]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - fs/ext4/super.c
projects / mirror_ubuntu-artful-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge remote-tracking branch 'mkp-scsi/4.8/scsi-fixes' into fixes
[mirror_ubuntu-artful-kernel.git] / fs / ext4 / super.c
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/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_group_t *first_not_zeroed)
2215 {
2216 struct ext4_sb_info *sbi = EXT4_SB(sb);
2217 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2218 ext4_fsblk_t last_block;
2219 ext4_fsblk_t block_bitmap;
2220 ext4_fsblk_t inode_bitmap;
2221 ext4_fsblk_t inode_table;
2222 int flexbg_flag = 0;
2223 ext4_group_t i, grp = sbi->s_groups_count;
2224
2225 if (ext4_has_feature_flex_bg(sb))
2226 flexbg_flag = 1;
2227
2228 ext4_debug("Checking group descriptors");
2229
2230 for (i = 0; i < sbi->s_groups_count; i++) {
2231 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2232
2233 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2234 last_block = ext4_blocks_count(sbi->s_es) - 1;
2235 else
2236 last_block = first_block +
2237 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2238
2239 if ((grp == sbi->s_groups_count) &&
2240 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2241 grp = i;
2242
2243 block_bitmap = ext4_block_bitmap(sb, gdp);
2244 if (block_bitmap < first_block || block_bitmap > last_block) {
2245 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2246 "Block bitmap for group %u not in group "
2247 "(block %llu)!", i, block_bitmap);
2248 return 0;
2249 }
2250 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2251 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2252 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2253 "Inode bitmap for group %u not in group "
2254 "(block %llu)!", i, inode_bitmap);
2255 return 0;
2256 }
2257 inode_table = ext4_inode_table(sb, gdp);
2258 if (inode_table < first_block ||
2259 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2260 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2261 "Inode table for group %u not in group "
2262 "(block %llu)!", i, inode_table);
2263 return 0;
2264 }
2265 ext4_lock_group(sb, i);
2266 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2267 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2268 "Checksum for group %u failed (%u!=%u)",
2269 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2270 gdp)), le16_to_cpu(gdp->bg_checksum));
2271 if (!(sb->s_flags & MS_RDONLY)) {
2272 ext4_unlock_group(sb, i);
2273 return 0;
2274 }
2275 }
2276 ext4_unlock_group(sb, i);
2277 if (!flexbg_flag)
2278 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2279 }
2280 if (NULL != first_not_zeroed)
2281 *first_not_zeroed = grp;
2282 return 1;
2283 }
2284
2285 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2286 * the superblock) which were deleted from all directories, but held open by
2287 * a process at the time of a crash. We walk the list and try to delete these
2288 * inodes at recovery time (only with a read-write filesystem).
2289 *
2290 * In order to keep the orphan inode chain consistent during traversal (in
2291 * case of crash during recovery), we link each inode into the superblock
2292 * orphan list_head and handle it the same way as an inode deletion during
2293 * normal operation (which journals the operations for us).
2294 *
2295 * We only do an iget() and an iput() on each inode, which is very safe if we
2296 * accidentally point at an in-use or already deleted inode. The worst that
2297 * can happen in this case is that we get a "bit already cleared" message from
2298 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2299 * e2fsck was run on this filesystem, and it must have already done the orphan
2300 * inode cleanup for us, so we can safely abort without any further action.
2301 */
2302 static void ext4_orphan_cleanup(struct super_block *sb,
2303 struct ext4_super_block *es)
2304 {
2305 unsigned int s_flags = sb->s_flags;
2306 int nr_orphans = 0, nr_truncates = 0;
2307 #ifdef CONFIG_QUOTA
2308 int i;
2309 #endif
2310 if (!es->s_last_orphan) {
2311 jbd_debug(4, "no orphan inodes to clean up\n");
2312 return;
2313 }
2314
2315 if (bdev_read_only(sb->s_bdev)) {
2316 ext4_msg(sb, KERN_ERR, "write access "
2317 "unavailable, skipping orphan cleanup");
2318 return;
2319 }
2320
2321 /* Check if feature set would not allow a r/w mount */
2322 if (!ext4_feature_set_ok(sb, 0)) {
2323 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2324 "unknown ROCOMPAT features");
2325 return;
2326 }
2327
2328 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2329 /* don't clear list on RO mount w/ errors */
2330 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2331 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2332 "clearing orphan list.\n");
2333 es->s_last_orphan = 0;
2334 }
2335 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2336 return;
2337 }
2338
2339 if (s_flags & MS_RDONLY) {
2340 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2341 sb->s_flags &= ~MS_RDONLY;
2342 }
2343 #ifdef CONFIG_QUOTA
2344 /* Needed for iput() to work correctly and not trash data */
2345 sb->s_flags |= MS_ACTIVE;
2346 /* Turn on quotas so that they are updated correctly */
2347 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2348 if (EXT4_SB(sb)->s_qf_names[i]) {
2349 int ret = ext4_quota_on_mount(sb, i);
2350 if (ret < 0)
2351 ext4_msg(sb, KERN_ERR,
2352 "Cannot turn on journaled "
2353 "quota: error %d", ret);
2354 }
2355 }
2356 #endif
2357
2358 while (es->s_last_orphan) {
2359 struct inode *inode;
2360
2361 /*
2362 * We may have encountered an error during cleanup; if
2363 * so, skip the rest.
2364 */
2365 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2366 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2367 es->s_last_orphan = 0;
2368 break;
2369 }
2370
2371 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2372 if (IS_ERR(inode)) {
2373 es->s_last_orphan = 0;
2374 break;
2375 }
2376
2377 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2378 dquot_initialize(inode);
2379 if (inode->i_nlink) {
2380 if (test_opt(sb, DEBUG))
2381 ext4_msg(sb, KERN_DEBUG,
2382 "%s: truncating inode %lu to %lld bytes",
2383 __func__, inode->i_ino, inode->i_size);
2384 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2385 inode->i_ino, inode->i_size);
2386 inode_lock(inode);
2387 truncate_inode_pages(inode->i_mapping, inode->i_size);
2388 ext4_truncate(inode);
2389 inode_unlock(inode);
2390 nr_truncates++;
2391 } else {
2392 if (test_opt(sb, DEBUG))
2393 ext4_msg(sb, KERN_DEBUG,
2394 "%s: deleting unreferenced inode %lu",
2395 __func__, inode->i_ino);
2396 jbd_debug(2, "deleting unreferenced inode %lu\n",
2397 inode->i_ino);
2398 nr_orphans++;
2399 }
2400 iput(inode); /* The delete magic happens here! */
2401 }
2402
2403 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2404
2405 if (nr_orphans)
2406 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2407 PLURAL(nr_orphans));
2408 if (nr_truncates)
2409 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2410 PLURAL(nr_truncates));
2411 #ifdef CONFIG_QUOTA
2412 /* Turn quotas off */
2413 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2414 if (sb_dqopt(sb)->files[i])
2415 dquot_quota_off(sb, i);
2416 }
2417 #endif
2418 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2419 }
2420
2421 /*
2422 * Maximal extent format file size.
2423 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2424 * extent format containers, within a sector_t, and within i_blocks
2425 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2426 * so that won't be a limiting factor.
2427 *
2428 * However there is other limiting factor. We do store extents in the form
2429 * of starting block and length, hence the resulting length of the extent
2430 * covering maximum file size must fit into on-disk format containers as
2431 * well. Given that length is always by 1 unit bigger than max unit (because
2432 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2433 *
2434 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2435 */
2436 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2437 {
2438 loff_t res;
2439 loff_t upper_limit = MAX_LFS_FILESIZE;
2440
2441 /* small i_blocks in vfs inode? */
2442 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2443 /*
2444 * CONFIG_LBDAF is not enabled implies the inode
2445 * i_block represent total blocks in 512 bytes
2446 * 32 == size of vfs inode i_blocks * 8
2447 */
2448 upper_limit = (1LL << 32) - 1;
2449
2450 /* total blocks in file system block size */
2451 upper_limit >>= (blkbits - 9);
2452 upper_limit <<= blkbits;
2453 }
2454
2455 /*
2456 * 32-bit extent-start container, ee_block. We lower the maxbytes
2457 * by one fs block, so ee_len can cover the extent of maximum file
2458 * size
2459 */
2460 res = (1LL << 32) - 1;
2461 res <<= blkbits;
2462
2463 /* Sanity check against vm- & vfs- imposed limits */
2464 if (res > upper_limit)
2465 res = upper_limit;
2466
2467 return res;
2468 }
2469
2470 /*
2471 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2472 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2473 * We need to be 1 filesystem block less than the 2^48 sector limit.
2474 */
2475 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2476 {
2477 loff_t res = EXT4_NDIR_BLOCKS;
2478 int meta_blocks;
2479 loff_t upper_limit;
2480 /* This is calculated to be the largest file size for a dense, block
2481 * mapped file such that the file's total number of 512-byte sectors,
2482 * including data and all indirect blocks, does not exceed (2^48 - 1).
2483 *
2484 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2485 * number of 512-byte sectors of the file.
2486 */
2487
2488 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2489 /*
2490 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2491 * the inode i_block field represents total file blocks in
2492 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2493 */
2494 upper_limit = (1LL << 32) - 1;
2495
2496 /* total blocks in file system block size */
2497 upper_limit >>= (bits - 9);
2498
2499 } else {
2500 /*
2501 * We use 48 bit ext4_inode i_blocks
2502 * With EXT4_HUGE_FILE_FL set the i_blocks
2503 * represent total number of blocks in
2504 * file system block size
2505 */
2506 upper_limit = (1LL << 48) - 1;
2507
2508 }
2509
2510 /* indirect blocks */
2511 meta_blocks = 1;
2512 /* double indirect blocks */
2513 meta_blocks += 1 + (1LL << (bits-2));
2514 /* tripple indirect blocks */
2515 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2516
2517 upper_limit -= meta_blocks;
2518 upper_limit <<= bits;
2519
2520 res += 1LL << (bits-2);
2521 res += 1LL << (2*(bits-2));
2522 res += 1LL << (3*(bits-2));
2523 res <<= bits;
2524 if (res > upper_limit)
2525 res = upper_limit;
2526
2527 if (res > MAX_LFS_FILESIZE)
2528 res = MAX_LFS_FILESIZE;
2529
2530 return res;
2531 }
2532
2533 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2534 ext4_fsblk_t logical_sb_block, int nr)
2535 {
2536 struct ext4_sb_info *sbi = EXT4_SB(sb);
2537 ext4_group_t bg, first_meta_bg;
2538 int has_super = 0;
2539
2540 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2541
2542 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2543 return logical_sb_block + nr + 1;
2544 bg = sbi->s_desc_per_block * nr;
2545 if (ext4_bg_has_super(sb, bg))
2546 has_super = 1;
2547
2548 /*
2549 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2550 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2551 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2552 * compensate.
2553 */
2554 if (sb->s_blocksize == 1024 && nr == 0 &&
2555 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2556 has_super++;
2557
2558 return (has_super + ext4_group_first_block_no(sb, bg));
2559 }
2560
2561 /**
2562 * ext4_get_stripe_size: Get the stripe size.
2563 * @sbi: In memory super block info
2564 *
2565 * If we have specified it via mount option, then
2566 * use the mount option value. If the value specified at mount time is
2567 * greater than the blocks per group use the super block value.
2568 * If the super block value is greater than blocks per group return 0.
2569 * Allocator needs it be less than blocks per group.
2570 *
2571 */
2572 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2573 {
2574 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2575 unsigned long stripe_width =
2576 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2577 int ret;
2578
2579 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2580 ret = sbi->s_stripe;
2581 else if (stripe_width <= sbi->s_blocks_per_group)
2582 ret = stripe_width;
2583 else if (stride <= sbi->s_blocks_per_group)
2584 ret = stride;
2585 else
2586 ret = 0;
2587
2588 /*
2589 * If the stripe width is 1, this makes no sense and
2590 * we set it to 0 to turn off stripe handling code.
2591 */
2592 if (ret <= 1)
2593 ret = 0;
2594
2595 return ret;
2596 }
2597
2598 /*
2599 * Check whether this filesystem can be mounted based on
2600 * the features present and the RDONLY/RDWR mount requested.
2601 * Returns 1 if this filesystem can be mounted as requested,
2602 * 0 if it cannot be.
