projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge remote-tracking branch 'iommu/next'
[deliverable/linux.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/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <asm/uaccess.h>
42
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
45
46 #include "ext4.h"
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
49 #include "xattr.h"
50 #include "acl.h"
51 #include "mballoc.h"
52
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
55
56 static struct ext4_lazy_init *ext4_li_info;
57 static struct mutex ext4_li_mtx;
58 static struct ratelimit_state ext4_mount_msg_ratelimit;
59
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 unsigned long journal_devnum);
62 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
63 static int ext4_commit_super(struct super_block *sb, int sync);
64 static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66 static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
68 static int ext4_sync_fs(struct super_block *sb, int wait);
69 static int ext4_remount(struct super_block *sb, int *flags, char *data);
70 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
71 static int ext4_unfreeze(struct super_block *sb);
72 static int ext4_freeze(struct super_block *sb);
73 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74 const char *dev_name, void *data);
75 static inline int ext2_feature_set_ok(struct super_block *sb);
76 static inline int ext3_feature_set_ok(struct super_block *sb);
77 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block *sb);
80 static void ext4_clear_request_list(void);
81
82 /*
83 * Lock ordering
84 *
85 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
86 * i_mmap_rwsem (inode->i_mmap_rwsem)!
87 *
88 * page fault path:
89 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
90 * page lock -> i_data_sem (rw)
91 *
92 * buffered write path:
93 * sb_start_write -> i_mutex -> mmap_sem
94 * sb_start_write -> i_mutex -> transaction start -> page lock ->
95 * i_data_sem (rw)
96 *
97 * truncate:
98 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
99 * i_mmap_rwsem (w) -> page lock
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * transaction start -> i_data_sem (rw)
102 *
103 * direct IO:
104 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
105 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
106 * transaction start -> i_data_sem (rw)
107 *
108 * writepages:
109 * transaction start -> page lock(s) -> i_data_sem (rw)
110 */
111
112 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
113 static struct file_system_type ext2_fs_type = {
114 .owner = THIS_MODULE,
115 .name = "ext2",
116 .mount = ext4_mount,
117 .kill_sb = kill_block_super,
118 .fs_flags = FS_REQUIRES_DEV,
119 };
120 MODULE_ALIAS_FS("ext2");
121 MODULE_ALIAS("ext2");
122 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
123 #else
124 #define IS_EXT2_SB(sb) (0)
125 #endif
126
127
128 static struct file_system_type ext3_fs_type = {
129 .owner = THIS_MODULE,
130 .name = "ext3",
131 .mount = ext4_mount,
132 .kill_sb = kill_block_super,
133 .fs_flags = FS_REQUIRES_DEV,
134 };
135 MODULE_ALIAS_FS("ext3");
136 MODULE_ALIAS("ext3");
137 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
138
139 static int ext4_verify_csum_type(struct super_block *sb,
140 struct ext4_super_block *es)
141 {
142 if (!ext4_has_feature_metadata_csum(sb))
143 return 1;
144
145 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
146 }
147
148 static __le32 ext4_superblock_csum(struct super_block *sb,
149 struct ext4_super_block *es)
150 {
151 struct ext4_sb_info *sbi = EXT4_SB(sb);
152 int offset = offsetof(struct ext4_super_block, s_checksum);
153 __u32 csum;
154
155 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
156
157 return cpu_to_le32(csum);
158 }
159
160 static int ext4_superblock_csum_verify(struct super_block *sb,
161 struct ext4_super_block *es)
162 {
163 if (!ext4_has_metadata_csum(sb))
164 return 1;
165
166 return es->s_checksum == ext4_superblock_csum(sb, es);
167 }
168
169 void ext4_superblock_csum_set(struct super_block *sb)
170 {
171 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
172
173 if (!ext4_has_metadata_csum(sb))
174 return;
175
176 es->s_checksum = ext4_superblock_csum(sb, es);
177 }
178
179 void *ext4_kvmalloc(size_t size, gfp_t flags)
180 {
181 void *ret;
182
183 ret = kmalloc(size, flags | __GFP_NOWARN);
184 if (!ret)
185 ret = __vmalloc(size, flags, PAGE_KERNEL);
186 return ret;
187 }
188
189 void *ext4_kvzalloc(size_t size, gfp_t flags)
190 {
191 void *ret;
192
193 ret = kzalloc(size, flags | __GFP_NOWARN);
194 if (!ret)
195 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
196 return ret;
197 }
198
199 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
200 struct ext4_group_desc *bg)
201 {
202 return le32_to_cpu(bg->bg_block_bitmap_lo) |
203 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
204 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
205 }
206
207 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
208 struct ext4_group_desc *bg)
209 {
210 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
212 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
213 }
214
215 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
216 struct ext4_group_desc *bg)
217 {
218 return le32_to_cpu(bg->bg_inode_table_lo) |
219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
220 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
221 }
222
223 __u32 ext4_free_group_clusters(struct super_block *sb,
224 struct ext4_group_desc *bg)
225 {
226 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
228 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
229 }
230
231 __u32 ext4_free_inodes_count(struct super_block *sb,
232 struct ext4_group_desc *bg)
233 {
234 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
236 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
237 }
238
239 __u32 ext4_used_dirs_count(struct super_block *sb,
240 struct ext4_group_desc *bg)
241 {
242 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
244 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
245 }
246
247 __u32 ext4_itable_unused_count(struct super_block *sb,
248 struct ext4_group_desc *bg)
249 {
250 return le16_to_cpu(bg->bg_itable_unused_lo) |
251 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
252 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
253 }
254
255 void ext4_block_bitmap_set(struct super_block *sb,
256 struct ext4_group_desc *bg, ext4_fsblk_t blk)
257 {
258 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
259 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
260 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
261 }
262
263 void ext4_inode_bitmap_set(struct super_block *sb,
264 struct ext4_group_desc *bg, ext4_fsblk_t blk)
265 {
266 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
269 }
270
271 void ext4_inode_table_set(struct super_block *sb,
272 struct ext4_group_desc *bg, ext4_fsblk_t blk)
273 {
274 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
277 }
278
279 void ext4_free_group_clusters_set(struct super_block *sb,
280 struct ext4_group_desc *bg, __u32 count)
281 {
282 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
285 }
286
287 void ext4_free_inodes_set(struct super_block *sb,
288 struct ext4_group_desc *bg, __u32 count)
289 {
290 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
293 }
294
295 void ext4_used_dirs_set(struct super_block *sb,
296 struct ext4_group_desc *bg, __u32 count)
297 {
298 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
301 }
302
303 void ext4_itable_unused_set(struct super_block *sb,
304 struct ext4_group_desc *bg, __u32 count)
305 {
306 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
307 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
308 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
309 }
310
311
312 static void __save_error_info(struct super_block *sb, const char *func,
313 unsigned int line)
314 {
315 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
316
317 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
318 if (bdev_read_only(sb->s_bdev))
319 return;
320 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
321 es->s_last_error_time = cpu_to_le32(get_seconds());
322 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
323 es->s_last_error_line = cpu_to_le32(line);
324 if (!es->s_first_error_time) {
325 es->s_first_error_time = es->s_last_error_time;
326 strncpy(es->s_first_error_func, func,
327 sizeof(es->s_first_error_func));
328 es->s_first_error_line = cpu_to_le32(line);
329 es->s_first_error_ino = es->s_last_error_ino;
330 es->s_first_error_block = es->s_last_error_block;
331 }
332 /*
333 * Start the daily error reporting function if it hasn't been
334 * started already
335 */
336 if (!es->s_error_count)
337 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
338 le32_add_cpu(&es->s_error_count, 1);
339 }
340
341 static void save_error_info(struct super_block *sb, const char *func,
342 unsigned int line)
343 {
344 __save_error_info(sb, func, line);
345 ext4_commit_super(sb, 1);
346 }
347
348 /*
349 * The del_gendisk() function uninitializes the disk-specific data
350 * structures, including the bdi structure, without telling anyone
351 * else. Once this happens, any attempt to call mark_buffer_dirty()
352 * (for example, by ext4_commit_super), will cause a kernel OOPS.
353 * This is a kludge to prevent these oops until we can put in a proper
354 * hook in del_gendisk() to inform the VFS and file system layers.
355 */
356 static int block_device_ejected(struct super_block *sb)
357 {
358 struct inode *bd_inode = sb->s_bdev->bd_inode;
359 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
360
361 return bdi->dev == NULL;
362 }
363
364 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
365 {
366 struct super_block *sb = journal->j_private;
367 struct ext4_sb_info *sbi = EXT4_SB(sb);
368 int error = is_journal_aborted(journal);
369 struct ext4_journal_cb_entry *jce;
370
371 BUG_ON(txn->t_state == T_FINISHED);
372 spin_lock(&sbi->s_md_lock);
373 while (!list_empty(&txn->t_private_list)) {
374 jce = list_entry(txn->t_private_list.next,
375 struct ext4_journal_cb_entry, jce_list);
376 list_del_init(&jce->jce_list);
377 spin_unlock(&sbi->s_md_lock);
378 jce->jce_func(sb, jce, error);
379 spin_lock(&sbi->s_md_lock);
380 }
381 spin_unlock(&sbi->s_md_lock);
382 }
383
384 /* Deal with the reporting of failure conditions on a filesystem such as
385 * inconsistencies detected or read IO failures.
386 *
387 * On ext2, we can store the error state of the filesystem in the
388 * superblock. That is not possible on ext4, because we may have other
389 * write ordering constraints on the superblock which prevent us from
390 * writing it out straight away; and given that the journal is about to
391 * be aborted, we can't rely on the current, or future, transactions to
392 * write out the superblock safely.
393 *
394 * We'll just use the jbd2_journal_abort() error code to record an error in
395 * the journal instead. On recovery, the journal will complain about
396 * that error until we've noted it down and cleared it.
397 */
398
399 static void ext4_handle_error(struct super_block *sb)
400 {
401 if (sb->s_flags & MS_RDONLY)
402 return;
403
404 if (!test_opt(sb, ERRORS_CONT)) {
405 journal_t *journal = EXT4_SB(sb)->s_journal;
406
407 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
408 if (journal)
409 jbd2_journal_abort(journal, -EIO);
410 }
411 if (test_opt(sb, ERRORS_RO)) {
412 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
413 /*
414 * Make sure updated value of ->s_mount_flags will be visible
415 * before ->s_flags update
416 */
417 smp_wmb();
418 sb->s_flags |= MS_RDONLY;
419 }
420 if (test_opt(sb, ERRORS_PANIC)) {
421 if (EXT4_SB(sb)->s_journal &&
422 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
423 return;
424 panic("EXT4-fs (device %s): panic forced after error\n",
425 sb->s_id);
426 }
427 }
428
429 #define ext4_error_ratelimit(sb) \
430 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
431 "EXT4-fs error")
432
433 void __ext4_error(struct super_block *sb, const char *function,
434 unsigned int line, const char *fmt, ...)
435 {
436 struct va_format vaf;
437 va_list args;
438
439 if (ext4_error_ratelimit(sb)) {
440 va_start(args, fmt);
441 vaf.fmt = fmt;
442 vaf.va = &args;
443 printk(KERN_CRIT
444 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
445 sb->s_id, function, line, current->comm, &vaf);
446 va_end(args);
447 }
448 save_error_info(sb, function, line);
449 ext4_handle_error(sb);
450 }
451
452 void __ext4_error_inode(struct inode *inode, const char *function,
453 unsigned int line, ext4_fsblk_t block,
454 const char *fmt, ...)
455 {
456 va_list args;
457 struct va_format vaf;
458 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
459
460 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
461 es->s_last_error_block = cpu_to_le64(block);
462 if (ext4_error_ratelimit(inode->i_sb)) {
463 va_start(args, fmt);
464 vaf.fmt = fmt;
465 vaf.va = &args;
466 if (block)
467 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
468 "inode #%lu: block %llu: comm %s: %pV\n",
469 inode->i_sb->s_id, function, line, inode->i_ino,
470 block, current->comm, &vaf);
471 else
472 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
473 "inode #%lu: comm %s: %pV\n",
474 inode->i_sb->s_id, function, line, inode->i_ino,
475 current->comm, &vaf);
476 va_end(args);
477 }
478 save_error_info(inode->i_sb, function, line);
479 ext4_handle_error(inode->i_sb);
480 }
481
482 void __ext4_error_file(struct file *file, const char *function,
483 unsigned int line, ext4_fsblk_t block,
484 const char *fmt, ...)
485 {
486 va_list args;
487 struct va_format vaf;
488 struct ext4_super_block *es;
489 struct inode *inode = file_inode(file);
490 char pathname[80], *path;
491
492 es = EXT4_SB(inode->i_sb)->s_es;
493 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
494 if (ext4_error_ratelimit(inode->i_sb)) {
495 path = file_path(file, pathname, sizeof(pathname));
496 if (IS_ERR(path))
497 path = "(unknown)";
498 va_start(args, fmt);
499 vaf.fmt = fmt;
500 vaf.va = &args;
501 if (block)
502 printk(KERN_CRIT
503 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
504 "block %llu: comm %s: path %s: %pV\n",
505 inode->i_sb->s_id, function, line, inode->i_ino,
506 block, current->comm, path, &vaf);
507 else
508 printk(KERN_CRIT
509 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
510 "comm %s: path %s: %pV\n",
511 inode->i_sb->s_id, function, line, inode->i_ino,
512 current->comm, path, &vaf);
513 va_end(args);
514 }
515 save_error_info(inode->i_sb, function, line);
516 ext4_handle_error(inode->i_sb);
517 }
518
519 const char *ext4_decode_error(struct super_block *sb, int errno,
520 char nbuf[16])
521 {
522 char *errstr = NULL;
523
524 switch (errno) {
525 case -EFSCORRUPTED:
526 errstr = "Corrupt filesystem";
527 break;
528 case -EFSBADCRC:
529 errstr = "Filesystem failed CRC";
530 break;
531 case -EIO:
532 errstr = "IO failure";
533 break;
534 case -ENOMEM:
535 errstr = "Out of memory";
536 break;
537 case -EROFS:
538 if (!sb || (EXT4_SB(sb)->s_journal &&
539 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
540 errstr = "Journal has aborted";
541 else
542 errstr = "Readonly filesystem";
543 break;
544 default:
545 /* If the caller passed in an extra buffer for unknown
546 * errors, textualise them now. Else we just return
547 * NULL. */
548 if (nbuf) {
549 /* Check for truncated error codes... */
550 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
551 errstr = nbuf;
552 }
553 break;
554 }
555
556 return errstr;
557 }
558
559 /* __ext4_std_error decodes expected errors from journaling functions
560 * automatically and invokes the appropriate error response. */
561
562 void __ext4_std_error(struct super_block *sb, const char *function,
563 unsigned int line, int errno)
564 {
565 char nbuf[16];
566 const char *errstr;
567
568 /* Special case: if the error is EROFS, and we're not already
569 * inside a transaction, then there's really no point in logging
570 * an error. */
571 if (errno == -EROFS && journal_current_handle() == NULL &&
572 (sb->s_flags & MS_RDONLY))
573 return;
574
575 if (ext4_error_ratelimit(sb)) {
576 errstr = ext4_decode_error(sb, errno, nbuf);
577 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
578 sb->s_id, function, line, errstr);
579 }
580
581 save_error_info(sb, function, line);
582 ext4_handle_error(sb);
583 }
584
585 /*
586 * ext4_abort is a much stronger failure handler than ext4_error. The
587 * abort function may be used to deal with unrecoverable failures such
588 * as journal IO errors or ENOMEM at a critical moment in log management.
589 *
590 * We unconditionally force the filesystem into an ABORT|READONLY state,
591 * unless the error response on the fs has been set to panic in which
592 * case we take the easy way out and panic immediately.
593 */
594
595 void __ext4_abort(struct super_block *sb, const char *function,
596 unsigned int line, const char *fmt, ...)
597 {
598 va_list args;
599
600 save_error_info(sb, function, line);
601 va_start(args, fmt);
602 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
603 function, line);
604 vprintk(fmt, args);
605 printk("\n");
606 va_end(args);
607
608 if ((sb->s_flags & MS_RDONLY) == 0) {
609 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
610 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
611 /*
612 * Make sure updated value of ->s_mount_flags will be visible
613 * before ->s_flags update
614 */
615 smp_wmb();
616 sb->s_flags |= MS_RDONLY;
617 if (EXT4_SB(sb)->s_journal)
618 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
619 save_error_info(sb, function, line);
620 }
621 if (test_opt(sb, ERRORS_PANIC)) {
622 if (EXT4_SB(sb)->s_journal &&
623 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
624 return;
625 panic("EXT4-fs panic from previous error\n");
626 }
627 }
628
629 void __ext4_msg(struct super_block *sb,
630 const char *prefix, const char *fmt, ...)
631 {
632 struct va_format vaf;
633 va_list args;
634
635 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
636 return;
637
638 va_start(args, fmt);
639 vaf.fmt = fmt;
640 vaf.va = &args;
641 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
642 va_end(args);
643 }
644
645 #define ext4_warning_ratelimit(sb) \
646 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
647 "EXT4-fs warning")
648
649 void __ext4_warning(struct super_block *sb, const char *function,
650 unsigned int line, const char *fmt, ...)
651 {
652 struct va_format vaf;
653 va_list args;
654
655 if (!ext4_warning_ratelimit(sb))
656 return;
657
658 va_start(args, fmt);
659 vaf.fmt = fmt;
660 vaf.va = &args;
661 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
662 sb->s_id, function, line, &vaf);
663 va_end(args);
664 }
665
666 void __ext4_warning_inode(const struct inode *inode, const char *function,
667 unsigned int line, const char *fmt, ...)
668 {
669 struct va_format vaf;
670 va_list args;
671
672 if (!ext4_warning_ratelimit(inode->i_sb))
673 return;
674
675 va_start(args, fmt);
676 vaf.fmt = fmt;
677 vaf.va = &args;
678 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
679 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
680 function, line, inode->i_ino, current->comm, &vaf);
681 va_end(args);
682 }
683
684 void __ext4_grp_locked_error(const char *function, unsigned int line,
685 struct super_block *sb, ext4_group_t grp,
686 unsigned long ino, ext4_fsblk_t block,
687 const char *fmt, ...)
688 __releases(bitlock)
689 __acquires(bitlock)
690 {
691 struct va_format vaf;
692 va_list args;
693 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
694
695 es->s_last_error_ino = cpu_to_le32(ino);
696 es->s_last_error_block = cpu_to_le64(block);
697 __save_error_info(sb, function, line);
698
699 if (ext4_error_ratelimit(sb)) {
700 va_start(args, fmt);
701 vaf.fmt = fmt;
702 vaf.va = &args;
703 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
704 sb->s_id, function, line, grp);
705 if (ino)
706 printk(KERN_CONT "inode %lu: ", ino);
707 if (block)
708 printk(KERN_CONT "block %llu:",
709 (unsigned long long) block);
710 printk(KERN_CONT "%pV\n", &vaf);
711 va_end(args);
712 }
713
714 if (test_opt(sb, ERRORS_CONT)) {
715 ext4_commit_super(sb, 0);
716 return;
717 }
718
719 ext4_unlock_group(sb, grp);
720 ext4_handle_error(sb);
721 /*
722 * We only get here in the ERRORS_RO case; relocking the group
723 * may be dangerous, but nothing bad will happen since the
724 * filesystem will have already been marked read/only and the
725 * journal has been aborted. We return 1 as a hint to callers
726 * who might what to use the return value from
727 * ext4_grp_locked_error() to distinguish between the
728 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
729 * aggressively from the ext4 function in question, with a
730 * more appropriate error code.
731 */
732 ext4_lock_group(sb, grp);
733 return;
734 }
735
736 void ext4_update_dynamic_rev(struct super_block *sb)
737 {
738 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
739
740 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
741 return;
742
743 ext4_warning(sb,
744 "updating to rev %d because of new feature flag, "
745 "running e2fsck is recommended",
746 EXT4_DYNAMIC_REV);
747
748 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
749 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
750 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
751 /* leave es->s_feature_*compat flags alone */
752 /* es->s_uuid will be set by e2fsck if empty */
753
754 /*
755 * The rest of the superblock fields should be zero, and if not it
756 * means they are likely already in use, so leave them alone. We
757 * can leave it up to e2fsck to clean up any inconsistencies there.
