?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
mmc: sdhci-msm: Fix error return code in sdhci_msm_probe()
[linux/fpc-iii.git] / fs / ext4 / super.c
blob3ec8708989ca016c3dd9e7e2d9bd479a1cd9beaf
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_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_barrier, "barrier=%u"},
1331 {Opt_barrier, "barrier"},
1332 {Opt_nobarrier, "nobarrier"},
1333 {Opt_i_version, "i_version"},
1334 {Opt_dax, "dax"},
1335 {Opt_stripe, "stripe=%u"},
1336 {Opt_delalloc, "delalloc"},
1337 {Opt_lazytime, "lazytime"},
1338 {Opt_nolazytime, "nolazytime"},
1339 {Opt_nodelalloc, "nodelalloc"},
1340 {Opt_removed, "mblk_io_submit"},
1341 {Opt_removed, "nomblk_io_submit"},
1342 {Opt_block_validity, "block_validity"},
1343 {Opt_noblock_validity, "noblock_validity"},
1344 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1345 {Opt_journal_ioprio, "journal_ioprio=%u"},
1346 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1347 {Opt_auto_da_alloc, "auto_da_alloc"},
1348 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1349 {Opt_dioread_nolock, "dioread_nolock"},
1350 {Opt_dioread_lock, "dioread_lock"},
1351 {Opt_discard, "discard"},
1352 {Opt_nodiscard, "nodiscard"},
1353 {Opt_init_itable, "init_itable=%u"},
1354 {Opt_init_itable, "init_itable"},
1355 {Opt_noinit_itable, "noinit_itable"},
1356 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1357 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1358 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1359 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1360 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1361 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1362 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1363 {Opt_err, NULL},
1364 };
1365
1366 static ext4_fsblk_t get_sb_block(void **data)
1367 {
1368 ext4_fsblk_t sb_block;
1369 char *options = (char *) *data;
1370
1371 if (!options || strncmp(options, "sb=", 3) != 0)
1372 return 1; /* Default location */
1373
1374 options += 3;
1375 /* TODO: use simple_strtoll with >32bit ext4 */
1376 sb_block = simple_strtoul(options, &options, 0);
1377 if (*options && *options != ',') {
1378 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1379 (char *) *data);
1380 return 1;
1381 }
1382 if (*options == ',')
1383 options++;
1384 *data = (void *) options;
1385
1386 return sb_block;
1387 }
1388
1389 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1390 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1391 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1392
1393 #ifdef CONFIG_QUOTA
1394 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1395 {
1396 struct ext4_sb_info *sbi = EXT4_SB(sb);
1397 char *qname;
1398 int ret = -1;
1399
1400 if (sb_any_quota_loaded(sb) &&
1401 !sbi->s_qf_names[qtype]) {
1402 ext4_msg(sb, KERN_ERR,
1403 "Cannot change journaled "
1404 "quota options when quota turned on");
1405 return -1;
1406 }
1407 if (ext4_has_feature_quota(sb)) {
1408 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1409 "ignored when QUOTA feature is enabled");
1410 return 1;
1411 }
1412 qname = match_strdup(args);
1413 if (!qname) {
1414 ext4_msg(sb, KERN_ERR,
1415 "Not enough memory for storing quotafile name");
1416 return -1;
1417 }
1418 if (sbi->s_qf_names[qtype]) {
1419 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1420 ret = 1;
1421 else
1422 ext4_msg(sb, KERN_ERR,
1423 "%s quota file already specified",
1424 QTYPE2NAME(qtype));
1425 goto errout;
1426 }
1427 if (strchr(qname, '/')) {
1428 ext4_msg(sb, KERN_ERR,
1429 "quotafile must be on filesystem root");
1430 goto errout;
1431 }
1432 sbi->s_qf_names[qtype] = qname;
1433 set_opt(sb, QUOTA);
1434 return 1;
1435 errout:
1436 kfree(qname);
1437 return ret;
1438 }
1439
1440 static int clear_qf_name(struct super_block *sb, int qtype)
1441 {
1442
1443 struct ext4_sb_info *sbi = EXT4_SB(sb);
1444
1445 if (sb_any_quota_loaded(sb) &&
1446 sbi->s_qf_names[qtype]) {
1447 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1448 " when quota turned on");
1449 return -1;
1450 }
1451 kfree(sbi->s_qf_names[qtype]);
1452 sbi->s_qf_names[qtype] = NULL;
1453 return 1;
1454 }
1455 #endif
1456
1457 #define MOPT_SET 0x0001
1458 #define MOPT_CLEAR 0x0002
1459 #define MOPT_NOSUPPORT 0x0004
1460 #define MOPT_EXPLICIT 0x0008
1461 #define MOPT_CLEAR_ERR 0x0010
1462 #define MOPT_GTE0 0x0020
1463 #ifdef CONFIG_QUOTA
1464 #define MOPT_Q 0
1465 #define MOPT_QFMT 0x0040
1466 #else
1467 #define MOPT_Q MOPT_NOSUPPORT
1468 #define MOPT_QFMT MOPT_NOSUPPORT
1469 #endif
1470 #define MOPT_DATAJ 0x0080
1471 #define MOPT_NO_EXT2 0x0100
1472 #define MOPT_NO_EXT3 0x0200
1473 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1474 #define MOPT_STRING 0x0400
1475
1476 static const struct mount_opts {
1477 int token;
1478 int mount_opt;
1479 int flags;
1480 } ext4_mount_opts[] = {
1481 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1482 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1483 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1484 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1485 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1486 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1487 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1488 MOPT_EXT4_ONLY | MOPT_SET},
1489 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1490 MOPT_EXT4_ONLY | MOPT_CLEAR},
1491 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1492 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1493 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1494 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1495 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1496 MOPT_EXT4_ONLY | MOPT_CLEAR},
1497 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1498 MOPT_EXT4_ONLY | MOPT_CLEAR},
1499 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1500 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1501 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1502 EXT4_MOUNT_JOURNAL_CHECKSUM),
1503 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1504 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1505 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1506 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1507 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1508 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1509 MOPT_NO_EXT2},
1510 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1511 MOPT_NO_EXT2},
1512 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1513 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1514 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1515 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1516 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1517 {Opt_commit, 0, MOPT_GTE0},
1518 {Opt_max_batch_time, 0, MOPT_GTE0},
1519 {Opt_min_batch_time, 0, MOPT_GTE0},
1520 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1521 {Opt_init_itable, 0, MOPT_GTE0},
1522 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1523 {Opt_stripe, 0, MOPT_GTE0},
1524 {Opt_resuid, 0, MOPT_GTE0},
1525 {Opt_resgid, 0, MOPT_GTE0},
1526 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1527 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1528 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1529 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1530 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1531 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1532 MOPT_NO_EXT2 | MOPT_DATAJ},
1533 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1534 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1535 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1536 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1537 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1538 #else
1539 {Opt_acl, 0, MOPT_NOSUPPORT},
1540 {Opt_noacl, 0, MOPT_NOSUPPORT},
1541 #endif
1542 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1543 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1544 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1545 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1546 MOPT_SET | MOPT_Q},
1547 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1548 MOPT_SET | MOPT_Q},
1549 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1550 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1551 {Opt_usrjquota, 0, MOPT_Q},
1552 {Opt_grpjquota, 0, MOPT_Q},
1553 {Opt_offusrjquota, 0, MOPT_Q},
1554 {Opt_offgrpjquota, 0, MOPT_Q},
1555 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1556 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1557 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1558 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1559 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1560 {Opt_err, 0, 0}
1561 };
1562
1563 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1564 substring_t *args, unsigned long *journal_devnum,
1565 unsigned int *journal_ioprio, int is_remount)
1566 {
1567 struct ext4_sb_info *sbi = EXT4_SB(sb);
1568 const struct mount_opts *m;
1569 kuid_t uid;
1570 kgid_t gid;
1571 int arg = 0;
1572
1573 #ifdef CONFIG_QUOTA
1574 if (token == Opt_usrjquota)
1575 return set_qf_name(sb, USRQUOTA, &args[0]);
1576 else if (token == Opt_grpjquota)
1577 return set_qf_name(sb, GRPQUOTA, &args[0]);
1578 else if (token == Opt_offusrjquota)
1579 return clear_qf_name(sb, USRQUOTA);
1580 else if (token == Opt_offgrpjquota)
1581 return clear_qf_name(sb, GRPQUOTA);
1582 #endif
1583 switch (token) {
1584 case Opt_noacl:
1585 case Opt_nouser_xattr:
1586 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1587 break;
1588 case Opt_sb:
1589 return 1; /* handled by get_sb_block() */
1590 case Opt_removed:
1591 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1592 return 1;
1593 case Opt_abort:
1594 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1595 return 1;
1596 case Opt_i_version:
1597 sb->s_flags |= MS_I_VERSION;
1598 return 1;
1599 case Opt_lazytime:
1600 sb->s_flags |= MS_LAZYTIME;
1601 return 1;
1602 case Opt_nolazytime:
1603 sb->s_flags &= ~MS_LAZYTIME;
1604 return 1;
1605 }
1606
1607 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1608 if (token == m->token)
1609 break;
1610
1611 if (m->token == Opt_err) {
1612 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1613 "or missing value", opt);
1614 return -1;
1615 }
1616
1617 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1618 ext4_msg(sb, KERN_ERR,
1619 "Mount option \"%s\" incompatible with ext2", opt);
1620 return -1;
1621 }
1622 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1623 ext4_msg(sb, KERN_ERR,
1624 "Mount option \"%s\" incompatible with ext3", opt);
1625 return -1;
1626 }
1627
1628 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1629 return -1;
1630 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1631 return -1;
1632 if (m->flags & MOPT_EXPLICIT) {
1633 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1634 set_opt2(sb, EXPLICIT_DELALLOC);
1635 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1636 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1637 } else
1638 return -1;
1639 }
1640 if (m->flags & MOPT_CLEAR_ERR)
1641 clear_opt(sb, ERRORS_MASK);
1642 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1643 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1644 "options when quota turned on");
1645 return -1;
1646 }
1647
1648 if (m->flags & MOPT_NOSUPPORT) {
1649 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1650 } else if (token == Opt_commit) {
1651 if (arg == 0)
1652 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1653 sbi->s_commit_interval = HZ * arg;
1654 } else if (token == Opt_max_batch_time) {
1655 sbi->s_max_batch_time = arg;
1656 } else if (token == Opt_min_batch_time) {
1657 sbi->s_min_batch_time = arg;
1658 } else if (token == Opt_inode_readahead_blks) {
1659 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1660 ext4_msg(sb, KERN_ERR,
1661 "EXT4-fs: inode_readahead_blks must be "
1662 "0 or a power of 2 smaller than 2^31");
1663 return -1;
1664 }
1665 sbi->s_inode_readahead_blks = arg;
1666 } else if (token == Opt_init_itable) {
1667 set_opt(sb, INIT_INODE_TABLE);
1668 if (!args->from)
1669 arg = EXT4_DEF_LI_WAIT_MULT;
1670 sbi->s_li_wait_mult = arg;
1671 } else if (token == Opt_max_dir_size_kb) {
1672 sbi->s_max_dir_size_kb = arg;
1673 } else if (token == Opt_stripe) {
1674 sbi->s_stripe = arg;
1675 } else if (token == Opt_resuid) {
1676 uid = make_kuid(current_user_ns(), arg);
1677 if (!uid_valid(uid)) {
1678 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1679 return -1;
1680 }
1681 sbi->s_resuid = uid;
1682 } else if (token == Opt_resgid) {
1683 gid = make_kgid(current_user_ns(), arg);
1684 if (!gid_valid(gid)) {
1685 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1686 return -1;
1687 }
1688 sbi->s_resgid = gid;
1689 } else if (token == Opt_journal_dev) {
1690 if (is_remount) {
1691 ext4_msg(sb, KERN_ERR,
1692 "Cannot specify journal on remount");
1693 return -1;
1694 }
1695 *journal_devnum = arg;
1696 } else if (token == Opt_journal_path) {
1697 char *journal_path;
1698 struct inode *journal_inode;
1699 struct path path;
1700 int error;
1701
1702 if (is_remount) {
1703 ext4_msg(sb, KERN_ERR,
1704 "Cannot specify journal on remount");
1705 return -1;
1706 }
1707 journal_path = match_strdup(&args[0]);
1708 if (!journal_path) {
1709 ext4_msg(sb, KERN_ERR, "error: could not dup "
1710 "journal device string");
1711 return -1;
1712 }
1713
1714 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1715 if (error) {
1716 ext4_msg(sb, KERN_ERR, "error: could not find "
1717 "journal device path: error %d", error);
1718 kfree(journal_path);
1719 return -1;
1720 }
1721
1722 journal_inode = d_inode(path.dentry);
1723 if (!S_ISBLK(journal_inode->i_mode)) {
1724 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1725 "is not a block device", journal_path);
1726 path_put(&path);
1727 kfree(journal_path);
1728 return -1;
1729 }
1730
1731 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1732 path_put(&path);
1733 kfree(journal_path);
1734 } else if (token == Opt_journal_ioprio) {
1735 if (arg > 7) {
1736 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1737 " (must be 0-7)");
1738 return -1;
1739 }
1740 *journal_ioprio =
1741 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1742 } else if (token == Opt_test_dummy_encryption) {
1743 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1744 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1745 ext4_msg(sb, KERN_WARNING,
1746 "Test dummy encryption mode enabled");
1747 #else
1748 ext4_msg(sb, KERN_WARNING,
1749 "Test dummy encryption mount option ignored");
1750 #endif
1751 } else if (m->flags & MOPT_DATAJ) {
1752 if (is_remount) {
1753 if (!sbi->s_journal)
1754 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1755 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1756 ext4_msg(sb, KERN_ERR,
1757 "Cannot change data mode on remount");
1758 return -1;
1759 }
1760 } else {
1761 clear_opt(sb, DATA_FLAGS);
1762 sbi->s_mount_opt |= m->mount_opt;
1763 }
1764 #ifdef CONFIG_QUOTA
1765 } else if (m->flags & MOPT_QFMT) {
1766 if (sb_any_quota_loaded(sb) &&
1767 sbi->s_jquota_fmt != m->mount_opt) {
1768 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1769 "quota options when quota turned on");
1770 return -1;
1771 }
1772 if (ext4_has_feature_quota(sb)) {
1773 ext4_msg(sb, KERN_INFO,
1774 "Quota format mount options ignored "
1775 "when QUOTA feature is enabled");
1776 return 1;
1777 }
1778 sbi->s_jquota_fmt = m->mount_opt;
1779 #endif
1780 } else if (token == Opt_dax) {
1781 #ifdef CONFIG_FS_DAX
1782 ext4_msg(sb, KERN_WARNING,
1783 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1784 sbi->s_mount_opt |= m->mount_opt;
1785 #else
1786 ext4_msg(sb, KERN_INFO, "dax option not supported");
1787 return -1;
1788 #endif
1789 } else if (token == Opt_data_err_abort) {
1790 sbi->s_mount_opt |= m->mount_opt;
1791 } else if (token == Opt_data_err_ignore) {
1792 sbi->s_mount_opt &= ~m->mount_opt;
1793 } else {
1794 if (!args->from)
1795 arg = 1;
1796 if (m->flags & MOPT_CLEAR)
1797 arg = !arg;
1798 else if (unlikely(!(m->flags & MOPT_SET))) {
1799 ext4_msg(sb, KERN_WARNING,
1800 "buggy handling of option %s", opt);
1801 WARN_ON(1);
1802 return -1;
1803 }
1804 if (arg != 0)
1805 sbi->s_mount_opt |= m->mount_opt;
1806 else
1807 sbi->s_mount_opt &= ~m->mount_opt;
1808 }
1809 return 1;
1810 }
1811
1812 static int parse_options(char *options, struct super_block *sb,
1813 unsigned long *journal_devnum,
1814 unsigned int *journal_ioprio,
1815 int is_remount)
1816 {
1817 struct ext4_sb_info *sbi = EXT4_SB(sb);
1818 char *p;
1819 substring_t args[MAX_OPT_ARGS];
1820 int token;
1821
1822 if (!options)
1823 return 1;
1824
1825 while ((p = strsep(&options, ",")) != NULL) {
1826 if (!*p)
1827 continue;
1828 /*
1829 * Initialize args struct so we know whether arg was
1830 * found; some options take optional arguments.
