check in advance whether we
4378 * can get read-write access to the device.
4379 */
4380 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4381 if (sb->s_flags & MS_RDONLY) {
4382 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4383 "required on readonly filesystem");
4384 if (really_read_only) {
4385 ext4_msg(sb, KERN_ERR, "write access "
4386 "unavailable, cannot proceed");
4387 return -EROFS;
4388 }
4389 ext4_msg(sb, KERN_INFO, "write access will "
4390 "be enabled during recovery");
4391 }
4392 }
4393
4394 if (journal_inum && journal_dev) {
4395 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4396 "and inode journals!");
4397 return -EINVAL;
4398 }
4399
4400 if (journal_inum) {
4401 if (!(journal = ext4_get_journal(sb, journal_inum)))
4402 return -EINVAL;
4403 } else {
4404 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4405 return -EINVAL;
4406 }
4407
4408 if (!(journal->j_flags & JBD2_BARRIER))
4409 ext4_msg(sb, KERN_INFO, "barriers disabled");
4410
4411 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4412 err = jbd2_journal_wipe(journal, !really_read_only);
4413 if (!err) {
4414 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4415 if (save)
4416 memcpy(save, ((char *) es) +
4417 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4418 err = jbd2_journal_load(journal);
4419 if (save)
4420 memcpy(((char *) es) + EXT4_S_ERR_START,
4421 save, EXT4_S_ERR_LEN);
4422 kfree(save);
4423 }
4424
4425 if (err) {
4426 ext4_msg(sb, KERN_ERR, "error loading journal");
4427 jbd2_journal_destroy(journal);
4428 return err;
4429 }
4430
4431 EXT4_SB(sb)->s_journal = journal;
4432 ext4_clear_journal_err(sb, es);
4433
4434 if (!really_read_only && journal_devnum &&
4435 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4436 es->s_journal_dev = cpu_to_le32(journal_devnum);
4437
4438 /* Make sure we flush the recovery flag to disk. */
4439 ext4_commit_super(sb, 1);
4440 }
4441
4442 return 0;
4443 }
4444
4445 static int ext4_commit_super(struct super_block *sb, int sync)
4446 {
4447 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4448 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4449 int error = 0;
4450
4451 if (!sbh || block_device_ejected(sb))
4452 return error;
4453 if (buffer_write_io_error(sbh)) {
4454 /*
4455 * Oh, dear. A previous attempt to write the
4456 * superblock failed. This could happen because the
4457 * USB device was yanked out. Or it could happen to
4458 * be a transient write error and maybe the block will
4459 * be remapped. Nothing we can do but to retry the
4460 * write and hope for the best.
4461 */
4462 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4463 "superblock detected");
4464 clear_buffer_write_io_error(sbh);
4465 set_buffer_uptodate(sbh);
4466 }
4467 /*
4468 * If the file system is mounted read-only, don't update the
4469 * superblock write time. This avoids updating the superblock
4470 * write time when we are mounting the root file system
4471 * read/only but we need to replay the journal; at that point,
4472 * for people who are east of GMT and who make their clock
4473 * tick in localtime for Windows bug-for-bug compatibility,
4474 * the clock is set in the future, and this will cause e2fsck
4475 * to complain and force a full file system check.
4476 */
4477 if (!(sb->s_flags & MS_RDONLY))
4478 es->s_wtime = cpu_to_le32(get_seconds());
4479 if (sb->s_bdev->bd_part)
4480 es->s_kbytes_written =
4481 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4482 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4483 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4484 else
4485 es->s_kbytes_written =
4486 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4487 ext4_free_blocks_count_set(es,
4488 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4489 &EXT4_SB(sb)->s_freeclusters_counter)));
4490 es->s_free_inodes_count =
4491 cpu_to_le32(percpu_counter_sum_positive(
4492 &EXT4_SB(sb)->s_freeinodes_counter));
4493 BUFFER_TRACE(sbh, "marking dirty");
4494 ext4_superblock_csum_set(sb);
4495 mark_buffer_dirty(sbh);
4496 if (sync) {
4497 error = sync_dirty_buffer(sbh);
4498 if (error)
4499 return error;
4500
4501 error = buffer_write_io_error(sbh);
4502 if (error) {
4503 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4504 "superblock");
4505 clear_buffer_write_io_error(sbh);
4506 set_buffer_uptodate(sbh);
4507 }
4508 }
4509 return error;
4510 }
4511
4512 /*
4513 * Have we just finished recovery? If so, and if we are mounting (or
4514 * remounting) the filesystem readonly, then we will end up with a
4515 * consistent fs on disk. Record that fact.