758 */
759 }
760
761 /*
762 * Open the external journal device
763 */
764 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
765 {
766 struct block_device *bdev;
767 char b[BDEVNAME_SIZE];
768
769 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
770 if (IS_ERR(bdev))
771 goto fail;
772 return bdev;
773
774 fail:
775 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
776 __bdevname(dev, b), PTR_ERR(bdev));
777 return NULL;
778 }
779
780 /*
781 * Release the journal device
782 */
783 static void ext4_blkdev_put(struct block_device *bdev)
784 {
785 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
786 }
787
788 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
789 {
790 struct block_device *bdev;
791 bdev = sbi->journal_bdev;
792 if (bdev) {
793 ext4_blkdev_put(bdev);
794 sbi->journal_bdev = NULL;
795 }
796 }
797
798 static inline struct inode *orphan_list_entry(struct list_head *l)
799 {
800 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
801 }
802
803 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
804 {
805 struct list_head *l;
806
807 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
808 le32_to_cpu(sbi->s_es->s_last_orphan));
809
810 printk(KERN_ERR "sb_info orphan list:\n");
811 list_for_each(l, &sbi->s_orphan) {
812 struct inode *inode = orphan_list_entry(l);
813 printk(KERN_ERR " "
814 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
815 inode->i_sb->s_id, inode->i_ino, inode,
816 inode->i_mode, inode->i_nlink,
817 NEXT_ORPHAN(inode));
818 }
819 }
820
821 static void ext4_put_super(struct super_block *sb)
822 {
823 struct ext4_sb_info *sbi = EXT4_SB(sb);
824 struct ext4_super_block *es = sbi->s_es;
825 int i, err;
826
827 ext4_unregister_li_request(sb);
828 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
829
830 flush_workqueue(sbi->rsv_conversion_wq);
831 destroy_workqueue(sbi->rsv_conversion_wq);
832
833 if (sbi->s_journal) {
834 err = jbd2_journal_destroy(sbi->s_journal);
835 sbi->s_journal = NULL;
836 if (err < 0)
837 ext4_abort(sb, "Couldn't clean up the journal");
838 }
839
840 ext4_unregister_sysfs(sb);
841 ext4_es_unregister_shrinker(sbi);
842 del_timer_sync(&sbi->s_err_report);
843 ext4_release_system_zone(sb);
844 ext4_mb_release(sb);
845 ext4_ext_release(sb);
846
847 if (!(sb->s_flags & MS_RDONLY)) {
848 ext4_clear_feature_journal_needs_recovery(sb);
849 es->s_state = cpu_to_le16(sbi->s_mount_state);
850 }
851 if (!(sb->s_flags & MS_RDONLY))
852 ext4_commit_super(sb, 1);
853
854 for (i = 0; i < sbi->s_gdb_count; i++)
855 brelse(sbi->s_group_desc[i]);
856 kvfree(sbi->s_group_desc);
857 kvfree(sbi->s_flex_groups);
858 percpu_counter_destroy(&sbi->s_freeclusters_counter);
859 percpu_counter_destroy(&sbi->s_freeinodes_counter);
860 percpu_counter_destroy(&sbi->s_dirs_counter);
861 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
862 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
863 brelse(sbi->s_sbh);
864 #ifdef CONFIG_QUOTA
865 for (i = 0; i < EXT4_MAXQUOTAS; i++)
866 kfree(sbi->s_qf_names[i]);
867 #endif
868
869 /* Debugging code just in case the in-memory inode orphan list
870 * isn't empty. The on-disk one can be non-empty if we've
871 * detected an error and taken the fs readonly, but the
872 * in-memory list had better be clean by this point. */
873 if (!list_empty(&sbi->s_orphan))
874 dump_orphan_list(sb, sbi);
875 J_ASSERT(list_empty(&sbi->s_orphan));
876
877 sync_blockdev(sb->s_bdev);
878 invalidate_bdev(sb->s_bdev);
879 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
880 /*
881 * Invalidate the journal device's buffers. We don't want them
882 * floating about in memory - the physical journal device may
883 * hotswapped, and it breaks the `ro-after' testing code.
884 */
885 sync_blockdev(sbi->journal_bdev);
886 invalidate_bdev(sbi->journal_bdev);
887 ext4_blkdev_remove(sbi);
888 }
889 if (sbi->s_mb_cache) {
890 ext4_xattr_destroy_cache(sbi->s_mb_cache);
891 sbi->s_mb_cache = NULL;
892 }
893 if (sbi->s_mmp_tsk)
894 kthread_stop(sbi->s_mmp_tsk);
895 sb->s_fs_info = NULL;
896 /*
897 * Now that we are completely done shutting down the
898 * superblock, we need to actually destroy the kobject.
899 */
900 kobject_put(&sbi->s_kobj);
901 wait_for_completion(&sbi->s_kobj_unregister);
902 if (sbi->s_chksum_driver)
903 crypto_free_shash(sbi->s_chksum_driver);
904 kfree(sbi->s_blockgroup_lock);
905 kfree(sbi);
906 }
907
908 static struct kmem_cache *ext4_inode_cachep;
909
910 /*
911 * Called inside transaction, so use GFP_NOFS
912 */
913 static struct inode *ext4_alloc_inode(struct super_block *sb)
914 {
915 struct ext4_inode_info *ei;
916
917 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
918 if (!ei)
919 return NULL;
920
921 ei->vfs_inode.i_version = 1;
922 spin_lock_init(&ei->i_raw_lock);
923 INIT_LIST_HEAD(&ei->i_prealloc_list);
924 spin_lock_init(&ei->i_prealloc_lock);
925 ext4_es_init_tree(&ei->i_es_tree);
926 rwlock_init(&ei->i_es_lock);
927 INIT_LIST_HEAD(&ei->i_es_list);
928 ei->i_es_all_nr = 0;
929 ei->i_es_shk_nr = 0;
930 ei->i_es_shrink_lblk = 0;
931 ei->i_reserved_data_blocks = 0;
932 ei->i_reserved_meta_blocks = 0;
933 ei->i_allocated_meta_blocks = 0;
934 ei->i_da_metadata_calc_len = 0;
935 ei->i_da_metadata_calc_last_lblock = 0;
936 spin_lock_init(&(ei->i_block_reservation_lock));
937 #ifdef CONFIG_QUOTA
938 ei->i_reserved_quota = 0;
939 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
940 #endif
941 ei->jinode = NULL;
942 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
943 spin_lock_init(&ei->i_completed_io_lock);
944 ei->i_sync_tid = 0;
945 ei->i_datasync_tid = 0;
946 atomic_set(&ei->i_unwritten, 0);
947 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
948 return &ei->vfs_inode;
949 }
950
951 static int ext4_drop_inode(struct inode *inode)
952 {
953 int drop = generic_drop_inode(inode);
954
955 trace_ext4_drop_inode(inode, drop);
956 return drop;
957 }
958
959 static void ext4_i_callback(struct rcu_head *head)
960 {
961 struct inode *inode = container_of(head, struct inode, i_rcu);
962 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
963 }
964
965 static void ext4_destroy_inode(struct inode *inode)
966 {
967 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
968 ext4_msg(inode->i_sb, KERN_ERR,
969 "Inode %lu (%p): orphan list check failed!",
970 inode->i_ino, EXT4_I(inode));
971 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
972 EXT4_I(inode), sizeof(struct ext4_inode_info),
973 true);
974 dump_stack();
975 }
976 call_rcu(&inode->i_rcu, ext4_i_callback);
977 }
978
979 static void init_once(void *foo)
980 {
981 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
982
983 INIT_LIST_HEAD(&ei->i_orphan);
984 init_rwsem(&ei->xattr_sem);
985 init_rwsem(&ei->i_data_sem);
986 init_rwsem(&ei->i_mmap_sem);
987 inode_init_once(&ei->vfs_inode);
988 }
989
990 static int __init init_inodecache(void)
991 {
992 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
993 sizeof(struct ext4_inode_info),
994 0, (SLAB_RECLAIM_ACCOUNT|
995 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
996 init_once);
997 if (ext4_inode_cachep == NULL)
998 return -ENOMEM;
999 return 0;
1000 }
1001
1002 static void destroy_inodecache(void)
1003 {
1004 /*
1005 * Make sure all delayed rcu free inodes are flushed before we
1006 * destroy cache.
1007 */
1008 rcu_barrier();
1009 kmem_cache_destroy(ext4_inode_cachep);
1010 }
1011
1012 void ext4_clear_inode(struct inode *inode)
1013 {
1014 invalidate_inode_buffers(inode);
1015 clear_inode(inode);
1016 dquot_drop(inode);
1017 ext4_discard_preallocations(inode);
1018 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1019 if (EXT4_I(inode)->jinode) {
1020 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1021 EXT4_I(inode)->jinode);
1022 jbd2_free_inode(EXT4_I(inode)->jinode);
1023 EXT4_I(inode)->jinode = NULL;
1024 }
1025 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1026 fscrypt_put_encryption_info(inode, NULL);
1027 #endif
1028 }
1029
1030 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1031 u64 ino, u32 generation)
1032 {
1033 struct inode *inode;
1034
1035 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1036 return ERR_PTR(-ESTALE);
1037 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1038 return ERR_PTR(-ESTALE);
1039
1040 /* iget isn't really right if the inode is currently unallocated!!
1041 *
1042 * ext4_read_inode will return a bad_inode if the inode had been
1043 * deleted, so we should be safe.
1044 *
1045 * Currently we don't know the generation for parent directory, so
1046 * a generation of 0 means "accept any"
1047 */
1048 inode = ext4_iget_normal(sb, ino);
1049 if (IS_ERR(inode))
1050 return ERR_CAST(inode);
1051 if (generation && inode->i_generation != generation) {
1052 iput(inode);
1053 return ERR_PTR(-ESTALE);
1054 }
1055
1056 return inode;
1057 }
1058
1059 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1060 int fh_len, int fh_type)
1061 {
1062 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1063 ext4_nfs_get_inode);
1064 }
1065
1066 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1067 int fh_len, int fh_type)
1068 {
1069 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1070 ext4_nfs_get_inode);
1071 }
1072
1073 /*
1074 * Try to release metadata pages (indirect blocks, directories) which are
1075 * mapped via the block device. Since these pages could have journal heads
1076 * which would prevent try_to_free_buffers() from freeing them, we must use
1077 * jbd2 layer's try_to_free_buffers() function to release them.
1078 */
1079 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1080 gfp_t wait)
1081 {
1082 journal_t *journal = EXT4_SB(sb)->s_journal;
1083
1084 WARN_ON(PageChecked(page));
1085 if (!page_has_buffers(page))
1086 return 0;
1087 if (journal)
1088 return jbd2_journal_try_to_free_buffers(journal, page,
1089 wait & ~__GFP_DIRECT_RECLAIM);
1090 return try_to_free_buffers(page);
1091 }
1092
1093 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1094 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1095 {
1096 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1097 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1098 }
1099
1100 static int ext4_key_prefix(struct inode *inode, u8 **key)
1101 {
1102 *key = EXT4_SB(inode->i_sb)->key_prefix;
1103 return EXT4_SB(inode->i_sb)->key_prefix_size;
1104 }
1105
1106 static int ext4_prepare_context(struct inode *inode)
1107 {
1108 return ext4_convert_inline_data(inode);
1109 }
1110
1111 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1112 void *fs_data)
1113 {
1114 handle_t *handle;
1115 int res, res2;
1116
1117 /* fs_data is null when internally used. */
1118 if (fs_data) {
1119 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1120 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx,
1121 len, 0);
1122 if (!res) {
1123 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1124 ext4_clear_inode_state(inode,
1125 EXT4_STATE_MAY_INLINE_DATA);
1126 }
1127 return res;
1128 }
1129
1130 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1131 ext4_jbd2_credits_xattr(inode));
1132 if (IS_ERR(handle))
1133 return PTR_ERR(handle);
1134
1135 res = ext4_xattr_set(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1136 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx,
1137 len, 0);
1138 if (!res) {
1139 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1140 res = ext4_mark_inode_dirty(handle, inode);
1141 if (res)
1142 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1143 }
1144 res2 = ext4_journal_stop(handle);
1145 if (!res)
1146 res = res2;
1147 return res;
1148 }
1149
1150 static int ext4_dummy_context(struct inode *inode)
1151 {
1152 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1153 }
1154
1155 static unsigned ext4_max_namelen(struct inode *inode)
1156 {
1157 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1158 EXT4_NAME_LEN;
1159 }
1160
1161 static struct fscrypt_operations ext4_cryptops = {
1162 .get_context = ext4_get_context,
1163 .key_prefix = ext4_key_prefix,
1164 .prepare_context = ext4_prepare_context,
1165 .set_context = ext4_set_context,
1166 .dummy_context = ext4_dummy_context,
1167 .is_encrypted = ext4_encrypted_inode,
1168 .empty_dir = ext4_empty_dir,
1169 .max_namelen = ext4_max_namelen,
1170 };
1171 #else
1172 static struct fscrypt_operations ext4_cryptops = {
1173 .is_encrypted = ext4_encrypted_inode,
1174 };
1175 #endif
1176
1177 #ifdef CONFIG_QUOTA
1178 static char *quotatypes[] = INITQFNAMES;
1179 #define QTYPE2NAME(t) (quotatypes[t])
1180
1181 static int ext4_write_dquot(struct dquot *dquot);
1182 static int ext4_acquire_dquot(struct dquot *dquot);
1183 static int ext4_release_dquot(struct dquot *dquot);
1184 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1185 static int ext4_write_info(struct super_block *sb, int type);
1186 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1187 struct path *path);
1188 static int ext4_quota_off(struct super_block *sb, int type);
1189 static int ext4_quota_on_mount(struct super_block *sb, int type);
1190 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1191 size_t len, loff_t off);
1192 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1193 const char *data, size_t len, loff_t off);
1194 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1195 unsigned int flags);
1196 static int ext4_enable_quotas(struct super_block *sb);
1197 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1198
1199 static struct dquot **ext4_get_dquots(struct inode *inode)
1200 {
1201 return EXT4_I(inode)->i_dquot;
1202 }
1203
1204 static const struct dquot_operations ext4_quota_operations = {
1205 .get_reserved_space = ext4_get_reserved_space,
1206 .write_dquot = ext4_write_dquot,
1207 .acquire_dquot = ext4_acquire_dquot,
1208 .release_dquot = ext4_release_dquot,
1209 .mark_dirty = ext4_mark_dquot_dirty,
1210 .write_info = ext4_write_info,
1211 .alloc_dquot = dquot_alloc,
1212 .destroy_dquot = dquot_destroy,
1213 .get_projid = ext4_get_projid,
1214 .get_next_id = ext4_get_next_id,
1215 };
1216
1217 static const struct quotactl_ops ext4_qctl_operations = {
1218 .quota_on = ext4_quota_on,
1219 .quota_off = ext4_quota_off,
1220 .quota_sync = dquot_quota_sync,
1221 .get_state = dquot_get_state,
1222 .set_info = dquot_set_dqinfo,
1223 .get_dqblk = dquot_get_dqblk,
1224 .set_dqblk = dquot_set_dqblk,
1225 .get_nextdqblk = dquot_get_next_dqblk,
1226 };
1227 #endif
1228
1229 static const struct super_operations ext4_sops = {
1230 .alloc_inode = ext4_alloc_inode,
1231 .destroy_inode = ext4_destroy_inode,
1232 .write_inode = ext4_write_inode,
1233 .dirty_inode = ext4_dirty_inode,
1234 .drop_inode = ext4_drop_inode,
1235 .evict_inode = ext4_evict_inode,
1236 .put_super = ext4_put_super,
1237 .sync_fs = ext4_sync_fs,
1238 .freeze_fs = ext4_freeze,
1239 .unfreeze_fs = ext4_unfreeze,
1240 .statfs = ext4_statfs,
1241 .remount_fs = ext4_remount,
1242 .show_options = ext4_show_options,
1243 #ifdef CONFIG_QUOTA
1244 .quota_read = ext4_quota_read,
1245 .quota_write = ext4_quota_write,
1246 .get_dquots = ext4_get_dquots,
1247 #endif
1248 .bdev_try_to_free_page = bdev_try_to_free_page,
1249 };
1250
1251 static const struct export_operations ext4_export_ops = {
1252 .fh_to_dentry = ext4_fh_to_dentry,
1253 .fh_to_parent = ext4_fh_to_parent,
1254 .get_parent = ext4_get_parent,
1255 };
1256
1257 enum {
1258 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1259 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1260 Opt_nouid32, Opt_debug, Opt_removed,
1261 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1262 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1263 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1264 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1265 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1266 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1267 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1268 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1269 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1270 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1271 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1272 Opt_lazytime, Opt_nolazytime,
1273 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1274 Opt_inode_readahead_blks, Opt_journal_ioprio,
1275 Opt_dioread_nolock, Opt_dioread_lock,
1276 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1277 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1278 };
1279
1280 static const match_table_t tokens = {
1281 {Opt_bsd_df, "bsddf"},
1282 {Opt_minix_df, "minixdf"},
1283 {Opt_grpid, "grpid"},
1284 {Opt_grpid, "bsdgroups"},
1285 {Opt_nogrpid, "nogrpid"},
1286 {Opt_nogrpid, "sysvgroups"},
1287 {Opt_resgid, "resgid=%u"},
1288 {Opt_resuid, "resuid=%u"},
1289 {Opt_sb, "sb=%u"},
1290 {Opt_err_cont, "errors=continue"},
1291 {Opt_err_panic, "errors=panic"},
1292 {Opt_err_ro, "errors=remount-ro"},
1293 {Opt_nouid32, "nouid32"},
1294 {Opt_debug, "debug"},
1295 {Opt_removed, "oldalloc"},
1296 {Opt_removed, "orlov"},
1297 {Opt_user_xattr, "user_xattr"},
1298 {Opt_nouser_xattr, "nouser_xattr"},
1299 {Opt_acl, "acl"},
1300 {Opt_noacl, "noacl"},
1301 {Opt_noload, "norecovery"},
1302 {Opt_noload, "noload"},
1303 {Opt_removed, "nobh"},
1304 {Opt_removed, "bh"},
1305 {Opt_commit, "commit=%u"},
1306 {Opt_min_batch_time, "min_batch_time=%u"},
1307 {Opt_max_batch_time, "max_batch_time=%u"},
1308 {Opt_journal_dev, "journal_dev=%u"},
1309 {Opt_journal_path, "journal_path=%s"},
1310 {Opt_journal_checksum, "journal_checksum"},
1311 {Opt_nojournal_checksum, "nojournal_checksum"},
1312 {Opt_journal_async_commit, "journal_async_commit"},
1313 {Opt_abort, "abort"},
1314 {Opt_data_journal, "data=journal"},
1315 {Opt_data_ordered, "data=ordered"},
1316 {Opt_data_writeback, "data=writeback"},
1317 {Opt_data_err_abort, "data_err=abort"},
1318 {Opt_data_err_ignore, "data_err=ignore"},
1319 {Opt_offusrjquota, "usrjquota="},
1320 {Opt_usrjquota, "usrjquota=%s"},
1321 {Opt_offgrpjquota, "grpjquota="},
1322 {Opt_grpjquota, "grpjquota=%s"},
1323 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1324 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1325 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1326 {Opt_grpquota, "grpquota"},
1327 {Opt_noquota, "noquota"},
1328 {Opt_quota, "quota"},
1329 {Opt_usrquota, "usrquota"},
1330 {Opt_prjquota, "prjquota"},
1331 {Opt_barrier, "barrier=%u"},
1332 {Opt_barrier, "barrier"},
1333 {Opt_nobarrier, "nobarrier"},
1334 {Opt_i_version, "i_version"},
1335 {Opt_dax, "dax"},
1336 {Opt_stripe, "stripe=%u"},
1337 {Opt_delalloc, "delalloc"},
1338 {Opt_lazytime, "lazytime"},
1339 {Opt_nolazytime, "nolazytime"},
1340 {Opt_nodelalloc, "nodelalloc"},
1341 {Opt_removed, "mblk_io_submit"},
1342 {Opt_removed, "nomblk_io_submit"},
1343 {Opt_block_validity, "block_validity"},
1344 {Opt_noblock_validity, "noblock_validity"},
1345 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1346 {Opt_journal_ioprio, "journal_ioprio=%u"},
1347 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1348 {Opt_auto_da_alloc, "auto_da_alloc"},
1349 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1350 {Opt_dioread_nolock, "dioread_nolock"},
1351 {Opt_dioread_lock, "dioread_lock"},
1352 {Opt_discard, "discard"},
1353 {Opt_nodiscard, "nodiscard"},
1354 {Opt_init_itable, "init_itable=%u"},
1355 {Opt_init_itable, "init_itable"},
1356 {Opt_noinit_itable, "noinit_itable"},
1357 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1358 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1359 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1360 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1361 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1362 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1363 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1364 {Opt_err, NULL},
1365 };
1366
1367 static ext4_fsblk_t get_sb_block(void **data)
1368 {
1369 ext4_fsblk_t sb_block;
1370 char *options = (char *) *data;
1371
1372 if (!options || strncmp(options, "sb=", 3) != 0)
1373 return 1; /* Default location */
1374
1375 options += 3;
1376 /* TODO: use simple_strtoll with >32bit ext4 */
1377 sb_block = simple_strtoul(options, &options, 0);
1378 if (*options && *options != ',') {
1379 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1380 (char *) *data);
1381 return 1;
1382 }
1383 if (*options == ',')
1384 options++;
1385 *data = (void *) options;
1386
1387 return sb_block;
1388 }
1389
1390 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1391 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1392 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1393
1394 #ifdef CONFIG_QUOTA
1395 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1396 {
1397 struct ext4_sb_info *sbi = EXT4_SB(sb);
1398 char *qname;
1399 int ret = -1;
1400
1401 if (sb_any_quota_loaded(sb) &&
1402 !sbi->s_qf_names[qtype]) {
1403 ext4_msg(sb, KERN_ERR,
1404 "Cannot change journaled "
1405 "quota options when quota turned on");
1406 return -1;
1407 }
1408 if (ext4_has_feature_quota(sb)) {
1409 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1410 "ignored when QUOTA feature is enabled");
1411 return 1;
1412 }
1413 qname = match_strdup(args);
1414 if (!qname) {
1415 ext4_msg(sb, KERN_ERR,
1416 "Not enough memory for storing quotafile name");
1417 return -1;
1418 }
1419 if (sbi->s_qf_names[qtype]) {
1420 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1421 ret = 1;
1422 else
1423 ext4_msg(sb, KERN_ERR,
1424 "%s quota file already specified",
1425 QTYPE2NAME(qtype));
1426 goto errout;
1427 }
1428 if (strchr(qname, '/')) {
1429 ext4_msg(sb, KERN_ERR,
1430 "quotafile must be on filesystem root");
1431 goto errout;
1432 }
1433 sbi->s_qf_names[qtype] = qname;
1434 set_opt(sb, QUOTA);
1435 return 1;
1436 errout:
1437 kfree(qname);
1438 return ret;
1439 }
1440
1441 static int clear_qf_name(struct super_block *sb, int qtype)
1442 {
1443
1444 struct ext4_sb_info *sbi = EXT4_SB(sb);
1445
1446 if (sb_any_quota_loaded(sb) &&
1447 sbi->s_qf_names[qtype]) {
1448 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1449 " when quota turned on");
1450 return -1;
1451 }
1452 kfree(sbi->s_qf_names[qtype]);
1453 sbi->s_qf_names[qtype] = NULL;
1454 return 1;
1455 }
1456 #endif
1457
1458 #define MOPT_SET 0x0001
1459 #define MOPT_CLEAR 0x0002
1460 #define MOPT_NOSUPPORT 0x0004
1461 #define MOPT_EXPLICIT 0x0008
1462 #define MOPT_CLEAR_ERR 0x0010
1463 #define MOPT_GTE0 0x0020
1464 #ifdef CONFIG_QUOTA
1465 #define MOPT_Q 0
1466 #define MOPT_QFMT 0x0040
1467 #else
1468 #define MOPT_Q MOPT_NOSUPPORT
1469 #define MOPT_QFMT MOPT_NOSUPPORT
1470 #endif
1471 #define MOPT_DATAJ 0x0080
1472 #define MOPT_NO_EXT2 0x0100
1473 #define MOPT_NO_EXT3 0x0200
1474 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1475 #define MOPT_STRING 0x0400
1476
1477 static const struct mount_opts {
1478 int token;
1479 int mount_opt;
1480 int flags;
1481 } ext4_mount_opts[] = {
1482 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1483 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1484 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1485 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1486 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1487 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1488 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1489 MOPT_EXT4_ONLY | MOPT_SET},
1490 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1491 MOPT_EXT4_ONLY | MOPT_CLEAR},
1492 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1493 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1494 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1495 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1496 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1497 MOPT_EXT4_ONLY | MOPT_CLEAR},
1498 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1499 MOPT_EXT4_ONLY | MOPT_CLEAR},
1500 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1501 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1502 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1503 EXT4_MOUNT_JOURNAL_CHECKSUM),
1504 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1505 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1506 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1507 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1508 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1509 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1510 MOPT_NO_EXT2},
1511 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1512 MOPT_NO_EXT2},
1513 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1514 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1515 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1516 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1517 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1518 {Opt_commit, 0, MOPT_GTE0},
1519 {Opt_max_batch_time, 0, MOPT_GTE0},
1520 {Opt_min_batch_time, 0, MOPT_GTE0},
1521 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1522 {Opt_init_itable, 0, MOPT_GTE0},
1523 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1524 {Opt_stripe, 0, MOPT_GTE0},
1525 {Opt_resuid, 0, MOPT_GTE0},
1526 {Opt_resgid, 0, MOPT_GTE0},
1527 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1528 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1529 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1530 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1531 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1532 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1533 MOPT_NO_EXT2 | MOPT_DATAJ},
