]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - fs/ext4/super.c
projects / mirror_ubuntu-zesty-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Revert "ext4: remove block_device_ejected"
[mirror_ubuntu-zesty-kernel.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/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.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 static int ext4_mballoc_ready;
63 static struct ratelimit_state ext4_mount_msg_ratelimit;
64
65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
66 unsigned long journal_devnum);
67 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
68 static int ext4_commit_super(struct super_block *sb, int sync);
69 static void ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71 static void ext4_clear_journal_err(struct super_block *sb,
72 struct ext4_super_block *es);
73 static int ext4_sync_fs(struct super_block *sb, int wait);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
87
88 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89 static struct file_system_type ext2_fs_type = {
90 .owner = THIS_MODULE,
91 .name = "ext2",
92 .mount = ext4_mount,
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
95 };
96 MODULE_ALIAS_FS("ext2");
97 MODULE_ALIAS("ext2");
98 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
99 #else
100 #define IS_EXT2_SB(sb) (0)
101 #endif
102
103
104 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
105 static struct file_system_type ext3_fs_type = {
106 .owner = THIS_MODULE,
107 .name = "ext3",
108 .mount = ext4_mount,
109 .kill_sb = kill_block_super,
110 .fs_flags = FS_REQUIRES_DEV,
111 };
112 MODULE_ALIAS_FS("ext3");
113 MODULE_ALIAS("ext3");
114 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
115 #else
116 #define IS_EXT3_SB(sb) (0)
117 #endif
118
119 static int ext4_verify_csum_type(struct super_block *sb,
120 struct ext4_super_block *es)
121 {
122 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
123 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
124 return 1;
125
126 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
127 }
128
129 static __le32 ext4_superblock_csum(struct super_block *sb,
130 struct ext4_super_block *es)
131 {
132 struct ext4_sb_info *sbi = EXT4_SB(sb);
133 int offset = offsetof(struct ext4_super_block, s_checksum);
134 __u32 csum;
135
136 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
137
138 return cpu_to_le32(csum);
139 }
140
141 static int ext4_superblock_csum_verify(struct super_block *sb,
142 struct ext4_super_block *es)
143 {
144 if (!ext4_has_metadata_csum(sb))
145 return 1;
146
147 return es->s_checksum == ext4_superblock_csum(sb, es);
148 }
149
150 void ext4_superblock_csum_set(struct super_block *sb)
151 {
152 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
153
154 if (!ext4_has_metadata_csum(sb))
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 | __GFP_NOWARN);
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 | __GFP_NOWARN);
175 if (!ret)
176 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
177 return ret;
178 }
179
180 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
181 struct ext4_group_desc *bg)
182 {
183 return le32_to_cpu(bg->bg_block_bitmap_lo) |
184 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
185 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
186 }
187
188 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
189 struct ext4_group_desc *bg)
190 {
191 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
192 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
193 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
194 }
195
196 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
197 struct ext4_group_desc *bg)
198 {
199 return le32_to_cpu(bg->bg_inode_table_lo) |
200 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
201 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
202 }
203
204 __u32 ext4_free_group_clusters(struct super_block *sb,
205 struct ext4_group_desc *bg)
206 {
207 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
208 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
209 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
210 }
211
212 __u32 ext4_free_inodes_count(struct super_block *sb,
213 struct ext4_group_desc *bg)
214 {
215 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
216 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
217 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
218 }
219
220 __u32 ext4_used_dirs_count(struct super_block *sb,
221 struct ext4_group_desc *bg)
222 {
223 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
224 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
225 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
226 }
227
228 __u32 ext4_itable_unused_count(struct super_block *sb,
229 struct ext4_group_desc *bg)
230 {
231 return le16_to_cpu(bg->bg_itable_unused_lo) |
232 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
233 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
234 }
235
236 void ext4_block_bitmap_set(struct super_block *sb,
237 struct ext4_group_desc *bg, ext4_fsblk_t blk)
238 {
239 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
240 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
241 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
242 }
243
244 void ext4_inode_bitmap_set(struct super_block *sb,
245 struct ext4_group_desc *bg, ext4_fsblk_t blk)
246 {
247 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
248 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
249 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
250 }
251
252 void ext4_inode_table_set(struct super_block *sb,
253 struct ext4_group_desc *bg, ext4_fsblk_t blk)
254 {
255 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
256 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
257 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
258 }
259
260 void ext4_free_group_clusters_set(struct super_block *sb,
261 struct ext4_group_desc *bg, __u32 count)
262 {
263 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
264 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
265 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
266 }
267
268 void ext4_free_inodes_set(struct super_block *sb,
269 struct ext4_group_desc *bg, __u32 count)
270 {
271 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
272 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
273 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
274 }
275
276 void ext4_used_dirs_set(struct super_block *sb,
277 struct ext4_group_desc *bg, __u32 count)
278 {
279 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
280 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
281 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
282 }
283
284 void ext4_itable_unused_set(struct super_block *sb,
285 struct ext4_group_desc *bg, __u32 count)
286 {
287 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
288 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
289 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
290 }
291
292
293 static void __save_error_info(struct super_block *sb, const char *func,
294 unsigned int line)
295 {
296 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
297
298 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
299 if (bdev_read_only(sb->s_bdev))
300 return;
301 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
302 es->s_last_error_time = cpu_to_le32(get_seconds());
303 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
304 es->s_last_error_line = cpu_to_le32(line);
305 if (!es->s_first_error_time) {
306 es->s_first_error_time = es->s_last_error_time;
307 strncpy(es->s_first_error_func, func,
308 sizeof(es->s_first_error_func));
309 es->s_first_error_line = cpu_to_le32(line);
310 es->s_first_error_ino = es->s_last_error_ino;
311 es->s_first_error_block = es->s_last_error_block;
312 }
313 /*
314 * Start the daily error reporting function if it hasn't been
315 * started already
316 */
317 if (!es->s_error_count)
318 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
319 le32_add_cpu(&es->s_error_count, 1);
320 }
321
322 static void save_error_info(struct super_block *sb, const char *func,
323 unsigned int line)
324 {
325 __save_error_info(sb, func, line);
326 ext4_commit_super(sb, 1);
327 }
328
329 /*
330 * The del_gendisk() function uninitializes the disk-specific data
331 * structures, including the bdi structure, without telling anyone
332 * else. Once this happens, any attempt to call mark_buffer_dirty()
333 * (for example, by ext4_commit_super), will cause a kernel OOPS.
334 * This is a kludge to prevent these oops until we can put in a proper
335 * hook in del_gendisk() to inform the VFS and file system layers.
336 */
337 static int block_device_ejected(struct super_block *sb)
338 {
339 struct inode *bd_inode = sb->s_bdev->bd_inode;
340 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
341
342 return bdi->dev == NULL;
343 }
344
345 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
346 {
347 struct super_block *sb = journal->j_private;
348 struct ext4_sb_info *sbi = EXT4_SB(sb);
349 int error = is_journal_aborted(journal);
350 struct ext4_journal_cb_entry *jce;
351
352 BUG_ON(txn->t_state == T_FINISHED);
353 spin_lock(&sbi->s_md_lock);
354 while (!list_empty(&txn->t_private_list)) {
355 jce = list_entry(txn->t_private_list.next,
356 struct ext4_journal_cb_entry, jce_list);
357 list_del_init(&jce->jce_list);
358 spin_unlock(&sbi->s_md_lock);
359 jce->jce_func(sb, jce, error);
360 spin_lock(&sbi->s_md_lock);
361 }
362 spin_unlock(&sbi->s_md_lock);
363 }
364
365 /* Deal with the reporting of failure conditions on a filesystem such as
366 * inconsistencies detected or read IO failures.
367 *
368 * On ext2, we can store the error state of the filesystem in the
369 * superblock. That is not possible on ext4, because we may have other
370 * write ordering constraints on the superblock which prevent us from
371 * writing it out straight away; and given that the journal is about to
372 * be aborted, we can't rely on the current, or future, transactions to
373 * write out the superblock safely.
374 *
375 * We'll just use the jbd2_journal_abort() error code to record an error in
376 * the journal instead. On recovery, the journal will complain about
377 * that error until we've noted it down and cleared it.
378 */
379
380 static void ext4_handle_error(struct super_block *sb)
381 {
382 if (sb->s_flags & MS_RDONLY)
383 return;
384
385 if (!test_opt(sb, ERRORS_CONT)) {
386 journal_t *journal = EXT4_SB(sb)->s_journal;
387
388 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
389 if (journal)
390 jbd2_journal_abort(journal, -EIO);
391 }
392 if (test_opt(sb, ERRORS_RO)) {
393 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
394 /*
395 * Make sure updated value of ->s_mount_flags will be visible
396 * before ->s_flags update
397 */
398 smp_wmb();
399 sb->s_flags |= MS_RDONLY;
400 }
401 if (test_opt(sb, ERRORS_PANIC))
402 panic("EXT4-fs (device %s): panic forced after error\n",
403 sb->s_id);
404 }
405
406 #define ext4_error_ratelimit(sb) \
407 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
408 "EXT4-fs error")
409
410 void __ext4_error(struct super_block *sb, const char *function,
411 unsigned int line, const char *fmt, ...)
412 {
413 struct va_format vaf;
414 va_list args;
415
416 if (ext4_error_ratelimit(sb)) {
417 va_start(args, fmt);
418 vaf.fmt = fmt;
419 vaf.va = &args;
420 printk(KERN_CRIT
421 "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 }
425 save_error_info(sb, function, line);
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 if (ext4_error_ratelimit(inode->i_sb)) {
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 save_error_info(inode->i_sb, function, line);
456 ext4_handle_error(inode->i_sb);
457 }
458
459 void __ext4_error_file(struct file *file, const char *function,
460 unsigned int line, ext4_fsblk_t block,
461 const char *fmt, ...)
462 {
463 va_list args;
464 struct va_format vaf;
465 struct ext4_super_block *es;
466 struct inode *inode = file_inode(file);
467 char pathname[80], *path;
468
469 es = EXT4_SB(inode->i_sb)->s_es;
470 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
471 if (ext4_error_ratelimit(inode->i_sb)) {
472 path = file_path(file, pathname, sizeof(pathname));
473 if (IS_ERR(path))
474 path = "(unknown)";
475 va_start(args, fmt);
476 vaf.fmt = fmt;
477 vaf.va = &args;
478 if (block)
479 printk(KERN_CRIT
480 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
481 "block %llu: comm %s: path %s: %pV\n",
482 inode->i_sb->s_id, function, line, inode->i_ino,
483 block, current->comm, path, &vaf);
484 else
485 printk(KERN_CRIT
486 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
487 "comm %s: path %s: %pV\n",
488 inode->i_sb->s_id, function, line, inode->i_ino,
489 current->comm, path, &vaf);
490 va_end(args);
491 }
492 save_error_info(inode->i_sb, function, line);
493 ext4_handle_error(inode->i_sb);
494 }
495
496 const char *ext4_decode_error(struct super_block *sb, int errno,
497 char nbuf[16])
498 {
499 char *errstr = NULL;
500
501 switch (errno) {
502 case -EIO:
503 errstr = "IO failure";
504 break;
505 case -ENOMEM:
506 errstr = "Out of memory";
507 break;
508 case -EROFS:
509 if (!sb || (EXT4_SB(sb)->s_journal &&
510 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
511 errstr = "Journal has aborted";
512 else
513 errstr = "Readonly filesystem";
514 break;
515 default:
516 /* If the caller passed in an extra buffer for unknown
517 * errors, textualise them now. Else we just return
518 * NULL. */
519 if (nbuf) {
520 /* Check for truncated error codes... */
521 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
522 errstr = nbuf;
523 }
524 break;
525 }
526
527 return errstr;
528 }
529
530 /* __ext4_std_error decodes expected errors from journaling functions
531 * automatically and invokes the appropriate error response. */
532
533 void __ext4_std_error(struct super_block *sb, const char *function,
534 unsigned int line, int errno)
535 {
536 char nbuf[16];
537 const char *errstr;
538
539 /* Special case: if the error is EROFS, and we're not already
540 * inside a transaction, then there's really no point in logging
541 * an error. */
542 if (errno == -EROFS && journal_current_handle() == NULL &&
543 (sb->s_flags & MS_RDONLY))
544 return;
545
546 if (ext4_error_ratelimit(sb)) {
547 errstr = ext4_decode_error(sb, errno, nbuf);
548 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
549 sb->s_id, function, line, errstr);
550 }
551
552 save_error_info(sb, function, line);
553 ext4_handle_error(sb);
554 }
555
556 /*
557 * ext4_abort is a much stronger failure handler than ext4_error. The
558 * abort function may be used to deal with unrecoverable failures such
559 * as journal IO errors or ENOMEM at a critical moment in log management.
560 *
561 * We unconditionally force the filesystem into an ABORT|READONLY state,
562 * unless the error response on the fs has been set to panic in which
563 * case we take the easy way out and panic immediately.
564 */
565
566 void __ext4_abort(struct super_block *sb, const char *function,
567 unsigned int line, const char *fmt, ...)
568 {
569 va_list args;
570
571 save_error_info(sb, function, line);
572 va_start(args, fmt);
573 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
574 function, line);
575 vprintk(fmt, args);
576 printk("\n");
577 va_end(args);
578
579 if ((sb->s_flags & MS_RDONLY) == 0) {
580 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
581 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
582 /*
583 * Make sure updated value of ->s_mount_flags will be visible
584 * before ->s_flags update
585 */
586 smp_wmb();
587 sb->s_flags |= MS_RDONLY;
588 if (EXT4_SB(sb)->s_journal)
589 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
590 save_error_info(sb, function, line);
591 }
592 if (test_opt(sb, ERRORS_PANIC))
593 panic("EXT4-fs panic from previous error\n");
594 }
595
596 void __ext4_msg(struct super_block *sb,
597 const char *prefix, const char *fmt, ...)
598 {
599 struct va_format vaf;
600 va_list args;
601
602 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
603 return;
604
605 va_start(args, fmt);
606 vaf.fmt = fmt;
607 vaf.va = &args;
608 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
609 va_end(args);
610 }
611
612 #define ext4_warning_ratelimit(sb) \
613 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
614 "EXT4-fs warning")
615
616 void __ext4_warning(struct super_block *sb, const char *function,
617 unsigned int line, const char *fmt, ...)
618 {
619 struct va_format vaf;
620 va_list args;
621
622 if (!ext4_warning_ratelimit(sb))
623 return;
624
625 va_start(args, fmt);
626 vaf.fmt = fmt;
627 vaf.va = &args;
628 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
629 sb->s_id, function, line, &vaf);
630 va_end(args);
631 }
632
633 void __ext4_warning_inode(const struct inode *inode, const char *function,
634 unsigned int line, const char *fmt, ...)
635 {
636 struct va_format vaf;
637 va_list args;
638
639 if (!ext4_warning_ratelimit(inode->i_sb))
640 return;
641
642 va_start(args, fmt);
643 vaf.fmt = fmt;
644 vaf.va = &args;
645 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
646 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
647 function, line, inode->i_ino, current->comm, &vaf);
648 va_end(args);
649 }
650
651 void __ext4_grp_locked_error(const char *function, unsigned int line,
652 struct super_block *sb, ext4_group_t grp,
653 unsigned long ino, ext4_fsblk_t block,
654 const char *fmt, ...)
655 __releases(bitlock)
656 __acquires(bitlock)
657 {
658 struct va_format vaf;
659 va_list args;
660 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
661
662 es->s_last_error_ino = cpu_to_le32(ino);
663 es->s_last_error_block = cpu_to_le64(block);
664 __save_error_info(sb, function, line);
665
666 if (ext4_error_ratelimit(sb)) {
667 va_start(args, fmt);
668 vaf.fmt = fmt;
669 vaf.va = &args;
670 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
671 sb->s_id, function, line, grp);
672 if (ino)
673 printk(KERN_CONT "inode %lu: ", ino);
674 if (block)
675 printk(KERN_CONT "block %llu:",
676 (unsigned long long) block);
677 printk(KERN_CONT "%pV\n", &vaf);
678 va_end(args);
679 }
680
681 if (test_opt(sb, ERRORS_CONT)) {
682 ext4_commit_super(sb, 0);
683 return;
684 }
685
686 ext4_unlock_group(sb, grp);
687 ext4_handle_error(sb);
688 /*
689 * We only get here in the ERRORS_RO case; relocking the group
690 * may be dangerous, but nothing bad will happen since the
691 * filesystem will have already been marked read/only and the
692 * journal has been aborted. We return 1 as a hint to callers
693 * who might what to use the return value from
694 * ext4_grp_locked_error() to distinguish between the
695 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
696 * aggressively from the ext4 function in question, with a
697 * more appropriate error code.
698 */
699 ext4_lock_group(sb, grp);
700 return;
701 }
702
703 void ext4_update_dynamic_rev(struct super_block *sb)
704 {
705 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
706
707 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
708 return;
709
710 ext4_warning(sb,
711 "updating to rev %d because of new feature flag, "
712 "running e2fsck is recommended",
713 EXT4_DYNAMIC_REV);
714
715 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
716 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
717 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
718 /* leave es->s_feature_*compat flags alone */
719 /* es->s_uuid will be set by e2fsck if empty */
720
721 /*
722 * The rest of the superblock fields should be zero, and if not it
723 * means they are likely already in use, so leave them alone. We
724 * can leave it up to e2fsck to clean up any inconsistencies there.
725 */
726 }
727
728 /*
729 * Open the external journal device
730 */
731 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
732 {
733 struct block_device *bdev;
734 char b[BDEVNAME_SIZE];
735
736 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
737 if (IS_ERR(bdev))
738 goto fail;
739 return bdev;
740
741 fail:
742 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
743 __bdevname(dev, b), PTR_ERR(bdev));
744 return NULL;
745 }
746
747 /*
748 * Release the journal device
749 */
750 static void ext4_blkdev_put(struct block_device *bdev)
751 {
752 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
753 }
754
755 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
756 {
757 struct block_device *bdev;
758 bdev = sbi->journal_bdev;
759 if (bdev) {
760 ext4_blkdev_put(bdev);
761 sbi->journal_bdev = NULL;
762 }
763 }
764
765 static inline struct inode *orphan_list_entry(struct list_head *l)
766 {
767 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
768 }
769
770 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
771 {
772 struct list_head *l;
773
774 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
775 le32_to_cpu(sbi->s_es->s_last_orphan));
776
777 printk(KERN_ERR "sb_info orphan list:\n");
778 list_for_each(l, &sbi->s_orphan) {
779 struct inode *inode = orphan_list_entry(l);
780 printk(KERN_ERR " "
781 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
782 inode->i_sb->s_id, inode->i_ino, inode,
783 inode->i_mode, inode->i_nlink,
784 NEXT_ORPHAN(inode));
785 }
786 }
787
788 static void ext4_put_super(struct super_block *sb)
789 {
790 struct ext4_sb_info *sbi = EXT4_SB(sb);
791 struct ext4_super_block *es = sbi->s_es;
792 int i, err;
793
794 ext4_unregister_li_request(sb);
795 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
796
797 flush_workqueue(sbi->rsv_conversion_wq);
798 destroy_workqueue(sbi->rsv_conversion_wq);
799
800 if (sbi->s_journal) {
801 err = jbd2_journal_destroy(sbi->s_journal);
802 sbi->s_journal = NULL;
803 if (err < 0)
804 ext4_abort(sb, "Couldn't clean up the journal");
805 }
806
807 ext4_es_unregister_shrinker(sbi);
808 del_timer_sync(&sbi->s_err_report);
809 ext4_release_system_zone(sb);
810 ext4_mb_release(sb);
811 ext4_ext_release(sb);
812 ext4_xattr_put_super(sb);
813
814 if (!(sb->s_flags & MS_RDONLY)) {
815 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
816 es->s_state = cpu_to_le16(sbi->s_mount_state);
817 }
818 if (!(sb->s_flags & MS_RDONLY))
819 ext4_commit_super(sb, 1);
820
821 if (sbi->s_proc) {
822 remove_proc_entry("options", sbi->s_proc);
823 remove_proc_entry(sb->s_id, ext4_proc_root);
824 }
825 kobject_del(&sbi->s_kobj);
826
827 for (i = 0; i < sbi->s_gdb_count; i++)
828 brelse(sbi->s_group_desc[i]);
829 kvfree(sbi->s_group_desc);
830 kvfree(sbi->s_flex_groups);
831 percpu_counter_destroy(&sbi->s_freeclusters_counter);
832 percpu_counter_destroy(&sbi->s_freeinodes_counter);
833 percpu_counter_destroy(&sbi->s_dirs_counter);
834 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
835 brelse(sbi->s_sbh);
836 #ifdef CONFIG_QUOTA
837 for (i = 0; i < EXT4_MAXQUOTAS; i++)
838 kfree(sbi->s_qf_names[i]);
839 #endif
840
841 /* Debugging code just in case the in-memory inode orphan list
842 * isn't empty. The on-disk one can be non-empty if we've
843 * detected an error and taken the fs readonly, but the
844 * in-memory list had better be clean by this point. */
845 if (!list_empty(&sbi->s_orphan))
846 dump_orphan_list(sb, sbi);
847 J_ASSERT(list_empty(&sbi->s_orphan));
848
849 sync_blockdev(sb->s_bdev);
850 invalidate_bdev(sb->s_bdev);
851 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
852 /*
853 * Invalidate the journal device's buffers. We don't want them
854 * floating about in memory - the physical journal device may
855 * hotswapped, and it breaks the `ro-after' testing code.