4516 */
4517 static void ext4_mark_recovery_complete(struct super_block *sb,
4518 struct ext4_super_block *es)
4519 {
4520 journal_t *journal = EXT4_SB(sb)->s_journal;
4521
4522 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4523 BUG_ON(journal != NULL);
4524 return;
4525 }
4526 jbd2_journal_lock_updates(journal);
4527 if (jbd2_journal_flush(journal) < 0)
4528 goto out;
4529
4530 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4531 sb->s_flags & MS_RDONLY) {
4532 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4533 ext4_commit_super(sb, 1);
4534 }
4535
4536 out:
4537 jbd2_journal_unlock_updates(journal);
4538 }
4539
4540 /*
4541 * If we are mounting (or read-write remounting) a filesystem whose journal
4542 * has recorded an error from a previous lifetime, move that error to the
4543 * main filesystem now.
4544 */
4545 static void ext4_clear_journal_err(struct super_block *sb,
4546 struct ext4_super_block *es)
4547 {
4548 journal_t *journal;
4549 int j_errno;
4550 const char *errstr;
4551
4552 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4553
4554 journal = EXT4_SB(sb)->s_journal;
4555
4556 /*
4557 * Now check for any error status which may have been recorded in the
4558 * journal by a prior ext4_error() or ext4_abort()
4559 */
4560
4561 j_errno = jbd2_journal_errno(journal);
4562 if (j_errno) {
4563 char nbuf[16];
4564
4565 errstr = ext4_decode_error(sb, j_errno, nbuf);
4566 ext4_warning(sb, "Filesystem error recorded "
4567 "from previous mount: %s", errstr);
4568 ext4_warning(sb, "Marking fs in need of filesystem check.");
4569
4570 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4571 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4572 ext4_commit_super(sb, 1);
4573
4574 jbd2_journal_clear_err(journal);
4575 jbd2_journal_update_sb_errno(journal);
4576 }
4577 }
4578
4579 /*
4580 * Force the running and committing transactions to commit,
4581 * and wait on the commit.
4582 */
4583 int ext4_force_commit(struct super_block *sb)
4584 {
4585 journal_t *journal;
4586
4587 if (sb->s_flags & MS_RDONLY)
4588 return 0;
4589
4590 journal = EXT4_SB(sb)->s_journal;
4591 return ext4_journal_force_commit(journal);
4592 }
4593
4594 static int ext4_sync_fs(struct super_block *sb, int wait)
4595 {
4596 int ret = 0;
4597 tid_t target;
4598 struct ext4_sb_info *sbi = EXT4_SB(sb);
4599
4600 trace_ext4_sync_fs(sb, wait);
4601 flush_workqueue(sbi->dio_unwritten_wq);
4602 /*
4603 * Writeback quota in non-journalled quota case - journalled quota has
4604 * no dirty dquots
4605 */
4606 dquot_writeback_dquots(sb, -1);
4607 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4608 if (wait)
4609 jbd2_log_wait_commit(sbi->s_journal, target);
4610 }
4611 return ret;
4612 }
4613
4614 /*
4615 * LVM calls this function before a (read-only) snapshot is created. This
4616 * gives us a chance to flush the journal completely and mark the fs clean.
4617 *
4618 * Note that only this function cannot bring a filesystem to be in a clean
4619 * state independently. It relies on upper layer to stop all data & metadata
4620 * modifications.
4621 */
4622 static int ext4_freeze(struct super_block *sb)
4623 {
4624 int error = 0;
4625 journal_t *journal;
4626
4627 if (sb->s_flags & MS_RDONLY)
4628 return 0;
4629
4630 journal = EXT4_SB(sb)->s_journal;
4631
4632 /* Now we set up the journal barrier. */
4633 jbd2_journal_lock_updates(journal);
4634
4635 /*
4636 * Don't clear the needs_recovery flag if we failed to flush
4637 * the journal.
4638 */
4639 error = jbd2_journal_flush(journal);
4640 if (error < 0)
4641 goto out;
4642
4643 /* Journal blocked and flushed, clear needs_recovery flag. */
4644 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4645 error = ext4_commit_super(sb, 1);
4646 out:
4647 /* we rely on upper layer to stop further updates */
4648 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4649 return error;
4650 }
4651
4652 /*
4653 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4654 * flag here, even though the filesystem is not technically dirty yet.