1534 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1535 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1536 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1537 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1538 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1539 #else
1540 {Opt_acl, 0, MOPT_NOSUPPORT},
1541 {Opt_noacl, 0, MOPT_NOSUPPORT},
1542 #endif
1543 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1544 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1545 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1546 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1547 MOPT_SET | MOPT_Q},
1548 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1549 MOPT_SET | MOPT_Q},
1550 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1551 MOPT_SET | MOPT_Q},
1552 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1553 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1554 MOPT_CLEAR | MOPT_Q},
1555 {Opt_usrjquota, 0, MOPT_Q},
1556 {Opt_grpjquota, 0, MOPT_Q},
1557 {Opt_offusrjquota, 0, MOPT_Q},
1558 {Opt_offgrpjquota, 0, MOPT_Q},
1559 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1560 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1561 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1562 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1563 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1564 {Opt_err, 0, 0}
1565 };
1566
1567 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1568 substring_t *args, unsigned long *journal_devnum,
1569 unsigned int *journal_ioprio, int is_remount)
1570 {
1571 struct ext4_sb_info *sbi = EXT4_SB(sb);
1572 const struct mount_opts *m;
1573 kuid_t uid;
1574 kgid_t gid;
1575 int arg = 0;
1576
1577 #ifdef CONFIG_QUOTA
1578 if (token == Opt_usrjquota)
1579 return set_qf_name(sb, USRQUOTA, &args[0]);
1580 else if (token == Opt_grpjquota)
1581 return set_qf_name(sb, GRPQUOTA, &args[0]);
1582 else if (token == Opt_offusrjquota)
1583 return clear_qf_name(sb, USRQUOTA);
1584 else if (token == Opt_offgrpjquota)
1585 return clear_qf_name(sb, GRPQUOTA);
1586 #endif
1587 switch (token) {
1588 case Opt_noacl:
1589 case Opt_nouser_xattr:
1590 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1591 break;
1592 case Opt_sb:
1593 return 1; /* handled by get_sb_block() */
1594 case Opt_removed:
1595 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1596 return 1;
1597 case Opt_abort:
1598 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1599 return 1;
1600 case Opt_i_version:
1601 sb->s_flags |= MS_I_VERSION;
1602 return 1;
1603 case Opt_lazytime:
1604 sb->s_flags |= MS_LAZYTIME;
1605 return 1;
1606 case Opt_nolazytime:
1607 sb->s_flags &= ~MS_LAZYTIME;
1608 return 1;
1609 }
1610
1611 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1612 if (token == m->token)
1613 break;
1614
1615 if (m->token == Opt_err) {
1616 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1617 "or missing value", opt);
1618 return -1;
1619 }
1620
1621 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1622 ext4_msg(sb, KERN_ERR,
1623 "Mount option \"%s\" incompatible with ext2", opt);
1624 return -1;
1625 }
1626 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1627 ext4_msg(sb, KERN_ERR,
1628 "Mount option \"%s\" incompatible with ext3", opt);
1629 return -1;
1630 }
1631
1632 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1633 return -1;
1634 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1635 return -1;
1636 if (m->flags & MOPT_EXPLICIT) {
1637 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1638 set_opt2(sb, EXPLICIT_DELALLOC);
1639 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1640 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1641 } else
1642 return -1;
1643 }
1644 if (m->flags & MOPT_CLEAR_ERR)
1645 clear_opt(sb, ERRORS_MASK);
1646 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1647 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1648 "options when quota turned on");
1649 return -1;
1650 }
1651
1652 if (m->flags & MOPT_NOSUPPORT) {
1653 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1654 } else if (token == Opt_commit) {
1655 if (arg == 0)
1656 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1657 sbi->s_commit_interval = HZ * arg;
1658 } else if (token == Opt_max_batch_time) {
1659 sbi->s_max_batch_time = arg;
1660 } else if (token == Opt_min_batch_time) {
1661 sbi->s_min_batch_time = arg;
1662 } else if (token == Opt_inode_readahead_blks) {
1663 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1664 ext4_msg(sb, KERN_ERR,
1665 "EXT4-fs: inode_readahead_blks must be "
1666 "0 or a power of 2 smaller than 2^31");
1667 return -1;
1668 }
1669 sbi->s_inode_readahead_blks = arg;
1670 } else if (token == Opt_init_itable) {
1671 set_opt(sb, INIT_INODE_TABLE);
1672 if (!args->from)
1673 arg = EXT4_DEF_LI_WAIT_MULT;
1674 sbi->s_li_wait_mult = arg;
1675 } else if (token == Opt_max_dir_size_kb) {
1676 sbi->s_max_dir_size_kb = arg;
1677 } else if (token == Opt_stripe) {
1678 sbi->s_stripe = arg;
1679 } else if (token == Opt_resuid) {
1680 uid = make_kuid(current_user_ns(), arg);
1681 if (!uid_valid(uid)) {
1682 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1683 return -1;
1684 }
1685 sbi->s_resuid = uid;
1686 } else if (token == Opt_resgid) {
1687 gid = make_kgid(current_user_ns(), arg);
1688 if (!gid_valid(gid)) {
1689 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1690 return -1;
1691 }
1692 sbi->s_resgid = gid;
1693 } else if (token == Opt_journal_dev) {
1694 if (is_remount) {
1695 ext4_msg(sb, KERN_ERR,
1696 "Cannot specify journal on remount");
1697 return -1;
1698 }
1699 *journal_devnum = arg;
1700 } else if (token == Opt_journal_path) {
1701 char *journal_path;
1702 struct inode *journal_inode;
1703 struct path path;
1704 int error;
1705
1706 if (is_remount) {
1707 ext4_msg(sb, KERN_ERR,
1708 "Cannot specify journal on remount");
1709 return -1;
1710 }
1711 journal_path = match_strdup(&args[0]);
1712 if (!journal_path) {
1713 ext4_msg(sb, KERN_ERR, "error: could not dup "
1714 "journal device string");
1715 return -1;
1716 }
1717
1718 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1719 if (error) {
1720 ext4_msg(sb, KERN_ERR, "error: could not find "
1721 "journal device path: error %d", error);
1722 kfree(journal_path);
1723 return -1;
1724 }
1725
1726 journal_inode = d_inode(path.dentry);
1727 if (!S_ISBLK(journal_inode->i_mode)) {
1728 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1729 "is not a block device", journal_path);
1730 path_put(&path);
1731 kfree(journal_path);
1732 return -1;
1733 }
1734
1735 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1736 path_put(&path);
1737 kfree(journal_path);
1738 } else if (token == Opt_journal_ioprio) {
1739 if (arg > 7) {
1740 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1741 " (must be 0-7)");
1742 return -1;
1743 }
1744 *journal_ioprio =
1745 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1746 } else if (token == Opt_test_dummy_encryption) {
1747 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1748 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1749 ext4_msg(sb, KERN_WARNING,
1750 "Test dummy encryption mode enabled");
1751 #else
1752 ext4_msg(sb, KERN_WARNING,
1753 "Test dummy encryption mount option ignored");
1754 #endif
1755 } else if (m->flags & MOPT_DATAJ) {
1756 if (is_remount) {
1757 if (!sbi->s_journal)
1758 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1759 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1760 ext4_msg(sb, KERN_ERR,
1761 "Cannot change data mode on remount");
1762 return -1;
1763 }
1764 } else {
1765 clear_opt(sb, DATA_FLAGS);
1766 sbi->s_mount_opt |= m->mount_opt;
1767 }
1768 #ifdef CONFIG_QUOTA
1769 } else if (m->flags & MOPT_QFMT) {
1770 if (sb_any_quota_loaded(sb) &&
1771 sbi->s_jquota_fmt != m->mount_opt) {
1772 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1773 "quota options when quota turned on");
1774 return -1;
1775 }
1776 if (ext4_has_feature_quota(sb)) {
1777 ext4_msg(sb, KERN_INFO,
1778 "Quota format mount options ignored "
1779 "when QUOTA feature is enabled");
1780 return 1;
1781 }
1782 sbi->s_jquota_fmt = m->mount_opt;
1783 #endif
1784 } else if (token == Opt_dax) {
1785 #ifdef CONFIG_FS_DAX
1786 ext4_msg(sb, KERN_WARNING,
1787 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1788 sbi->s_mount_opt |= m->mount_opt;
1789 #else
1790 ext4_msg(sb, KERN_INFO, "dax option not supported");
1791 return -1;
1792 #endif
1793 } else if (token == Opt_data_err_abort) {
1794 sbi->s_mount_opt |= m->mount_opt;
1795 } else if (token == Opt_data_err_ignore) {
1796 sbi->s_mount_opt &= ~m->mount_opt;
1797 } else {
1798 if (!args->from)
1799 arg = 1;
1800 if (m->flags & MOPT_CLEAR)
1801 arg = !arg;
1802 else if (unlikely(!(m->flags & MOPT_SET))) {
1803 ext4_msg(sb, KERN_WARNING,
1804 "buggy handling of option %s", opt);
1805 WARN_ON(1);
1806 return -1;
1807 }
1808 if (arg != 0)
1809 sbi->s_mount_opt |= m->mount_opt;
1810 else
1811 sbi->s_mount_opt &= ~m->mount_opt;
1812 }
1813 return 1;
1814 }
1815
1816 static int parse_options(char *options, struct super_block *sb,
1817 unsigned long *journal_devnum,
1818 unsigned int *journal_ioprio,
1819 int is_remount)
1820 {
1821 struct ext4_sb_info *sbi = EXT4_SB(sb);
1822 char *p;
1823 substring_t args[MAX_OPT_ARGS];
1824 int token;
1825
1826 if (!options)
1827 return 1;
1828
1829 while ((p = strsep(&options, ",")) != NULL) {
1830 if (!*p)
1831 continue;
1832 /*
1833 * Initialize args struct so we know whether arg was
1834 * found; some options take optional arguments.
1835 */
1836 args[0].to = args[0].from = NULL;
1837 token = match_token(p, tokens, args);
1838 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1839 journal_ioprio, is_remount) < 0)
1840 return 0;
1841 }
1842 #ifdef CONFIG_QUOTA
1843 /*
1844 * We do the test below only for project quotas. 'usrquota' and
1845 * 'grpquota' mount options are allowed even without quota feature
1846 * to support legacy quotas in quota files.
1847 */
1848 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1849 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1850 "Cannot enable project quota enforcement.");
1851 return 0;
1852 }
1853 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1854 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1855 clear_opt(sb, USRQUOTA);
1856
1857 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1858 clear_opt(sb, GRPQUOTA);
1859
1860 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1861 ext4_msg(sb, KERN_ERR, "old and new quota "
1862 "format mixing");
1863 return 0;
1864 }
1865
1866 if (!sbi->s_jquota_fmt) {
1867 ext4_msg(sb, KERN_ERR, "journaled quota format "
1868 "not specified");
1869 return 0;
1870 }
1871 }
1872 #endif
1873 if (test_opt(sb, DIOREAD_NOLOCK)) {
1874 int blocksize =
1875 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1876
1877 if (blocksize < PAGE_SIZE) {
1878 ext4_msg(sb, KERN_ERR, "can't mount with "
1879 "dioread_nolock if block size != PAGE_SIZE");
1880 return 0;
1881 }
1882 }
1883 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1884 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1885 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1886 "in data=ordered mode");
1887 return 0;
1888 }
1889 return 1;
1890 }
1891
1892 static inline void ext4_show_quota_options(struct seq_file *seq,
1893 struct super_block *sb)
1894 {
1895 #if defined(CONFIG_QUOTA)
1896 struct ext4_sb_info *sbi = EXT4_SB(sb);
1897
1898 if (sbi->s_jquota_fmt) {
1899 char *fmtname = "";
1900
1901 switch (sbi->s_jquota_fmt) {
1902 case QFMT_VFS_OLD:
1903 fmtname = "vfsold";
1904 break;
1905 case QFMT_VFS_V0:
1906 fmtname = "vfsv0";
1907 break;
1908 case QFMT_VFS_V1:
1909 fmtname = "vfsv1";
1910 break;
1911 }
1912 seq_printf(seq, ",jqfmt=%s", fmtname);
1913 }
1914
1915 if (sbi->s_qf_names[USRQUOTA])
1916 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1917
1918 if (sbi->s_qf_names[GRPQUOTA])
1919 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1920 #endif
1921 }
1922
1923 static const char *token2str(int token)
1924 {
1925 const struct match_token *t;
1926
1927 for (t = tokens; t->token != Opt_err; t++)
1928 if (t->token == token && !strchr(t->pattern, '='))
1929 break;
1930 return t->pattern;
1931 }
1932
1933 /*
1934 * Show an option if
1935 * - it's set to a non-default value OR
1936 * - if the per-sb default is different from the global default
1937 */
1938 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1939 int nodefs)
1940 {
1941 struct ext4_sb_info *sbi = EXT4_SB(sb);
1942 struct ext4_super_block *es = sbi->s_es;
1943 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1944 const struct mount_opts *m;
1945 char sep = nodefs ? '\n' : ',';
1946
1947 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1948 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1949
1950 if (sbi->s_sb_block != 1)
1951 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1952
1953 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1954 int want_set = m->flags & MOPT_SET;
1955 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1956 (m->flags & MOPT_CLEAR_ERR))
1957 continue;
1958 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1959 continue; /* skip if same as the default */
1960 if ((want_set &&
1961 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1962 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1963 continue; /* select Opt_noFoo vs Opt_Foo */
1964 SEQ_OPTS_PRINT("%s", token2str(m->token));
1965 }
1966
1967 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1968 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1969 SEQ_OPTS_PRINT("resuid=%u",
1970 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1971 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1972 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1973 SEQ_OPTS_PRINT("resgid=%u",
1974 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1975 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1976 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1977 SEQ_OPTS_PUTS("errors=remount-ro");
1978 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1979 SEQ_OPTS_PUTS("errors=continue");
1980 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1981 SEQ_OPTS_PUTS("errors=panic");
1982 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1983 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1984 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1985 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1986 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1987 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1988 if (sb->s_flags & MS_I_VERSION)
1989 SEQ_OPTS_PUTS("i_version");
1990 if (nodefs || sbi->s_stripe)
1991 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1992 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1993 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1994 SEQ_OPTS_PUTS("data=journal");
1995 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1996 SEQ_OPTS_PUTS("data=ordered");
1997 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1998 SEQ_OPTS_PUTS("data=writeback");
1999 }
2000 if (nodefs ||
2001 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2002 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2003 sbi->s_inode_readahead_blks);
2004
2005 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2006 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2007 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2008 if (nodefs || sbi->s_max_dir_size_kb)
2009 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2010 if (test_opt(sb, DATA_ERR_ABORT))
2011 SEQ_OPTS_PUTS("data_err=abort");
2012
2013 ext4_show_quota_options(seq, sb);
2014 return 0;
2015 }
2016
2017 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2018 {
2019 return _ext4_show_options(seq, root->d_sb, 0);
2020 }
2021
2022 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2023 {
2024 struct super_block *sb = seq->private;
2025 int rc;
2026
2027 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2028 rc = _ext4_show_options(seq, sb, 1);
2029 seq_puts(seq, "\n");
2030 return rc;
2031 }
2032
2033 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2034 int read_only)
2035 {
2036 struct ext4_sb_info *sbi = EXT4_SB(sb);
2037 int res = 0;
2038
2039 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2040 ext4_msg(sb, KERN_ERR, "revision level too high, "
2041 "forcing read-only mode");
2042 res = MS_RDONLY;
2043 }
2044 if (read_only)
2045 goto done;
2046 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2047 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2048 "running e2fsck is recommended");
2049 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2050 ext4_msg(sb, KERN_WARNING,
2051 "warning: mounting fs with errors, "
2052 "running e2fsck is recommended");
2053 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2054 le16_to_cpu(es->s_mnt_count) >=
2055 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2056 ext4_msg(sb, KERN_WARNING,
2057 "warning: maximal mount count reached, "
2058 "running e2fsck is recommended");
2059 else if (le32_to_cpu(es->s_checkinterval) &&
2060 (le32_to_cpu(es->s_lastcheck) +
2061 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2062 ext4_msg(sb, KERN_WARNING,
2063 "warning: checktime reached, "
2064 "running e2fsck is recommended");
2065 if (!sbi->s_journal)
2066 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2067 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2068 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2069 le16_add_cpu(&es->s_mnt_count, 1);
2070 es->s_mtime = cpu_to_le32(get_seconds());
2071 ext4_update_dynamic_rev(sb);
2072 if (sbi->s_journal)
2073 ext4_set_feature_journal_needs_recovery(sb);
2074
2075 ext4_commit_super(sb, 1);
2076 done:
2077 if (test_opt(sb, DEBUG))
2078 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2079 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2080 sb->s_blocksize,
2081 sbi->s_groups_count,
2082 EXT4_BLOCKS_PER_GROUP(sb),
2083 EXT4_INODES_PER_GROUP(sb),
2084 sbi->s_mount_opt, sbi->s_mount_opt2);
2085
2086 cleancache_init_fs(sb);
2087 return res;
2088 }
2089
2090 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2091 {
2092 struct ext4_sb_info *sbi = EXT4_SB(sb);
2093 struct flex_groups *new_groups;
2094 int size;
2095
2096 if (!sbi->s_log_groups_per_flex)
2097 return 0;
2098
2099 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2100 if (size <= sbi->s_flex_groups_allocated)
2101 return 0;
2102
2103 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2104 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2105 if (!new_groups) {
2106 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2107 size / (int) sizeof(struct flex_groups));
2108 return -ENOMEM;
2109 }
2110
2111 if (sbi->s_flex_groups) {
2112 memcpy(new_groups, sbi->s_flex_groups,
2113 (sbi->s_flex_groups_allocated *
2114 sizeof(struct flex_groups)));
2115 kvfree(sbi->s_flex_groups);
2116 }
2117 sbi->s_flex_groups = new_groups;
2118 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2119 return 0;
2120 }
2121
2122 static int ext4_fill_flex_info(struct super_block *sb)
2123 {
2124 struct ext4_sb_info *sbi = EXT4_SB(sb);
2125 struct ext4_group_desc *gdp = NULL;
2126 ext4_group_t flex_group;
2127 int i, err;
2128
2129 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2130 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2131 sbi->s_log_groups_per_flex = 0;
2132 return 1;
2133 }
2134
2135 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2136 if (err)
2137 goto failed;
2138
2139 for (i = 0; i < sbi->s_groups_count; i++) {
2140 gdp = ext4_get_group_desc(sb, i, NULL);
2141
2142 flex_group = ext4_flex_group(sbi, i);
2143 atomic_add(ext4_free_inodes_count(sb, gdp),
2144 &sbi->s_flex_groups[flex_group].free_inodes);
2145 atomic64_add(ext4_free_group_clusters(sb, gdp),
2146 &sbi->s_flex_groups[flex_group].free_clusters);
2147 atomic_add(ext4_used_dirs_count(sb, gdp),
2148 &sbi->s_flex_groups[flex_group].used_dirs);
2149 }
2150
2151 return 1;
2152 failed:
2153 return 0;
2154 }
2155
2156 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2157 struct ext4_group_desc *gdp)
2158 {
2159 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2160 __u16 crc = 0;
2161 __le32 le_group = cpu_to_le32(block_group);
2162 struct ext4_sb_info *sbi = EXT4_SB(sb);
2163
2164 if (ext4_has_metadata_csum(sbi->s_sb)) {
2165 /* Use new metadata_csum algorithm */
2166 __u32 csum32;
2167 __u16 dummy_csum = 0;
2168
2169 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2170 sizeof(le_group));
2171 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2172 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2173 sizeof(dummy_csum));
2174 offset += sizeof(dummy_csum);
2175 if (offset < sbi->s_desc_size)
2176 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2177 sbi->s_desc_size - offset);
2178
2179 crc = csum32 & 0xFFFF;
2180 goto out;
2181 }
2182
2183 /* old crc16 code */
2184 if (!ext4_has_feature_gdt_csum(sb))
2185 return 0;
2186
2187 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2188 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2189 crc = crc16(crc, (__u8 *)gdp, offset);
2190 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2191 /* for checksum of struct ext4_group_desc do the rest...*/
2192 if (ext4_has_feature_64bit(sb) &&
2193 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2194 crc = crc16(crc, (__u8 *)gdp + offset,
2195 le16_to_cpu(sbi->s_es->s_desc_size) -
2196 offset);
2197
2198 out:
2199 return cpu_to_le16(crc);
2200 }
2201
2202 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2203 struct ext4_group_desc *gdp)
2204 {
2205 if (ext4_has_group_desc_csum(sb) &&
2206 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2207 return 0;
2208
2209 return 1;
2210 }
2211
2212 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2213 struct ext4_group_desc *gdp)
2214 {
2215 if (!ext4_has_group_desc_csum(sb))
2216 return;
2217 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2218 }
2219
2220 /* Called at mount-time, super-block is locked */
2221 static int ext4_check_descriptors(struct super_block *sb,
2222 ext4_fsblk_t sb_block,
2223 ext4_group_t *first_not_zeroed)
2224 {
2225 struct ext4_sb_info *sbi = EXT4_SB(sb);
2226 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2227 ext4_fsblk_t last_block;
2228 ext4_fsblk_t block_bitmap;
2229 ext4_fsblk_t inode_bitmap;
2230 ext4_fsblk_t inode_table;
2231 int flexbg_flag = 0;
2232 ext4_group_t i, grp = sbi->s_groups_count;
2233
2234 if (ext4_has_feature_flex_bg(sb))
2235 flexbg_flag = 1;
2236
2237 ext4_debug("Checking group descriptors");
2238
2239 for (i = 0; i < sbi->s_groups_count; i++) {
2240 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2241
2242 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2243 last_block = ext4_blocks_count(sbi->s_es) - 1;
2244 else
2245 last_block = first_block +
2246 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2247
2248 if ((grp == sbi->s_groups_count) &&
2249 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2250 grp = i;
2251
2252 block_bitmap = ext4_block_bitmap(sb, gdp);
2253 if (block_bitmap == sb_block) {
2254 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2255 "Block bitmap for group %u overlaps "
2256 "superblock", i);
2257 }
2258 if (block_bitmap < first_block || block_bitmap > last_block) {
2259 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2260 "Block bitmap for group %u not in group "
2261 "(block %llu)!", i, block_bitmap);
2262 return 0;
2263 }
2264 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2265 if (inode_bitmap == sb_block) {
2266 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2267 "Inode bitmap for group %u overlaps "
2268 "superblock", i);
2269 }
2270 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2271 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2272 "Inode bitmap for group %u not in group "
2273 "(block %llu)!", i, inode_bitmap);
2274 return 0;
2275 }
2276 inode_table = ext4_inode_table(sb, gdp);
2277 if (inode_table == sb_block) {
2278 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2279 "Inode table for group %u overlaps "
2280 "superblock", i);
2281 }
2282 if (inode_table < first_block ||
2283 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2284 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2285 "Inode table for group %u not in group "
2286 "(block %llu)!", i, inode_table);
2287 return 0;
2288 }
2289 ext4_lock_group(sb, i);
2290 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2291 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2292 "Checksum for group %u failed (%u!=%u)",
2293 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2294 gdp)), le16_to_cpu(gdp->bg_checksum));
2295 if (!(sb->s_flags & MS_RDONLY)) {
2296 ext4_unlock_group(sb, i);
2297 return 0;
2298 }
2299 }
2300 ext4_unlock_group(sb, i);
2301 if (!flexbg_flag)
2302 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2303 }
2304 if (NULL != first_not_zeroed)
2305 *first_not_zeroed = grp;
2306 return 1;
2307 }
2308
2309 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2310 * the superblock) which were deleted from all directories, but held open by
2311 * a process at the time of a crash. We walk the list and try to delete these
2312 * inodes at recovery time (only with a read-write filesystem).