856 */
857 sync_blockdev(sbi->journal_bdev);
858 invalidate_bdev(sbi->journal_bdev);
859 ext4_blkdev_remove(sbi);
860 }
861 if (sbi->s_mb_cache) {
862 ext4_xattr_destroy_cache(sbi->s_mb_cache);
863 sbi->s_mb_cache = NULL;
864 }
865 if (sbi->s_mmp_tsk)
866 kthread_stop(sbi->s_mmp_tsk);
867 sb->s_fs_info = NULL;
868 /*
869 * Now that we are completely done shutting down the
870 * superblock, we need to actually destroy the kobject.
871 */
872 kobject_put(&sbi->s_kobj);
873 wait_for_completion(&sbi->s_kobj_unregister);
874 if (sbi->s_chksum_driver)
875 crypto_free_shash(sbi->s_chksum_driver);
876 kfree(sbi->s_blockgroup_lock);
877 kfree(sbi);
878 }
879
880 static struct kmem_cache *ext4_inode_cachep;
881
882 /*
883 * Called inside transaction, so use GFP_NOFS
884 */
885 static struct inode *ext4_alloc_inode(struct super_block *sb)
886 {
887 struct ext4_inode_info *ei;
888
889 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
890 if (!ei)
891 return NULL;
892
893 ei->vfs_inode.i_version = 1;
894 spin_lock_init(&ei->i_raw_lock);
895 INIT_LIST_HEAD(&ei->i_prealloc_list);
896 spin_lock_init(&ei->i_prealloc_lock);
897 ext4_es_init_tree(&ei->i_es_tree);
898 rwlock_init(&ei->i_es_lock);
899 INIT_LIST_HEAD(&ei->i_es_list);
900 ei->i_es_all_nr = 0;
901 ei->i_es_shk_nr = 0;
902 ei->i_es_shrink_lblk = 0;
903 ei->i_reserved_data_blocks = 0;
904 ei->i_reserved_meta_blocks = 0;
905 ei->i_allocated_meta_blocks = 0;
906 ei->i_da_metadata_calc_len = 0;
907 ei->i_da_metadata_calc_last_lblock = 0;
908 spin_lock_init(&(ei->i_block_reservation_lock));
909 #ifdef CONFIG_QUOTA
910 ei->i_reserved_quota = 0;
911 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
912 #endif
913 ei->jinode = NULL;
914 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
915 spin_lock_init(&ei->i_completed_io_lock);
916 ei->i_sync_tid = 0;
917 ei->i_datasync_tid = 0;
918 atomic_set(&ei->i_ioend_count, 0);
919 atomic_set(&ei->i_unwritten, 0);
920 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
921 #ifdef CONFIG_EXT4_FS_ENCRYPTION
922 ei->i_crypt_info = NULL;
923 #endif
924 return &ei->vfs_inode;
925 }
926
927 static int ext4_drop_inode(struct inode *inode)
928 {
929 int drop = generic_drop_inode(inode);
930
931 trace_ext4_drop_inode(inode, drop);
932 return drop;
933 }
934
935 static void ext4_i_callback(struct rcu_head *head)
936 {
937 struct inode *inode = container_of(head, struct inode, i_rcu);
938 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
939 }
940
941 static void ext4_destroy_inode(struct inode *inode)
942 {
943 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
944 ext4_msg(inode->i_sb, KERN_ERR,
945 "Inode %lu (%p): orphan list check failed!",
946 inode->i_ino, EXT4_I(inode));
947 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
948 EXT4_I(inode), sizeof(struct ext4_inode_info),
949 true);
950 dump_stack();
951 }
952 call_rcu(&inode->i_rcu, ext4_i_callback);
953 }
954
955 static void init_once(void *foo)
956 {
957 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
958
959 INIT_LIST_HEAD(&ei->i_orphan);
960 init_rwsem(&ei->xattr_sem);
961 init_rwsem(&ei->i_data_sem);
962 inode_init_once(&ei->vfs_inode);
963 }
964
965 static int __init init_inodecache(void)
966 {
967 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
968 sizeof(struct ext4_inode_info),
969 0, (SLAB_RECLAIM_ACCOUNT|
970 SLAB_MEM_SPREAD),
971 init_once);
972 if (ext4_inode_cachep == NULL)
973 return -ENOMEM;
974 return 0;
975 }
976
977 static void destroy_inodecache(void)
978 {
979 /*
980 * Make sure all delayed rcu free inodes are flushed before we
981 * destroy cache.
982 */
983 rcu_barrier();
984 kmem_cache_destroy(ext4_inode_cachep);
985 }
986
987 void ext4_clear_inode(struct inode *inode)
988 {
989 invalidate_inode_buffers(inode);
990 clear_inode(inode);
991 dquot_drop(inode);
992 ext4_discard_preallocations(inode);
993 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
994 if (EXT4_I(inode)->jinode) {
995 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
996 EXT4_I(inode)->jinode);
997 jbd2_free_inode(EXT4_I(inode)->jinode);
998 EXT4_I(inode)->jinode = NULL;
999 }
1000 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1001 if (EXT4_I(inode)->i_crypt_info)
1002 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
1003 #endif
1004 }
1005
1006 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1007 u64 ino, u32 generation)
1008 {
1009 struct inode *inode;
1010
1011 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1012 return ERR_PTR(-ESTALE);
1013 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1014 return ERR_PTR(-ESTALE);
1015
1016 /* iget isn't really right if the inode is currently unallocated!!
1017 *
1018 * ext4_read_inode will return a bad_inode if the inode had been
1019 * deleted, so we should be safe.
1020 *
1021 * Currently we don't know the generation for parent directory, so
1022 * a generation of 0 means "accept any"
1023 */
1024 inode = ext4_iget_normal(sb, ino);
1025 if (IS_ERR(inode))
1026 return ERR_CAST(inode);
1027 if (generation && inode->i_generation != generation) {
1028 iput(inode);
1029 return ERR_PTR(-ESTALE);
1030 }
1031
1032 return inode;
1033 }
1034
1035 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1036 int fh_len, int fh_type)
1037 {
1038 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1039 ext4_nfs_get_inode);
1040 }
1041
1042 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1043 int fh_len, int fh_type)
1044 {
1045 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1046 ext4_nfs_get_inode);
1047 }
1048
1049 /*
1050 * Try to release metadata pages (indirect blocks, directories) which are
1051 * mapped via the block device. Since these pages could have journal heads
1052 * which would prevent try_to_free_buffers() from freeing them, we must use
1053 * jbd2 layer's try_to_free_buffers() function to release them.
1054 */
1055 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1056 gfp_t wait)
1057 {
1058 journal_t *journal = EXT4_SB(sb)->s_journal;
1059
1060 WARN_ON(PageChecked(page));
1061 if (!page_has_buffers(page))
1062 return 0;
1063 if (journal)
1064 return jbd2_journal_try_to_free_buffers(journal, page,
1065 wait & ~__GFP_WAIT);
1066 return try_to_free_buffers(page);
1067 }
1068
1069 #ifdef CONFIG_QUOTA
1070 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1071 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1072
1073 static int ext4_write_dquot(struct dquot *dquot);
1074 static int ext4_acquire_dquot(struct dquot *dquot);
1075 static int ext4_release_dquot(struct dquot *dquot);
1076 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1077 static int ext4_write_info(struct super_block *sb, int type);
1078 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1079 struct path *path);
1080 static int ext4_quota_off(struct super_block *sb, int type);
1081 static int ext4_quota_on_mount(struct super_block *sb, int type);
1082 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1083 size_t len, loff_t off);
1084 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1085 const char *data, size_t len, loff_t off);
1086 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1087 unsigned int flags);
1088 static int ext4_enable_quotas(struct super_block *sb);
1089
1090 static struct dquot **ext4_get_dquots(struct inode *inode)
1091 {
1092 return EXT4_I(inode)->i_dquot;
1093 }
1094
1095 static const struct dquot_operations ext4_quota_operations = {
1096 .get_reserved_space = ext4_get_reserved_space,
1097 .write_dquot = ext4_write_dquot,
1098 .acquire_dquot = ext4_acquire_dquot,
1099 .release_dquot = ext4_release_dquot,
1100 .mark_dirty = ext4_mark_dquot_dirty,
1101 .write_info = ext4_write_info,
1102 .alloc_dquot = dquot_alloc,
1103 .destroy_dquot = dquot_destroy,
1104 };
1105
1106 static const struct quotactl_ops ext4_qctl_operations = {
1107 .quota_on = ext4_quota_on,
1108 .quota_off = ext4_quota_off,
1109 .quota_sync = dquot_quota_sync,
1110 .get_state = dquot_get_state,
1111 .set_info = dquot_set_dqinfo,
1112 .get_dqblk = dquot_get_dqblk,
1113 .set_dqblk = dquot_set_dqblk
1114 };
1115 #endif
1116
1117 static const struct super_operations ext4_sops = {
1118 .alloc_inode = ext4_alloc_inode,
1119 .destroy_inode = ext4_destroy_inode,
1120 .write_inode = ext4_write_inode,
1121 .dirty_inode = ext4_dirty_inode,
1122 .drop_inode = ext4_drop_inode,
1123 .evict_inode = ext4_evict_inode,
1124 .put_super = ext4_put_super,
1125 .sync_fs = ext4_sync_fs,
1126 .freeze_fs = ext4_freeze,
1127 .unfreeze_fs = ext4_unfreeze,
1128 .statfs = ext4_statfs,
1129 .remount_fs = ext4_remount,
1130 .show_options = ext4_show_options,
1131 #ifdef CONFIG_QUOTA
1132 .quota_read = ext4_quota_read,
1133 .quota_write = ext4_quota_write,
1134 .get_dquots = ext4_get_dquots,
1135 #endif
1136 .bdev_try_to_free_page = bdev_try_to_free_page,
1137 };
1138
1139 static const struct export_operations ext4_export_ops = {
1140 .fh_to_dentry = ext4_fh_to_dentry,
1141 .fh_to_parent = ext4_fh_to_parent,
1142 .get_parent = ext4_get_parent,
1143 };
1144
1145 enum {
1146 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1147 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1148 Opt_nouid32, Opt_debug, Opt_removed,
1149 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1150 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1151 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1152 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1153 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1154 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1155 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1156 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1157 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1158 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1159 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1160 Opt_lazytime, Opt_nolazytime,
1161 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1162 Opt_inode_readahead_blks, Opt_journal_ioprio,
1163 Opt_dioread_nolock, Opt_dioread_lock,
1164 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1165 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1166 };
1167
1168 static const match_table_t tokens = {
1169 {Opt_bsd_df, "bsddf"},
1170 {Opt_minix_df, "minixdf"},
1171 {Opt_grpid, "grpid"},
1172 {Opt_grpid, "bsdgroups"},
1173 {Opt_nogrpid, "nogrpid"},
1174 {Opt_nogrpid, "sysvgroups"},
1175 {Opt_resgid, "resgid=%u"},
1176 {Opt_resuid, "resuid=%u"},
1177 {Opt_sb, "sb=%u"},
1178 {Opt_err_cont, "errors=continue"},
1179 {Opt_err_panic, "errors=panic"},
1180 {Opt_err_ro, "errors=remount-ro"},
1181 {Opt_nouid32, "nouid32"},
1182 {Opt_debug, "debug"},
1183 {Opt_removed, "oldalloc"},
1184 {Opt_removed, "orlov"},
1185 {Opt_user_xattr, "user_xattr"},
1186 {Opt_nouser_xattr, "nouser_xattr"},
1187 {Opt_acl, "acl"},
1188 {Opt_noacl, "noacl"},
1189 {Opt_noload, "norecovery"},
1190 {Opt_noload, "noload"},
1191 {Opt_removed, "nobh"},
1192 {Opt_removed, "bh"},
1193 {Opt_commit, "commit=%u"},
1194 {Opt_min_batch_time, "min_batch_time=%u"},
1195 {Opt_max_batch_time, "max_batch_time=%u"},
1196 {Opt_journal_dev, "journal_dev=%u"},
1197 {Opt_journal_path, "journal_path=%s"},
1198 {Opt_journal_checksum, "journal_checksum"},
1199 {Opt_nojournal_checksum, "nojournal_checksum"},
1200 {Opt_journal_async_commit, "journal_async_commit"},
1201 {Opt_abort, "abort"},
1202 {Opt_data_journal, "data=journal"},
1203 {Opt_data_ordered, "data=ordered"},
1204 {Opt_data_writeback, "data=writeback"},
1205 {Opt_data_err_abort, "data_err=abort"},
1206 {Opt_data_err_ignore, "data_err=ignore"},
1207 {Opt_offusrjquota, "usrjquota="},
1208 {Opt_usrjquota, "usrjquota=%s"},
1209 {Opt_offgrpjquota, "grpjquota="},
1210 {Opt_grpjquota, "grpjquota=%s"},
1211 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1212 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1213 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1214 {Opt_grpquota, "grpquota"},
1215 {Opt_noquota, "noquota"},
1216 {Opt_quota, "quota"},
1217 {Opt_usrquota, "usrquota"},
1218 {Opt_barrier, "barrier=%u"},
1219 {Opt_barrier, "barrier"},
1220 {Opt_nobarrier, "nobarrier"},
1221 {Opt_i_version, "i_version"},
1222 {Opt_dax, "dax"},
1223 {Opt_stripe, "stripe=%u"},
1224 {Opt_delalloc, "delalloc"},
1225 {Opt_lazytime, "lazytime"},
1226 {Opt_nolazytime, "nolazytime"},
1227 {Opt_nodelalloc, "nodelalloc"},
1228 {Opt_removed, "mblk_io_submit"},
1229 {Opt_removed, "nomblk_io_submit"},
1230 {Opt_block_validity, "block_validity"},
1231 {Opt_noblock_validity, "noblock_validity"},
1232 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1233 {Opt_journal_ioprio, "journal_ioprio=%u"},
1234 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1235 {Opt_auto_da_alloc, "auto_da_alloc"},
1236 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1237 {Opt_dioread_nolock, "dioread_nolock"},
1238 {Opt_dioread_lock, "dioread_lock"},
1239 {Opt_discard, "discard"},
1240 {Opt_nodiscard, "nodiscard"},
1241 {Opt_init_itable, "init_itable=%u"},
1242 {Opt_init_itable, "init_itable"},
1243 {Opt_noinit_itable, "noinit_itable"},
1244 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1245 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1246 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1247 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1248 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1249 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1250 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1251 {Opt_err, NULL},
1252 };
1253
1254 static ext4_fsblk_t get_sb_block(void **data)
1255 {
1256 ext4_fsblk_t sb_block;
1257 char *options = (char *) *data;
1258
1259 if (!options || strncmp(options, "sb=", 3) != 0)
1260 return 1; /* Default location */
1261
1262 options += 3;
1263 /* TODO: use simple_strtoll with >32bit ext4 */
1264 sb_block = simple_strtoul(options, &options, 0);
1265 if (*options && *options != ',') {
1266 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1267 (char *) *data);
1268 return 1;
1269 }
1270 if (*options == ',')
1271 options++;
1272 *data = (void *) options;
1273
1274 return sb_block;
1275 }
1276
1277 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1278 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1279 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1280
1281 #ifdef CONFIG_QUOTA
1282 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1283 {
1284 struct ext4_sb_info *sbi = EXT4_SB(sb);
1285 char *qname;
1286 int ret = -1;
1287
1288 if (sb_any_quota_loaded(sb) &&
1289 !sbi->s_qf_names[qtype]) {
1290 ext4_msg(sb, KERN_ERR,
1291 "Cannot change journaled "
1292 "quota options when quota turned on");
1293 return -1;
1294 }
1295 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1296 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1297 "when QUOTA feature is enabled");
1298 return -1;
1299 }
1300 qname = match_strdup(args);
1301 if (!qname) {
1302 ext4_msg(sb, KERN_ERR,
1303 "Not enough memory for storing quotafile name");
1304 return -1;
1305 }
1306 if (sbi->s_qf_names[qtype]) {
1307 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1308 ret = 1;
1309 else
1310 ext4_msg(sb, KERN_ERR,
1311 "%s quota file already specified",
1312 QTYPE2NAME(qtype));
1313 goto errout;
1314 }
1315 if (strchr(qname, '/')) {
1316 ext4_msg(sb, KERN_ERR,
1317 "quotafile must be on filesystem root");
1318 goto errout;
1319 }
1320 sbi->s_qf_names[qtype] = qname;
1321 set_opt(sb, QUOTA);
1322 return 1;
1323 errout:
1324 kfree(qname);
1325 return ret;
1326 }
1327
1328 static int clear_qf_name(struct super_block *sb, int qtype)
1329 {
1330
1331 struct ext4_sb_info *sbi = EXT4_SB(sb);
1332
1333 if (sb_any_quota_loaded(sb) &&
1334 sbi->s_qf_names[qtype]) {
1335 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1336 " when quota turned on");
1337 return -1;
1338 }
1339 kfree(sbi->s_qf_names[qtype]);
1340 sbi->s_qf_names[qtype] = NULL;
1341 return 1;
1342 }
1343 #endif
1344
1345 #define MOPT_SET 0x0001
1346 #define MOPT_CLEAR 0x0002
1347 #define MOPT_NOSUPPORT 0x0004
1348 #define MOPT_EXPLICIT 0x0008
1349 #define MOPT_CLEAR_ERR 0x0010
1350 #define MOPT_GTE0 0x0020
1351 #ifdef CONFIG_QUOTA
1352 #define MOPT_Q 0
1353 #define MOPT_QFMT 0x0040
1354 #else
1355 #define MOPT_Q MOPT_NOSUPPORT
1356 #define MOPT_QFMT MOPT_NOSUPPORT
1357 #endif
1358 #define MOPT_DATAJ 0x0080
1359 #define MOPT_NO_EXT2 0x0100
1360 #define MOPT_NO_EXT3 0x0200
1361 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1362 #define MOPT_STRING 0x0400
1363
1364 static const struct mount_opts {
1365 int token;
1366 int mount_opt;
1367 int flags;
1368 } ext4_mount_opts[] = {
1369 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1370 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1371 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1372 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1373 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1374 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1375 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1376 MOPT_EXT4_ONLY | MOPT_SET},
1377 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1378 MOPT_EXT4_ONLY | MOPT_CLEAR},
1379 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1380 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1381 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1382 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1383 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1384 MOPT_EXT4_ONLY | MOPT_CLEAR},
1385 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1386 MOPT_EXT4_ONLY | MOPT_CLEAR},
1387 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1388 MOPT_EXT4_ONLY | MOPT_SET},
1389 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1390 EXT4_MOUNT_JOURNAL_CHECKSUM),
1391 MOPT_EXT4_ONLY | MOPT_SET},
1392 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1393 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1395 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1396 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1397 MOPT_NO_EXT2 | MOPT_SET},
1398 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1399 MOPT_NO_EXT2 | MOPT_CLEAR},
1400 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1401 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1402 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1403 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1404 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1405 {Opt_commit, 0, MOPT_GTE0},
1406 {Opt_max_batch_time, 0, MOPT_GTE0},
1407 {Opt_min_batch_time, 0, MOPT_GTE0},
1408 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1409 {Opt_init_itable, 0, MOPT_GTE0},
1410 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1411 {Opt_stripe, 0, MOPT_GTE0},
1412 {Opt_resuid, 0, MOPT_GTE0},
1413 {Opt_resgid, 0, MOPT_GTE0},
1414 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1415 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1416 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1417 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1418 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1419 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1420 MOPT_NO_EXT2 | MOPT_DATAJ},
1421 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1422 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1423 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1424 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1425 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1426 #else
1427 {Opt_acl, 0, MOPT_NOSUPPORT},
1428 {Opt_noacl, 0, MOPT_NOSUPPORT},
1429 #endif
1430 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1431 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1432 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1433 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1434 MOPT_SET | MOPT_Q},
1435 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1436 MOPT_SET | MOPT_Q},
1437 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1438 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1439 {Opt_usrjquota, 0, MOPT_Q},
1440 {Opt_grpjquota, 0, MOPT_Q},
1441 {Opt_offusrjquota, 0, MOPT_Q},
1442 {Opt_offgrpjquota, 0, MOPT_Q},
1443 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1444 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1445 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1446 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1447 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1448 {Opt_err, 0, 0}
1449 };
1450
1451 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1452 substring_t *args, unsigned long *journal_devnum,
1453 unsigned int *journal_ioprio, int is_remount)
1454 {
1455 struct ext4_sb_info *sbi = EXT4_SB(sb);
1456 const struct mount_opts *m;
1457 kuid_t uid;
1458 kgid_t gid;
1459 int arg = 0;
1460
1461 #ifdef CONFIG_QUOTA
1462 if (token == Opt_usrjquota)
1463 return set_qf_name(sb, USRQUOTA, &args[0]);
1464 else if (token == Opt_grpjquota)
1465 return set_qf_name(sb, GRPQUOTA, &args[0]);
1466 else if (token == Opt_offusrjquota)
1467 return clear_qf_name(sb, USRQUOTA);
1468 else if (token == Opt_offgrpjquota)
1469 return clear_qf_name(sb, GRPQUOTA);
1470 #endif
1471 switch (token) {
1472 case Opt_noacl:
1473 case Opt_nouser_xattr:
1474 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1475 break;
1476 case Opt_sb:
1477 return 1; /* handled by get_sb_block() */
1478 case Opt_removed:
1479 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1480 return 1;
1481 case Opt_abort:
1482 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1483 return 1;
1484 case Opt_i_version:
1485 sb->s_flags |= MS_I_VERSION;
1486 return 1;
1487 case Opt_lazytime:
1488 sb->s_flags |= MS_LAZYTIME;
1489 return 1;
1490 case Opt_nolazytime:
1491 sb->s_flags &= ~MS_LAZYTIME;
1492 return 1;
1493 }
1494
1495 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1496 if (token == m->token)
1497 break;
1498
1499 if (m->token == Opt_err) {
1500 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1501 "or missing value", opt);