4655 */
4656 static int ext4_unfreeze(struct super_block *sb)
4657 {
4658 if (sb->s_flags & MS_RDONLY)
4659 return 0;
4660
4661 /* Reset the needs_recovery flag before the fs is unlocked. */
4662 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4663 ext4_commit_super(sb, 1);
4664 return 0;
4665 }
4666
4667 /*
4668 * Structure to save mount options for ext4_remount's benefit
4669 */
4670 struct ext4_mount_options {
4671 unsigned long s_mount_opt;
4672 unsigned long s_mount_opt2;
4673 kuid_t s_resuid;
4674 kgid_t s_resgid;
4675 unsigned long s_commit_interval;
4676 u32 s_min_batch_time, s_max_batch_time;
4677 #ifdef CONFIG_QUOTA
4678 int s_jquota_fmt;
4679 char *s_qf_names[MAXQUOTAS];
4680 #endif
4681 };
4682
4683 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4684 {
4685 struct ext4_super_block *es;
4686 struct ext4_sb_info *sbi = EXT4_SB(sb);
4687 unsigned long old_sb_flags;
4688 struct ext4_mount_options old_opts;
4689 int enable_quota = 0;
4690 ext4_group_t g;
4691 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4692 int err = 0;
4693 #ifdef CONFIG_QUOTA
4694 int i, j;
4695 #endif
4696 char *orig_data = kstrdup(data, GFP_KERNEL);
4697
4698 /* Store the original options */
4699 old_sb_flags = sb->s_flags;
4700 old_opts.s_mount_opt = sbi->s_mount_opt;
4701 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4702 old_opts.s_resuid = sbi->s_resuid;
4703 old_opts.s_resgid = sbi->s_resgid;
4704 old_opts.s_commit_interval = sbi->s_commit_interval;
4705 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4706 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4707 #ifdef CONFIG_QUOTA
4708 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4709 for (i = 0; i < MAXQUOTAS; i++)
4710 if (sbi->s_qf_names[i]) {
4711 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4712 GFP_KERNEL);
4713 if (!old_opts.s_qf_names[i]) {
4714 for (j = 0; j < i; j++)
4715 kfree(old_opts.s_qf_names[j]);
4716 kfree(orig_data);
4717 return -ENOMEM;
4718 }
4719 } else
4720 old_opts.s_qf_names[i] = NULL;
4721 #endif
4722 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4723 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4724
4725 /*
4726 * Allow the "check" option to be passed as a remount option.
4727 */
4728 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4729 err = -EINVAL;
4730 goto restore_opts;
4731 }
4732
4733 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4734 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4735 ext4_msg(sb, KERN_ERR, "can't mount with "
4736 "both data=journal and delalloc");
4737 err = -EINVAL;
4738 goto restore_opts;
4739 }
4740 if (test_opt(sb, DIOREAD_NOLOCK)) {
4741 ext4_msg(sb, KERN_ERR, "can't mount with "
4742 "both data=journal and dioread_nolock");
4743 err = -EINVAL;
4744 goto restore_opts;
4745 }
4746 }
4747
4748 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4749 ext4_abort(sb, "Abort forced by user");
4750
4751 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4752 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4753
4754 es = sbi->s_es;
4755
4756 if (sbi->s_journal) {
4757 ext4_init_journal_params(sb, sbi->s_journal);
4758 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4759 }
4760
4761 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4762 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4763 err = -EROFS;
4764 goto restore_opts;
4765 }
4766
4767 if (*flags & MS_RDONLY) {
4768 err = dquot_suspend(sb, -1);
4769 if (err < 0)
4770 goto restore_opts;
4771
4772 /*
4773 * First of all, the unconditional stuff we have to do
4774 * to disable replay of the journal when we next remount
4775 */
4776 sb->s_flags |= MS_RDONLY;
4777
4778 /*
4779 * OK, test if we are remounting a valid rw partition
4780 * readonly, and if so set the rdonly flag and then
4781 * mark the partition as valid again.