2603 */
2604 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2605 {
2606 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2607 ext4_msg(sb, KERN_ERR,
2608 "Couldn't mount because of "
2609 "unsupported optional features (%x)",
2610 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2611 ~EXT4_FEATURE_INCOMPAT_SUPP));
2612 return 0;
2613 }
2614
2615 if (readonly)
2616 return 1;
2617
2618 if (ext4_has_feature_readonly(sb)) {
2619 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2620 sb->s_flags |= MS_RDONLY;
2621 return 1;
2622 }
2623
2624 /* Check that feature set is OK for a read-write mount */
2625 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2626 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2627 "unsupported optional features (%x)",
2628 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2629 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2630 return 0;
2631 }
2632 /*
2633 * Large file size enabled file system can only be mounted
2634 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2635 */
2636 if (ext4_has_feature_huge_file(sb)) {
2637 if (sizeof(blkcnt_t) < sizeof(u64)) {
2638 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2639 "cannot be mounted RDWR without "
2640 "CONFIG_LBDAF");
2641 return 0;
2642 }
2643 }
2644 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2645 ext4_msg(sb, KERN_ERR,
2646 "Can't support bigalloc feature without "
2647 "extents feature\n");
2648 return 0;
2649 }
2650
2651 #ifndef CONFIG_QUOTA
2652 if (ext4_has_feature_quota(sb) && !readonly) {
2653 ext4_msg(sb, KERN_ERR,
2654 "Filesystem with quota feature cannot be mounted RDWR "
2655 "without CONFIG_QUOTA");
2656 return 0;
2657 }
2658 if (ext4_has_feature_project(sb) && !readonly) {
2659 ext4_msg(sb, KERN_ERR,
2660 "Filesystem with project quota feature cannot be mounted RDWR "
2661 "without CONFIG_QUOTA");
2662 return 0;
2663 }
2664 #endif /* CONFIG_QUOTA */
2665 return 1;
2666 }
2667
2668 /*
2669 * This function is called once a day if we have errors logged
2670 * on the file system
2671 */
2672 static void print_daily_error_info(unsigned long arg)
2673 {
2674 struct super_block *sb = (struct super_block *) arg;
2675 struct ext4_sb_info *sbi;
2676 struct ext4_super_block *es;
2677
2678 sbi = EXT4_SB(sb);
2679 es = sbi->s_es;
2680
2681 if (es->s_error_count)
2682 /* fsck newer than v1.41.13 is needed to clean this condition. */
2683 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2684 le32_to_cpu(es->s_error_count));
2685 if (es->s_first_error_time) {
2686 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2687 sb->s_id, le32_to_cpu(es->s_first_error_time),
2688 (int) sizeof(es->s_first_error_func),
2689 es->s_first_error_func,
2690 le32_to_cpu(es->s_first_error_line));
2691 if (es->s_first_error_ino)
2692 printk(": inode %u",
2693 le32_to_cpu(es->s_first_error_ino));
2694 if (es->s_first_error_block)
2695 printk(": block %llu", (unsigned long long)
2696 le64_to_cpu(es->s_first_error_block));
2697 printk("\n");
2698 }
2699 if (es->s_last_error_time) {
2700 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2701 sb->s_id, le32_to_cpu(es->s_last_error_time),
2702 (int) sizeof(es->s_last_error_func),
2703 es->s_last_error_func,
2704 le32_to_cpu(es->s_last_error_line));
2705 if (es->s_last_error_ino)
2706 printk(": inode %u",
2707 le32_to_cpu(es->s_last_error_ino));
2708 if (es->s_last_error_block)
2709 printk(": block %llu", (unsigned long long)
2710 le64_to_cpu(es->s_last_error_block));
2711 printk("\n");
2712 }
2713 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2714 }
2715
2716 /* Find next suitable group and run ext4_init_inode_table */
2717 static int ext4_run_li_request(struct ext4_li_request *elr)
2718 {
2719 struct ext4_group_desc *gdp = NULL;
2720 ext4_group_t group, ngroups;
2721 struct super_block *sb;
2722 unsigned long timeout = 0;
2723 int ret = 0;
2724
2725 sb = elr->lr_super;
2726 ngroups = EXT4_SB(sb)->s_groups_count;
2727
2728 sb_start_write(sb);
2729 for (group = elr->lr_next_group; group < ngroups; group++) {
2730 gdp = ext4_get_group_desc(sb, group, NULL);
2731 if (!gdp) {
2732 ret = 1;
2733 break;
2734 }
2735
2736 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2737 break;
2738 }
2739
2740 if (group >= ngroups)
2741 ret = 1;
2742
2743 if (!ret) {
2744 timeout = jiffies;
2745 ret = ext4_init_inode_table(sb, group,
2746 elr->lr_timeout ? 0 : 1);
2747 if (elr->lr_timeout == 0) {
2748 timeout = (jiffies - timeout) *
2749 elr->lr_sbi->s_li_wait_mult;
2750 elr->lr_timeout = timeout;
2751 }
2752 elr->lr_next_sched = jiffies + elr->lr_timeout;
2753 elr->lr_next_group = group + 1;
2754 }
2755 sb_end_write(sb);
2756
2757 return ret;
2758 }
2759
2760 /*
2761 * Remove lr_request from the list_request and free the
2762 * request structure. Should be called with li_list_mtx held
2763 */
2764 static void ext4_remove_li_request(struct ext4_li_request *elr)
2765 {
2766 struct ext4_sb_info *sbi;
2767
2768 if (!elr)
2769 return;
2770
2771 sbi = elr->lr_sbi;
2772
2773 list_del(&elr->lr_request);
2774 sbi->s_li_request = NULL;
2775 kfree(elr);
2776 }
2777
2778 static void ext4_unregister_li_request(struct super_block *sb)
2779 {
2780 mutex_lock(&ext4_li_mtx);
2781 if (!ext4_li_info) {
2782 mutex_unlock(&ext4_li_mtx);
2783 return;
2784 }
2785
2786 mutex_lock(&ext4_li_info->li_list_mtx);
2787 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2788 mutex_unlock(&ext4_li_info->li_list_mtx);
2789 mutex_unlock(&ext4_li_mtx);
2790 }
2791
2792 static struct task_struct *ext4_lazyinit_task;
2793
2794 /*
2795 * This is the function where ext4lazyinit thread lives. It walks
2796 * through the request list searching for next scheduled filesystem.
2797 * When such a fs is found, run the lazy initialization request
2798 * (ext4_rn_li_request) and keep track of the time spend in this
2799 * function. Based on that time we compute next schedule time of
2800 * the request. When walking through the list is complete, compute
2801 * next waking time and put itself into sleep.
2802 */
2803 static int ext4_lazyinit_thread(void *arg)
2804 {
2805 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2806 struct list_head *pos, *n;
2807 struct ext4_li_request *elr;
2808 unsigned long next_wakeup, cur;
2809
2810 BUG_ON(NULL == eli);
2811
2812 cont_thread:
2813 while (true) {
2814 next_wakeup = MAX_JIFFY_OFFSET;
2815
2816 mutex_lock(&eli->li_list_mtx);
2817 if (list_empty(&eli->li_request_list)) {
2818 mutex_unlock(&eli->li_list_mtx);
2819 goto exit_thread;
2820 }
2821
2822 list_for_each_safe(pos, n, &eli->li_request_list) {
2823 elr = list_entry(pos, struct ext4_li_request,
2824 lr_request);
2825
2826 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2827 if (ext4_run_li_request(elr) != 0) {
2828 /* error, remove the lazy_init job */
2829 ext4_remove_li_request(elr);
2830 continue;
2831 }
2832 }
2833
2834 if (time_before(elr->lr_next_sched, next_wakeup))
2835 next_wakeup = elr->lr_next_sched;
2836 }
2837 mutex_unlock(&eli->li_list_mtx);
2838
2839 try_to_freeze();
2840
2841 cur = jiffies;
2842 if ((time_after_eq(cur, next_wakeup)) ||
2843 (MAX_JIFFY_OFFSET == next_wakeup)) {
2844 cond_resched();
2845 continue;
2846 }
2847
2848 schedule_timeout_interruptible(next_wakeup - cur);
2849
2850 if (kthread_should_stop()) {
2851 ext4_clear_request_list();
2852 goto exit_thread;
2853 }
2854 }
2855
2856 exit_thread:
2857 /*
2858 * It looks like the request list is empty, but we need
2859 * to check it under the li_list_mtx lock, to prevent any
2860 * additions into it, and of course we should lock ext4_li_mtx
2861 * to atomically free the list and ext4_li_info, because at
2862 * this point another ext4 filesystem could be registering
2863 * new one.
2864 */
2865 mutex_lock(&ext4_li_mtx);
2866 mutex_lock(&eli->li_list_mtx);
2867 if (!list_empty(&eli->li_request_list)) {
2868 mutex_unlock(&eli->li_list_mtx);
2869 mutex_unlock(&ext4_li_mtx);
2870 goto cont_thread;
2871 }
2872 mutex_unlock(&eli->li_list_mtx);
2873 kfree(ext4_li_info);
2874 ext4_li_info = NULL;
2875 mutex_unlock(&ext4_li_mtx);
2876
2877 return 0;
2878 }
2879
2880 static void ext4_clear_request_list(void)
2881 {
2882 struct list_head *pos, *n;
2883 struct ext4_li_request *elr;
2884
2885 mutex_lock(&ext4_li_info->li_list_mtx);
2886 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2887 elr = list_entry(pos, struct ext4_li_request,
2888 lr_request);
2889 ext4_remove_li_request(elr);
2890 }
2891 mutex_unlock(&ext4_li_info->li_list_mtx);
2892 }
2893
2894 static int ext4_run_lazyinit_thread(void)
2895 {
2896 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2897 ext4_li_info, "ext4lazyinit");
2898 if (IS_ERR(ext4_lazyinit_task)) {
2899 int err = PTR_ERR(ext4_lazyinit_task);
2900 ext4_clear_request_list();
2901 kfree(ext4_li_info);
2902 ext4_li_info = NULL;
2903 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2904 "initialization thread\n",
2905 err);
2906 return err;
2907 }
2908 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2909 return 0;
2910 }
2911
2912 /*
2913 * Check whether it make sense to run itable init. thread or not.
2914 * If there is at least one uninitialized inode table, return
2915 * corresponding group number, else the loop goes through all
2916 * groups and return total number of groups.
2917 */
2918 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2919 {
2920 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2921 struct ext4_group_desc *gdp = NULL;
2922
2923 for (group = 0; group < ngroups; group++) {
2924 gdp = ext4_get_group_desc(sb, group, NULL);
2925 if (!gdp)
2926 continue;
2927
2928 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2929 break;
2930 }
2931
2932 return group;
2933 }
2934
2935 static int ext4_li_info_new(void)
2936 {
2937 struct ext4_lazy_init *eli = NULL;
2938
2939 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2940 if (!eli)
2941 return -ENOMEM;
2942
2943 INIT_LIST_HEAD(&eli->li_request_list);
2944 mutex_init(&eli->li_list_mtx);
2945
2946 eli->li_state |= EXT4_LAZYINIT_QUIT;
2947
2948 ext4_li_info = eli;
2949
2950 return 0;
2951 }
2952
2953 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2954 ext4_group_t start)
2955 {
2956 struct ext4_sb_info *sbi = EXT4_SB(sb);
2957 struct ext4_li_request *elr;
2958
2959 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2960 if (!elr)
2961 return NULL;
2962
2963 elr->lr_super = sb;
2964 elr->lr_sbi = sbi;
2965 elr->lr_next_group = start;
2966
2967 /*
2968 * Randomize first schedule time of the request to
2969 * spread the inode table initialization requests
2970 * better.
2971 */
2972 elr->lr_next_sched = jiffies + (prandom_u32() %
2973 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2974 return elr;
2975 }
2976
2977 int ext4_register_li_request(struct super_block *sb,
2978 ext4_group_t first_not_zeroed)
2979 {
2980 struct ext4_sb_info *sbi = EXT4_SB(sb);
2981 struct ext4_li_request *elr = NULL;
2982 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2983 int ret = 0;
2984
2985 mutex_lock(&ext4_li_mtx);
2986 if (sbi->s_li_request != NULL) {
2987 /*
2988 * Reset timeout so it can be computed again, because
2989 * s_li_wait_mult might have changed.
2990 */
2991 sbi->s_li_request->lr_timeout = 0;
2992 goto out;
2993 }
2994
2995 if (first_not_zeroed == ngroups ||
2996 (sb->s_flags & MS_RDONLY) ||
2997 !test_opt(sb, INIT_INODE_TABLE))
2998 goto out;
2999
3000 elr = ext4_li_request_new(sb, first_not_zeroed);
3001 if (!elr) {
3002 ret = -ENOMEM;
3003 goto out;
3004 }
3005
3006 if (NULL == ext4_li_info) {
3007 ret = ext4_li_info_new();
3008 if (ret)
3009 goto out;
3010 }
3011
3012 mutex_lock(&ext4_li_info->li_list_mtx);
3013 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3014 mutex_unlock(&ext4_li_info->li_list_mtx);
3015
3016 sbi->s_li_request = elr;
3017 /*
3018 * set elr to NULL here since it has been inserted to
3019 * the request_list and the removal and free of it is
3020 * handled by ext4_clear_request_list from now on.
3021 */
3022 elr = NULL;
3023
3024 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3025 ret = ext4_run_lazyinit_thread();
3026 if (ret)
3027 goto out;
3028 }
3029 out:
3030 mutex_unlock(&ext4_li_mtx);
3031 if (ret)
3032 kfree(elr);
3033 return ret;
3034 }
3035
3036 /*
3037 * We do not need to lock anything since this is called on
3038 * module unload.