1502 return -1;
1503 }
1504
1505 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1506 ext4_msg(sb, KERN_ERR,
1507 "Mount option \"%s\" incompatible with ext2", opt);
1508 return -1;
1509 }
1510 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1511 ext4_msg(sb, KERN_ERR,
1512 "Mount option \"%s\" incompatible with ext3", opt);
1513 return -1;
1514 }
1515
1516 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1517 return -1;
1518 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1519 return -1;
1520 if (m->flags & MOPT_EXPLICIT)
1521 set_opt2(sb, EXPLICIT_DELALLOC);
1522 if (m->flags & MOPT_CLEAR_ERR)
1523 clear_opt(sb, ERRORS_MASK);
1524 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1525 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1526 "options when quota turned on");
1527 return -1;
1528 }
1529
1530 if (m->flags & MOPT_NOSUPPORT) {
1531 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1532 } else if (token == Opt_commit) {
1533 if (arg == 0)
1534 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1535 sbi->s_commit_interval = HZ * arg;
1536 } else if (token == Opt_max_batch_time) {
1537 sbi->s_max_batch_time = arg;
1538 } else if (token == Opt_min_batch_time) {
1539 sbi->s_min_batch_time = arg;
1540 } else if (token == Opt_inode_readahead_blks) {
1541 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1542 ext4_msg(sb, KERN_ERR,
1543 "EXT4-fs: inode_readahead_blks must be "
1544 "0 or a power of 2 smaller than 2^31");
1545 return -1;
1546 }
1547 sbi->s_inode_readahead_blks = arg;
1548 } else if (token == Opt_init_itable) {
1549 set_opt(sb, INIT_INODE_TABLE);
1550 if (!args->from)
1551 arg = EXT4_DEF_LI_WAIT_MULT;
1552 sbi->s_li_wait_mult = arg;
1553 } else if (token == Opt_max_dir_size_kb) {
1554 sbi->s_max_dir_size_kb = arg;
1555 } else if (token == Opt_stripe) {
1556 sbi->s_stripe = arg;
1557 } else if (token == Opt_resuid) {
1558 uid = make_kuid(current_user_ns(), arg);
1559 if (!uid_valid(uid)) {
1560 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1561 return -1;
1562 }
1563 sbi->s_resuid = uid;
1564 } else if (token == Opt_resgid) {
1565 gid = make_kgid(current_user_ns(), arg);
1566 if (!gid_valid(gid)) {
1567 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1568 return -1;
1569 }
1570 sbi->s_resgid = gid;
1571 } else if (token == Opt_journal_dev) {
1572 if (is_remount) {
1573 ext4_msg(sb, KERN_ERR,
1574 "Cannot specify journal on remount");
1575 return -1;
1576 }
1577 *journal_devnum = arg;
1578 } else if (token == Opt_journal_path) {
1579 char *journal_path;
1580 struct inode *journal_inode;
1581 struct path path;
1582 int error;
1583
1584 if (is_remount) {
1585 ext4_msg(sb, KERN_ERR,
1586 "Cannot specify journal on remount");
1587 return -1;
1588 }
1589 journal_path = match_strdup(&args[0]);
1590 if (!journal_path) {
1591 ext4_msg(sb, KERN_ERR, "error: could not dup "
1592 "journal device string");
1593 return -1;
1594 }
1595
1596 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1597 if (error) {
1598 ext4_msg(sb, KERN_ERR, "error: could not find "
1599 "journal device path: error %d", error);
1600 kfree(journal_path);
1601 return -1;
1602 }
1603
1604 journal_inode = d_inode(path.dentry);
1605 if (!S_ISBLK(journal_inode->i_mode)) {
1606 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1607 "is not a block device", journal_path);
1608 path_put(&path);
1609 kfree(journal_path);
1610 return -1;
1611 }
1612
1613 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1614 path_put(&path);
1615 kfree(journal_path);
1616 } else if (token == Opt_journal_ioprio) {
1617 if (arg > 7) {
1618 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1619 " (must be 0-7)");
1620 return -1;
1621 }
1622 *journal_ioprio =
1623 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1624 } else if (token == Opt_test_dummy_encryption) {
1625 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1626 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1627 ext4_msg(sb, KERN_WARNING,
1628 "Test dummy encryption mode enabled");
1629 #else
1630 ext4_msg(sb, KERN_WARNING,
1631 "Test dummy encryption mount option ignored");
1632 #endif
1633 } else if (m->flags & MOPT_DATAJ) {
1634 if (is_remount) {
1635 if (!sbi->s_journal)
1636 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1637 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1638 ext4_msg(sb, KERN_ERR,
1639 "Cannot change data mode on remount");
1640 return -1;
1641 }
1642 } else {
1643 clear_opt(sb, DATA_FLAGS);
1644 sbi->s_mount_opt |= m->mount_opt;
1645 }
1646 #ifdef CONFIG_QUOTA
1647 } else if (m->flags & MOPT_QFMT) {
1648 if (sb_any_quota_loaded(sb) &&
1649 sbi->s_jquota_fmt != m->mount_opt) {
1650 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1651 "quota options when quota turned on");
1652 return -1;
1653 }
1654 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1655 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1656 ext4_msg(sb, KERN_ERR,
1657 "Cannot set journaled quota options "
1658 "when QUOTA feature is enabled");
1659 return -1;
1660 }
1661 sbi->s_jquota_fmt = m->mount_opt;
1662 #endif
1663 #ifndef CONFIG_FS_DAX
1664 } else if (token == Opt_dax) {
1665 ext4_msg(sb, KERN_INFO, "dax option not supported");
1666 return -1;
1667 #endif
1668 } else {
1669 if (!args->from)
1670 arg = 1;
1671 if (m->flags & MOPT_CLEAR)
1672 arg = !arg;
1673 else if (unlikely(!(m->flags & MOPT_SET))) {
1674 ext4_msg(sb, KERN_WARNING,
1675 "buggy handling of option %s", opt);
1676 WARN_ON(1);
1677 return -1;
1678 }
1679 if (arg != 0)
1680 sbi->s_mount_opt |= m->mount_opt;
1681 else
1682 sbi->s_mount_opt &= ~m->mount_opt;
1683 }
1684 return 1;
1685 }
1686
1687 static int parse_options(char *options, struct super_block *sb,
1688 unsigned long *journal_devnum,
1689 unsigned int *journal_ioprio,
1690 int is_remount)
1691 {
1692 struct ext4_sb_info *sbi = EXT4_SB(sb);
1693 char *p;
1694 substring_t args[MAX_OPT_ARGS];
1695 int token;
1696
1697 if (!options)
1698 return 1;
1699
1700 while ((p = strsep(&options, ",")) != NULL) {
1701 if (!*p)
1702 continue;
1703 /*
1704 * Initialize args struct so we know whether arg was
1705 * found; some options take optional arguments.
1706 */
1707 args[0].to = args[0].from = NULL;
1708 token = match_token(p, tokens, args);
1709 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1710 journal_ioprio, is_remount) < 0)
1711 return 0;
1712 }
1713 #ifdef CONFIG_QUOTA
1714 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1715 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1716 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1717 "feature is enabled");
1718 return 0;
1719 }
1720 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1721 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1722 clear_opt(sb, USRQUOTA);
1723
1724 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1725 clear_opt(sb, GRPQUOTA);
1726
1727 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1728 ext4_msg(sb, KERN_ERR, "old and new quota "
1729 "format mixing");
1730 return 0;
1731 }
1732
1733 if (!sbi->s_jquota_fmt) {
1734 ext4_msg(sb, KERN_ERR, "journaled quota format "
1735 "not specified");
1736 return 0;
1737 }
1738 }
1739 #endif
1740 if (test_opt(sb, DIOREAD_NOLOCK)) {
1741 int blocksize =
1742 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1743
1744 if (blocksize < PAGE_CACHE_SIZE) {
1745 ext4_msg(sb, KERN_ERR, "can't mount with "
1746 "dioread_nolock if block size != PAGE_SIZE");
1747 return 0;
1748 }
1749 }
1750 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1751 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1752 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1753 "in data=ordered mode");
1754 return 0;
1755 }
1756 return 1;
1757 }
1758
1759 static inline void ext4_show_quota_options(struct seq_file *seq,
1760 struct super_block *sb)
1761 {
1762 #if defined(CONFIG_QUOTA)
1763 struct ext4_sb_info *sbi = EXT4_SB(sb);
1764
1765 if (sbi->s_jquota_fmt) {
1766 char *fmtname = "";
1767
1768 switch (sbi->s_jquota_fmt) {
1769 case QFMT_VFS_OLD:
1770 fmtname = "vfsold";
1771 break;
1772 case QFMT_VFS_V0:
1773 fmtname = "vfsv0";
1774 break;
1775 case QFMT_VFS_V1:
1776 fmtname = "vfsv1";
1777 break;
1778 }
1779 seq_printf(seq, ",jqfmt=%s", fmtname);
1780 }
1781
1782 if (sbi->s_qf_names[USRQUOTA])
1783 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1784
1785 if (sbi->s_qf_names[GRPQUOTA])
1786 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1787 #endif
1788 }
1789
1790 static const char *token2str(int token)
1791 {
1792 const struct match_token *t;
1793
1794 for (t = tokens; t->token != Opt_err; t++)
1795 if (t->token == token && !strchr(t->pattern, '='))
1796 break;
1797 return t->pattern;
1798 }
1799
1800 /*
1801 * Show an option if
1802 * - it's set to a non-default value OR
1803 * - if the per-sb default is different from the global default
1804 */
1805 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1806 int nodefs)
1807 {
1808 struct ext4_sb_info *sbi = EXT4_SB(sb);
1809 struct ext4_super_block *es = sbi->s_es;
1810 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1811 const struct mount_opts *m;
1812 char sep = nodefs ? '\n' : ',';
1813
1814 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1815 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1816
1817 if (sbi->s_sb_block != 1)
1818 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1819
1820 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1821 int want_set = m->flags & MOPT_SET;
1822 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1823 (m->flags & MOPT_CLEAR_ERR))
1824 continue;
1825 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1826 continue; /* skip if same as the default */
1827 if ((want_set &&
1828 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1829 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1830 continue; /* select Opt_noFoo vs Opt_Foo */
1831 SEQ_OPTS_PRINT("%s", token2str(m->token));
1832 }
1833
1834 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1835 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1836 SEQ_OPTS_PRINT("resuid=%u",
1837 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1838 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1839 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1840 SEQ_OPTS_PRINT("resgid=%u",
1841 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1842 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1843 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1844 SEQ_OPTS_PUTS("errors=remount-ro");
1845 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1846 SEQ_OPTS_PUTS("errors=continue");
1847 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1848 SEQ_OPTS_PUTS("errors=panic");
1849 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1850 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1851 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1852 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1853 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1854 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1855 if (sb->s_flags & MS_I_VERSION)
1856 SEQ_OPTS_PUTS("i_version");
1857 if (nodefs || sbi->s_stripe)
1858 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1859 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1860 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1861 SEQ_OPTS_PUTS("data=journal");
1862 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1863 SEQ_OPTS_PUTS("data=ordered");
1864 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1865 SEQ_OPTS_PUTS("data=writeback");
1866 }
1867 if (nodefs ||
1868 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1869 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1870 sbi->s_inode_readahead_blks);
1871
1872 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1873 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1874 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1875 if (nodefs || sbi->s_max_dir_size_kb)
1876 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1877
1878 ext4_show_quota_options(seq, sb);
1879 return 0;
1880 }
1881
1882 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1883 {
1884 return _ext4_show_options(seq, root->d_sb, 0);
1885 }
1886
1887 static int options_seq_show(struct seq_file *seq, void *offset)
1888 {
1889 struct super_block *sb = seq->private;
1890 int rc;
1891
1892 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1893 rc = _ext4_show_options(seq, sb, 1);
1894 seq_puts(seq, "\n");
1895 return rc;
1896 }
1897
1898 static int options_open_fs(struct inode *inode, struct file *file)
1899 {
1900 return single_open(file, options_seq_show, PDE_DATA(inode));
1901 }
1902
1903 static const struct file_operations ext4_seq_options_fops = {
1904 .owner = THIS_MODULE,
1905 .open = options_open_fs,
1906 .read = seq_read,
1907 .llseek = seq_lseek,
1908 .release = single_release,
1909 };
1910
1911 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1912 int read_only)
1913 {
1914 struct ext4_sb_info *sbi = EXT4_SB(sb);
1915 int res = 0;
1916
1917 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1918 ext4_msg(sb, KERN_ERR, "revision level too high, "
1919 "forcing read-only mode");
1920 res = MS_RDONLY;
1921 }
1922 if (read_only)
1923 goto done;
1924 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1925 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1926 "running e2fsck is recommended");
1927 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1928 ext4_msg(sb, KERN_WARNING,
1929 "warning: mounting fs with errors, "
1930 "running e2fsck is recommended");
1931 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1932 le16_to_cpu(es->s_mnt_count) >=
1933 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1934 ext4_msg(sb, KERN_WARNING,
1935 "warning: maximal mount count reached, "
1936 "running e2fsck is recommended");
1937 else if (le32_to_cpu(es->s_checkinterval) &&
1938 (le32_to_cpu(es->s_lastcheck) +
1939 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1940 ext4_msg(sb, KERN_WARNING,
1941 "warning: checktime reached, "
1942 "running e2fsck is recommended");
1943 if (!sbi->s_journal)
1944 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1945 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1946 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1947 le16_add_cpu(&es->s_mnt_count, 1);
1948 es->s_mtime = cpu_to_le32(get_seconds());
1949 ext4_update_dynamic_rev(sb);
1950 if (sbi->s_journal)
1951 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1952
1953 ext4_commit_super(sb, 1);
1954 done:
1955 if (test_opt(sb, DEBUG))
1956 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1957 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1958 sb->s_blocksize,
1959 sbi->s_groups_count,
1960 EXT4_BLOCKS_PER_GROUP(sb),
1961 EXT4_INODES_PER_GROUP(sb),
1962 sbi->s_mount_opt, sbi->s_mount_opt2);
1963
1964 cleancache_init_fs(sb);
1965 return res;
1966 }
1967
1968 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1969 {
1970 struct ext4_sb_info *sbi = EXT4_SB(sb);
1971 struct flex_groups *new_groups;
1972 int size;
1973
1974 if (!sbi->s_log_groups_per_flex)
1975 return 0;
1976
1977 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1978 if (size <= sbi->s_flex_groups_allocated)
1979 return 0;
1980
1981 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1982 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1983 if (!new_groups) {
1984 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1985 size / (int) sizeof(struct flex_groups));
1986 return -ENOMEM;
1987 }
1988
1989 if (sbi->s_flex_groups) {
1990 memcpy(new_groups, sbi->s_flex_groups,
1991 (sbi->s_flex_groups_allocated *
1992 sizeof(struct flex_groups)));
1993 kvfree(sbi->s_flex_groups);
1994 }
1995 sbi->s_flex_groups = new_groups;
1996 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1997 return 0;
1998 }
1999
2000 static int ext4_fill_flex_info(struct super_block *sb)
2001 {
2002 struct ext4_sb_info *sbi = EXT4_SB(sb);
2003 struct ext4_group_desc *gdp = NULL;
2004 ext4_group_t flex_group;
2005 int i, err;
2006
2007 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2008 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2009 sbi->s_log_groups_per_flex = 0;
2010 return 1;
2011 }
2012
2013 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2014 if (err)
2015 goto failed;
2016
2017 for (i = 0; i < sbi->s_groups_count; i++) {
2018 gdp = ext4_get_group_desc(sb, i, NULL);
2019
2020 flex_group = ext4_flex_group(sbi, i);
2021 atomic_add(ext4_free_inodes_count(sb, gdp),
2022 &sbi->s_flex_groups[flex_group].free_inodes);
2023 atomic64_add(ext4_free_group_clusters(sb, gdp),
2024 &sbi->s_flex_groups[flex_group].free_clusters);
2025 atomic_add(ext4_used_dirs_count(sb, gdp),
2026 &sbi->s_flex_groups[flex_group].used_dirs);
2027 }
2028
2029 return 1;
2030 failed:
2031 return 0;
2032 }
2033
2034 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2035 struct ext4_group_desc *gdp)
2036 {
2037 int offset;
2038 __u16 crc = 0;
2039 __le32 le_group = cpu_to_le32(block_group);
2040
2041 if (ext4_has_metadata_csum(sbi->s_sb)) {
2042 /* Use new metadata_csum algorithm */
2043 __le16 save_csum;
2044 __u32 csum32;
2045
2046 save_csum = gdp->bg_checksum;
2047 gdp->bg_checksum = 0;
2048 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2049 sizeof(le_group));
2050 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2051 sbi->s_desc_size);
2052 gdp->bg_checksum = save_csum;
2053
2054 crc = csum32 & 0xFFFF;
2055 goto out;
2056 }
2057
2058 /* old crc16 code */
2059 if (!(sbi->s_es->s_feature_ro_compat &
2060 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2061 return 0;
2062
2063 offset = offsetof(struct ext4_group_desc, bg_checksum);
2064
2065 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2066 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2067 crc = crc16(crc, (__u8 *)gdp, offset);
2068 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2069 /* for checksum of struct ext4_group_desc do the rest...*/
2070 if ((sbi->s_es->s_feature_incompat &
2071 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2072 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2073 crc = crc16(crc, (__u8 *)gdp + offset,
2074 le16_to_cpu(sbi->s_es->s_desc_size) -
2075 offset);
2076
2077 out:
2078 return cpu_to_le16(crc);
2079 }
2080
2081 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2082 struct ext4_group_desc *gdp)
2083 {
2084 if (ext4_has_group_desc_csum(sb) &&
2085 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2086 block_group, gdp)))
2087 return 0;
2088
2089 return 1;
2090 }
2091
2092 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2093 struct ext4_group_desc *gdp)
2094 {
2095 if (!ext4_has_group_desc_csum(sb))
2096 return;
2097 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2098 }
2099
2100 /* Called at mount-time, super-block is locked */
2101 static int ext4_check_descriptors(struct super_block *sb,
2102 ext4_group_t *first_not_zeroed)
2103 {
2104 struct ext4_sb_info *sbi = EXT4_SB(sb);
2105 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2106 ext4_fsblk_t last_block;
2107 ext4_fsblk_t block_bitmap;
2108 ext4_fsblk_t inode_bitmap;
2109 ext4_fsblk_t inode_table;
2110 int flexbg_flag = 0;
2111 ext4_group_t i, grp = sbi->s_groups_count;
2112
2113 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2114 flexbg_flag = 1;
2115
2116 ext4_debug("Checking group descriptors");
2117
2118 for (i = 0; i < sbi->s_groups_count; i++) {
2119 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2120
2121 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2122 last_block = ext4_blocks_count(sbi->s_es) - 1;
2123 else
2124 last_block = first_block +
2125 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2126
2127 if ((grp == sbi->s_groups_count) &&
2128 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2129 grp = i;
2130
2131 block_bitmap = ext4_block_bitmap(sb, gdp);
2132 if (block_bitmap < first_block || block_bitmap > last_block) {
2133 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2134 "Block bitmap for group %u not in group "
2135 "(block %llu)!", i, block_bitmap);
2136 return 0;
2137 }
2138 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2139 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2140 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2141 "Inode bitmap for group %u not in group "
2142 "(block %llu)!", i, inode_bitmap);
2143 return 0;
2144 }
2145 inode_table = ext4_inode_table(sb, gdp);
2146 if (inode_table < first_block ||
2147 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2148 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2149 "Inode table for group %u not in group "
2150 "(block %llu)!", i, inode_table);
2151 return 0;
2152 }
2153 ext4_lock_group(sb, i);
2154 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2155 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2156 "Checksum for group %u failed (%u!=%u)",
2157 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2158 gdp)), le16_to_cpu(gdp->bg_checksum));
2159 if (!(sb->s_flags & MS_RDONLY)) {
2160 ext4_unlock_group(sb, i);
2161 return 0;
2162 }
2163 }
2164 ext4_unlock_group(sb, i);
2165 if (!flexbg_flag)
2166 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2167 }
2168 if (NULL != first_not_zeroed)
2169 *first_not_zeroed = grp;
2170 return 1;
2171 }
2172
2173 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2174 * the superblock) which were deleted from all directories, but held open by
2175 * a process at the time of a crash. We walk the list and try to delete these
2176 * inodes at recovery time (only with a read-write filesystem).
2177 *
2178 * In order to keep the orphan inode chain consistent during traversal (in
2179 * case of crash during recovery), we link each inode into the superblock
2180 * orphan list_head and handle it the same way as an inode deletion during
2181 * normal operation (which journals the operations for us).