2313 *
2314 * In order to keep the orphan inode chain consistent during traversal (in
2315 * case of crash during recovery), we link each inode into the superblock
2316 * orphan list_head and handle it the same way as an inode deletion during
2317 * normal operation (which journals the operations for us).
2318 *
2319 * We only do an iget() and an iput() on each inode, which is very safe if we
2320 * accidentally point at an in-use or already deleted inode. The worst that
2321 * can happen in this case is that we get a "bit already cleared" message from
2322 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2323 * e2fsck was run on this filesystem, and it must have already done the orphan
2324 * inode cleanup for us, so we can safely abort without any further action.
2325 */
2326 static void ext4_orphan_cleanup(struct super_block *sb,
2327 struct ext4_super_block *es)
2328 {
2329 unsigned int s_flags = sb->s_flags;
2330 int nr_orphans = 0, nr_truncates = 0;
2331 #ifdef CONFIG_QUOTA
2332 int i;
2333 #endif
2334 if (!es->s_last_orphan) {
2335 jbd_debug(4, "no orphan inodes to clean up\n");
2336 return;
2337 }
2338
2339 if (bdev_read_only(sb->s_bdev)) {
2340 ext4_msg(sb, KERN_ERR, "write access "
2341 "unavailable, skipping orphan cleanup");
2342 return;
2343 }
2344
2345 /* Check if feature set would not allow a r/w mount */
2346 if (!ext4_feature_set_ok(sb, 0)) {
2347 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2348 "unknown ROCOMPAT features");
2349 return;
2350 }
2351
2352 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2353 /* don't clear list on RO mount w/ errors */
2354 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2355 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2356 "clearing orphan list.\n");
2357 es->s_last_orphan = 0;
2358 }
2359 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2360 return;
2361 }
2362
2363 if (s_flags & MS_RDONLY) {
2364 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2365 sb->s_flags &= ~MS_RDONLY;
2366 }
2367 #ifdef CONFIG_QUOTA
2368 /* Needed for iput() to work correctly and not trash data */
2369 sb->s_flags |= MS_ACTIVE;
2370 /* Turn on quotas so that they are updated correctly */
2371 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2372 if (EXT4_SB(sb)->s_qf_names[i]) {
2373 int ret = ext4_quota_on_mount(sb, i);
2374 if (ret < 0)
2375 ext4_msg(sb, KERN_ERR,
2376 "Cannot turn on journaled "
2377 "quota: error %d", ret);
2378 }
2379 }
2380 #endif
2381
2382 while (es->s_last_orphan) {
2383 struct inode *inode;
2384
2385 /*
2386 * We may have encountered an error during cleanup; if
2387 * so, skip the rest.
2388 */
2389 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2390 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2391 es->s_last_orphan = 0;
2392 break;
2393 }
2394
2395 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2396 if (IS_ERR(inode)) {
2397 es->s_last_orphan = 0;
2398 break;
2399 }
2400
2401 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2402 dquot_initialize(inode);
2403 if (inode->i_nlink) {
2404 if (test_opt(sb, DEBUG))
2405 ext4_msg(sb, KERN_DEBUG,
2406 "%s: truncating inode %lu to %lld bytes",
2407 __func__, inode->i_ino, inode->i_size);
2408 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2409 inode->i_ino, inode->i_size);
2410 inode_lock(inode);
2411 truncate_inode_pages(inode->i_mapping, inode->i_size);
2412 ext4_truncate(inode);
2413 inode_unlock(inode);
2414 nr_truncates++;
2415 } else {
2416 if (test_opt(sb, DEBUG))
2417 ext4_msg(sb, KERN_DEBUG,
2418 "%s: deleting unreferenced inode %lu",
2419 __func__, inode->i_ino);
2420 jbd_debug(2, "deleting unreferenced inode %lu\n",
2421 inode->i_ino);
2422 nr_orphans++;
2423 }
2424 iput(inode); /* The delete magic happens here! */
2425 }
2426
2427 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2428
2429 if (nr_orphans)
2430 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2431 PLURAL(nr_orphans));
2432 if (nr_truncates)
2433 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2434 PLURAL(nr_truncates));
2435 #ifdef CONFIG_QUOTA
2436 /* Turn quotas off */
2437 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2438 if (sb_dqopt(sb)->files[i])
2439 dquot_quota_off(sb, i);
2440 }
2441 #endif
2442 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2443 }
2444
2445 /*
2446 * Maximal extent format file size.
2447 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2448 * extent format containers, within a sector_t, and within i_blocks
2449 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2450 * so that won't be a limiting factor.
2451 *
2452 * However there is other limiting factor. We do store extents in the form
2453 * of starting block and length, hence the resulting length of the extent
2454 * covering maximum file size must fit into on-disk format containers as
2455 * well. Given that length is always by 1 unit bigger than max unit (because
2456 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2457 *
2458 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2459 */
2460 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2461 {
2462 loff_t res;
2463 loff_t upper_limit = MAX_LFS_FILESIZE;
2464
2465 /* small i_blocks in vfs inode? */
2466 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2467 /*
2468 * CONFIG_LBDAF is not enabled implies the inode
2469 * i_block represent total blocks in 512 bytes
2470 * 32 == size of vfs inode i_blocks * 8
2471 */
2472 upper_limit = (1LL << 32) - 1;
2473
2474 /* total blocks in file system block size */
2475 upper_limit >>= (blkbits - 9);
2476 upper_limit <<= blkbits;
2477 }
2478
2479 /*
2480 * 32-bit extent-start container, ee_block. We lower the maxbytes
2481 * by one fs block, so ee_len can cover the extent of maximum file
2482 * size
2483 */
2484 res = (1LL << 32) - 1;
2485 res <<= blkbits;
2486
2487 /* Sanity check against vm- & vfs- imposed limits */
2488 if (res > upper_limit)
2489 res = upper_limit;
2490
2491 return res;
2492 }
2493
2494 /*
2495 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2496 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2497 * We need to be 1 filesystem block less than the 2^48 sector limit.
2498 */
2499 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2500 {
2501 loff_t res = EXT4_NDIR_BLOCKS;
2502 int meta_blocks;
2503 loff_t upper_limit;
2504 /* This is calculated to be the largest file size for a dense, block
2505 * mapped file such that the file's total number of 512-byte sectors,
2506 * including data and all indirect blocks, does not exceed (2^48 - 1).
2507 *
2508 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2509 * number of 512-byte sectors of the file.
2510 */
2511
2512 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2513 /*
2514 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2515 * the inode i_block field represents total file blocks in
2516 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2517 */
2518 upper_limit = (1LL << 32) - 1;
2519
2520 /* total blocks in file system block size */
2521 upper_limit >>= (bits - 9);
2522
2523 } else {
2524 /*
2525 * We use 48 bit ext4_inode i_blocks
2526 * With EXT4_HUGE_FILE_FL set the i_blocks
2527 * represent total number of blocks in
2528 * file system block size
2529 */
2530 upper_limit = (1LL << 48) - 1;
2531
2532 }
2533
2534 /* indirect blocks */
2535 meta_blocks = 1;
2536 /* double indirect blocks */
2537 meta_blocks += 1 + (1LL << (bits-2));
2538 /* tripple indirect blocks */
2539 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2540
2541 upper_limit -= meta_blocks;
2542 upper_limit <<= bits;
2543
2544 res += 1LL << (bits-2);
2545 res += 1LL << (2*(bits-2));
2546 res += 1LL << (3*(bits-2));
2547 res <<= bits;
2548 if (res > upper_limit)
2549 res = upper_limit;
2550
2551 if (res > MAX_LFS_FILESIZE)
2552 res = MAX_LFS_FILESIZE;
2553
2554 return res;
2555 }
2556
2557 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2558 ext4_fsblk_t logical_sb_block, int nr)
2559 {
2560 struct ext4_sb_info *sbi = EXT4_SB(sb);
2561 ext4_group_t bg, first_meta_bg;
2562 int has_super = 0;
2563
2564 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2565
2566 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2567 return logical_sb_block + nr + 1;
2568 bg = sbi->s_desc_per_block * nr;
2569 if (ext4_bg_has_super(sb, bg))
2570 has_super = 1;
2571
2572 /*
2573 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2574 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2575 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2576 * compensate.
2577 */
2578 if (sb->s_blocksize == 1024 && nr == 0 &&
2579 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2580 has_super++;
2581
2582 return (has_super + ext4_group_first_block_no(sb, bg));
2583 }
2584
2585 /**
2586 * ext4_get_stripe_size: Get the stripe size.
2587 * @sbi: In memory super block info
2588 *
2589 * If we have specified it via mount option, then
2590 * use the mount option value. If the value specified at mount time is
2591 * greater than the blocks per group use the super block value.
2592 * If the super block value is greater than blocks per group return 0.
2593 * Allocator needs it be less than blocks per group.
2594 *
2595 */
2596 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2597 {
2598 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2599 unsigned long stripe_width =
2600 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2601 int ret;
2602
2603 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2604 ret = sbi->s_stripe;
2605 else if (stripe_width <= sbi->s_blocks_per_group)
2606 ret = stripe_width;
2607 else if (stride <= sbi->s_blocks_per_group)
2608 ret = stride;
2609 else
2610 ret = 0;
2611
2612 /*
2613 * If the stripe width is 1, this makes no sense and
2614 * we set it to 0 to turn off stripe handling code.
2615 */
2616 if (ret <= 1)
2617 ret = 0;
2618
2619 return ret;
2620 }
2621
2622 /*
2623 * Check whether this filesystem can be mounted based on
2624 * the features present and the RDONLY/RDWR mount requested.
2625 * Returns 1 if this filesystem can be mounted as requested,
2626 * 0 if it cannot be.
2627 */
2628 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2629 {
2630 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2631 ext4_msg(sb, KERN_ERR,
2632 "Couldn't mount because of "
2633 "unsupported optional features (%x)",
2634 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2635 ~EXT4_FEATURE_INCOMPAT_SUPP));
2636 return 0;
2637 }
2638
2639 if (readonly)
2640 return 1;
2641
2642 if (ext4_has_feature_readonly(sb)) {
2643 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2644 sb->s_flags |= MS_RDONLY;
2645 return 1;
2646 }
2647
2648 /* Check that feature set is OK for a read-write mount */
2649 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2650 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2651 "unsupported optional features (%x)",
2652 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2653 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2654 return 0;
2655 }
2656 /*
2657 * Large file size enabled file system can only be mounted
2658 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2659 */
2660 if (ext4_has_feature_huge_file(sb)) {
2661 if (sizeof(blkcnt_t) < sizeof(u64)) {
2662 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2663 "cannot be mounted RDWR without "
2664 "CONFIG_LBDAF");
2665 return 0;
2666 }
2667 }
2668 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2669 ext4_msg(sb, KERN_ERR,
2670 "Can't support bigalloc feature without "
2671 "extents feature\n");
2672 return 0;
2673 }
2674
2675 #ifndef CONFIG_QUOTA
2676 if (ext4_has_feature_quota(sb) && !readonly) {
2677 ext4_msg(sb, KERN_ERR,
2678 "Filesystem with quota feature cannot be mounted RDWR "
2679 "without CONFIG_QUOTA");
2680 return 0;
2681 }
2682 if (ext4_has_feature_project(sb) && !readonly) {
2683 ext4_msg(sb, KERN_ERR,
2684 "Filesystem with project quota feature cannot be mounted RDWR "
2685 "without CONFIG_QUOTA");
2686 return 0;
2687 }
2688 #endif /* CONFIG_QUOTA */
2689 return 1;
2690 }
2691
2692 /*
2693 * This function is called once a day if we have errors logged
2694 * on the file system
2695 */
2696 static void print_daily_error_info(unsigned long arg)
2697 {
2698 struct super_block *sb = (struct super_block *) arg;
2699 struct ext4_sb_info *sbi;
2700 struct ext4_super_block *es;
2701
2702 sbi = EXT4_SB(sb);
2703 es = sbi->s_es;
2704
2705 if (es->s_error_count)
2706 /* fsck newer than v1.41.13 is needed to clean this condition. */
2707 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2708 le32_to_cpu(es->s_error_count));
2709 if (es->s_first_error_time) {
2710 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2711 sb->s_id, le32_to_cpu(es->s_first_error_time),
2712 (int) sizeof(es->s_first_error_func),
2713 es->s_first_error_func,
2714 le32_to_cpu(es->s_first_error_line));
2715 if (es->s_first_error_ino)
2716 printk(": inode %u",
2717 le32_to_cpu(es->s_first_error_ino));
2718 if (es->s_first_error_block)
2719 printk(": block %llu", (unsigned long long)
2720 le64_to_cpu(es->s_first_error_block));
2721 printk("\n");
2722 }
2723 if (es->s_last_error_time) {
2724 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2725 sb->s_id, le32_to_cpu(es->s_last_error_time),
2726 (int) sizeof(es->s_last_error_func),
2727 es->s_last_error_func,
2728 le32_to_cpu(es->s_last_error_line));
2729 if (es->s_last_error_ino)
2730 printk(": inode %u",
2731 le32_to_cpu(es->s_last_error_ino));
2732 if (es->s_last_error_block)
2733 printk(": block %llu", (unsigned long long)
2734 le64_to_cpu(es->s_last_error_block));
2735 printk("\n");
2736 }
2737 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2738 }
2739
2740 /* Find next suitable group and run ext4_init_inode_table */
2741 static int ext4_run_li_request(struct ext4_li_request *elr)
2742 {
2743 struct ext4_group_desc *gdp = NULL;
2744 ext4_group_t group, ngroups;
2745 struct super_block *sb;
2746 unsigned long timeout = 0;
2747 int ret = 0;
2748
2749 sb = elr->lr_super;
2750 ngroups = EXT4_SB(sb)->s_groups_count;
2751
2752 for (group = elr->lr_next_group; group < ngroups; group++) {
2753 gdp = ext4_get_group_desc(sb, group, NULL);
2754 if (!gdp) {
2755 ret = 1;
2756 break;
2757 }
2758
2759 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2760 break;
2761 }
2762
2763 if (group >= ngroups)
2764 ret = 1;
2765
2766 if (!ret) {
2767 timeout = jiffies;
2768 ret = ext4_init_inode_table(sb, group,
2769 elr->lr_timeout ? 0 : 1);
2770 if (elr->lr_timeout == 0) {
2771 timeout = (jiffies - timeout) *
2772 elr->lr_sbi->s_li_wait_mult;
2773 elr->lr_timeout = timeout;
2774 }
2775 elr->lr_next_sched = jiffies + elr->lr_timeout;
2776 elr->lr_next_group = group + 1;
2777 }
2778 return ret;
2779 }
2780
2781 /*
2782 * Remove lr_request from the list_request and free the
2783 * request structure. Should be called with li_list_mtx held
2784 */
2785 static void ext4_remove_li_request(struct ext4_li_request *elr)
2786 {
2787 struct ext4_sb_info *sbi;
2788
2789 if (!elr)
2790 return;
2791
2792 sbi = elr->lr_sbi;
2793
2794 list_del(&elr->lr_request);
2795 sbi->s_li_request = NULL;
2796 kfree(elr);
2797 }
2798
2799 static void ext4_unregister_li_request(struct super_block *sb)
2800 {
2801 mutex_lock(&ext4_li_mtx);
2802 if (!ext4_li_info) {
2803 mutex_unlock(&ext4_li_mtx);
2804 return;
2805 }
2806
2807 mutex_lock(&ext4_li_info->li_list_mtx);
2808 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2809 mutex_unlock(&ext4_li_info->li_list_mtx);
2810 mutex_unlock(&ext4_li_mtx);
2811 }
2812
2813 static struct task_struct *ext4_lazyinit_task;
2814
2815 /*
2816 * This is the function where ext4lazyinit thread lives. It walks
2817 * through the request list searching for next scheduled filesystem.
2818 * When such a fs is found, run the lazy initialization request
2819 * (ext4_rn_li_request) and keep track of the time spend in this
2820 * function. Based on that time we compute next schedule time of
2821 * the request. When walking through the list is complete, compute
2822 * next waking time and put itself into sleep.