1831 */
1832 args[0].to = args[0].from = NULL;
1833 token = match_token(p, tokens, args);
1834 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1835 journal_ioprio, is_remount) < 0)
1836 return 0;
1837 }
1838 #ifdef CONFIG_QUOTA
1839 if (ext4_has_feature_quota(sb) &&
1840 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1841 ext4_msg(sb, KERN_INFO, "Quota feature enabled, usrquota and grpquota "
1842 "mount options ignored.");
1843 clear_opt(sb, USRQUOTA);
1844 clear_opt(sb, GRPQUOTA);
1845 } else if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1846 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1847 clear_opt(sb, USRQUOTA);
1848
1849 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1850 clear_opt(sb, GRPQUOTA);
1851
1852 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1853 ext4_msg(sb, KERN_ERR, "old and new quota "
1854 "format mixing");
1855 return 0;
1856 }
1857
1858 if (!sbi->s_jquota_fmt) {
1859 ext4_msg(sb, KERN_ERR, "journaled quota format "
1860 "not specified");
1861 return 0;
1862 }
1863 }
1864 #endif
1865 if (test_opt(sb, DIOREAD_NOLOCK)) {
1866 int blocksize =
1867 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1868
1869 if (blocksize < PAGE_SIZE) {
1870 ext4_msg(sb, KERN_ERR, "can't mount with "
1871 "dioread_nolock if block size != PAGE_SIZE");
1872 return 0;
1873 }
1874 }
1875 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1876 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1877 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1878 "in data=ordered mode");
1879 return 0;
1880 }
1881 return 1;
1882 }
1883
1884 static inline void ext4_show_quota_options(struct seq_file *seq,
1885 struct super_block *sb)
1886 {
1887 #if defined(CONFIG_QUOTA)
1888 struct ext4_sb_info *sbi = EXT4_SB(sb);
1889
1890 if (sbi->s_jquota_fmt) {
1891 char *fmtname = "";
1892
1893 switch (sbi->s_jquota_fmt) {
1894 case QFMT_VFS_OLD:
1895 fmtname = "vfsold";
1896 break;
1897 case QFMT_VFS_V0:
1898 fmtname = "vfsv0";
1899 break;
1900 case QFMT_VFS_V1:
1901 fmtname = "vfsv1";
1902 break;
1903 }
1904 seq_printf(seq, ",jqfmt=%s", fmtname);
1905 }
1906
1907 if (sbi->s_qf_names[USRQUOTA])
1908 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1909
1910 if (sbi->s_qf_names[GRPQUOTA])
1911 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1912 #endif
1913 }
1914
1915 static const char *token2str(int token)
1916 {
1917 const struct match_token *t;
1918
1919 for (t = tokens; t->token != Opt_err; t++)
1920 if (t->token == token && !strchr(t->pattern, '='))
1921 break;
1922 return t->pattern;
1923 }
1924
1925 /*
1926 * Show an option if
1927 * - it's set to a non-default value OR
1928 * - if the per-sb default is different from the global default
1929 */
1930 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1931 int nodefs)
1932 {
1933 struct ext4_sb_info *sbi = EXT4_SB(sb);
1934 struct ext4_super_block *es = sbi->s_es;
1935 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1936 const struct mount_opts *m;
1937 char sep = nodefs ? '\n' : ',';
1938
1939 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1940 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1941
1942 if (sbi->s_sb_block != 1)
1943 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1944
1945 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1946 int want_set = m->flags & MOPT_SET;
1947 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1948 (m->flags & MOPT_CLEAR_ERR))
1949 continue;
1950 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1951 continue; /* skip if same as the default */
1952 if ((want_set &&
1953 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1954 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1955 continue; /* select Opt_noFoo vs Opt_Foo */
1956 SEQ_OPTS_PRINT("%s", token2str(m->token));
1957 }
1958
1959 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1960 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1961 SEQ_OPTS_PRINT("resuid=%u",
1962 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1963 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1964 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1965 SEQ_OPTS_PRINT("resgid=%u",
1966 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1967 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1968 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1969 SEQ_OPTS_PUTS("errors=remount-ro");
1970 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1971 SEQ_OPTS_PUTS("errors=continue");
1972 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1973 SEQ_OPTS_PUTS("errors=panic");
1974 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1975 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1976 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1977 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1978 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1979 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1980 if (sb->s_flags & MS_I_VERSION)
1981 SEQ_OPTS_PUTS("i_version");
1982 if (nodefs || sbi->s_stripe)
1983 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1984 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1985 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1986 SEQ_OPTS_PUTS("data=journal");
1987 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1988 SEQ_OPTS_PUTS("data=ordered");
1989 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1990 SEQ_OPTS_PUTS("data=writeback");
1991 }
1992 if (nodefs ||
1993 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1994 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1995 sbi->s_inode_readahead_blks);
1996
1997 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1998 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1999 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2000 if (nodefs || sbi->s_max_dir_size_kb)
2001 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2002 if (test_opt(sb, DATA_ERR_ABORT))
2003 SEQ_OPTS_PUTS("data_err=abort");
2004
2005 ext4_show_quota_options(seq, sb);
2006 return 0;
2007 }
2008
2009 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2010 {
2011 return _ext4_show_options(seq, root->d_sb, 0);
2012 }
2013
2014 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2015 {
2016 struct super_block *sb = seq->private;
2017 int rc;
2018
2019 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2020 rc = _ext4_show_options(seq, sb, 1);
2021 seq_puts(seq, "\n");
2022 return rc;
2023 }
2024
2025 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2026 int read_only)
2027 {
2028 struct ext4_sb_info *sbi = EXT4_SB(sb);
2029 int res = 0;
2030
2031 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2032 ext4_msg(sb, KERN_ERR, "revision level too high, "
2033 "forcing read-only mode");
2034 res = MS_RDONLY;
2035 }
2036 if (read_only)
2037 goto done;
2038 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2039 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2040 "running e2fsck is recommended");
2041 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2042 ext4_msg(sb, KERN_WARNING,
2043 "warning: mounting fs with errors, "
2044 "running e2fsck is recommended");
2045 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2046 le16_to_cpu(es->s_mnt_count) >=
2047 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2048 ext4_msg(sb, KERN_WARNING,
2049 "warning: maximal mount count reached, "
2050 "running e2fsck is recommended");
2051 else if (le32_to_cpu(es->s_checkinterval) &&
2052 (le32_to_cpu(es->s_lastcheck) +
2053 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2054 ext4_msg(sb, KERN_WARNING,
2055 "warning: checktime reached, "
2056 "running e2fsck is recommended");
2057 if (!sbi->s_journal)
2058 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2059 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2060 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2061 le16_add_cpu(&es->s_mnt_count, 1);
2062 es->s_mtime = cpu_to_le32(get_seconds());
2063 ext4_update_dynamic_rev(sb);
2064 if (sbi->s_journal)
2065 ext4_set_feature_journal_needs_recovery(sb);
2066
2067 ext4_commit_super(sb, 1);
2068 done:
2069 if (test_opt(sb, DEBUG))
2070 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2071 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2072 sb->s_blocksize,
2073 sbi->s_groups_count,
2074 EXT4_BLOCKS_PER_GROUP(sb),
2075 EXT4_INODES_PER_GROUP(sb),
2076 sbi->s_mount_opt, sbi->s_mount_opt2);
2077
2078 cleancache_init_fs(sb);
2079 return res;
2080 }
2081
2082 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2083 {
2084 struct ext4_sb_info *sbi = EXT4_SB(sb);
2085 struct flex_groups *new_groups;
2086 int size;
2087
2088 if (!sbi->s_log_groups_per_flex)
2089 return 0;
2090
2091 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2092 if (size <= sbi->s_flex_groups_allocated)
2093 return 0;
2094
2095 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2096 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2097 if (!new_groups) {
2098 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2099 size / (int) sizeof(struct flex_groups));
2100 return -ENOMEM;
2101 }
2102
2103 if (sbi->s_flex_groups) {
2104 memcpy(new_groups, sbi->s_flex_groups,
2105 (sbi->s_flex_groups_allocated *
2106 sizeof(struct flex_groups)));
2107 kvfree(sbi->s_flex_groups);
2108 }
2109 sbi->s_flex_groups = new_groups;
2110 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2111 return 0;
2112 }
2113
2114 static int ext4_fill_flex_info(struct super_block *sb)
2115 {
2116 struct ext4_sb_info *sbi = EXT4_SB(sb);
2117 struct ext4_group_desc *gdp = NULL;
2118 ext4_group_t flex_group;
2119 int i, err;
2120
2121 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2122 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2123 sbi->s_log_groups_per_flex = 0;
2124 return 1;
2125 }
2126
2127 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2128 if (err)
2129 goto failed;
2130
2131 for (i = 0; i < sbi->s_groups_count; i++) {
2132 gdp = ext4_get_group_desc(sb, i, NULL);
2133
2134 flex_group = ext4_flex_group(sbi, i);
2135 atomic_add(ext4_free_inodes_count(sb, gdp),
2136 &sbi->s_flex_groups[flex_group].free_inodes);
2137 atomic64_add(ext4_free_group_clusters(sb, gdp),
2138 &sbi->s_flex_groups[flex_group].free_clusters);
2139 atomic_add(ext4_used_dirs_count(sb, gdp),
2140 &sbi->s_flex_groups[flex_group].used_dirs);
2141 }
2142
2143 return 1;
2144 failed:
2145 return 0;
2146 }
2147
2148 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2149 struct ext4_group_desc *gdp)
2150 {
2151 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2152 __u16 crc = 0;
2153 __le32 le_group = cpu_to_le32(block_group);
2154 struct ext4_sb_info *sbi = EXT4_SB(sb);
2155
2156 if (ext4_has_metadata_csum(sbi->s_sb)) {
2157 /* Use new metadata_csum algorithm */
2158 __u32 csum32;
2159 __u16 dummy_csum = 0;
2160
2161 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2162 sizeof(le_group));
2163 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2164 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2165 sizeof(dummy_csum));
2166 offset += sizeof(dummy_csum);
2167 if (offset < sbi->s_desc_size)
2168 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2169 sbi->s_desc_size - offset);
2170
2171 crc = csum32 & 0xFFFF;
2172 goto out;
2173 }
2174
2175 /* old crc16 code */
2176 if (!ext4_has_feature_gdt_csum(sb))
2177 return 0;
2178
2179 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2180 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2181 crc = crc16(crc, (__u8 *)gdp, offset);
2182 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2183 /* for checksum of struct ext4_group_desc do the rest...*/
2184 if (ext4_has_feature_64bit(sb) &&
2185 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2186 crc = crc16(crc, (__u8 *)gdp + offset,
2187 le16_to_cpu(sbi->s_es->s_desc_size) -
2188 offset);
2189
2190 out:
2191 return cpu_to_le16(crc);
2192 }
2193
2194 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2195 struct ext4_group_desc *gdp)
2196 {
2197 if (ext4_has_group_desc_csum(sb) &&
2198 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2199 return 0;
2200
2201 return 1;
2202 }
2203
2204 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2205 struct ext4_group_desc *gdp)
2206 {
2207 if (!ext4_has_group_desc_csum(sb))
2208 return;
2209 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2210 }
2211
2212 /* Called at mount-time, super-block is locked */
2213 static int ext4_check_descriptors(struct super_block *sb,
2214 ext4_fsblk_t sb_block,
2215 ext4_group_t *first_not_zeroed)
2216 {
2217 struct ext4_sb_info *sbi = EXT4_SB(sb);
2218 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2219 ext4_fsblk_t last_block;
2220 ext4_fsblk_t block_bitmap;
2221 ext4_fsblk_t inode_bitmap;
2222 ext4_fsblk_t inode_table;
2223 int flexbg_flag = 0;
2224 ext4_group_t i, grp = sbi->s_groups_count;
2225
2226 if (ext4_has_feature_flex_bg(sb))
2227 flexbg_flag = 1;
2228
2229 ext4_debug("Checking group descriptors");
2230
2231 for (i = 0; i < sbi->s_groups_count; i++) {
2232 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2233
2234 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2235 last_block = ext4_blocks_count(sbi->s_es) - 1;
2236 else
2237 last_block = first_block +
2238 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2239
2240 if ((grp == sbi->s_groups_count) &&
2241 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2242 grp = i;
2243
2244 block_bitmap = ext4_block_bitmap(sb, gdp);
2245 if (block_bitmap == sb_block) {
2246 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2247 "Block bitmap for group %u overlaps "
2248 "superblock", i);
2249 }
2250 if (block_bitmap < first_block || block_bitmap > last_block) {
2251 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2252 "Block bitmap for group %u not in group "
2253 "(block %llu)!", i, block_bitmap);
2254 return 0;
2255 }
2256 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2257 if (inode_bitmap == sb_block) {
2258 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2259 "Inode bitmap for group %u overlaps "
2260 "superblock", i);
2261 }
2262 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2263 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2264 "Inode bitmap for group %u not in group "
2265 "(block %llu)!", i, inode_bitmap);
2266 return 0;
2267 }
2268 inode_table = ext4_inode_table(sb, gdp);
2269 if (inode_table == sb_block) {
2270 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2271 "Inode table for group %u overlaps "
2272 "superblock", i);
2273 }
2274 if (inode_table < first_block ||
2275 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2276 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2277 "Inode table for group %u not in group "
2278 "(block %llu)!", i, inode_table);
2279 return 0;
2280 }
2281 ext4_lock_group(sb, i);
2282 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2283 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2284 "Checksum for group %u failed (%u!=%u)",
2285 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2286 gdp)), le16_to_cpu(gdp->bg_checksum));
2287 if (!(sb->s_flags & MS_RDONLY)) {
2288 ext4_unlock_group(sb, i);
2289 return 0;
2290 }
2291 }
2292 ext4_unlock_group(sb, i);
2293 if (!flexbg_flag)
2294 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2295 }
2296 if (NULL != first_not_zeroed)
2297 *first_not_zeroed = grp;
2298 return 1;
2299 }
2300
2301 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2302 * the superblock) which were deleted from all directories, but held open by
2303 * a process at the time of a crash. We walk the list and try to delete these
2304 * inodes at recovery time (only with a read-write filesystem).
2305 *
2306 * In order to keep the orphan inode chain consistent during traversal (in
2307 * case of crash during recovery), we link each inode into the superblock
2308 * orphan list_head and handle it the same way as an inode deletion during
2309 * normal operation (which journals the operations for us).
2310 *
2311 * We only do an iget() and an iput() on each inode, which is very safe if we
2312 * accidentally point at an in-use or already deleted inode. The worst that
2313 * can happen in this case is that we get a "bit already cleared" message from
2314 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2315 * e2fsck was run on this filesystem, and it must have already done the orphan
2316 * inode cleanup for us, so we can safely abort without any further action.
2317 */
2318 static void ext4_orphan_cleanup(struct super_block *sb,
2319 struct ext4_super_block *es)
2320 {
2321 unsigned int s_flags = sb->s_flags;
2322 int nr_orphans = 0, nr_truncates = 0;
2323 #ifdef CONFIG_QUOTA
2324 int i;
2325 #endif
2326 if (!es->s_last_orphan) {
2327 jbd_debug(4, "no orphan inodes to clean up\n");
2328 return;
2329 }
2330
2331 if (bdev_read_only(sb->s_bdev)) {
2332 ext4_msg(sb, KERN_ERR, "write access "
2333 "unavailable, skipping orphan cleanup");
2334 return;
2335 }
2336
2337 /* Check if feature set would not allow a r/w mount */
2338 if (!ext4_feature_set_ok(sb, 0)) {
2339 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2340 "unknown ROCOMPAT features");
2341 return;
2342 }
2343
2344 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2345 /* don't clear list on RO mount w/ errors */
2346 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2347 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2348 "clearing orphan list.\n");
2349 es->s_last_orphan = 0;
2350 }
2351 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2352 return;
2353 }
2354
2355 if (s_flags & MS_RDONLY) {
2356 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2357 sb->s_flags &= ~MS_RDONLY;
2358 }
2359 #ifdef CONFIG_QUOTA
2360 /* Needed for iput() to work correctly and not trash data */
2361 sb->s_flags |= MS_ACTIVE;
2362 /* Turn on quotas so that they are updated correctly */
2363 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2364 if (EXT4_SB(sb)->s_qf_names[i]) {
2365 int ret = ext4_quota_on_mount(sb, i);
2366 if (ret < 0)
2367 ext4_msg(sb, KERN_ERR,
2368 "Cannot turn on journaled "
2369 "quota: error %d", ret);
2370 }
2371 }
2372 #endif
2373
2374 while (es->s_last_orphan) {
2375 struct inode *inode;
2376
2377 /*
2378 * We may have encountered an error during cleanup; if
2379 * so, skip the rest.
