1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * linux/fs/ext4/super.c
4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/fs/minix/inode.c
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 *
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
18 */
20 #include <linux/module.h>
21 #include <linux/string.h>
22 #include <linux/fs.h>
23 #include <linux/time.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/dax.h>
42 #include <linux/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
46 #include <linux/kthread.h>
47 #include <linux/freezer.h>
49 #include "ext4.h"
50 #include "ext4_extents.h" /* Needed for trace points definition */
51 #include "ext4_jbd2.h"
52 #include "xattr.h"
53 #include "acl.h"
54 #include "mballoc.h"
55 #include "fsmap.h"
57 #define CREATE_TRACE_POINTS
58 #include <trace/events/ext4.h>
60 static struct ext4_lazy_init *ext4_li_info;
61 static struct mutex ext4_li_mtx;
62 static struct ratelimit_state ext4_mount_msg_ratelimit;
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static struct inode *ext4_get_journal_inode(struct super_block *sb,
86 unsigned int journal_inum);
88 /*
89 * Lock ordering
90 *
91 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
92 * i_mmap_rwsem (inode->i_mmap_rwsem)!
93 *
94 * page fault path:
95 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
96 * page lock -> i_data_sem (rw)
97 *
98 * buffered write path:
99 * sb_start_write -> i_mutex -> mmap_sem
100 * sb_start_write -> i_mutex -> transaction start -> page lock ->
101 * i_data_sem (rw)
102 *
103 * truncate:
104 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
106 * i_data_sem (rw)
107 *
108 * direct IO:
109 * sb_start_write -> i_mutex -> mmap_sem
110 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
111 *
112 * writepages:
113 * transaction start -> page lock(s) -> i_data_sem (rw)
114 */
116 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
117 static struct file_system_type ext2_fs_type = {
118 .owner = THIS_MODULE,
119 .name = "ext2",
120 .mount = ext4_mount,
121 .kill_sb = kill_block_super,
122 .fs_flags = FS_REQUIRES_DEV,
123 };
124 MODULE_ALIAS_FS("ext2");
125 MODULE_ALIAS("ext2");
126 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
127 #else
128 #define IS_EXT2_SB(sb) (0)
129 #endif
132 static struct file_system_type ext3_fs_type = {
133 .owner = THIS_MODULE,
134 .name = "ext3",
135 .mount = ext4_mount,
136 .kill_sb = kill_block_super,
137 .fs_flags = FS_REQUIRES_DEV,
138 };
139 MODULE_ALIAS_FS("ext3");
140 MODULE_ALIAS("ext3");
141 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
143 static int ext4_verify_csum_type(struct super_block *sb,
144 struct ext4_super_block *es)
145 {
146 if (!ext4_has_feature_metadata_csum(sb))
147 return 1;
149 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
150 }
152 static __le32 ext4_superblock_csum(struct super_block *sb,
153 struct ext4_super_block *es)
154 {
155 struct ext4_sb_info *sbi = EXT4_SB(sb);
156 int offset = offsetof(struct ext4_super_block, s_checksum);
157 __u32 csum;
159 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
161 return cpu_to_le32(csum);
162 }
164 static int ext4_superblock_csum_verify(struct super_block *sb,
165 struct ext4_super_block *es)
166 {
167 if (!ext4_has_metadata_csum(sb))
168 return 1;
170 return es->s_checksum == ext4_superblock_csum(sb, es);
171 }
173 void ext4_superblock_csum_set(struct super_block *sb)
174 {
175 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
177 if (!ext4_has_metadata_csum(sb))
178 return;
180 es->s_checksum = ext4_superblock_csum(sb, es);
181 }
183 void *ext4_kvmalloc(size_t size, gfp_t flags)
184 {
185 void *ret;
187 ret = kmalloc(size, flags | __GFP_NOWARN);
188 if (!ret)
189 ret = __vmalloc(size, flags, PAGE_KERNEL);
190 return ret;
191 }
193 void *ext4_kvzalloc(size_t size, gfp_t flags)
194 {
195 void *ret;
197 ret = kzalloc(size, flags | __GFP_NOWARN);
198 if (!ret)
199 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
200 return ret;
201 }
203 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
204 struct ext4_group_desc *bg)
205 {
206 return le32_to_cpu(bg->bg_block_bitmap_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
209 }
211 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
212 struct ext4_group_desc *bg)
213 {
214 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
217 }
219 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
220 struct ext4_group_desc *bg)
221 {
222 return le32_to_cpu(bg->bg_inode_table_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
225 }
227 __u32 ext4_free_group_clusters(struct super_block *sb,
228 struct ext4_group_desc *bg)
229 {
230 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
233 }
235 __u32 ext4_free_inodes_count(struct super_block *sb,
236 struct ext4_group_desc *bg)
237 {
238 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
239 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
240 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
241 }
243 __u32 ext4_used_dirs_count(struct super_block *sb,
244 struct ext4_group_desc *bg)
245 {
246 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
247 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
248 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
249 }
251 __u32 ext4_itable_unused_count(struct super_block *sb,
252 struct ext4_group_desc *bg)
253 {
254 return le16_to_cpu(bg->bg_itable_unused_lo) |
255 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
256 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
257 }
259 void ext4_block_bitmap_set(struct super_block *sb,
260 struct ext4_group_desc *bg, ext4_fsblk_t blk)
261 {
262 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
265 }
267 void ext4_inode_bitmap_set(struct super_block *sb,
268 struct ext4_group_desc *bg, ext4_fsblk_t blk)
269 {
270 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
273 }
275 void ext4_inode_table_set(struct super_block *sb,
276 struct ext4_group_desc *bg, ext4_fsblk_t blk)
277 {
278 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
281 }
283 void ext4_free_group_clusters_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
285 {
286 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
289 }
291 void ext4_free_inodes_set(struct super_block *sb,
292 struct ext4_group_desc *bg, __u32 count)
293 {
294 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
295 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
296 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
297 }
299 void ext4_used_dirs_set(struct super_block *sb,
300 struct ext4_group_desc *bg, __u32 count)
301 {
302 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
303 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
304 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
305 }
307 void ext4_itable_unused_set(struct super_block *sb,
308 struct ext4_group_desc *bg, __u32 count)
309 {
310 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
311 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
312 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
313 }
315 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
316 {
317 time64_t now = ktime_get_real_seconds();
319 now = clamp_val(now, 0, (1ull << 40) - 1);
321 *lo = cpu_to_le32(lower_32_bits(now));
322 *hi = upper_32_bits(now);
323 }
325 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
326 {
327 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
328 }
329 #define ext4_update_tstamp(es, tstamp) \
330 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
331 #define ext4_get_tstamp(es, tstamp) \
332 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
334 static void __save_error_info(struct super_block *sb, const char *func,
335 unsigned int line)
336 {
337 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
339 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
340 if (bdev_read_only(sb->s_bdev))
341 return;
342 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
343 ext4_update_tstamp(es, s_last_error_time);
344 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
345 es->s_last_error_line = cpu_to_le32(line);
346 if (!es->s_first_error_time) {
347 es->s_first_error_time = es->s_last_error_time;
348 es->s_first_error_time_hi = es->s_last_error_time_hi;
349 strncpy(es->s_first_error_func, func,
350 sizeof(es->s_first_error_func));
351 es->s_first_error_line = cpu_to_le32(line);
352 es->s_first_error_ino = es->s_last_error_ino;
353 es->s_first_error_block = es->s_last_error_block;
354 }
355 /*
356 * Start the daily error reporting function if it hasn't been
357 * started already
358 */
359 if (!es->s_error_count)
360 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
361 le32_add_cpu(&es->s_error_count, 1);
362 }
364 static void save_error_info(struct super_block *sb, const char *func,
365 unsigned int line)
366 {
367 __save_error_info(sb, func, line);
368 ext4_commit_super(sb, 1);
369 }
371 /*
372 * The del_gendisk() function uninitializes the disk-specific data
373 * structures, including the bdi structure, without telling anyone
374 * else. Once this happens, any attempt to call mark_buffer_dirty()
375 * (for example, by ext4_commit_super), will cause a kernel OOPS.
376 * This is a kludge to prevent these oops until we can put in a proper
377 * hook in del_gendisk() to inform the VFS and file system layers.
378 */
379 static int block_device_ejected(struct super_block *sb)
380 {
381 struct inode *bd_inode = sb->s_bdev->bd_inode;
382 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
384 return bdi->dev == NULL;
385 }
387 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
388 {
389 struct super_block *sb = journal->j_private;
390 struct ext4_sb_info *sbi = EXT4_SB(sb);
391 int error = is_journal_aborted(journal);
392 struct ext4_journal_cb_entry *jce;
394 BUG_ON(txn->t_state == T_FINISHED);
396 ext4_process_freed_data(sb, txn->t_tid);
398 spin_lock(&sbi->s_md_lock);
399 while (!list_empty(&txn->t_private_list)) {
400 jce = list_entry(txn->t_private_list.next,
401 struct ext4_journal_cb_entry, jce_list);
402 list_del_init(&jce->jce_list);
403 spin_unlock(&sbi->s_md_lock);
404 jce->jce_func(sb, jce, error);
405 spin_lock(&sbi->s_md_lock);
406 }
407 spin_unlock(&sbi->s_md_lock);
408 }
410 /* Deal with the reporting of failure conditions on a filesystem such as
411 * inconsistencies detected or read IO failures.
412 *
413 * On ext2, we can store the error state of the filesystem in the
414 * superblock. That is not possible on ext4, because we may have other
415 * write ordering constraints on the superblock which prevent us from
416 * writing it out straight away; and given that the journal is about to
417 * be aborted, we can't rely on the current, or future, transactions to
418 * write out the superblock safely.
419 *
420 * We'll just use the jbd2_journal_abort() error code to record an error in
421 * the journal instead. On recovery, the journal will complain about
422 * that error until we've noted it down and cleared it.
423 */
425 static void ext4_handle_error(struct super_block *sb)
426 {
427 if (test_opt(sb, WARN_ON_ERROR))
428 WARN_ON_ONCE(1);
430 if (sb_rdonly(sb))
431 return;
433 if (!test_opt(sb, ERRORS_CONT)) {
434 journal_t *journal = EXT4_SB(sb)->s_journal;
436 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
437 if (journal)
438 jbd2_journal_abort(journal, -EIO);
439 }
440 if (test_opt(sb, ERRORS_RO)) {
441 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
442 /*
443 * Make sure updated value of ->s_mount_flags will be visible
444 * before ->s_flags update
445 */
446 smp_wmb();
447 sb->s_flags |= SB_RDONLY;
448 }
449 if (test_opt(sb, ERRORS_PANIC)) {
450 if (EXT4_SB(sb)->s_journal &&
451 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
452 return;
453 panic("EXT4-fs (device %s): panic forced after error\n",
454 sb->s_id);
455 }
456 }
458 #define ext4_error_ratelimit(sb) \
459 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
460 "EXT4-fs error")
462 void __ext4_error(struct super_block *sb, const char *function,
463 unsigned int line, const char *fmt, ...)