2182 *
2183 * We only do an iget() and an iput() on each inode, which is very safe if we
2184 * accidentally point at an in-use or already deleted inode. The worst that
2185 * can happen in this case is that we get a "bit already cleared" message from
2186 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2187 * e2fsck was run on this filesystem, and it must have already done the orphan
2188 * inode cleanup for us, so we can safely abort without any further action.
2189 */
2190 static void ext4_orphan_cleanup(struct super_block *sb,
2191 struct ext4_super_block *es)
2192 {
2193 unsigned int s_flags = sb->s_flags;
2194 int nr_orphans = 0, nr_truncates = 0;
2195 #ifdef CONFIG_QUOTA
2196 int i;
2197 #endif
2198 if (!es->s_last_orphan) {
2199 jbd_debug(4, "no orphan inodes to clean up\n");
2200 return;
2201 }
2202
2203 if (bdev_read_only(sb->s_bdev)) {
2204 ext4_msg(sb, KERN_ERR, "write access "
2205 "unavailable, skipping orphan cleanup");
2206 return;
2207 }
2208
2209 /* Check if feature set would not allow a r/w mount */
2210 if (!ext4_feature_set_ok(sb, 0)) {
2211 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2212 "unknown ROCOMPAT features");
2213 return;
2214 }
2215
2216 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2217 /* don't clear list on RO mount w/ errors */
2218 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2219 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2220 "clearing orphan list.\n");
2221 es->s_last_orphan = 0;
2222 }
2223 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2224 return;
2225 }
2226
2227 if (s_flags & MS_RDONLY) {
2228 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2229 sb->s_flags &= ~MS_RDONLY;
2230 }
2231 #ifdef CONFIG_QUOTA
2232 /* Needed for iput() to work correctly and not trash data */
2233 sb->s_flags |= MS_ACTIVE;
2234 /* Turn on quotas so that they are updated correctly */
2235 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2236 if (EXT4_SB(sb)->s_qf_names[i]) {
2237 int ret = ext4_quota_on_mount(sb, i);
2238 if (ret < 0)
2239 ext4_msg(sb, KERN_ERR,
2240 "Cannot turn on journaled "
2241 "quota: error %d", ret);
2242 }
2243 }
2244 #endif
2245
2246 while (es->s_last_orphan) {
2247 struct inode *inode;
2248
2249 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2250 if (IS_ERR(inode)) {
2251 es->s_last_orphan = 0;
2252 break;
2253 }
2254
2255 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2256 dquot_initialize(inode);
2257 if (inode->i_nlink) {
2258 if (test_opt(sb, DEBUG))
2259 ext4_msg(sb, KERN_DEBUG,
2260 "%s: truncating inode %lu to %lld bytes",
2261 __func__, inode->i_ino, inode->i_size);
2262 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2263 inode->i_ino, inode->i_size);
2264 mutex_lock(&inode->i_mutex);
2265 truncate_inode_pages(inode->i_mapping, inode->i_size);
2266 ext4_truncate(inode);
2267 mutex_unlock(&inode->i_mutex);
2268 nr_truncates++;
2269 } else {
2270 if (test_opt(sb, DEBUG))
2271 ext4_msg(sb, KERN_DEBUG,
2272 "%s: deleting unreferenced inode %lu",
2273 __func__, inode->i_ino);
2274 jbd_debug(2, "deleting unreferenced inode %lu\n",
2275 inode->i_ino);
2276 nr_orphans++;
2277 }
2278 iput(inode); /* The delete magic happens here! */
2279 }
2280
2281 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2282
2283 if (nr_orphans)
2284 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2285 PLURAL(nr_orphans));
2286 if (nr_truncates)
2287 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2288 PLURAL(nr_truncates));
2289 #ifdef CONFIG_QUOTA
2290 /* Turn quotas off */
2291 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2292 if (sb_dqopt(sb)->files[i])
2293 dquot_quota_off(sb, i);
2294 }
2295 #endif
2296 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2297 }
2298
2299 /*
2300 * Maximal extent format file size.
2301 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2302 * extent format containers, within a sector_t, and within i_blocks
2303 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2304 * so that won't be a limiting factor.
2305 *
2306 * However there is other limiting factor. We do store extents in the form
2307 * of starting block and length, hence the resulting length of the extent
2308 * covering maximum file size must fit into on-disk format containers as
2309 * well. Given that length is always by 1 unit bigger than max unit (because
2310 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2311 *
2312 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2313 */
2314 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2315 {
2316 loff_t res;
2317 loff_t upper_limit = MAX_LFS_FILESIZE;
2318
2319 /* small i_blocks in vfs inode? */
2320 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2321 /*
2322 * CONFIG_LBDAF is not enabled implies the inode
2323 * i_block represent total blocks in 512 bytes
2324 * 32 == size of vfs inode i_blocks * 8
2325 */
2326 upper_limit = (1LL << 32) - 1;
2327
2328 /* total blocks in file system block size */
2329 upper_limit >>= (blkbits - 9);
2330 upper_limit <<= blkbits;
2331 }
2332
2333 /*
2334 * 32-bit extent-start container, ee_block. We lower the maxbytes
2335 * by one fs block, so ee_len can cover the extent of maximum file
2336 * size
2337 */
2338 res = (1LL << 32) - 1;
2339 res <<= blkbits;
2340
2341 /* Sanity check against vm- & vfs- imposed limits */
2342 if (res > upper_limit)
2343 res = upper_limit;
2344
2345 return res;
2346 }
2347
2348 /*
2349 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2350 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2351 * We need to be 1 filesystem block less than the 2^48 sector limit.
2352 */
2353 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2354 {
2355 loff_t res = EXT4_NDIR_BLOCKS;
2356 int meta_blocks;
2357 loff_t upper_limit;
2358 /* This is calculated to be the largest file size for a dense, block
2359 * mapped file such that the file's total number of 512-byte sectors,
2360 * including data and all indirect blocks, does not exceed (2^48 - 1).
2361 *
2362 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2363 * number of 512-byte sectors of the file.
2364 */
2365
2366 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2367 /*
2368 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2369 * the inode i_block field represents total file blocks in
2370 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2371 */
2372 upper_limit = (1LL << 32) - 1;
2373
2374 /* total blocks in file system block size */
2375 upper_limit >>= (bits - 9);
2376
2377 } else {
2378 /*
2379 * We use 48 bit ext4_inode i_blocks
2380 * With EXT4_HUGE_FILE_FL set the i_blocks
2381 * represent total number of blocks in
2382 * file system block size
2383 */
2384 upper_limit = (1LL << 48) - 1;
2385
2386 }
2387
2388 /* indirect blocks */
2389 meta_blocks = 1;
2390 /* double indirect blocks */
2391 meta_blocks += 1 + (1LL << (bits-2));
2392 /* tripple indirect blocks */
2393 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2394
2395 upper_limit -= meta_blocks;
2396 upper_limit <<= bits;
2397
2398 res += 1LL << (bits-2);
2399 res += 1LL << (2*(bits-2));
2400 res += 1LL << (3*(bits-2));
2401 res <<= bits;
2402 if (res > upper_limit)
2403 res = upper_limit;
2404
2405 if (res > MAX_LFS_FILESIZE)
2406 res = MAX_LFS_FILESIZE;
2407
2408 return res;
2409 }
2410
2411 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2412 ext4_fsblk_t logical_sb_block, int nr)
2413 {
2414 struct ext4_sb_info *sbi = EXT4_SB(sb);
2415 ext4_group_t bg, first_meta_bg;
2416 int has_super = 0;
2417
2418 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2419
2420 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2421 nr < first_meta_bg)
2422 return logical_sb_block + nr + 1;
2423 bg = sbi->s_desc_per_block * nr;
2424 if (ext4_bg_has_super(sb, bg))
2425 has_super = 1;
2426
2427 /*
2428 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2429 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2430 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2431 * compensate.
2432 */
2433 if (sb->s_blocksize == 1024 && nr == 0 &&
2434 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2435 has_super++;
2436
2437 return (has_super + ext4_group_first_block_no(sb, bg));
2438 }
2439
2440 /**
2441 * ext4_get_stripe_size: Get the stripe size.
2442 * @sbi: In memory super block info
2443 *
2444 * If we have specified it via mount option, then
2445 * use the mount option value. If the value specified at mount time is
2446 * greater than the blocks per group use the super block value.
2447 * If the super block value is greater than blocks per group return 0.
2448 * Allocator needs it be less than blocks per group.
2449 *
2450 */
2451 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2452 {
2453 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2454 unsigned long stripe_width =
2455 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2456 int ret;
2457
2458 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2459 ret = sbi->s_stripe;
2460 else if (stripe_width <= sbi->s_blocks_per_group)
2461 ret = stripe_width;
2462 else if (stride <= sbi->s_blocks_per_group)
2463 ret = stride;
2464 else
2465 ret = 0;
2466
2467 /*
2468 * If the stripe width is 1, this makes no sense and
2469 * we set it to 0 to turn off stripe handling code.
2470 */
2471 if (ret <= 1)
2472 ret = 0;
2473
2474 return ret;
2475 }
2476
2477 /* sysfs supprt */
2478
2479 struct ext4_attr {
2480 struct attribute attr;
2481 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2482 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2483 const char *, size_t);
2484 union {
2485 int offset;
2486 int deprecated_val;
2487 } u;
2488 };
2489
2490 static int parse_strtoull(const char *buf,
2491 unsigned long long max, unsigned long long *value)
2492 {
2493 int ret;
2494
2495 ret = kstrtoull(skip_spaces(buf), 0, value);
2496 if (!ret && *value > max)
2497 ret = -EINVAL;
2498 return ret;
2499 }
2500
2501 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2502 struct ext4_sb_info *sbi,
2503 char *buf)
2504 {
2505 return snprintf(buf, PAGE_SIZE, "%llu\n",
2506 (s64) EXT4_C2B(sbi,
2507 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2508 }
2509
2510 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2511 struct ext4_sb_info *sbi, char *buf)
2512 {
2513 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2514
2515 if (!sb->s_bdev->bd_part)
2516 return snprintf(buf, PAGE_SIZE, "0\n");
2517 return snprintf(buf, PAGE_SIZE, "%lu\n",
2518 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2519 sbi->s_sectors_written_start) >> 1);
2520 }
2521
2522 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2523 struct ext4_sb_info *sbi, char *buf)
2524 {
2525 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2526
2527 if (!sb->s_bdev->bd_part)
2528 return snprintf(buf, PAGE_SIZE, "0\n");
2529 return snprintf(buf, PAGE_SIZE, "%llu\n",
2530 (unsigned long long)(sbi->s_kbytes_written +
2531 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2532 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2533 }
2534
2535 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2536 struct ext4_sb_info *sbi,
2537 const char *buf, size_t count)
2538 {
2539 unsigned long t;
2540 int ret;
2541
2542 ret = kstrtoul(skip_spaces(buf), 0, &t);
2543 if (ret)
2544 return ret;
2545
2546 if (t && (!is_power_of_2(t) || t > 0x40000000))
2547 return -EINVAL;
2548
2549 sbi->s_inode_readahead_blks = t;
2550 return count;
2551 }
2552
2553 static ssize_t sbi_ui_show(struct ext4_attr *a,
2554 struct ext4_sb_info *sbi, char *buf)
2555 {
2556 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2557
2558 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2559 }
2560
2561 static ssize_t sbi_ui_store(struct ext4_attr *a,
2562 struct ext4_sb_info *sbi,
2563 const char *buf, size_t count)
2564 {
2565 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2566 unsigned long t;
2567 int ret;
2568
2569 ret = kstrtoul(skip_spaces(buf), 0, &t);
2570 if (ret)
2571 return ret;
2572 *ui = t;
2573 return count;
2574 }
2575
2576 static ssize_t es_ui_show(struct ext4_attr *a,
2577 struct ext4_sb_info *sbi, char *buf)
2578 {
2579
2580 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2581 a->u.offset);
2582
2583 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2584 }
2585
2586 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2587 struct ext4_sb_info *sbi, char *buf)
2588 {
2589 return snprintf(buf, PAGE_SIZE, "%llu\n",
2590 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2591 }
2592
2593 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2594 struct ext4_sb_info *sbi,
2595 const char *buf, size_t count)
2596 {
2597 unsigned long long val;
2598 int ret;
2599
2600 if (parse_strtoull(buf, -1ULL, &val))
2601 return -EINVAL;
2602 ret = ext4_reserve_clusters(sbi, val);
2603
2604 return ret ? ret : count;
2605 }
2606
2607 static ssize_t trigger_test_error(struct ext4_attr *a,
2608 struct ext4_sb_info *sbi,
2609 const char *buf, size_t count)
2610 {
2611 int len = count;
2612
2613 if (!capable(CAP_SYS_ADMIN))
2614 return -EPERM;
2615
2616 if (len && buf[len-1] == '\n')
2617 len--;
2618
2619 if (len)
2620 ext4_error(sbi->s_sb, "%.*s", len, buf);
2621 return count;
2622 }
2623
2624 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2625 struct ext4_sb_info *sbi, char *buf)
2626 {
2627 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2628 }
2629
2630 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2631 static struct ext4_attr ext4_attr_##_name = { \
2632 .attr = {.name = __stringify(_name), .mode = _mode }, \
2633 .show = _show, \
2634 .store = _store, \
2635 .u = { \
2636 .offset = offsetof(struct ext4_sb_info, _elname),\
2637 }, \
2638 }
2639
2640 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2641 static struct ext4_attr ext4_attr_##_name = { \
2642 .attr = {.name = __stringify(_name), .mode = _mode }, \
2643 .show = _show, \
2644 .store = _store, \
2645 .u = { \
2646 .offset = offsetof(struct ext4_super_block, _elname), \
2647 }, \
2648 }
2649
2650 #define EXT4_ATTR(name, mode, show, store) \
2651 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2652
2653 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2654 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2655 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2656
2657 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2658 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2659 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2660 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2661
2662 #define ATTR_LIST(name) &ext4_attr_##name.attr
2663 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2664 static struct ext4_attr ext4_attr_##_name = { \
2665 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2666 .show = sbi_deprecated_show, \
2667 .u = { \
2668 .deprecated_val = _val, \
2669 }, \
2670 }
2671
2672 EXT4_RO_ATTR(delayed_allocation_blocks);
2673 EXT4_RO_ATTR(session_write_kbytes);
2674 EXT4_RO_ATTR(lifetime_write_kbytes);
2675 EXT4_RW_ATTR(reserved_clusters);
2676 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2677 inode_readahead_blks_store, s_inode_readahead_blks);
2678 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2679 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2680 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2681 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2682 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2683 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2684 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2685 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2686 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2687 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2688 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2689 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2690 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2691 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2692 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2693 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2694 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2695 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2696 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2697
2698 static struct attribute *ext4_attrs[] = {
2699 ATTR_LIST(delayed_allocation_blocks),
2700 ATTR_LIST(session_write_kbytes),
2701 ATTR_LIST(lifetime_write_kbytes),
2702 ATTR_LIST(reserved_clusters),
2703 ATTR_LIST(inode_readahead_blks),
2704 ATTR_LIST(inode_goal),
2705 ATTR_LIST(mb_stats),
2706 ATTR_LIST(mb_max_to_scan),
2707 ATTR_LIST(mb_min_to_scan),
2708 ATTR_LIST(mb_order2_req),
2709 ATTR_LIST(mb_stream_req),
2710 ATTR_LIST(mb_group_prealloc),
2711 ATTR_LIST(max_writeback_mb_bump),
2712 ATTR_LIST(extent_max_zeroout_kb),
2713 ATTR_LIST(trigger_fs_error),
2714 ATTR_LIST(err_ratelimit_interval_ms),
2715 ATTR_LIST(err_ratelimit_burst),
2716 ATTR_LIST(warning_ratelimit_interval_ms),
2717 ATTR_LIST(warning_ratelimit_burst),
2718 ATTR_LIST(msg_ratelimit_interval_ms),
2719 ATTR_LIST(msg_ratelimit_burst),
2720 ATTR_LIST(errors_count),
2721 ATTR_LIST(first_error_time),
2722 ATTR_LIST(last_error_time),
2723 NULL,
2724 };
2725
2726 /* Features this copy of ext4 supports */
2727 EXT4_INFO_ATTR(lazy_itable_init);
2728 EXT4_INFO_ATTR(batched_discard);
2729 EXT4_INFO_ATTR(meta_bg_resize);
2730 EXT4_INFO_ATTR(encryption);
2731
2732 static struct attribute *ext4_feat_attrs[] = {
2733 ATTR_LIST(lazy_itable_init),
2734 ATTR_LIST(batched_discard),
2735 ATTR_LIST(meta_bg_resize),
2736 ATTR_LIST(encryption),
2737 NULL,
2738 };
2739
2740 static ssize_t ext4_attr_show(struct kobject *kobj,
2741 struct attribute *attr, char *buf)
2742 {
2743 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2744 s_kobj);
2745 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2746
2747 return a->show ? a->show(a, sbi, buf) : 0;
2748 }
2749
2750 static ssize_t ext4_attr_store(struct kobject *kobj,
2751 struct attribute *attr,
2752 const char *buf, size_t len)
2753 {
2754 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2755 s_kobj);
2756 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2757
2758 return a->store ? a->store(a, sbi, buf, len) : 0;
2759 }
2760
2761 static void ext4_sb_release(struct kobject *kobj)
2762 {
2763 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2764 s_kobj);
2765 complete(&sbi->s_kobj_unregister);
2766 }
2767
2768 static const struct sysfs_ops ext4_attr_ops = {
2769 .show = ext4_attr_show,
2770 .store = ext4_attr_store,
2771 };
2772
2773 static struct kobj_type ext4_ktype = {
2774 .default_attrs = ext4_attrs,
2775 .sysfs_ops = &ext4_attr_ops,
2776 .release = ext4_sb_release,
2777 };
2778
2779 static void ext4_feat_release(struct kobject *kobj)
2780 {
2781 complete(&ext4_feat->f_kobj_unregister);
2782 }
2783
2784 static ssize_t ext4_feat_show(struct kobject *kobj,
2785 struct attribute *attr, char *buf)
2786 {
2787 return snprintf(buf, PAGE_SIZE, "supported\n");
2788 }
2789
2790 /*
2791 * We can not use ext4_attr_show/store because it relies on the kobject
2792 * being embedded in the ext4_sb_info structure which is definitely not
2793 * true in this case.
2794 */
2795 static const struct sysfs_ops ext4_feat_ops = {
2796 .show = ext4_feat_show,
2797 .store = NULL,
2798 };
2799
2800 static struct kobj_type ext4_feat_ktype = {
2801 .default_attrs = ext4_feat_attrs,
2802 .sysfs_ops = &ext4_feat_ops,
2803 .release = ext4_feat_release,
2804 };
2805
2806 /*
2807 * Check whether this filesystem can be mounted based on
2808 * the features present and the RDONLY/RDWR mount requested.
2809 * Returns 1 if this filesystem can be mounted as requested,
2810 * 0 if it cannot be.