2380 */
2381 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2382 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2383 es->s_last_orphan = 0;
2384 break;
2385 }
2386
2387 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2388 if (IS_ERR(inode)) {
2389 es->s_last_orphan = 0;
2390 break;
2391 }
2392
2393 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2394 dquot_initialize(inode);
2395 if (inode->i_nlink) {
2396 if (test_opt(sb, DEBUG))
2397 ext4_msg(sb, KERN_DEBUG,
2398 "%s: truncating inode %lu to %lld bytes",
2399 __func__, inode->i_ino, inode->i_size);
2400 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2401 inode->i_ino, inode->i_size);
2402 inode_lock(inode);
2403 truncate_inode_pages(inode->i_mapping, inode->i_size);
2404 ext4_truncate(inode);
2405 inode_unlock(inode);
2406 nr_truncates++;
2407 } else {
2408 if (test_opt(sb, DEBUG))
2409 ext4_msg(sb, KERN_DEBUG,
2410 "%s: deleting unreferenced inode %lu",
2411 __func__, inode->i_ino);
2412 jbd_debug(2, "deleting unreferenced inode %lu\n",
2413 inode->i_ino);
2414 nr_orphans++;
2415 }
2416 iput(inode); /* The delete magic happens here! */
2417 }
2418
2419 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2420
2421 if (nr_orphans)
2422 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2423 PLURAL(nr_orphans));
2424 if (nr_truncates)
2425 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2426 PLURAL(nr_truncates));
2427 #ifdef CONFIG_QUOTA
2428 /* Turn quotas off */
2429 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2430 if (sb_dqopt(sb)->files[i])
2431 dquot_quota_off(sb, i);
2432 }
2433 #endif
2434 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2435 }
2436
2437 /*
2438 * Maximal extent format file size.
2439 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2440 * extent format containers, within a sector_t, and within i_blocks
2441 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2442 * so that won't be a limiting factor.
2443 *
2444 * However there is other limiting factor. We do store extents in the form
2445 * of starting block and length, hence the resulting length of the extent
2446 * covering maximum file size must fit into on-disk format containers as
2447 * well. Given that length is always by 1 unit bigger than max unit (because
2448 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2449 *
2450 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2451 */
2452 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2453 {
2454 loff_t res;
2455 loff_t upper_limit = MAX_LFS_FILESIZE;
2456
2457 /* small i_blocks in vfs inode? */
2458 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2459 /*
2460 * CONFIG_LBDAF is not enabled implies the inode
2461 * i_block represent total blocks in 512 bytes
2462 * 32 == size of vfs inode i_blocks * 8
2463 */
2464 upper_limit = (1LL << 32) - 1;
2465
2466 /* total blocks in file system block size */
2467 upper_limit >>= (blkbits - 9);
2468 upper_limit <<= blkbits;
2469 }
2470
2471 /*
2472 * 32-bit extent-start container, ee_block. We lower the maxbytes
2473 * by one fs block, so ee_len can cover the extent of maximum file
2474 * size
2475 */
2476 res = (1LL << 32) - 1;
2477 res <<= blkbits;
2478
2479 /* Sanity check against vm- & vfs- imposed limits */
2480 if (res > upper_limit)
2481 res = upper_limit;
2482
2483 return res;
2484 }
2485
2486 /*
2487 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2488 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2489 * We need to be 1 filesystem block less than the 2^48 sector limit.
2490 */
2491 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2492 {
2493 loff_t res = EXT4_NDIR_BLOCKS;
2494 int meta_blocks;
2495 loff_t upper_limit;
2496 /* This is calculated to be the largest file size for a dense, block
2497 * mapped file such that the file's total number of 512-byte sectors,
2498 * including data and all indirect blocks, does not exceed (2^48 - 1).
2499 *
2500 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2501 * number of 512-byte sectors of the file.
2502 */
2503
2504 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2505 /*
2506 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2507 * the inode i_block field represents total file blocks in
2508 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2509 */
2510 upper_limit = (1LL << 32) - 1;
2511
2512 /* total blocks in file system block size */
2513 upper_limit >>= (bits - 9);
2514
2515 } else {
2516 /*
2517 * We use 48 bit ext4_inode i_blocks
2518 * With EXT4_HUGE_FILE_FL set the i_blocks
2519 * represent total number of blocks in
2520 * file system block size
2521 */
2522 upper_limit = (1LL << 48) - 1;
2523
2524 }
2525
2526 /* indirect blocks */
2527 meta_blocks = 1;
2528 /* double indirect blocks */
2529 meta_blocks += 1 + (1LL << (bits-2));
2530 /* tripple indirect blocks */
2531 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2532
2533 upper_limit -= meta_blocks;
2534 upper_limit <<= bits;
2535
2536 res += 1LL << (bits-2);
2537 res += 1LL << (2*(bits-2));
2538 res += 1LL << (3*(bits-2));
2539 res <<= bits;
2540 if (res > upper_limit)
2541 res = upper_limit;
2542
2543 if (res > MAX_LFS_FILESIZE)
2544 res = MAX_LFS_FILESIZE;
2545
2546 return res;
2547 }
2548
2549 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2550 ext4_fsblk_t logical_sb_block, int nr)
2551 {
2552 struct ext4_sb_info *sbi = EXT4_SB(sb);
2553 ext4_group_t bg, first_meta_bg;
2554 int has_super = 0;
2555
2556 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2557
2558 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2559 return logical_sb_block + nr + 1;
2560 bg = sbi->s_desc_per_block * nr;
2561 if (ext4_bg_has_super(sb, bg))
2562 has_super = 1;
2563
2564 /*
2565 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2566 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2567 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2568 * compensate.
2569 */
2570 if (sb->s_blocksize == 1024 && nr == 0 &&
2571 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2572 has_super++;
2573
2574 return (has_super + ext4_group_first_block_no(sb, bg));
2575 }
2576
2577 /**
2578 * ext4_get_stripe_size: Get the stripe size.
2579 * @sbi: In memory super block info
2580 *
2581 * If we have specified it via mount option, then
2582 * use the mount option value. If the value specified at mount time is
2583 * greater than the blocks per group use the super block value.
2584 * If the super block value is greater than blocks per group return 0.
2585 * Allocator needs it be less than blocks per group.
2586 *
2587 */
2588 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2589 {
2590 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2591 unsigned long stripe_width =
2592 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2593 int ret;
2594
2595 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2596 ret = sbi->s_stripe;
2597 else if (stripe_width <= sbi->s_blocks_per_group)
2598 ret = stripe_width;
2599 else if (stride <= sbi->s_blocks_per_group)
2600 ret = stride;
2601 else
2602 ret = 0;
2603
2604 /*
2605 * If the stripe width is 1, this makes no sense and
2606 * we set it to 0 to turn off stripe handling code.
2607 */
2608 if (ret <= 1)
2609 ret = 0;
2610
2611 return ret;
2612 }
2613
2614 /*
2615 * Check whether this filesystem can be mounted based on
2616 * the features present and the RDONLY/RDWR mount requested.
2617 * Returns 1 if this filesystem can be mounted as requested,
2618 * 0 if it cannot be.
2619 */
2620 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2621 {
2622 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2623 ext4_msg(sb, KERN_ERR,
2624 "Couldn't mount because of "
2625 "unsupported optional features (%x)",
2626 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2627 ~EXT4_FEATURE_INCOMPAT_SUPP));
2628 return 0;
2629 }
2630
2631 if (readonly)
2632 return 1;
2633
2634 if (ext4_has_feature_readonly(sb)) {
2635 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2636 sb->s_flags |= MS_RDONLY;
2637 return 1;
2638 }
2639
2640 /* Check that feature set is OK for a read-write mount */
2641 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2642 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2643 "unsupported optional features (%x)",
2644 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2645 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2646 return 0;
2647 }
2648 /*
2649 * Large file size enabled file system can only be mounted
2650 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2651 */
2652 if (ext4_has_feature_huge_file(sb)) {
2653 if (sizeof(blkcnt_t) < sizeof(u64)) {
2654 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2655 "cannot be mounted RDWR without "
2656 "CONFIG_LBDAF");
2657 return 0;
2658 }
2659 }
2660 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2661 ext4_msg(sb, KERN_ERR,
2662 "Can't support bigalloc feature without "
2663 "extents feature\n");
2664 return 0;
2665 }
2666
2667 #ifndef CONFIG_QUOTA
2668 if (ext4_has_feature_quota(sb) && !readonly) {
2669 ext4_msg(sb, KERN_ERR,
2670 "Filesystem with quota feature cannot be mounted RDWR "
2671 "without CONFIG_QUOTA");
2672 return 0;
2673 }
2674 if (ext4_has_feature_project(sb) && !readonly) {
2675 ext4_msg(sb, KERN_ERR,
2676 "Filesystem with project quota feature cannot be mounted RDWR "
2677 "without CONFIG_QUOTA");
2678 return 0;
2679 }
2680 #endif /* CONFIG_QUOTA */
2681 return 1;
2682 }
2683
2684 /*
2685 * This function is called once a day if we have errors logged
2686 * on the file system
2687 */
2688 static void print_daily_error_info(unsigned long arg)
2689 {
2690 struct super_block *sb = (struct super_block *) arg;
2691 struct ext4_sb_info *sbi;
2692 struct ext4_super_block *es;
2693
2694 sbi = EXT4_SB(sb);
2695 es = sbi->s_es;
2696
2697 if (es->s_error_count)
2698 /* fsck newer than v1.41.13 is needed to clean this condition. */
2699 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2700 le32_to_cpu(es->s_error_count));
2701 if (es->s_first_error_time) {
2702 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2703 sb->s_id, le32_to_cpu(es->s_first_error_time),
2704 (int) sizeof(es->s_first_error_func),
2705 es->s_first_error_func,
2706 le32_to_cpu(es->s_first_error_line));
2707 if (es->s_first_error_ino)
2708 printk(": inode %u",
2709 le32_to_cpu(es->s_first_error_ino));
2710 if (es->s_first_error_block)
2711 printk(": block %llu", (unsigned long long)
2712 le64_to_cpu(es->s_first_error_block));
2713 printk("\n");
2714 }
2715 if (es->s_last_error_time) {
2716 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2717 sb->s_id, le32_to_cpu(es->s_last_error_time),
2718 (int) sizeof(es->s_last_error_func),
2719 es->s_last_error_func,
2720 le32_to_cpu(es->s_last_error_line));
2721 if (es->s_last_error_ino)
2722 printk(": inode %u",
2723 le32_to_cpu(es->s_last_error_ino));
2724 if (es->s_last_error_block)
2725 printk(": block %llu", (unsigned long long)
2726 le64_to_cpu(es->s_last_error_block));
2727 printk("\n");
2728 }
2729 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2730 }
2731
2732 /* Find next suitable group and run ext4_init_inode_table */
2733 static int ext4_run_li_request(struct ext4_li_request *elr)
2734 {
2735 struct ext4_group_desc *gdp = NULL;
2736 ext4_group_t group, ngroups;
2737 struct super_block *sb;
2738 unsigned long timeout = 0;
2739 int ret = 0;
2740
2741 sb = elr->lr_super;
2742 ngroups = EXT4_SB(sb)->s_groups_count;
2743
2744 sb_start_write(sb);
2745 for (group = elr->lr_next_group; group < ngroups; group++) {
2746 gdp = ext4_get_group_desc(sb, group, NULL);
2747 if (!gdp) {
2748 ret = 1;
2749 break;
2750 }
2751
2752 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2753 break;
2754 }
2755
2756 if (group >= ngroups)
2757 ret = 1;
2758
2759 if (!ret) {
2760 timeout = jiffies;
2761 ret = ext4_init_inode_table(sb, group,
2762 elr->lr_timeout ? 0 : 1);
2763 if (elr->lr_timeout == 0) {
2764 timeout = (jiffies - timeout) *
2765 elr->lr_sbi->s_li_wait_mult;
2766 elr->lr_timeout = timeout;
2767 }
2768 elr->lr_next_sched = jiffies + elr->lr_timeout;
2769 elr->lr_next_group = group + 1;
2770 }
2771 sb_end_write(sb);
2772
2773 return ret;
2774 }
2775
2776 /*
2777 * Remove lr_request from the list_request and free the
2778 * request structure. Should be called with li_list_mtx held
2779 */
2780 static void ext4_remove_li_request(struct ext4_li_request *elr)
2781 {
2782 struct ext4_sb_info *sbi;
2783
2784 if (!elr)
2785 return;
2786
2787 sbi = elr->lr_sbi;
2788
2789 list_del(&elr->lr_request);
2790 sbi->s_li_request = NULL;
2791 kfree(elr);
2792 }
2793
2794 static void ext4_unregister_li_request(struct super_block *sb)
2795 {
2796 mutex_lock(&ext4_li_mtx);
2797 if (!ext4_li_info) {
2798 mutex_unlock(&ext4_li_mtx);
2799 return;
2800 }
2801
2802 mutex_lock(&ext4_li_info->li_list_mtx);
2803 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2804 mutex_unlock(&ext4_li_info->li_list_mtx);
2805 mutex_unlock(&ext4_li_mtx);
2806 }
2807
2808 static struct task_struct *ext4_lazyinit_task;
2809
2810 /*
2811 * This is the function where ext4lazyinit thread lives. It walks
2812 * through the request list searching for next scheduled filesystem.
2813 * When such a fs is found, run the lazy initialization request
2814 * (ext4_rn_li_request) and keep track of the time spend in this
2815 * function. Based on that time we compute next schedule time of
2816 * the request. When walking through the list is complete, compute
2817 * next waking time and put itself into sleep.
2818 */
2819 static int ext4_lazyinit_thread(void *arg)
2820 {
2821 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2822 struct list_head *pos, *n;
2823 struct ext4_li_request *elr;
2824 unsigned long next_wakeup, cur;
2825
2826 BUG_ON(NULL == eli);
2827
2828 cont_thread:
2829 while (true) {
2830 next_wakeup = MAX_JIFFY_OFFSET;
2831
2832 mutex_lock(&eli->li_list_mtx);
2833 if (list_empty(&eli->li_request_list)) {
2834 mutex_unlock(&eli->li_list_mtx);
2835 goto exit_thread;
2836 }
2837
2838 list_for_each_safe(pos, n, &eli->li_request_list) {
2839 elr = list_entry(pos, struct ext4_li_request,
2840 lr_request);
2841
2842 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2843 if (ext4_run_li_request(elr) != 0) {
2844 /* error, remove the lazy_init job */
2845 ext4_remove_li_request(elr);
2846 continue;
2847 }
2848 }
2849
2850 if (time_before(elr->lr_next_sched, next_wakeup))
2851 next_wakeup = elr->lr_next_sched;
2852 }
2853 mutex_unlock(&eli->li_list_mtx);
2854
2855 try_to_freeze();
2856
2857 cur = jiffies;
2858 if ((time_after_eq(cur, next_wakeup)) ||
2859 (MAX_JIFFY_OFFSET == next_wakeup)) {
2860 cond_resched();
2861 continue;
2862 }
2863
2864 schedule_timeout_interruptible(next_wakeup - cur);
2865
2866 if (kthread_should_stop()) {
2867 ext4_clear_request_list();
2868 goto exit_thread;
2869 }
2870 }
2871
2872 exit_thread:
2873 /*
2874 * It looks like the request list is empty, but we need
2875 * to check it under the li_list_mtx lock, to prevent any
2876 * additions into it, and of course we should lock ext4_li_mtx
2877 * to atomically free the list and ext4_li_info, because at
2878 * this point another ext4 filesystem could be registering
2879 * new one.