3039 */
3040 static void ext4_destroy_lazyinit_thread(void)
3041 {
3042 /*
3043 * If thread exited earlier
3044 * there's nothing to be done.
3045 */
3046 if (!ext4_li_info || !ext4_lazyinit_task)
3047 return;
3048
3049 kthread_stop(ext4_lazyinit_task);
3050 }
3051
3052 static int set_journal_csum_feature_set(struct super_block *sb)
3053 {
3054 int ret = 1;
3055 int compat, incompat;
3056 struct ext4_sb_info *sbi = EXT4_SB(sb);
3057
3058 if (ext4_has_metadata_csum(sb)) {
3059 /* journal checksum v3 */
3060 compat = 0;
3061 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3062 } else {
3063 /* journal checksum v1 */
3064 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3065 incompat = 0;
3066 }
3067
3068 jbd2_journal_clear_features(sbi->s_journal,
3069 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3070 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3071 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3072 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3073 ret = jbd2_journal_set_features(sbi->s_journal,
3074 compat, 0,
3075 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3076 incompat);
3077 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3078 ret = jbd2_journal_set_features(sbi->s_journal,
3079 compat, 0,
3080 incompat);
3081 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3082 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3083 } else {
3084 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3085 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3086 }
3087
3088 return ret;
3089 }
3090
3091 /*
3092 * Note: calculating the overhead so we can be compatible with
3093 * historical BSD practice is quite difficult in the face of
3094 * clusters/bigalloc. This is because multiple metadata blocks from
3095 * different block group can end up in the same allocation cluster.
3096 * Calculating the exact overhead in the face of clustered allocation
3097 * requires either O(all block bitmaps) in memory or O(number of block
3098 * groups**2) in time. We will still calculate the superblock for
3099 * older file systems --- and if we come across with a bigalloc file
3100 * system with zero in s_overhead_clusters the estimate will be close to
3101 * correct especially for very large cluster sizes --- but for newer
3102 * file systems, it's better to calculate this figure once at mkfs
3103 * time, and store it in the superblock. If the superblock value is
3104 * present (even for non-bigalloc file systems), we will use it.
3105 */
3106 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3107 char *buf)
3108 {
3109 struct ext4_sb_info *sbi = EXT4_SB(sb);
3110 struct ext4_group_desc *gdp;
3111 ext4_fsblk_t first_block, last_block, b;
3112 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3113 int s, j, count = 0;
3114
3115 if (!ext4_has_feature_bigalloc(sb))
3116 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3117 sbi->s_itb_per_group + 2);
3118
3119 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3120 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3121 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3122 for (i = 0; i < ngroups; i++) {
3123 gdp = ext4_get_group_desc(sb, i, NULL);
3124 b = ext4_block_bitmap(sb, gdp);
3125 if (b >= first_block && b <= last_block) {
3126 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3127 count++;
3128 }
3129 b = ext4_inode_bitmap(sb, gdp);
3130 if (b >= first_block && b <= last_block) {
3131 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3132 count++;
3133 }
3134 b = ext4_inode_table(sb, gdp);
3135 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3136 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3137 int c = EXT4_B2C(sbi, b - first_block);
3138 ext4_set_bit(c, buf);
3139 count++;
3140 }
3141 if (i != grp)
3142 continue;
3143 s = 0;
3144 if (ext4_bg_has_super(sb, grp)) {
3145 ext4_set_bit(s++, buf);
3146 count++;
3147 }
3148 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3149 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3150 count++;
3151 }
3152 }
3153 if (!count)
3154 return 0;
3155 return EXT4_CLUSTERS_PER_GROUP(sb) -
3156 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3157 }
3158
3159 /*
3160 * Compute the overhead and stash it in sbi->s_overhead
3161 */
3162 int ext4_calculate_overhead(struct super_block *sb)
3163 {
3164 struct ext4_sb_info *sbi = EXT4_SB(sb);
3165 struct ext4_super_block *es = sbi->s_es;
3166 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3167 ext4_fsblk_t overhead = 0;
3168 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3169
3170 if (!buf)
3171 return -ENOMEM;
3172
3173 /*
3174 * Compute the overhead (FS structures). This is constant
3175 * for a given filesystem unless the number of block groups
3176 * changes so we cache the previous value until it does.
3177 */
3178
3179 /*
3180 * All of the blocks before first_data_block are overhead
3181 */
3182 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3183
3184 /*
3185 * Add the overhead found in each block group
3186 */
3187 for (i = 0; i < ngroups; i++) {
3188 int blks;
3189
3190 blks = count_overhead(sb, i, buf);
3191 overhead += blks;
3192 if (blks)
3193 memset(buf, 0, PAGE_SIZE);
3194 cond_resched();
3195 }
3196 /* Add the internal journal blocks as well */
3197 if (sbi->s_journal && !sbi->journal_bdev)
3198 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3199
3200 sbi->s_overhead = overhead;
3201 smp_wmb();
3202 free_page((unsigned long) buf);
3203 return 0;
3204 }
3205
3206 static void ext4_set_resv_clusters(struct super_block *sb)
3207 {
3208 ext4_fsblk_t resv_clusters;
3209 struct ext4_sb_info *sbi = EXT4_SB(sb);
3210
3211 /*
3212 * There's no need to reserve anything when we aren't using extents.
3213 * The space estimates are exact, there are no unwritten extents,
3214 * hole punching doesn't need new metadata... This is needed especially
3215 * to keep ext2/3 backward compatibility.
3216 */
3217 if (!ext4_has_feature_extents(sb))
3218 return;
3219 /*
3220 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3221 * This should cover the situations where we can not afford to run
3222 * out of space like for example punch hole, or converting
3223 * unwritten extents in delalloc path. In most cases such
3224 * allocation would require 1, or 2 blocks, higher numbers are
3225 * very rare.
3226 */
3227 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3228 sbi->s_cluster_bits);
3229
3230 do_div(resv_clusters, 50);
3231 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3232
3233 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3234 }
3235
3236 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3237 {
3238 char *orig_data = kstrdup(data, GFP_KERNEL);
3239 struct buffer_head *bh;
3240 struct ext4_super_block *es = NULL;
3241 struct ext4_sb_info *sbi;
3242 ext4_fsblk_t block;
3243 ext4_fsblk_t sb_block = get_sb_block(&data);
3244 ext4_fsblk_t logical_sb_block;
3245 unsigned long offset = 0;
3246 unsigned long journal_devnum = 0;
3247 unsigned long def_mount_opts;
3248 struct inode *root;
3249 const char *descr;
3250 int ret = -ENOMEM;
3251 int blocksize, clustersize;
3252 unsigned int db_count;
3253 unsigned int i;
3254 int needs_recovery, has_huge_files, has_bigalloc;
3255 __u64 blocks_count;
3256 int err = 0;
3257 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3258 ext4_group_t first_not_zeroed;
3259
3260 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3261 if (!sbi)
3262 goto out_free_orig;
3263
3264 sbi->s_blockgroup_lock =
3265 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3266 if (!sbi->s_blockgroup_lock) {
3267 kfree(sbi);
3268 goto out_free_orig;
3269 }
3270 sb->s_fs_info = sbi;
3271 sbi->s_sb = sb;
3272 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3273 sbi->s_sb_block = sb_block;
3274 if (sb->s_bdev->bd_part)
3275 sbi->s_sectors_written_start =
3276 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3277
3278 /* Cleanup superblock name */
3279 strreplace(sb->s_id, '/', '!');
3280
3281 /* -EINVAL is default */
3282 ret = -EINVAL;
3283 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3284 if (!blocksize) {
3285 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3286 goto out_fail;
3287 }
3288
3289 /*
3290 * The ext4 superblock will not be buffer aligned for other than 1kB
3291 * block sizes. We need to calculate the offset from buffer start.
3292 */
3293 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3294 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3295 offset = do_div(logical_sb_block, blocksize);
3296 } else {
3297 logical_sb_block = sb_block;
3298 }
3299
3300 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3301 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3302 goto out_fail;
3303 }
3304 /*
3305 * Note: s_es must be initialized as soon as possible because
3306 * some ext4 macro-instructions depend on its value
3307 */
3308 es = (struct ext4_super_block *) (bh->b_data + offset);
3309 sbi->s_es = es;
3310 sb->s_magic = le16_to_cpu(es->s_magic);
3311 if (sb->s_magic != EXT4_SUPER_MAGIC)
3312 goto cantfind_ext4;
3313 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3314
3315 /* Warn if metadata_csum and gdt_csum are both set. */
3316 if (ext4_has_feature_metadata_csum(sb) &&
3317 ext4_has_feature_gdt_csum(sb))
3318 ext4_warning(sb, "metadata_csum and uninit_bg are "
3319 "redundant flags; please run fsck.");
3320
3321 /* Check for a known checksum algorithm */
3322 if (!ext4_verify_csum_type(sb, es)) {
3323 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3324 "unknown checksum algorithm.");
3325 silent = 1;
3326 goto cantfind_ext4;
3327 }
3328
3329 /* Load the checksum driver */
3330 if (ext4_has_feature_metadata_csum(sb)) {
3331 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3332 if (IS_ERR(sbi->s_chksum_driver)) {
3333 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3334 ret = PTR_ERR(sbi->s_chksum_driver);
3335 sbi->s_chksum_driver = NULL;
3336 goto failed_mount;
3337 }
3338 }
3339
3340 /* Check superblock checksum */
3341 if (!ext4_superblock_csum_verify(sb, es)) {
3342 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3343 "invalid superblock checksum. Run e2fsck?");
3344 silent = 1;
3345 ret = -EFSBADCRC;
3346 goto cantfind_ext4;
3347 }
3348
3349 /* Precompute checksum seed for all metadata */
3350 if (ext4_has_feature_csum_seed(sb))
3351 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3352 else if (ext4_has_metadata_csum(sb))
3353 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3354 sizeof(es->s_uuid));
3355
3356 /* Set defaults before we parse the mount options */
3357 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3358 set_opt(sb, INIT_INODE_TABLE);
3359 if (def_mount_opts & EXT4_DEFM_DEBUG)
3360 set_opt(sb, DEBUG);
3361 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3362 set_opt(sb, GRPID);
3363 if (def_mount_opts & EXT4_DEFM_UID16)
3364 set_opt(sb, NO_UID32);
3365 /* xattr user namespace & acls are now defaulted on */
3366 set_opt(sb, XATTR_USER);
3367 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3368 set_opt(sb, POSIX_ACL);
3369 #endif
3370 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3371 if (ext4_has_metadata_csum(sb))
3372 set_opt(sb, JOURNAL_CHECKSUM);
3373
3374 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3375 set_opt(sb, JOURNAL_DATA);
3376 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3377 set_opt(sb, ORDERED_DATA);
3378 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3379 set_opt(sb, WRITEBACK_DATA);
3380
3381 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3382 set_opt(sb, ERRORS_PANIC);
3383 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3384 set_opt(sb, ERRORS_CONT);
3385 else
3386 set_opt(sb, ERRORS_RO);
3387 /* block_validity enabled by default; disable with noblock_validity */
3388 set_opt(sb, BLOCK_VALIDITY);
3389 if (def_mount_opts & EXT4_DEFM_DISCARD)
3390 set_opt(sb, DISCARD);
3391
3392 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3393 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3394 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3395 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3396 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3397
3398 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3399 set_opt(sb, BARRIER);
3400
3401 /*
3402 * enable delayed allocation by default
3403 * Use -o nodelalloc to turn it off
3404 */
3405 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3406 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3407 set_opt(sb, DELALLOC);
3408
3409 /*
3410 * set default s_li_wait_mult for lazyinit, for the case there is
3411 * no mount option specified.