2811 */
2812 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2813 {
2814 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2815 ext4_msg(sb, KERN_ERR,
2816 "Couldn't mount because of "
2817 "unsupported optional features (%x)",
2818 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2819 ~EXT4_FEATURE_INCOMPAT_SUPP));
2820 return 0;
2821 }
2822
2823 if (readonly)
2824 return 1;
2825
2826 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2827 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2828 sb->s_flags |= MS_RDONLY;
2829 return 1;
2830 }
2831
2832 /* Check that feature set is OK for a read-write mount */
2833 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2834 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2835 "unsupported optional features (%x)",
2836 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2837 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2838 return 0;
2839 }
2840 /*
2841 * Large file size enabled file system can only be mounted
2842 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2843 */
2844 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2845 if (sizeof(blkcnt_t) < sizeof(u64)) {
2846 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2847 "cannot be mounted RDWR without "
2848 "CONFIG_LBDAF");
2849 return 0;
2850 }
2851 }
2852 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2853 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2854 ext4_msg(sb, KERN_ERR,
2855 "Can't support bigalloc feature without "
2856 "extents feature\n");
2857 return 0;
2858 }
2859
2860 #ifndef CONFIG_QUOTA
2861 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2862 !readonly) {
2863 ext4_msg(sb, KERN_ERR,
2864 "Filesystem with quota feature cannot be mounted RDWR "
2865 "without CONFIG_QUOTA");
2866 return 0;
2867 }
2868 #endif /* CONFIG_QUOTA */
2869 return 1;
2870 }
2871
2872 /*
2873 * This function is called once a day if we have errors logged
2874 * on the file system
2875 */
2876 static void print_daily_error_info(unsigned long arg)
2877 {
2878 struct super_block *sb = (struct super_block *) arg;
2879 struct ext4_sb_info *sbi;
2880 struct ext4_super_block *es;
2881
2882 sbi = EXT4_SB(sb);
2883 es = sbi->s_es;
2884
2885 if (es->s_error_count)
2886 /* fsck newer than v1.41.13 is needed to clean this condition. */
2887 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2888 le32_to_cpu(es->s_error_count));
2889 if (es->s_first_error_time) {
2890 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2891 sb->s_id, le32_to_cpu(es->s_first_error_time),
2892 (int) sizeof(es->s_first_error_func),
2893 es->s_first_error_func,
2894 le32_to_cpu(es->s_first_error_line));
2895 if (es->s_first_error_ino)
2896 printk(": inode %u",
2897 le32_to_cpu(es->s_first_error_ino));
2898 if (es->s_first_error_block)
2899 printk(": block %llu", (unsigned long long)
2900 le64_to_cpu(es->s_first_error_block));
2901 printk("\n");
2902 }
2903 if (es->s_last_error_time) {
2904 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2905 sb->s_id, le32_to_cpu(es->s_last_error_time),
2906 (int) sizeof(es->s_last_error_func),
2907 es->s_last_error_func,
2908 le32_to_cpu(es->s_last_error_line));
2909 if (es->s_last_error_ino)
2910 printk(": inode %u",
2911 le32_to_cpu(es->s_last_error_ino));
2912 if (es->s_last_error_block)
2913 printk(": block %llu", (unsigned long long)
2914 le64_to_cpu(es->s_last_error_block));
2915 printk("\n");
2916 }
2917 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2918 }
2919
2920 /* Find next suitable group and run ext4_init_inode_table */
2921 static int ext4_run_li_request(struct ext4_li_request *elr)
2922 {
2923 struct ext4_group_desc *gdp = NULL;
2924 ext4_group_t group, ngroups;
2925 struct super_block *sb;
2926 unsigned long timeout = 0;
2927 int ret = 0;
2928
2929 sb = elr->lr_super;
2930 ngroups = EXT4_SB(sb)->s_groups_count;
2931
2932 sb_start_write(sb);
2933 for (group = elr->lr_next_group; group < ngroups; group++) {
2934 gdp = ext4_get_group_desc(sb, group, NULL);
2935 if (!gdp) {
2936 ret = 1;
2937 break;
2938 }
2939
2940 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2941 break;
2942 }
2943
2944 if (group >= ngroups)
2945 ret = 1;
2946
2947 if (!ret) {
2948 timeout = jiffies;
2949 ret = ext4_init_inode_table(sb, group,
2950 elr->lr_timeout ? 0 : 1);
2951 if (elr->lr_timeout == 0) {
2952 timeout = (jiffies - timeout) *
2953 elr->lr_sbi->s_li_wait_mult;
2954 elr->lr_timeout = timeout;
2955 }
2956 elr->lr_next_sched = jiffies + elr->lr_timeout;
2957 elr->lr_next_group = group + 1;
2958 }
2959 sb_end_write(sb);
2960
2961 return ret;
2962 }
2963
2964 /*
2965 * Remove lr_request from the list_request and free the
2966 * request structure. Should be called with li_list_mtx held
2967 */
2968 static void ext4_remove_li_request(struct ext4_li_request *elr)
2969 {
2970 struct ext4_sb_info *sbi;
2971
2972 if (!elr)
2973 return;
2974
2975 sbi = elr->lr_sbi;
2976
2977 list_del(&elr->lr_request);
2978 sbi->s_li_request = NULL;
2979 kfree(elr);
2980 }
2981
2982 static void ext4_unregister_li_request(struct super_block *sb)
2983 {
2984 mutex_lock(&ext4_li_mtx);
2985 if (!ext4_li_info) {
2986 mutex_unlock(&ext4_li_mtx);
2987 return;
2988 }
2989
2990 mutex_lock(&ext4_li_info->li_list_mtx);
2991 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2992 mutex_unlock(&ext4_li_info->li_list_mtx);
2993 mutex_unlock(&ext4_li_mtx);
2994 }
2995
2996 static struct task_struct *ext4_lazyinit_task;
2997
2998 /*
2999 * This is the function where ext4lazyinit thread lives. It walks
3000 * through the request list searching for next scheduled filesystem.
3001 * When such a fs is found, run the lazy initialization request
3002 * (ext4_rn_li_request) and keep track of the time spend in this
3003 * function. Based on that time we compute next schedule time of
3004 * the request. When walking through the list is complete, compute
3005 * next waking time and put itself into sleep.
3006 */
3007 static int ext4_lazyinit_thread(void *arg)
3008 {
3009 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3010 struct list_head *pos, *n;
3011 struct ext4_li_request *elr;
3012 unsigned long next_wakeup, cur;
3013
3014 BUG_ON(NULL == eli);
3015
3016 cont_thread:
3017 while (true) {
3018 next_wakeup = MAX_JIFFY_OFFSET;
3019
3020 mutex_lock(&eli->li_list_mtx);
3021 if (list_empty(&eli->li_request_list)) {
3022 mutex_unlock(&eli->li_list_mtx);
3023 goto exit_thread;
3024 }
3025
3026 list_for_each_safe(pos, n, &eli->li_request_list) {
3027 elr = list_entry(pos, struct ext4_li_request,
3028 lr_request);
3029
3030 if (time_after_eq(jiffies, elr->lr_next_sched)) {
3031 if (ext4_run_li_request(elr) != 0) {
3032 /* error, remove the lazy_init job */
3033 ext4_remove_li_request(elr);
3034 continue;
3035 }
3036 }
3037
3038 if (time_before(elr->lr_next_sched, next_wakeup))
3039 next_wakeup = elr->lr_next_sched;
3040 }
3041 mutex_unlock(&eli->li_list_mtx);
3042
3043 try_to_freeze();
3044
3045 cur = jiffies;
3046 if ((time_after_eq(cur, next_wakeup)) ||
3047 (MAX_JIFFY_OFFSET == next_wakeup)) {
3048 cond_resched();
3049 continue;
3050 }
3051
3052 schedule_timeout_interruptible(next_wakeup - cur);
3053
3054 if (kthread_should_stop()) {
3055 ext4_clear_request_list();
3056 goto exit_thread;
3057 }
3058 }
3059
3060 exit_thread:
3061 /*
3062 * It looks like the request list is empty, but we need
3063 * to check it under the li_list_mtx lock, to prevent any
3064 * additions into it, and of course we should lock ext4_li_mtx
3065 * to atomically free the list and ext4_li_info, because at
3066 * this point another ext4 filesystem could be registering
3067 * new one.
3068 */
3069 mutex_lock(&ext4_li_mtx);
3070 mutex_lock(&eli->li_list_mtx);
3071 if (!list_empty(&eli->li_request_list)) {
3072 mutex_unlock(&eli->li_list_mtx);
3073 mutex_unlock(&ext4_li_mtx);
3074 goto cont_thread;
3075 }
3076 mutex_unlock(&eli->li_list_mtx);
3077 kfree(ext4_li_info);
3078 ext4_li_info = NULL;
3079 mutex_unlock(&ext4_li_mtx);
3080
3081 return 0;
3082 }
3083
3084 static void ext4_clear_request_list(void)
3085 {
3086 struct list_head *pos, *n;
3087 struct ext4_li_request *elr;
3088
3089 mutex_lock(&ext4_li_info->li_list_mtx);
3090 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3091 elr = list_entry(pos, struct ext4_li_request,
3092 lr_request);
3093 ext4_remove_li_request(elr);
3094 }
3095 mutex_unlock(&ext4_li_info->li_list_mtx);
3096 }
3097
3098 static int ext4_run_lazyinit_thread(void)
3099 {
3100 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3101 ext4_li_info, "ext4lazyinit");
3102 if (IS_ERR(ext4_lazyinit_task)) {
3103 int err = PTR_ERR(ext4_lazyinit_task);
3104 ext4_clear_request_list();
3105 kfree(ext4_li_info);
3106 ext4_li_info = NULL;
3107 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3108 "initialization thread\n",
3109 err);
3110 return err;
3111 }
3112 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3113 return 0;
3114 }
3115
3116 /*
3117 * Check whether it make sense to run itable init. thread or not.
3118 * If there is at least one uninitialized inode table, return
3119 * corresponding group number, else the loop goes through all
3120 * groups and return total number of groups.
3121 */
3122 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3123 {
3124 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3125 struct ext4_group_desc *gdp = NULL;
3126
3127 for (group = 0; group < ngroups; group++) {
3128 gdp = ext4_get_group_desc(sb, group, NULL);
3129 if (!gdp)
3130 continue;
3131
3132 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3133 break;
3134 }
3135
3136 return group;
3137 }
3138
3139 static int ext4_li_info_new(void)
3140 {
3141 struct ext4_lazy_init *eli = NULL;
3142
3143 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3144 if (!eli)
3145 return -ENOMEM;
3146
3147 INIT_LIST_HEAD(&eli->li_request_list);
3148 mutex_init(&eli->li_list_mtx);
3149
3150 eli->li_state |= EXT4_LAZYINIT_QUIT;
3151
3152 ext4_li_info = eli;
3153
3154 return 0;
3155 }
3156
3157 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3158 ext4_group_t start)
3159 {
3160 struct ext4_sb_info *sbi = EXT4_SB(sb);
3161 struct ext4_li_request *elr;
3162
3163 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3164 if (!elr)
3165 return NULL;
3166
3167 elr->lr_super = sb;
3168 elr->lr_sbi = sbi;
3169 elr->lr_next_group = start;
3170
3171 /*
3172 * Randomize first schedule time of the request to
3173 * spread the inode table initialization requests
3174 * better.
3175 */
3176 elr->lr_next_sched = jiffies + (prandom_u32() %
3177 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3178 return elr;
3179 }
3180
3181 int ext4_register_li_request(struct super_block *sb,
3182 ext4_group_t first_not_zeroed)
3183 {
3184 struct ext4_sb_info *sbi = EXT4_SB(sb);
3185 struct ext4_li_request *elr = NULL;
3186 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3187 int ret = 0;
3188
3189 mutex_lock(&ext4_li_mtx);
3190 if (sbi->s_li_request != NULL) {
3191 /*
3192 * Reset timeout so it can be computed again, because
3193 * s_li_wait_mult might have changed.
3194 */
3195 sbi->s_li_request->lr_timeout = 0;
3196 goto out;
3197 }
3198
3199 if (first_not_zeroed == ngroups ||
3200 (sb->s_flags & MS_RDONLY) ||
3201 !test_opt(sb, INIT_INODE_TABLE))
3202 goto out;
3203
3204 elr = ext4_li_request_new(sb, first_not_zeroed);
3205 if (!elr) {
3206 ret = -ENOMEM;
3207 goto out;
3208 }
3209
3210 if (NULL == ext4_li_info) {
3211 ret = ext4_li_info_new();
3212 if (ret)
3213 goto out;
3214 }
3215
3216 mutex_lock(&ext4_li_info->li_list_mtx);
3217 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3218 mutex_unlock(&ext4_li_info->li_list_mtx);
3219
3220 sbi->s_li_request = elr;
3221 /*
3222 * set elr to NULL here since it has been inserted to
3223 * the request_list and the removal and free of it is
3224 * handled by ext4_clear_request_list from now on.
3225 */
3226 elr = NULL;
3227
3228 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3229 ret = ext4_run_lazyinit_thread();
3230 if (ret)
3231 goto out;
3232 }
3233 out:
3234 mutex_unlock(&ext4_li_mtx);
3235 if (ret)
3236 kfree(elr);
3237 return ret;
3238 }
3239
3240 /*
3241 * We do not need to lock anything since this is called on
3242 * module unload.
3243 */
3244 static void ext4_destroy_lazyinit_thread(void)
3245 {
3246 /*
3247 * If thread exited earlier
3248 * there's nothing to be done.
3249 */
3250 if (!ext4_li_info || !ext4_lazyinit_task)
3251 return;
3252
3253 kthread_stop(ext4_lazyinit_task);
3254 }
3255
3256 static int set_journal_csum_feature_set(struct super_block *sb)
3257 {
3258 int ret = 1;
3259 int compat, incompat;
3260 struct ext4_sb_info *sbi = EXT4_SB(sb);
3261
3262 if (ext4_has_metadata_csum(sb)) {
3263 /* journal checksum v3 */
3264 compat = 0;
3265 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3266 } else {
3267 /* journal checksum v1 */
3268 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3269 incompat = 0;
3270 }
3271
3272 jbd2_journal_clear_features(sbi->s_journal,
3273 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3274 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3275 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3276 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3277 ret = jbd2_journal_set_features(sbi->s_journal,
3278 compat, 0,
3279 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3280 incompat);
3281 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3282 ret = jbd2_journal_set_features(sbi->s_journal,
3283 compat, 0,
3284 incompat);
3285 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3286 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3287 } else {
3288 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3289 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3290 }
3291
3292 return ret;
3293 }
3294
3295 /*
3296 * Note: calculating the overhead so we can be compatible with
3297 * historical BSD practice is quite difficult in the face of
3298 * clusters/bigalloc. This is because multiple metadata blocks from
3299 * different block group can end up in the same allocation cluster.
3300 * Calculating the exact overhead in the face of clustered allocation
3301 * requires either O(all block bitmaps) in memory or O(number of block
3302 * groups**2) in time. We will still calculate the superblock for
3303 * older file systems --- and if we come across with a bigalloc file
3304 * system with zero in s_overhead_clusters the estimate will be close to
3305 * correct especially for very large cluster sizes --- but for newer
3306 * file systems, it's better to calculate this figure once at mkfs
3307 * time, and store it in the superblock. If the superblock value is
3308 * present (even for non-bigalloc file systems), we will use it.
3309 */
3310 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3311 char *buf)
3312 {
3313 struct ext4_sb_info *sbi = EXT4_SB(sb);
3314 struct ext4_group_desc *gdp;
3315 ext4_fsblk_t first_block, last_block, b;
3316 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3317 int s, j, count = 0;
3318
3319 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3320 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3321 sbi->s_itb_per_group + 2);
3322
3323 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3324 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3325 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3326 for (i = 0; i < ngroups; i++) {
3327 gdp = ext4_get_group_desc(sb, i, NULL);
3328 b = ext4_block_bitmap(sb, gdp);
3329 if (b >= first_block && b <= last_block) {
3330 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3331 count++;
3332 }
3333 b = ext4_inode_bitmap(sb, gdp);
3334 if (b >= first_block && b <= last_block) {
3335 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3336 count++;
3337 }
3338 b = ext4_inode_table(sb, gdp);
3339 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3340 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3341 int c = EXT4_B2C(sbi, b - first_block);
3342 ext4_set_bit(c, buf);
3343 count++;
3344 }
3345 if (i != grp)
3346 continue;
3347 s = 0;
3348 if (ext4_bg_has_super(sb, grp)) {
3349 ext4_set_bit(s++, buf);
3350 count++;
3351 }
3352 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3353 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3354 count++;
3355 }
3356 }
3357 if (!count)
3358 return 0;
3359 return EXT4_CLUSTERS_PER_GROUP(sb) -
3360 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3361 }
3362
3363 /*
3364 * Compute the overhead and stash it in sbi->s_overhead
3365 */
3366 int ext4_calculate_overhead(struct super_block *sb)
3367 {
3368 struct ext4_sb_info *sbi = EXT4_SB(sb);
3369 struct ext4_super_block *es = sbi->s_es;
3370 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3371 ext4_fsblk_t overhead = 0;
3372 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3373
3374 if (!buf)
3375 return -ENOMEM;
3376
3377 /*
3378 * Compute the overhead (FS structures). This is constant
3379 * for a given filesystem unless the number of block groups
3380 * changes so we cache the previous value until it does.
3381 */
3382
3383 /*
3384 * All of the blocks before first_data_block are overhead
3385 */
3386 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3387
3388 /*
3389 * Add the overhead found in each block group
3390 */
3391 for (i = 0; i < ngroups; i++) {
3392 int blks;
3393
3394 blks = count_overhead(sb, i, buf);
3395 overhead += blks;
3396 if (blks)
3397 memset(buf, 0, PAGE_SIZE);
3398 cond_resched();
3399 }
3400 /* Add the internal journal blocks as well */
3401 if (sbi->s_journal && !sbi->journal_bdev)
3402 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3403
3404 sbi->s_overhead = overhead;
3405 smp_wmb();
3406 free_page((unsigned long) buf);
3407 return 0;
3408 }
3409
3410
3411 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3412 {
3413 ext4_fsblk_t resv_clusters;
3414
3415 /*
3416 * There's no need to reserve anything when we aren't using extents.
3417 * The space estimates are exact, there are no unwritten extents,
3418 * hole punching doesn't need new metadata... This is needed especially
3419 * to keep ext2/3 backward compatibility.
3420 */
3421 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3422 return 0;
3423 /*
3424 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3425 * This should cover the situations where we can not afford to run
3426 * out of space like for example punch hole, or converting
3427 * unwritten extents in delalloc path. In most cases such
3428 * allocation would require 1, or 2 blocks, higher numbers are
3429 * very rare.
3430 */
3431 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3432 EXT4_SB(sb)->s_cluster_bits;
3433
3434 do_div(resv_clusters, 50);
3435 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3436
3437 return resv_clusters;
3438 }
3439
3440
3441 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3442 {
3443 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3444 sbi->s_cluster_bits;
3445
3446 if (count >= clusters)
3447 return -EINVAL;
3448
3449 atomic64_set(&sbi->s_resv_clusters, count);
3450 return 0;
3451 }
3452
3453 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3454 {
3455 char *orig_data = kstrdup(data, GFP_KERNEL);
3456 struct buffer_head *bh;
3457 struct ext4_super_block *es = NULL;
3458 struct ext4_sb_info *sbi;
3459 ext4_fsblk_t block;
3460 ext4_fsblk_t sb_block = get_sb_block(&data);
3461 ext4_fsblk_t logical_sb_block;
3462 unsigned long offset = 0;
3463 unsigned long journal_devnum = 0;
3464 unsigned long def_mount_opts;
3465 struct inode *root;
3466 const char *descr;
3467 int ret = -ENOMEM;
3468 int blocksize, clustersize;
3469 unsigned int db_count;
3470 unsigned int i;
3471 int needs_recovery, has_huge_files, has_bigalloc;
3472 __u64 blocks_count;
3473 int err = 0;
3474 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3475 ext4_group_t first_not_zeroed;
3476
3477 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3478 if (!sbi)
3479 goto out_free_orig;
3480
3481 sbi->s_blockgroup_lock =
3482 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3483 if (!sbi->s_blockgroup_lock) {
3484 kfree(sbi);
3485 goto out_free_orig;
3486 }
3487 sb->s_fs_info = sbi;
3488 sbi->s_sb = sb;
3489 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3490 sbi->s_sb_block = sb_block;
3491 if (sb->s_bdev->bd_part)
3492 sbi->s_sectors_written_start =
3493 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3494
3495 /* Cleanup superblock name */
3496 strreplace(sb->s_id, '/', '!');
3497
3498 /* -EINVAL is default */
3499 ret = -EINVAL;
3500 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3501 if (!blocksize) {
3502 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3503 goto out_fail;
3504 }
3505
3506 /*
3507 * The ext4 superblock will not be buffer aligned for other than 1kB
3508 * block sizes. We need to calculate the offset from buffer start.
3509 */
3510 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3511 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3512 offset = do_div(logical_sb_block, blocksize);
3513 } else {
3514 logical_sb_block = sb_block;
3515 }
3516
3517 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3518 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3519 goto out_fail;
3520 }
3521 /*
3522 * Note: s_es must be initialized as soon as possible because
3523 * some ext4 macro-instructions depend on its value
3524 */
3525 es = (struct ext4_super_block *) (bh->b_data + offset);
3526 sbi->s_es = es;
3527 sb->s_magic = le16_to_cpu(es->s_magic);
3528 if (sb->s_magic != EXT4_SUPER_MAGIC)
3529 goto cantfind_ext4;
3530 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3531
3532 /* Warn if metadata_csum and gdt_csum are both set. */
3533 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3534 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3535 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3536 ext4_warning(sb, "metadata_csum and uninit_bg are "
3537 "redundant flags; please run fsck.");
3538
3539 /* Check for a known checksum algorithm */
3540 if (!ext4_verify_csum_type(sb, es)) {
3541 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3542 "unknown checksum algorithm.");
3543 silent = 1;
3544 goto cantfind_ext4;
3545 }
3546
3547 /* Load the checksum driver */
3548 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3549 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3550 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3551 if (IS_ERR(sbi->s_chksum_driver)) {
3552 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3553 ret = PTR_ERR(sbi->s_chksum_driver);
3554 sbi->s_chksum_driver = NULL;
3555 goto failed_mount;
3556 }
3557 }
3558
3559 /* Check superblock checksum */
3560 if (!ext4_superblock_csum_verify(sb, es)) {
3561 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3562 "invalid superblock checksum. Run e2fsck?");
3563 silent = 1;
3564 goto cantfind_ext4;
3565 }
3566
3567 /* Precompute checksum seed for all metadata */
3568 if (ext4_has_metadata_csum(sb))
3569 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3570 sizeof(es->s_uuid));
3571
3572 /* Set defaults before we parse the mount options */
3573 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3574 set_opt(sb, INIT_INODE_TABLE);
3575 if (def_mount_opts & EXT4_DEFM_DEBUG)
3576 set_opt(sb, DEBUG);
3577 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3578 set_opt(sb, GRPID);
3579 if (def_mount_opts & EXT4_DEFM_UID16)
3580 set_opt(sb, NO_UID32);
3581 /* xattr user namespace & acls are now defaulted on */
3582 set_opt(sb, XATTR_USER);
3583 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3584 set_opt(sb, POSIX_ACL);
3585 #endif
3586 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3587 if (ext4_has_metadata_csum(sb))
3588 set_opt(sb, JOURNAL_CHECKSUM);
3589
3590 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3591 set_opt(sb, JOURNAL_DATA);
3592 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3593 set_opt(sb, ORDERED_DATA);
3594 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3595 set_opt(sb, WRITEBACK_DATA);
3596
3597 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3598 set_opt(sb, ERRORS_PANIC);
3599 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3600 set_opt(sb, ERRORS_CONT);
3601 else
3602 set_opt(sb, ERRORS_RO);
3603 /* block_validity enabled by default; disable with noblock_validity */
3604 set_opt(sb, BLOCK_VALIDITY);
3605 if (def_mount_opts & EXT4_DEFM_DISCARD)
3606 set_opt(sb, DISCARD);
3607
3608 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3609 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3610 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3611 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3612 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3613
3614 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3615 set_opt(sb, BARRIER);
3616
3617 /*
3618 * enable delayed allocation by default
3619 * Use -o nodelalloc to turn it off
3620 */
3621 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3622 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3623 set_opt(sb, DELALLOC);
3624
3625 /*
3626 * set default s_li_wait_mult for lazyinit, for the case there is
3627 * no mount option specified.