2880 */
2881 mutex_lock(&ext4_li_mtx);
2882 mutex_lock(&eli->li_list_mtx);
2883 if (!list_empty(&eli->li_request_list)) {
2884 mutex_unlock(&eli->li_list_mtx);
2885 mutex_unlock(&ext4_li_mtx);
2886 goto cont_thread;
2887 }
2888 mutex_unlock(&eli->li_list_mtx);
2889 kfree(ext4_li_info);
2890 ext4_li_info = NULL;
2891 mutex_unlock(&ext4_li_mtx);
2892
2893 return 0;
2894 }
2895
2896 static void ext4_clear_request_list(void)
2897 {
2898 struct list_head *pos, *n;
2899 struct ext4_li_request *elr;
2900
2901 mutex_lock(&ext4_li_info->li_list_mtx);
2902 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2903 elr = list_entry(pos, struct ext4_li_request,
2904 lr_request);
2905 ext4_remove_li_request(elr);
2906 }
2907 mutex_unlock(&ext4_li_info->li_list_mtx);
2908 }
2909
2910 static int ext4_run_lazyinit_thread(void)
2911 {
2912 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2913 ext4_li_info, "ext4lazyinit");
2914 if (IS_ERR(ext4_lazyinit_task)) {
2915 int err = PTR_ERR(ext4_lazyinit_task);
2916 ext4_clear_request_list();
2917 kfree(ext4_li_info);
2918 ext4_li_info = NULL;
2919 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2920 "initialization thread\n",
2921 err);
2922 return err;
2923 }
2924 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2925 return 0;
2926 }
2927
2928 /*
2929 * Check whether it make sense to run itable init. thread or not.
2930 * If there is at least one uninitialized inode table, return
2931 * corresponding group number, else the loop goes through all
2932 * groups and return total number of groups.
2933 */
2934 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2935 {
2936 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2937 struct ext4_group_desc *gdp = NULL;
2938
2939 for (group = 0; group < ngroups; group++) {
2940 gdp = ext4_get_group_desc(sb, group, NULL);
2941 if (!gdp)
2942 continue;
2943
2944 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2945 break;
2946 }
2947
2948 return group;
2949 }
2950
2951 static int ext4_li_info_new(void)
2952 {
2953 struct ext4_lazy_init *eli = NULL;
2954
2955 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2956 if (!eli)
2957 return -ENOMEM;
2958
2959 INIT_LIST_HEAD(&eli->li_request_list);
2960 mutex_init(&eli->li_list_mtx);
2961
2962 eli->li_state |= EXT4_LAZYINIT_QUIT;
2963
2964 ext4_li_info = eli;
2965
2966 return 0;
2967 }
2968
2969 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2970 ext4_group_t start)
2971 {
2972 struct ext4_sb_info *sbi = EXT4_SB(sb);
2973 struct ext4_li_request *elr;
2974
2975 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2976 if (!elr)
2977 return NULL;
2978
2979 elr->lr_super = sb;
2980 elr->lr_sbi = sbi;
2981 elr->lr_next_group = start;
2982
2983 /*
2984 * Randomize first schedule time of the request to
2985 * spread the inode table initialization requests
2986 * better.
2987 */
2988 elr->lr_next_sched = jiffies + (prandom_u32() %
2989 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2990 return elr;
2991 }
2992
2993 int ext4_register_li_request(struct super_block *sb,
2994 ext4_group_t first_not_zeroed)
2995 {
2996 struct ext4_sb_info *sbi = EXT4_SB(sb);
2997 struct ext4_li_request *elr = NULL;
2998 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2999 int ret = 0;
3000
3001 mutex_lock(&ext4_li_mtx);
3002 if (sbi->s_li_request != NULL) {
3003 /*
3004 * Reset timeout so it can be computed again, because
3005 * s_li_wait_mult might have changed.
3006 */
3007 sbi->s_li_request->lr_timeout = 0;
3008 goto out;
3009 }
3010
3011 if (first_not_zeroed == ngroups ||
3012 (sb->s_flags & MS_RDONLY) ||
3013 !test_opt(sb, INIT_INODE_TABLE))
3014 goto out;
3015
3016 elr = ext4_li_request_new(sb, first_not_zeroed);
3017 if (!elr) {
3018 ret = -ENOMEM;
3019 goto out;
3020 }
3021
3022 if (NULL == ext4_li_info) {
3023 ret = ext4_li_info_new();
3024 if (ret)
3025 goto out;
3026 }
3027
3028 mutex_lock(&ext4_li_info->li_list_mtx);
3029 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3030 mutex_unlock(&ext4_li_info->li_list_mtx);
3031
3032 sbi->s_li_request = elr;
3033 /*
3034 * set elr to NULL here since it has been inserted to
3035 * the request_list and the removal and free of it is
3036 * handled by ext4_clear_request_list from now on.
3037 */
3038 elr = NULL;
3039
3040 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3041 ret = ext4_run_lazyinit_thread();
3042 if (ret)
3043 goto out;
3044 }
3045 out:
3046 mutex_unlock(&ext4_li_mtx);
3047 if (ret)
3048 kfree(elr);
3049 return ret;
3050 }
3051
3052 /*
3053 * We do not need to lock anything since this is called on
3054 * module unload.
3055 */
3056 static void ext4_destroy_lazyinit_thread(void)
3057 {
3058 /*
3059 * If thread exited earlier
3060 * there's nothing to be done.
3061 */
3062 if (!ext4_li_info || !ext4_lazyinit_task)
3063 return;
3064
3065 kthread_stop(ext4_lazyinit_task);
3066 }
3067
3068 static int set_journal_csum_feature_set(struct super_block *sb)
3069 {
3070 int ret = 1;
3071 int compat, incompat;
3072 struct ext4_sb_info *sbi = EXT4_SB(sb);
3073
3074 if (ext4_has_metadata_csum(sb)) {
3075 /* journal checksum v3 */
3076 compat = 0;
3077 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3078 } else {
3079 /* journal checksum v1 */
3080 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3081 incompat = 0;
3082 }
3083
3084 jbd2_journal_clear_features(sbi->s_journal,
3085 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3086 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3087 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3088 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3089 ret = jbd2_journal_set_features(sbi->s_journal,
3090 compat, 0,
3091 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3092 incompat);
3093 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3094 ret = jbd2_journal_set_features(sbi->s_journal,
3095 compat, 0,
3096 incompat);
3097 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3098 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3099 } else {
3100 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3101 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3102 }
3103
3104 return ret;
3105 }
3106
3107 /*
3108 * Note: calculating the overhead so we can be compatible with
3109 * historical BSD practice is quite difficult in the face of
3110 * clusters/bigalloc. This is because multiple metadata blocks from
3111 * different block group can end up in the same allocation cluster.
3112 * Calculating the exact overhead in the face of clustered allocation
3113 * requires either O(all block bitmaps) in memory or O(number of block
3114 * groups**2) in time. We will still calculate the superblock for
3115 * older file systems --- and if we come across with a bigalloc file
3116 * system with zero in s_overhead_clusters the estimate will be close to
3117 * correct especially for very large cluster sizes --- but for newer
3118 * file systems, it's better to calculate this figure once at mkfs
3119 * time, and store it in the superblock. If the superblock value is
3120 * present (even for non-bigalloc file systems), we will use it.
3121 */
3122 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3123 char *buf)
3124 {
3125 struct ext4_sb_info *sbi = EXT4_SB(sb);
3126 struct ext4_group_desc *gdp;
3127 ext4_fsblk_t first_block, last_block, b;
3128 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3129 int s, j, count = 0;
3130
3131 if (!ext4_has_feature_bigalloc(sb))
3132 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3133 sbi->s_itb_per_group + 2);
3134
3135 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3136 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3137 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3138 for (i = 0; i < ngroups; i++) {
3139 gdp = ext4_get_group_desc(sb, i, NULL);
3140 b = ext4_block_bitmap(sb, gdp);
3141 if (b >= first_block && b <= last_block) {
3142 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3143 count++;
3144 }
3145 b = ext4_inode_bitmap(sb, gdp);
3146 if (b >= first_block && b <= last_block) {
3147 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3148 count++;
3149 }
3150 b = ext4_inode_table(sb, gdp);
3151 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3152 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3153 int c = EXT4_B2C(sbi, b - first_block);
3154 ext4_set_bit(c, buf);
3155 count++;
3156 }
3157 if (i != grp)
3158 continue;
3159 s = 0;
3160 if (ext4_bg_has_super(sb, grp)) {
3161 ext4_set_bit(s++, buf);
3162 count++;
3163 }
3164 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3165 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3166 count++;
3167 }
3168 }
3169 if (!count)
3170 return 0;
3171 return EXT4_CLUSTERS_PER_GROUP(sb) -
3172 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3173 }
3174
3175 /*
3176 * Compute the overhead and stash it in sbi->s_overhead
3177 */
3178 int ext4_calculate_overhead(struct super_block *sb)
3179 {
3180 struct ext4_sb_info *sbi = EXT4_SB(sb);
3181 struct ext4_super_block *es = sbi->s_es;
3182 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3183 ext4_fsblk_t overhead = 0;
3184 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3185
3186 if (!buf)
3187 return -ENOMEM;
3188
3189 /*
3190 * Compute the overhead (FS structures). This is constant
3191 * for a given filesystem unless the number of block groups
3192 * changes so we cache the previous value until it does.
3193 */
3194
3195 /*
3196 * All of the blocks before first_data_block are overhead
3197 */
3198 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3199
3200 /*
3201 * Add the overhead found in each block group
3202 */
3203 for (i = 0; i < ngroups; i++) {
3204 int blks;
3205
3206 blks = count_overhead(sb, i, buf);
3207 overhead += blks;
3208 if (blks)
3209 memset(buf, 0, PAGE_SIZE);
3210 cond_resched();
3211 }
3212 /* Add the internal journal blocks as well */
3213 if (sbi->s_journal && !sbi->journal_bdev)
3214 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3215
3216 sbi->s_overhead = overhead;
3217 smp_wmb();
3218 free_page((unsigned long) buf);
3219 return 0;
3220 }
3221
3222 static void ext4_set_resv_clusters(struct super_block *sb)
3223 {
3224 ext4_fsblk_t resv_clusters;
3225 struct ext4_sb_info *sbi = EXT4_SB(sb);
3226
3227 /*
3228 * There's no need to reserve anything when we aren't using extents.
3229 * The space estimates are exact, there are no unwritten extents,
3230 * hole punching doesn't need new metadata... This is needed especially
3231 * to keep ext2/3 backward compatibility.
3232 */
3233 if (!ext4_has_feature_extents(sb))
3234 return;
3235 /*
3236 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3237 * This should cover the situations where we can not afford to run
3238 * out of space like for example punch hole, or converting
3239 * unwritten extents in delalloc path. In most cases such
3240 * allocation would require 1, or 2 blocks, higher numbers are
3241 * very rare.
3242 */
3243 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3244 sbi->s_cluster_bits);
3245
3246 do_div(resv_clusters, 50);
3247 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3248
3249 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3250 }
3251
3252 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3253 {
3254 char *orig_data = kstrdup(data, GFP_KERNEL);
3255 struct buffer_head *bh;
3256 struct ext4_super_block *es = NULL;
3257 struct ext4_sb_info *sbi;
3258 ext4_fsblk_t block;
3259 ext4_fsblk_t sb_block = get_sb_block(&data);
3260 ext4_fsblk_t logical_sb_block;
3261 unsigned long offset = 0;
3262 unsigned long journal_devnum = 0;
3263 unsigned long def_mount_opts;
3264 struct inode *root;
3265 const char *descr;
3266 int ret = -ENOMEM;
3267 int blocksize, clustersize;
3268 unsigned int db_count;
3269 unsigned int i;
3270 int needs_recovery, has_huge_files, has_bigalloc;
3271 __u64 blocks_count;
3272 int err = 0;
3273 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3274 ext4_group_t first_not_zeroed;
3275
3276 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3277 if (!sbi)
3278 goto out_free_orig;
3279
3280 sbi->s_blockgroup_lock =
3281 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3282 if (!sbi->s_blockgroup_lock) {
3283 kfree(sbi);
3284 goto out_free_orig;
3285 }
3286 sb->s_fs_info = sbi;
3287 sbi->s_sb = sb;
3288 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3289 sbi->s_sb_block = sb_block;
3290 if (sb->s_bdev->bd_part)
3291 sbi->s_sectors_written_start =
3292 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3293
3294 /* Cleanup superblock name */
3295 strreplace(sb->s_id, '/', '!');
3296
3297 /* -EINVAL is default */
3298 ret = -EINVAL;
3299 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3300 if (!blocksize) {
3301 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3302 goto out_fail;
3303 }
3304
3305 /*
3306 * The ext4 superblock will not be buffer aligned for other than 1kB
3307 * block sizes. We need to calculate the offset from buffer start.
3308 */
3309 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3310 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3311 offset = do_div(logical_sb_block, blocksize);
3312 } else {
3313 logical_sb_block = sb_block;
3314 }
3315
3316 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3317 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3318 goto out_fail;
3319 }
3320 /*
3321 * Note: s_es must be initialized as soon as possible because
3322 * some ext4 macro-instructions depend on its value
3323 */
3324 es = (struct ext4_super_block *) (bh->b_data + offset);
3325 sbi->s_es = es;
3326 sb->s_magic = le16_to_cpu(es->s_magic);
3327 if (sb->s_magic != EXT4_SUPER_MAGIC)
3328 goto cantfind_ext4;
3329 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3330
3331 /* Warn if metadata_csum and gdt_csum are both set. */
3332 if (ext4_has_feature_metadata_csum(sb) &&
3333 ext4_has_feature_gdt_csum(sb))
3334 ext4_warning(sb, "metadata_csum and uninit_bg are "
3335 "redundant flags; please run fsck.");
3336
3337 /* Check for a known checksum algorithm */
3338 if (!ext4_verify_csum_type(sb, es)) {
3339 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3340 "unknown checksum algorithm.");
3341 silent = 1;
3342 goto cantfind_ext4;
3343 }
3344
3345 /* Load the checksum driver */
3346 if (ext4_has_feature_metadata_csum(sb)) {
3347 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3348 if (IS_ERR(sbi->s_chksum_driver)) {
3349 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3350 ret = PTR_ERR(sbi->s_chksum_driver);
3351 sbi->s_chksum_driver = NULL;
3352 goto failed_mount;
3353 }
3354 }
3355
3356 /* Check superblock checksum */
3357 if (!ext4_superblock_csum_verify(sb, es)) {
3358 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3359 "invalid superblock checksum. Run e2fsck?");
3360 silent = 1;
3361 ret = -EFSBADCRC;
3362 goto cantfind_ext4;
3363 }
3364
3365 /* Precompute checksum seed for all metadata */
3366 if (ext4_has_feature_csum_seed(sb))
3367 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3368 else if (ext4_has_metadata_csum(sb))
3369 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3370 sizeof(es->s_uuid));
3371
3372 /* Set defaults before we parse the mount options */
3373 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3374 set_opt(sb, INIT_INODE_TABLE);
3375 if (def_mount_opts & EXT4_DEFM_DEBUG)
3376 set_opt(sb, DEBUG);
3377 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3378 set_opt(sb, GRPID);
3379 if (def_mount_opts & EXT4_DEFM_UID16)
3380 set_opt(sb, NO_UID32);
3381 /* xattr user namespace & acls are now defaulted on */
3382 set_opt(sb, XATTR_USER);
3383 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3384 set_opt(sb, POSIX_ACL);
3385 #endif
3386 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3387 if (ext4_has_metadata_csum(sb))
3388 set_opt(sb, JOURNAL_CHECKSUM);
3389
3390 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3391 set_opt(sb, JOURNAL_DATA);
3392 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3393 set_opt(sb, ORDERED_DATA);
3394 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3395 set_opt(sb, WRITEBACK_DATA);
3396
3397 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3398 set_opt(sb, ERRORS_PANIC);
3399 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3400 set_opt(sb, ERRORS_CONT);
3401 else
3402 set_opt(sb, ERRORS_RO);
3403 /* block_validity enabled by default; disable with noblock_validity */
3404 set_opt(sb, BLOCK_VALIDITY);
3405 if (def_mount_opts & EXT4_DEFM_DISCARD)
3406 set_opt(sb, DISCARD);
3407
3408 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3409 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3410 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3411 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3412 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3413
3414 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3415 set_opt(sb, BARRIER);
3416
3417 /*
3418 * enable delayed allocation by default
3419 * Use -o nodelalloc to turn it off
3420 */
3421 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3422 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3423 set_opt(sb, DELALLOC);
3424
3425 /*
3426 * set default s_li_wait_mult for lazyinit, for the case there is
3427 * no mount option specified.