464 {
465 struct va_format vaf;
466 va_list args;
468 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
469 return;
471 trace_ext4_error(sb, function, line);
472 if (ext4_error_ratelimit(sb)) {
473 va_start(args, fmt);
474 vaf.fmt = fmt;
475 vaf.va = &args;
476 printk(KERN_CRIT
477 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
478 sb->s_id, function, line, current->comm, &vaf);
479 va_end(args);
480 }
481 save_error_info(sb, function, line);
482 ext4_handle_error(sb);
483 }
485 void __ext4_error_inode(struct inode *inode, const char *function,
486 unsigned int line, ext4_fsblk_t block,
487 const char *fmt, ...)
488 {
489 va_list args;
490 struct va_format vaf;
491 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
493 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
494 return;
496 trace_ext4_error(inode->i_sb, function, line);
497 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
498 es->s_last_error_block = cpu_to_le64(block);
499 if (ext4_error_ratelimit(inode->i_sb)) {
500 va_start(args, fmt);
501 vaf.fmt = fmt;
502 vaf.va = &args;
503 if (block)
504 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
505 "inode #%lu: block %llu: comm %s: %pV\n",
506 inode->i_sb->s_id, function, line, inode->i_ino,
507 block, current->comm, &vaf);
508 else
509 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
510 "inode #%lu: comm %s: %pV\n",
511 inode->i_sb->s_id, function, line, inode->i_ino,
512 current->comm, &vaf);
513 va_end(args);
514 }
515 save_error_info(inode->i_sb, function, line);
516 ext4_handle_error(inode->i_sb);
517 }
519 void __ext4_error_file(struct file *file, const char *function,
520 unsigned int line, ext4_fsblk_t block,
521 const char *fmt, ...)
522 {
523 va_list args;
524 struct va_format vaf;
525 struct ext4_super_block *es;
526 struct inode *inode = file_inode(file);
527 char pathname[80], *path;
529 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
530 return;
532 trace_ext4_error(inode->i_sb, function, line);
533 es = EXT4_SB(inode->i_sb)->s_es;
534 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
535 if (ext4_error_ratelimit(inode->i_sb)) {
536 path = file_path(file, pathname, sizeof(pathname));
537 if (IS_ERR(path))
538 path = "(unknown)";
539 va_start(args, fmt);
540 vaf.fmt = fmt;
541 vaf.va = &args;
542 if (block)
543 printk(KERN_CRIT
544 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
545 "block %llu: comm %s: path %s: %pV\n",
546 inode->i_sb->s_id, function, line, inode->i_ino,
547 block, current->comm, path, &vaf);
548 else
549 printk(KERN_CRIT
550 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
551 "comm %s: path %s: %pV\n",
552 inode->i_sb->s_id, function, line, inode->i_ino,
553 current->comm, path, &vaf);
554 va_end(args);
555 }
556 save_error_info(inode->i_sb, function, line);
557 ext4_handle_error(inode->i_sb);
558 }
560 const char *ext4_decode_error(struct super_block *sb, int errno,
561 char nbuf[16])
562 {
563 char *errstr = NULL;
565 switch (errno) {
566 case -EFSCORRUPTED:
567 errstr = "Corrupt filesystem";
568 break;
569 case -EFSBADCRC:
570 errstr = "Filesystem failed CRC";
571 break;
572 case -EIO:
573 errstr = "IO failure";
574 break;
575 case -ENOMEM:
576 errstr = "Out of memory";
577 break;
578 case -EROFS:
579 if (!sb || (EXT4_SB(sb)->s_journal &&
580 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
581 errstr = "Journal has aborted";
582 else
583 errstr = "Readonly filesystem";
584 break;
585 default:
586 /* If the caller passed in an extra buffer for unknown
587 * errors, textualise them now. Else we just return
588 * NULL. */
589 if (nbuf) {
590 /* Check for truncated error codes... */
591 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
592 errstr = nbuf;
593 }
594 break;
595 }
597 return errstr;
598 }
600 /* __ext4_std_error decodes expected errors from journaling functions
601 * automatically and invokes the appropriate error response. */
603 void __ext4_std_error(struct super_block *sb, const char *function,
604 unsigned int line, int errno)
605 {
606 char nbuf[16];
607 const char *errstr;
609 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
610 return;
612 /* Special case: if the error is EROFS, and we're not already
613 * inside a transaction, then there's really no point in logging
614 * an error. */
615 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
616 return;
618 if (ext4_error_ratelimit(sb)) {
619 errstr = ext4_decode_error(sb, errno, nbuf);
620 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
621 sb->s_id, function, line, errstr);
622 }
624 save_error_info(sb, function, line);
625 ext4_handle_error(sb);
626 }
628 /*
629 * ext4_abort is a much stronger failure handler than ext4_error. The
630 * abort function may be used to deal with unrecoverable failures such
631 * as journal IO errors or ENOMEM at a critical moment in log management.
632 *
633 * We unconditionally force the filesystem into an ABORT|READONLY state,
634 * unless the error response on the fs has been set to panic in which
635 * case we take the easy way out and panic immediately.
636 */
638 void __ext4_abort(struct super_block *sb, const char *function,
639 unsigned int line, const char *fmt, ...)
640 {
641 struct va_format vaf;
642 va_list args;
644 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
645 return;
647 save_error_info(sb, function, line);
648 va_start(args, fmt);
649 vaf.fmt = fmt;
650 vaf.va = &args;
651 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
652 sb->s_id, function, line, &vaf);
653 va_end(args);
655 if (sb_rdonly(sb) == 0) {
656 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
657 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
658 /*
659 * Make sure updated value of ->s_mount_flags will be visible
660 * before ->s_flags update
661 */
662 smp_wmb();
663 sb->s_flags |= SB_RDONLY;
664 if (EXT4_SB(sb)->s_journal)
665 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
666 save_error_info(sb, function, line);
667 }
668 if (test_opt(sb, ERRORS_PANIC)) {
669 if (EXT4_SB(sb)->s_journal &&
670 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
671 return;
672 panic("EXT4-fs panic from previous error\n");
673 }
674 }
676 void __ext4_msg(struct super_block *sb,
677 const char *prefix, const char *fmt, ...)
678 {
679 struct va_format vaf;
680 va_list args;
682 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
683 return;
685 va_start(args, fmt);
686 vaf.fmt = fmt;
687 vaf.va = &args;
688 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
689 va_end(args);
690 }
692 #define ext4_warning_ratelimit(sb) \
693 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
694 "EXT4-fs warning")
696 void __ext4_warning(struct super_block *sb, const char *function,
697 unsigned int line, const char *fmt, ...)
698 {
699 struct va_format vaf;
700 va_list args;
702 if (!ext4_warning_ratelimit(sb))
703 return;
705 va_start(args, fmt);
706 vaf.fmt = fmt;
707 vaf.va = &args;
708 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
709 sb->s_id, function, line, &vaf);
710 va_end(args);
711 }
713 void __ext4_warning_inode(const struct inode *inode, const char *function,
714 unsigned int line, const char *fmt, ...)
715 {
716 struct va_format vaf;
717 va_list args;
719 if (!ext4_warning_ratelimit(inode->i_sb))
720 return;
722 va_start(args, fmt);
723 vaf.fmt = fmt;
724 vaf.va = &args;
725 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
726 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
727 function, line, inode->i_ino, current->comm, &vaf);
728 va_end(args);
729 }
731 void __ext4_grp_locked_error(const char *function, unsigned int line,
732 struct super_block *sb, ext4_group_t grp,
733 unsigned long ino, ext4_fsblk_t block,
734 const char *fmt, ...)
735 __releases(bitlock)
736 __acquires(bitlock)
737 {
738 struct va_format vaf;
739 va_list args;
740 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
742 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
743 return;
745 trace_ext4_error(sb, function, line);
746 es->s_last_error_ino = cpu_to_le32(ino);
747 es->s_last_error_block = cpu_to_le64(block);
748 __save_error_info(sb, function, line);
750 if (ext4_error_ratelimit(sb)) {
751 va_start(args, fmt);
752 vaf.fmt = fmt;
753 vaf.va = &args;
754 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
755 sb->s_id, function, line, grp);
756 if (ino)
757 printk(KERN_CONT "inode %lu: ", ino);
758 if (block)
759 printk(KERN_CONT "block %llu:",
760 (unsigned long long) block);
761 printk(KERN_CONT "%pV\n", &vaf);
762 va_end(args);
763 }
765 if (test_opt(sb, WARN_ON_ERROR))
766 WARN_ON_ONCE(1);
768 if (test_opt(sb, ERRORS_CONT)) {
769 ext4_commit_super(sb, 0);
770 return;
771 }
773 ext4_unlock_group(sb, grp);
774 ext4_commit_super(sb, 1);
775 ext4_handle_error(sb);
776 /*
777 * We only get here in the ERRORS_RO case; relocking the group
778 * may be dangerous, but nothing bad will happen since the
779 * filesystem will have already been marked read/only and the
780 * journal has been aborted. We return 1 as a hint to callers
781 * who might what to use the return value from
782 * ext4_grp_locked_error() to distinguish between the
783 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
784 * aggressively from the ext4 function in question, with a
785 * more appropriate error code.
786 */
787 ext4_lock_group(sb, grp);
788 return;
789 }
791 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
792 ext4_group_t group,
793 unsigned int flags)
794 {
795 struct ext4_sb_info *sbi = EXT4_SB(sb);
796 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
797 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
798 int ret;
800 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
801 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
802 &grp->bb_state);
803 if (!ret)
804 percpu_counter_sub(&sbi->s_freeclusters_counter,
805 grp->bb_free);
806 }
808 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
809 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
810 &grp->bb_state);
811 if (!ret && gdp) {
812 int count;
814 count = ext4_free_inodes_count(sb, gdp);
815 percpu_counter_sub(&sbi->s_freeinodes_counter,
816 count);
817 }
818 }
819 }
821 void ext4_update_dynamic_rev(struct super_block *sb)
822 {
823 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
825 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
826 return;
828 ext4_warning(sb,
829 "updating to rev %d because of new feature flag, "
830 "running e2fsck is recommended",
831 EXT4_DYNAMIC_REV);
833 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
834 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
835 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
836 /* leave es->s_feature_*compat flags alone */
837 /* es->s_uuid will be set by e2fsck if empty */
839 /*
840 * The rest of the superblock fields should be zero, and if not it
841 * means they are likely already in use, so leave them alone. We
842 * can leave it up to e2fsck to clean up any inconsistencies there.