2823 */
2824 static int ext4_lazyinit_thread(void *arg)
2825 {
2826 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2827 struct list_head *pos, *n;
2828 struct ext4_li_request *elr;
2829 unsigned long next_wakeup, cur;
2830
2831 BUG_ON(NULL == eli);
2832
2833 cont_thread:
2834 while (true) {
2835 next_wakeup = MAX_JIFFY_OFFSET;
2836
2837 mutex_lock(&eli->li_list_mtx);
2838 if (list_empty(&eli->li_request_list)) {
2839 mutex_unlock(&eli->li_list_mtx);
2840 goto exit_thread;
2841 }
2842 list_for_each_safe(pos, n, &eli->li_request_list) {
2843 int err = 0;
2844 int progress = 0;
2845 elr = list_entry(pos, struct ext4_li_request,
2846 lr_request);
2847
2848 if (time_before(jiffies, elr->lr_next_sched)) {
2849 if (time_before(elr->lr_next_sched, next_wakeup))
2850 next_wakeup = elr->lr_next_sched;
2851 continue;
2852 }
2853 if (down_read_trylock(&elr->lr_super->s_umount)) {
2854 if (sb_start_write_trylock(elr->lr_super)) {
2855 progress = 1;
2856 /*
2857 * We hold sb->s_umount, sb can not
2858 * be removed from the list, it is
2859 * now safe to drop li_list_mtx
2860 */
2861 mutex_unlock(&eli->li_list_mtx);
2862 err = ext4_run_li_request(elr);
2863 sb_end_write(elr->lr_super);
2864 mutex_lock(&eli->li_list_mtx);
2865 n = pos->next;
2866 }
2867 up_read((&elr->lr_super->s_umount));
2868 }
2869 /* error, remove the lazy_init job */
2870 if (err) {
2871 ext4_remove_li_request(elr);
2872 continue;
2873 }
2874 if (!progress) {
2875 elr->lr_next_sched = jiffies +
2876 (prandom_u32()
2877 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2878 if (time_before(elr->lr_next_sched,
2879 next_wakeup))
2880 next_wakeup = elr->lr_next_sched;
2881 }
2882 }
2883 mutex_unlock(&eli->li_list_mtx);
2884
2885 try_to_freeze();
2886
2887 cur = jiffies;
2888 if ((time_after_eq(cur, next_wakeup)) ||
2889 (MAX_JIFFY_OFFSET == next_wakeup)) {
2890 cond_resched();
2891 continue;
2892 }
2893
2894 schedule_timeout_interruptible(next_wakeup - cur);
2895
2896 if (kthread_should_stop()) {
2897 ext4_clear_request_list();
2898 goto exit_thread;
2899 }
2900 }
2901
2902 exit_thread:
2903 /*
2904 * It looks like the request list is empty, but we need
2905 * to check it under the li_list_mtx lock, to prevent any
2906 * additions into it, and of course we should lock ext4_li_mtx
2907 * to atomically free the list and ext4_li_info, because at
2908 * this point another ext4 filesystem could be registering
2909 * new one.
2910 */
2911 mutex_lock(&ext4_li_mtx);
2912 mutex_lock(&eli->li_list_mtx);
2913 if (!list_empty(&eli->li_request_list)) {
2914 mutex_unlock(&eli->li_list_mtx);
2915 mutex_unlock(&ext4_li_mtx);
2916 goto cont_thread;
2917 }
2918 mutex_unlock(&eli->li_list_mtx);
2919 kfree(ext4_li_info);
2920 ext4_li_info = NULL;
2921 mutex_unlock(&ext4_li_mtx);
2922
2923 return 0;
2924 }
2925
2926 static void ext4_clear_request_list(void)
2927 {
2928 struct list_head *pos, *n;
2929 struct ext4_li_request *elr;
2930
2931 mutex_lock(&ext4_li_info->li_list_mtx);
2932 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2933 elr = list_entry(pos, struct ext4_li_request,
2934 lr_request);
2935 ext4_remove_li_request(elr);
2936 }
2937 mutex_unlock(&ext4_li_info->li_list_mtx);
2938 }
2939
2940 static int ext4_run_lazyinit_thread(void)
2941 {
2942 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2943 ext4_li_info, "ext4lazyinit");
2944 if (IS_ERR(ext4_lazyinit_task)) {
2945 int err = PTR_ERR(ext4_lazyinit_task);
2946 ext4_clear_request_list();
2947 kfree(ext4_li_info);
2948 ext4_li_info = NULL;
2949 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2950 "initialization thread\n",
2951 err);
2952 return err;
2953 }
2954 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2955 return 0;
2956 }
2957
2958 /*
2959 * Check whether it make sense to run itable init. thread or not.
2960 * If there is at least one uninitialized inode table, return
2961 * corresponding group number, else the loop goes through all
2962 * groups and return total number of groups.
2963 */
2964 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2965 {
2966 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2967 struct ext4_group_desc *gdp = NULL;
2968
2969 for (group = 0; group < ngroups; group++) {
2970 gdp = ext4_get_group_desc(sb, group, NULL);
2971 if (!gdp)
2972 continue;
2973
2974 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2975 break;
2976 }
2977
2978 return group;
2979 }
2980
2981 static int ext4_li_info_new(void)
2982 {
2983 struct ext4_lazy_init *eli = NULL;
2984
2985 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2986 if (!eli)
2987 return -ENOMEM;
2988
2989 INIT_LIST_HEAD(&eli->li_request_list);
2990 mutex_init(&eli->li_list_mtx);
2991
2992 eli->li_state |= EXT4_LAZYINIT_QUIT;
2993
2994 ext4_li_info = eli;
2995
2996 return 0;
2997 }
2998
2999 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3000 ext4_group_t start)
3001 {
3002 struct ext4_sb_info *sbi = EXT4_SB(sb);
3003 struct ext4_li_request *elr;
3004
3005 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3006 if (!elr)
3007 return NULL;
3008
3009 elr->lr_super = sb;
3010 elr->lr_sbi = sbi;
3011 elr->lr_next_group = start;
3012
3013 /*
3014 * Randomize first schedule time of the request to
3015 * spread the inode table initialization requests
3016 * better.
3017 */
3018 elr->lr_next_sched = jiffies + (prandom_u32() %
3019 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3020 return elr;
3021 }
3022
3023 int ext4_register_li_request(struct super_block *sb,
3024 ext4_group_t first_not_zeroed)
3025 {
3026 struct ext4_sb_info *sbi = EXT4_SB(sb);
3027 struct ext4_li_request *elr = NULL;
3028 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3029 int ret = 0;
3030
3031 mutex_lock(&ext4_li_mtx);
3032 if (sbi->s_li_request != NULL) {
3033 /*
3034 * Reset timeout so it can be computed again, because
3035 * s_li_wait_mult might have changed.
3036 */
3037 sbi->s_li_request->lr_timeout = 0;
3038 goto out;
3039 }
3040
3041 if (first_not_zeroed == ngroups ||
3042 (sb->s_flags & MS_RDONLY) ||
3043 !test_opt(sb, INIT_INODE_TABLE))
3044 goto out;
3045
3046 elr = ext4_li_request_new(sb, first_not_zeroed);
3047 if (!elr) {
3048 ret = -ENOMEM;
3049 goto out;
3050 }
3051
3052 if (NULL == ext4_li_info) {
3053 ret = ext4_li_info_new();
3054 if (ret)
3055 goto out;
3056 }
3057
3058 mutex_lock(&ext4_li_info->li_list_mtx);
3059 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3060 mutex_unlock(&ext4_li_info->li_list_mtx);
3061
3062 sbi->s_li_request = elr;
3063 /*
3064 * set elr to NULL here since it has been inserted to
3065 * the request_list and the removal and free of it is
3066 * handled by ext4_clear_request_list from now on.
3067 */
3068 elr = NULL;
3069
3070 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3071 ret = ext4_run_lazyinit_thread();
3072 if (ret)
3073 goto out;
3074 }
3075 out:
3076 mutex_unlock(&ext4_li_mtx);
3077 if (ret)
3078 kfree(elr);
3079 return ret;
3080 }
3081
3082 /*
3083 * We do not need to lock anything since this is called on
3084 * module unload.
3085 */
3086 static void ext4_destroy_lazyinit_thread(void)
3087 {
3088 /*
3089 * If thread exited earlier
3090 * there's nothing to be done.
3091 */
3092 if (!ext4_li_info || !ext4_lazyinit_task)
3093 return;
3094
3095 kthread_stop(ext4_lazyinit_task);
3096 }
3097
3098 static int set_journal_csum_feature_set(struct super_block *sb)
3099 {
3100 int ret = 1;
3101 int compat, incompat;
3102 struct ext4_sb_info *sbi = EXT4_SB(sb);
3103
3104 if (ext4_has_metadata_csum(sb)) {
3105 /* journal checksum v3 */
3106 compat = 0;
3107 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3108 } else {
3109 /* journal checksum v1 */
3110 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3111 incompat = 0;
3112 }
3113
3114 jbd2_journal_clear_features(sbi->s_journal,
3115 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3116 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3117 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3118 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3119 ret = jbd2_journal_set_features(sbi->s_journal,
3120 compat, 0,
3121 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3122 incompat);
3123 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3124 ret = jbd2_journal_set_features(sbi->s_journal,
3125 compat, 0,
3126 incompat);
3127 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3128 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3129 } else {
3130 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3131 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3132 }
3133
3134 return ret;
3135 }
3136
3137 /*
3138 * Note: calculating the overhead so we can be compatible with
3139 * historical BSD practice is quite difficult in the face of
3140 * clusters/bigalloc. This is because multiple metadata blocks from
3141 * different block group can end up in the same allocation cluster.
3142 * Calculating the exact overhead in the face of clustered allocation
3143 * requires either O(all block bitmaps) in memory or O(number of block
3144 * groups**2) in time. We will still calculate the superblock for
3145 * older file systems --- and if we come across with a bigalloc file
3146 * system with zero in s_overhead_clusters the estimate will be close to
3147 * correct especially for very large cluster sizes --- but for newer
3148 * file systems, it's better to calculate this figure once at mkfs
3149 * time, and store it in the superblock. If the superblock value is
3150 * present (even for non-bigalloc file systems), we will use it.
3151 */
3152 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3153 char *buf)
3154 {
3155 struct ext4_sb_info *sbi = EXT4_SB(sb);
3156 struct ext4_group_desc *gdp;
3157 ext4_fsblk_t first_block, last_block, b;
3158 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3159 int s, j, count = 0;
3160
3161 if (!ext4_has_feature_bigalloc(sb))
3162 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3163 sbi->s_itb_per_group + 2);
3164
3165 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3166 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3167 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3168 for (i = 0; i < ngroups; i++) {
3169 gdp = ext4_get_group_desc(sb, i, NULL);
3170 b = ext4_block_bitmap(sb, gdp);
3171 if (b >= first_block && b <= last_block) {
3172 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3173 count++;
3174 }
3175 b = ext4_inode_bitmap(sb, gdp);
3176 if (b >= first_block && b <= last_block) {
3177 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3178 count++;
3179 }
3180 b = ext4_inode_table(sb, gdp);
3181 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3182 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3183 int c = EXT4_B2C(sbi, b - first_block);
3184 ext4_set_bit(c, buf);
3185 count++;
3186 }
3187 if (i != grp)
3188 continue;
3189 s = 0;
3190 if (ext4_bg_has_super(sb, grp)) {
3191 ext4_set_bit(s++, buf);
3192 count++;
3193 }
3194 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3195 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3196 count++;
3197 }
3198 }
3199 if (!count)
3200 return 0;
3201 return EXT4_CLUSTERS_PER_GROUP(sb) -
3202 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3203 }
3204
3205 /*
3206 * Compute the overhead and stash it in sbi->s_overhead
3207 */
3208 int ext4_calculate_overhead(struct super_block *sb)
3209 {
3210 struct ext4_sb_info *sbi = EXT4_SB(sb);
3211 struct ext4_super_block *es = sbi->s_es;
3212 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3213 ext4_fsblk_t overhead = 0;
3214 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3215
3216 if (!buf)
3217 return -ENOMEM;
3218
3219 /*
3220 * Compute the overhead (FS structures). This is constant
3221 * for a given filesystem unless the number of block groups
3222 * changes so we cache the previous value until it does.
3223 */
3224
3225 /*
3226 * All of the blocks before first_data_block are overhead
3227 */
3228 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3229
3230 /*
3231 * Add the overhead found in each block group
3232 */
3233 for (i = 0; i < ngroups; i++) {
3234 int blks;
3235
3236 blks = count_overhead(sb, i, buf);
3237 overhead += blks;
3238 if (blks)
3239 memset(buf, 0, PAGE_SIZE);
3240 cond_resched();
3241 }
3242 /* Add the internal journal blocks as well */
3243 if (sbi->s_journal && !sbi->journal_bdev)
3244 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3245
3246 sbi->s_overhead = overhead;
3247 smp_wmb();
3248 free_page((unsigned long) buf);
3249 return 0;
3250 }
3251
3252 static void ext4_set_resv_clusters(struct super_block *sb)
3253 {
3254 ext4_fsblk_t resv_clusters;
3255 struct ext4_sb_info *sbi = EXT4_SB(sb);
3256
3257 /*
3258 * There's no need to reserve anything when we aren't using extents.
3259 * The space estimates are exact, there are no unwritten extents,
3260 * hole punching doesn't need new metadata... This is needed especially
3261 * to keep ext2/3 backward compatibility.
3262 */
3263 if (!ext4_has_feature_extents(sb))
3264 return;
3265 /*
3266 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3267 * This should cover the situations where we can not afford to run
3268 * out of space like for example punch hole, or converting
3269 * unwritten extents in delalloc path. In most cases such
3270 * allocation would require 1, or 2 blocks, higher numbers are
3271 * very rare.
3272 */
3273 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3274 sbi->s_cluster_bits);
3275
3276 do_div(resv_clusters, 50);
3277 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3278
3279 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3280 }
3281
3282 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3283 {
3284 char *orig_data = kstrdup(data, GFP_KERNEL);
3285 struct buffer_head *bh;
3286 struct ext4_super_block *es = NULL;
3287 struct ext4_sb_info *sbi;
3288 ext4_fsblk_t block;
3289 ext4_fsblk_t sb_block = get_sb_block(&data);
3290 ext4_fsblk_t logical_sb_block;
3291 unsigned long offset = 0;
3292 unsigned long journal_devnum = 0;
3293 unsigned long def_mount_opts;
3294 struct inode *root;
3295 const char *descr;
3296 int ret = -ENOMEM;
3297 int blocksize, clustersize;
3298 unsigned int db_count;
3299 unsigned int i;
3300 int needs_recovery, has_huge_files, has_bigalloc;
3301 __u64 blocks_count;
3302 int err = 0;
3303 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3304 ext4_group_t first_not_zeroed;
3305
3306 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3307 if (!sbi)
3308 goto out_free_orig;
3309
3310 sbi->s_blockgroup_lock =
3311 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3312 if (!sbi->s_blockgroup_lock) {
3313 kfree(sbi);
3314 goto out_free_orig;
3315 }
3316 sb->s_fs_info = sbi;
3317 sbi->s_sb = sb;
3318 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3319 sbi->s_sb_block = sb_block;
3320 if (sb->s_bdev->bd_part)
3321 sbi->s_sectors_written_start =
3322 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3323
3324 /* Cleanup superblock name */
3325 strreplace(sb->s_id, '/', '!');
3326
3327 /* -EINVAL is default */
3328 ret = -EINVAL;
3329 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3330 if (!blocksize) {
3331 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3332 goto out_fail;
3333 }
3334
3335 /*
3336 * The ext4 superblock will not be buffer aligned for other than 1kB
3337 * block sizes. We need to calculate the offset from buffer start.
3338 */
3339 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3340 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3341 offset = do_div(logical_sb_block, blocksize);
3342 } else {
3343 logical_sb_block = sb_block;
3344 }
3345
3346 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3347 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3348 goto out_fail;
3349 }
3350 /*
3351 * Note: s_es must be initialized as soon as possible because
3352 * some ext4 macro-instructions depend on its value
3353 */
3354 es = (struct ext4_super_block *) (bh->b_data + offset);
3355 sbi->s_es = es;
3356 sb->s_magic = le16_to_cpu(es->s_magic);
3357 if (sb->s_magic != EXT4_SUPER_MAGIC)
3358 goto cantfind_ext4;
3359 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3360
3361 /* Warn if metadata_csum and gdt_csum are both set. */
3362 if (ext4_has_feature_metadata_csum(sb) &&
3363 ext4_has_feature_gdt_csum(sb))
3364 ext4_warning(sb, "metadata_csum and uninit_bg are "
3365 "redundant flags; please run fsck.");
3366
3367 /* Check for a known checksum algorithm */
3368 if (!ext4_verify_csum_type(sb, es)) {
3369 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3370 "unknown checksum algorithm.");
3371 silent = 1;
3372 goto cantfind_ext4;
3373 }
3374
3375 /* Load the checksum driver */
3376 if (ext4_has_feature_metadata_csum(sb)) {
3377 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3378 if (IS_ERR(sbi->s_chksum_driver)) {
3379 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3380 ret = PTR_ERR(sbi->s_chksum_driver);
3381 sbi->s_chksum_driver = NULL;
3382 goto failed_mount;
3383 }
3384 }
3385
3386 /* Check superblock checksum */
3387 if (!ext4_superblock_csum_verify(sb, es)) {
3388 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3389 "invalid superblock checksum. Run e2fsck?");
3390 silent = 1;
3391 ret = -EFSBADCRC;
3392 goto cantfind_ext4;
3393 }
3394
3395 /* Precompute checksum seed for all metadata */
3396 if (ext4_has_feature_csum_seed(sb))
3397 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3398 else if (ext4_has_metadata_csum(sb))
3399 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3400 sizeof(es->s_uuid));
3401
3402 /* Set defaults before we parse the mount options */
3403 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3404 set_opt(sb, INIT_INODE_TABLE);
3405 if (def_mount_opts & EXT4_DEFM_DEBUG)
3406 set_opt(sb, DEBUG);
3407 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3408 set_opt(sb, GRPID);
3409 if (def_mount_opts & EXT4_DEFM_UID16)
3410 set_opt(sb, NO_UID32);
3411 /* xattr user namespace & acls are now defaulted on */
3412 set_opt(sb, XATTR_USER);
3413 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3414 set_opt(sb, POSIX_ACL);
3415 #endif
3416 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3417 if (ext4_has_metadata_csum(sb))
3418 set_opt(sb, JOURNAL_CHECKSUM);
3419
3420 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3421 set_opt(sb, JOURNAL_DATA);
3422 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3423 set_opt(sb, ORDERED_DATA);
3424 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3425 set_opt(sb, WRITEBACK_DATA);
3426
3427 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3428 set_opt(sb, ERRORS_PANIC);
3429 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3430 set_opt(sb, ERRORS_CONT);
3431 else
3432 set_opt(sb, ERRORS_RO);
3433 /* block_validity enabled by default; disable with noblock_validity */
3434 set_opt(sb, BLOCK_VALIDITY);
3435 if (def_mount_opts & EXT4_DEFM_DISCARD)
3436 set_opt(sb, DISCARD);
3437
3438 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3439 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3440 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3441 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3442 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3443
3444 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3445 set_opt(sb, BARRIER);
3446
3447 /*
3448 * enable delayed allocation by default
3449 * Use -o nodelalloc to turn it off
3450 */
3451 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3452 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3453 set_opt(sb, DELALLOC);
3454
3455 /*
3456 * set default s_li_wait_mult for lazyinit, for the case there is
3457 * no mount option specified.
3458 */
3459 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3460
3461 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3462 &journal_devnum, &journal_ioprio, 0)) {
3463 ext4_msg(sb, KERN_WARNING,
3464 "failed to parse options in superblock: %s",
3465 sbi->s_es->s_mount_opts);
3466 }
3467 sbi->s_def_mount_opt = sbi->s_mount_opt;
3468 if (!parse_options((char *) data, sb, &journal_devnum,
3469 &journal_ioprio, 0))
3470 goto failed_mount;
3471
3472 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3473 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3474 "with data=journal disables delayed "
3475 "allocation and O_DIRECT support!\n");
3476 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3477 ext4_msg(sb, KERN_ERR, "can't mount with "
3478 "both data=journal and delalloc");
3479 goto failed_mount;
3480 }
3481 if (test_opt(sb, DIOREAD_NOLOCK)) {
3482 ext4_msg(sb, KERN_ERR, "can't mount with "
3483 "both data=journal and dioread_nolock");
3484 goto failed_mount;
3485 }
3486 if (test_opt(sb, DAX)) {
3487 ext4_msg(sb, KERN_ERR, "can't mount with "
3488 "both data=journal and dax");
3489 goto failed_mount;
3490 }
3491 if (test_opt(sb, DELALLOC))
3492 clear_opt(sb, DELALLOC);
3493 } else {
3494 sb->s_iflags |= SB_I_CGROUPWB;
3495 }
3496
3497 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3498 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3499
3500 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3501 (ext4_has_compat_features(sb) ||
3502 ext4_has_ro_compat_features(sb) ||
3503 ext4_has_incompat_features(sb)))
3504 ext4_msg(sb, KERN_WARNING,
3505 "feature flags set on rev 0 fs, "
3506 "running e2fsck is recommended");
3507
3508 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3509 set_opt2(sb, HURD_COMPAT);
3510 if (ext4_has_feature_64bit(sb)) {
3511 ext4_msg(sb, KERN_ERR,
3512 "The Hurd can't support 64-bit file systems");
3513 goto failed_mount;
3514 }
3515 }
3516
3517 if (IS_EXT2_SB(sb)) {
3518 if (ext2_feature_set_ok(sb))
3519 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3520 "using the ext4 subsystem");
3521 else {
3522 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3523 "to feature incompatibilities");
3524 goto failed_mount;
3525 }
3526 }
3527
3528 if (IS_EXT3_SB(sb)) {
3529 if (ext3_feature_set_ok(sb))
3530 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3531 "using the ext4 subsystem");
3532 else {
3533 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3534 "to feature incompatibilities");
3535 goto failed_mount;
3536 }
3537 }
3538
3539 /*
3540 * Check feature flags regardless of the revision level, since we
3541 * previously didn't change the revision level when setting the flags,
3542 * so there is a chance incompat flags are set on a rev 0 filesystem.