3412 */
3413 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3414
3415 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3416 &journal_devnum, &journal_ioprio, 0)) {
3417 ext4_msg(sb, KERN_WARNING,
3418 "failed to parse options in superblock: %s",
3419 sbi->s_es->s_mount_opts);
3420 }
3421 sbi->s_def_mount_opt = sbi->s_mount_opt;
3422 if (!parse_options((char *) data, sb, &journal_devnum,
3423 &journal_ioprio, 0))
3424 goto failed_mount;
3425
3426 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3427 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3428 "with data=journal disables delayed "
3429 "allocation and O_DIRECT support!\n");
3430 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3431 ext4_msg(sb, KERN_ERR, "can't mount with "
3432 "both data=journal and delalloc");
3433 goto failed_mount;
3434 }
3435 if (test_opt(sb, DIOREAD_NOLOCK)) {
3436 ext4_msg(sb, KERN_ERR, "can't mount with "
3437 "both data=journal and dioread_nolock");
3438 goto failed_mount;
3439 }
3440 if (test_opt(sb, DAX)) {
3441 ext4_msg(sb, KERN_ERR, "can't mount with "
3442 "both data=journal and dax");
3443 goto failed_mount;
3444 }
3445 if (test_opt(sb, DELALLOC))
3446 clear_opt(sb, DELALLOC);
3447 } else {
3448 sb->s_iflags |= SB_I_CGROUPWB;
3449 }
3450
3451 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3452 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3453
3454 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3455 (ext4_has_compat_features(sb) ||
3456 ext4_has_ro_compat_features(sb) ||
3457 ext4_has_incompat_features(sb)))
3458 ext4_msg(sb, KERN_WARNING,
3459 "feature flags set on rev 0 fs, "
3460 "running e2fsck is recommended");
3461
3462 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3463 set_opt2(sb, HURD_COMPAT);
3464 if (ext4_has_feature_64bit(sb)) {
3465 ext4_msg(sb, KERN_ERR,
3466 "The Hurd can't support 64-bit file systems");
3467 goto failed_mount;
3468 }
3469 }
3470
3471 if (IS_EXT2_SB(sb)) {
3472 if (ext2_feature_set_ok(sb))
3473 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3474 "using the ext4 subsystem");
3475 else {
3476 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3477 "to feature incompatibilities");
3478 goto failed_mount;
3479 }
3480 }
3481
3482 if (IS_EXT3_SB(sb)) {
3483 if (ext3_feature_set_ok(sb))
3484 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3485 "using the ext4 subsystem");
3486 else {
3487 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3488 "to feature incompatibilities");
3489 goto failed_mount;
3490 }
3491 }
3492
3493 /*
3494 * Check feature flags regardless of the revision level, since we
3495 * previously didn't change the revision level when setting the flags,
3496 * so there is a chance incompat flags are set on a rev 0 filesystem.
3497 */
3498 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3499 goto failed_mount;
3500
3501 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3502 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3503 blocksize > EXT4_MAX_BLOCK_SIZE) {
3504 ext4_msg(sb, KERN_ERR,
3505 "Unsupported filesystem blocksize %d", blocksize);
3506 goto failed_mount;
3507 }
3508
3509 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3510 ext4_msg(sb, KERN_ERR,
3511 "Number of reserved GDT blocks insanely large: %d",
3512 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3513 goto failed_mount;
3514 }
3515
3516 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3517 err = bdev_dax_supported(sb, blocksize);
3518 if (err)
3519 goto failed_mount;
3520 }
3521
3522 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3523 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3524 es->s_encryption_level);
3525 goto failed_mount;
3526 }
3527
3528 if (sb->s_blocksize != blocksize) {
3529 /* Validate the filesystem blocksize */
3530 if (!sb_set_blocksize(sb, blocksize)) {
3531 ext4_msg(sb, KERN_ERR, "bad block size %d",
3532 blocksize);
3533 goto failed_mount;
3534 }
3535
3536 brelse(bh);
3537 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3538 offset = do_div(logical_sb_block, blocksize);
3539 bh = sb_bread_unmovable(sb, logical_sb_block);
3540 if (!bh) {
3541 ext4_msg(sb, KERN_ERR,
3542 "Can't read superblock on 2nd try");
3543 goto failed_mount;
3544 }
3545 es = (struct ext4_super_block *)(bh->b_data + offset);
3546 sbi->s_es = es;
3547 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3548 ext4_msg(sb, KERN_ERR,
3549 "Magic mismatch, very weird!");
3550 goto failed_mount;
3551 }
3552 }
3553
3554 has_huge_files = ext4_has_feature_huge_file(sb);
3555 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3556 has_huge_files);
3557 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3558
3559 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3560 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3561 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3562 } else {
3563 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3564 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3565 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3566 (!is_power_of_2(sbi->s_inode_size)) ||
3567 (sbi->s_inode_size > blocksize)) {
3568 ext4_msg(sb, KERN_ERR,
3569 "unsupported inode size: %d",
3570 sbi->s_inode_size);
3571 goto failed_mount;
3572 }
3573 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3574 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3575 }
3576
3577 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3578 if (ext4_has_feature_64bit(sb)) {
3579 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3580 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3581 !is_power_of_2(sbi->s_desc_size)) {
3582 ext4_msg(sb, KERN_ERR,
3583 "unsupported descriptor size %lu",
3584 sbi->s_desc_size);
3585 goto failed_mount;
3586 }
3587 } else
3588 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3589
3590 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3591 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3592 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3593 goto cantfind_ext4;
3594
3595 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3596 if (sbi->s_inodes_per_block == 0)
3597 goto cantfind_ext4;
3598 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3599 sbi->s_inodes_per_block;
3600 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3601 sbi->s_sbh = bh;
3602 sbi->s_mount_state = le16_to_cpu(es->s_state);
3603 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3604 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3605
3606 for (i = 0; i < 4; i++)
3607 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3608 sbi->s_def_hash_version = es->s_def_hash_version;
3609 if (ext4_has_feature_dir_index(sb)) {
3610 i = le32_to_cpu(es->s_flags);
3611 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3612 sbi->s_hash_unsigned = 3;
3613 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3614 #ifdef __CHAR_UNSIGNED__
3615 if (!(sb->s_flags & MS_RDONLY))
3616 es->s_flags |=
3617 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3618 sbi->s_hash_unsigned = 3;
3619 #else
3620 if (!(sb->s_flags & MS_RDONLY))
3621 es->s_flags |=
3622 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3623 #endif
3624 }
3625 }
3626
3627 /* Handle clustersize */
3628 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3629 has_bigalloc = ext4_has_feature_bigalloc(sb);
3630 if (has_bigalloc) {
3631 if (clustersize < blocksize) {
3632 ext4_msg(sb, KERN_ERR,
3633 "cluster size (%d) smaller than "
3634 "block size (%d)", clustersize, blocksize);
3635 goto failed_mount;
3636 }
3637 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3638 le32_to_cpu(es->s_log_block_size);
3639 sbi->s_clusters_per_group =
3640 le32_to_cpu(es->s_clusters_per_group);
3641 if (sbi->s_clusters_per_group > blocksize * 8) {
3642 ext4_msg(sb, KERN_ERR,
3643 "#clusters per group too big: %lu",
3644 sbi->s_clusters_per_group);
3645 goto failed_mount;
3646 }
3647 if (sbi->s_blocks_per_group !=
3648 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3649 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3650 "clusters per group (%lu) inconsistent",
3651 sbi->s_blocks_per_group,
3652 sbi->s_clusters_per_group);
3653 goto failed_mount;
3654 }
3655 } else {
3656 if (clustersize != blocksize) {
3657 ext4_warning(sb, "fragment/cluster size (%d) != "
3658 "block size (%d)", clustersize,
3659 blocksize);
3660 clustersize = blocksize;
3661 }
3662 if (sbi->s_blocks_per_group > blocksize * 8) {
3663 ext4_msg(sb, KERN_ERR,
3664 "#blocks per group too big: %lu",
3665 sbi->s_blocks_per_group);
3666 goto failed_mount;
3667 }
3668 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3669 sbi->s_cluster_bits = 0;
3670 }
3671 sbi->s_cluster_ratio = clustersize / blocksize;
3672
3673 if (sbi->s_inodes_per_group > blocksize * 8) {
3674 ext4_msg(sb, KERN_ERR,
3675 "#inodes per group too big: %lu",
3676 sbi->s_inodes_per_group);
3677 goto failed_mount;
3678 }
3679
3680 /* Do we have standard group size of clustersize * 8 blocks ? */
3681 if (sbi->s_blocks_per_group == clustersize << 3)
3682 set_opt2(sb, STD_GROUP_SIZE);
3683
3684 /*
3685 * Test whether we have more sectors than will fit in sector_t,
3686 * and whether the max offset is addressable by the page cache.
3687 */
3688 err = generic_check_addressable(sb->s_blocksize_bits,
3689 ext4_blocks_count(es));
3690 if (err) {
3691 ext4_msg(sb, KERN_ERR, "filesystem"
3692 " too large to mount safely on this system");
3693 if (sizeof(sector_t) < 8)
3694 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3695 goto failed_mount;
3696 }
3697
3698 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3699 goto cantfind_ext4;
3700
3701 /* check blocks count against device size */
3702 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3703 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3704 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3705 "exceeds size of device (%llu blocks)",
3706 ext4_blocks_count(es), blocks_count);
3707 goto failed_mount;
3708 }
3709
3710 /*
3711 * It makes no sense for the first data block to be beyond the end
3712 * of the filesystem.
3713 */
3714 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3715 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3716 "block %u is beyond end of filesystem (%llu)",
3717 le32_to_cpu(es->s_first_data_block),
3718 ext4_blocks_count(es));
3719 goto failed_mount;
3720 }
3721 blocks_count = (ext4_blocks_count(es) -
3722 le32_to_cpu(es->s_first_data_block) +
3723 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3724 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3725 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3726 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3727 "(block count %llu, first data block %u, "
3728 "blocks per group %lu)", sbi->s_groups_count,
3729 ext4_blocks_count(es),
3730 le32_to_cpu(es->s_first_data_block),
3731 EXT4_BLOCKS_PER_GROUP(sb));
3732 goto failed_mount;
3733 }
3734 sbi->s_groups_count = blocks_count;
3735 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3736 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3737 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3738 EXT4_DESC_PER_BLOCK(sb);
3739 sbi->s_group_desc = ext4_kvmalloc(db_count *
3740 sizeof(struct buffer_head *),
3741 GFP_KERNEL);
3742 if (sbi->s_group_desc == NULL) {
3743 ext4_msg(sb, KERN_ERR, "not enough memory");
3744 ret = -ENOMEM;
3745 goto failed_mount;
3746 }
3747
3748 bgl_lock_init(sbi->s_blockgroup_lock);
3749
3750 for (i = 0; i < db_count; i++) {
3751 block = descriptor_loc(sb, logical_sb_block, i);
3752 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3753 if (!sbi->s_group_desc[i]) {
3754 ext4_msg(sb, KERN_ERR,
3755 "can't read group descriptor %d", i);
3756 db_count = i;
3757 goto failed_mount2;
3758 }
3759 }
3760 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3761 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3762 ret = -EFSCORRUPTED;
3763 goto failed_mount2;
3764 }
3765
3766 sbi->s_gdb_count = db_count;
3767 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3768 spin_lock_init(&sbi->s_next_gen_lock);
3769
3770 setup_timer(&sbi->s_err_report, print_daily_error_info,
3771 (unsigned long) sb);
3772
3773 /* Register extent status tree shrinker */
3774 if (ext4_es_register_shrinker(sbi))
3775 goto failed_mount3;
3776
3777 sbi->s_stripe = ext4_get_stripe_size(sbi);
3778 sbi->s_extent_max_zeroout_kb = 32;
3779
3780 /*
3781 * set up enough so that it can read an inode
3782 */
3783 sb->s_op = &ext4_sops;
3784 sb->s_export_op = &ext4_export_ops;
3785 sb->s_xattr = ext4_xattr_handlers;
3786 sb->s_cop = &ext4_cryptops;
3787 #ifdef CONFIG_QUOTA
3788 sb->dq_op = &ext4_quota_operations;
3789 if (ext4_has_feature_quota(sb))
3790 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3791 else
3792 sb->s_qcop = &ext4_qctl_operations;
3793 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3794 #endif
3795 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3796
3797 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3798 mutex_init(&sbi->s_orphan_lock);
3799
3800 sb->s_root = NULL;
3801
3802 needs_recovery = (es->s_last_orphan != 0 ||
3803 ext4_has_feature_journal_needs_recovery(sb));
3804
3805 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3806 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3807 goto failed_mount3a;
3808
3809 /*
3810 * The first inode we look at is the journal inode. Don't try
3811 * root first: it may be modified in the journal!