843 */
844 }
846 /*
847 * Open the external journal device
848 */
849 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
850 {
851 struct block_device *bdev;
852 char b[BDEVNAME_SIZE];
854 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
855 if (IS_ERR(bdev))
856 goto fail;
857 return bdev;
859 fail:
860 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
861 __bdevname(dev, b), PTR_ERR(bdev));
862 return NULL;
863 }
865 /*
866 * Release the journal device
867 */
868 static void ext4_blkdev_put(struct block_device *bdev)
869 {
870 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
871 }
873 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
874 {
875 struct block_device *bdev;
876 bdev = sbi->journal_bdev;
877 if (bdev) {
878 ext4_blkdev_put(bdev);
879 sbi->journal_bdev = NULL;
880 }
881 }
883 static inline struct inode *orphan_list_entry(struct list_head *l)
884 {
885 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
886 }
888 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
889 {
890 struct list_head *l;
892 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
893 le32_to_cpu(sbi->s_es->s_last_orphan));
895 printk(KERN_ERR "sb_info orphan list:\n");
896 list_for_each(l, &sbi->s_orphan) {
897 struct inode *inode = orphan_list_entry(l);
898 printk(KERN_ERR " "
899 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
900 inode->i_sb->s_id, inode->i_ino, inode,
901 inode->i_mode, inode->i_nlink,
902 NEXT_ORPHAN(inode));
903 }
904 }
906 #ifdef CONFIG_QUOTA
907 static int ext4_quota_off(struct super_block *sb, int type);
909 static inline void ext4_quota_off_umount(struct super_block *sb)
910 {
911 int type;
913 /* Use our quota_off function to clear inode flags etc. */
914 for (type = 0; type < EXT4_MAXQUOTAS; type++)
915 ext4_quota_off(sb, type);
916 }
917 #else
918 static inline void ext4_quota_off_umount(struct super_block *sb)
919 {
920 }
921 #endif
923 static void ext4_put_super(struct super_block *sb)
924 {
925 struct ext4_sb_info *sbi = EXT4_SB(sb);
926 struct ext4_super_block *es = sbi->s_es;
927 int aborted = 0;
928 int i, err;
930 ext4_unregister_li_request(sb);
931 ext4_quota_off_umount(sb);
933 destroy_workqueue(sbi->rsv_conversion_wq);
935 if (sbi->s_journal) {
936 aborted = is_journal_aborted(sbi->s_journal);
937 err = jbd2_journal_destroy(sbi->s_journal);
938 sbi->s_journal = NULL;
939 if ((err < 0) && !aborted)
940 ext4_abort(sb, "Couldn't clean up the journal");
941 }
943 ext4_unregister_sysfs(sb);
944 ext4_es_unregister_shrinker(sbi);
945 del_timer_sync(&sbi->s_err_report);
946 ext4_release_system_zone(sb);
947 ext4_mb_release(sb);
948 ext4_ext_release(sb);
950 if (!sb_rdonly(sb) && !aborted) {
951 ext4_clear_feature_journal_needs_recovery(sb);
952 es->s_state = cpu_to_le16(sbi->s_mount_state);
953 }
954 if (!sb_rdonly(sb))
955 ext4_commit_super(sb, 1);
957 for (i = 0; i < sbi->s_gdb_count; i++)
958 brelse(sbi->s_group_desc[i]);
959 kvfree(sbi->s_group_desc);
960 kvfree(sbi->s_flex_groups);
961 percpu_counter_destroy(&sbi->s_freeclusters_counter);
962 percpu_counter_destroy(&sbi->s_freeinodes_counter);
963 percpu_counter_destroy(&sbi->s_dirs_counter);
964 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
965 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
966 #ifdef CONFIG_QUOTA
967 for (i = 0; i < EXT4_MAXQUOTAS; i++)
968 kfree(sbi->s_qf_names[i]);
969 #endif
971 /* Debugging code just in case the in-memory inode orphan list
972 * isn't empty. The on-disk one can be non-empty if we've
973 * detected an error and taken the fs readonly, but the
974 * in-memory list had better be clean by this point. */
975 if (!list_empty(&sbi->s_orphan))
976 dump_orphan_list(sb, sbi);
977 J_ASSERT(list_empty(&sbi->s_orphan));
979 sync_blockdev(sb->s_bdev);
980 invalidate_bdev(sb->s_bdev);
981 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
982 /*
983 * Invalidate the journal device's buffers. We don't want them
984 * floating about in memory - the physical journal device may
985 * hotswapped, and it breaks the `ro-after' testing code.
986 */
987 sync_blockdev(sbi->journal_bdev);
988 invalidate_bdev(sbi->journal_bdev);
989 ext4_blkdev_remove(sbi);
990 }
991 if (sbi->s_ea_inode_cache) {
992 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
993 sbi->s_ea_inode_cache = NULL;
994 }
995 if (sbi->s_ea_block_cache) {
996 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
997 sbi->s_ea_block_cache = NULL;
998 }
999 if (sbi->s_mmp_tsk)
1000 kthread_stop(sbi->s_mmp_tsk);
1001 brelse(sbi->s_sbh);
1002 sb->s_fs_info = NULL;
1003 /*
1004 * Now that we are completely done shutting down the
1005 * superblock, we need to actually destroy the kobject.
1006 */
1007 kobject_put(&sbi->s_kobj);
1008 wait_for_completion(&sbi->s_kobj_unregister);
1009 if (sbi->s_chksum_driver)
1010 crypto_free_shash(sbi->s_chksum_driver);
1011 kfree(sbi->s_blockgroup_lock);
1012 fs_put_dax(sbi->s_daxdev);
1013 kfree(sbi);
1014 }
1016 static struct kmem_cache *ext4_inode_cachep;
1018 /*
1019 * Called inside transaction, so use GFP_NOFS
1020 */
1021 static struct inode *ext4_alloc_inode(struct super_block *sb)
1022 {
1023 struct ext4_inode_info *ei;
1025 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1026 if (!ei)
1027 return NULL;
1029 inode_set_iversion(&ei->vfs_inode, 1);
1030 spin_lock_init(&ei->i_raw_lock);
1031 INIT_LIST_HEAD(&ei->i_prealloc_list);
1032 spin_lock_init(&ei->i_prealloc_lock);
1033 ext4_es_init_tree(&ei->i_es_tree);
1034 rwlock_init(&ei->i_es_lock);
1035 INIT_LIST_HEAD(&ei->i_es_list);
1036 ei->i_es_all_nr = 0;
1037 ei->i_es_shk_nr = 0;
1038 ei->i_es_shrink_lblk = 0;
1039 ei->i_reserved_data_blocks = 0;
1040 ei->i_da_metadata_calc_len = 0;
1041 ei->i_da_metadata_calc_last_lblock = 0;
1042 spin_lock_init(&(ei->i_block_reservation_lock));
1043 #ifdef CONFIG_QUOTA
1044 ei->i_reserved_quota = 0;
1045 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1046 #endif
1047 ei->jinode = NULL;
1048 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1049 spin_lock_init(&ei->i_completed_io_lock);
1050 ei->i_sync_tid = 0;
1051 ei->i_datasync_tid = 0;
1052 atomic_set(&ei->i_unwritten, 0);
1053 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1054 return &ei->vfs_inode;
1055 }
1057 static int ext4_drop_inode(struct inode *inode)
1058 {
1059 int drop = generic_drop_inode(inode);
1061 trace_ext4_drop_inode(inode, drop);
1062 return drop;
1063 }
1065 static void ext4_i_callback(struct rcu_head *head)
1066 {
1067 struct inode *inode = container_of(head, struct inode, i_rcu);
1068 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1069 }
1071 static void ext4_destroy_inode(struct inode *inode)
1072 {
1073 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1074 ext4_msg(inode->i_sb, KERN_ERR,
1075 "Inode %lu (%p): orphan list check failed!",
1076 inode->i_ino, EXT4_I(inode));
1077 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1078 EXT4_I(inode), sizeof(struct ext4_inode_info),
1079 true);
1080 dump_stack();
1081 }
1082 call_rcu(&inode->i_rcu, ext4_i_callback);
1083 }
1085 static void init_once(void *foo)
1086 {
1087 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1089 INIT_LIST_HEAD(&ei->i_orphan);
1090 init_rwsem(&ei->xattr_sem);
1091 init_rwsem(&ei->i_data_sem);
1092 init_rwsem(&ei->i_mmap_sem);
1093 inode_init_once(&ei->vfs_inode);
1094 }
1096 static int __init init_inodecache(void)
1097 {
1098 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1099 sizeof(struct ext4_inode_info), 0,
1100 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1101 SLAB_ACCOUNT),
1102 offsetof(struct ext4_inode_info, i_data),
1103 sizeof_field(struct ext4_inode_info, i_data),
1104 init_once);
1105 if (ext4_inode_cachep == NULL)
1106 return -ENOMEM;
1107 return 0;
1108 }
1110 static void destroy_inodecache(void)
1111 {
1112 /*
1113 * Make sure all delayed rcu free inodes are flushed before we
1114 * destroy cache.
1115 */
1116 rcu_barrier();
1117 kmem_cache_destroy(ext4_inode_cachep);
1118 }
1120 void ext4_clear_inode(struct inode *inode)
1121 {
1122 invalidate_inode_buffers(inode);
1123 clear_inode(inode);
1124 dquot_drop(inode);
1125 ext4_discard_preallocations(inode);
1126 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1127 if (EXT4_I(inode)->jinode) {
1128 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1129 EXT4_I(inode)->jinode);
1130 jbd2_free_inode(EXT4_I(inode)->jinode);
1131 EXT4_I(inode)->jinode = NULL;
1132 }
1133 fscrypt_put_encryption_info(inode);
1134 }
1136 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1137 u64 ino, u32 generation)
1138 {
1139 struct inode *inode;
1141 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1142 return ERR_PTR(-ESTALE);
1143 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1144 return ERR_PTR(-ESTALE);
1146 /* iget isn't really right if the inode is currently unallocated!!
1147 *
1148 * ext4_read_inode will return a bad_inode if the inode had been
1149 * deleted, so we should be safe.
1150 *
1151 * Currently we don't know the generation for parent directory, so
1152 * a generation of 0 means "accept any"
1153 */
1154 inode = ext4_iget_normal(sb, ino);
1155 if (IS_ERR(inode))
1156 return ERR_CAST(inode);
1157 if (generation && inode->i_generation != generation) {
1158 iput(inode);
1159 return ERR_PTR(-ESTALE);
1160 }
1162 return inode;
1163 }
1165 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1166 int fh_len, int fh_type)
1167 {
1168 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1169 ext4_nfs_get_inode);
1170 }
1172 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1173 int fh_len, int fh_type)
1174 {
1175 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1176 ext4_nfs_get_inode);
1177 }
1179 /*
1180 * Try to release metadata pages (indirect blocks, directories) which are
1181 * mapped via the block device. Since these pages could have journal heads
1182 * which would prevent try_to_free_buffers() from freeing them, we must use
1183 * jbd2 layer's try_to_free_buffers() function to release them.