4782 */
4783 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4784 (sbi->s_mount_state & EXT4_VALID_FS))
4785 es->s_state = cpu_to_le16(sbi->s_mount_state);
4786
4787 if (sbi->s_journal)
4788 ext4_mark_recovery_complete(sb, es);
4789 } else {
4790 /* Make sure we can mount this feature set readwrite */
4791 if (!ext4_feature_set_ok(sb, 0)) {
4792 err = -EROFS;
4793 goto restore_opts;
4794 }
4795 /*
4796 * Make sure the group descriptor checksums
4797 * are sane. If they aren't, refuse to remount r/w.
4798 */
4799 for (g = 0; g < sbi->s_groups_count; g++) {
4800 struct ext4_group_desc *gdp =
4801 ext4_get_group_desc(sb, g, NULL);
4802
4803 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4804 ext4_msg(sb, KERN_ERR,
4805 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4806 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4807 le16_to_cpu(gdp->bg_checksum));
4808 err = -EINVAL;
4809 goto restore_opts;
4810 }
4811 }
4812
4813 /*
4814 * If we have an unprocessed orphan list hanging
4815 * around from a previously readonly bdev mount,
4816 * require a full umount/remount for now.
4817 */
4818 if (es->s_last_orphan) {
4819 ext4_msg(sb, KERN_WARNING, "Couldn't "
4820 "remount RDWR because of unprocessed "
4821 "orphan inode list. Please "
4822 "umount/remount instead");
4823 err = -EINVAL;
4824 goto restore_opts;
4825 }
4826
4827 /*
4828 * Mounting a RDONLY partition read-write, so reread
4829 * and store the current valid flag. (It may have
4830 * been changed by e2fsck since we originally mounted
4831 * the partition.)
4832 */
4833 if (sbi->s_journal)
4834 ext4_clear_journal_err(sb, es);
4835 sbi->s_mount_state = le16_to_cpu(es->s_state);
4836 if (!ext4_setup_super(sb, es, 0))
4837 sb->s_flags &= ~MS_RDONLY;
4838 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4839 EXT4_FEATURE_INCOMPAT_MMP))
4840 if (ext4_multi_mount_protect(sb,
4841 le64_to_cpu(es->s_mmp_block))) {
4842 err = -EROFS;
4843 goto restore_opts;
4844 }
4845 enable_quota = 1;
4846 }
4847 }
4848
4849 /*
4850 * Reinitialize lazy itable initialization thread based on
4851 * current settings
4852 */
4853 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4854 ext4_unregister_li_request(sb);
4855 else {
4856 ext4_group_t first_not_zeroed;
4857 first_not_zeroed = ext4_has_uninit_itable(sb);
4858 ext4_register_li_request(sb, first_not_zeroed);
4859 }
4860
4861 ext4_setup_system_zone(sb);
4862 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4863 ext4_commit_super(sb, 1);
4864
4865 #ifdef CONFIG_QUOTA
4866 /* Release old quota file names */
4867 for (i = 0; i < MAXQUOTAS; i++)
4868 kfree(old_opts.s_qf_names[i]);
4869 if (enable_quota) {
4870 if (sb_any_quota_suspended(sb))
4871 dquot_resume(sb, -1);
4872 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4873 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4874 err = ext4_enable_quotas(sb);
4875 if (err)
4876 goto restore_opts;
4877 }
4878 }
4879 #endif
4880
4881 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4882 kfree(orig_data);
4883 return 0;
4884
4885 restore_opts:
4886 sb->s_flags = old_sb_flags;
4887 sbi->s_mount_opt = old_opts.s_mount_opt;
4888 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4889 sbi->s_resuid = old_opts.s_resuid;
4890 sbi->s_resgid = old_opts.s_resgid;
4891 sbi->s_commit_interval = old_opts.s_commit_interval;
4892 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4893 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4894 #ifdef CONFIG_QUOTA
4895 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4896 for (i = 0; i < MAXQUOTAS; i++) {
4897 kfree(sbi->s_qf_names[i]);
4898 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4899 }
4900 #endif
4901 kfree(orig_data);
4902 return err;
4903 }
4904
4905 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4906 {
4907 struct super_block *sb = dentry->d_sb;
4908 struct ext4_sb_info *sbi = EXT4_SB(sb);
4909 struct ext4_super_block *es = sbi->s_es;
4910 ext4_fsblk_t overhead = 0, resv_blocks;
4911 u64 fsid;
4912 s64 bfree;
4913 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4914
4915 if (!test_opt(sb, MINIX_DF))
4916 overhead = sbi->s_overhead;
4917
4918 buf->f_type = EXT4_SUPER_MAGIC;
4919 buf->f_bsize = sb->s_blocksize;