3543 */
3544 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3545 goto failed_mount;
3546
3547 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3548 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3549 blocksize > EXT4_MAX_BLOCK_SIZE) {
3550 ext4_msg(sb, KERN_ERR,
3551 "Unsupported filesystem blocksize %d", blocksize);
3552 goto failed_mount;
3553 }
3554
3555 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3556 ext4_msg(sb, KERN_ERR,
3557 "Number of reserved GDT blocks insanely large: %d",
3558 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3559 goto failed_mount;
3560 }
3561
3562 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3563 err = bdev_dax_supported(sb, blocksize);
3564 if (err)
3565 goto failed_mount;
3566 }
3567
3568 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3569 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3570 es->s_encryption_level);
3571 goto failed_mount;
3572 }
3573
3574 if (sb->s_blocksize != blocksize) {
3575 /* Validate the filesystem blocksize */
3576 if (!sb_set_blocksize(sb, blocksize)) {
3577 ext4_msg(sb, KERN_ERR, "bad block size %d",
3578 blocksize);
3579 goto failed_mount;
3580 }
3581
3582 brelse(bh);
3583 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3584 offset = do_div(logical_sb_block, blocksize);
3585 bh = sb_bread_unmovable(sb, logical_sb_block);
3586 if (!bh) {
3587 ext4_msg(sb, KERN_ERR,
3588 "Can't read superblock on 2nd try");
3589 goto failed_mount;
3590 }
3591 es = (struct ext4_super_block *)(bh->b_data + offset);
3592 sbi->s_es = es;
3593 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3594 ext4_msg(sb, KERN_ERR,
3595 "Magic mismatch, very weird!");
3596 goto failed_mount;
3597 }
3598 }
3599
3600 has_huge_files = ext4_has_feature_huge_file(sb);
3601 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3602 has_huge_files);
3603 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3604
3605 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3606 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3607 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3608 } else {
3609 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3610 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3611 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3612 (!is_power_of_2(sbi->s_inode_size)) ||
3613 (sbi->s_inode_size > blocksize)) {
3614 ext4_msg(sb, KERN_ERR,
3615 "unsupported inode size: %d",
3616 sbi->s_inode_size);
3617 goto failed_mount;
3618 }
3619 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3620 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3621 }
3622
3623 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3624 if (ext4_has_feature_64bit(sb)) {
3625 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3626 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3627 !is_power_of_2(sbi->s_desc_size)) {
3628 ext4_msg(sb, KERN_ERR,
3629 "unsupported descriptor size %lu",
3630 sbi->s_desc_size);
3631 goto failed_mount;
3632 }
3633 } else
3634 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3635
3636 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3637 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3638 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3639 goto cantfind_ext4;
3640
3641 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3642 if (sbi->s_inodes_per_block == 0)
3643 goto cantfind_ext4;
3644 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3645 sbi->s_inodes_per_block;
3646 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3647 sbi->s_sbh = bh;
3648 sbi->s_mount_state = le16_to_cpu(es->s_state);
3649 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3650 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3651
3652 for (i = 0; i < 4; i++)
3653 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3654 sbi->s_def_hash_version = es->s_def_hash_version;
3655 if (ext4_has_feature_dir_index(sb)) {
3656 i = le32_to_cpu(es->s_flags);
3657 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3658 sbi->s_hash_unsigned = 3;
3659 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3660 #ifdef __CHAR_UNSIGNED__
3661 if (!(sb->s_flags & MS_RDONLY))
3662 es->s_flags |=
3663 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3664 sbi->s_hash_unsigned = 3;
3665 #else
3666 if (!(sb->s_flags & MS_RDONLY))
3667 es->s_flags |=
3668 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3669 #endif
3670 }
3671 }
3672
3673 /* Handle clustersize */
3674 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3675 has_bigalloc = ext4_has_feature_bigalloc(sb);
3676 if (has_bigalloc) {
3677 if (clustersize < blocksize) {
3678 ext4_msg(sb, KERN_ERR,
3679 "cluster size (%d) smaller than "
3680 "block size (%d)", clustersize, blocksize);
3681 goto failed_mount;
3682 }
3683 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3684 le32_to_cpu(es->s_log_block_size);
3685 sbi->s_clusters_per_group =
3686 le32_to_cpu(es->s_clusters_per_group);
3687 if (sbi->s_clusters_per_group > blocksize * 8) {
3688 ext4_msg(sb, KERN_ERR,
3689 "#clusters per group too big: %lu",
3690 sbi->s_clusters_per_group);
3691 goto failed_mount;
3692 }
3693 if (sbi->s_blocks_per_group !=
3694 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3695 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3696 "clusters per group (%lu) inconsistent",
3697 sbi->s_blocks_per_group,
3698 sbi->s_clusters_per_group);
3699 goto failed_mount;
3700 }
3701 } else {
3702 if (clustersize != blocksize) {
3703 ext4_warning(sb, "fragment/cluster size (%d) != "
3704 "block size (%d)", clustersize,
3705 blocksize);
3706 clustersize = blocksize;
3707 }
3708 if (sbi->s_blocks_per_group > blocksize * 8) {
3709 ext4_msg(sb, KERN_ERR,
3710 "#blocks per group too big: %lu",
3711 sbi->s_blocks_per_group);
3712 goto failed_mount;
3713 }
3714 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3715 sbi->s_cluster_bits = 0;
3716 }
3717 sbi->s_cluster_ratio = clustersize / blocksize;
3718
3719 if (sbi->s_inodes_per_group > blocksize * 8) {
3720 ext4_msg(sb, KERN_ERR,
3721 "#inodes per group too big: %lu",
3722 sbi->s_inodes_per_group);
3723 goto failed_mount;
3724 }
3725
3726 /* Do we have standard group size of clustersize * 8 blocks ? */
3727 if (sbi->s_blocks_per_group == clustersize << 3)
3728 set_opt2(sb, STD_GROUP_SIZE);
3729
3730 /*
3731 * Test whether we have more sectors than will fit in sector_t,
3732 * and whether the max offset is addressable by the page cache.
3733 */
3734 err = generic_check_addressable(sb->s_blocksize_bits,
3735 ext4_blocks_count(es));
3736 if (err) {
3737 ext4_msg(sb, KERN_ERR, "filesystem"
3738 " too large to mount safely on this system");
3739 if (sizeof(sector_t) < 8)
3740 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3741 goto failed_mount;
3742 }
3743
3744 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3745 goto cantfind_ext4;
3746
3747 /* check blocks count against device size */
3748 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3749 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3750 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3751 "exceeds size of device (%llu blocks)",
3752 ext4_blocks_count(es), blocks_count);
3753 goto failed_mount;
3754 }
3755
3756 /*
3757 * It makes no sense for the first data block to be beyond the end
3758 * of the filesystem.
3759 */
3760 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3761 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3762 "block %u is beyond end of filesystem (%llu)",
3763 le32_to_cpu(es->s_first_data_block),
3764 ext4_blocks_count(es));
3765 goto failed_mount;
3766 }
3767 blocks_count = (ext4_blocks_count(es) -
3768 le32_to_cpu(es->s_first_data_block) +
3769 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3770 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3771 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3772 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3773 "(block count %llu, first data block %u, "
3774 "blocks per group %lu)", sbi->s_groups_count,
3775 ext4_blocks_count(es),
3776 le32_to_cpu(es->s_first_data_block),
3777 EXT4_BLOCKS_PER_GROUP(sb));
3778 goto failed_mount;
3779 }
3780 sbi->s_groups_count = blocks_count;
3781 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3782 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3783 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3784 EXT4_DESC_PER_BLOCK(sb);
3785 sbi->s_group_desc = ext4_kvmalloc(db_count *
3786 sizeof(struct buffer_head *),
3787 GFP_KERNEL);
3788 if (sbi->s_group_desc == NULL) {
3789 ext4_msg(sb, KERN_ERR, "not enough memory");
3790 ret = -ENOMEM;
3791 goto failed_mount;
3792 }
3793
3794 bgl_lock_init(sbi->s_blockgroup_lock);
3795
3796 for (i = 0; i < db_count; i++) {
3797 block = descriptor_loc(sb, logical_sb_block, i);
3798 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3799 if (!sbi->s_group_desc[i]) {
3800 ext4_msg(sb, KERN_ERR,
3801 "can't read group descriptor %d", i);
3802 db_count = i;
3803 goto failed_mount2;
3804 }
3805 }
3806 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3807 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3808 ret = -EFSCORRUPTED;
3809 goto failed_mount2;
3810 }
3811
3812 sbi->s_gdb_count = db_count;
3813 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3814 spin_lock_init(&sbi->s_next_gen_lock);
3815
3816 setup_timer(&sbi->s_err_report, print_daily_error_info,
3817 (unsigned long) sb);
3818
3819 /* Register extent status tree shrinker */
3820 if (ext4_es_register_shrinker(sbi))
3821 goto failed_mount3;
3822
3823 sbi->s_stripe = ext4_get_stripe_size(sbi);
3824 sbi->s_extent_max_zeroout_kb = 32;
3825
3826 /*
3827 * set up enough so that it can read an inode
3828 */
3829 sb->s_op = &ext4_sops;
3830 sb->s_export_op = &ext4_export_ops;
3831 sb->s_xattr = ext4_xattr_handlers;
3832 sb->s_cop = &ext4_cryptops;
3833 #ifdef CONFIG_QUOTA
3834 sb->dq_op = &ext4_quota_operations;
3835 if (ext4_has_feature_quota(sb))
3836 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3837 else
3838 sb->s_qcop = &ext4_qctl_operations;
3839 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3840 #endif
3841 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3842
3843 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3844 mutex_init(&sbi->s_orphan_lock);
3845
3846 sb->s_root = NULL;
3847
3848 needs_recovery = (es->s_last_orphan != 0 ||
3849 ext4_has_feature_journal_needs_recovery(sb));
3850
3851 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3852 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3853 goto failed_mount3a;
3854
3855 /*
3856 * The first inode we look at is the journal inode. Don't try
3857 * root first: it may be modified in the journal!
3858 */
3859 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3860 if (ext4_load_journal(sb, es, journal_devnum))
3861 goto failed_mount3a;
3862 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3863 ext4_has_feature_journal_needs_recovery(sb)) {
3864 ext4_msg(sb, KERN_ERR, "required journal recovery "
3865 "suppressed and not mounted read-only");
3866 goto failed_mount_wq;
3867 } else {
3868 /* Nojournal mode, all journal mount options are illegal */
3869 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3870 ext4_msg(sb, KERN_ERR, "can't mount with "
3871 "journal_checksum, fs mounted w/o journal");
3872 goto failed_mount_wq;
3873 }
3874 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3875 ext4_msg(sb, KERN_ERR, "can't mount with "
3876 "journal_async_commit, fs mounted w/o journal");
3877 goto failed_mount_wq;
3878 }
3879 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3880 ext4_msg(sb, KERN_ERR, "can't mount with "
3881 "commit=%lu, fs mounted w/o journal",
3882 sbi->s_commit_interval / HZ);
3883 goto failed_mount_wq;
3884 }
3885 if (EXT4_MOUNT_DATA_FLAGS &
3886 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3887 ext4_msg(sb, KERN_ERR, "can't mount with "
3888 "data=, fs mounted w/o journal");
3889 goto failed_mount_wq;
3890 }
3891 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3892 clear_opt(sb, JOURNAL_CHECKSUM);
3893 clear_opt(sb, DATA_FLAGS);
3894 sbi->s_journal = NULL;
3895 needs_recovery = 0;
3896 goto no_journal;
3897 }
3898
3899 if (ext4_has_feature_64bit(sb) &&
3900 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3901 JBD2_FEATURE_INCOMPAT_64BIT)) {
3902 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3903 goto failed_mount_wq;
3904 }
3905
3906 if (!set_journal_csum_feature_set(sb)) {
3907 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3908 "feature set");
3909 goto failed_mount_wq;
3910 }
3911
3912 /* We have now updated the journal if required, so we can
3913 * validate the data journaling mode. */
3914 switch (test_opt(sb, DATA_FLAGS)) {
3915 case 0:
3916 /* No mode set, assume a default based on the journal
3917 * capabilities: ORDERED_DATA if the journal can
3918 * cope, else JOURNAL_DATA
3919 */
3920 if (jbd2_journal_check_available_features
3921 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3922 set_opt(sb, ORDERED_DATA);
3923 else
3924 set_opt(sb, JOURNAL_DATA);
3925 break;
3926
3927 case EXT4_MOUNT_ORDERED_DATA:
3928 case EXT4_MOUNT_WRITEBACK_DATA:
3929 if (!jbd2_journal_check_available_features
3930 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3931 ext4_msg(sb, KERN_ERR, "Journal does not support "
3932 "requested data journaling mode");
3933 goto failed_mount_wq;
3934 }
3935 default:
3936 break;
3937 }
3938 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3939
3940 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3941
3942 no_journal:
3943 sbi->s_mb_cache = ext4_xattr_create_cache();
3944 if (!sbi->s_mb_cache) {
3945 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3946 goto failed_mount_wq;
3947 }
3948
3949 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3950 (blocksize != PAGE_SIZE)) {
3951 ext4_msg(sb, KERN_ERR,
3952 "Unsupported blocksize for fs encryption");
3953 goto failed_mount_wq;
3954 }
3955
3956 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3957 !ext4_has_feature_encrypt(sb)) {
3958 ext4_set_feature_encrypt(sb);
3959 ext4_commit_super(sb, 1);
3960 }
3961
3962 /*
3963 * Get the # of file system overhead blocks from the
3964 * superblock if present.
3965 */
3966 if (es->s_overhead_clusters)
3967 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3968 else {
3969 err = ext4_calculate_overhead(sb);
3970 if (err)
3971 goto failed_mount_wq;
3972 }
3973
3974 /*
3975 * The maximum number of concurrent works can be high and
3976 * concurrency isn't really necessary. Limit it to 1.
3977 */
3978 EXT4_SB(sb)->rsv_conversion_wq =
3979 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3980 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3981 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3982 ret = -ENOMEM;
3983 goto failed_mount4;
3984 }
3985
3986 /*
3987 * The jbd2_journal_load will have done any necessary log recovery,
3988 * so we can safely mount the rest of the filesystem now.
3989 */
3990
3991 root = ext4_iget(sb, EXT4_ROOT_INO);
3992 if (IS_ERR(root)) {
3993 ext4_msg(sb, KERN_ERR, "get root inode failed");
3994 ret = PTR_ERR(root);
3995 root = NULL;
3996 goto failed_mount4;
3997 }
3998 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3999 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4000 iput(root);
4001 goto failed_mount4;
4002 }
4003 sb->s_root = d_make_root(root);
4004 if (!sb->s_root) {
4005 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4006 ret = -ENOMEM;
4007 goto failed_mount4;
4008 }
4009
4010 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4011 sb->s_flags |= MS_RDONLY;
4012
4013 /* determine the minimum size of new large inodes, if present */
4014 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4015 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4016 EXT4_GOOD_OLD_INODE_SIZE;
4017 if (ext4_has_feature_extra_isize(sb)) {
4018 if (sbi->s_want_extra_isize <
4019 le16_to_cpu(es->s_want_extra_isize))
4020 sbi->s_want_extra_isize =
4021 le16_to_cpu(es->s_want_extra_isize);
4022 if (sbi->s_want_extra_isize <
4023 le16_to_cpu(es->s_min_extra_isize))
4024 sbi->s_want_extra_isize =
4025 le16_to_cpu(es->s_min_extra_isize);
4026 }
4027 }
4028 /* Check if enough inode space is available */
4029 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4030 sbi->s_inode_size) {
4031 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4032 EXT4_GOOD_OLD_INODE_SIZE;
4033 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4034 "available");
4035 }
4036
4037 ext4_set_resv_clusters(sb);
4038
4039 err = ext4_setup_system_zone(sb);
4040 if (err) {
4041 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4042 "zone (%d)", err);
4043 goto failed_mount4a;
4044 }
4045
4046 ext4_ext_init(sb);
4047 err = ext4_mb_init(sb);
4048 if (err) {
4049 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4050 err);
4051 goto failed_mount5;
4052 }
4053
4054 block = ext4_count_free_clusters(sb);
4055 ext4_free_blocks_count_set(sbi->s_es,
4056 EXT4_C2B(sbi, block));
4057 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4058 GFP_KERNEL);
4059 if (!err) {
4060 unsigned long freei = ext4_count_free_inodes(sb);
4061 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4062 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4063 GFP_KERNEL);
4064 }
4065 if (!err)
4066 err = percpu_counter_init(&sbi->s_dirs_counter,
4067 ext4_count_dirs(sb), GFP_KERNEL);
4068 if (!err)
4069 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4070 GFP_KERNEL);
4071 if (!err)
4072 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4073
4074 if (err) {
4075 ext4_msg(sb, KERN_ERR, "insufficient memory");
4076 goto failed_mount6;
4077 }
4078
4079 if (ext4_has_feature_flex_bg(sb))
4080 if (!ext4_fill_flex_info(sb)) {
4081 ext4_msg(sb, KERN_ERR,
4082 "unable to initialize "
4083 "flex_bg meta info!");
4084 goto failed_mount6;
4085 }
4086
4087 err = ext4_register_li_request(sb, first_not_zeroed);
4088 if (err)
4089 goto failed_mount6;
4090
4091 err = ext4_register_sysfs(sb);
4092 if (err)
4093 goto failed_mount7;
4094
4095 #ifdef CONFIG_QUOTA
4096 /* Enable quota usage during mount. */
4097 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4098 err = ext4_enable_quotas(sb);
4099 if (err)
4100 goto failed_mount8;
4101 }
4102 #endif /* CONFIG_QUOTA */
4103
4104 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4105 ext4_orphan_cleanup(sb, es);
4106 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4107 if (needs_recovery) {
4108 ext4_msg(sb, KERN_INFO, "recovery complete");
4109 ext4_mark_recovery_complete(sb, es);
4110 }
4111 if (EXT4_SB(sb)->s_journal) {
4112 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4113 descr = " journalled data mode";
4114 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4115 descr = " ordered data mode";
4116 else
4117 descr = " writeback data mode";
4118 } else
4119 descr = "out journal";
4120
4121 if (test_opt(sb, DISCARD)) {
4122 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4123 if (!blk_queue_discard(q))
4124 ext4_msg(sb, KERN_WARNING,
4125 "mounting with \"discard\" option, but "
4126 "the device does not support discard");
4127 }
4128
4129 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4130 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4131 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4132 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4133
4134 if (es->s_error_count)
4135 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4136
4137 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4138 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4139 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4140 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4141
4142 kfree(orig_data);
4143 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4144 memcpy(sbi->key_prefix, EXT4_KEY_DESC_PREFIX,
4145 EXT4_KEY_DESC_PREFIX_SIZE);
4146 sbi->key_prefix_size = EXT4_KEY_DESC_PREFIX_SIZE;
4147 #endif
4148 return 0;
4149
4150 cantfind_ext4:
4151 if (!silent)
4152 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4153 goto failed_mount;
4154
4155 #ifdef CONFIG_QUOTA
4156 failed_mount8:
4157 ext4_unregister_sysfs(sb);
4158 #endif
4159 failed_mount7:
4160 ext4_unregister_li_request(sb);
4161 failed_mount6:
4162 ext4_mb_release(sb);
4163 if (sbi->s_flex_groups)
4164 kvfree(sbi->s_flex_groups);
4165 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4166 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4167 percpu_counter_destroy(&sbi->s_dirs_counter);
4168 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4169 failed_mount5:
4170 ext4_ext_release(sb);
4171 ext4_release_system_zone(sb);
4172 failed_mount4a:
4173 dput(sb->s_root);
4174 sb->s_root = NULL;
4175 failed_mount4:
4176 ext4_msg(sb, KERN_ERR, "mount failed");
4177 if (EXT4_SB(sb)->rsv_conversion_wq)
4178 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4179 failed_mount_wq:
4180 if (sbi->s_mb_cache) {
4181 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4182 sbi->s_mb_cache = NULL;
4183 }
4184 if (sbi->s_journal) {
4185 jbd2_journal_destroy(sbi->s_journal);
4186 sbi->s_journal = NULL;
4187 }
4188 failed_mount3a:
4189 ext4_es_unregister_shrinker(sbi);
4190 failed_mount3:
4191 del_timer_sync(&sbi->s_err_report);
4192 if (sbi->s_mmp_tsk)
4193 kthread_stop(sbi->s_mmp_tsk);
4194 failed_mount2:
4195 for (i = 0; i < db_count; i++)
4196 brelse(sbi->s_group_desc[i]);
4197 kvfree(sbi->s_group_desc);
4198 failed_mount:
4199 if (sbi->s_chksum_driver)
4200 crypto_free_shash(sbi->s_chksum_driver);
4201 #ifdef CONFIG_QUOTA
4202 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4203 kfree(sbi->s_qf_names[i]);
4204 #endif
4205 ext4_blkdev_remove(sbi);
4206 brelse(bh);
4207 out_fail:
4208 sb->s_fs_info = NULL;
4209 kfree(sbi->s_blockgroup_lock);
4210 kfree(sbi);
4211 out_free_orig:
4212 kfree(orig_data);
4213 return err ? err : ret;
4214 }
4215
4216 /*
4217 * Setup any per-fs journal parameters now. We'll do this both on
4218 * initial mount, once the journal has been initialised but before we've
4219 * done any recovery; and again on any subsequent remount.