1184 */
1185 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1186 gfp_t wait)
1187 {
1188 journal_t *journal = EXT4_SB(sb)->s_journal;
1190 WARN_ON(PageChecked(page));
1191 if (!page_has_buffers(page))
1192 return 0;
1193 if (journal)
1194 return jbd2_journal_try_to_free_buffers(journal, page,
1195 wait & ~__GFP_DIRECT_RECLAIM);
1196 return try_to_free_buffers(page);
1197 }
1199 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1200 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1201 {
1202 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1203 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1204 }
1206 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1207 void *fs_data)
1208 {
1209 handle_t *handle = fs_data;
1210 int res, res2, credits, retries = 0;
1212 /*
1213 * Encrypting the root directory is not allowed because e2fsck expects
1214 * lost+found to exist and be unencrypted, and encrypting the root
1215 * directory would imply encrypting the lost+found directory as well as
1216 * the filename "lost+found" itself.
1217 */
1218 if (inode->i_ino == EXT4_ROOT_INO)
1219 return -EPERM;
1221 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1222 return -EINVAL;
1224 res = ext4_convert_inline_data(inode);
1225 if (res)
1226 return res;
1228 /*
1229 * If a journal handle was specified, then the encryption context is
1230 * being set on a new inode via inheritance and is part of a larger
1231 * transaction to create the inode. Otherwise the encryption context is
1232 * being set on an existing inode in its own transaction. Only in the
1233 * latter case should the "retry on ENOSPC" logic be used.
1234 */
1236 if (handle) {
1237 res = ext4_xattr_set_handle(handle, inode,
1238 EXT4_XATTR_INDEX_ENCRYPTION,
1239 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1240 ctx, len, 0);
1241 if (!res) {
1242 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1243 ext4_clear_inode_state(inode,
1244 EXT4_STATE_MAY_INLINE_DATA);
1245 /*
1246 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1247 * S_DAX may be disabled
1248 */
1249 ext4_set_inode_flags(inode);
1250 }
1251 return res;
1252 }
1254 res = dquot_initialize(inode);
1255 if (res)
1256 return res;
1257 retry:
1258 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1259 &credits);
1260 if (res)
1261 return res;
1263 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1264 if (IS_ERR(handle))
1265 return PTR_ERR(handle);
1267 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1268 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1269 ctx, len, 0);
1270 if (!res) {
1271 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1272 /*
1273 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1274 * S_DAX may be disabled
1275 */
1276 ext4_set_inode_flags(inode);
1277 res = ext4_mark_inode_dirty(handle, inode);
1278 if (res)
1279 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1280 }
1281 res2 = ext4_journal_stop(handle);
1283 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1284 goto retry;
1285 if (!res)
1286 res = res2;
1287 return res;
1288 }
1290 static bool ext4_dummy_context(struct inode *inode)
1291 {
1292 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1293 }
1295 static const struct fscrypt_operations ext4_cryptops = {
1296 .key_prefix = "ext4:",
1297 .get_context = ext4_get_context,
1298 .set_context = ext4_set_context,
1299 .dummy_context = ext4_dummy_context,
1300 .empty_dir = ext4_empty_dir,
1301 .max_namelen = EXT4_NAME_LEN,
1302 };
1303 #endif
1305 #ifdef CONFIG_QUOTA
1306 static const char * const quotatypes[] = INITQFNAMES;
1307 #define QTYPE2NAME(t) (quotatypes[t])
1309 static int ext4_write_dquot(struct dquot *dquot);
1310 static int ext4_acquire_dquot(struct dquot *dquot);
1311 static int ext4_release_dquot(struct dquot *dquot);
1312 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1313 static int ext4_write_info(struct super_block *sb, int type);
1314 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1315 const struct path *path);
1316 static int ext4_quota_on_mount(struct super_block *sb, int type);
1317 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1318 size_t len, loff_t off);
1319 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1320 const char *data, size_t len, loff_t off);
1321 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1322 unsigned int flags);
1323 static int ext4_enable_quotas(struct super_block *sb);
1324 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1326 static struct dquot **ext4_get_dquots(struct inode *inode)
1327 {
1328 return EXT4_I(inode)->i_dquot;
1329 }
1331 static const struct dquot_operations ext4_quota_operations = {
1332 .get_reserved_space = ext4_get_reserved_space,
1333 .write_dquot = ext4_write_dquot,
1334 .acquire_dquot = ext4_acquire_dquot,
1335 .release_dquot = ext4_release_dquot,
1336 .mark_dirty = ext4_mark_dquot_dirty,
1337 .write_info = ext4_write_info,
1338 .alloc_dquot = dquot_alloc,
1339 .destroy_dquot = dquot_destroy,
1340 .get_projid = ext4_get_projid,
1341 .get_inode_usage = ext4_get_inode_usage,
1342 .get_next_id = ext4_get_next_id,
1343 };
1345 static const struct quotactl_ops ext4_qctl_operations = {
1346 .quota_on = ext4_quota_on,
1347 .quota_off = ext4_quota_off,
1348 .quota_sync = dquot_quota_sync,
1349 .get_state = dquot_get_state,
1350 .set_info = dquot_set_dqinfo,
1351 .get_dqblk = dquot_get_dqblk,
1352 .set_dqblk = dquot_set_dqblk,
1353 .get_nextdqblk = dquot_get_next_dqblk,
1354 };
1355 #endif
1357 static const struct super_operations ext4_sops = {
1358 .alloc_inode = ext4_alloc_inode,
1359 .destroy_inode = ext4_destroy_inode,
1360 .write_inode = ext4_write_inode,
1361 .dirty_inode = ext4_dirty_inode,
1362 .drop_inode = ext4_drop_inode,
1363 .evict_inode = ext4_evict_inode,
1364 .put_super = ext4_put_super,
1365 .sync_fs = ext4_sync_fs,
1366 .freeze_fs = ext4_freeze,
1367 .unfreeze_fs = ext4_unfreeze,
1368 .statfs = ext4_statfs,
1369 .remount_fs = ext4_remount,
1370 .show_options = ext4_show_options,
1371 #ifdef CONFIG_QUOTA
1372 .quota_read = ext4_quota_read,
1373 .quota_write = ext4_quota_write,
1374 .get_dquots = ext4_get_dquots,
1375 #endif
1376 .bdev_try_to_free_page = bdev_try_to_free_page,
1377 };
1379 static const struct export_operations ext4_export_ops = {
1380 .fh_to_dentry = ext4_fh_to_dentry,
1381 .fh_to_parent = ext4_fh_to_parent,
1382 .get_parent = ext4_get_parent,
1383 };
1385 enum {
1386 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1387 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1388 Opt_nouid32, Opt_debug, Opt_removed,
1389 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1390 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1391 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1392 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1393 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1394 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1395 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1396 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1397 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1398 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1399 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1400 Opt_nowarn_on_error, Opt_mblk_io_submit,
1401 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1402 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1403 Opt_inode_readahead_blks, Opt_journal_ioprio,
1404 Opt_dioread_nolock, Opt_dioread_lock,
1405 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1406 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1407 };
1409 static const match_table_t tokens = {
1410 {Opt_bsd_df, "bsddf"},
1411 {Opt_minix_df, "minixdf"},
1412 {Opt_grpid, "grpid"},
1413 {Opt_grpid, "bsdgroups"},
1414 {Opt_nogrpid, "nogrpid"},
1415 {Opt_nogrpid, "sysvgroups"},
1416 {Opt_resgid, "resgid=%u"},
1417 {Opt_resuid, "resuid=%u"},
1418 {Opt_sb, "sb=%u"},
1419 {Opt_err_cont, "errors=continue"},
1420 {Opt_err_panic, "errors=panic"},
1421 {Opt_err_ro, "errors=remount-ro"},
1422 {Opt_nouid32, "nouid32"},
1423 {Opt_debug, "debug"},
1424 {Opt_removed, "oldalloc"},
1425 {Opt_removed, "orlov"},
1426 {Opt_user_xattr, "user_xattr"},
1427 {Opt_nouser_xattr, "nouser_xattr"},
1428 {Opt_acl, "acl"},
1429 {Opt_noacl, "noacl"},
1430 {Opt_noload, "norecovery"},
1431 {Opt_noload, "noload"},
1432 {Opt_removed, "nobh"},
1433 {Opt_removed, "bh"},
1434 {Opt_commit, "commit=%u"},
1435 {Opt_min_batch_time, "min_batch_time=%u"},
1436 {Opt_max_batch_time, "max_batch_time=%u"},
1437 {Opt_journal_dev, "journal_dev=%u"},
1438 {Opt_journal_path, "journal_path=%s"},
1439 {Opt_journal_checksum, "journal_checksum"},
1440 {Opt_nojournal_checksum, "nojournal_checksum"},
1441 {Opt_journal_async_commit, "journal_async_commit"},
1442 {Opt_abort, "abort"},
1443 {Opt_data_journal, "data=journal"},
1444 {Opt_data_ordered, "data=ordered"},
1445 {Opt_data_writeback, "data=writeback"},
1446 {Opt_data_err_abort, "data_err=abort"},
1447 {Opt_data_err_ignore, "data_err=ignore"},
1448 {Opt_offusrjquota, "usrjquota="},
1449 {Opt_usrjquota, "usrjquota=%s"},
1450 {Opt_offgrpjquota, "grpjquota="},
1451 {Opt_grpjquota, "grpjquota=%s"},
1452 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1453 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1454 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1455 {Opt_grpquota, "grpquota"},
1456 {Opt_noquota, "noquota"},
1457 {Opt_quota, "quota"},
1458 {Opt_usrquota, "usrquota"},
1459 {Opt_prjquota, "prjquota"},
1460 {Opt_barrier, "barrier=%u"},
1461 {Opt_barrier, "barrier"},
1462 {Opt_nobarrier, "nobarrier"},
1463 {Opt_i_version, "i_version"},
1464 {Opt_dax, "dax"},
1465 {Opt_stripe, "stripe=%u"},
1466 {Opt_delalloc, "delalloc"},
1467 {Opt_warn_on_error, "warn_on_error"},
1468 {Opt_nowarn_on_error, "nowarn_on_error"},
1469 {Opt_lazytime, "lazytime"},
1470 {Opt_nolazytime, "nolazytime"},
1471 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1472 {Opt_nodelalloc, "nodelalloc"},
1473 {Opt_removed, "mblk_io_submit"},
1474 {Opt_removed, "nomblk_io_submit"},
1475 {Opt_block_validity, "block_validity"},
1476 {Opt_noblock_validity, "noblock_validity"},
1477 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1478 {Opt_journal_ioprio, "journal_ioprio=%u"},
1479 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1480 {Opt_auto_da_alloc, "auto_da_alloc"},
1481 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1482 {Opt_dioread_nolock, "dioread_nolock"},
1483 {Opt_dioread_lock, "dioread_lock"},
1484 {Opt_discard, "discard"},
1485 {Opt_nodiscard, "nodiscard"},
1486 {Opt_init_itable, "init_itable=%u"},
1487 {Opt_init_itable, "init_itable"},
1488 {Opt_noinit_itable, "noinit_itable"},
1489 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1490 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1491 {Opt_nombcache, "nombcache"},
1492 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1493 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1494 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1495 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1496 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1497 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1498 {Opt_err, NULL},
1499 };
1501 static ext4_fsblk_t get_sb_block(void **data)
1502 {
1503 ext4_fsblk_t sb_block;
1504 char *options = (char *) *data;