4920 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4921 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4922 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4923 /* prevent underflow in case that few free space is available */
4924 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4925 buf->f_bavail = buf->f_bfree -
4926 (ext4_r_blocks_count(es) + resv_blocks);
4927 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4928 buf->f_bavail = 0;
4929 buf->f_files = le32_to_cpu(es->s_inodes_count);
4930 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4931 buf->f_namelen = EXT4_NAME_LEN;
4932 fsid = le64_to_cpup((void *)es->s_uuid) ^
4933 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4934 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4935 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4936
4937 return 0;
4938 }
4939
4940 /* Helper function for writing quotas on sync - we need to start transaction
4941 * before quota file is locked for write. Otherwise the are possible deadlocks:
4942 * Process 1 Process 2
4943 * ext4_create() quota_sync()
4944 * jbd2_journal_start() write_dquot()
4945 * dquot_initialize() down(dqio_mutex)
4946 * down(dqio_mutex) jbd2_journal_start()
4947 *
4948 */
4949
4950 #ifdef CONFIG_QUOTA
4951
4952 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4953 {
4954 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4955 }
4956
4957 static int ext4_write_dquot(struct dquot *dquot)
4958 {
4959 int ret, err;
4960 handle_t *handle;
4961 struct inode *inode;
4962
4963 inode = dquot_to_inode(dquot);
4964 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4965 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4966 if (IS_ERR(handle))
4967 return PTR_ERR(handle);
4968 ret = dquot_commit(dquot);
4969 err = ext4_journal_stop(handle);
4970 if (!ret)
4971 ret = err;
4972 return ret;
4973 }
4974
4975 static int ext4_acquire_dquot(struct dquot *dquot)
4976 {
4977 int ret, err;
4978 handle_t *handle;
4979
4980 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4981 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4982 if (IS_ERR(handle))
4983 return PTR_ERR(handle);
4984 ret = dquot_acquire(dquot);
4985 err = ext4_journal_stop(handle);
4986 if (!ret)
4987 ret = err;
4988 return ret;
4989 }
4990
4991 static int ext4_release_dquot(struct dquot *dquot)
4992 {
4993 int ret, err;
4994 handle_t *handle;
4995
4996 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
4997 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4998 if (IS_ERR(handle)) {
4999 /* Release dquot anyway to avoid endless cycle in dqput() */
5000 dquot_release(dquot);
5001 return PTR_ERR(handle);
5002 }
5003 ret = dquot_release(dquot);
5004 err = ext4_journal_stop(handle);
5005 if (!ret)
5006 ret = err;
5007 return ret;
5008 }
5009
5010 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5011 {
5012 struct super_block *sb = dquot->dq_sb;
5013 struct ext4_sb_info *sbi = EXT4_SB(sb);
5014
5015 /* Are we journaling quotas? */
5016 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5017 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5018 dquot_mark_dquot_dirty(dquot);
5019 return ext4_write_dquot(dquot);
5020 } else {
5021 return dquot_mark_dquot_dirty(dquot);
5022 }
5023 }
5024
5025 static int ext4_write_info(struct super_block *sb, int type)
5026 {
5027 int ret, err;
5028 handle_t *handle;
5029
5030 /* Data block + inode block */
5031 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5032 if (IS_ERR(handle))
5033 return PTR_ERR(handle);
5034 ret = dquot_commit_info(sb, type);
5035 err = ext4_journal_stop(handle);
5036 if (!ret)
5037 ret = err;
5038 return ret;
5039 }
5040
5041 /*
5042 * Turn on quotas during mount time - we need to find
5043 * the quota file and such...
5044 */
5045 static int ext4_quota_on_mount(struct super_block *sb, int type)
5046 {
5047 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5048 EXT4_SB(sb)->s_jquota_fmt, type);
5049 }
5050
5051 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5052 {
5053 struct ext4_inode_info *ei = EXT4_I(inode);
5054
5055 /* The first argument of lockdep_set_subclass has to be
5056 * *exactly* the same as the argument to init_rwsem() --- in
5057 * this case, in init_once() --- or lockdep gets unhappy
5058 * because the name of the lock is set using the
5059 * stringification of the argument to init_rwsem().