4220 */
4221 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4222 {
4223 struct ext4_sb_info *sbi = EXT4_SB(sb);
4224
4225 journal->j_commit_interval = sbi->s_commit_interval;
4226 journal->j_min_batch_time = sbi->s_min_batch_time;
4227 journal->j_max_batch_time = sbi->s_max_batch_time;
4228
4229 write_lock(&journal->j_state_lock);
4230 if (test_opt(sb, BARRIER))
4231 journal->j_flags |= JBD2_BARRIER;
4232 else
4233 journal->j_flags &= ~JBD2_BARRIER;
4234 if (test_opt(sb, DATA_ERR_ABORT))
4235 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4236 else
4237 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4238 write_unlock(&journal->j_state_lock);
4239 }
4240
4241 static journal_t *ext4_get_journal(struct super_block *sb,
4242 unsigned int journal_inum)
4243 {
4244 struct inode *journal_inode;
4245 journal_t *journal;
4246
4247 BUG_ON(!ext4_has_feature_journal(sb));
4248
4249 /* First, test for the existence of a valid inode on disk. Bad
4250 * things happen if we iget() an unused inode, as the subsequent
4251 * iput() will try to delete it. */
4252
4253 journal_inode = ext4_iget(sb, journal_inum);
4254 if (IS_ERR(journal_inode)) {
4255 ext4_msg(sb, KERN_ERR, "no journal found");
4256 return NULL;
4257 }
4258 if (!journal_inode->i_nlink) {
4259 make_bad_inode(journal_inode);
4260 iput(journal_inode);
4261 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4262 return NULL;
4263 }
4264
4265 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4266 journal_inode, journal_inode->i_size);
4267 if (!S_ISREG(journal_inode->i_mode)) {
4268 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4269 iput(journal_inode);
4270 return NULL;
4271 }
4272
4273 journal = jbd2_journal_init_inode(journal_inode);
4274 if (!journal) {
4275 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4276 iput(journal_inode);
4277 return NULL;
4278 }
4279 journal->j_private = sb;
4280 ext4_init_journal_params(sb, journal);
4281 return journal;
4282 }
4283
4284 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4285 dev_t j_dev)
4286 {
4287 struct buffer_head *bh;
4288 journal_t *journal;
4289 ext4_fsblk_t start;
4290 ext4_fsblk_t len;
4291 int hblock, blocksize;
4292 ext4_fsblk_t sb_block;
4293 unsigned long offset;
4294 struct ext4_super_block *es;
4295 struct block_device *bdev;
4296
4297 BUG_ON(!ext4_has_feature_journal(sb));
4298
4299 bdev = ext4_blkdev_get(j_dev, sb);
4300 if (bdev == NULL)
4301 return NULL;
4302
4303 blocksize = sb->s_blocksize;
4304 hblock = bdev_logical_block_size(bdev);
4305 if (blocksize < hblock) {
4306 ext4_msg(sb, KERN_ERR,
4307 "blocksize too small for journal device");
4308 goto out_bdev;
4309 }
4310
4311 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4312 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4313 set_blocksize(bdev, blocksize);
4314 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4315 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4316 "external journal");
4317 goto out_bdev;
4318 }
4319
4320 es = (struct ext4_super_block *) (bh->b_data + offset);
4321 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4322 !(le32_to_cpu(es->s_feature_incompat) &
4323 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4324 ext4_msg(sb, KERN_ERR, "external journal has "
4325 "bad superblock");
4326 brelse(bh);
4327 goto out_bdev;
4328 }
4329
4330 if ((le32_to_cpu(es->s_feature_ro_compat) &
4331 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4332 es->s_checksum != ext4_superblock_csum(sb, es)) {
4333 ext4_msg(sb, KERN_ERR, "external journal has "
4334 "corrupt superblock");
4335 brelse(bh);
4336 goto out_bdev;
4337 }
4338
4339 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4340 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4341 brelse(bh);
4342 goto out_bdev;
4343 }
4344
4345 len = ext4_blocks_count(es);
4346 start = sb_block + 1;
4347 brelse(bh); /* we're done with the superblock */
4348
4349 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4350 start, len, blocksize);
4351 if (!journal) {
4352 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4353 goto out_bdev;
4354 }
4355 journal->j_private = sb;
4356 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4357 wait_on_buffer(journal->j_sb_buffer);
4358 if (!buffer_uptodate(journal->j_sb_buffer)) {
4359 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4360 goto out_journal;
4361 }
4362 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4363 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4364 "user (unsupported) - %d",
4365 be32_to_cpu(journal->j_superblock->s_nr_users));
4366 goto out_journal;
4367 }
4368 EXT4_SB(sb)->journal_bdev = bdev;
4369 ext4_init_journal_params(sb, journal);
4370 return journal;
4371
4372 out_journal:
4373 jbd2_journal_destroy(journal);
4374 out_bdev:
4375 ext4_blkdev_put(bdev);
4376 return NULL;
4377 }
4378
4379 static int ext4_load_journal(struct super_block *sb,
4380 struct ext4_super_block *es,
4381 unsigned long journal_devnum)
4382 {
4383 journal_t *journal;
4384 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4385 dev_t journal_dev;
4386 int err = 0;
4387 int really_read_only;
4388
4389 BUG_ON(!ext4_has_feature_journal(sb));
4390
4391 if (journal_devnum &&
4392 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4393 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4394 "numbers have changed");
4395 journal_dev = new_decode_dev(journal_devnum);
4396 } else
4397 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4398
4399 really_read_only = bdev_read_only(sb->s_bdev);
4400
4401 /*
4402 * Are we loading a blank journal or performing recovery after a
4403 * crash? For recovery, we need to check in advance whether we
4404 * can get read-write access to the device.
4405 */
4406 if (ext4_has_feature_journal_needs_recovery(sb)) {
4407 if (sb->s_flags & MS_RDONLY) {
4408 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4409 "required on readonly filesystem");
4410 if (really_read_only) {
4411 ext4_msg(sb, KERN_ERR, "write access "
4412 "unavailable, cannot proceed");
4413 return -EROFS;
4414 }
4415 ext4_msg(sb, KERN_INFO, "write access will "
4416 "be enabled during recovery");
4417 }
4418 }
4419
4420 if (journal_inum && journal_dev) {
4421 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4422 "and inode journals!");
4423 return -EINVAL;
4424 }
4425
4426 if (journal_inum) {
4427 if (!(journal = ext4_get_journal(sb, journal_inum)))
4428 return -EINVAL;
4429 } else {
4430 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4431 return -EINVAL;
4432 }
4433
4434 if (!(journal->j_flags & JBD2_BARRIER))
4435 ext4_msg(sb, KERN_INFO, "barriers disabled");
4436
4437 if (!ext4_has_feature_journal_needs_recovery(sb))
4438 err = jbd2_journal_wipe(journal, !really_read_only);
4439 if (!err) {
4440 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4441 if (save)
4442 memcpy(save, ((char *) es) +
4443 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4444 err = jbd2_journal_load(journal);
4445 if (save)
4446 memcpy(((char *) es) + EXT4_S_ERR_START,
4447 save, EXT4_S_ERR_LEN);
4448 kfree(save);
4449 }
4450
4451 if (err) {
4452 ext4_msg(sb, KERN_ERR, "error loading journal");
4453 jbd2_journal_destroy(journal);
4454 return err;
4455 }
4456
4457 EXT4_SB(sb)->s_journal = journal;
4458 ext4_clear_journal_err(sb, es);
4459
4460 if (!really_read_only && journal_devnum &&
4461 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4462 es->s_journal_dev = cpu_to_le32(journal_devnum);
4463
4464 /* Make sure we flush the recovery flag to disk. */
4465 ext4_commit_super(sb, 1);
4466 }
4467
4468 return 0;
4469 }
4470
4471 static int ext4_commit_super(struct super_block *sb, int sync)
4472 {
4473 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4474 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4475 int error = 0;
4476
4477 if (!sbh || block_device_ejected(sb))
4478 return error;
4479 /*
4480 * If the file system is mounted read-only, don't update the
4481 * superblock write time. This avoids updating the superblock
4482 * write time when we are mounting the root file system
4483 * read/only but we need to replay the journal; at that point,
4484 * for people who are east of GMT and who make their clock
4485 * tick in localtime for Windows bug-for-bug compatibility,
4486 * the clock is set in the future, and this will cause e2fsck
4487 * to complain and force a full file system check.
4488 */
4489 if (!(sb->s_flags & MS_RDONLY))
4490 es->s_wtime = cpu_to_le32(get_seconds());
4491 if (sb->s_bdev->bd_part)
4492 es->s_kbytes_written =
4493 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4494 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4495 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4496 else
4497 es->s_kbytes_written =
4498 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4499 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4500 ext4_free_blocks_count_set(es,
4501 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4502 &EXT4_SB(sb)->s_freeclusters_counter)));
4503 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4504 es->s_free_inodes_count =
4505 cpu_to_le32(percpu_counter_sum_positive(
4506 &EXT4_SB(sb)->s_freeinodes_counter));
4507 BUFFER_TRACE(sbh, "marking dirty");
4508 ext4_superblock_csum_set(sb);
4509 lock_buffer(sbh);
4510 if (buffer_write_io_error(sbh)) {
4511 /*
4512 * Oh, dear. A previous attempt to write the
4513 * superblock failed. This could happen because the
4514 * USB device was yanked out. Or it could happen to
4515 * be a transient write error and maybe the block will
4516 * be remapped. Nothing we can do but to retry the
4517 * write and hope for the best.
4518 */
4519 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4520 "superblock detected");
4521 clear_buffer_write_io_error(sbh);
4522 set_buffer_uptodate(sbh);
4523 }
4524 mark_buffer_dirty(sbh);
4525 unlock_buffer(sbh);
4526 if (sync) {
4527 error = __sync_dirty_buffer(sbh,
4528 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4529 if (error)
4530 return error;
4531
4532 error = buffer_write_io_error(sbh);
4533 if (error) {
4534 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4535 "superblock");
4536 clear_buffer_write_io_error(sbh);
4537 set_buffer_uptodate(sbh);
4538 }
4539 }
4540 return error;
4541 }
4542
4543 /*
4544 * Have we just finished recovery? If so, and if we are mounting (or
4545 * remounting) the filesystem readonly, then we will end up with a
4546 * consistent fs on disk. Record that fact.
4547 */
4548 static void ext4_mark_recovery_complete(struct super_block *sb,
4549 struct ext4_super_block *es)
4550 {
4551 journal_t *journal = EXT4_SB(sb)->s_journal;
4552
4553 if (!ext4_has_feature_journal(sb)) {
4554 BUG_ON(journal != NULL);
4555 return;
4556 }
4557 jbd2_journal_lock_updates(journal);
4558 if (jbd2_journal_flush(journal) < 0)
4559 goto out;
4560
4561 if (ext4_has_feature_journal_needs_recovery(sb) &&
4562 sb->s_flags & MS_RDONLY) {
4563 ext4_clear_feature_journal_needs_recovery(sb);
4564 ext4_commit_super(sb, 1);
4565 }
4566
4567 out:
4568 jbd2_journal_unlock_updates(journal);
4569 }
4570
4571 /*
4572 * If we are mounting (or read-write remounting) a filesystem whose journal
4573 * has recorded an error from a previous lifetime, move that error to the
4574 * main filesystem now.
4575 */
4576 static void ext4_clear_journal_err(struct super_block *sb,
4577 struct ext4_super_block *es)
4578 {
4579 journal_t *journal;
4580 int j_errno;
4581 const char *errstr;
4582
4583 BUG_ON(!ext4_has_feature_journal(sb));
4584
4585 journal = EXT4_SB(sb)->s_journal;
4586
4587 /*
4588 * Now check for any error status which may have been recorded in the
4589 * journal by a prior ext4_error() or ext4_abort()
4590 */
4591
4592 j_errno = jbd2_journal_errno(journal);
4593 if (j_errno) {
4594 char nbuf[16];
4595
4596 errstr = ext4_decode_error(sb, j_errno, nbuf);
4597 ext4_warning(sb, "Filesystem error recorded "
4598 "from previous mount: %s", errstr);
4599 ext4_warning(sb, "Marking fs in need of filesystem check.");
4600
4601 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4602 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4603 ext4_commit_super(sb, 1);
4604
4605 jbd2_journal_clear_err(journal);
4606 jbd2_journal_update_sb_errno(journal);
4607 }
4608 }
4609
4610 /*
4611 * Force the running and committing transactions to commit,
4612 * and wait on the commit.
4613 */
4614 int ext4_force_commit(struct super_block *sb)
4615 {
4616 journal_t *journal;
4617
4618 if (sb->s_flags & MS_RDONLY)
4619 return 0;
4620
4621 journal = EXT4_SB(sb)->s_journal;
4622 return ext4_journal_force_commit(journal);
4623 }
4624
4625 static int ext4_sync_fs(struct super_block *sb, int wait)
4626 {
4627 int ret = 0;
4628 tid_t target;
4629 bool needs_barrier = false;
4630 struct ext4_sb_info *sbi = EXT4_SB(sb);
4631
4632 trace_ext4_sync_fs(sb, wait);
4633 flush_workqueue(sbi->rsv_conversion_wq);
4634 /*
4635 * Writeback quota in non-journalled quota case - journalled quota has
4636 * no dirty dquots
4637 */
4638 dquot_writeback_dquots(sb, -1);
4639 /*
4640 * Data writeback is possible w/o journal transaction, so barrier must
4641 * being sent at the end of the function. But we can skip it if
4642 * transaction_commit will do it for us.
4643 */
4644 if (sbi->s_journal) {
4645 target = jbd2_get_latest_transaction(sbi->s_journal);
4646 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4647 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4648 needs_barrier = true;
4649
4650 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4651 if (wait)
4652 ret = jbd2_log_wait_commit(sbi->s_journal,
4653 target);
4654 }
4655 } else if (wait && test_opt(sb, BARRIER))
4656 needs_barrier = true;
4657 if (needs_barrier) {
4658 int err;
4659 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4660 if (!ret)
4661 ret = err;
4662 }
4663
4664 return ret;
4665 }
4666
4667 /*
4668 * LVM calls this function before a (read-only) snapshot is created. This
4669 * gives us a chance to flush the journal completely and mark the fs clean.
4670 *
4671 * Note that only this function cannot bring a filesystem to be in a clean
4672 * state independently. It relies on upper layer to stop all data & metadata
4673 * modifications.
4674 */
4675 static int ext4_freeze(struct super_block *sb)
4676 {
4677 int error = 0;
4678 journal_t *journal;
4679
4680 if (sb->s_flags & MS_RDONLY)
4681 return 0;
4682
4683 journal = EXT4_SB(sb)->s_journal;
4684
4685 if (journal) {
4686 /* Now we set up the journal barrier. */
4687 jbd2_journal_lock_updates(journal);
4688
4689 /*
4690 * Don't clear the needs_recovery flag if we failed to
4691 * flush the journal.
4692 */
4693 error = jbd2_journal_flush(journal);
4694 if (error < 0)
4695 goto out;
4696
4697 /* Journal blocked and flushed, clear needs_recovery flag. */
4698 ext4_clear_feature_journal_needs_recovery(sb);
4699 }
4700
4701 error = ext4_commit_super(sb, 1);
4702 out:
4703 if (journal)
4704 /* we rely on upper layer to stop further updates */
4705 jbd2_journal_unlock_updates(journal);
4706 return error;
4707 }
4708
4709 /*
4710 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4711 * flag here, even though the filesystem is not technically dirty yet.
4712 */
4713 static int ext4_unfreeze(struct super_block *sb)
4714 {
4715 if (sb->s_flags & MS_RDONLY)
4716 return 0;
4717
4718 if (EXT4_SB(sb)->s_journal) {
4719 /* Reset the needs_recovery flag before the fs is unlocked. */
4720 ext4_set_feature_journal_needs_recovery(sb);
4721 }
4722
4723 ext4_commit_super(sb, 1);
4724 return 0;
4725 }
4726
4727 /*
4728 * Structure to save mount options for ext4_remount's benefit
4729 */
4730 struct ext4_mount_options {
4731 unsigned long s_mount_opt;
4732 unsigned long s_mount_opt2;
4733 kuid_t s_resuid;
4734 kgid_t s_resgid;
4735 unsigned long s_commit_interval;
4736 u32 s_min_batch_time, s_max_batch_time;
4737 #ifdef CONFIG_QUOTA
4738 int s_jquota_fmt;
4739 char *s_qf_names[EXT4_MAXQUOTAS];
4740 #endif
4741 };
4742
4743 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4744 {
4745 struct ext4_super_block *es;
4746 struct ext4_sb_info *sbi = EXT4_SB(sb);
4747 unsigned long old_sb_flags;
4748 struct ext4_mount_options old_opts;
4749 int enable_quota = 0;
4750 ext4_group_t g;
4751 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4752 int err = 0;
4753 #ifdef CONFIG_QUOTA
4754 int i, j;
4755 #endif
4756 char *orig_data = kstrdup(data, GFP_KERNEL);
4757
4758 /* Store the original options */
4759 old_sb_flags = sb->s_flags;
4760 old_opts.s_mount_opt = sbi->s_mount_opt;
4761 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4762 old_opts.s_resuid = sbi->s_resuid;
4763 old_opts.s_resgid = sbi->s_resgid;
4764 old_opts.s_commit_interval = sbi->s_commit_interval;
4765 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4766 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4767 #ifdef CONFIG_QUOTA
4768 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4769 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4770 if (sbi->s_qf_names[i]) {
4771 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4772 GFP_KERNEL);
4773 if (!old_opts.s_qf_names[i]) {
4774 for (j = 0; j < i; j++)
4775 kfree(old_opts.s_qf_names[j]);
4776 kfree(orig_data);
4777 return -ENOMEM;
4778 }
4779 } else
4780 old_opts.s_qf_names[i] = NULL;
4781 #endif
4782 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4783 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4784
4785 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4786 err = -EINVAL;
4787 goto restore_opts;
4788 }
4789
4790 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4791 test_opt(sb, JOURNAL_CHECKSUM)) {
4792 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4793 "during remount not supported; ignoring");
4794 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4795 }
4796
4797 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4798 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4799 ext4_msg(sb, KERN_ERR, "can't mount with "
4800 "both data=journal and delalloc");
4801 err = -EINVAL;
4802 goto restore_opts;
4803 }
4804 if (test_opt(sb, DIOREAD_NOLOCK)) {
4805 ext4_msg(sb, KERN_ERR, "can't mount with "
4806 "both data=journal and dioread_nolock");
4807 err = -EINVAL;
4808 goto restore_opts;
4809 }
4810 if (test_opt(sb, DAX)) {
4811 ext4_msg(sb, KERN_ERR, "can't mount with "
4812 "both data=journal and dax");
4813 err = -EINVAL;
4814 goto restore_opts;
4815 }
4816 }
4817
4818 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4819 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4820 "dax flag with busy inodes while remounting");
4821 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4822 }
4823
4824 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4825 ext4_abort(sb, "Abort forced by user");
4826
4827 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4828 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4829
4830 es = sbi->s_es;
4831
4832 if (sbi->s_journal) {
4833 ext4_init_journal_params(sb, sbi->s_journal);
4834 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4835 }
4836
4837 if (*flags & MS_LAZYTIME)
4838 sb->s_flags |= MS_LAZYTIME;
4839
4840 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4841 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4842 err = -EROFS;
4843 goto restore_opts;
4844 }
4845
4846 if (*flags & MS_RDONLY) {
4847 err = sync_filesystem(sb);
4848 if (err < 0)
4849 goto restore_opts;
4850 err = dquot_suspend(sb, -1);
4851 if (err < 0)
4852 goto restore_opts;
4853
4854 /*
4855 * First of all, the unconditional stuff we have to do
4856 * to disable replay of the journal when we next remount
4857 */
4858 sb->s_flags |= MS_RDONLY;
4859
4860 /*
4861 * OK, test if we are remounting a valid rw partition
4862 * readonly, and if so set the rdonly flag and then
4863 * mark the partition as valid again.
4864 */
4865 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4866 (sbi->s_mount_state & EXT4_VALID_FS))
4867 es->s_state = cpu_to_le16(sbi->s_mount_state);
4868
4869 if (sbi->s_journal)
4870 ext4_mark_recovery_complete(sb, es);
4871 } else {
4872 /* Make sure we can mount this feature set readwrite */
4873 if (ext4_has_feature_readonly(sb) ||
4874 !ext4_feature_set_ok(sb, 0)) {
4875 err = -EROFS;
4876 goto restore_opts;
4877 }
4878 /*
4879 * Make sure the group descriptor checksums
4880 * are sane. If they aren't, refuse to remount r/w.