1506 if (!options || strncmp(options, "sb=", 3) != 0)
1507 return 1; /* Default location */
1509 options += 3;
1510 /* TODO: use simple_strtoll with >32bit ext4 */
1511 sb_block = simple_strtoul(options, &options, 0);
1512 if (*options && *options != ',') {
1513 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1514 (char *) *data);
1515 return 1;
1516 }
1517 if (*options == ',')
1518 options++;
1519 *data = (void *) options;
1521 return sb_block;
1522 }
1524 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1525 static const char deprecated_msg[] =
1526 "Mount option \"%s\" will be removed by %s\n"
1527 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1529 #ifdef CONFIG_QUOTA
1530 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1531 {
1532 struct ext4_sb_info *sbi = EXT4_SB(sb);
1533 char *qname;
1534 int ret = -1;
1536 if (sb_any_quota_loaded(sb) &&
1537 !sbi->s_qf_names[qtype]) {
1538 ext4_msg(sb, KERN_ERR,
1539 "Cannot change journaled "
1540 "quota options when quota turned on");
1541 return -1;
1542 }
1543 if (ext4_has_feature_quota(sb)) {
1544 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1545 "ignored when QUOTA feature is enabled");
1546 return 1;
1547 }
1548 qname = match_strdup(args);
1549 if (!qname) {
1550 ext4_msg(sb, KERN_ERR,
1551 "Not enough memory for storing quotafile name");
1552 return -1;
1553 }
1554 if (sbi->s_qf_names[qtype]) {
1555 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1556 ret = 1;
1557 else
1558 ext4_msg(sb, KERN_ERR,
1559 "%s quota file already specified",
1560 QTYPE2NAME(qtype));
1561 goto errout;
1562 }
1563 if (strchr(qname, '/')) {
1564 ext4_msg(sb, KERN_ERR,
1565 "quotafile must be on filesystem root");
1566 goto errout;
1567 }
1568 sbi->s_qf_names[qtype] = qname;
1569 set_opt(sb, QUOTA);
1570 return 1;
1571 errout:
1572 kfree(qname);
1573 return ret;
1574 }
1576 static int clear_qf_name(struct super_block *sb, int qtype)
1577 {
1579 struct ext4_sb_info *sbi = EXT4_SB(sb);
1581 if (sb_any_quota_loaded(sb) &&
1582 sbi->s_qf_names[qtype]) {
1583 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1584 " when quota turned on");
1585 return -1;
1586 }
1587 kfree(sbi->s_qf_names[qtype]);
1588 sbi->s_qf_names[qtype] = NULL;
1589 return 1;
1590 }
1591 #endif
1593 #define MOPT_SET 0x0001
1594 #define MOPT_CLEAR 0x0002
1595 #define MOPT_NOSUPPORT 0x0004
1596 #define MOPT_EXPLICIT 0x0008
1597 #define MOPT_CLEAR_ERR 0x0010
1598 #define MOPT_GTE0 0x0020
1599 #ifdef CONFIG_QUOTA
1600 #define MOPT_Q 0
1601 #define MOPT_QFMT 0x0040
1602 #else
1603 #define MOPT_Q MOPT_NOSUPPORT
1604 #define MOPT_QFMT MOPT_NOSUPPORT
1605 #endif
1606 #define MOPT_DATAJ 0x0080
1607 #define MOPT_NO_EXT2 0x0100
1608 #define MOPT_NO_EXT3 0x0200
1609 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1610 #define MOPT_STRING 0x0400
1612 static const struct mount_opts {
1613 int token;
1614 int mount_opt;
1615 int flags;
1616 } ext4_mount_opts[] = {
1617 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1618 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1619 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1620 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1621 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1622 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1623 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1624 MOPT_EXT4_ONLY | MOPT_SET},
1625 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1626 MOPT_EXT4_ONLY | MOPT_CLEAR},
1627 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1628 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1629 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1630 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1631 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1632 MOPT_EXT4_ONLY | MOPT_CLEAR},
1633 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1634 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1635 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1636 MOPT_EXT4_ONLY | MOPT_CLEAR},
1637 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1638 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1639 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1640 EXT4_MOUNT_JOURNAL_CHECKSUM),
1641 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1642 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1643 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1644 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1645 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1646 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1647 MOPT_NO_EXT2},
1648 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1649 MOPT_NO_EXT2},
1650 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1651 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1652 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1653 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1654 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1655 {Opt_commit, 0, MOPT_GTE0},
1656 {Opt_max_batch_time, 0, MOPT_GTE0},
1657 {Opt_min_batch_time, 0, MOPT_GTE0},
1658 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1659 {Opt_init_itable, 0, MOPT_GTE0},
1660 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1661 {Opt_stripe, 0, MOPT_GTE0},
1662 {Opt_resuid, 0, MOPT_GTE0},
1663 {Opt_resgid, 0, MOPT_GTE0},
1664 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1665 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1666 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1667 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1668 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1669 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1670 MOPT_NO_EXT2 | MOPT_DATAJ},
1671 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1672 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1673 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1674 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1675 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1676 #else
1677 {Opt_acl, 0, MOPT_NOSUPPORT},
1678 {Opt_noacl, 0, MOPT_NOSUPPORT},
1679 #endif
1680 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1681 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1682 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1683 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1684 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1685 MOPT_SET | MOPT_Q},
1686 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1687 MOPT_SET | MOPT_Q},
1688 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1689 MOPT_SET | MOPT_Q},
1690 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1691 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1692 MOPT_CLEAR | MOPT_Q},
1693 {Opt_usrjquota, 0, MOPT_Q},
1694 {Opt_grpjquota, 0, MOPT_Q},
1695 {Opt_offusrjquota, 0, MOPT_Q},
1696 {Opt_offgrpjquota, 0, MOPT_Q},
1697 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1698 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1699 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1700 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1701 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1702 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1703 {Opt_err, 0, 0}
1704 };
1706 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1707 substring_t *args, unsigned long *journal_devnum,
1708 unsigned int *journal_ioprio, int is_remount)
1709 {
1710 struct ext4_sb_info *sbi = EXT4_SB(sb);
1711 const struct mount_opts *m;
1712 kuid_t uid;
1713 kgid_t gid;
1714 int arg = 0;
1716 #ifdef CONFIG_QUOTA
1717 if (token == Opt_usrjquota)
1718 return set_qf_name(sb, USRQUOTA, &args[0]);
1719 else if (token == Opt_grpjquota)
1720 return set_qf_name(sb, GRPQUOTA, &args[0]);
1721 else if (token == Opt_offusrjquota)
1722 return clear_qf_name(sb, USRQUOTA);
1723 else if (token == Opt_offgrpjquota)
1724 return clear_qf_name(sb, GRPQUOTA);
1725 #endif
1726 switch (token) {
1727 case Opt_noacl:
1728 case Opt_nouser_xattr:
1729 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1730 break;
1731 case Opt_sb:
1732 return 1; /* handled by get_sb_block() */
1733 case Opt_removed:
1734 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1735 return 1;
1736 case Opt_abort:
1737 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1738 return 1;
1739 case Opt_i_version:
1740 sb->s_flags |= SB_I_VERSION;
1741 return 1;
1742 case Opt_lazytime:
1743 sb->s_flags |= SB_LAZYTIME;
1744 return 1;
1745 case Opt_nolazytime:
1746 sb->s_flags &= ~SB_LAZYTIME;
1747 return 1;
1748 }
1750 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1751 if (token == m->token)
1752 break;
1754 if (m->token == Opt_err) {
1755 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1756 "or missing value", opt);
1757 return -1;
1758 }
1760 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1761 ext4_msg(sb, KERN_ERR,
1762 "Mount option \"%s\" incompatible with ext2", opt);
1763 return -1;
1764 }
1765 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1766 ext4_msg(sb, KERN_ERR,
1767 "Mount option \"%s\" incompatible with ext3", opt);
1768 return -1;
1769 }
1771 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1772 return -1;
1773 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1774 return -1;
1775 if (m->flags & MOPT_EXPLICIT) {
1776 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1777 set_opt2(sb, EXPLICIT_DELALLOC);
1778 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1779 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1780 } else
1781 return -1;
1782 }
1783 if (m->flags & MOPT_CLEAR_ERR)
1784 clear_opt(sb, ERRORS_MASK);
1785 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1786 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1787 "options when quota turned on");
1788 return -1;
1789 }
1791 if (m->flags & MOPT_NOSUPPORT) {
1792 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1793 } else if (token == Opt_commit) {
1794 if (arg == 0)
1795 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1796 sbi->s_commit_interval = HZ * arg;
1797 } else if (token == Opt_debug_want_extra_isize) {
1798 sbi->s_want_extra_isize = arg;
1799 } else if (token == Opt_max_batch_time) {
1800 sbi->s_max_batch_time = arg;
1801 } else if (token == Opt_min_batch_time) {
1802 sbi->s_min_batch_time = arg;
1803 } else if (token == Opt_inode_readahead_blks) {
1804 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1805 ext4_msg(sb, KERN_ERR,
1806 "EXT4-fs: inode_readahead_blks must be "
1807 "0 or a power of 2 smaller than 2^31");
1808 return -1;
1809 }
1810 sbi->s_inode_readahead_blks = arg;
1811 } else if (token == Opt_init_itable) {
1812 set_opt(sb, INIT_INODE_TABLE);
1813 if (!args->from)
1814 arg = EXT4_DEF_LI_WAIT_MULT;
1815 sbi->s_li_wait_mult = arg;
1816 } else if (token == Opt_max_dir_size_kb) {
1817 sbi->s_max_dir_size_kb = arg;
1818 } else if (token == Opt_stripe) {
1819 sbi->s_stripe = arg;
1820 } else if (token == Opt_resuid) {
1821 uid = make_kuid(current_user_ns(), arg);
1822 if (!uid_valid(uid)) {
1823 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1824 return -1;
1825 }
1826 sbi->s_resuid = uid;
1827 } else if (token == Opt_resgid) {
1828 gid = make_kgid(current_user_ns(), arg);
1829 if (!gid_valid(gid)) {
1830 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1831 return -1;
1832 }
1833 sbi->s_resgid = gid;
1834 } else if (token == Opt_journal_dev) {