3812 */
3813 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3814 if (ext4_load_journal(sb, es, journal_devnum))
3815 goto failed_mount3a;
3816 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3817 ext4_has_feature_journal_needs_recovery(sb)) {
3818 ext4_msg(sb, KERN_ERR, "required journal recovery "
3819 "suppressed and not mounted read-only");
3820 goto failed_mount_wq;
3821 } else {
3822 /* Nojournal mode, all journal mount options are illegal */
3823 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3824 ext4_msg(sb, KERN_ERR, "can't mount with "
3825 "journal_checksum, fs mounted w/o journal");
3826 goto failed_mount_wq;
3827 }
3828 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3829 ext4_msg(sb, KERN_ERR, "can't mount with "
3830 "journal_async_commit, fs mounted w/o journal");
3831 goto failed_mount_wq;
3832 }
3833 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3834 ext4_msg(sb, KERN_ERR, "can't mount with "
3835 "commit=%lu, fs mounted w/o journal",
3836 sbi->s_commit_interval / HZ);
3837 goto failed_mount_wq;
3838 }
3839 if (EXT4_MOUNT_DATA_FLAGS &
3840 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3841 ext4_msg(sb, KERN_ERR, "can't mount with "
3842 "data=, fs mounted w/o journal");
3843 goto failed_mount_wq;
3844 }
3845 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3846 clear_opt(sb, JOURNAL_CHECKSUM);
3847 clear_opt(sb, DATA_FLAGS);
3848 sbi->s_journal = NULL;
3849 needs_recovery = 0;
3850 goto no_journal;
3851 }
3852
3853 if (ext4_has_feature_64bit(sb) &&
3854 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3855 JBD2_FEATURE_INCOMPAT_64BIT)) {
3856 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3857 goto failed_mount_wq;
3858 }
3859
3860 if (!set_journal_csum_feature_set(sb)) {
3861 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3862 "feature set");
3863 goto failed_mount_wq;
3864 }
3865
3866 /* We have now updated the journal if required, so we can
3867 * validate the data journaling mode. */
3868 switch (test_opt(sb, DATA_FLAGS)) {
3869 case 0:
3870 /* No mode set, assume a default based on the journal
3871 * capabilities: ORDERED_DATA if the journal can
3872 * cope, else JOURNAL_DATA
3873 */
3874 if (jbd2_journal_check_available_features
3875 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3876 set_opt(sb, ORDERED_DATA);
3877 else
3878 set_opt(sb, JOURNAL_DATA);
3879 break;
3880
3881 case EXT4_MOUNT_ORDERED_DATA:
3882 case EXT4_MOUNT_WRITEBACK_DATA:
3883 if (!jbd2_journal_check_available_features
3884 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3885 ext4_msg(sb, KERN_ERR, "Journal does not support "
3886 "requested data journaling mode");
3887 goto failed_mount_wq;
3888 }
3889 default:
3890 break;
3891 }
3892 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3893
3894 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3895
3896 no_journal:
3897 sbi->s_mb_cache = ext4_xattr_create_cache();
3898 if (!sbi->s_mb_cache) {
3899 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3900 goto failed_mount_wq;
3901 }
3902
3903 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3904 (blocksize != PAGE_SIZE)) {
3905 ext4_msg(sb, KERN_ERR,
3906 "Unsupported blocksize for fs encryption");
3907 goto failed_mount_wq;
3908 }
3909
3910 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3911 !ext4_has_feature_encrypt(sb)) {
3912 ext4_set_feature_encrypt(sb);
3913 ext4_commit_super(sb, 1);
3914 }
3915
3916 /*
3917 * Get the # of file system overhead blocks from the
3918 * superblock if present.
3919 */
3920 if (es->s_overhead_clusters)
3921 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3922 else {
3923 err = ext4_calculate_overhead(sb);
3924 if (err)
3925 goto failed_mount_wq;
3926 }
3927
3928 /*
3929 * The maximum number of concurrent works can be high and
3930 * concurrency isn't really necessary. Limit it to 1.
3931 */
3932 EXT4_SB(sb)->rsv_conversion_wq =
3933 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3934 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3935 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3936 ret = -ENOMEM;
3937 goto failed_mount4;
3938 }
3939
3940 /*
3941 * The jbd2_journal_load will have done any necessary log recovery,
3942 * so we can safely mount the rest of the filesystem now.
3943 */
3944
3945 root = ext4_iget(sb, EXT4_ROOT_INO);
3946 if (IS_ERR(root)) {
3947 ext4_msg(sb, KERN_ERR, "get root inode failed");
3948 ret = PTR_ERR(root);
3949 root = NULL;
3950 goto failed_mount4;
3951 }
3952 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3953 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3954 iput(root);
3955 goto failed_mount4;
3956 }
3957 sb->s_root = d_make_root(root);
3958 if (!sb->s_root) {
3959 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3960 ret = -ENOMEM;
3961 goto failed_mount4;
3962 }
3963
3964 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3965 sb->s_flags |= MS_RDONLY;
3966
3967 /* determine the minimum size of new large inodes, if present */
3968 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3969 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3970 EXT4_GOOD_OLD_INODE_SIZE;
3971 if (ext4_has_feature_extra_isize(sb)) {
3972 if (sbi->s_want_extra_isize <
3973 le16_to_cpu(es->s_want_extra_isize))
3974 sbi->s_want_extra_isize =
3975 le16_to_cpu(es->s_want_extra_isize);
3976 if (sbi->s_want_extra_isize <
3977 le16_to_cpu(es->s_min_extra_isize))
3978 sbi->s_want_extra_isize =
3979 le16_to_cpu(es->s_min_extra_isize);
3980 }
3981 }
3982 /* Check if enough inode space is available */
3983 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3984 sbi->s_inode_size) {
3985 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3986 EXT4_GOOD_OLD_INODE_SIZE;
3987 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3988 "available");
3989 }
3990
3991 ext4_set_resv_clusters(sb);
3992
3993 err = ext4_setup_system_zone(sb);
3994 if (err) {
3995 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3996 "zone (%d)", err);
3997 goto failed_mount4a;
3998 }
3999
4000 ext4_ext_init(sb);
4001 err = ext4_mb_init(sb);
4002 if (err) {
4003 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4004 err);
4005 goto failed_mount5;
4006 }
4007
4008 block = ext4_count_free_clusters(sb);
4009 ext4_free_blocks_count_set(sbi->s_es,
4010 EXT4_C2B(sbi, block));
4011 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4012 GFP_KERNEL);
4013 if (!err) {
4014 unsigned long freei = ext4_count_free_inodes(sb);
4015 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4016 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4017 GFP_KERNEL);
4018 }
4019 if (!err)
4020 err = percpu_counter_init(&sbi->s_dirs_counter,
4021 ext4_count_dirs(sb), GFP_KERNEL);
4022 if (!err)
4023 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4024 GFP_KERNEL);
4025 if (!err)
4026 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4027
4028 if (err) {
4029 ext4_msg(sb, KERN_ERR, "insufficient memory");
4030 goto failed_mount6;
4031 }
4032
4033 if (ext4_has_feature_flex_bg(sb))
4034 if (!ext4_fill_flex_info(sb)) {
4035 ext4_msg(sb, KERN_ERR,
4036 "unable to initialize "
4037 "flex_bg meta info!");
4038 goto failed_mount6;
4039 }
4040
4041 err = ext4_register_li_request(sb, first_not_zeroed);
4042 if (err)
4043 goto failed_mount6;
4044
4045 err = ext4_register_sysfs(sb);
4046 if (err)
4047 goto failed_mount7;
4048
4049 #ifdef CONFIG_QUOTA
4050 /* Enable quota usage during mount. */
4051 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4052 err = ext4_enable_quotas(sb);
4053 if (err)
4054 goto failed_mount8;
4055 }
4056 #endif /* CONFIG_QUOTA */
4057
4058 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4059 ext4_orphan_cleanup(sb, es);
4060 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4061 if (needs_recovery) {
4062 ext4_msg(sb, KERN_INFO, "recovery complete");
4063 ext4_mark_recovery_complete(sb, es);
4064 }
4065 if (EXT4_SB(sb)->s_journal) {
4066 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4067 descr = " journalled data mode";
4068 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4069 descr = " ordered data mode";
4070 else
4071 descr = " writeback data mode";
4072 } else
4073 descr = "out journal";
4074
4075 if (test_opt(sb, DISCARD)) {
4076 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4077 if (!blk_queue_discard(q))
4078 ext4_msg(sb, KERN_WARNING,
4079 "mounting with \"discard\" option, but "
4080 "the device does not support discard");
4081 }
4082
4083 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4084 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4085 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4086 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4087
4088 if (es->s_error_count)
4089 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4090
4091 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4092 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4093 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4094 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4095
4096 kfree(orig_data);
4097 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4098 memcpy(sbi->key_prefix, EXT4_KEY_DESC_PREFIX,
4099 EXT4_KEY_DESC_PREFIX_SIZE);
4100 sbi->key_prefix_size = EXT4_KEY_DESC_PREFIX_SIZE;
4101 #endif
4102 return 0;
4103
4104 cantfind_ext4:
4105 if (!silent)
4106 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4107 goto failed_mount;
4108
4109 #ifdef CONFIG_QUOTA
4110 failed_mount8:
4111 ext4_unregister_sysfs(sb);
4112 #endif
4113 failed_mount7:
4114 ext4_unregister_li_request(sb);
4115 failed_mount6:
4116 ext4_mb_release(sb);
4117 if (sbi->s_flex_groups)
4118 kvfree(sbi->s_flex_groups);
4119 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4120 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4121 percpu_counter_destroy(&sbi->s_dirs_counter);
4122 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4123 failed_mount5:
4124 ext4_ext_release(sb);
4125 ext4_release_system_zone(sb);
4126 failed_mount4a:
4127 dput(sb->s_root);
4128 sb->s_root = NULL;
4129 failed_mount4:
4130 ext4_msg(sb, KERN_ERR, "mount failed");
4131 if (EXT4_SB(sb)->rsv_conversion_wq)
4132 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4133 failed_mount_wq:
4134 if (sbi->s_mb_cache) {
4135 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4136 sbi->s_mb_cache = NULL;
4137 }
4138 if (sbi->s_journal) {
4139 jbd2_journal_destroy(sbi->s_journal);
4140 sbi->s_journal = NULL;
4141 }
4142 failed_mount3a:
4143 ext4_es_unregister_shrinker(sbi);
4144 failed_mount3:
4145 del_timer_sync(&sbi->s_err_report);
4146 if (sbi->s_mmp_tsk)
4147 kthread_stop(sbi->s_mmp_tsk);
4148 failed_mount2:
4149 for (i = 0; i < db_count; i++)
4150 brelse(sbi->s_group_desc[i]);
4151 kvfree(sbi->s_group_desc);
4152 failed_mount:
4153 if (sbi->s_chksum_driver)
4154 crypto_free_shash(sbi->s_chksum_driver);
4155 #ifdef CONFIG_QUOTA
4156 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4157 kfree(sbi->s_qf_names[i]);
4158 #endif
4159 ext4_blkdev_remove(sbi);
4160 brelse(bh);
4161 out_fail:
4162 sb->s_fs_info = NULL;
4163 kfree(sbi->s_blockgroup_lock);
4164 kfree(sbi);
4165 out_free_orig:
4166 kfree(orig_data);
4167 return err ? err : ret;
4168 }
4169
4170 /*
4171 * Setup any per-fs journal parameters now. We'll do this both on
4172 * initial mount, once the journal has been initialised but before we've
4173 * done any recovery; and again on any subsequent remount.
4174 */
4175 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4176 {
4177 struct ext4_sb_info *sbi = EXT4_SB(sb);
4178
4179 journal->j_commit_interval = sbi->s_commit_interval;
4180 journal->j_min_batch_time = sbi->s_min_batch_time;
4181 journal->j_max_batch_time = sbi->s_max_batch_time;
4182
4183 write_lock(&journal->j_state_lock);
4184 if (test_opt(sb, BARRIER))
4185 journal->j_flags |= JBD2_BARRIER;
4186 else
4187 journal->j_flags &= ~JBD2_BARRIER;
4188 if (test_opt(sb, DATA_ERR_ABORT))
4189 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4190 else
4191 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4192 write_unlock(&journal->j_state_lock);
4193 }
4194
4195 static journal_t *ext4_get_journal(struct super_block *sb,
4196 unsigned int journal_inum)
4197 {
4198 struct inode *journal_inode;
4199 journal_t *journal;
4200
4201 BUG_ON(!ext4_has_feature_journal(sb));
4202
4203 /* First, test for the existence of a valid inode on disk. Bad
4204 * things happen if we iget() an unused inode, as the subsequent
4205 * iput() will try to delete it. */
4206
4207 journal_inode = ext4_iget(sb, journal_inum);
4208 if (IS_ERR(journal_inode)) {
4209 ext4_msg(sb, KERN_ERR, "no journal found");
4210 return NULL;
4211 }
4212 if (!journal_inode->i_nlink) {
4213 make_bad_inode(journal_inode);
4214 iput(journal_inode);
4215 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4216 return NULL;
4217 }
4218
4219 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4220 journal_inode, journal_inode->i_size);
4221 if (!S_ISREG(journal_inode->i_mode)) {
4222 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4223 iput(journal_inode);
4224 return NULL;
4225 }
4226
4227 journal = jbd2_journal_init_inode(journal_inode);
4228 if (!journal) {
4229 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4230 iput(journal_inode);
4231 return NULL;
4232 }
4233 journal->j_private = sb;
4234 ext4_init_journal_params(sb, journal);
4235 return journal;
4236 }
4237
4238 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4239 dev_t j_dev)
4240 {
4241 struct buffer_head *bh;
4242 journal_t *journal;
4243 ext4_fsblk_t start;
4244 ext4_fsblk_t len;
4245 int hblock, blocksize;
4246 ext4_fsblk_t sb_block;
4247 unsigned long offset;
4248 struct ext4_super_block *es;
4249 struct block_device *bdev;
4250
4251 BUG_ON(!ext4_has_feature_journal(sb));
4252
4253 bdev = ext4_blkdev_get(j_dev, sb);
4254 if (bdev == NULL)
4255 return NULL;
4256
4257 blocksize = sb->s_blocksize;
4258 hblock = bdev_logical_block_size(bdev);
4259 if (blocksize < hblock) {
4260 ext4_msg(sb, KERN_ERR,
4261 "blocksize too small for journal device");
4262 goto out_bdev;
4263 }
4264
4265 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4266 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4267 set_blocksize(bdev, blocksize);
4268 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4269 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4270 "external journal");
4271 goto out_bdev;
4272 }
4273
4274 es = (struct ext4_super_block *) (bh->b_data + offset);
4275 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4276 !(le32_to_cpu(es->s_feature_incompat) &
4277 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4278 ext4_msg(sb, KERN_ERR, "external journal has "
4279 "bad superblock");
4280 brelse(bh);
4281 goto out_bdev;
4282 }
4283
4284 if ((le32_to_cpu(es->s_feature_ro_compat) &
4285 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4286 es->s_checksum != ext4_superblock_csum(sb, es)) {
4287 ext4_msg(sb, KERN_ERR, "external journal has "
4288 "corrupt superblock");
4289 brelse(bh);
4290 goto out_bdev;
4291 }
4292
4293 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4294 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4295 brelse(bh);
4296 goto out_bdev;
4297 }
4298
4299 len = ext4_blocks_count(es);
4300 start = sb_block + 1;
4301 brelse(bh); /* we're done with the superblock */
4302
4303 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4304 start, len, blocksize);
4305 if (!journal) {
4306 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4307 goto out_bdev;
4308 }
4309 journal->j_private = sb;
4310 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4311 wait_on_buffer(journal->j_sb_buffer);
4312 if (!buffer_uptodate(journal->j_sb_buffer)) {
4313 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4314 goto out_journal;
4315 }
4316 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4317 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4318 "user (unsupported) - %d",
4319 be32_to_cpu(journal->j_superblock->s_nr_users));
4320 goto out_journal;
4321 }
4322 EXT4_SB(sb)->journal_bdev = bdev;
4323 ext4_init_journal_params(sb, journal);
4324 return journal;
4325
4326 out_journal:
4327 jbd2_journal_destroy(journal);
4328 out_bdev:
4329 ext4_blkdev_put(bdev);
4330 return NULL;
4331 }
4332
4333 static int ext4_load_journal(struct super_block *sb,
4334 struct ext4_super_block *es,
4335 unsigned long journal_devnum)
4336 {
4337 journal_t *journal;
4338 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4339 dev_t journal_dev;
4340 int err = 0;
4341 int really_read_only;
4342
4343 BUG_ON(!ext4_has_feature_journal(sb));
4344
4345 if (journal_devnum &&
4346 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4347 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4348 "numbers have changed");
4349 journal_dev = new_decode_dev(journal_devnum);
4350 } else
4351 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4352
4353 really_read_only = bdev_read_only(sb->s_bdev);
4354
4355 /*
4356 * Are we loading a blank journal or performing recovery after a
4357 * crash? For recovery, we need to check in advance whether we
4358 * can get read-write access to the device.