4881 */
4882 for (g = 0; g < sbi->s_groups_count; g++) {
4883 struct ext4_group_desc *gdp =
4884 ext4_get_group_desc(sb, g, NULL);
4885
4886 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4887 ext4_msg(sb, KERN_ERR,
4888 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4889 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4890 le16_to_cpu(gdp->bg_checksum));
4891 err = -EFSBADCRC;
4892 goto restore_opts;
4893 }
4894 }
4895
4896 /*
4897 * If we have an unprocessed orphan list hanging
4898 * around from a previously readonly bdev mount,
4899 * require a full umount/remount for now.
4900 */
4901 if (es->s_last_orphan) {
4902 ext4_msg(sb, KERN_WARNING, "Couldn't "
4903 "remount RDWR because of unprocessed "
4904 "orphan inode list. Please "
4905 "umount/remount instead");
4906 err = -EINVAL;
4907 goto restore_opts;
4908 }
4909
4910 /*
4911 * Mounting a RDONLY partition read-write, so reread
4912 * and store the current valid flag. (It may have
4913 * been changed by e2fsck since we originally mounted
4914 * the partition.)
4915 */
4916 if (sbi->s_journal)
4917 ext4_clear_journal_err(sb, es);
4918 sbi->s_mount_state = le16_to_cpu(es->s_state);
4919 if (!ext4_setup_super(sb, es, 0))
4920 sb->s_flags &= ~MS_RDONLY;
4921 if (ext4_has_feature_mmp(sb))
4922 if (ext4_multi_mount_protect(sb,
4923 le64_to_cpu(es->s_mmp_block))) {
4924 err = -EROFS;
4925 goto restore_opts;
4926 }
4927 enable_quota = 1;
4928 }
4929 }
4930
4931 /*
4932 * Reinitialize lazy itable initialization thread based on
4933 * current settings
4934 */
4935 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4936 ext4_unregister_li_request(sb);
4937 else {
4938 ext4_group_t first_not_zeroed;
4939 first_not_zeroed = ext4_has_uninit_itable(sb);
4940 ext4_register_li_request(sb, first_not_zeroed);
4941 }
4942
4943 ext4_setup_system_zone(sb);
4944 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4945 ext4_commit_super(sb, 1);
4946
4947 #ifdef CONFIG_QUOTA
4948 /* Release old quota file names */
4949 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4950 kfree(old_opts.s_qf_names[i]);
4951 if (enable_quota) {
4952 if (sb_any_quota_suspended(sb))
4953 dquot_resume(sb, -1);
4954 else if (ext4_has_feature_quota(sb)) {
4955 err = ext4_enable_quotas(sb);
4956 if (err)
4957 goto restore_opts;
4958 }
4959 }
4960 #endif
4961
4962 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4963 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4964 kfree(orig_data);
4965 return 0;
4966
4967 restore_opts:
4968 sb->s_flags = old_sb_flags;
4969 sbi->s_mount_opt = old_opts.s_mount_opt;
4970 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4971 sbi->s_resuid = old_opts.s_resuid;
4972 sbi->s_resgid = old_opts.s_resgid;
4973 sbi->s_commit_interval = old_opts.s_commit_interval;
4974 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4975 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4976 #ifdef CONFIG_QUOTA
4977 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4978 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
4979 kfree(sbi->s_qf_names[i]);
4980 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4981 }
4982 #endif
4983 kfree(orig_data);
4984 return err;
4985 }
4986
4987 #ifdef CONFIG_QUOTA
4988 static int ext4_statfs_project(struct super_block *sb,
4989 kprojid_t projid, struct kstatfs *buf)
4990 {
4991 struct kqid qid;
4992 struct dquot *dquot;
4993 u64 limit;
4994 u64 curblock;
4995
4996 qid = make_kqid_projid(projid);
4997 dquot = dqget(sb, qid);
4998 if (IS_ERR(dquot))
4999 return PTR_ERR(dquot);
5000 spin_lock(&dq_data_lock);
5001
5002 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5003 dquot->dq_dqb.dqb_bsoftlimit :
5004 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5005 if (limit && buf->f_blocks > limit) {
5006 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5007 buf->f_blocks = limit;
5008 buf->f_bfree = buf->f_bavail =
5009 (buf->f_blocks > curblock) ?
5010 (buf->f_blocks - curblock) : 0;
5011 }
5012
5013 limit = dquot->dq_dqb.dqb_isoftlimit ?
5014 dquot->dq_dqb.dqb_isoftlimit :
5015 dquot->dq_dqb.dqb_ihardlimit;
5016 if (limit && buf->f_files > limit) {
5017 buf->f_files = limit;
5018 buf->f_ffree =
5019 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5020 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5021 }
5022
5023 spin_unlock(&dq_data_lock);
5024 dqput(dquot);
5025 return 0;
5026 }
5027 #endif
5028
5029 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5030 {
5031 struct super_block *sb = dentry->d_sb;
5032 struct ext4_sb_info *sbi = EXT4_SB(sb);
5033 struct ext4_super_block *es = sbi->s_es;
5034 ext4_fsblk_t overhead = 0, resv_blocks;
5035 u64 fsid;
5036 s64 bfree;
5037 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5038
5039 if (!test_opt(sb, MINIX_DF))
5040 overhead = sbi->s_overhead;
5041
5042 buf->f_type = EXT4_SUPER_MAGIC;
5043 buf->f_bsize = sb->s_blocksize;
5044 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5045 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5046 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5047 /* prevent underflow in case that few free space is available */
5048 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5049 buf->f_bavail = buf->f_bfree -
5050 (ext4_r_blocks_count(es) + resv_blocks);
5051 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5052 buf->f_bavail = 0;
5053 buf->f_files = le32_to_cpu(es->s_inodes_count);
5054 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5055 buf->f_namelen = EXT4_NAME_LEN;
5056 fsid = le64_to_cpup((void *)es->s_uuid) ^
5057 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5058 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5059 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5060
5061 #ifdef CONFIG_QUOTA
5062 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5063 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5064 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5065 #endif
5066 return 0;
5067 }
5068
5069 /* Helper function for writing quotas on sync - we need to start transaction
5070 * before quota file is locked for write. Otherwise the are possible deadlocks:
5071 * Process 1 Process 2
5072 * ext4_create() quota_sync()
5073 * jbd2_journal_start() write_dquot()
5074 * dquot_initialize() down(dqio_mutex)
5075 * down(dqio_mutex) jbd2_journal_start()
5076 *
5077 */
5078
5079 #ifdef CONFIG_QUOTA
5080
5081 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5082 {
5083 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5084 }
5085
5086 static int ext4_write_dquot(struct dquot *dquot)
5087 {
5088 int ret, err;
5089 handle_t *handle;
5090 struct inode *inode;
5091
5092 inode = dquot_to_inode(dquot);
5093 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5094 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5095 if (IS_ERR(handle))
5096 return PTR_ERR(handle);
5097 ret = dquot_commit(dquot);
5098 err = ext4_journal_stop(handle);
5099 if (!ret)
5100 ret = err;
5101 return ret;
5102 }
5103
5104 static int ext4_acquire_dquot(struct dquot *dquot)
5105 {
5106 int ret, err;
5107 handle_t *handle;
5108
5109 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5110 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5111 if (IS_ERR(handle))
5112 return PTR_ERR(handle);
5113 ret = dquot_acquire(dquot);
5114 err = ext4_journal_stop(handle);
5115 if (!ret)
5116 ret = err;
5117 return ret;
5118 }
5119
5120 static int ext4_release_dquot(struct dquot *dquot)
5121 {
5122 int ret, err;
5123 handle_t *handle;
5124
5125 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5126 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5127 if (IS_ERR(handle)) {
5128 /* Release dquot anyway to avoid endless cycle in dqput() */
5129 dquot_release(dquot);
5130 return PTR_ERR(handle);
5131 }
5132 ret = dquot_release(dquot);
5133 err = ext4_journal_stop(handle);
5134 if (!ret)
5135 ret = err;
5136 return ret;
5137 }
5138
5139 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5140 {
5141 struct super_block *sb = dquot->dq_sb;
5142 struct ext4_sb_info *sbi = EXT4_SB(sb);
5143
5144 /* Are we journaling quotas? */
5145 if (ext4_has_feature_quota(sb) ||
5146 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5147 dquot_mark_dquot_dirty(dquot);
5148 return ext4_write_dquot(dquot);
5149 } else {
5150 return dquot_mark_dquot_dirty(dquot);
5151 }
5152 }
5153
5154 static int ext4_write_info(struct super_block *sb, int type)
5155 {
5156 int ret, err;
5157 handle_t *handle;
5158
5159 /* Data block + inode block */
5160 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5161 if (IS_ERR(handle))
5162 return PTR_ERR(handle);
5163 ret = dquot_commit_info(sb, type);
5164 err = ext4_journal_stop(handle);
5165 if (!ret)
5166 ret = err;
5167 return ret;
5168 }
5169
5170 /*
5171 * Turn on quotas during mount time - we need to find
5172 * the quota file and such...
5173 */
5174 static int ext4_quota_on_mount(struct super_block *sb, int type)
5175 {
5176 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5177 EXT4_SB(sb)->s_jquota_fmt, type);
5178 }
5179
5180 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5181 {
5182 struct ext4_inode_info *ei = EXT4_I(inode);
5183
5184 /* The first argument of lockdep_set_subclass has to be
5185 * *exactly* the same as the argument to init_rwsem() --- in
5186 * this case, in init_once() --- or lockdep gets unhappy
5187 * because the name of the lock is set using the
5188 * stringification of the argument to init_rwsem().
5189 */
5190 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5191 lockdep_set_subclass(&ei->i_data_sem, subclass);
5192 }
5193
5194 /*
5195 * Standard function to be called on quota_on
5196 */
5197 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5198 struct path *path)
5199 {
5200 int err;
5201
5202 if (!test_opt(sb, QUOTA))
5203 return -EINVAL;
5204
5205 /* Quotafile not on the same filesystem? */
5206 if (path->dentry->d_sb != sb)
5207 return -EXDEV;
5208 /* Journaling quota? */
5209 if (EXT4_SB(sb)->s_qf_names[type]) {
5210 /* Quotafile not in fs root? */
5211 if (path->dentry->d_parent != sb->s_root)
5212 ext4_msg(sb, KERN_WARNING,
5213 "Quota file not on filesystem root. "
5214 "Journaled quota will not work");
5215 }
5216
5217 /*
5218 * When we journal data on quota file, we have to flush journal to see
5219 * all updates to the file when we bypass pagecache...
5220 */
5221 if (EXT4_SB(sb)->s_journal &&
5222 ext4_should_journal_data(d_inode(path->dentry))) {
5223 /*
5224 * We don't need to lock updates but journal_flush() could
5225 * otherwise be livelocked...
5226 */
5227 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5228 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5229 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5230 if (err)
5231 return err;
5232 }
5233 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5234 err = dquot_quota_on(sb, type, format_id, path);
5235 if (err)
5236 lockdep_set_quota_inode(path->dentry->d_inode,
5237 I_DATA_SEM_NORMAL);
5238 return err;
5239 }
5240
5241 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5242 unsigned int flags)
5243 {
5244 int err;
5245 struct inode *qf_inode;
5246 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5247 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5248 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5249 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5250 };
5251
5252 BUG_ON(!ext4_has_feature_quota(sb));
5253
5254 if (!qf_inums[type])
5255 return -EPERM;
5256
5257 qf_inode = ext4_iget(sb, qf_inums[type]);
5258 if (IS_ERR(qf_inode)) {
5259 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5260 return PTR_ERR(qf_inode);
5261 }
5262
5263 /* Don't account quota for quota files to avoid recursion */
5264 qf_inode->i_flags |= S_NOQUOTA;
5265 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5266 err = dquot_enable(qf_inode, type, format_id, flags);
5267 iput(qf_inode);
5268 if (err)
5269 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5270
5271 return err;
5272 }
5273
5274 /* Enable usage tracking for all quota types. */
5275 static int ext4_enable_quotas(struct super_block *sb)
5276 {
5277 int type, err = 0;
5278 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5279 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5280 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5281 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5282 };
5283 bool quota_mopt[EXT4_MAXQUOTAS] = {
5284 test_opt(sb, USRQUOTA),
5285 test_opt(sb, GRPQUOTA),
5286 test_opt(sb, PRJQUOTA),
5287 };
5288
5289 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5290 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5291 if (qf_inums[type]) {
5292 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5293 DQUOT_USAGE_ENABLED |
5294 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5295 if (err) {
5296 ext4_warning(sb,
5297 "Failed to enable quota tracking "
5298 "(type=%d, err=%d). Please run "
5299 "e2fsck to fix.", type, err);
5300 return err;
5301 }
5302 }
5303 }
5304 return 0;
5305 }
5306
5307 static int ext4_quota_off(struct super_block *sb, int type)
5308 {
5309 struct inode *inode = sb_dqopt(sb)->files[type];
5310 handle_t *handle;
5311
5312 /* Force all delayed allocation blocks to be allocated.
5313 * Caller already holds s_umount sem */
5314 if (test_opt(sb, DELALLOC))
5315 sync_filesystem(sb);
5316
5317 if (!inode)
5318 goto out;
5319
5320 /* Update modification times of quota files when userspace can
5321 * start looking at them */
5322 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5323 if (IS_ERR(handle))
5324 goto out;
5325 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5326 ext4_mark_inode_dirty(handle, inode);
5327 ext4_journal_stop(handle);
5328
5329 out:
5330 return dquot_quota_off(sb, type);
5331 }
5332
5333 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5334 * acquiring the locks... As quota files are never truncated and quota code
5335 * itself serializes the operations (and no one else should touch the files)
5336 * we don't have to be afraid of races */
5337 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5338 size_t len, loff_t off)
5339 {
5340 struct inode *inode = sb_dqopt(sb)->files[type];
5341 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5342 int offset = off & (sb->s_blocksize - 1);
5343 int tocopy;
5344 size_t toread;
5345 struct buffer_head *bh;
5346 loff_t i_size = i_size_read(inode);
5347
5348 if (off > i_size)
5349 return 0;
5350 if (off+len > i_size)
5351 len = i_size-off;
5352 toread = len;
5353 while (toread > 0) {
5354 tocopy = sb->s_blocksize - offset < toread ?
5355 sb->s_blocksize - offset : toread;
5356 bh = ext4_bread(NULL, inode, blk, 0);
5357 if (IS_ERR(bh))
5358 return PTR_ERR(bh);
5359 if (!bh) /* A hole? */
5360 memset(data, 0, tocopy);
5361 else
5362 memcpy(data, bh->b_data+offset, tocopy);
5363 brelse(bh);
5364 offset = 0;
5365 toread -= tocopy;
5366 data += tocopy;
5367 blk++;
5368 }
5369 return len;
5370 }
5371
5372 /* Write to quotafile (we know the transaction is already started and has
5373 * enough credits) */
5374 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5375 const char *data, size_t len, loff_t off)
5376 {
5377 struct inode *inode = sb_dqopt(sb)->files[type];
5378 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5379 int err, offset = off & (sb->s_blocksize - 1);
5380 int retries = 0;
5381 struct buffer_head *bh;
5382 handle_t *handle = journal_current_handle();
5383
5384 if (EXT4_SB(sb)->s_journal && !handle) {
5385 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5386 " cancelled because transaction is not started",
5387 (unsigned long long)off, (unsigned long long)len);
5388 return -EIO;
5389 }
5390 /*
5391 * Since we account only one data block in transaction credits,
5392 * then it is impossible to cross a block boundary.
5393 */
5394 if (sb->s_blocksize - offset < len) {
5395 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5396 " cancelled because not block aligned",
5397 (unsigned long long)off, (unsigned long long)len);
5398 return -EIO;
5399 }
5400
5401 do {
5402 bh = ext4_bread(handle, inode, blk,
5403 EXT4_GET_BLOCKS_CREATE |
5404 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5405 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5406 ext4_should_retry_alloc(inode->i_sb, &retries));
5407 if (IS_ERR(bh))
5408 return PTR_ERR(bh);
5409 if (!bh)
5410 goto out;
5411 BUFFER_TRACE(bh, "get write access");
5412 err = ext4_journal_get_write_access(handle, bh);
5413 if (err) {
5414 brelse(bh);
5415 return err;
5416 }
5417 lock_buffer(bh);
5418 memcpy(bh->b_data+offset, data, len);
5419 flush_dcache_page(bh->b_page);
5420 unlock_buffer(bh);
5421 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5422 brelse(bh);
5423 out:
5424 if (inode->i_size < off + len) {
5425 i_size_write(inode, off + len);
5426 EXT4_I(inode)->i_disksize = inode->i_size;
5427 ext4_mark_inode_dirty(handle, inode);
5428 }
5429 return len;
5430 }
5431
5432 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5433 {
5434 const struct quota_format_ops *ops;
5435
5436 if (!sb_has_quota_loaded(sb, qid->type))
5437 return -ESRCH;
5438 ops = sb_dqopt(sb)->ops[qid->type];
5439 if (!ops || !ops->get_next_id)
5440 return -ENOSYS;
5441 return dquot_get_next_id(sb, qid);
5442 }
5443 #endif
5444
5445 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5446 const char *dev_name, void *data)
5447 {
5448 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5449 }
5450
5451 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5452 static inline void register_as_ext2(void)
5453 {
5454 int err = register_filesystem(&ext2_fs_type);
5455 if (err)
5456 printk(KERN_WARNING
5457 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5458 }
5459
5460 static inline void unregister_as_ext2(void)
5461 {
5462 unregister_filesystem(&ext2_fs_type);
5463 }
5464
5465 static inline int ext2_feature_set_ok(struct super_block *sb)
5466 {
5467 if (ext4_has_unknown_ext2_incompat_features(sb))
5468 return 0;
5469 if (sb->s_flags & MS_RDONLY)
5470 return 1;
5471 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5472 return 0;
5473 return 1;
5474 }
5475 #else
5476 static inline void register_as_ext2(void) { }
5477 static inline void unregister_as_ext2(void) { }
5478 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5479 #endif
5480
5481 static inline void register_as_ext3(void)
5482 {
5483 int err = register_filesystem(&ext3_fs_type);
5484 if (err)
5485 printk(KERN_WARNING
5486 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5487 }
5488
5489 static inline void unregister_as_ext3(void)
5490 {
5491 unregister_filesystem(&ext3_fs_type);
5492 }
5493
5494 static inline int ext3_feature_set_ok(struct super_block *sb)
5495 {
5496 if (ext4_has_unknown_ext3_incompat_features(sb))
5497 return 0;
5498 if (!ext4_has_feature_journal(sb))
5499 return 0;
5500 if (sb->s_flags & MS_RDONLY)
5501 return 1;
5502 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5503 return 0;
5504 return 1;
5505 }
5506
5507 static struct file_system_type ext4_fs_type = {
5508 .owner = THIS_MODULE,
5509 .name = "ext4",
5510 .mount = ext4_mount,
5511 .kill_sb = kill_block_super,
5512 .fs_flags = FS_REQUIRES_DEV,
5513 };
5514 MODULE_ALIAS_FS("ext4");
5515
5516 /* Shared across all ext4 file systems */
5517 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5518
5519 static int __init ext4_init_fs(void)
5520 {
5521 int i, err;
5522
5523 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5524 ext4_li_info = NULL;
5525 mutex_init(&ext4_li_mtx);
5526
5527 /* Build-time check for flags consistency */
5528 ext4_check_flag_values();
5529
5530 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5531 init_waitqueue_head(&ext4__ioend_wq[i]);
5532
5533 err = ext4_init_es();
5534 if (err)
5535 return err;
5536
5537 err = ext4_init_pageio();
5538 if (err)
5539 goto out5;
5540
5541 err = ext4_init_system_zone();
5542 if (err)
5543 goto out4;
5544
5545 err = ext4_init_sysfs();
5546 if (err)
5547 goto out3;
5548
5549 err = ext4_init_mballoc();
5550 if (err)
5551 goto out2;
5552 err = init_inodecache();
5553 if (err)
5554 goto out1;
5555 register_as_ext3();
5556 register_as_ext2();
5557 err = register_filesystem(&ext4_fs_type);
5558 if (err)
5559 goto out;
5560
5561 return 0;
5562 out:
5563 unregister_as_ext2();
5564 unregister_as_ext3();
5565 destroy_inodecache();
5566 out1:
5567 ext4_exit_mballoc();
5568 out2:
5569 ext4_exit_sysfs();
5570 out3:
5571 ext4_exit_system_zone();
5572 out4:
5573 ext4_exit_pageio();
5574 out5:
5575 ext4_exit_es();
5576
5577 return err;
5578 }
5579
5580 static void __exit ext4_exit_fs(void)
5581 {
5582 ext4_destroy_lazyinit_thread();
5583 unregister_as_ext2();
5584 unregister_as_ext3();
5585 unregister_filesystem(&ext4_fs_type);
5586 destroy_inodecache();
5587 ext4_exit_mballoc();
5588 ext4_exit_sysfs();
5589 ext4_exit_system_zone();
5590 ext4_exit_pageio();
5591 ext4_exit_es();
5592 }
5593
5594 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5595 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5596 MODULE_LICENSE("GPL");
5597 module_init(ext4_init_fs)
5598 module_exit(ext4_exit_fs)
Linux kernel - Deliverables
This page took 0.176781 seconds and 5 git commands to generate.