3628 */
3629 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3630
3631 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3632 &journal_devnum, &journal_ioprio, 0)) {
3633 ext4_msg(sb, KERN_WARNING,
3634 "failed to parse options in superblock: %s",
3635 sbi->s_es->s_mount_opts);
3636 }
3637 sbi->s_def_mount_opt = sbi->s_mount_opt;
3638 if (!parse_options((char *) data, sb, &journal_devnum,
3639 &journal_ioprio, 0))
3640 goto failed_mount;
3641
3642 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3643 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3644 "with data=journal disables delayed "
3645 "allocation and O_DIRECT support!\n");
3646 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3647 ext4_msg(sb, KERN_ERR, "can't mount with "
3648 "both data=journal and delalloc");
3649 goto failed_mount;
3650 }
3651 if (test_opt(sb, DIOREAD_NOLOCK)) {
3652 ext4_msg(sb, KERN_ERR, "can't mount with "
3653 "both data=journal and dioread_nolock");
3654 goto failed_mount;
3655 }
3656 if (test_opt(sb, DAX)) {
3657 ext4_msg(sb, KERN_ERR, "can't mount with "
3658 "both data=journal and dax");
3659 goto failed_mount;
3660 }
3661 if (test_opt(sb, DELALLOC))
3662 clear_opt(sb, DELALLOC);
3663 } else {
3664 sb->s_iflags |= SB_I_CGROUPWB;
3665 }
3666
3667 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3668 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3669
3670 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3671 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3672 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3673 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3674 ext4_msg(sb, KERN_WARNING,
3675 "feature flags set on rev 0 fs, "
3676 "running e2fsck is recommended");
3677
3678 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3679 set_opt2(sb, HURD_COMPAT);
3680 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3681 EXT4_FEATURE_INCOMPAT_64BIT)) {
3682 ext4_msg(sb, KERN_ERR,
3683 "The Hurd can't support 64-bit file systems");
3684 goto failed_mount;
3685 }
3686 }
3687
3688 if (IS_EXT2_SB(sb)) {
3689 if (ext2_feature_set_ok(sb))
3690 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3691 "using the ext4 subsystem");
3692 else {
3693 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3694 "to feature incompatibilities");
3695 goto failed_mount;
3696 }
3697 }
3698
3699 if (IS_EXT3_SB(sb)) {
3700 if (ext3_feature_set_ok(sb))
3701 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3702 "using the ext4 subsystem");
3703 else {
3704 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3705 "to feature incompatibilities");
3706 goto failed_mount;
3707 }
3708 }
3709
3710 /*
3711 * Check feature flags regardless of the revision level, since we
3712 * previously didn't change the revision level when setting the flags,
3713 * so there is a chance incompat flags are set on a rev 0 filesystem.
3714 */
3715 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3716 goto failed_mount;
3717
3718 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3719 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3720 blocksize > EXT4_MAX_BLOCK_SIZE) {
3721 ext4_msg(sb, KERN_ERR,
3722 "Unsupported filesystem blocksize %d", blocksize);
3723 goto failed_mount;
3724 }
3725
3726 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3727 if (blocksize != PAGE_SIZE) {
3728 ext4_msg(sb, KERN_ERR,
3729 "error: unsupported blocksize for dax");
3730 goto failed_mount;
3731 }
3732 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3733 ext4_msg(sb, KERN_ERR,
3734 "error: device does not support dax");
3735 goto failed_mount;
3736 }
3737 }
3738
3739 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3740 es->s_encryption_level) {
3741 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3742 es->s_encryption_level);
3743 goto failed_mount;
3744 }
3745
3746 if (sb->s_blocksize != blocksize) {
3747 /* Validate the filesystem blocksize */
3748 if (!sb_set_blocksize(sb, blocksize)) {
3749 ext4_msg(sb, KERN_ERR, "bad block size %d",
3750 blocksize);
3751 goto failed_mount;
3752 }
3753
3754 brelse(bh);
3755 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3756 offset = do_div(logical_sb_block, blocksize);
3757 bh = sb_bread_unmovable(sb, logical_sb_block);
3758 if (!bh) {
3759 ext4_msg(sb, KERN_ERR,
3760 "Can't read superblock on 2nd try");
3761 goto failed_mount;
3762 }
3763 es = (struct ext4_super_block *)(bh->b_data + offset);
3764 sbi->s_es = es;
3765 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3766 ext4_msg(sb, KERN_ERR,
3767 "Magic mismatch, very weird!");
3768 goto failed_mount;
3769 }
3770 }
3771
3772 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3773 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3774 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3775 has_huge_files);
3776 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3777
3778 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3779 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3780 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3781 } else {
3782 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3783 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3784 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3785 (!is_power_of_2(sbi->s_inode_size)) ||
3786 (sbi->s_inode_size > blocksize)) {
3787 ext4_msg(sb, KERN_ERR,
3788 "unsupported inode size: %d",
3789 sbi->s_inode_size);
3790 goto failed_mount;
3791 }
3792 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3793 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3794 }
3795
3796 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3797 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3798 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3799 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3800 !is_power_of_2(sbi->s_desc_size)) {
3801 ext4_msg(sb, KERN_ERR,
3802 "unsupported descriptor size %lu",
3803 sbi->s_desc_size);
3804 goto failed_mount;
3805 }
3806 } else
3807 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3808
3809 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3810 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3811 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3812 goto cantfind_ext4;
3813
3814 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3815 if (sbi->s_inodes_per_block == 0)
3816 goto cantfind_ext4;
3817 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3818 sbi->s_inodes_per_block;
3819 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3820 sbi->s_sbh = bh;
3821 sbi->s_mount_state = le16_to_cpu(es->s_state);
3822 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3823 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3824
3825 for (i = 0; i < 4; i++)
3826 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3827 sbi->s_def_hash_version = es->s_def_hash_version;
3828 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3829 i = le32_to_cpu(es->s_flags);
3830 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3831 sbi->s_hash_unsigned = 3;
3832 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3833 #ifdef __CHAR_UNSIGNED__
3834 if (!(sb->s_flags & MS_RDONLY))
3835 es->s_flags |=
3836 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3837 sbi->s_hash_unsigned = 3;
3838 #else
3839 if (!(sb->s_flags & MS_RDONLY))
3840 es->s_flags |=
3841 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3842 #endif
3843 }
3844 }
3845
3846 /* Handle clustersize */
3847 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3848 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3849 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3850 if (has_bigalloc) {
3851 if (clustersize < blocksize) {
3852 ext4_msg(sb, KERN_ERR,
3853 "cluster size (%d) smaller than "
3854 "block size (%d)", clustersize, blocksize);
3855 goto failed_mount;
3856 }
3857 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3858 le32_to_cpu(es->s_log_block_size);
3859 sbi->s_clusters_per_group =
3860 le32_to_cpu(es->s_clusters_per_group);
3861 if (sbi->s_clusters_per_group > blocksize * 8) {
3862 ext4_msg(sb, KERN_ERR,
3863 "#clusters per group too big: %lu",
3864 sbi->s_clusters_per_group);
3865 goto failed_mount;
3866 }
3867 if (sbi->s_blocks_per_group !=
3868 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3869 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3870 "clusters per group (%lu) inconsistent",
3871 sbi->s_blocks_per_group,
3872 sbi->s_clusters_per_group);
3873 goto failed_mount;
3874 }
3875 } else {
3876 if (clustersize != blocksize) {
3877 ext4_warning(sb, "fragment/cluster size (%d) != "
3878 "block size (%d)", clustersize,
3879 blocksize);
3880 clustersize = blocksize;
3881 }
3882 if (sbi->s_blocks_per_group > blocksize * 8) {
3883 ext4_msg(sb, KERN_ERR,
3884 "#blocks per group too big: %lu",
3885 sbi->s_blocks_per_group);
3886 goto failed_mount;
3887 }
3888 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3889 sbi->s_cluster_bits = 0;
3890 }
3891 sbi->s_cluster_ratio = clustersize / blocksize;
3892
3893 if (sbi->s_inodes_per_group > blocksize * 8) {
3894 ext4_msg(sb, KERN_ERR,
3895 "#inodes per group too big: %lu",
3896 sbi->s_inodes_per_group);
3897 goto failed_mount;
3898 }
3899
3900 /* Do we have standard group size of clustersize * 8 blocks ? */
3901 if (sbi->s_blocks_per_group == clustersize << 3)
3902 set_opt2(sb, STD_GROUP_SIZE);
3903
3904 /*
3905 * Test whether we have more sectors than will fit in sector_t,
3906 * and whether the max offset is addressable by the page cache.
3907 */
3908 err = generic_check_addressable(sb->s_blocksize_bits,
3909 ext4_blocks_count(es));
3910 if (err) {
3911 ext4_msg(sb, KERN_ERR, "filesystem"
3912 " too large to mount safely on this system");
3913 if (sizeof(sector_t) < 8)
3914 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3915 goto failed_mount;
3916 }
3917
3918 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3919 goto cantfind_ext4;
3920
3921 /* check blocks count against device size */
3922 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3923 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3924 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3925 "exceeds size of device (%llu blocks)",
3926 ext4_blocks_count(es), blocks_count);
3927 goto failed_mount;
3928 }
3929
3930 /*
3931 * It makes no sense for the first data block to be beyond the end
3932 * of the filesystem.
3933 */
3934 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3935 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3936 "block %u is beyond end of filesystem (%llu)",
3937 le32_to_cpu(es->s_first_data_block),
3938 ext4_blocks_count(es));
3939 goto failed_mount;
3940 }
3941 blocks_count = (ext4_blocks_count(es) -
3942 le32_to_cpu(es->s_first_data_block) +
3943 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3944 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3945 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3946 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3947 "(block count %llu, first data block %u, "
3948 "blocks per group %lu)", sbi->s_groups_count,
3949 ext4_blocks_count(es),
3950 le32_to_cpu(es->s_first_data_block),
3951 EXT4_BLOCKS_PER_GROUP(sb));
3952 goto failed_mount;
3953 }
3954 sbi->s_groups_count = blocks_count;
3955 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3956 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3957 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3958 EXT4_DESC_PER_BLOCK(sb);
3959 sbi->s_group_desc = ext4_kvmalloc(db_count *
3960 sizeof(struct buffer_head *),
3961 GFP_KERNEL);
3962 if (sbi->s_group_desc == NULL) {
3963 ext4_msg(sb, KERN_ERR, "not enough memory");
3964 ret = -ENOMEM;
3965 goto failed_mount;
3966 }
3967
3968 if (ext4_proc_root)
3969 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3970
3971 if (sbi->s_proc)
3972 proc_create_data("options", S_IRUGO, sbi->s_proc,
3973 &ext4_seq_options_fops, sb);
3974
3975 bgl_lock_init(sbi->s_blockgroup_lock);
3976
3977 for (i = 0; i < db_count; i++) {
3978 block = descriptor_loc(sb, logical_sb_block, i);
3979 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3980 if (!sbi->s_group_desc[i]) {
3981 ext4_msg(sb, KERN_ERR,
3982 "can't read group descriptor %d", i);
3983 db_count = i;
3984 goto failed_mount2;
3985 }
3986 }
3987 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3988 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3989 goto failed_mount2;
3990 }
3991
3992 sbi->s_gdb_count = db_count;
3993 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3994 spin_lock_init(&sbi->s_next_gen_lock);
3995
3996 setup_timer(&sbi->s_err_report, print_daily_error_info,
3997 (unsigned long) sb);
3998
3999 /* Register extent status tree shrinker */
4000 if (ext4_es_register_shrinker(sbi))
4001 goto failed_mount3;
4002
4003 sbi->s_stripe = ext4_get_stripe_size(sbi);
4004 sbi->s_extent_max_zeroout_kb = 32;
4005
4006 /*
4007 * set up enough so that it can read an inode
4008 */
4009 sb->s_op = &ext4_sops;
4010 sb->s_export_op = &ext4_export_ops;
4011 sb->s_xattr = ext4_xattr_handlers;
4012 #ifdef CONFIG_QUOTA
4013 sb->dq_op = &ext4_quota_operations;
4014 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4015 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4016 else
4017 sb->s_qcop = &ext4_qctl_operations;
4018 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
4019 #endif
4020 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4021
4022 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4023 mutex_init(&sbi->s_orphan_lock);
4024
4025 sb->s_root = NULL;
4026
4027 needs_recovery = (es->s_last_orphan != 0 ||
4028 EXT4_HAS_INCOMPAT_FEATURE(sb,
4029 EXT4_FEATURE_INCOMPAT_RECOVER));
4030
4031 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
4032 !(sb->s_flags & MS_RDONLY))
4033 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4034 goto failed_mount3a;
4035
4036 /*
4037 * The first inode we look at is the journal inode. Don't try
4038 * root first: it may be modified in the journal!
4039 */
4040 if (!test_opt(sb, NOLOAD) &&
4041 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4042 if (ext4_load_journal(sb, es, journal_devnum))
4043 goto failed_mount3a;
4044 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4045 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4046 ext4_msg(sb, KERN_ERR, "required journal recovery "
4047 "suppressed and not mounted read-only");
4048 goto failed_mount_wq;
4049 } else {
4050 clear_opt(sb, DATA_FLAGS);
4051 sbi->s_journal = NULL;
4052 needs_recovery = 0;
4053 goto no_journal;
4054 }
4055
4056 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
4057 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4058 JBD2_FEATURE_INCOMPAT_64BIT)) {
4059 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4060 goto failed_mount_wq;
4061 }
4062
4063 if (!set_journal_csum_feature_set(sb)) {
4064 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4065 "feature set");
4066 goto failed_mount_wq;
4067 }
4068
4069 /* We have now updated the journal if required, so we can
4070 * validate the data journaling mode. */
4071 switch (test_opt(sb, DATA_FLAGS)) {
4072 case 0:
4073 /* No mode set, assume a default based on the journal
4074 * capabilities: ORDERED_DATA if the journal can
4075 * cope, else JOURNAL_DATA
4076 */
4077 if (jbd2_journal_check_available_features
4078 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4079 set_opt(sb, ORDERED_DATA);
4080 else
4081 set_opt(sb, JOURNAL_DATA);
4082 break;
4083
4084 case EXT4_MOUNT_ORDERED_DATA:
4085 case EXT4_MOUNT_WRITEBACK_DATA:
4086 if (!jbd2_journal_check_available_features
4087 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4088 ext4_msg(sb, KERN_ERR, "Journal does not support "
4089 "requested data journaling mode");
4090 goto failed_mount_wq;
4091 }
4092 default:
4093 break;
4094 }
4095 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4096
4097 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4098
4099 no_journal:
4100 if (ext4_mballoc_ready) {
4101 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4102 if (!sbi->s_mb_cache) {
4103 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4104 goto failed_mount_wq;
4105 }
4106 }
4107
4108 if ((DUMMY_ENCRYPTION_ENABLED(sbi) ||
4109 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) &&
4110 (blocksize != PAGE_CACHE_SIZE)) {
4111 ext4_msg(sb, KERN_ERR,
4112 "Unsupported blocksize for fs encryption");
4113 goto failed_mount_wq;
4114 }
4115
4116 if (DUMMY_ENCRYPTION_ENABLED(sbi) &&
4117 !(sb->s_flags & MS_RDONLY) &&
4118 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
4119 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
4120 ext4_commit_super(sb, 1);
4121 }
4122
4123 /*
4124 * Get the # of file system overhead blocks from the
4125 * superblock if present.
4126 */
4127 if (es->s_overhead_clusters)
4128 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4129 else {
4130 err = ext4_calculate_overhead(sb);
4131 if (err)
4132 goto failed_mount_wq;
4133 }
4134
4135 /*
4136 * The maximum number of concurrent works can be high and
4137 * concurrency isn't really necessary. Limit it to 1.
4138 */
4139 EXT4_SB(sb)->rsv_conversion_wq =
4140 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4141 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4142 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4143 ret = -ENOMEM;
4144 goto failed_mount4;
4145 }
4146
4147 /*
4148 * The jbd2_journal_load will have done any necessary log recovery,
4149 * so we can safely mount the rest of the filesystem now.
4150 */
4151
4152 root = ext4_iget(sb, EXT4_ROOT_INO);
4153 if (IS_ERR(root)) {
4154 ext4_msg(sb, KERN_ERR, "get root inode failed");
4155 ret = PTR_ERR(root);
4156 root = NULL;
4157 goto failed_mount4;
4158 }
4159 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4160 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4161 iput(root);
4162 goto failed_mount4;
4163 }
4164 sb->s_root = d_make_root(root);
4165 if (!sb->s_root) {
4166 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4167 ret = -ENOMEM;
4168 goto failed_mount4;
4169 }
4170
4171 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4172 sb->s_flags |= MS_RDONLY;
4173
4174 /* determine the minimum size of new large inodes, if present */
4175 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4176 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4177 EXT4_GOOD_OLD_INODE_SIZE;
4178 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4179 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4180 if (sbi->s_want_extra_isize <
4181 le16_to_cpu(es->s_want_extra_isize))
4182 sbi->s_want_extra_isize =
4183 le16_to_cpu(es->s_want_extra_isize);
4184 if (sbi->s_want_extra_isize <
4185 le16_to_cpu(es->s_min_extra_isize))
4186 sbi->s_want_extra_isize =
4187 le16_to_cpu(es->s_min_extra_isize);
4188 }
4189 }
4190 /* Check if enough inode space is available */
4191 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4192 sbi->s_inode_size) {
4193 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4194 EXT4_GOOD_OLD_INODE_SIZE;
4195 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4196 "available");
4197 }
4198
4199 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4200 if (err) {
4201 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4202 "reserved pool", ext4_calculate_resv_clusters(sb));
4203 goto failed_mount4a;
4204 }
4205
4206 err = ext4_setup_system_zone(sb);
4207 if (err) {
4208 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4209 "zone (%d)", err);
4210 goto failed_mount4a;
4211 }
4212
4213 ext4_ext_init(sb);
4214 err = ext4_mb_init(sb);
4215 if (err) {
4216 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4217 err);
4218 goto failed_mount5;
4219 }
4220
4221 block = ext4_count_free_clusters(sb);
4222 ext4_free_blocks_count_set(sbi->s_es,
4223 EXT4_C2B(sbi, block));
4224 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4225 GFP_KERNEL);
4226 if (!err) {
4227 unsigned long freei = ext4_count_free_inodes(sb);
4228 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4229 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4230 GFP_KERNEL);
4231 }
4232 if (!err)
4233 err = percpu_counter_init(&sbi->s_dirs_counter,
4234 ext4_count_dirs(sb), GFP_KERNEL);
4235 if (!err)
4236 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4237 GFP_KERNEL);
4238 if (err) {
4239 ext4_msg(sb, KERN_ERR, "insufficient memory");
4240 goto failed_mount6;
4241 }
4242
4243 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4244 if (!ext4_fill_flex_info(sb)) {
4245 ext4_msg(sb, KERN_ERR,
4246 "unable to initialize "
4247 "flex_bg meta info!");
4248 goto failed_mount6;
4249 }
4250
4251 err = ext4_register_li_request(sb, first_not_zeroed);
4252 if (err)
4253 goto failed_mount6;
4254
4255 sbi->s_kobj.kset = ext4_kset;
4256 init_completion(&sbi->s_kobj_unregister);
4257 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4258 "%s", sb->s_id);
4259 if (err)
4260 goto failed_mount7;
4261
4262 #ifdef CONFIG_QUOTA
4263 /* Enable quota usage during mount. */
4264 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4265 !(sb->s_flags & MS_RDONLY)) {
4266 err = ext4_enable_quotas(sb);
4267 if (err)
4268 goto failed_mount8;
4269 }
4270 #endif /* CONFIG_QUOTA */
4271
4272 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4273 ext4_orphan_cleanup(sb, es);
4274 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4275 if (needs_recovery) {
4276 ext4_msg(sb, KERN_INFO, "recovery complete");
4277 ext4_mark_recovery_complete(sb, es);
4278 }
4279 if (EXT4_SB(sb)->s_journal) {
4280 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4281 descr = " journalled data mode";
4282 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4283 descr = " ordered data mode";
4284 else
4285 descr = " writeback data mode";
4286 } else
4287 descr = "out journal";
4288
4289 if (test_opt(sb, DISCARD)) {
4290 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4291 if (!blk_queue_discard(q))
4292 ext4_msg(sb, KERN_WARNING,
4293 "mounting with \"discard\" option, but "
4294 "the device does not support discard");
4295 }
4296
4297 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4298 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4299 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4300 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4301
4302 if (es->s_error_count)
4303 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4304
4305 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4306 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4307 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4308 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4309
4310 kfree(orig_data);
4311 return 0;
4312
4313 cantfind_ext4:
4314 if (!silent)
4315 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4316 goto failed_mount;
4317
4318 #ifdef CONFIG_QUOTA
4319 failed_mount8:
4320 kobject_del(&sbi->s_kobj);
4321 #endif
4322 failed_mount7:
4323 ext4_unregister_li_request(sb);
4324 failed_mount6:
4325 ext4_mb_release(sb);
4326 if (sbi->s_flex_groups)
4327 kvfree(sbi->s_flex_groups);
4328 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4329 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4330 percpu_counter_destroy(&sbi->s_dirs_counter);
4331 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4332 failed_mount5:
4333 ext4_ext_release(sb);
4334 ext4_release_system_zone(sb);
4335 failed_mount4a:
4336 dput(sb->s_root);
4337 sb->s_root = NULL;
4338 failed_mount4:
4339 ext4_msg(sb, KERN_ERR, "mount failed");
4340 if (EXT4_SB(sb)->rsv_conversion_wq)
4341 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4342 failed_mount_wq:
4343 if (sbi->s_journal) {
4344 jbd2_journal_destroy(sbi->s_journal);
4345 sbi->s_journal = NULL;
4346 }
4347 failed_mount3a:
4348 ext4_es_unregister_shrinker(sbi);
4349 failed_mount3:
4350 del_timer_sync(&sbi->s_err_report);
4351 if (sbi->s_mmp_tsk)
4352 kthread_stop(sbi->s_mmp_tsk);
4353 failed_mount2:
4354 for (i = 0; i < db_count; i++)
4355 brelse(sbi->s_group_desc[i]);
4356 kvfree(sbi->s_group_desc);
4357 failed_mount:
4358 if (sbi->s_chksum_driver)
4359 crypto_free_shash(sbi->s_chksum_driver);
4360 if (sbi->s_proc) {
4361 remove_proc_entry("options", sbi->s_proc);
4362 remove_proc_entry(sb->s_id, ext4_proc_root);
4363 }
4364 #ifdef CONFIG_QUOTA
4365 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4366 kfree(sbi->s_qf_names[i]);
4367 #endif
4368 ext4_blkdev_remove(sbi);
4369 brelse(bh);
4370 out_fail:
4371 sb->s_fs_info = NULL;
4372 kfree(sbi->s_blockgroup_lock);
4373 kfree(sbi);
4374 out_free_orig:
4375 kfree(orig_data);
4376 return err ? err : ret;
4377 }
4378
4379 /*
4380 * Setup any per-fs journal parameters now. We'll do this both on
4381 * initial mount, once the journal has been initialised but before we've
4382 * done any recovery; and again on any subsequent remount.