3428 */
3429 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3430
3431 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3432 &journal_devnum, &journal_ioprio, 0)) {
3433 ext4_msg(sb, KERN_WARNING,
3434 "failed to parse options in superblock: %s",
3435 sbi->s_es->s_mount_opts);
3436 }
3437 sbi->s_def_mount_opt = sbi->s_mount_opt;
3438 if (!parse_options((char *) data, sb, &journal_devnum,
3439 &journal_ioprio, 0))
3440 goto failed_mount;
3441
3442 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3443 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3444 "with data=journal disables delayed "
3445 "allocation and O_DIRECT support!\n");
3446 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3447 ext4_msg(sb, KERN_ERR, "can't mount with "
3448 "both data=journal and delalloc");
3449 goto failed_mount;
3450 }
3451 if (test_opt(sb, DIOREAD_NOLOCK)) {
3452 ext4_msg(sb, KERN_ERR, "can't mount with "
3453 "both data=journal and dioread_nolock");
3454 goto failed_mount;
3455 }
3456 if (test_opt(sb, DAX)) {
3457 ext4_msg(sb, KERN_ERR, "can't mount with "
3458 "both data=journal and dax");
3459 goto failed_mount;
3460 }
3461 if (test_opt(sb, DELALLOC))
3462 clear_opt(sb, DELALLOC);
3463 } else {
3464 sb->s_iflags |= SB_I_CGROUPWB;
3465 }
3466
3467 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3468 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3469
3470 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3471 (ext4_has_compat_features(sb) ||
3472 ext4_has_ro_compat_features(sb) ||
3473 ext4_has_incompat_features(sb)))
3474 ext4_msg(sb, KERN_WARNING,
3475 "feature flags set on rev 0 fs, "
3476 "running e2fsck is recommended");
3477
3478 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3479 set_opt2(sb, HURD_COMPAT);
3480 if (ext4_has_feature_64bit(sb)) {
3481 ext4_msg(sb, KERN_ERR,
3482 "The Hurd can't support 64-bit file systems");
3483 goto failed_mount;
3484 }
3485 }
3486
3487 if (IS_EXT2_SB(sb)) {
3488 if (ext2_feature_set_ok(sb))
3489 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3490 "using the ext4 subsystem");
3491 else {
3492 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3493 "to feature incompatibilities");
3494 goto failed_mount;
3495 }
3496 }
3497
3498 if (IS_EXT3_SB(sb)) {
3499 if (ext3_feature_set_ok(sb))
3500 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3501 "using the ext4 subsystem");
3502 else {
3503 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3504 "to feature incompatibilities");
3505 goto failed_mount;
3506 }
3507 }
3508
3509 /*
3510 * Check feature flags regardless of the revision level, since we
3511 * previously didn't change the revision level when setting the flags,
3512 * so there is a chance incompat flags are set on a rev 0 filesystem.
3513 */
3514 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3515 goto failed_mount;
3516
3517 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3518 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3519 blocksize > EXT4_MAX_BLOCK_SIZE) {
3520 ext4_msg(sb, KERN_ERR,
3521 "Unsupported filesystem blocksize %d", blocksize);
3522 goto failed_mount;
3523 }
3524
3525 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3526 ext4_msg(sb, KERN_ERR,
3527 "Number of reserved GDT blocks insanely large: %d",
3528 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3529 goto failed_mount;
3530 }
3531
3532 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3533 err = bdev_dax_supported(sb, blocksize);
3534 if (err)
3535 goto failed_mount;
3536 }
3537
3538 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3539 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3540 es->s_encryption_level);
3541 goto failed_mount;
3542 }
3543
3544 if (sb->s_blocksize != blocksize) {
3545 /* Validate the filesystem blocksize */
3546 if (!sb_set_blocksize(sb, blocksize)) {
3547 ext4_msg(sb, KERN_ERR, "bad block size %d",
3548 blocksize);
3549 goto failed_mount;
3550 }
3551
3552 brelse(bh);
3553 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3554 offset = do_div(logical_sb_block, blocksize);
3555 bh = sb_bread_unmovable(sb, logical_sb_block);
3556 if (!bh) {
3557 ext4_msg(sb, KERN_ERR,
3558 "Can't read superblock on 2nd try");
3559 goto failed_mount;
3560 }
3561 es = (struct ext4_super_block *)(bh->b_data + offset);
3562 sbi->s_es = es;
3563 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3564 ext4_msg(sb, KERN_ERR,
3565 "Magic mismatch, very weird!");
3566 goto failed_mount;
3567 }
3568 }
3569
3570 has_huge_files = ext4_has_feature_huge_file(sb);
3571 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3572 has_huge_files);
3573 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3574
3575 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3576 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3577 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3578 } else {
3579 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3580 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3581 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3582 (!is_power_of_2(sbi->s_inode_size)) ||
3583 (sbi->s_inode_size > blocksize)) {
3584 ext4_msg(sb, KERN_ERR,
3585 "unsupported inode size: %d",
3586 sbi->s_inode_size);
3587 goto failed_mount;
3588 }
3589 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3590 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3591 }
3592
3593 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3594 if (ext4_has_feature_64bit(sb)) {
3595 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3596 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3597 !is_power_of_2(sbi->s_desc_size)) {
3598 ext4_msg(sb, KERN_ERR,
3599 "unsupported descriptor size %lu",
3600 sbi->s_desc_size);
3601 goto failed_mount;
3602 }
3603 } else
3604 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3605
3606 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3607 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3608 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3609 goto cantfind_ext4;
3610
3611 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3612 if (sbi->s_inodes_per_block == 0)
3613 goto cantfind_ext4;
3614 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3615 sbi->s_inodes_per_block;
3616 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3617 sbi->s_sbh = bh;
3618 sbi->s_mount_state = le16_to_cpu(es->s_state);
3619 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3620 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3621
3622 for (i = 0; i < 4; i++)
3623 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3624 sbi->s_def_hash_version = es->s_def_hash_version;
3625 if (ext4_has_feature_dir_index(sb)) {
3626 i = le32_to_cpu(es->s_flags);
3627 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3628 sbi->s_hash_unsigned = 3;
3629 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3630 #ifdef __CHAR_UNSIGNED__
3631 if (!(sb->s_flags & MS_RDONLY))
3632 es->s_flags |=
3633 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3634 sbi->s_hash_unsigned = 3;
3635 #else
3636 if (!(sb->s_flags & MS_RDONLY))
3637 es->s_flags |=
3638 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3639 #endif
3640 }
3641 }
3642
3643 /* Handle clustersize */
3644 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3645 has_bigalloc = ext4_has_feature_bigalloc(sb);
3646 if (has_bigalloc) {
3647 if (clustersize < blocksize) {
3648 ext4_msg(sb, KERN_ERR,
3649 "cluster size (%d) smaller than "
3650 "block size (%d)", clustersize, blocksize);
3651 goto failed_mount;
3652 }
3653 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3654 le32_to_cpu(es->s_log_block_size);
3655 sbi->s_clusters_per_group =
3656 le32_to_cpu(es->s_clusters_per_group);
3657 if (sbi->s_clusters_per_group > blocksize * 8) {
3658 ext4_msg(sb, KERN_ERR,
3659 "#clusters per group too big: %lu",
3660 sbi->s_clusters_per_group);
3661 goto failed_mount;
3662 }
3663 if (sbi->s_blocks_per_group !=
3664 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3665 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3666 "clusters per group (%lu) inconsistent",
3667 sbi->s_blocks_per_group,
3668 sbi->s_clusters_per_group);
3669 goto failed_mount;
3670 }
3671 } else {
3672 if (clustersize != blocksize) {
3673 ext4_warning(sb, "fragment/cluster size (%d) != "
3674 "block size (%d)", clustersize,
3675 blocksize);
3676 clustersize = blocksize;
3677 }
3678 if (sbi->s_blocks_per_group > blocksize * 8) {
3679 ext4_msg(sb, KERN_ERR,
3680 "#blocks per group too big: %lu",
3681 sbi->s_blocks_per_group);
3682 goto failed_mount;
3683 }
3684 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3685 sbi->s_cluster_bits = 0;
3686 }
3687 sbi->s_cluster_ratio = clustersize / blocksize;
3688
3689 if (sbi->s_inodes_per_group > blocksize * 8) {
3690 ext4_msg(sb, KERN_ERR,
3691 "#inodes per group too big: %lu",
3692 sbi->s_inodes_per_group);
3693 goto failed_mount;
3694 }
3695
3696 /* Do we have standard group size of clustersize * 8 blocks ? */
3697 if (sbi->s_blocks_per_group == clustersize << 3)
3698 set_opt2(sb, STD_GROUP_SIZE);
3699
3700 /*
3701 * Test whether we have more sectors than will fit in sector_t,
3702 * and whether the max offset is addressable by the page cache.
3703 */
3704 err = generic_check_addressable(sb->s_blocksize_bits,
3705 ext4_blocks_count(es));
3706 if (err) {
3707 ext4_msg(sb, KERN_ERR, "filesystem"
3708 " too large to mount safely on this system");
3709 if (sizeof(sector_t) < 8)
3710 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3711 goto failed_mount;
3712 }
3713
3714 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3715 goto cantfind_ext4;
3716
3717 /* check blocks count against device size */
3718 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3719 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3720 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3721 "exceeds size of device (%llu blocks)",
3722 ext4_blocks_count(es), blocks_count);
3723 goto failed_mount;
3724 }
3725
3726 /*
3727 * It makes no sense for the first data block to be beyond the end
3728 * of the filesystem.
3729 */
3730 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3731 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3732 "block %u is beyond end of filesystem (%llu)",
3733 le32_to_cpu(es->s_first_data_block),
3734 ext4_blocks_count(es));
3735 goto failed_mount;
3736 }
3737 blocks_count = (ext4_blocks_count(es) -
3738 le32_to_cpu(es->s_first_data_block) +
3739 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3740 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3741 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3742 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3743 "(block count %llu, first data block %u, "
3744 "blocks per group %lu)", sbi->s_groups_count,
3745 ext4_blocks_count(es),
3746 le32_to_cpu(es->s_first_data_block),
3747 EXT4_BLOCKS_PER_GROUP(sb));
3748 goto failed_mount;
3749 }
3750 sbi->s_groups_count = blocks_count;
3751 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3752 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3753 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3754 EXT4_DESC_PER_BLOCK(sb);
3755 sbi->s_group_desc = ext4_kvmalloc(db_count *
3756 sizeof(struct buffer_head *),
3757 GFP_KERNEL);
3758 if (sbi->s_group_desc == NULL) {
3759 ext4_msg(sb, KERN_ERR, "not enough memory");
3760 ret = -ENOMEM;
3761 goto failed_mount;
3762 }
3763
3764 bgl_lock_init(sbi->s_blockgroup_lock);
3765
3766 for (i = 0; i < db_count; i++) {
3767 block = descriptor_loc(sb, logical_sb_block, i);
3768 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3769 if (!sbi->s_group_desc[i]) {
3770 ext4_msg(sb, KERN_ERR,
3771 "can't read group descriptor %d", i);
3772 db_count = i;
3773 goto failed_mount2;
3774 }
3775 }
3776 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3777 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3778 ret = -EFSCORRUPTED;
3779 goto failed_mount2;
3780 }
3781
3782 sbi->s_gdb_count = db_count;
3783 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3784 spin_lock_init(&sbi->s_next_gen_lock);
3785
3786 setup_timer(&sbi->s_err_report, print_daily_error_info,
3787 (unsigned long) sb);
3788
3789 /* Register extent status tree shrinker */
3790 if (ext4_es_register_shrinker(sbi))
3791 goto failed_mount3;
3792
3793 sbi->s_stripe = ext4_get_stripe_size(sbi);
3794 sbi->s_extent_max_zeroout_kb = 32;
3795
3796 /*
3797 * set up enough so that it can read an inode
3798 */
3799 sb->s_op = &ext4_sops;
3800 sb->s_export_op = &ext4_export_ops;
3801 sb->s_xattr = ext4_xattr_handlers;
3802 sb->s_cop = &ext4_cryptops;
3803 #ifdef CONFIG_QUOTA
3804 sb->dq_op = &ext4_quota_operations;
3805 if (ext4_has_feature_quota(sb))
3806 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3807 else
3808 sb->s_qcop = &ext4_qctl_operations;
3809 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3810 #endif
3811 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3812
3813 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3814 mutex_init(&sbi->s_orphan_lock);
3815
3816 sb->s_root = NULL;
3817
3818 needs_recovery = (es->s_last_orphan != 0 ||
3819 ext4_has_feature_journal_needs_recovery(sb));
3820
3821 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3822 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3823 goto failed_mount3a;
3824
3825 /*
3826 * The first inode we look at is the journal inode. Don't try
3827 * root first: it may be modified in the journal!
3828 */
3829 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3830 if (ext4_load_journal(sb, es, journal_devnum))
3831 goto failed_mount3a;
3832 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3833 ext4_has_feature_journal_needs_recovery(sb)) {
3834 ext4_msg(sb, KERN_ERR, "required journal recovery "
3835 "suppressed and not mounted read-only");
3836 goto failed_mount_wq;
3837 } else {
3838 /* Nojournal mode, all journal mount options are illegal */
3839 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3840 ext4_msg(sb, KERN_ERR, "can't mount with "
3841 "journal_checksum, fs mounted w/o journal");
3842 goto failed_mount_wq;
3843 }
3844 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3845 ext4_msg(sb, KERN_ERR, "can't mount with "
3846 "journal_async_commit, fs mounted w/o journal");
3847 goto failed_mount_wq;
3848 }
3849 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3850 ext4_msg(sb, KERN_ERR, "can't mount with "
3851 "commit=%lu, fs mounted w/o journal",
3852 sbi->s_commit_interval / HZ);
3853 goto failed_mount_wq;
3854 }
3855 if (EXT4_MOUNT_DATA_FLAGS &
3856 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3857 ext4_msg(sb, KERN_ERR, "can't mount with "
3858 "data=, fs mounted w/o journal");
3859 goto failed_mount_wq;
3860 }
3861 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3862 clear_opt(sb, JOURNAL_CHECKSUM);
3863 clear_opt(sb, DATA_FLAGS);
3864 sbi->s_journal = NULL;
3865 needs_recovery = 0;
3866 goto no_journal;
3867 }
3868
3869 if (ext4_has_feature_64bit(sb) &&
3870 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3871 JBD2_FEATURE_INCOMPAT_64BIT)) {
3872 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3873 goto failed_mount_wq;
3874 }
3875
3876 if (!set_journal_csum_feature_set(sb)) {
3877 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3878 "feature set");
3879 goto failed_mount_wq;
3880 }
3881
3882 /* We have now updated the journal if required, so we can
3883 * validate the data journaling mode. */
3884 switch (test_opt(sb, DATA_FLAGS)) {
3885 case 0:
3886 /* No mode set, assume a default based on the journal
3887 * capabilities: ORDERED_DATA if the journal can
3888 * cope, else JOURNAL_DATA
3889 */
3890 if (jbd2_journal_check_available_features
3891 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3892 set_opt(sb, ORDERED_DATA);
3893 else
3894 set_opt(sb, JOURNAL_DATA);
3895 break;
3896
3897 case EXT4_MOUNT_ORDERED_DATA:
3898 case EXT4_MOUNT_WRITEBACK_DATA:
3899 if (!jbd2_journal_check_available_features
3900 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3901 ext4_msg(sb, KERN_ERR, "Journal does not support "
3902 "requested data journaling mode");
3903 goto failed_mount_wq;
3904 }
3905 default:
3906 break;
3907 }
3908 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3909
3910 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3911
3912 no_journal:
3913 sbi->s_mb_cache = ext4_xattr_create_cache();
3914 if (!sbi->s_mb_cache) {
3915 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3916 goto failed_mount_wq;
3917 }
3918
3919 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3920 (blocksize != PAGE_SIZE)) {
3921 ext4_msg(sb, KERN_ERR,
3922 "Unsupported blocksize for fs encryption");
3923 goto failed_mount_wq;
3924 }
3925
3926 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3927 !ext4_has_feature_encrypt(sb)) {
3928 ext4_set_feature_encrypt(sb);
3929 ext4_commit_super(sb, 1);
3930 }
3931
3932 /*
3933 * Get the # of file system overhead blocks from the
3934 * superblock if present.
3935 */
3936 if (es->s_overhead_clusters)
3937 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3938 else {
3939 err = ext4_calculate_overhead(sb);
3940 if (err)
3941 goto failed_mount_wq;
3942 }
3943
3944 /*
3945 * The maximum number of concurrent works can be high and
3946 * concurrency isn't really necessary. Limit it to 1.
3947 */
3948 EXT4_SB(sb)->rsv_conversion_wq =
3949 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3950 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3951 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3952 ret = -ENOMEM;
3953 goto failed_mount4;
3954 }
3955
3956 /*
3957 * The jbd2_journal_load will have done any necessary log recovery,
3958 * so we can safely mount the rest of the filesystem now.