5060 */
5061 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5062 lockdep_set_subclass(&ei->i_data_sem, subclass);
5063 }
5064
5065 /*
5066 * Standard function to be called on quota_on
5067 */
5068 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5069 struct path *path)
5070 {
5071 int err;
5072
5073 if (!test_opt(sb, QUOTA))
5074 return -EINVAL;
5075
5076 /* Quotafile not on the same filesystem? */
5077 if (path->dentry->d_sb != sb)
5078 return -EXDEV;
5079 /* Journaling quota? */
5080 if (EXT4_SB(sb)->s_qf_names[type]) {
5081 /* Quotafile not in fs root? */
5082 if (path->dentry->d_parent != sb->s_root)
5083 ext4_msg(sb, KERN_WARNING,
5084 "Quota file not on filesystem root. "
5085 "Journaled quota will not work");
5086 }
5087
5088 /*
5089 * When we journal data on quota file, we have to flush journal to see
5090 * all updates to the file when we bypass pagecache...
5091 */
5092 if (EXT4_SB(sb)->s_journal &&
5093 ext4_should_journal_data(path->dentry->d_inode)) {
5094 /*
5095 * We don't need to lock updates but journal_flush() could
5096 * otherwise be livelocked...
5097 */
5098 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5099 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5100 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5101 if (err)
5102 return err;
5103 }
5104 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5105 err = dquot_quota_on(sb, type, format_id, path);
5106 if (err)
5107 lockdep_set_quota_inode(path->dentry->d_inode,
5108 I_DATA_SEM_NORMAL);
5109 return err;
5110 }
5111
5112 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5113 unsigned int flags)
5114 {
5115 int err;
5116 struct inode *qf_inode;
5117 unsigned long qf_inums[MAXQUOTAS] = {
5118 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5119 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5120 };
5121
5122 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5123
5124 if (!qf_inums[type])
5125 return -EPERM;
5126
5127 qf_inode = ext4_iget(sb, qf_inums[type]);
5128 if (IS_ERR(qf_inode)) {
5129 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5130 return PTR_ERR(qf_inode);
5131 }
5132
5133 /* Don't account quota for quota files to avoid recursion */
5134 qf_inode->i_flags |= S_NOQUOTA;
5135 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5136 err = dquot_enable(qf_inode, type, format_id, flags);
5137 iput(qf_inode);
5138 if (err)
5139 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5140
5141 return err;
5142 }
5143
5144 /* Enable usage tracking for all quota types. */
5145 static int ext4_enable_quotas(struct super_block *sb)
5146 {
5147 int type, err = 0;
5148 unsigned long qf_inums[MAXQUOTAS] = {
5149 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5150 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5151 };
5152
5153 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5154 for (type = 0; type < MAXQUOTAS; type++) {
5155 if (qf_inums[type]) {
5156 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5157 DQUOT_USAGE_ENABLED);
5158 if (err) {
5159 ext4_warning(sb,
5160 "Failed to enable quota tracking "
5161 "(type=%d, err=%d). Please run "
5162 "e2fsck to fix.", type, err);
5163 return err;
5164 }
5165 }
5166 }
5167 return 0;
5168 }
5169
5170 /*
5171 * quota_on function that is used when QUOTA feature is set.
5172 */
5173 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5174 int format_id)
5175 {
5176 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5177 return -EINVAL;
5178
5179 /*
5180 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5181 */
5182 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5183 }
5184
5185 static int ext4_quota_off(struct super_block *sb, int type)
5186 {
5187 struct inode *inode = sb_dqopt(sb)->files[type];
5188 handle_t *handle;
5189
5190 /* Force all delayed allocation blocks to be allocated.
5191 * Caller already holds s_umount sem */
5192 if (test_opt(sb, DELALLOC))
5193 sync_filesystem(sb);
5194
5195 if (!inode)
5196 goto out;
5197
5198 /* Update modification times of quota files when userspace can
5199 * start looking at them */
5200 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5201 if (IS_ERR(handle))
5202 goto out;
5203 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5204 ext4_mark_inode_dirty(handle, inode);
5205 ext4_journal_stop(handle);
5206
5207 out:
5208 return dquot_quota_off(sb, type);
5209 }
5210
5211 /*
5212 * quota_off function that is used when QUOTA feature is set.