1835 if (is_remount) {
1836 ext4_msg(sb, KERN_ERR,
1837 "Cannot specify journal on remount");
1838 return -1;
1839 }
1840 *journal_devnum = arg;
1841 } else if (token == Opt_journal_path) {
1842 char *journal_path;
1843 struct inode *journal_inode;
1844 struct path path;
1845 int error;
1847 if (is_remount) {
1848 ext4_msg(sb, KERN_ERR,
1849 "Cannot specify journal on remount");
1850 return -1;
1851 }
1852 journal_path = match_strdup(&args[0]);
1853 if (!journal_path) {
1854 ext4_msg(sb, KERN_ERR, "error: could not dup "
1855 "journal device string");
1856 return -1;
1857 }
1859 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1860 if (error) {
1861 ext4_msg(sb, KERN_ERR, "error: could not find "
1862 "journal device path: error %d", error);
1863 kfree(journal_path);
1864 return -1;
1865 }
1867 journal_inode = d_inode(path.dentry);
1868 if (!S_ISBLK(journal_inode->i_mode)) {
1869 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1870 "is not a block device", journal_path);
1871 path_put(&path);
1872 kfree(journal_path);
1873 return -1;
1874 }
1876 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1877 path_put(&path);
1878 kfree(journal_path);
1879 } else if (token == Opt_journal_ioprio) {
1880 if (arg > 7) {
1881 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1882 " (must be 0-7)");
1883 return -1;
1884 }
1885 *journal_ioprio =
1886 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1887 } else if (token == Opt_test_dummy_encryption) {
1888 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1889 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1890 ext4_msg(sb, KERN_WARNING,
1891 "Test dummy encryption mode enabled");
1892 #else
1893 ext4_msg(sb, KERN_WARNING,
1894 "Test dummy encryption mount option ignored");
1895 #endif
1896 } else if (m->flags & MOPT_DATAJ) {
1897 if (is_remount) {
1898 if (!sbi->s_journal)
1899 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1900 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1901 ext4_msg(sb, KERN_ERR,
1902 "Cannot change data mode on remount");
1903 return -1;
1904 }
1905 } else {
1906 clear_opt(sb, DATA_FLAGS);
1907 sbi->s_mount_opt |= m->mount_opt;
1908 }
1909 #ifdef CONFIG_QUOTA
1910 } else if (m->flags & MOPT_QFMT) {
1911 if (sb_any_quota_loaded(sb) &&
1912 sbi->s_jquota_fmt != m->mount_opt) {
1913 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1914 "quota options when quota turned on");
1915 return -1;
1916 }
1917 if (ext4_has_feature_quota(sb)) {
1918 ext4_msg(sb, KERN_INFO,
1919 "Quota format mount options ignored "
1920 "when QUOTA feature is enabled");
1921 return 1;
1922 }
1923 sbi->s_jquota_fmt = m->mount_opt;
1924 #endif
1925 } else if (token == Opt_dax) {
1926 #ifdef CONFIG_FS_DAX
1927 ext4_msg(sb, KERN_WARNING,
1928 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1929 sbi->s_mount_opt |= m->mount_opt;
1930 #else
1931 ext4_msg(sb, KERN_INFO, "dax option not supported");
1932 return -1;
1933 #endif
1934 } else if (token == Opt_data_err_abort) {
1935 sbi->s_mount_opt |= m->mount_opt;
1936 } else if (token == Opt_data_err_ignore) {
1937 sbi->s_mount_opt &= ~m->mount_opt;
1938 } else {
1939 if (!args->from)
1940 arg = 1;
1941 if (m->flags & MOPT_CLEAR)
1942 arg = !arg;
1943 else if (unlikely(!(m->flags & MOPT_SET))) {
1944 ext4_msg(sb, KERN_WARNING,
1945 "buggy handling of option %s", opt);
1946 WARN_ON(1);
1947 return -1;
1948 }
1949 if (arg != 0)
1950 sbi->s_mount_opt |= m->mount_opt;
1951 else
1952 sbi->s_mount_opt &= ~m->mount_opt;
1953 }
1954 return 1;
1955 }
1957 static int parse_options(char *options, struct super_block *sb,
1958 unsigned long *journal_devnum,
1959 unsigned int *journal_ioprio,
1960 int is_remount)
1961 {
1962 struct ext4_sb_info *sbi = EXT4_SB(sb);
1963 char *p;
1964 substring_t args[MAX_OPT_ARGS];
1965 int token;
1967 if (!options)
1968 return 1;
1970 while ((p = strsep(&options, ",")) != NULL) {
1971 if (!*p)
1972 continue;
1973 /*
1974 * Initialize args struct so we know whether arg was
1975 * found; some options take optional arguments.
1976 */
1977 args[0].to = args[0].from = NULL;
1978 token = match_token(p, tokens, args);
1979 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1980 journal_ioprio, is_remount) < 0)
1981 return 0;
1982 }
1983 #ifdef CONFIG_QUOTA
1984 /*
1985 * We do the test below only for project quotas. 'usrquota' and
1986 * 'grpquota' mount options are allowed even without quota feature
1987 * to support legacy quotas in quota files.
1988 */
1989 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1990 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1991 "Cannot enable project quota enforcement.");
1992 return 0;
1993 }
1994 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1995 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1996 clear_opt(sb, USRQUOTA);
1998 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1999 clear_opt(sb, GRPQUOTA);
2001 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2002 ext4_msg(sb, KERN_ERR, "old and new quota "
2003 "format mixing");
2004 return 0;
2005 }
2007 if (!sbi->s_jquota_fmt) {
2008 ext4_msg(sb, KERN_ERR, "journaled quota format "
2009 "not specified");
2010 return 0;
2011 }
2012 }
2013 #endif
2014 if (test_opt(sb, DIOREAD_NOLOCK)) {
2015 int blocksize =
2016 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2018 if (blocksize < PAGE_SIZE) {
2019 ext4_msg(sb, KERN_ERR, "can't mount with "
2020 "dioread_nolock if block size != PAGE_SIZE");
2021 return 0;
2022 }
2023 }
2024 return 1;
2025 }
2027 static inline void ext4_show_quota_options(struct seq_file *seq,
2028 struct super_block *sb)
2029 {
2030 #if defined(CONFIG_QUOTA)
2031 struct ext4_sb_info *sbi = EXT4_SB(sb);
2033 if (sbi->s_jquota_fmt) {
2034 char *fmtname = "";
2036 switch (sbi->s_jquota_fmt) {
2037 case QFMT_VFS_OLD:
2038 fmtname = "vfsold";
2039 break;
2040 case QFMT_VFS_V0:
2041 fmtname = "vfsv0";
2042 break;
2043 case QFMT_VFS_V1:
2044 fmtname = "vfsv1";
2045 break;
2046 }
2047 seq_printf(seq, ",jqfmt=%s", fmtname);
2048 }
2050 if (sbi->s_qf_names[USRQUOTA])
2051 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
2053 if (sbi->s_qf_names[GRPQUOTA])
2054 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2055 #endif
2056 }
2058 static const char *token2str(int token)
2059 {
2060 const struct match_token *t;
2062 for (t = tokens; t->token != Opt_err; t++)
2063 if (t->token == token && !strchr(t->pattern, '='))
2064 break;
2065 return t->pattern;
2066 }
2068 /*
2069 * Show an option if
2070 * - it's set to a non-default value OR
2071 * - if the per-sb default is different from the global default
2072 */
2073 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2074 int nodefs)
2075 {
2076 struct ext4_sb_info *sbi = EXT4_SB(sb);
2077 struct ext4_super_block *es = sbi->s_es;
2078 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2079 const struct mount_opts *m;
2080 char sep = nodefs ? '\n' : ',';
2082 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2083 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2085 if (sbi->s_sb_block != 1)
2086 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2088 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2089 int want_set = m->flags & MOPT_SET;
2090 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2091 (m->flags & MOPT_CLEAR_ERR))
2092 continue;
2093 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2094 continue; /* skip if same as the default */
2095 if ((want_set &&
2096 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2097 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2098 continue; /* select Opt_noFoo vs Opt_Foo */
2099 SEQ_OPTS_PRINT("%s", token2str(m->token));
2100 }
2102 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2103 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2104 SEQ_OPTS_PRINT("resuid=%u",
2105 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2106 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2107 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2108 SEQ_OPTS_PRINT("resgid=%u",
2109 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2110 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2111 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2112 SEQ_OPTS_PUTS("errors=remount-ro");
2113 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2114 SEQ_OPTS_PUTS("errors=continue");
2115 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2116 SEQ_OPTS_PUTS("errors=panic");
2117 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2118 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2119 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2120 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2121 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2122 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2123 if (sb->s_flags & SB_I_VERSION)
2124 SEQ_OPTS_PUTS("i_version");
2125 if (nodefs || sbi->s_stripe)
2126 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2127 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2128 (sbi->s_mount_opt ^ def_mount_opt)) {
2129 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2130 SEQ_OPTS_PUTS("data=journal");
2131 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2132 SEQ_OPTS_PUTS("data=ordered");
2133 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2134 SEQ_OPTS_PUTS("data=writeback");
2135 }
2136 if (nodefs ||
2137 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2138 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2139 sbi->s_inode_readahead_blks);
2141 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2142 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2143 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2144 if (nodefs || sbi->s_max_dir_size_kb)
2145 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2146 if (test_opt(sb, DATA_ERR_ABORT))
2147 SEQ_OPTS_PUTS("data_err=abort");
2148 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2149 SEQ_OPTS_PUTS("test_dummy_encryption");
2151 ext4_show_quota_options(seq, sb);
2152 return 0;
2153 }
2155 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2156 {
2157 return _ext4_show_options(seq, root->d_sb, 0);
2158 }
2160 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2161 {
2162 struct super_block *sb = seq->private;
2163 int rc;
2165 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2166 rc = _ext4_show_options(seq, sb, 1);
2167 seq_puts(seq, "\n");
2168 return rc;
2169 }
2171 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2172 int read_only)
2173 {
2174 struct ext4_sb_info *sbi = EXT4_SB(sb);
2175 int err = 0;
2177 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2178 ext4_msg(sb, KERN_ERR, "revision level too high, "
2179 "forcing read-only mode");
2180 err = -EROFS;
2181 }
2182 if (read_only)
2183 goto done;
2184 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2185 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2186 "running e2fsck is recommended");
2187 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2188 ext4_msg(sb, KERN_WARNING,
2189 "warning: mounting fs with errors, "
2190 "running e2fsck is recommended");
2191 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2192 le16_to_cpu(es->s_mnt_count) >=
2193 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2194 ext4_msg(sb, KERN_WARNING,
2195 "warning: maximal mount count reached, "
2196 "running e2fsck is recommended");