3959 */
3960
3961 root = ext4_iget(sb, EXT4_ROOT_INO);
3962 if (IS_ERR(root)) {
3963 ext4_msg(sb, KERN_ERR, "get root inode failed");
3964 ret = PTR_ERR(root);
3965 root = NULL;
3966 goto failed_mount4;
3967 }
3968 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3969 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3970 iput(root);
3971 goto failed_mount4;
3972 }
3973 sb->s_root = d_make_root(root);
3974 if (!sb->s_root) {
3975 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3976 ret = -ENOMEM;
3977 goto failed_mount4;
3978 }
3979
3980 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3981 sb->s_flags |= MS_RDONLY;
3982
3983 /* determine the minimum size of new large inodes, if present */
3984 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3985 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3986 EXT4_GOOD_OLD_INODE_SIZE;
3987 if (ext4_has_feature_extra_isize(sb)) {
3988 if (sbi->s_want_extra_isize <
3989 le16_to_cpu(es->s_want_extra_isize))
3990 sbi->s_want_extra_isize =
3991 le16_to_cpu(es->s_want_extra_isize);
3992 if (sbi->s_want_extra_isize <
3993 le16_to_cpu(es->s_min_extra_isize))
3994 sbi->s_want_extra_isize =
3995 le16_to_cpu(es->s_min_extra_isize);
3996 }
3997 }
3998 /* Check if enough inode space is available */
3999 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4000 sbi->s_inode_size) {
4001 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4002 EXT4_GOOD_OLD_INODE_SIZE;
4003 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4004 "available");
4005 }
4006
4007 ext4_set_resv_clusters(sb);
4008
4009 err = ext4_setup_system_zone(sb);
4010 if (err) {
4011 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4012 "zone (%d)", err);
4013 goto failed_mount4a;
4014 }
4015
4016 ext4_ext_init(sb);
4017 err = ext4_mb_init(sb);
4018 if (err) {
4019 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4020 err);
4021 goto failed_mount5;
4022 }
4023
4024 block = ext4_count_free_clusters(sb);
4025 ext4_free_blocks_count_set(sbi->s_es,
4026 EXT4_C2B(sbi, block));
4027 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4028 GFP_KERNEL);
4029 if (!err) {
4030 unsigned long freei = ext4_count_free_inodes(sb);
4031 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4032 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4033 GFP_KERNEL);
4034 }
4035 if (!err)
4036 err = percpu_counter_init(&sbi->s_dirs_counter,
4037 ext4_count_dirs(sb), GFP_KERNEL);
4038 if (!err)
4039 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4040 GFP_KERNEL);
4041 if (!err)
4042 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4043
4044 if (err) {
4045 ext4_msg(sb, KERN_ERR, "insufficient memory");
4046 goto failed_mount6;
4047 }
4048
4049 if (ext4_has_feature_flex_bg(sb))
4050 if (!ext4_fill_flex_info(sb)) {
4051 ext4_msg(sb, KERN_ERR,
4052 "unable to initialize "
4053 "flex_bg meta info!");
4054 goto failed_mount6;
4055 }
4056
4057 err = ext4_register_li_request(sb, first_not_zeroed);
4058 if (err)
4059 goto failed_mount6;
4060
4061 err = ext4_register_sysfs(sb);
4062 if (err)
4063 goto failed_mount7;
4064
4065 #ifdef CONFIG_QUOTA
4066 /* Enable quota usage during mount. */
4067 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4068 err = ext4_enable_quotas(sb);
4069 if (err)
4070 goto failed_mount8;
4071 }
4072 #endif /* CONFIG_QUOTA */
4073
4074 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4075 ext4_orphan_cleanup(sb, es);
4076 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4077 if (needs_recovery) {
4078 ext4_msg(sb, KERN_INFO, "recovery complete");
4079 ext4_mark_recovery_complete(sb, es);
4080 }
4081 if (EXT4_SB(sb)->s_journal) {
4082 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4083 descr = " journalled data mode";
4084 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4085 descr = " ordered data mode";
4086 else
4087 descr = " writeback data mode";
4088 } else
4089 descr = "out journal";
4090
4091 if (test_opt(sb, DISCARD)) {
4092 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4093 if (!blk_queue_discard(q))
4094 ext4_msg(sb, KERN_WARNING,
4095 "mounting with \"discard\" option, but "
4096 "the device does not support discard");
4097 }
4098
4099 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4100 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4101 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4102 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4103
4104 if (es->s_error_count)
4105 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4106
4107 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4108 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4109 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4110 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4111
4112 kfree(orig_data);
4113 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4114 memcpy(sbi->key_prefix, EXT4_KEY_DESC_PREFIX,
4115 EXT4_KEY_DESC_PREFIX_SIZE);
4116 sbi->key_prefix_size = EXT4_KEY_DESC_PREFIX_SIZE;
4117 #endif
4118 return 0;
4119
4120 cantfind_ext4:
4121 if (!silent)
4122 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4123 goto failed_mount;
4124
4125 #ifdef CONFIG_QUOTA
4126 failed_mount8:
4127 ext4_unregister_sysfs(sb);
4128 #endif
4129 failed_mount7:
4130 ext4_unregister_li_request(sb);
4131 failed_mount6:
4132 ext4_mb_release(sb);
4133 if (sbi->s_flex_groups)
4134 kvfree(sbi->s_flex_groups);
4135 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4136 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4137 percpu_counter_destroy(&sbi->s_dirs_counter);
4138 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4139 failed_mount5:
4140 ext4_ext_release(sb);
4141 ext4_release_system_zone(sb);
4142 failed_mount4a:
4143 dput(sb->s_root);
4144 sb->s_root = NULL;
4145 failed_mount4:
4146 ext4_msg(sb, KERN_ERR, "mount failed");
4147 if (EXT4_SB(sb)->rsv_conversion_wq)
4148 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4149 failed_mount_wq:
4150 if (sbi->s_mb_cache) {
4151 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4152 sbi->s_mb_cache = NULL;
4153 }
4154 if (sbi->s_journal) {
4155 jbd2_journal_destroy(sbi->s_journal);
4156 sbi->s_journal = NULL;
4157 }
4158 failed_mount3a:
4159 ext4_es_unregister_shrinker(sbi);
4160 failed_mount3:
4161 del_timer_sync(&sbi->s_err_report);
4162 if (sbi->s_mmp_tsk)
4163 kthread_stop(sbi->s_mmp_tsk);
4164 failed_mount2:
4165 for (i = 0; i < db_count; i++)
4166 brelse(sbi->s_group_desc[i]);
4167 kvfree(sbi->s_group_desc);
4168 failed_mount:
4169 if (sbi->s_chksum_driver)
4170 crypto_free_shash(sbi->s_chksum_driver);
4171 #ifdef CONFIG_QUOTA
4172 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4173 kfree(sbi->s_qf_names[i]);
4174 #endif
4175 ext4_blkdev_remove(sbi);
4176 brelse(bh);
4177 out_fail:
4178 sb->s_fs_info = NULL;
4179 kfree(sbi->s_blockgroup_lock);
4180 kfree(sbi);
4181 out_free_orig:
4182 kfree(orig_data);
4183 return err ? err : ret;
4184 }
4185
4186 /*
4187 * Setup any per-fs journal parameters now. We'll do this both on
4188 * initial mount, once the journal has been initialised but before we've
4189 * done any recovery; and again on any subsequent remount.
4190 */
4191 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4192 {
4193 struct ext4_sb_info *sbi = EXT4_SB(sb);
4194
4195 journal->j_commit_interval = sbi->s_commit_interval;
4196 journal->j_min_batch_time = sbi->s_min_batch_time;
4197 journal->j_max_batch_time = sbi->s_max_batch_time;
4198
4199 write_lock(&journal->j_state_lock);
4200 if (test_opt(sb, BARRIER))
4201 journal->j_flags |= JBD2_BARRIER;
4202 else
4203 journal->j_flags &= ~JBD2_BARRIER;
4204 if (test_opt(sb, DATA_ERR_ABORT))
4205 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4206 else
4207 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4208 write_unlock(&journal->j_state_lock);
4209 }
4210
4211 static journal_t *ext4_get_journal(struct super_block *sb,
4212 unsigned int journal_inum)
4213 {
4214 struct inode *journal_inode;
4215 journal_t *journal;
4216
4217 BUG_ON(!ext4_has_feature_journal(sb));
4218
4219 /* First, test for the existence of a valid inode on disk. Bad
4220 * things happen if we iget() an unused inode, as the subsequent
4221 * iput() will try to delete it. */
4222
4223 journal_inode = ext4_iget(sb, journal_inum);
4224 if (IS_ERR(journal_inode)) {
4225 ext4_msg(sb, KERN_ERR, "no journal found");
4226 return NULL;
4227 }
4228 if (!journal_inode->i_nlink) {
4229 make_bad_inode(journal_inode);
4230 iput(journal_inode);
4231 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4232 return NULL;
4233 }
4234
4235 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4236 journal_inode, journal_inode->i_size);
4237 if (!S_ISREG(journal_inode->i_mode)) {
4238 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4239 iput(journal_inode);
4240 return NULL;
4241 }
4242
4243 journal = jbd2_journal_init_inode(journal_inode);
4244 if (!journal) {
4245 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4246 iput(journal_inode);
4247 return NULL;
4248 }
4249 journal->j_private = sb;
4250 ext4_init_journal_params(sb, journal);
4251 return journal;
4252 }
4253
4254 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4255 dev_t j_dev)
4256 {
4257 struct buffer_head *bh;
4258 journal_t *journal;
4259 ext4_fsblk_t start;
4260 ext4_fsblk_t len;
4261 int hblock, blocksize;
4262 ext4_fsblk_t sb_block;
4263 unsigned long offset;
4264 struct ext4_super_block *es;
4265 struct block_device *bdev;
4266
4267 BUG_ON(!ext4_has_feature_journal(sb));
4268
4269 bdev = ext4_blkdev_get(j_dev, sb);
4270 if (bdev == NULL)
4271 return NULL;
4272
4273 blocksize = sb->s_blocksize;
4274 hblock = bdev_logical_block_size(bdev);
4275 if (blocksize < hblock) {
4276 ext4_msg(sb, KERN_ERR,
4277 "blocksize too small for journal device");
4278 goto out_bdev;
4279 }
4280
4281 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4282 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4283 set_blocksize(bdev, blocksize);
4284 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4285 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4286 "external journal");
4287 goto out_bdev;
4288 }
4289
4290 es = (struct ext4_super_block *) (bh->b_data + offset);
4291 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4292 !(le32_to_cpu(es->s_feature_incompat) &
4293 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4294 ext4_msg(sb, KERN_ERR, "external journal has "
4295 "bad superblock");
4296 brelse(bh);
4297 goto out_bdev;
4298 }
4299
4300 if ((le32_to_cpu(es->s_feature_ro_compat) &
4301 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4302 es->s_checksum != ext4_superblock_csum(sb, es)) {
4303 ext4_msg(sb, KERN_ERR, "external journal has "
4304 "corrupt superblock");
4305 brelse(bh);
4306 goto out_bdev;
4307 }
4308
4309 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4310 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4311 brelse(bh);
4312 goto out_bdev;
4313 }
4314
4315 len = ext4_blocks_count(es);
4316 start = sb_block + 1;
4317 brelse(bh); /* we're done with the superblock */
4318
4319 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4320 start, len, blocksize);
4321 if (!journal) {
4322 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4323 goto out_bdev;
4324 }
4325 journal->j_private = sb;
4326 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4327 wait_on_buffer(journal->j_sb_buffer);
4328 if (!buffer_uptodate(journal->j_sb_buffer)) {
4329 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4330 goto out_journal;
4331 }
4332 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4333 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4334 "user (unsupported) - %d",
4335 be32_to_cpu(journal->j_superblock->s_nr_users));
4336 goto out_journal;
4337 }
4338 EXT4_SB(sb)->journal_bdev = bdev;
4339 ext4_init_journal_params(sb, journal);
4340 return journal;
4341
4342 out_journal:
4343 jbd2_journal_destroy(journal);
4344 out_bdev:
4345 ext4_blkdev_put(bdev);
4346 return NULL;
4347 }
4348
4349 static int ext4_load_journal(struct super_block *sb,
4350 struct ext4_super_block *es,
4351 unsigned long journal_devnum)
4352 {
4353 journal_t *journal;
4354 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4355 dev_t journal_dev;
4356 int err = 0;
4357 int really_read_only;
4358
4359 BUG_ON(!ext4_has_feature_journal(sb));
4360
4361 if (journal_devnum &&
4362 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4363 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4364 "numbers have changed");
4365 journal_dev = new_decode_dev(journal_devnum);
4366 } else
4367 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4368
4369 really_read_only = bdev_read_only(sb->s_bdev);
4370
4371 /*
4372 * Are we loading a blank journal or performing recovery after a
4373 * crash? For recovery, we need to check in advance whether we
4374 * can get read-write access to the device.
4375 */
4376 if (ext4_has_feature_journal_needs_recovery(sb)) {
4377 if (sb->s_flags & MS_RDONLY) {
4378 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4379 "required on readonly filesystem");
4380 if (really_read_only) {
4381 ext4_msg(sb, KERN_ERR, "write access "
4382 "unavailable, cannot proceed");
4383 return -EROFS;
4384 }
4385 ext4_msg(sb, KERN_INFO, "write access will "
4386 "be enabled during recovery");
4387 }
4388 }
4389
4390 if (journal_inum && journal_dev) {
4391 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4392 "and inode journals!");
4393 return -EINVAL;
4394 }
4395
4396 if (journal_inum) {
4397 if (!(journal = ext4_get_journal(sb, journal_inum)))
4398 return -EINVAL;
4399 } else {
4400 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4401 return -EINVAL;
4402 }
4403
4404 if (!(journal->j_flags & JBD2_BARRIER))
4405 ext4_msg(sb, KERN_INFO, "barriers disabled");
4406
4407 if (!ext4_has_feature_journal_needs_recovery(sb))
4408 err = jbd2_journal_wipe(journal, !really_read_only);
4409 if (!err) {
4410 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4411 if (save)
4412 memcpy(save, ((char *) es) +
4413 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4414 err = jbd2_journal_load(journal);
4415 if (save)
4416 memcpy(((char *) es) + EXT4_S_ERR_START,
4417 save, EXT4_S_ERR_LEN);
4418 kfree(save);
4419 }
4420
4421 if (err) {
4422 ext4_msg(sb, KERN_ERR, "error loading journal");
4423 jbd2_journal_destroy(journal);
4424 return err;
4425 }
4426
4427 EXT4_SB(sb)->s_journal = journal;
4428 ext4_clear_journal_err(sb, es);
4429
4430 if (!really_read_only && journal_devnum &&
4431 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4432 es->s_journal_dev = cpu_to_le32(journal_devnum);
4433
4434 /* Make sure we flush the recovery flag to disk. */
4435 ext4_commit_super(sb, 1);
4436 }
4437
4438 return 0;
4439 }
4440
4441 static int ext4_commit_super(struct super_block *sb, int sync)
4442 {
4443 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4444 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4445 int error = 0;
4446
4447 if (!sbh || block_device_ejected(sb))
4448 return error;
4449 /*
4450 * If the file system is mounted read-only, don't update the
4451 * superblock write time. This avoids updating the superblock
4452 * write time when we are mounting the root file system
4453 * read/only but we need to replay the journal; at that point,
4454 * for people who are east of GMT and who make their clock
4455 * tick in localtime for Windows bug-for-bug compatibility,
4456 * the clock is set in the future, and this will cause e2fsck
4457 * to complain and force a full file system check.
4458 */
4459 if (!(sb->s_flags & MS_RDONLY))
4460 es->s_wtime = cpu_to_le32(get_seconds());
4461 if (sb->s_bdev->bd_part)
4462 es->s_kbytes_written =
4463 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4464 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4465 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4466 else
4467 es->s_kbytes_written =
4468 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4469 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4470 ext4_free_blocks_count_set(es,
4471 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4472 &EXT4_SB(sb)->s_freeclusters_counter)));
4473 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4474 es->s_free_inodes_count =
4475 cpu_to_le32(percpu_counter_sum_positive(
4476 &EXT4_SB(sb)->s_freeinodes_counter));
4477 BUFFER_TRACE(sbh, "marking dirty");
4478 ext4_superblock_csum_set(sb);
4479 lock_buffer(sbh);
4480 if (buffer_write_io_error(sbh)) {
4481 /*
4482 * Oh, dear. A previous attempt to write the
4483 * superblock failed. This could happen because the
4484 * USB device was yanked out. Or it could happen to
4485 * be a transient write error and maybe the block will
4486 * be remapped. Nothing we can do but to retry the
4487 * write and hope for the best.