4359 */
4360 if (ext4_has_feature_journal_needs_recovery(sb)) {
4361 if (sb->s_flags & MS_RDONLY) {
4362 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4363 "required on readonly filesystem");
4364 if (really_read_only) {
4365 ext4_msg(sb, KERN_ERR, "write access "
4366 "unavailable, cannot proceed");
4367 return -EROFS;
4368 }
4369 ext4_msg(sb, KERN_INFO, "write access will "
4370 "be enabled during recovery");
4371 }
4372 }
4373
4374 if (journal_inum && journal_dev) {
4375 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4376 "and inode journals!");
4377 return -EINVAL;
4378 }
4379
4380 if (journal_inum) {
4381 if (!(journal = ext4_get_journal(sb, journal_inum)))
4382 return -EINVAL;
4383 } else {
4384 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4385 return -EINVAL;
4386 }
4387
4388 if (!(journal->j_flags & JBD2_BARRIER))
4389 ext4_msg(sb, KERN_INFO, "barriers disabled");
4390
4391 if (!ext4_has_feature_journal_needs_recovery(sb))
4392 err = jbd2_journal_wipe(journal, !really_read_only);
4393 if (!err) {
4394 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4395 if (save)
4396 memcpy(save, ((char *) es) +
4397 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4398 err = jbd2_journal_load(journal);
4399 if (save)
4400 memcpy(((char *) es) + EXT4_S_ERR_START,
4401 save, EXT4_S_ERR_LEN);
4402 kfree(save);
4403 }
4404
4405 if (err) {
4406 ext4_msg(sb, KERN_ERR, "error loading journal");
4407 jbd2_journal_destroy(journal);
4408 return err;
4409 }
4410
4411 EXT4_SB(sb)->s_journal = journal;
4412 ext4_clear_journal_err(sb, es);
4413
4414 if (!really_read_only && journal_devnum &&
4415 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4416 es->s_journal_dev = cpu_to_le32(journal_devnum);
4417
4418 /* Make sure we flush the recovery flag to disk. */
4419 ext4_commit_super(sb, 1);
4420 }
4421
4422 return 0;
4423 }
4424
4425 static int ext4_commit_super(struct super_block *sb, int sync)
4426 {
4427 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4428 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4429 int error = 0;
4430
4431 if (!sbh || block_device_ejected(sb))
4432 return error;
4433 /*
4434 * If the file system is mounted read-only, don't update the
4435 * superblock write time. This avoids updating the superblock
4436 * write time when we are mounting the root file system
4437 * read/only but we need to replay the journal; at that point,
4438 * for people who are east of GMT and who make their clock
4439 * tick in localtime for Windows bug-for-bug compatibility,
4440 * the clock is set in the future, and this will cause e2fsck
4441 * to complain and force a full file system check.
4442 */
4443 if (!(sb->s_flags & MS_RDONLY))
4444 es->s_wtime = cpu_to_le32(get_seconds());
4445 if (sb->s_bdev->bd_part)
4446 es->s_kbytes_written =
4447 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4448 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4449 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4450 else
4451 es->s_kbytes_written =
4452 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4453 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4454 ext4_free_blocks_count_set(es,
4455 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4456 &EXT4_SB(sb)->s_freeclusters_counter)));
4457 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4458 es->s_free_inodes_count =
4459 cpu_to_le32(percpu_counter_sum_positive(
4460 &EXT4_SB(sb)->s_freeinodes_counter));
4461 BUFFER_TRACE(sbh, "marking dirty");
4462 ext4_superblock_csum_set(sb);
4463 lock_buffer(sbh);
4464 if (buffer_write_io_error(sbh)) {
4465 /*
4466 * Oh, dear. A previous attempt to write the
4467 * superblock failed. This could happen because the
4468 * USB device was yanked out. Or it could happen to
4469 * be a transient write error and maybe the block will
4470 * be remapped. Nothing we can do but to retry the
4471 * write and hope for the best.
4472 */
4473 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4474 "superblock detected");
4475 clear_buffer_write_io_error(sbh);
4476 set_buffer_uptodate(sbh);
4477 }
4478 mark_buffer_dirty(sbh);
4479 unlock_buffer(sbh);
4480 if (sync) {
4481 error = __sync_dirty_buffer(sbh,
4482 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4483 if (error)
4484 return error;
4485
4486 error = buffer_write_io_error(sbh);
4487 if (error) {
4488 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4489 "superblock");
4490 clear_buffer_write_io_error(sbh);
4491 set_buffer_uptodate(sbh);
4492 }
4493 }
4494 return error;
4495 }
4496
4497 /*
4498 * Have we just finished recovery? If so, and if we are mounting (or
4499 * remounting) the filesystem readonly, then we will end up with a
4500 * consistent fs on disk. Record that fact.
4501 */
4502 static void ext4_mark_recovery_complete(struct super_block *sb,
4503 struct ext4_super_block *es)
4504 {
4505 journal_t *journal = EXT4_SB(sb)->s_journal;
4506
4507 if (!ext4_has_feature_journal(sb)) {
4508 BUG_ON(journal != NULL);
4509 return;
4510 }
4511 jbd2_journal_lock_updates(journal);
4512 if (jbd2_journal_flush(journal) < 0)
4513 goto out;
4514
4515 if (ext4_has_feature_journal_needs_recovery(sb) &&
4516 sb->s_flags & MS_RDONLY) {
4517 ext4_clear_feature_journal_needs_recovery(sb);
4518 ext4_commit_super(sb, 1);
4519 }
4520
4521 out:
4522 jbd2_journal_unlock_updates(journal);
4523 }
4524
4525 /*
4526 * If we are mounting (or read-write remounting) a filesystem whose journal
4527 * has recorded an error from a previous lifetime, move that error to the
4528 * main filesystem now.
4529 */
4530 static void ext4_clear_journal_err(struct super_block *sb,
4531 struct ext4_super_block *es)
4532 {
4533 journal_t *journal;
4534 int j_errno;
4535 const char *errstr;
4536
4537 BUG_ON(!ext4_has_feature_journal(sb));
4538
4539 journal = EXT4_SB(sb)->s_journal;
4540
4541 /*
4542 * Now check for any error status which may have been recorded in the
4543 * journal by a prior ext4_error() or ext4_abort()
4544 */
4545
4546 j_errno = jbd2_journal_errno(journal);
4547 if (j_errno) {
4548 char nbuf[16];
4549
4550 errstr = ext4_decode_error(sb, j_errno, nbuf);
4551 ext4_warning(sb, "Filesystem error recorded "
4552 "from previous mount: %s", errstr);
4553 ext4_warning(sb, "Marking fs in need of filesystem check.");
4554
4555 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4556 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4557 ext4_commit_super(sb, 1);
4558
4559 jbd2_journal_clear_err(journal);
4560 jbd2_journal_update_sb_errno(journal);
4561 }
4562 }
4563
4564 /*
4565 * Force the running and committing transactions to commit,
4566 * and wait on the commit.
4567 */
4568 int ext4_force_commit(struct super_block *sb)
4569 {
4570 journal_t *journal;
4571
4572 if (sb->s_flags & MS_RDONLY)
4573 return 0;
4574
4575 journal = EXT4_SB(sb)->s_journal;
4576 return ext4_journal_force_commit(journal);
4577 }
4578
4579 static int ext4_sync_fs(struct super_block *sb, int wait)
4580 {
4581 int ret = 0;
4582 tid_t target;
4583 bool needs_barrier = false;
4584 struct ext4_sb_info *sbi = EXT4_SB(sb);
4585
4586 trace_ext4_sync_fs(sb, wait);
4587 flush_workqueue(sbi->rsv_conversion_wq);
4588 /*
4589 * Writeback quota in non-journalled quota case - journalled quota has
4590 * no dirty dquots
4591 */
4592 dquot_writeback_dquots(sb, -1);
4593 /*
4594 * Data writeback is possible w/o journal transaction, so barrier must
4595 * being sent at the end of the function. But we can skip it if
4596 * transaction_commit will do it for us.
4597 */
4598 if (sbi->s_journal) {
4599 target = jbd2_get_latest_transaction(sbi->s_journal);
4600 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4601 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4602 needs_barrier = true;
4603
4604 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4605 if (wait)
4606 ret = jbd2_log_wait_commit(sbi->s_journal,
4607 target);
4608 }
4609 } else if (wait && test_opt(sb, BARRIER))
4610 needs_barrier = true;
4611 if (needs_barrier) {
4612 int err;
4613 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4614 if (!ret)
4615 ret = err;
4616 }
4617
4618 return ret;
4619 }
4620
4621 /*
4622 * LVM calls this function before a (read-only) snapshot is created. This
4623 * gives us a chance to flush the journal completely and mark the fs clean.
4624 *
4625 * Note that only this function cannot bring a filesystem to be in a clean
4626 * state independently. It relies on upper layer to stop all data & metadata
4627 * modifications.
4628 */
4629 static int ext4_freeze(struct super_block *sb)
4630 {
4631 int error = 0;
4632 journal_t *journal;
4633
4634 if (sb->s_flags & MS_RDONLY)
4635 return 0;
4636
4637 journal = EXT4_SB(sb)->s_journal;
4638
4639 if (journal) {
4640 /* Now we set up the journal barrier. */
4641 jbd2_journal_lock_updates(journal);
4642
4643 /*
4644 * Don't clear the needs_recovery flag if we failed to
4645 * flush the journal.
4646 */
4647 error = jbd2_journal_flush(journal);
4648 if (error < 0)
4649 goto out;
4650
4651 /* Journal blocked and flushed, clear needs_recovery flag. */
4652 ext4_clear_feature_journal_needs_recovery(sb);
4653 }
4654
4655 error = ext4_commit_super(sb, 1);
4656 out:
4657 if (journal)
4658 /* we rely on upper layer to stop further updates */
4659 jbd2_journal_unlock_updates(journal);
4660 return error;
4661 }
4662
4663 /*
4664 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4665 * flag here, even though the filesystem is not technically dirty yet.