2197 else if (le32_to_cpu(es->s_checkinterval) &&
2198 (ext4_get_tstamp(es, s_lastcheck) +
2199 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2200 ext4_msg(sb, KERN_WARNING,
2201 "warning: checktime reached, "
2202 "running e2fsck is recommended");
2203 if (!sbi->s_journal)
2204 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2205 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2206 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2207 le16_add_cpu(&es->s_mnt_count, 1);
2208 ext4_update_tstamp(es, s_mtime);
2209 ext4_update_dynamic_rev(sb);
2210 if (sbi->s_journal)
2211 ext4_set_feature_journal_needs_recovery(sb);
2213 err = ext4_commit_super(sb, 1);
2214 done:
2215 if (test_opt(sb, DEBUG))
2216 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2217 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2218 sb->s_blocksize,
2219 sbi->s_groups_count,
2220 EXT4_BLOCKS_PER_GROUP(sb),
2221 EXT4_INODES_PER_GROUP(sb),
2222 sbi->s_mount_opt, sbi->s_mount_opt2);
2224 cleancache_init_fs(sb);
2225 return err;
2226 }
2228 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2229 {
2230 struct ext4_sb_info *sbi = EXT4_SB(sb);
2231 struct flex_groups *new_groups;
2232 int size;
2234 if (!sbi->s_log_groups_per_flex)
2235 return 0;
2237 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2238 if (size <= sbi->s_flex_groups_allocated)
2239 return 0;
2241 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2242 new_groups = kvzalloc(size, GFP_KERNEL);
2243 if (!new_groups) {
2244 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2245 size / (int) sizeof(struct flex_groups));
2246 return -ENOMEM;
2247 }
2249 if (sbi->s_flex_groups) {
2250 memcpy(new_groups, sbi->s_flex_groups,
2251 (sbi->s_flex_groups_allocated *
2252 sizeof(struct flex_groups)));
2253 kvfree(sbi->s_flex_groups);
2254 }
2255 sbi->s_flex_groups = new_groups;
2256 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2257 return 0;
2258 }
2260 static int ext4_fill_flex_info(struct super_block *sb)
2261 {
2262 struct ext4_sb_info *sbi = EXT4_SB(sb);
2263 struct ext4_group_desc *gdp = NULL;
2264 ext4_group_t flex_group;
2265 int i, err;
2267 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2268 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2269 sbi->s_log_groups_per_flex = 0;
2270 return 1;
2271 }
2273 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2274 if (err)
2275 goto failed;
2277 for (i = 0; i < sbi->s_groups_count; i++) {
2278 gdp = ext4_get_group_desc(sb, i, NULL);
2280 flex_group = ext4_flex_group(sbi, i);
2281 atomic_add(ext4_free_inodes_count(sb, gdp),
2282 &sbi->s_flex_groups[flex_group].free_inodes);
2283 atomic64_add(ext4_free_group_clusters(sb, gdp),
2284 &sbi->s_flex_groups[flex_group].free_clusters);
2285 atomic_add(ext4_used_dirs_count(sb, gdp),
2286 &sbi->s_flex_groups[flex_group].used_dirs);
2287 }
2289 return 1;
2290 failed:
2291 return 0;
2292 }
2294 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2295 struct ext4_group_desc *gdp)
2296 {
2297 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2298 __u16 crc = 0;
2299 __le32 le_group = cpu_to_le32(block_group);
2300 struct ext4_sb_info *sbi = EXT4_SB(sb);
2302 if (ext4_has_metadata_csum(sbi->s_sb)) {
2303 /* Use new metadata_csum algorithm */
2304 __u32 csum32;
2305 __u16 dummy_csum = 0;
2307 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2308 sizeof(le_group));
2309 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2310 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2311 sizeof(dummy_csum));
2312 offset += sizeof(dummy_csum);
2313 if (offset < sbi->s_desc_size)
2314 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2315 sbi->s_desc_size - offset);
2317 crc = csum32 & 0xFFFF;
2318 goto out;
2319 }
2321 /* old crc16 code */
2322 if (!ext4_has_feature_gdt_csum(sb))
2323 return 0;
2325 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2326 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2327 crc = crc16(crc, (__u8 *)gdp, offset);
2328 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2329 /* for checksum of struct ext4_group_desc do the rest...*/
2330 if (ext4_has_feature_64bit(sb) &&
2331 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2332 crc = crc16(crc, (__u8 *)gdp + offset,
2333 le16_to_cpu(sbi->s_es->s_desc_size) -
2334 offset);
2336 out:
2337 return cpu_to_le16(crc);
2338 }
2340 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2341 struct ext4_group_desc *gdp)
2342 {
2343 if (ext4_has_group_desc_csum(sb) &&
2344 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2345 return 0;
2347 return 1;
2348 }
2350 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2351 struct ext4_group_desc *gdp)
2352 {
2353 if (!ext4_has_group_desc_csum(sb))
2354 return;
2355 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2356 }
2358 /* Called at mount-time, super-block is locked */
2359 static int ext4_check_descriptors(struct super_block *sb,
2360 ext4_fsblk_t sb_block,
2361 ext4_group_t *first_not_zeroed)
2362 {
2363 struct ext4_sb_info *sbi = EXT4_SB(sb);
2364 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2365 ext4_fsblk_t last_block;
2366 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2367 ext4_fsblk_t block_bitmap;
2368 ext4_fsblk_t inode_bitmap;
2369 ext4_fsblk_t inode_table;
2370 int flexbg_flag = 0;
2371 ext4_group_t i, grp = sbi->s_groups_count;
2373 if (ext4_has_feature_flex_bg(sb))
2374 flexbg_flag = 1;
2376 ext4_debug("Checking group descriptors");
2378 for (i = 0; i < sbi->s_groups_count; i++) {
2379 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2381 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2382 last_block = ext4_blocks_count(sbi->s_es) - 1;
2383 else
2384 last_block = first_block +
2385 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2387 if ((grp == sbi->s_groups_count) &&
2388 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2389 grp = i;
2391 block_bitmap = ext4_block_bitmap(sb, gdp);
2392 if (block_bitmap == sb_block) {
2393 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2394 "Block bitmap for group %u overlaps "
2395 "superblock", i);
2396 if (!sb_rdonly(sb))
2397 return 0;
2398 }
2399 if (block_bitmap >= sb_block + 1 &&
2400 block_bitmap <= last_bg_block) {
2401 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2402 "Block bitmap for group %u overlaps "
2403 "block group descriptors", i);
2404 if (!sb_rdonly(sb))
2405 return 0;
2406 }
2407 if (block_bitmap < first_block || block_bitmap > last_block) {
2408 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2409 "Block bitmap for group %u not in group "
2410 "(block %llu)!", i, block_bitmap);
2411 return 0;
2412 }
2413 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2414 if (inode_bitmap == sb_block) {
2415 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2416 "Inode bitmap for group %u overlaps "
2417 "superblock", i);
2418 if (!sb_rdonly(sb))
2419 return 0;
2420 }
2421 if (inode_bitmap >= sb_block + 1 &&
2422 inode_bitmap <= last_bg_block) {
2423 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2424 "Inode bitmap for group %u overlaps "
2425 "block group descriptors", i);
2426 if (!sb_rdonly(sb))
2427 return 0;
2428 }
2429 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2430 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2431 "Inode bitmap for group %u not in group "
2432 "(block %llu)!", i, inode_bitmap);
2433 return 0;
2434 }
2435 inode_table = ext4_inode_table(sb, gdp);
2436 if (inode_table == sb_block) {
2437 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2438 "Inode table for group %u overlaps "
2439 "superblock", i);
2440 if (!sb_rdonly(sb))
2441 return 0;
2442 }
2443 if (inode_table >= sb_block + 1 &&
2444 inode_table <= last_bg_block) {
2445 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2446 "Inode table for group %u overlaps "
2447 "block group descriptors", i);
2448 if (!sb_rdonly(sb))
2449 return 0;
2450 }
2451 if (inode_table < first_block ||
2452 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2453 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2454 "Inode table for group %u not in group "
2455 "(block %llu)!", i, inode_table);
2456 return 0;
2457 }
2458 ext4_lock_group(sb, i);
2459 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2460 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2461 "Checksum for group %u failed (%u!=%u)",
2462 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2463 gdp)), le16_to_cpu(gdp->bg_checksum));
2464 if (!sb_rdonly(sb)) {
2465 ext4_unlock_group(sb, i);
2466 return 0;
2467 }
2468 }
2469 ext4_unlock_group(sb, i);
2470 if (!flexbg_flag)
2471 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2472 }
2473 if (NULL != first_not_zeroed)
2474 *first_not_zeroed = grp;
2475 return 1;
2476 }
2478 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2479 * the superblock) which were deleted from all directories, but held open by
2480 * a process at the time of a crash. We walk the list and try to delete these
2481 * inodes at recovery time (only with a read-write filesystem).
2482 *
2483 * In order to keep the orphan inode chain consistent during traversal (in
2484 * case of crash during recovery), we link each inode into the superblock
2485 * orphan list_head and handle it the same way as an inode deletion during
2486 * normal operation (which journals the operations for us).
2487 *
2488 * We only do an iget() and an iput() on each inode, which is very safe if we
2489 * accidentally point at an in-use or already deleted inode. The worst that
2490 * can happen in this case is that we get a "bit already cleared" message from
2491 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2492 * e2fsck was run on this filesystem, and it must have already done the orphan
2493 * inode cleanup for us, so we can safely abort without any further action.
2494 */
2495 static void ext4_orphan_cleanup(struct super_block *sb,
2496 struct ext4_super_block *es)
2497 {
2498 unsigned int s_flags = sb->s_flags;
2499 int ret, nr_orphans = 0, nr_truncates = 0;
2500 #ifdef CONFIG_QUOTA
2501 int quota_update = 0;
2502 int i;
2503 #endif
2504 if (!es->s_last_orphan) {
2505 jbd_debug(4, "no orphan inodes to clean up\n");
2506 return;
2507 }
2509 if (bdev_read_only(sb->s_bdev)) {
2510 ext4_msg(sb, KERN_ERR, "write access "
2511 "unavailable, skipping orphan cleanup");
2512 return;
2513 }
2515 /* Check if feature set would not allow a r/w mount */
2516 if (!ext4_feature_set_ok(sb, 0)) {
2517 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2518 "unknown ROCOMPAT features");
2519 return;
2520 }
2522 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2523 /* don't clear list on RO mount w/ errors */
2524 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2525 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2526 "clearing orphan list.\n");
2527 es->s_last_orphan = 0;
2528 }
2529 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2530 return;
2531 }
2533 if (s_flags & SB_RDONLY) {
2534 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2535 sb->s_flags &= ~SB_RDONLY;
2536 }
2537 #ifdef CONFIG_QUOTA
2538 /* Needed for iput() to work correctly and not trash data */
2539 sb->s_flags |= SB_ACTIVE;
2541 /*
2542 * Turn on quotas which were not enabled for read-only mounts if
2543 * filesystem has quota feature, so that they are updated correctly.