4488 */
4489 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4490 "superblock detected");
4491 clear_buffer_write_io_error(sbh);
4492 set_buffer_uptodate(sbh);
4493 }
4494 mark_buffer_dirty(sbh);
4495 unlock_buffer(sbh);
4496 if (sync) {
4497 error = __sync_dirty_buffer(sbh,
4498 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4499 if (error)
4500 return error;
4501
4502 error = buffer_write_io_error(sbh);
4503 if (error) {
4504 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4505 "superblock");
4506 clear_buffer_write_io_error(sbh);
4507 set_buffer_uptodate(sbh);
4508 }
4509 }
4510 return error;
4511 }
4512
4513 /*
4514 * Have we just finished recovery? If so, and if we are mounting (or
4515 * remounting) the filesystem readonly, then we will end up with a
4516 * consistent fs on disk. Record that fact.
4517 */
4518 static void ext4_mark_recovery_complete(struct super_block *sb,
4519 struct ext4_super_block *es)
4520 {
4521 journal_t *journal = EXT4_SB(sb)->s_journal;
4522
4523 if (!ext4_has_feature_journal(sb)) {
4524 BUG_ON(journal != NULL);
4525 return;
4526 }
4527 jbd2_journal_lock_updates(journal);
4528 if (jbd2_journal_flush(journal) < 0)
4529 goto out;
4530
4531 if (ext4_has_feature_journal_needs_recovery(sb) &&
4532 sb->s_flags & MS_RDONLY) {
4533 ext4_clear_feature_journal_needs_recovery(sb);
4534 ext4_commit_super(sb, 1);
4535 }
4536
4537 out:
4538 jbd2_journal_unlock_updates(journal);
4539 }
4540
4541 /*
4542 * If we are mounting (or read-write remounting) a filesystem whose journal
4543 * has recorded an error from a previous lifetime, move that error to the
4544 * main filesystem now.
4545 */
4546 static void ext4_clear_journal_err(struct super_block *sb,
4547 struct ext4_super_block *es)
4548 {
4549 journal_t *journal;
4550 int j_errno;
4551 const char *errstr;
4552
4553 BUG_ON(!ext4_has_feature_journal(sb));
4554
4555 journal = EXT4_SB(sb)->s_journal;
4556
4557 /*
4558 * Now check for any error status which may have been recorded in the
4559 * journal by a prior ext4_error() or ext4_abort()
4560 */
4561
4562 j_errno = jbd2_journal_errno(journal);
4563 if (j_errno) {
4564 char nbuf[16];
4565
4566 errstr = ext4_decode_error(sb, j_errno, nbuf);
4567 ext4_warning(sb, "Filesystem error recorded "
4568 "from previous mount: %s", errstr);
4569 ext4_warning(sb, "Marking fs in need of filesystem check.");
4570
4571 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4572 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4573 ext4_commit_super(sb, 1);
4574
4575 jbd2_journal_clear_err(journal);
4576 jbd2_journal_update_sb_errno(journal);
4577 }
4578 }
4579
4580 /*
4581 * Force the running and committing transactions to commit,
4582 * and wait on the commit.
4583 */
4584 int ext4_force_commit(struct super_block *sb)
4585 {
4586 journal_t *journal;
4587
4588 if (sb->s_flags & MS_RDONLY)
4589 return 0;
4590
4591 journal = EXT4_SB(sb)->s_journal;
4592 return ext4_journal_force_commit(journal);
4593 }
4594
4595 static int ext4_sync_fs(struct super_block *sb, int wait)
4596 {
4597 int ret = 0;
4598 tid_t target;
4599 bool needs_barrier = false;
4600 struct ext4_sb_info *sbi = EXT4_SB(sb);
4601
4602 trace_ext4_sync_fs(sb, wait);
4603 flush_workqueue(sbi->rsv_conversion_wq);
4604 /*
4605 * Writeback quota in non-journalled quota case - journalled quota has
4606 * no dirty dquots
4607 */
4608 dquot_writeback_dquots(sb, -1);
4609 /*
4610 * Data writeback is possible w/o journal transaction, so barrier must
4611 * being sent at the end of the function. But we can skip it if
4612 * transaction_commit will do it for us.
4613 */
4614 if (sbi->s_journal) {
4615 target = jbd2_get_latest_transaction(sbi->s_journal);
4616 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4617 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4618 needs_barrier = true;
4619
4620 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4621 if (wait)
4622 ret = jbd2_log_wait_commit(sbi->s_journal,
4623 target);
4624 }
4625 } else if (wait && test_opt(sb, BARRIER))
4626 needs_barrier = true;
4627 if (needs_barrier) {
4628 int err;
4629 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4630 if (!ret)
4631 ret = err;
4632 }
4633
4634 return ret;
4635 }
4636
4637 /*
4638 * LVM calls this function before a (read-only) snapshot is created. This
4639 * gives us a chance to flush the journal completely and mark the fs clean.
4640 *
4641 * Note that only this function cannot bring a filesystem to be in a clean
4642 * state independently. It relies on upper layer to stop all data & metadata
4643 * modifications.
4644 */
4645 static int ext4_freeze(struct super_block *sb)
4646 {
4647 int error = 0;
4648 journal_t *journal;
4649
4650 if (sb->s_flags & MS_RDONLY)
4651 return 0;
4652
4653 journal = EXT4_SB(sb)->s_journal;
4654
4655 if (journal) {
4656 /* Now we set up the journal barrier. */
4657 jbd2_journal_lock_updates(journal);
4658
4659 /*
4660 * Don't clear the needs_recovery flag if we failed to
4661 * flush the journal.
4662 */
4663 error = jbd2_journal_flush(journal);
4664 if (error < 0)
4665 goto out;
4666
4667 /* Journal blocked and flushed, clear needs_recovery flag. */
4668 ext4_clear_feature_journal_needs_recovery(sb);
4669 }
4670
4671 error = ext4_commit_super(sb, 1);
4672 out:
4673 if (journal)
4674 /* we rely on upper layer to stop further updates */
4675 jbd2_journal_unlock_updates(journal);
4676 return error;
4677 }
4678
4679 /*
4680 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4681 * flag here, even though the filesystem is not technically dirty yet.
4682 */
4683 static int ext4_unfreeze(struct super_block *sb)
4684 {
4685 if (sb->s_flags & MS_RDONLY)
4686 return 0;
4687
4688 if (EXT4_SB(sb)->s_journal) {
4689 /* Reset the needs_recovery flag before the fs is unlocked. */
4690 ext4_set_feature_journal_needs_recovery(sb);
4691 }
4692
4693 ext4_commit_super(sb, 1);
4694 return 0;
4695 }
4696
4697 /*
4698 * Structure to save mount options for ext4_remount's benefit
4699 */
4700 struct ext4_mount_options {
4701 unsigned long s_mount_opt;
4702 unsigned long s_mount_opt2;
4703 kuid_t s_resuid;
4704 kgid_t s_resgid;
4705 unsigned long s_commit_interval;
4706 u32 s_min_batch_time, s_max_batch_time;
4707 #ifdef CONFIG_QUOTA
4708 int s_jquota_fmt;
4709 char *s_qf_names[EXT4_MAXQUOTAS];
4710 #endif
4711 };
4712
4713 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4714 {
4715 struct ext4_super_block *es;
4716 struct ext4_sb_info *sbi = EXT4_SB(sb);
4717 unsigned long old_sb_flags;
4718 struct ext4_mount_options old_opts;
4719 int enable_quota = 0;
4720 ext4_group_t g;
4721 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4722 int err = 0;
4723 #ifdef CONFIG_QUOTA
4724 int i, j;
4725 #endif
4726 char *orig_data = kstrdup(data, GFP_KERNEL);
4727
4728 /* Store the original options */
4729 old_sb_flags = sb->s_flags;
4730 old_opts.s_mount_opt = sbi->s_mount_opt;
4731 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4732 old_opts.s_resuid = sbi->s_resuid;
4733 old_opts.s_resgid = sbi->s_resgid;
4734 old_opts.s_commit_interval = sbi->s_commit_interval;
4735 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4736 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4737 #ifdef CONFIG_QUOTA
4738 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4739 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4740 if (sbi->s_qf_names[i]) {
4741 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4742 GFP_KERNEL);
4743 if (!old_opts.s_qf_names[i]) {
4744 for (j = 0; j < i; j++)
4745 kfree(old_opts.s_qf_names[j]);
4746 kfree(orig_data);
4747 return -ENOMEM;
4748 }
4749 } else
4750 old_opts.s_qf_names[i] = NULL;
4751 #endif
4752 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4753 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4754
4755 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4756 err = -EINVAL;
4757 goto restore_opts;
4758 }
4759
4760 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4761 test_opt(sb, JOURNAL_CHECKSUM)) {
4762 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4763 "during remount not supported; ignoring");
4764 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4765 }
4766
4767 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4768 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4769 ext4_msg(sb, KERN_ERR, "can't mount with "
4770 "both data=journal and delalloc");
4771 err = -EINVAL;
4772 goto restore_opts;
4773 }
4774 if (test_opt(sb, DIOREAD_NOLOCK)) {
4775 ext4_msg(sb, KERN_ERR, "can't mount with "
4776 "both data=journal and dioread_nolock");
4777 err = -EINVAL;
4778 goto restore_opts;
4779 }
4780 if (test_opt(sb, DAX)) {
4781 ext4_msg(sb, KERN_ERR, "can't mount with "
4782 "both data=journal and dax");
4783 err = -EINVAL;
4784 goto restore_opts;
4785 }
4786 }
4787
4788 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4789 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4790 "dax flag with busy inodes while remounting");
4791 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4792 }
4793
4794 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4795 ext4_abort(sb, "Abort forced by user");
4796
4797 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4798 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4799
4800 es = sbi->s_es;
4801
4802 if (sbi->s_journal) {
4803 ext4_init_journal_params(sb, sbi->s_journal);
4804 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4805 }
4806
4807 if (*flags & MS_LAZYTIME)
4808 sb->s_flags |= MS_LAZYTIME;
4809
4810 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4811 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4812 err = -EROFS;
4813 goto restore_opts;
4814 }
4815
4816 if (*flags & MS_RDONLY) {
4817 err = sync_filesystem(sb);
4818 if (err < 0)
4819 goto restore_opts;
4820 err = dquot_suspend(sb, -1);
4821 if (err < 0)
4822 goto restore_opts;
4823
4824 /*
4825 * First of all, the unconditional stuff we have to do
4826 * to disable replay of the journal when we next remount
4827 */
4828 sb->s_flags |= MS_RDONLY;
4829
4830 /*
4831 * OK, test if we are remounting a valid rw partition
4832 * readonly, and if so set the rdonly flag and then
4833 * mark the partition as valid again.
4834 */
4835 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4836 (sbi->s_mount_state & EXT4_VALID_FS))
4837 es->s_state = cpu_to_le16(sbi->s_mount_state);
4838
4839 if (sbi->s_journal)
4840 ext4_mark_recovery_complete(sb, es);
4841 } else {
4842 /* Make sure we can mount this feature set readwrite */
4843 if (ext4_has_feature_readonly(sb) ||
4844 !ext4_feature_set_ok(sb, 0)) {
4845 err = -EROFS;
4846 goto restore_opts;
4847 }
4848 /*
4849 * Make sure the group descriptor checksums
4850 * are sane. If they aren't, refuse to remount r/w.
4851 */
4852 for (g = 0; g < sbi->s_groups_count; g++) {
4853 struct ext4_group_desc *gdp =
4854 ext4_get_group_desc(sb, g, NULL);
4855
4856 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4857 ext4_msg(sb, KERN_ERR,
4858 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4859 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4860 le16_to_cpu(gdp->bg_checksum));
4861 err = -EFSBADCRC;
4862 goto restore_opts;
4863 }
4864 }
4865
4866 /*
4867 * If we have an unprocessed orphan list hanging
4868 * around from a previously readonly bdev mount,
4869 * require a full umount/remount for now.
4870 */
4871 if (es->s_last_orphan) {
4872 ext4_msg(sb, KERN_WARNING, "Couldn't "
4873 "remount RDWR because of unprocessed "
4874 "orphan inode list. Please "
4875 "umount/remount instead");
4876 err = -EINVAL;
4877 goto restore_opts;
4878 }
4879
4880 /*
4881 * Mounting a RDONLY partition read-write, so reread
4882 * and store the current valid flag. (It may have
4883 * been changed by e2fsck since we originally mounted
4884 * the partition.)
4885 */
4886 if (sbi->s_journal)
4887 ext4_clear_journal_err(sb, es);
4888 sbi->s_mount_state = le16_to_cpu(es->s_state);
4889 if (!ext4_setup_super(sb, es, 0))
4890 sb->s_flags &= ~MS_RDONLY;
4891 if (ext4_has_feature_mmp(sb))
4892 if (ext4_multi_mount_protect(sb,
4893 le64_to_cpu(es->s_mmp_block))) {
4894 err = -EROFS;
4895 goto restore_opts;
4896 }
4897 enable_quota = 1;
4898 }
4899 }
4900
4901 /*
4902 * Reinitialize lazy itable initialization thread based on
4903 * current settings
4904 */
4905 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4906 ext4_unregister_li_request(sb);
4907 else {
4908 ext4_group_t first_not_zeroed;
4909 first_not_zeroed = ext4_has_uninit_itable(sb);
4910 ext4_register_li_request(sb, first_not_zeroed);
4911 }
4912
4913 ext4_setup_system_zone(sb);
4914 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4915 ext4_commit_super(sb, 1);
4916
4917 #ifdef CONFIG_QUOTA
4918 /* Release old quota file names */
4919 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4920 kfree(old_opts.s_qf_names[i]);
4921 if (enable_quota) {
4922 if (sb_any_quota_suspended(sb))
4923 dquot_resume(sb, -1);
4924 else if (ext4_has_feature_quota(sb)) {
4925 err = ext4_enable_quotas(sb);
4926 if (err)
4927 goto restore_opts;
4928 }
4929 }
4930 #endif
4931
4932 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4933 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4934 kfree(orig_data);
4935 return 0;
4936
4937 restore_opts:
4938 sb->s_flags = old_sb_flags;
4939 sbi->s_mount_opt = old_opts.s_mount_opt;
4940 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4941 sbi->s_resuid = old_opts.s_resuid;
4942 sbi->s_resgid = old_opts.s_resgid;
4943 sbi->s_commit_interval = old_opts.s_commit_interval;
4944 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4945 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4946 #ifdef CONFIG_QUOTA
4947 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4948 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
4949 kfree(sbi->s_qf_names[i]);
4950 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4951 }
4952 #endif
4953 kfree(orig_data);
4954 return err;
4955 }
4956
4957 #ifdef CONFIG_QUOTA
4958 static int ext4_statfs_project(struct super_block *sb,
4959 kprojid_t projid, struct kstatfs *buf)
4960 {
4961 struct kqid qid;
4962 struct dquot *dquot;
4963 u64 limit;
4964 u64 curblock;
4965
4966 qid = make_kqid_projid(projid);
4967 dquot = dqget(sb, qid);
4968 if (IS_ERR(dquot))
4969 return PTR_ERR(dquot);
4970 spin_lock(&dq_data_lock);
4971
4972 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
4973 dquot->dq_dqb.dqb_bsoftlimit :
4974 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
4975 if (limit && buf->f_blocks > limit) {
4976 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
4977 buf->f_blocks = limit;
4978 buf->f_bfree = buf->f_bavail =
4979 (buf->f_blocks > curblock) ?
4980 (buf->f_blocks - curblock) : 0;
4981 }
4982
4983 limit = dquot->dq_dqb.dqb_isoftlimit ?