5213 */
5214 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5215 {
5216 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5217 return -EINVAL;
5218
5219 /* Disable only the limits. */
5220 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5221 }
5222
5223 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5224 * acquiring the locks... As quota files are never truncated and quota code
5225 * itself serializes the operations (and no one else should touch the files)
5226 * we don't have to be afraid of races */
5227 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5228 size_t len, loff_t off)
5229 {
5230 struct inode *inode = sb_dqopt(sb)->files[type];
5231 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5232 int err = 0;
5233 int offset = off & (sb->s_blocksize - 1);
5234 int tocopy;
5235 size_t toread;
5236 struct buffer_head *bh;
5237 loff_t i_size = i_size_read(inode);
5238
5239 if (off > i_size)
5240 return 0;
5241 if (off+len > i_size)
5242 len = i_size-off;
5243 toread = len;
5244 while (toread > 0) {
5245 tocopy = sb->s_blocksize - offset < toread ?
5246 sb->s_blocksize - offset : toread;
5247 bh = ext4_bread(NULL, inode, blk, 0, &err);
5248 if (err)
5249 return err;
5250 if (!bh) /* A hole? */
5251 memset(data, 0, tocopy);
5252 else
5253 memcpy(data, bh->b_data+offset, tocopy);
5254 brelse(bh);
5255 offset = 0;
5256 toread -= tocopy;
5257 data += tocopy;
5258 blk++;
5259 }
5260 return len;
5261 }
5262
5263 /* Write to quotafile (we know the transaction is already started and has
5264 * enough credits) */
5265 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5266 const char *data, size_t len, loff_t off)
5267 {
5268 struct inode *inode = sb_dqopt(sb)->files[type];
5269 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5270 int err = 0;
5271 int offset = off & (sb->s_blocksize - 1);
5272 struct buffer_head *bh;
5273 handle_t *handle = journal_current_handle();
5274
5275 if (EXT4_SB(sb)->s_journal && !handle) {
5276 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5277 " cancelled because transaction is not started",
5278 (unsigned long long)off, (unsigned long long)len);
5279 return -EIO;
5280 }
5281 /*
5282 * Since we account only one data block in transaction credits,
5283 * then it is impossible to cross a block boundary.
5284 */
5285 if (sb->s_blocksize - offset < len) {
5286 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5287 " cancelled because not block aligned",
5288 (unsigned long long)off, (unsigned long long)len);
5289 return -EIO;
5290 }
5291
5292 bh = ext4_bread(handle, inode, blk, 1, &err);
5293 if (!bh)
5294 goto out;
5295 err = ext4_journal_get_write_access(handle, bh);
5296 if (err) {
5297 brelse(bh);
5298 goto out;
5299 }
5300 lock_buffer(bh);
5301 memcpy(bh->b_data+offset, data, len);
5302 flush_dcache_page(bh->b_page);
5303 unlock_buffer(bh);
5304 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5305 brelse(bh);
5306 out:
5307 if (err)
5308 return err;
5309 if (inode->i_size < off + len) {
5310 i_size_write(inode, off + len);
5311 EXT4_I(inode)->i_disksize = inode->i_size;
5312 ext4_mark_inode_dirty(handle, inode);
5313 }
5314 return len;
5315 }
5316
5317 #endif
5318
5319 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5320 const char *dev_name, void *data)
5321 {
5322 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5323 }
5324
5325 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5326 static inline void register_as_ext2(void)
5327 {
5328 int err = register_filesystem(&ext2_fs_type);
5329 if (err)
5330 printk(KERN_WARNING
5331 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5332 }
5333
5334 static inline void unregister_as_ext2(void)
5335 {
5336 unregister_filesystem(&ext2_fs_type);
5337 }
5338
5339 static inline int ext2_feature_set_ok(struct super_block *sb)
5340 {
5341 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5342 return 0;
5343 if (sb->s_flags & MS_RDONLY)
5344 return 1;
5345 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5346 return 0;
5347 return 1;
5348 }
5349 #else
5350 static inline void register_as_ext2(void) { }
5351 static inline void unregister_as_ext2(void) { }
5352 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5353 #endif
5354
5355 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5356 static inline void register_as_ext3(void)
