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