4666 */
4667 static int ext4_unfreeze(struct super_block *sb)
4668 {
4669 if (sb->s_flags & MS_RDONLY)
4670 return 0;
4671
4672 if (EXT4_SB(sb)->s_journal) {
4673 /* Reset the needs_recovery flag before the fs is unlocked. */
4674 ext4_set_feature_journal_needs_recovery(sb);
4675 }
4676
4677 ext4_commit_super(sb, 1);
4678 return 0;
4679 }
4680
4681 /*
4682 * Structure to save mount options for ext4_remount's benefit
4683 */
4684 struct ext4_mount_options {
4685 unsigned long s_mount_opt;
4686 unsigned long s_mount_opt2;
4687 kuid_t s_resuid;
4688 kgid_t s_resgid;
4689 unsigned long s_commit_interval;
4690 u32 s_min_batch_time, s_max_batch_time;
4691 #ifdef CONFIG_QUOTA
4692 int s_jquota_fmt;
4693 char *s_qf_names[EXT4_MAXQUOTAS];
4694 #endif
4695 };
4696
4697 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4698 {
4699 struct ext4_super_block *es;
4700 struct ext4_sb_info *sbi = EXT4_SB(sb);
4701 unsigned long old_sb_flags;
4702 struct ext4_mount_options old_opts;
4703 int enable_quota = 0;
4704 ext4_group_t g;
4705 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4706 int err = 0;
4707 #ifdef CONFIG_QUOTA
4708 int i, j;
4709 #endif
4710 char *orig_data = kstrdup(data, GFP_KERNEL);
4711
4712 /* Store the original options */
4713 old_sb_flags = sb->s_flags;
4714 old_opts.s_mount_opt = sbi->s_mount_opt;
4715 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4716 old_opts.s_resuid = sbi->s_resuid;
4717 old_opts.s_resgid = sbi->s_resgid;
4718 old_opts.s_commit_interval = sbi->s_commit_interval;
4719 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4720 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4721 #ifdef CONFIG_QUOTA
4722 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4723 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4724 if (sbi->s_qf_names[i]) {
4725 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4726 GFP_KERNEL);
4727 if (!old_opts.s_qf_names[i]) {
4728 for (j = 0; j < i; j++)
4729 kfree(old_opts.s_qf_names[j]);
4730 kfree(orig_data);
4731 return -ENOMEM;
4732 }
4733 } else
4734 old_opts.s_qf_names[i] = NULL;
4735 #endif
4736 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4737 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4738
4739 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4740 err = -EINVAL;
4741 goto restore_opts;
4742 }
4743
4744 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4745 test_opt(sb, JOURNAL_CHECKSUM)) {
4746 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4747 "during remount not supported; ignoring");
4748 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4749 }
4750
4751 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4752 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4753 ext4_msg(sb, KERN_ERR, "can't mount with "
4754 "both data=journal and delalloc");
4755 err = -EINVAL;
4756 goto restore_opts;
4757 }
4758 if (test_opt(sb, DIOREAD_NOLOCK)) {
4759 ext4_msg(sb, KERN_ERR, "can't mount with "
4760 "both data=journal and dioread_nolock");
4761 err = -EINVAL;
4762 goto restore_opts;
4763 }
4764 if (test_opt(sb, DAX)) {
4765 ext4_msg(sb, KERN_ERR, "can't mount with "
4766 "both data=journal and dax");
4767 err = -EINVAL;
4768 goto restore_opts;
4769 }
4770 }
4771
4772 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4773 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4774 "dax flag with busy inodes while remounting");
4775 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4776 }
4777
4778 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4779 ext4_abort(sb, "Abort forced by user");
4780
4781 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4782 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4783
4784 es = sbi->s_es;
4785
4786 if (sbi->s_journal) {
4787 ext4_init_journal_params(sb, sbi->s_journal);
4788 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4789 }
4790
4791 if (*flags & MS_LAZYTIME)
4792 sb->s_flags |= MS_LAZYTIME;
4793
4794 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4795 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4796 err = -EROFS;
4797 goto restore_opts;
4798 }
4799
4800 if (*flags & MS_RDONLY) {
4801 err = sync_filesystem(sb);
4802 if (err < 0)
4803 goto restore_opts;
4804 err = dquot_suspend(sb, -1);
4805 if (err < 0)
4806 goto restore_opts;
4807
4808 /*
4809 * First of all, the unconditional stuff we have to do
4810 * to disable replay of the journal when we next remount
4811 */
4812 sb->s_flags |= MS_RDONLY;
4813
4814 /*
4815 * OK, test if we are remounting a valid rw partition
4816 * readonly, and if so set the rdonly flag and then
4817 * mark the partition as valid again.
4818 */
4819 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4820 (sbi->s_mount_state & EXT4_VALID_FS))
4821 es->s_state = cpu_to_le16(sbi->s_mount_state);
4822
4823 if (sbi->s_journal)
4824 ext4_mark_recovery_complete(sb, es);
4825 } else {
4826 /* Make sure we can mount this feature set readwrite */
4827 if (ext4_has_feature_readonly(sb) ||
4828 !ext4_feature_set_ok(sb, 0)) {
4829 err = -EROFS;
4830 goto restore_opts;
4831 }
4832 /*
4833 * Make sure the group descriptor checksums
4834 * are sane. If they aren't, refuse to remount r/w.
4835 */
4836 for (g = 0; g < sbi->s_groups_count; g++) {
4837 struct ext4_group_desc *gdp =
4838 ext4_get_group_desc(sb, g, NULL);
4839
4840 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4841 ext4_msg(sb, KERN_ERR,
4842 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4843 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4844 le16_to_cpu(gdp->bg_checksum));
4845 err = -EFSBADCRC;
4846 goto restore_opts;
4847 }
4848 }
4849
4850 /*
4851 * If we have an unprocessed orphan list hanging
4852 * around from a previously readonly bdev mount,
4853 * require a full umount/remount for now.
4854 */
4855 if (es->s_last_orphan) {
4856 ext4_msg(sb, KERN_WARNING, "Couldn't "
4857 "remount RDWR because of unprocessed "
4858 "orphan inode list. Please "
4859 "umount/remount instead");
4860 err = -EINVAL;
4861 goto restore_opts;
4862 }
4863
4864 /*
4865 * Mounting a RDONLY partition read-write, so reread
4866 * and store the current valid flag. (It may have
4867 * been changed by e2fsck since we originally mounted
4868 * the partition.)
4869 */
4870 if (sbi->s_journal)
4871 ext4_clear_journal_err(sb, es);
4872 sbi->s_mount_state = le16_to_cpu(es->s_state);
4873 if (!ext4_setup_super(sb, es, 0))
4874 sb->s_flags &= ~MS_RDONLY;
4875 if (ext4_has_feature_mmp(sb))
4876 if (ext4_multi_mount_protect(sb,
4877 le64_to_cpu(es->s_mmp_block))) {
4878 err = -EROFS;
4879 goto restore_opts;
4880 }
4881 enable_quota = 1;
4882 }
4883 }
4884
4885 /*
4886 * Reinitialize lazy itable initialization thread based on
4887 * current settings
4888 */
4889 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4890 ext4_unregister_li_request(sb);
4891 else {
4892 ext4_group_t first_not_zeroed;
4893 first_not_zeroed = ext4_has_uninit_itable(sb);
4894 ext4_register_li_request(sb, first_not_zeroed);
4895 }
4896
4897 ext4_setup_system_zone(sb);
4898 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4899 ext4_commit_super(sb, 1);
4900
4901 #ifdef CONFIG_QUOTA
4902 /* Release old quota file names */
4903 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4904 kfree(old_opts.s_qf_names[i]);
4905 if (enable_quota) {
4906 if (sb_any_quota_suspended(sb))
4907 dquot_resume(sb, -1);
4908 else if (ext4_has_feature_quota(sb)) {
4909 err = ext4_enable_quotas(sb);
4910 if (err)
4911 goto restore_opts;
4912 }
4913 }
4914 #endif
4915
4916 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4917 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4918 kfree(orig_data);
4919 return 0;
4920
4921 restore_opts:
4922 sb->s_flags = old_sb_flags;
4923 sbi->s_mount_opt = old_opts.s_mount_opt;
4924 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4925 sbi->s_resuid = old_opts.s_resuid;
4926 sbi->s_resgid = old_opts.s_resgid;
4927 sbi->s_commit_interval = old_opts.s_commit_interval;
4928 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4929 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4930 #ifdef CONFIG_QUOTA
4931 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4932 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
4933 kfree(sbi->s_qf_names[i]);
4934 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4935 }
4936 #endif
4937 kfree(orig_data);
4938 return err;
4939 }
4940
4941 #ifdef CONFIG_QUOTA
4942 static int ext4_statfs_project(struct super_block *sb,
4943 kprojid_t projid, struct kstatfs *buf)
4944 {
4945 struct kqid qid;
4946 struct dquot *dquot;
4947 u64 limit;
4948 u64 curblock;
4949
4950 qid = make_kqid_projid(projid);
4951 dquot = dqget(sb, qid);
4952 if (IS_ERR(dquot))
4953 return PTR_ERR(dquot);
4954 spin_lock(&dq_data_lock);
4955
4956 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
4957 dquot->dq_dqb.dqb_bsoftlimit :
4958 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
4959 if (limit && buf->f_blocks > limit) {
4960 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
4961 buf->f_blocks = limit;
4962 buf->f_bfree = buf->f_bavail =
4963 (buf->f_blocks > curblock) ?
4964 (buf->f_blocks - curblock) : 0;
4965 }
4966
4967 limit = dquot->dq_dqb.dqb_isoftlimit ?
4968 dquot->dq_dqb.dqb_isoftlimit :
4969 dquot->dq_dqb.dqb_ihardlimit;
4970 if (limit && buf->f_files > limit) {
4971 buf->f_files = limit;
4972 buf->f_ffree =
4973 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
4974 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
4975 }
4976
4977 spin_unlock(&dq_data_lock);
4978 dqput(dquot);
4979 return 0;
4980 }
4981 #endif
4982
4983 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4984 {
4985 struct super_block *sb = dentry->d_sb;
4986 struct ext4_sb_info *sbi = EXT4_SB(sb);
4987 struct ext4_super_block *es = sbi->s_es;
4988 ext4_fsblk_t overhead = 0, resv_blocks;
4989 u64 fsid;
4990 s64 bfree;
4991 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4992
4993 if (!test_opt(sb, MINIX_DF))
4994 overhead = sbi->s_overhead;
4995
4996 buf->f_type = EXT4_SUPER_MAGIC;
4997 buf->f_bsize = sb->s_blocksize;
4998 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4999 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5000 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5001 /* prevent underflow in case that few free space is available */
5002 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5003 buf->f_bavail = buf->f_bfree -
5004 (ext4_r_blocks_count(es) + resv_blocks);
5005 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5006 buf->f_bavail = 0;
5007 buf->f_files = le32_to_cpu(es->s_inodes_count);
5008 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5009 buf->f_namelen = EXT4_NAME_LEN;
5010 fsid = le64_to_cpup((void *)es->s_uuid) ^
5011 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5012 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5013 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5014
5015 #ifdef CONFIG_QUOTA
5016 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5017 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5018 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5019 #endif
5020 return 0;
5021 }
5022
5023 /* Helper function for writing quotas on sync - we need to start transaction
5024 * before quota file is locked for write. Otherwise the are possible deadlocks:
5025 * Process 1 Process 2
5026 * ext4_create() quota_sync()
5027 * jbd2_journal_start() write_dquot()
5028 * dquot_initialize() down(dqio_mutex)
5029 * down(dqio_mutex) jbd2_journal_start()
5030 *
5031 */
5032
5033 #ifdef CONFIG_QUOTA
5034
5035 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5036 {
5037 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5038 }
5039
5040 static int ext4_write_dquot(struct dquot *dquot)
5041 {
5042 int ret, err;
5043 handle_t *handle;
5044 struct inode *inode;
5045
5046 inode = dquot_to_inode(dquot);
5047 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5048 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5049 if (IS_ERR(handle))
5050 return PTR_ERR(handle);
5051 ret = dquot_commit(dquot);
5052 err = ext4_journal_stop(handle);
5053 if (!ret)
5054 ret = err;
5055 return ret;
5056 }
5057
5058 static int ext4_acquire_dquot(struct dquot *dquot)
5059 {
5060 int ret, err;
5061 handle_t *handle;
5062
5063 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5064 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5065 if (IS_ERR(handle))
5066 return PTR_ERR(handle);
5067 ret = dquot_acquire(dquot);
5068 err = ext4_journal_stop(handle);
5069 if (!ret)
5070 ret = err;
5071 return ret;
5072 }
5073
5074 static int ext4_release_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_DEL_BLOCKS(dquot->dq_sb));
5081 if (IS_ERR(handle)) {
5082 /* Release dquot anyway to avoid endless cycle in dqput() */
5083 dquot_release(dquot);
5084 return PTR_ERR(handle);
5085 }
5086 ret = dquot_release(dquot);
5087 err = ext4_journal_stop(handle);
5088 if (!ret)
5089 ret = err;
5090 return ret;
5091 }
5092
5093 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5094 {
5095 struct super_block *sb = dquot->dq_sb;
5096 struct ext4_sb_info *sbi = EXT4_SB(sb);
5097
5098 /* Are we journaling quotas? */
5099 if (ext4_has_feature_quota(sb) ||
5100 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5101 dquot_mark_dquot_dirty(dquot);
5102 return ext4_write_dquot(dquot);
5103 } else {
5104 return dquot_mark_dquot_dirty(dquot);
5105 }
5106 }
5107
5108 static int ext4_write_info(struct super_block *sb, int type)
5109 {
5110 int ret, err;
5111 handle_t *handle;
5112
5113 /* Data block + inode block */
5114 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5115 if (IS_ERR(handle))
5116 return PTR_ERR(handle);
5117 ret = dquot_commit_info(sb, type);
5118 err = ext4_journal_stop(handle);
5119 if (!ret)
5120 ret = err;
5121 return ret;
5122 }
5123
5124 /*
5125 * Turn on quotas during mount time - we need to find
5126 * the quota file and such...