4383 */
4384 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4385 {
4386 struct ext4_sb_info *sbi = EXT4_SB(sb);
4387
4388 journal->j_commit_interval = sbi->s_commit_interval;
4389 journal->j_min_batch_time = sbi->s_min_batch_time;
4390 journal->j_max_batch_time = sbi->s_max_batch_time;
4391
4392 write_lock(&journal->j_state_lock);
4393 if (test_opt(sb, BARRIER))
4394 journal->j_flags |= JBD2_BARRIER;
4395 else
4396 journal->j_flags &= ~JBD2_BARRIER;
4397 if (test_opt(sb, DATA_ERR_ABORT))
4398 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4399 else
4400 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4401 write_unlock(&journal->j_state_lock);
4402 }
4403
4404 static journal_t *ext4_get_journal(struct super_block *sb,
4405 unsigned int journal_inum)
4406 {
4407 struct inode *journal_inode;
4408 journal_t *journal;
4409
4410 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4411
4412 /* First, test for the existence of a valid inode on disk. Bad
4413 * things happen if we iget() an unused inode, as the subsequent
4414 * iput() will try to delete it. */
4415
4416 journal_inode = ext4_iget(sb, journal_inum);
4417 if (IS_ERR(journal_inode)) {
4418 ext4_msg(sb, KERN_ERR, "no journal found");
4419 return NULL;
4420 }
4421 if (!journal_inode->i_nlink) {
4422 make_bad_inode(journal_inode);
4423 iput(journal_inode);
4424 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4425 return NULL;
4426 }
4427
4428 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4429 journal_inode, journal_inode->i_size);
4430 if (!S_ISREG(journal_inode->i_mode)) {
4431 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4432 iput(journal_inode);
4433 return NULL;
4434 }
4435
4436 journal = jbd2_journal_init_inode(journal_inode);
4437 if (!journal) {
4438 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4439 iput(journal_inode);
4440 return NULL;
4441 }
4442 journal->j_private = sb;
4443 ext4_init_journal_params(sb, journal);
4444 return journal;
4445 }
4446
4447 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4448 dev_t j_dev)
4449 {
4450 struct buffer_head *bh;
4451 journal_t *journal;
4452 ext4_fsblk_t start;
4453 ext4_fsblk_t len;
4454 int hblock, blocksize;
4455 ext4_fsblk_t sb_block;
4456 unsigned long offset;
4457 struct ext4_super_block *es;
4458 struct block_device *bdev;
4459
4460 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4461
4462 bdev = ext4_blkdev_get(j_dev, sb);
4463 if (bdev == NULL)
4464 return NULL;
4465
4466 blocksize = sb->s_blocksize;
4467 hblock = bdev_logical_block_size(bdev);
4468 if (blocksize < hblock) {
4469 ext4_msg(sb, KERN_ERR,
4470 "blocksize too small for journal device");
4471 goto out_bdev;
4472 }
4473
4474 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4475 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4476 set_blocksize(bdev, blocksize);
4477 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4478 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4479 "external journal");
4480 goto out_bdev;
4481 }
4482
4483 es = (struct ext4_super_block *) (bh->b_data + offset);
4484 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4485 !(le32_to_cpu(es->s_feature_incompat) &
4486 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4487 ext4_msg(sb, KERN_ERR, "external journal has "
4488 "bad superblock");
4489 brelse(bh);
4490 goto out_bdev;
4491 }
4492
4493 if ((le32_to_cpu(es->s_feature_ro_compat) &
4494 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4495 es->s_checksum != ext4_superblock_csum(sb, es)) {
4496 ext4_msg(sb, KERN_ERR, "external journal has "
4497 "corrupt superblock");
4498 brelse(bh);
4499 goto out_bdev;
4500 }
4501
4502 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4503 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4504 brelse(bh);
4505 goto out_bdev;
4506 }
4507
4508 len = ext4_blocks_count(es);
4509 start = sb_block + 1;
4510 brelse(bh); /* we're done with the superblock */
4511
4512 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4513 start, len, blocksize);
4514 if (!journal) {
4515 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4516 goto out_bdev;
4517 }
4518 journal->j_private = sb;
4519 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4520 wait_on_buffer(journal->j_sb_buffer);
4521 if (!buffer_uptodate(journal->j_sb_buffer)) {
4522 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4523 goto out_journal;
4524 }
4525 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4526 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4527 "user (unsupported) - %d",
4528 be32_to_cpu(journal->j_superblock->s_nr_users));
4529 goto out_journal;
4530 }
4531 EXT4_SB(sb)->journal_bdev = bdev;
4532 ext4_init_journal_params(sb, journal);
4533 return journal;
4534
4535 out_journal:
4536 jbd2_journal_destroy(journal);
4537 out_bdev:
4538 ext4_blkdev_put(bdev);
4539 return NULL;
4540 }
4541
4542 static int ext4_load_journal(struct super_block *sb,
4543 struct ext4_super_block *es,
4544 unsigned long journal_devnum)
4545 {
4546 journal_t *journal;
4547 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4548 dev_t journal_dev;
4549 int err = 0;
4550 int really_read_only;
4551
4552 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4553
4554 if (journal_devnum &&
4555 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4556 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4557 "numbers have changed");
4558 journal_dev = new_decode_dev(journal_devnum);
4559 } else
4560 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4561
4562 really_read_only = bdev_read_only(sb->s_bdev);
4563
4564 /*
4565 * Are we loading a blank journal or performing recovery after a
4566 * crash? For recovery, we need to check in advance whether we
4567 * can get read-write access to the device.
4568 */
4569 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4570 if (sb->s_flags & MS_RDONLY) {
4571 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4572 "required on readonly filesystem");
4573 if (really_read_only) {
4574 ext4_msg(sb, KERN_ERR, "write access "
4575 "unavailable, cannot proceed");
4576 return -EROFS;
4577 }
4578 ext4_msg(sb, KERN_INFO, "write access will "
4579 "be enabled during recovery");
4580 }
4581 }
4582
4583 if (journal_inum && journal_dev) {
4584 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4585 "and inode journals!");
4586 return -EINVAL;
4587 }
4588
4589 if (journal_inum) {
4590 if (!(journal = ext4_get_journal(sb, journal_inum)))
4591 return -EINVAL;
4592 } else {
4593 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4594 return -EINVAL;
4595 }
4596
4597 if (!(journal->j_flags & JBD2_BARRIER))
4598 ext4_msg(sb, KERN_INFO, "barriers disabled");
4599
4600 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4601 err = jbd2_journal_wipe(journal, !really_read_only);
4602 if (!err) {
4603 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4604 if (save)
4605 memcpy(save, ((char *) es) +
4606 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4607 err = jbd2_journal_load(journal);
4608 if (save)
4609 memcpy(((char *) es) + EXT4_S_ERR_START,
4610 save, EXT4_S_ERR_LEN);
4611 kfree(save);
4612 }
4613
4614 if (err) {
4615 ext4_msg(sb, KERN_ERR, "error loading journal");
4616 jbd2_journal_destroy(journal);
4617 return err;
4618 }
4619
4620 EXT4_SB(sb)->s_journal = journal;
4621 ext4_clear_journal_err(sb, es);
4622
4623 if (!really_read_only && journal_devnum &&
4624 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4625 es->s_journal_dev = cpu_to_le32(journal_devnum);
4626
4627 /* Make sure we flush the recovery flag to disk. */
4628 ext4_commit_super(sb, 1);
4629 }
4630
4631 return 0;
4632 }
4633
4634 static int ext4_commit_super(struct super_block *sb, int sync)
4635 {
4636 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4637 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4638 int error = 0;
4639
4640 if (!sbh || block_device_ejected(sb))
4641 return error;
4642 if (buffer_write_io_error(sbh)) {
4643 /*
4644 * Oh, dear. A previous attempt to write the
4645 * superblock failed. This could happen because the
4646 * USB device was yanked out. Or it could happen to
4647 * be a transient write error and maybe the block will
4648 * be remapped. Nothing we can do but to retry the
4649 * write and hope for the best.
4650 */
4651 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4652 "superblock detected");
4653 clear_buffer_write_io_error(sbh);
4654 set_buffer_uptodate(sbh);
4655 }
4656 /*
4657 * If the file system is mounted read-only, don't update the
4658 * superblock write time. This avoids updating the superblock
4659 * write time when we are mounting the root file system
4660 * read/only but we need to replay the journal; at that point,
4661 * for people who are east of GMT and who make their clock
4662 * tick in localtime for Windows bug-for-bug compatibility,
4663 * the clock is set in the future, and this will cause e2fsck
4664 * to complain and force a full file system check.
4665 */
4666 if (!(sb->s_flags & MS_RDONLY))
4667 es->s_wtime = cpu_to_le32(get_seconds());
4668 if (sb->s_bdev->bd_part)
4669 es->s_kbytes_written =
4670 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4671 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4672 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4673 else
4674 es->s_kbytes_written =
4675 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4676 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4677 ext4_free_blocks_count_set(es,
4678 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4679 &EXT4_SB(sb)->s_freeclusters_counter)));
4680 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4681 es->s_free_inodes_count =
4682 cpu_to_le32(percpu_counter_sum_positive(
4683 &EXT4_SB(sb)->s_freeinodes_counter));
4684 BUFFER_TRACE(sbh, "marking dirty");
4685 ext4_superblock_csum_set(sb);
4686 mark_buffer_dirty(sbh);
4687 if (sync) {
4688 error = __sync_dirty_buffer(sbh,
4689 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4690 if (error)
4691 return error;
4692
4693 error = buffer_write_io_error(sbh);
4694 if (error) {
4695 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4696 "superblock");
4697 clear_buffer_write_io_error(sbh);
4698 set_buffer_uptodate(sbh);
4699 }
4700 }
4701 return error;
4702 }
4703
4704 /*
4705 * Have we just finished recovery? If so, and if we are mounting (or
4706 * remounting) the filesystem readonly, then we will end up with a
4707 * consistent fs on disk. Record that fact.
4708 */
4709 static void ext4_mark_recovery_complete(struct super_block *sb,
4710 struct ext4_super_block *es)
4711 {
4712 journal_t *journal = EXT4_SB(sb)->s_journal;
4713
4714 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4715 BUG_ON(journal != NULL);
4716 return;
4717 }
4718 jbd2_journal_lock_updates(journal);
4719 if (jbd2_journal_flush(journal) < 0)
4720 goto out;
4721
4722 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4723 sb->s_flags & MS_RDONLY) {
4724 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4725 ext4_commit_super(sb, 1);
4726 }
4727
4728 out:
4729 jbd2_journal_unlock_updates(journal);
4730 }
4731
4732 /*
4733 * If we are mounting (or read-write remounting) a filesystem whose journal
4734 * has recorded an error from a previous lifetime, move that error to the
4735 * main filesystem now.
4736 */
4737 static void ext4_clear_journal_err(struct super_block *sb,
4738 struct ext4_super_block *es)
4739 {
4740 journal_t *journal;
4741 int j_errno;
4742 const char *errstr;
4743
4744 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4745
4746 journal = EXT4_SB(sb)->s_journal;
4747
4748 /*
4749 * Now check for any error status which may have been recorded in the
4750 * journal by a prior ext4_error() or ext4_abort()
4751 */
4752
4753 j_errno = jbd2_journal_errno(journal);
4754 if (j_errno) {
4755 char nbuf[16];
4756
4757 errstr = ext4_decode_error(sb, j_errno, nbuf);
4758 ext4_warning(sb, "Filesystem error recorded "
4759 "from previous mount: %s", errstr);
4760 ext4_warning(sb, "Marking fs in need of filesystem check.");
4761
4762 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4763 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4764 ext4_commit_super(sb, 1);
4765
4766 jbd2_journal_clear_err(journal);
4767 jbd2_journal_update_sb_errno(journal);
4768 }
4769 }
4770
4771 /*
4772 * Force the running and committing transactions to commit,
4773 * and wait on the commit.
4774 */
4775 int ext4_force_commit(struct super_block *sb)
4776 {
4777 journal_t *journal;
4778
4779 if (sb->s_flags & MS_RDONLY)
4780 return 0;
4781
4782 journal = EXT4_SB(sb)->s_journal;
4783 return ext4_journal_force_commit(journal);
4784 }
4785
4786 static int ext4_sync_fs(struct super_block *sb, int wait)
4787 {
4788 int ret = 0;
4789 tid_t target;
4790 bool needs_barrier = false;
4791 struct ext4_sb_info *sbi = EXT4_SB(sb);
4792
4793 trace_ext4_sync_fs(sb, wait);
4794 flush_workqueue(sbi->rsv_conversion_wq);
4795 /*
4796 * Writeback quota in non-journalled quota case - journalled quota has
4797 * no dirty dquots
4798 */
4799 dquot_writeback_dquots(sb, -1);
4800 /*
4801 * Data writeback is possible w/o journal transaction, so barrier must
4802 * being sent at the end of the function. But we can skip it if
4803 * transaction_commit will do it for us.
4804 */
4805 if (sbi->s_journal) {
4806 target = jbd2_get_latest_transaction(sbi->s_journal);
4807 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4808 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4809 needs_barrier = true;
4810
4811 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4812 if (wait)
4813 ret = jbd2_log_wait_commit(sbi->s_journal,
4814 target);
4815 }
4816 } else if (wait && test_opt(sb, BARRIER))
4817 needs_barrier = true;
4818 if (needs_barrier) {
4819 int err;
4820 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4821 if (!ret)
4822 ret = err;
4823 }
4824
4825 return ret;
4826 }
4827
4828 /*
4829 * LVM calls this function before a (read-only) snapshot is created. This
4830 * gives us a chance to flush the journal completely and mark the fs clean.
4831 *
4832 * Note that only this function cannot bring a filesystem to be in a clean
4833 * state independently. It relies on upper layer to stop all data & metadata
4834 * modifications.
4835 */
4836 static int ext4_freeze(struct super_block *sb)
4837 {
4838 int error = 0;
4839 journal_t *journal;
4840
4841 if (sb->s_flags & MS_RDONLY)
4842 return 0;
4843
4844 journal = EXT4_SB(sb)->s_journal;
4845
4846 if (journal) {
4847 /* Now we set up the journal barrier. */
4848 jbd2_journal_lock_updates(journal);
4849
4850 /*
4851 * Don't clear the needs_recovery flag if we failed to
4852 * flush the journal.
4853 */
4854 error = jbd2_journal_flush(journal);
4855 if (error < 0)
4856 goto out;
4857
4858 /* Journal blocked and flushed, clear needs_recovery flag. */
4859 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4860 }
4861
4862 error = ext4_commit_super(sb, 1);
4863 out:
4864 if (journal)
4865 /* we rely on upper layer to stop further updates */
4866 jbd2_journal_unlock_updates(journal);
4867 return error;
4868 }
4869
4870 /*
4871 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4872 * flag here, even though the filesystem is not technically dirty yet.
4873 */
4874 static int ext4_unfreeze(struct super_block *sb)
4875 {
4876 if (sb->s_flags & MS_RDONLY)
4877 return 0;
4878
4879 if (EXT4_SB(sb)->s_journal) {
4880 /* Reset the needs_recovery flag before the fs is unlocked. */
4881 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4882 }
4883
4884 ext4_commit_super(sb, 1);
4885 return 0;
4886 }
4887
4888 /*
4889 * Structure to save mount options for ext4_remount's benefit
4890 */
4891 struct ext4_mount_options {
4892 unsigned long s_mount_opt;
4893 unsigned long s_mount_opt2;
4894 kuid_t s_resuid;
4895 kgid_t s_resgid;
4896 unsigned long s_commit_interval;
4897 u32 s_min_batch_time, s_max_batch_time;
4898 #ifdef CONFIG_QUOTA
4899 int s_jquota_fmt;
4900 char *s_qf_names[EXT4_MAXQUOTAS];
4901 #endif
4902 };
4903
4904 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4905 {
4906 struct ext4_super_block *es;
4907 struct ext4_sb_info *sbi = EXT4_SB(sb);
4908 unsigned long old_sb_flags;
4909 struct ext4_mount_options old_opts;
4910 int enable_quota = 0;
4911 ext4_group_t g;
4912 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4913 int err = 0;
4914 #ifdef CONFIG_QUOTA
4915 int i, j;
4916 #endif
4917 char *orig_data = kstrdup(data, GFP_KERNEL);
4918
4919 /* Store the original options */
4920 old_sb_flags = sb->s_flags;
4921 old_opts.s_mount_opt = sbi->s_mount_opt;
4922 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4923 old_opts.s_resuid = sbi->s_resuid;
4924 old_opts.s_resgid = sbi->s_resgid;
4925 old_opts.s_commit_interval = sbi->s_commit_interval;
4926 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4927 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4928 #ifdef CONFIG_QUOTA
4929 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4930 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4931 if (sbi->s_qf_names[i]) {
4932 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4933 GFP_KERNEL);
4934 if (!old_opts.s_qf_names[i]) {
4935 for (j = 0; j < i; j++)
4936 kfree(old_opts.s_qf_names[j]);
4937 kfree(orig_data);
4938 return -ENOMEM;
4939 }
4940 } else
4941 old_opts.s_qf_names[i] = NULL;
4942 #endif
4943 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4944 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4945
4946 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4947 err = -EINVAL;
4948 goto restore_opts;
4949 }
4950
4951 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4952 test_opt(sb, JOURNAL_CHECKSUM)) {
4953 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4954 "during remount not supported; ignoring");
4955 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4956 }
4957
4958 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4959 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4960 ext4_msg(sb, KERN_ERR, "can't mount with "
4961 "both data=journal and delalloc");
4962 err = -EINVAL;
4963 goto restore_opts;
4964 }
4965 if (test_opt(sb, DIOREAD_NOLOCK)) {
4966 ext4_msg(sb, KERN_ERR, "can't mount with "
4967 "both data=journal and dioread_nolock");
4968 err = -EINVAL;
4969 goto restore_opts;
4970 }
4971 if (test_opt(sb, DAX)) {
4972 ext4_msg(sb, KERN_ERR, "can't mount with "
4973 "both data=journal and dax");
4974 err = -EINVAL;
4975 goto restore_opts;
4976 }
4977 }
4978
4979 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4980 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4981 "dax flag with busy inodes while remounting");
4982 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4983 }
4984
4985 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4986 ext4_abort(sb, "Abort forced by user");
4987
4988 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4989 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4990
4991 es = sbi->s_es;
4992
4993 if (sbi->s_journal) {
4994 ext4_init_journal_params(sb, sbi->s_journal);
4995 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4996 }
4997
4998 if (*flags & MS_LAZYTIME)
4999 sb->s_flags |= MS_LAZYTIME;
5000
5001 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
5002 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5003 err = -EROFS;
5004 goto restore_opts;
5005 }
5006
5007 if (*flags & MS_RDONLY) {
5008 err = sync_filesystem(sb);
5009 if (err < 0)
5010 goto restore_opts;
5011 err = dquot_suspend(sb, -1);
5012 if (err < 0)
5013 goto restore_opts;
5014
5015 /*
5016 * First of all, the unconditional stuff we have to do
5017 * to disable replay of the journal when we next remount
5018 */
5019 sb->s_flags |= MS_RDONLY;
5020
5021 /*
5022 * OK, test if we are remounting a valid rw partition
5023 * readonly, and if so set the rdonly flag and then
5024 * mark the partition as valid again.