5357 {
5358 int err = register_filesystem(&ext3_fs_type);
5359 if (err)
5360 printk(KERN_WARNING
5361 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5362 }
5363
5364 static inline void unregister_as_ext3(void)
5365 {
5366 unregister_filesystem(&ext3_fs_type);
5367 }
5368
5369 static inline int ext3_feature_set_ok(struct super_block *sb)
5370 {
5371 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5372 return 0;
5373 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5374 return 0;
5375 if (sb->s_flags & MS_RDONLY)
5376 return 1;
5377 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5378 return 0;
5379 return 1;
5380 }
5381 #else
5382 static inline void register_as_ext3(void) { }
5383 static inline void unregister_as_ext3(void) { }
5384 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5385 #endif
5386
5387 static struct file_system_type ext4_fs_type = {
5388 .owner = THIS_MODULE,
5389 .name = "ext4",
5390 .mount = ext4_mount,
5391 .kill_sb = kill_block_super,
5392 .fs_flags = FS_REQUIRES_DEV,
5393 };
5394 MODULE_ALIAS_FS("ext4");
5395
5396 static int __init ext4_init_feat_adverts(void)
5397 {
5398 struct ext4_features *ef;
5399 int ret = -ENOMEM;
5400
5401 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5402 if (!ef)
5403 goto out;
5404
5405 ef->f_kobj.kset = ext4_kset;
5406 init_completion(&ef->f_kobj_unregister);
5407 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5408 "features");
5409 if (ret) {
5410 kfree(ef);
5411 goto out;
5412 }
5413
5414 ext4_feat = ef;
5415 ret = 0;
5416 out:
5417 return ret;
5418 }
5419
5420 static void ext4_exit_feat_adverts(void)
5421 {
5422 kobject_put(&ext4_feat->f_kobj);
5423 wait_for_completion(&ext4_feat->f_kobj_unregister);
5424 kfree(ext4_feat);
5425 }
5426
5427 /* Shared across all ext4 file systems */
5428 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5429 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5430
5431 static int __init ext4_init_fs(void)
5432 {
5433 int i, err;
5434
5435 ext4_li_info = NULL;
5436 mutex_init(&ext4_li_mtx);
5437
5438 /* Build-time check for flags consistency */
5439 ext4_check_flag_values();
5440
5441 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5442 mutex_init(&ext4__aio_mutex[i]);
5443 init_waitqueue_head(&ext4__ioend_wq[i]);
5444 }
5445
5446 err = ext4_init_es();
5447 if (err)
5448 return err;
5449
5450 err = ext4_init_pageio();
5451 if (err)
5452 goto out7;
5453
5454 err = ext4_init_system_zone();
5455 if (err)
5456 goto out6;
5457 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5458 if (!ext4_kset) {
5459 err = -ENOMEM;
5460 goto out5;
5461 }
5462 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5463
5464 err = ext4_init_feat_adverts();
5465 if (err)
5466 goto out4;
5467
5468 err = ext4_init_mballoc();
5469 if (err)
5470 goto out3;
5471
5472 err = ext4_init_xattr();
5473 if (err)
5474 goto out2;
5475 err = init_inodecache();
5476 if (err)
5477 goto out1;
5478 register_as_ext3();
5479 register_as_ext2();
5480 err = register_filesystem(&ext4_fs_type);
5481 if (err)
5482 goto out;
5483
5484 return 0;
5485 out:
5486 unregister_as_ext2();
5487 unregister_as_ext3();
5488 destroy_inodecache();
5489 out1:
5490 ext4_exit_xattr();
5491 out2:
5492 ext4_exit_mballoc();
5493 out3:
5494 ext4_exit_feat_adverts();
5495 out4:
5496 if (ext4_proc_root)
5497 remove_proc_entry("fs/ext4", NULL);
5498 kset_unregister(ext4_kset);
5499 out5:
5500 ext4_exit_system_zone();
5501 out6:
5502 ext4_exit_pageio();
5503 out7:
5504 ext4_exit_es();
5505
5506 return err;
5507 }
5508
5509 static void __exit ext4_exit_fs(void)
5510 {
5511 ext4_destroy_lazyinit_thread();
5512 unregister_as_ext2();
5513 unregister_as_ext3();
5514 unregister_filesystem(&ext4_fs_type);
5515 destroy_inodecache();
5516 ext4_exit_xattr();
5517 ext4_exit_mballoc();
5518 ext4_exit_feat_adverts();
5519 remove_proc_entry("fs/ext4", NULL);
5520 kset_unregister(ext4_kset);
5521 ext4_exit_system_zone();
5522 ext4_exit_pageio();
5523 ext4_exit_es();
5524 }
5525
5526 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5527 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5528 MODULE_LICENSE("GPL");
5529 module_init(ext4_init_fs)
5530 module_exit(ext4_exit_fs)
kernel source for telekom puls (alcatel/mediatek)