5127 */
5128 static int ext4_quota_on_mount(struct super_block *sb, int type)
5129 {
5130 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5131 EXT4_SB(sb)->s_jquota_fmt, type);
5132 }
5133
5134 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5135 {
5136 struct ext4_inode_info *ei = EXT4_I(inode);
5137
5138 /* The first argument of lockdep_set_subclass has to be
5139 * *exactly* the same as the argument to init_rwsem() --- in
5140 * this case, in init_once() --- or lockdep gets unhappy
5141 * because the name of the lock is set using the
5142 * stringification of the argument to init_rwsem().
5143 */
5144 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5145 lockdep_set_subclass(&ei->i_data_sem, subclass);
5146 }
5147
5148 /*
5149 * Standard function to be called on quota_on
5150 */
5151 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5152 struct path *path)
5153 {
5154 int err;
5155
5156 if (!test_opt(sb, QUOTA))
5157 return -EINVAL;
5158
5159 /* Quotafile not on the same filesystem? */
5160 if (path->dentry->d_sb != sb)
5161 return -EXDEV;
5162 /* Journaling quota? */
5163 if (EXT4_SB(sb)->s_qf_names[type]) {
5164 /* Quotafile not in fs root? */
5165 if (path->dentry->d_parent != sb->s_root)
5166 ext4_msg(sb, KERN_WARNING,
5167 "Quota file not on filesystem root. "
5168 "Journaled quota will not work");
5169 }
5170
5171 /*
5172 * When we journal data on quota file, we have to flush journal to see
5173 * all updates to the file when we bypass pagecache...
5174 */
5175 if (EXT4_SB(sb)->s_journal &&
5176 ext4_should_journal_data(d_inode(path->dentry))) {
5177 /*
5178 * We don't need to lock updates but journal_flush() could
5179 * otherwise be livelocked...
5180 */
5181 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5182 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5183 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5184 if (err)
5185 return err;
5186 }
5187 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5188 err = dquot_quota_on(sb, type, format_id, path);
5189 if (err)
5190 lockdep_set_quota_inode(path->dentry->d_inode,
5191 I_DATA_SEM_NORMAL);
5192 return err;
5193 }
5194
5195 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5196 unsigned int flags)
5197 {
5198 int err;
5199 struct inode *qf_inode;
5200 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5201 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5202 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5203 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5204 };
5205
5206 BUG_ON(!ext4_has_feature_quota(sb));
5207
5208 if (!qf_inums[type])
5209 return -EPERM;
5210
5211 qf_inode = ext4_iget(sb, qf_inums[type]);
5212 if (IS_ERR(qf_inode)) {
5213 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5214 return PTR_ERR(qf_inode);
5215 }
5216
5217 /* Don't account quota for quota files to avoid recursion */
5218 qf_inode->i_flags |= S_NOQUOTA;
5219 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5220 err = dquot_enable(qf_inode, type, format_id, flags);
5221 iput(qf_inode);
5222 if (err)
5223 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5224
5225 return err;
5226 }
5227
5228 /* Enable usage tracking for all quota types. */
5229 static int ext4_enable_quotas(struct super_block *sb)
5230 {
5231 int type, err = 0;
5232 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5233 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5234 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5235 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5236 };
5237
5238 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5239 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5240 if (qf_inums[type]) {
5241 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5242 DQUOT_USAGE_ENABLED);
5243 if (err) {
5244 ext4_warning(sb,
5245 "Failed to enable quota tracking "
5246 "(type=%d, err=%d). Please run "
5247 "e2fsck to fix.", type, err);
5248 return err;
5249 }
5250 }
5251 }
5252 return 0;
5253 }
5254
5255 static int ext4_quota_off(struct super_block *sb, int type)
5256 {
5257 struct inode *inode = sb_dqopt(sb)->files[type];
5258 handle_t *handle;
5259
5260 /* Force all delayed allocation blocks to be allocated.
5261 * Caller already holds s_umount sem */
5262 if (test_opt(sb, DELALLOC))
5263 sync_filesystem(sb);
5264
5265 if (!inode)
5266 goto out;
5267
5268 /* Update modification times of quota files when userspace can
5269 * start looking at them */
5270 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5271 if (IS_ERR(handle))
5272 goto out;
5273 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5274 ext4_mark_inode_dirty(handle, inode);
5275 ext4_journal_stop(handle);
5276
5277 out:
5278 return dquot_quota_off(sb, type);
5279 }
5280
5281 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5282 * acquiring the locks... As quota files are never truncated and quota code
5283 * itself serializes the operations (and no one else should touch the files)
5284 * we don't have to be afraid of races */
5285 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5286 size_t len, loff_t off)
5287 {
5288 struct inode *inode = sb_dqopt(sb)->files[type];
5289 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5290 int offset = off & (sb->s_blocksize - 1);
5291 int tocopy;
5292 size_t toread;
5293 struct buffer_head *bh;
5294 loff_t i_size = i_size_read(inode);
5295
5296 if (off > i_size)
5297 return 0;
5298 if (off+len > i_size)
5299 len = i_size-off;
5300 toread = len;
5301 while (toread > 0) {
5302 tocopy = sb->s_blocksize - offset < toread ?
5303 sb->s_blocksize - offset : toread;
5304 bh = ext4_bread(NULL, inode, blk, 0);
5305 if (IS_ERR(bh))
5306 return PTR_ERR(bh);
5307 if (!bh) /* A hole? */
5308 memset(data, 0, tocopy);
5309 else
5310 memcpy(data, bh->b_data+offset, tocopy);
5311 brelse(bh);
5312 offset = 0;
5313 toread -= tocopy;
5314 data += tocopy;
5315 blk++;
5316 }
5317 return len;
5318 }
5319
5320 /* Write to quotafile (we know the transaction is already started and has
5321 * enough credits) */
5322 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5323 const char *data, size_t len, loff_t off)
5324 {
5325 struct inode *inode = sb_dqopt(sb)->files[type];
5326 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5327 int err, offset = off & (sb->s_blocksize - 1);
5328 int retries = 0;
5329 struct buffer_head *bh;
5330 handle_t *handle = journal_current_handle();
5331
5332 if (EXT4_SB(sb)->s_journal && !handle) {
5333 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5334 " cancelled because transaction is not started",
5335 (unsigned long long)off, (unsigned long long)len);
5336 return -EIO;
5337 }
5338 /*
5339 * Since we account only one data block in transaction credits,
5340 * then it is impossible to cross a block boundary.
5341 */
5342 if (sb->s_blocksize - offset < len) {
5343 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5344 " cancelled because not block aligned",
5345 (unsigned long long)off, (unsigned long long)len);
5346 return -EIO;
5347 }
5348
5349 do {
5350 bh = ext4_bread(handle, inode, blk,
5351 EXT4_GET_BLOCKS_CREATE |
5352 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5353 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5354 ext4_should_retry_alloc(inode->i_sb, &retries));
5355 if (IS_ERR(bh))
5356 return PTR_ERR(bh);
5357 if (!bh)
5358 goto out;
5359 BUFFER_TRACE(bh, "get write access");
5360 err = ext4_journal_get_write_access(handle, bh);
5361 if (err) {
5362 brelse(bh);
5363 return err;
5364 }
5365 lock_buffer(bh);
5366 memcpy(bh->b_data+offset, data, len);
5367 flush_dcache_page(bh->b_page);
5368 unlock_buffer(bh);
5369 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5370 brelse(bh);
5371 out:
5372 if (inode->i_size < off + len) {
5373 i_size_write(inode, off + len);
5374 EXT4_I(inode)->i_disksize = inode->i_size;
5375 ext4_mark_inode_dirty(handle, inode);
5376 }
5377 return len;
5378 }
5379
5380 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5381 {
5382 const struct quota_format_ops *ops;
5383
5384 if (!sb_has_quota_loaded(sb, qid->type))
5385 return -ESRCH;
5386 ops = sb_dqopt(sb)->ops[qid->type];
5387 if (!ops || !ops->get_next_id)
5388 return -ENOSYS;
5389 return dquot_get_next_id(sb, qid);
5390 }
5391 #endif
5392
5393 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5394 const char *dev_name, void *data)
5395 {
5396 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5397 }
5398
5399 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5400 static inline void register_as_ext2(void)
5401 {
5402 int err = register_filesystem(&ext2_fs_type);
5403 if (err)
5404 printk(KERN_WARNING
5405 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5406 }
5407
5408 static inline void unregister_as_ext2(void)
5409 {
5410 unregister_filesystem(&ext2_fs_type);
5411 }
5412
5413 static inline int ext2_feature_set_ok(struct super_block *sb)
5414 {
5415 if (ext4_has_unknown_ext2_incompat_features(sb))
5416 return 0;
5417 if (sb->s_flags & MS_RDONLY)
5418 return 1;
5419 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5420 return 0;
5421 return 1;
5422 }
5423 #else
5424 static inline void register_as_ext2(void) { }
5425 static inline void unregister_as_ext2(void) { }
5426 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5427 #endif
5428
5429 static inline void register_as_ext3(void)
5430 {
5431 int err = register_filesystem(&ext3_fs_type);
5432 if (err)
5433 printk(KERN_WARNING
5434 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5435 }
5436
5437 static inline void unregister_as_ext3(void)
5438 {
5439 unregister_filesystem(&ext3_fs_type);
5440 }
5441
5442 static inline int ext3_feature_set_ok(struct super_block *sb)
5443 {
5444 if (ext4_has_unknown_ext3_incompat_features(sb))
5445 return 0;
5446 if (!ext4_has_feature_journal(sb))
5447 return 0;
5448 if (sb->s_flags & MS_RDONLY)
5449 return 1;
5450 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5451 return 0;
5452 return 1;
5453 }
5454
5455 static struct file_system_type ext4_fs_type = {
5456 .owner = THIS_MODULE,
5457 .name = "ext4",
5458 .mount = ext4_mount,
5459 .kill_sb = kill_block_super,
5460 .fs_flags = FS_REQUIRES_DEV,
5461 };
5462 MODULE_ALIAS_FS("ext4");
5463
5464 /* Shared across all ext4 file systems */
5465 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5466
5467 static int __init ext4_init_fs(void)
5468 {
5469 int i, err;
5470
5471 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5472 ext4_li_info = NULL;
5473 mutex_init(&ext4_li_mtx);
5474
5475 /* Build-time check for flags consistency */
5476 ext4_check_flag_values();
5477
5478 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5479 init_waitqueue_head(&ext4__ioend_wq[i]);
5480
5481 err = ext4_init_es();
5482 if (err)
5483 return err;
5484
5485 err = ext4_init_pageio();
5486 if (err)
5487 goto out5;
5488
5489 err = ext4_init_system_zone();
5490 if (err)
5491 goto out4;
5492
5493 err = ext4_init_sysfs();
5494 if (err)
5495 goto out3;
5496
5497 err = ext4_init_mballoc();
5498 if (err)
5499 goto out2;
5500 err = init_inodecache();
5501 if (err)
5502 goto out1;
5503 register_as_ext3();
5504 register_as_ext2();
5505 err = register_filesystem(&ext4_fs_type);
5506 if (err)
5507 goto out;
5508
5509 return 0;
5510 out:
5511 unregister_as_ext2();
5512 unregister_as_ext3();
5513 destroy_inodecache();
5514 out1:
5515 ext4_exit_mballoc();
5516 out2:
5517 ext4_exit_sysfs();
5518 out3:
5519 ext4_exit_system_zone();
5520 out4:
5521 ext4_exit_pageio();
5522 out5:
5523 ext4_exit_es();
5524
5525 return err;
5526 }
5527
5528 static void __exit ext4_exit_fs(void)
5529 {
5530 ext4_destroy_lazyinit_thread();
5531 unregister_as_ext2();
5532 unregister_as_ext3();
5533 unregister_filesystem(&ext4_fs_type);
5534 destroy_inodecache();
5535 ext4_exit_mballoc();
5536 ext4_exit_sysfs();
5537 ext4_exit_system_zone();
5538 ext4_exit_pageio();
5539 ext4_exit_es();
5540 }
5541
5542 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5543 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5544 MODULE_LICENSE("GPL");
5545 module_init(ext4_init_fs)
5546 module_exit(ext4_exit_fs)
Ubuntu Artful kernel mirror