5025 */
5026 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5027 (sbi->s_mount_state & EXT4_VALID_FS))
5028 es->s_state = cpu_to_le16(sbi->s_mount_state);
5029
5030 if (sbi->s_journal)
5031 ext4_mark_recovery_complete(sb, es);
5032 } else {
5033 /* Make sure we can mount this feature set readwrite */
5034 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5035 EXT4_FEATURE_RO_COMPAT_READONLY) ||
5036 !ext4_feature_set_ok(sb, 0)) {
5037 err = -EROFS;
5038 goto restore_opts;
5039 }
5040 /*
5041 * Make sure the group descriptor checksums
5042 * are sane. If they aren't, refuse to remount r/w.
5043 */
5044 for (g = 0; g < sbi->s_groups_count; g++) {
5045 struct ext4_group_desc *gdp =
5046 ext4_get_group_desc(sb, g, NULL);
5047
5048 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5049 ext4_msg(sb, KERN_ERR,
5050 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5051 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
5052 le16_to_cpu(gdp->bg_checksum));
5053 err = -EINVAL;
5054 goto restore_opts;
5055 }
5056 }
5057
5058 /*
5059 * If we have an unprocessed orphan list hanging
5060 * around from a previously readonly bdev mount,
5061 * require a full umount/remount for now.
5062 */
5063 if (es->s_last_orphan) {
5064 ext4_msg(sb, KERN_WARNING, "Couldn't "
5065 "remount RDWR because of unprocessed "
5066 "orphan inode list. Please "
5067 "umount/remount instead");
5068 err = -EINVAL;
5069 goto restore_opts;
5070 }
5071
5072 /*
5073 * Mounting a RDONLY partition read-write, so reread
5074 * and store the current valid flag. (It may have
5075 * been changed by e2fsck since we originally mounted
5076 * the partition.)
5077 */
5078 if (sbi->s_journal)
5079 ext4_clear_journal_err(sb, es);
5080 sbi->s_mount_state = le16_to_cpu(es->s_state);
5081 if (!ext4_setup_super(sb, es, 0))
5082 sb->s_flags &= ~MS_RDONLY;
5083 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5084 EXT4_FEATURE_INCOMPAT_MMP))
5085 if (ext4_multi_mount_protect(sb,
5086 le64_to_cpu(es->s_mmp_block))) {
5087 err = -EROFS;
5088 goto restore_opts;
5089 }
5090 enable_quota = 1;
5091 }
5092 }
5093
5094 /*
5095 * Reinitialize lazy itable initialization thread based on
5096 * current settings
5097 */
5098 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5099 ext4_unregister_li_request(sb);
5100 else {
5101 ext4_group_t first_not_zeroed;
5102 first_not_zeroed = ext4_has_uninit_itable(sb);
5103 ext4_register_li_request(sb, first_not_zeroed);
5104 }
5105
5106 ext4_setup_system_zone(sb);
5107 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5108 ext4_commit_super(sb, 1);
5109
5110 #ifdef CONFIG_QUOTA
5111 /* Release old quota file names */
5112 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5113 kfree(old_opts.s_qf_names[i]);
5114 if (enable_quota) {
5115 if (sb_any_quota_suspended(sb))
5116 dquot_resume(sb, -1);
5117 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5118 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5119 err = ext4_enable_quotas(sb);
5120 if (err)
5121 goto restore_opts;
5122 }
5123 }
5124 #endif
5125
5126 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5127 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5128 kfree(orig_data);
5129 return 0;
5130
5131 restore_opts:
5132 sb->s_flags = old_sb_flags;
5133 sbi->s_mount_opt = old_opts.s_mount_opt;
5134 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5135 sbi->s_resuid = old_opts.s_resuid;
5136 sbi->s_resgid = old_opts.s_resgid;
5137 sbi->s_commit_interval = old_opts.s_commit_interval;
5138 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5139 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5140 #ifdef CONFIG_QUOTA
5141 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5142 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5143 kfree(sbi->s_qf_names[i]);
5144 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5145 }
5146 #endif
5147 kfree(orig_data);
5148 return err;
5149 }
5150
5151 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5152 {
5153 struct super_block *sb = dentry->d_sb;
5154 struct ext4_sb_info *sbi = EXT4_SB(sb);
5155 struct ext4_super_block *es = sbi->s_es;
5156 ext4_fsblk_t overhead = 0, resv_blocks;
5157 u64 fsid;
5158 s64 bfree;
5159 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5160
5161 if (!test_opt(sb, MINIX_DF))
5162 overhead = sbi->s_overhead;
5163
5164 buf->f_type = EXT4_SUPER_MAGIC;
5165 buf->f_bsize = sb->s_blocksize;
5166 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5167 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5168 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5169 /* prevent underflow in case that few free space is available */
5170 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5171 buf->f_bavail = buf->f_bfree -
5172 (ext4_r_blocks_count(es) + resv_blocks);
5173 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5174 buf->f_bavail = 0;
5175 buf->f_files = le32_to_cpu(es->s_inodes_count);
5176 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5177 buf->f_namelen = EXT4_NAME_LEN;
5178 fsid = le64_to_cpup((void *)es->s_uuid) ^
5179 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5180 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5181 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5182
5183 return 0;
5184 }
5185
5186 /* Helper function for writing quotas on sync - we need to start transaction
5187 * before quota file is locked for write. Otherwise the are possible deadlocks:
5188 * Process 1 Process 2
5189 * ext4_create() quota_sync()
5190 * jbd2_journal_start() write_dquot()
5191 * dquot_initialize() down(dqio_mutex)
5192 * down(dqio_mutex) jbd2_journal_start()
5193 *
5194 */
5195
5196 #ifdef CONFIG_QUOTA
5197
5198 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5199 {
5200 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5201 }
5202
5203 static int ext4_write_dquot(struct dquot *dquot)
5204 {
5205 int ret, err;
5206 handle_t *handle;
5207 struct inode *inode;
5208
5209 inode = dquot_to_inode(dquot);
5210 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5211 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5212 if (IS_ERR(handle))
5213 return PTR_ERR(handle);
5214 ret = dquot_commit(dquot);
5215 err = ext4_journal_stop(handle);
5216 if (!ret)
5217 ret = err;
5218 return ret;
5219 }
5220
5221 static int ext4_acquire_dquot(struct dquot *dquot)
5222 {
5223 int ret, err;
5224 handle_t *handle;
5225
5226 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5227 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5228 if (IS_ERR(handle))
5229 return PTR_ERR(handle);
5230 ret = dquot_acquire(dquot);
5231 err = ext4_journal_stop(handle);
5232 if (!ret)
5233 ret = err;
5234 return ret;
5235 }
5236
5237 static int ext4_release_dquot(struct dquot *dquot)
5238 {
5239 int ret, err;
5240 handle_t *handle;
5241
5242 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5243 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5244 if (IS_ERR(handle)) {
5245 /* Release dquot anyway to avoid endless cycle in dqput() */
5246 dquot_release(dquot);
5247 return PTR_ERR(handle);
5248 }
5249 ret = dquot_release(dquot);
5250 err = ext4_journal_stop(handle);
5251 if (!ret)
5252 ret = err;
5253 return ret;
5254 }
5255
5256 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5257 {
5258 struct super_block *sb = dquot->dq_sb;
5259 struct ext4_sb_info *sbi = EXT4_SB(sb);
5260
5261 /* Are we journaling quotas? */
5262 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5263 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5264 dquot_mark_dquot_dirty(dquot);
5265 return ext4_write_dquot(dquot);
5266 } else {
5267 return dquot_mark_dquot_dirty(dquot);
5268 }
5269 }
5270
5271 static int ext4_write_info(struct super_block *sb, int type)
5272 {
5273 int ret, err;
5274 handle_t *handle;
5275
5276 /* Data block + inode block */
5277 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5278 if (IS_ERR(handle))
5279 return PTR_ERR(handle);
5280 ret = dquot_commit_info(sb, type);
5281 err = ext4_journal_stop(handle);
5282 if (!ret)
5283 ret = err;
5284 return ret;
5285 }
5286
5287 /*
5288 * Turn on quotas during mount time - we need to find
5289 * the quota file and such...
5290 */
5291 static int ext4_quota_on_mount(struct super_block *sb, int type)
5292 {
5293 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5294 EXT4_SB(sb)->s_jquota_fmt, type);
5295 }
5296
5297 /*
5298 * Standard function to be called on quota_on
5299 */
5300 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5301 struct path *path)
5302 {
5303 int err;
5304
5305 if (!test_opt(sb, QUOTA))
5306 return -EINVAL;
5307
5308 /* Quotafile not on the same filesystem? */
5309 if (path->dentry->d_sb != sb)
5310 return -EXDEV;
5311 /* Journaling quota? */
5312 if (EXT4_SB(sb)->s_qf_names[type]) {
5313 /* Quotafile not in fs root? */
5314 if (path->dentry->d_parent != sb->s_root)
5315 ext4_msg(sb, KERN_WARNING,
5316 "Quota file not on filesystem root. "
5317 "Journaled quota will not work");
5318 }
5319
5320 /*
5321 * When we journal data on quota file, we have to flush journal to see
5322 * all updates to the file when we bypass pagecache...
5323 */
5324 if (EXT4_SB(sb)->s_journal &&
5325 ext4_should_journal_data(d_inode(path->dentry))) {
5326 /*
5327 * We don't need to lock updates but journal_flush() could
5328 * otherwise be livelocked...
5329 */
5330 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5331 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5332 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5333 if (err)
5334 return err;
5335 }
5336
5337 return dquot_quota_on(sb, type, format_id, path);
5338 }
5339
5340 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5341 unsigned int flags)
5342 {
5343 int err;
5344 struct inode *qf_inode;
5345 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5346 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5347 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5348 };
5349
5350 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5351
5352 if (!qf_inums[type])
5353 return -EPERM;
5354
5355 qf_inode = ext4_iget(sb, qf_inums[type]);
5356 if (IS_ERR(qf_inode)) {
5357 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5358 return PTR_ERR(qf_inode);
5359 }
5360
5361 /* Don't account quota for quota files to avoid recursion */
5362 qf_inode->i_flags |= S_NOQUOTA;
5363 err = dquot_enable(qf_inode, type, format_id, flags);
5364 iput(qf_inode);
5365
5366 return err;
5367 }
5368
5369 /* Enable usage tracking for all quota types. */
5370 static int ext4_enable_quotas(struct super_block *sb)
5371 {
5372 int type, err = 0;
5373 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5374 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5375 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5376 };
5377
5378 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5379 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5380 if (qf_inums[type]) {
5381 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5382 DQUOT_USAGE_ENABLED);
5383 if (err) {
5384 ext4_warning(sb,
5385 "Failed to enable quota tracking "
5386 "(type=%d, err=%d). Please run "
5387 "e2fsck to fix.", type, err);
5388 return err;
5389 }
5390 }
5391 }
5392 return 0;
5393 }
5394
5395 static int ext4_quota_off(struct super_block *sb, int type)
5396 {
5397 struct inode *inode = sb_dqopt(sb)->files[type];
5398 handle_t *handle;
5399
5400 /* Force all delayed allocation blocks to be allocated.
5401 * Caller already holds s_umount sem */
5402 if (test_opt(sb, DELALLOC))
5403 sync_filesystem(sb);
5404
5405 if (!inode)
5406 goto out;
5407
5408 /* Update modification times of quota files when userspace can
5409 * start looking at them */
5410 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5411 if (IS_ERR(handle))
5412 goto out;
5413 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5414 ext4_mark_inode_dirty(handle, inode);
5415 ext4_journal_stop(handle);
5416
5417 out:
5418 return dquot_quota_off(sb, type);
5419 }
5420
5421 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5422 * acquiring the locks... As quota files are never truncated and quota code
5423 * itself serializes the operations (and no one else should touch the files)
5424 * we don't have to be afraid of races */
5425 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5426 size_t len, loff_t off)
5427 {
5428 struct inode *inode = sb_dqopt(sb)->files[type];
5429 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5430 int offset = off & (sb->s_blocksize - 1);
5431 int tocopy;
5432 size_t toread;
5433 struct buffer_head *bh;
5434 loff_t i_size = i_size_read(inode);
5435
5436 if (off > i_size)
5437 return 0;
5438 if (off+len > i_size)
5439 len = i_size-off;
5440 toread = len;
5441 while (toread > 0) {
5442 tocopy = sb->s_blocksize - offset < toread ?
5443 sb->s_blocksize - offset : toread;
5444 bh = ext4_bread(NULL, inode, blk, 0);
5445 if (IS_ERR(bh))
5446 return PTR_ERR(bh);
5447 if (!bh) /* A hole? */
5448 memset(data, 0, tocopy);
5449 else
5450 memcpy(data, bh->b_data+offset, tocopy);
5451 brelse(bh);
5452 offset = 0;
5453 toread -= tocopy;
5454 data += tocopy;
5455 blk++;
5456 }
5457 return len;
5458 }
5459
5460 /* Write to quotafile (we know the transaction is already started and has
5461 * enough credits) */
5462 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5463 const char *data, size_t len, loff_t off)
5464 {
5465 struct inode *inode = sb_dqopt(sb)->files[type];
5466 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5467 int err, offset = off & (sb->s_blocksize - 1);
5468 int retries = 0;
5469 struct buffer_head *bh;
5470 handle_t *handle = journal_current_handle();
5471
5472 if (EXT4_SB(sb)->s_journal && !handle) {
5473 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5474 " cancelled because transaction is not started",
5475 (unsigned long long)off, (unsigned long long)len);
5476 return -EIO;
5477 }
5478 /*
5479 * Since we account only one data block in transaction credits,
5480 * then it is impossible to cross a block boundary.
5481 */
5482 if (sb->s_blocksize - offset < len) {
5483 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5484 " cancelled because not block aligned",
5485 (unsigned long long)off, (unsigned long long)len);
5486 return -EIO;
5487 }
5488
5489 do {
5490 bh = ext4_bread(handle, inode, blk,
5491 EXT4_GET_BLOCKS_CREATE |
5492 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5493 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5494 ext4_should_retry_alloc(inode->i_sb, &retries));
5495 if (IS_ERR(bh))
5496 return PTR_ERR(bh);
5497 if (!bh)
5498 goto out;
5499 BUFFER_TRACE(bh, "get write access");
5500 err = ext4_journal_get_write_access(handle, bh);
5501 if (err) {
5502 brelse(bh);
5503 return err;
5504 }
5505 lock_buffer(bh);
5506 memcpy(bh->b_data+offset, data, len);
5507 flush_dcache_page(bh->b_page);
5508 unlock_buffer(bh);
5509 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5510 brelse(bh);
5511 out:
5512 if (inode->i_size < off + len) {
5513 i_size_write(inode, off + len);
5514 EXT4_I(inode)->i_disksize = inode->i_size;
5515 ext4_mark_inode_dirty(handle, inode);
5516 }
5517 return len;
5518 }
5519
5520 #endif
5521
5522 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5523 const char *dev_name, void *data)
5524 {
5525 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5526 }
5527
5528 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5529 static inline void register_as_ext2(void)
5530 {
5531 int err = register_filesystem(&ext2_fs_type);
5532 if (err)
5533 printk(KERN_WARNING
5534 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5535 }
5536
5537 static inline void unregister_as_ext2(void)
5538 {
5539 unregister_filesystem(&ext2_fs_type);
5540 }
5541
5542 static inline int ext2_feature_set_ok(struct super_block *sb)
5543 {
5544 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5545 return 0;
5546 if (sb->s_flags & MS_RDONLY)
5547 return 1;
5548 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5549 return 0;
5550 return 1;
5551 }
5552 #else
5553 static inline void register_as_ext2(void) { }
5554 static inline void unregister_as_ext2(void) { }
5555 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5556 #endif
5557
5558 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5559 static inline void register_as_ext3(void)
5560 {
5561 int err = register_filesystem(&ext3_fs_type);
5562 if (err)
5563 printk(KERN_WARNING
5564 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5565 }
5566
5567 static inline void unregister_as_ext3(void)
5568 {
5569 unregister_filesystem(&ext3_fs_type);
5570 }
5571
5572 static inline int ext3_feature_set_ok(struct super_block *sb)
5573 {
5574 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5575 return 0;
5576 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5577 return 0;
5578 if (sb->s_flags & MS_RDONLY)
5579 return 1;
5580 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5581 return 0;
5582 return 1;
5583 }
5584 #else
5585 static inline void register_as_ext3(void) { }
5586 static inline void unregister_as_ext3(void) { }
5587 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5588 #endif
5589
5590 static struct file_system_type ext4_fs_type = {
5591 .owner = THIS_MODULE,
5592 .name = "ext4",
5593 .mount = ext4_mount,
5594 .kill_sb = kill_block_super,
5595 .fs_flags = FS_REQUIRES_DEV,
5596 };
5597 MODULE_ALIAS_FS("ext4");
5598
5599 static int __init ext4_init_feat_adverts(void)
5600 {
5601 struct ext4_features *ef;
5602 int ret = -ENOMEM;
5603
5604 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5605 if (!ef)
5606 goto out;
5607
5608 ef->f_kobj.kset = ext4_kset;
5609 init_completion(&ef->f_kobj_unregister);
5610 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5611 "features");
5612 if (ret) {
5613 kfree(ef);
5614 goto out;
5615 }
5616
5617 ext4_feat = ef;
5618 ret = 0;
5619 out:
5620 return ret;
5621 }
5622
5623 static void ext4_exit_feat_adverts(void)
5624 {
5625 kobject_put(&ext4_feat->f_kobj);
5626 wait_for_completion(&ext4_feat->f_kobj_unregister);
5627 kfree(ext4_feat);
5628 }
5629
5630 /* Shared across all ext4 file systems */
5631 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5632 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5633
5634 static int __init ext4_init_fs(void)
5635 {
5636 int i, err;
5637
5638 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5639 ext4_li_info = NULL;
5640 mutex_init(&ext4_li_mtx);
5641
5642 /* Build-time check for flags consistency */
5643 ext4_check_flag_values();
5644
5645 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5646 mutex_init(&ext4__aio_mutex[i]);
5647 init_waitqueue_head(&ext4__ioend_wq[i]);
5648 }
5649
5650 err = ext4_init_es();
5651 if (err)
5652 return err;
5653
5654 err = ext4_init_pageio();
5655 if (err)
5656 goto out7;
5657
5658 err = ext4_init_system_zone();
5659 if (err)
5660 goto out6;
5661 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5662 if (!ext4_kset) {
5663 err = -ENOMEM;
5664 goto out5;
5665 }
5666 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5667
5668 err = ext4_init_feat_adverts();
5669 if (err)
5670 goto out4;
5671
5672 err = ext4_init_mballoc();
5673 if (err)
5674 goto out2;
5675 else
5676 ext4_mballoc_ready = 1;
5677 err = init_inodecache();
5678 if (err)
5679 goto out1;
5680 register_as_ext3();
5681 register_as_ext2();
5682 err = register_filesystem(&ext4_fs_type);
5683 if (err)
5684 goto out;
5685
5686 return 0;
5687 out:
5688 unregister_as_ext2();
5689 unregister_as_ext3();
5690 destroy_inodecache();
5691 out1:
5692 ext4_mballoc_ready = 0;
5693 ext4_exit_mballoc();
5694 out2:
5695 ext4_exit_feat_adverts();
5696 out4:
5697 if (ext4_proc_root)
5698 remove_proc_entry("fs/ext4", NULL);
5699 kset_unregister(ext4_kset);
5700 out5:
5701 ext4_exit_system_zone();
5702 out6:
5703 ext4_exit_pageio();
5704 out7:
5705 ext4_exit_es();
5706
5707 return err;
5708 }
5709
5710 static void __exit ext4_exit_fs(void)
5711 {
5712 ext4_exit_crypto();
5713 ext4_destroy_lazyinit_thread();
5714 unregister_as_ext2();
5715 unregister_as_ext3();
5716 unregister_filesystem(&ext4_fs_type);
5717 destroy_inodecache();
5718 ext4_exit_mballoc();
5719 ext4_exit_feat_adverts();
5720 remove_proc_entry("fs/ext4", NULL);
5721 kset_unregister(ext4_kset);
5722 ext4_exit_system_zone();
5723 ext4_exit_pageio();
5724 ext4_exit_es();
5725 }
5726
5727 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5728 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5729 MODULE_LICENSE("GPL");
5730 module_init(ext4_init_fs)
5731 module_exit(ext4_exit_fs)
ubuntu-zesty-kernel mirror