4984 dquot->dq_dqb.dqb_isoftlimit :
4985 dquot->dq_dqb.dqb_ihardlimit;
4986 if (limit && buf->f_files > limit) {
4987 buf->f_files = limit;
4988 buf->f_ffree =
4989 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
4990 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
4991 }
4992
4993 spin_unlock(&dq_data_lock);
4994 dqput(dquot);
4995 return 0;
4996 }
4997 #endif
4998
4999 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5000 {
5001 struct super_block *sb = dentry->d_sb;
5002 struct ext4_sb_info *sbi = EXT4_SB(sb);
5003 struct ext4_super_block *es = sbi->s_es;
5004 ext4_fsblk_t overhead = 0, resv_blocks;
5005 u64 fsid;
5006 s64 bfree;
5007 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5008
5009 if (!test_opt(sb, MINIX_DF))
5010 overhead = sbi->s_overhead;
5011
5012 buf->f_type = EXT4_SUPER_MAGIC;
5013 buf->f_bsize = sb->s_blocksize;
5014 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5015 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5016 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5017 /* prevent underflow in case that few free space is available */
5018 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5019 buf->f_bavail = buf->f_bfree -
5020 (ext4_r_blocks_count(es) + resv_blocks);
5021 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5022 buf->f_bavail = 0;
5023 buf->f_files = le32_to_cpu(es->s_inodes_count);
5024 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5025 buf->f_namelen = EXT4_NAME_LEN;
5026 fsid = le64_to_cpup((void *)es->s_uuid) ^
5027 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5028 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5029 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5030
5031 #ifdef CONFIG_QUOTA
5032 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5033 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5034 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5035 #endif
5036 return 0;
5037 }
5038
5039 /* Helper function for writing quotas on sync - we need to start transaction
5040 * before quota file is locked for write. Otherwise the are possible deadlocks:
5041 * Process 1 Process 2
5042 * ext4_create() quota_sync()
5043 * jbd2_journal_start() write_dquot()
5044 * dquot_initialize() down(dqio_mutex)
5045 * down(dqio_mutex) jbd2_journal_start()
5046 *
5047 */
5048
5049 #ifdef CONFIG_QUOTA
5050
5051 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5052 {
5053 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5054 }
5055
5056 static int ext4_write_dquot(struct dquot *dquot)
5057 {
5058 int ret, err;
5059 handle_t *handle;
5060 struct inode *inode;
5061
5062 inode = dquot_to_inode(dquot);
5063 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5064 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5065 if (IS_ERR(handle))
5066 return PTR_ERR(handle);
5067 ret = dquot_commit(dquot);
5068 err = ext4_journal_stop(handle);
5069 if (!ret)
5070 ret = err;
5071 return ret;
5072 }
5073
5074 static int ext4_acquire_dquot(struct dquot *dquot)
5075 {
5076 int ret, err;
5077 handle_t *handle;
5078
5079 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5080 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5081 if (IS_ERR(handle))
5082 return PTR_ERR(handle);
5083 ret = dquot_acquire(dquot);
5084 err = ext4_journal_stop(handle);
5085 if (!ret)
5086 ret = err;
5087 return ret;
5088 }
5089
5090 static int ext4_release_dquot(struct dquot *dquot)
5091 {
5092 int ret, err;
5093 handle_t *handle;
5094
5095 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5096 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5097 if (IS_ERR(handle)) {
5098 /* Release dquot anyway to avoid endless cycle in dqput() */
5099 dquot_release(dquot);
5100 return PTR_ERR(handle);
5101 }
5102 ret = dquot_release(dquot);
5103 err = ext4_journal_stop(handle);
5104 if (!ret)
5105 ret = err;
5106 return ret;
5107 }
5108
5109 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5110 {
5111 struct super_block *sb = dquot->dq_sb;
5112 struct ext4_sb_info *sbi = EXT4_SB(sb);
5113
5114 /* Are we journaling quotas? */
5115 if (ext4_has_feature_quota(sb) ||
5116 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5117 dquot_mark_dquot_dirty(dquot);
5118 return ext4_write_dquot(dquot);
5119 } else {
5120 return dquot_mark_dquot_dirty(dquot);
5121 }
5122 }
5123
5124 static int ext4_write_info(struct super_block *sb, int type)
5125 {
5126 int ret, err;
5127 handle_t *handle;
5128
5129 /* Data block + inode block */
5130 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5131 if (IS_ERR(handle))
5132 return PTR_ERR(handle);
5133 ret = dquot_commit_info(sb, type);
5134 err = ext4_journal_stop(handle);
5135 if (!ret)
5136 ret = err;
5137 return ret;
5138 }
5139
5140 /*
5141 * Turn on quotas during mount time - we need to find
5142 * the quota file and such...
5143 */
5144 static int ext4_quota_on_mount(struct super_block *sb, int type)
5145 {
5146 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5147 EXT4_SB(sb)->s_jquota_fmt, type);
5148 }
5149
5150 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5151 {
5152 struct ext4_inode_info *ei = EXT4_I(inode);
5153
5154 /* The first argument of lockdep_set_subclass has to be
5155 * *exactly* the same as the argument to init_rwsem() --- in
5156 * this case, in init_once() --- or lockdep gets unhappy
5157 * because the name of the lock is set using the
5158 * stringification of the argument to init_rwsem().
5159 */
5160 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5161 lockdep_set_subclass(&ei->i_data_sem, subclass);
5162 }
5163
5164 /*
5165 * Standard function to be called on quota_on
5166 */
5167 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5168 struct path *path)
5169 {
5170 int err;
5171
5172 if (!test_opt(sb, QUOTA))
5173 return -EINVAL;
5174
5175 /* Quotafile not on the same filesystem? */
5176 if (path->dentry->d_sb != sb)
5177 return -EXDEV;
5178 /* Journaling quota? */
5179 if (EXT4_SB(sb)->s_qf_names[type]) {
5180 /* Quotafile not in fs root? */
5181 if (path->dentry->d_parent != sb->s_root)
5182 ext4_msg(sb, KERN_WARNING,
5183 "Quota file not on filesystem root. "
5184 "Journaled quota will not work");
5185 }
5186
5187 /*
5188 * When we journal data on quota file, we have to flush journal to see
5189 * all updates to the file when we bypass pagecache...
5190 */
5191 if (EXT4_SB(sb)->s_journal &&
5192 ext4_should_journal_data(d_inode(path->dentry))) {
5193 /*
5194 * We don't need to lock updates but journal_flush() could
5195 * otherwise be livelocked...
5196 */
5197 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5198 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5199 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5200 if (err)
5201 return err;
5202 }
5203 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5204 err = dquot_quota_on(sb, type, format_id, path);
5205 if (err)
5206 lockdep_set_quota_inode(path->dentry->d_inode,
5207 I_DATA_SEM_NORMAL);
5208 return err;
5209 }
5210
5211 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5212 unsigned int flags)
5213 {
5214 int err;
5215 struct inode *qf_inode;
5216 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5217 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5218 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5219 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5220 };
5221
5222 BUG_ON(!ext4_has_feature_quota(sb));
5223
5224 if (!qf_inums[type])
5225 return -EPERM;
5226
5227 qf_inode = ext4_iget(sb, qf_inums[type]);
5228 if (IS_ERR(qf_inode)) {
5229 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5230 return PTR_ERR(qf_inode);
5231 }
5232
5233 /* Don't account quota for quota files to avoid recursion */
5234 qf_inode->i_flags |= S_NOQUOTA;
5235 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5236 err = dquot_enable(qf_inode, type, format_id, flags);
5237 iput(qf_inode);
5238 if (err)
5239 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5240
5241 return err;
5242 }
5243
5244 /* Enable usage tracking for all quota types. */
5245 static int ext4_enable_quotas(struct super_block *sb)
5246 {
5247 int type, err = 0;
5248 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5249 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5250 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5251 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5252 };
5253
5254 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5255 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5256 if (qf_inums[type]) {
5257 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5258 DQUOT_USAGE_ENABLED);
5259 if (err) {
5260 ext4_warning(sb,
5261 "Failed to enable quota tracking "
5262 "(type=%d, err=%d). Please run "
5263 "e2fsck to fix.", type, err);
5264 return err;
5265 }
5266 }
5267 }
5268 return 0;
5269 }
5270
5271 static int ext4_quota_off(struct super_block *sb, int type)
5272 {
5273 struct inode *inode = sb_dqopt(sb)->files[type];
5274 handle_t *handle;
5275
5276 /* Force all delayed allocation blocks to be allocated.
5277 * Caller already holds s_umount sem */
5278 if (test_opt(sb, DELALLOC))
5279 sync_filesystem(sb);
5280
5281 if (!inode)
5282 goto out;
5283
5284 /* Update modification times of quota files when userspace can
5285 * start looking at them */
5286 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5287 if (IS_ERR(handle))
5288 goto out;
5289 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5290 ext4_mark_inode_dirty(handle, inode);
5291 ext4_journal_stop(handle);
5292
5293 out:
5294 return dquot_quota_off(sb, type);
5295 }
5296
5297 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5298 * acquiring the locks... As quota files are never truncated and quota code
5299 * itself serializes the operations (and no one else should touch the files)
5300 * we don't have to be afraid of races */
5301 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5302 size_t len, loff_t off)
5303 {
5304 struct inode *inode = sb_dqopt(sb)->files[type];
5305 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5306 int offset = off & (sb->s_blocksize - 1);
5307 int tocopy;
5308 size_t toread;
5309 struct buffer_head *bh;
5310 loff_t i_size = i_size_read(inode);
5311
5312 if (off > i_size)
5313 return 0;
5314 if (off+len > i_size)
5315 len = i_size-off;
5316 toread = len;
5317 while (toread > 0) {
5318 tocopy = sb->s_blocksize - offset < toread ?
5319 sb->s_blocksize - offset : toread;
5320 bh = ext4_bread(NULL, inode, blk, 0);
5321 if (IS_ERR(bh))
5322 return PTR_ERR(bh);
5323 if (!bh) /* A hole? */
5324 memset(data, 0, tocopy);
5325 else
5326 memcpy(data, bh->b_data+offset, tocopy);
5327 brelse(bh);
5328 offset = 0;
5329 toread -= tocopy;
5330 data += tocopy;
5331 blk++;
5332 }
5333 return len;
5334 }
5335
5336 /* Write to quotafile (we know the transaction is already started and has
5337 * enough credits) */
5338 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5339 const char *data, size_t len, loff_t off)
5340 {
5341 struct inode *inode = sb_dqopt(sb)->files[type];
5342 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5343 int err, offset = off & (sb->s_blocksize - 1);
5344 int retries = 0;
5345 struct buffer_head *bh;
5346 handle_t *handle = journal_current_handle();
5347
5348 if (EXT4_SB(sb)->s_journal && !handle) {
5349 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5350 " cancelled because transaction is not started",
5351 (unsigned long long)off, (unsigned long long)len);
5352 return -EIO;
5353 }
5354 /*
5355 * Since we account only one data block in transaction credits,
5356 * then it is impossible to cross a block boundary.
5357 */
5358 if (sb->s_blocksize - offset < len) {
5359 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5360 " cancelled because not block aligned",
5361 (unsigned long long)off, (unsigned long long)len);
5362 return -EIO;
5363 }
5364
5365 do {
5366 bh = ext4_bread(handle, inode, blk,
5367 EXT4_GET_BLOCKS_CREATE |
5368 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5369 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5370 ext4_should_retry_alloc(inode->i_sb, &retries));
5371 if (IS_ERR(bh))
5372 return PTR_ERR(bh);
5373 if (!bh)
5374 goto out;
5375 BUFFER_TRACE(bh, "get write access");
5376 err = ext4_journal_get_write_access(handle, bh);
5377 if (err) {
5378 brelse(bh);
5379 return err;
5380 }
5381 lock_buffer(bh);
5382 memcpy(bh->b_data+offset, data, len);
5383 flush_dcache_page(bh->b_page);
5384 unlock_buffer(bh);
5385 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5386 brelse(bh);
5387 out:
5388 if (inode->i_size < off + len) {
5389 i_size_write(inode, off + len);
5390 EXT4_I(inode)->i_disksize = inode->i_size;
5391 ext4_mark_inode_dirty(handle, inode);
5392 }
5393 return len;
5394 }
5395
5396 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5397 {
5398 const struct quota_format_ops *ops;
5399
5400 if (!sb_has_quota_loaded(sb, qid->type))
5401 return -ESRCH;
5402 ops = sb_dqopt(sb)->ops[qid->type];
5403 if (!ops || !ops->get_next_id)
5404 return -ENOSYS;
5405 return dquot_get_next_id(sb, qid);
5406 }
5407 #endif
5408
5409 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5410 const char *dev_name, void *data)
5411 {
5412 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5413 }
5414
5415 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5416 static inline void register_as_ext2(void)
5417 {
5418 int err = register_filesystem(&ext2_fs_type);
5419 if (err)
5420 printk(KERN_WARNING
5421 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5422 }
5423
5424 static inline void unregister_as_ext2(void)
5425 {
5426 unregister_filesystem(&ext2_fs_type);
5427 }
5428
5429 static inline int ext2_feature_set_ok(struct super_block *sb)
5430 {
5431 if (ext4_has_unknown_ext2_incompat_features(sb))
5432 return 0;
5433 if (sb->s_flags & MS_RDONLY)
5434 return 1;
5435 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5436 return 0;
5437 return 1;
5438 }
5439 #else
5440 static inline void register_as_ext2(void) { }
5441 static inline void unregister_as_ext2(void) { }
5442 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5443 #endif
5444
5445 static inline void register_as_ext3(void)
5446 {
5447 int err = register_filesystem(&ext3_fs_type);
5448 if (err)
5449 printk(KERN_WARNING
5450 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5451 }
5452
5453 static inline void unregister_as_ext3(void)
5454 {
5455 unregister_filesystem(&ext3_fs_type);
5456 }
5457
5458 static inline int ext3_feature_set_ok(struct super_block *sb)
5459 {
5460 if (ext4_has_unknown_ext3_incompat_features(sb))
5461 return 0;
5462 if (!ext4_has_feature_journal(sb))
5463 return 0;
5464 if (sb->s_flags & MS_RDONLY)
5465 return 1;
5466 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5467 return 0;
5468 return 1;
5469 }
5470
5471 static struct file_system_type ext4_fs_type = {
5472 .owner = THIS_MODULE,
5473 .name = "ext4",
5474 .mount = ext4_mount,
5475 .kill_sb = kill_block_super,
5476 .fs_flags = FS_REQUIRES_DEV,
5477 };
5478 MODULE_ALIAS_FS("ext4");
5479
5480 /* Shared across all ext4 file systems */
5481 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5482
5483 static int __init ext4_init_fs(void)
5484 {
5485 int i, err;
5486
5487 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5488 ext4_li_info = NULL;
5489 mutex_init(&ext4_li_mtx);
5490
5491 /* Build-time check for flags consistency */
5492 ext4_check_flag_values();
5493
5494 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5495 init_waitqueue_head(&ext4__ioend_wq[i]);
5496
5497 err = ext4_init_es();
5498 if (err)
5499 return err;
5500
5501 err = ext4_init_pageio();
5502 if (err)
5503 goto out5;
5504
5505 err = ext4_init_system_zone();
5506 if (err)
5507 goto out4;
5508
5509 err = ext4_init_sysfs();
5510 if (err)
5511 goto out3;
5512
5513 err = ext4_init_mballoc();
5514 if (err)
5515 goto out2;
5516 err = init_inodecache();
5517 if (err)
5518 goto out1;
5519 register_as_ext3();
5520 register_as_ext2();
5521 err = register_filesystem(&ext4_fs_type);
5522 if (err)
5523 goto out;
5524
5525 return 0;
5526 out:
5527 unregister_as_ext2();
5528 unregister_as_ext3();
5529 destroy_inodecache();
5530 out1:
5531 ext4_exit_mballoc();
5532 out2:
5533 ext4_exit_sysfs();
5534 out3:
5535 ext4_exit_system_zone();
5536 out4:
5537 ext4_exit_pageio();
5538 out5:
5539 ext4_exit_es();
5540
5541 return err;
5542 }
5543
5544 static void __exit ext4_exit_fs(void)
5545 {
5546 ext4_destroy_lazyinit_thread();
5547 unregister_as_ext2();
5548 unregister_as_ext3();
5549 unregister_filesystem(&ext4_fs_type);
5550 destroy_inodecache();
5551 ext4_exit_mballoc();
5552 ext4_exit_sysfs();
5553 ext4_exit_system_zone();
5554 ext4_exit_pageio();
5555 ext4_exit_es();
5556 }
5557
5558 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5559 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5560 MODULE_LICENSE("GPL");
5561 module_init(ext4_init_fs)
5562 module_exit(ext4_exit_fs)
Linux with added FPC-III support