2544 */
2545 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2546 int ret = ext4_enable_quotas(sb);
2548 if (!ret)
2549 quota_update = 1;
2550 else
2551 ext4_msg(sb, KERN_ERR,
2552 "Cannot turn on quotas: error %d", ret);
2553 }
2555 /* Turn on journaled quotas used for old sytle */
2556 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2557 if (EXT4_SB(sb)->s_qf_names[i]) {
2558 int ret = ext4_quota_on_mount(sb, i);
2560 if (!ret)
2561 quota_update = 1;
2562 else
2563 ext4_msg(sb, KERN_ERR,
2564 "Cannot turn on journaled "
2565 "quota: type %d: error %d", i, ret);
2566 }
2567 }
2568 #endif
2570 while (es->s_last_orphan) {
2571 struct inode *inode;
2573 /*
2574 * We may have encountered an error during cleanup; if
2575 * so, skip the rest.
2576 */
2577 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2578 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2579 es->s_last_orphan = 0;
2580 break;
2581 }
2583 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2584 if (IS_ERR(inode)) {
2585 es->s_last_orphan = 0;
2586 break;
2587 }
2589 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2590 dquot_initialize(inode);
2591 if (inode->i_nlink) {
2592 if (test_opt(sb, DEBUG))
2593 ext4_msg(sb, KERN_DEBUG,
2594 "%s: truncating inode %lu to %lld bytes",
2595 __func__, inode->i_ino, inode->i_size);
2596 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2597 inode->i_ino, inode->i_size);
2598 inode_lock(inode);
2599 truncate_inode_pages(inode->i_mapping, inode->i_size);
2600 ret = ext4_truncate(inode);
2601 if (ret)
2602 ext4_std_error(inode->i_sb, ret);
2603 inode_unlock(inode);
2604 nr_truncates++;
2605 } else {
2606 if (test_opt(sb, DEBUG))
2607 ext4_msg(sb, KERN_DEBUG,
2608 "%s: deleting unreferenced inode %lu",
2609 __func__, inode->i_ino);
2610 jbd_debug(2, "deleting unreferenced inode %lu\n",
2611 inode->i_ino);
2612 nr_orphans++;
2613 }
2614 iput(inode); /* The delete magic happens here! */
2615 }
2617 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2619 if (nr_orphans)
2620 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2621 PLURAL(nr_orphans));
2622 if (nr_truncates)
2623 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2624 PLURAL(nr_truncates));
2625 #ifdef CONFIG_QUOTA
2626 /* Turn off quotas if they were enabled for orphan cleanup */
2627 if (quota_update) {
2628 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2629 if (sb_dqopt(sb)->files[i])
2630 dquot_quota_off(sb, i);
2631 }
2632 }
2633 #endif
2634 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2635 }
2637 /*
2638 * Maximal extent format file size.
2639 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2640 * extent format containers, within a sector_t, and within i_blocks
2641 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2642 * so that won't be a limiting factor.
2643 *
2644 * However there is other limiting factor. We do store extents in the form
2645 * of starting block and length, hence the resulting length of the extent
2646 * covering maximum file size must fit into on-disk format containers as
2647 * well. Given that length is always by 1 unit bigger than max unit (because
2648 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2649 *
2650 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2651 */
2652 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2653 {
2654 loff_t res;
2655 loff_t upper_limit = MAX_LFS_FILESIZE;
2657 /* small i_blocks in vfs inode? */
2658 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2659 /*
2660 * CONFIG_LBDAF is not enabled implies the inode
2661 * i_block represent total blocks in 512 bytes
2662 * 32 == size of vfs inode i_blocks * 8
2663 */
2664 upper_limit = (1LL << 32) - 1;
2666 /* total blocks in file system block size */
2667 upper_limit >>= (blkbits - 9);
2668 upper_limit <<= blkbits;
2669 }
2671 /*
2672 * 32-bit extent-start container, ee_block. We lower the maxbytes
2673 * by one fs block, so ee_len can cover the extent of maximum file
2674 * size
2675 */
2676 res = (1LL << 32) - 1;
2677 res <<= blkbits;
2679 /* Sanity check against vm- & vfs- imposed limits */
2680 if (res > upper_limit)
2681 res = upper_limit;
2683 return res;
2684 }
2686 /*
2687 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2688 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2689 * We need to be 1 filesystem block less than the 2^48 sector limit.
2690 */
2691 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2692 {
2693 loff_t res = EXT4_NDIR_BLOCKS;
2694 int meta_blocks;
2695 loff_t upper_limit;
2696 /* This is calculated to be the largest file size for a dense, block
2697 * mapped file such that the file's total number of 512-byte sectors,
2698 * including data and all indirect blocks, does not exceed (2^48 - 1).
2699 *
2700 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2701 * number of 512-byte sectors of the file.
2702 */
2704 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2705 /*
2706 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2707 * the inode i_block field represents total file blocks in
2708 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2709 */
2710 upper_limit = (1LL << 32) - 1;
2712 /* total blocks in file system block size */
2713 upper_limit >>= (bits - 9);
2715 } else {
2716 /*
2717 * We use 48 bit ext4_inode i_blocks
2718 * With EXT4_HUGE_FILE_FL set the i_blocks
2719 * represent total number of blocks in
2720 * file system block size
2721 */
2722 upper_limit = (1LL << 48) - 1;
2724 }
2726 /* indirect blocks */
2727 meta_blocks = 1;
2728 /* double indirect blocks */
2729 meta_blocks += 1 + (1LL << (bits-2));
2730 /* tripple indirect blocks */
2731 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2733 upper_limit -= meta_blocks;
2734 upper_limit <<= bits;
2736 res += 1LL << (bits-2);
2737 res += 1LL << (2*(bits-2));
2738 res += 1LL << (3*(bits-2));
2739 res <<= bits;
2740 if (res > upper_limit)
2741 res = upper_limit;
2743 if (res > MAX_LFS_FILESIZE)
2744 res = MAX_LFS_FILESIZE;
2746 return res;
2747 }
2749 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2750 ext4_fsblk_t logical_sb_block, int nr)
2751 {
2752 struct ext4_sb_info *sbi = EXT4_SB(sb);
2753 ext4_group_t bg, first_meta_bg;
2754 int has_super = 0;
2756 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2758 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2759 return logical_sb_block + nr + 1;
2760 bg = sbi->s_desc_per_block * nr;
2761 if (ext4_bg_has_super(sb, bg))
2762 has_super = 1;
2764 /*
2765 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2766 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2767 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2768 * compensate.
2769 */
2770 if (sb->s_blocksize == 1024 && nr == 0 &&
2771 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
2772 has_super++;
2774 return (has_super + ext4_group_first_block_no(sb, bg));
2775 }
2777 /**
2778 * ext4_get_stripe_size: Get the stripe size.
2779 * @sbi: In memory super block info
2780 *
2781 * If we have specified it via mount option, then
2782 * use the mount option value. If the value specified at mount time is
2783 * greater than the blocks per group use the super block value.
2784 * If the super block value is greater than blocks per group return 0.
2785 * Allocator needs it be less than blocks per group.
2786 *
2787 */
2788 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2789 {
2790 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2791 unsigned long stripe_width =
2792 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2793 int ret;
2795 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2796 ret = sbi->s_stripe;
2797 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2798 ret = stripe_width;
2799 else if (stride && stride <= sbi->s_blocks_per_group)
2800 ret = stride;
2801 else
2802 ret = 0;
2804 /*
2805 * If the stripe width is 1, this makes no sense and
2806 * we set it to 0 to turn off stripe handling code.
2807 */
2808 if (ret <= 1)
2809 ret = 0;
2811 return ret;
2812 }
2814 /*
2815 * Check whether this filesystem can be mounted based on
2816 * the features present and the RDONLY/RDWR mount requested.
2817 * Returns 1 if this filesystem can be mounted as requested,
2818 * 0 if it cannot be.
2819 */
2820 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2821 {
2822 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2823 ext4_msg(sb, KERN_ERR,
2824 "Couldn't mount because of "
2825 "unsupported optional features (%x)",
2826 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2827 ~EXT4_FEATURE_INCOMPAT_SUPP));
2828 return 0;
2829 }
2831 if (readonly)
2832 return 1;
2834 if (ext4_has_feature_readonly(sb)) {
2835 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2836 sb->s_flags |= SB_RDONLY;
2837 return 1;
2838 }
2840 /* Check that feature set is OK for a read-write mount */
2841 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2842 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2843 "unsupported optional features (%x)",
2844 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2845 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2846 return 0;
2847 }
2848 /*
2849 * Large file size enabled file system can only be mounted
2850 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2851 */
2852 if (ext4_has_feature_huge_file(sb)) {
2853 if (sizeof(blkcnt_t) < sizeof(u64)) {
2854 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2855 "cannot be mounted RDWR without "
2856 "CONFIG_LBDAF");
2857 return 0;
2858 }
2859 }
2860 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2861 ext4_msg(sb, KERN_ERR,
2862 "Can't support bigalloc feature without "
2863 "extents feature\n");
2864 return 0;
2865 }
2867 #ifndef CONFIG_QUOTA
2868 if (ext4_has_feature_quota(sb) && !readonly) {
2869 ext4_msg(sb, KERN_ERR,
2870 "Filesystem with quota feature cannot be mounted RDWR "
2871 "without CONFIG_QUOTA");
2872 return 0;
2873 }
2874 if (ext4_has_feature_project(sb) && !readonly) {
2875 ext4_msg(sb, KERN_ERR,
2876 "Filesystem with project quota feature cannot be mounted RDWR "
2877 "without CONFIG_QUOTA");
2878 return 0;
2879 }
2880 #endif /* CONFIG_QUOTA */
2881 return 1;
2882 }
2884 /*
2885 * This function is called once a day if we have errors logged
2886 * on the file system
2887 */
2888 static void print_daily_error_info(struct timer_list *t)
2889 {
2890 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2891 struct super_block *sb = sbi->s_sb;
2892 struct ext4_super_block *es = sbi->s_es;
2894 if (es->s_error_count)
2895 /* fsck newer than v1.41.13 is needed to clean this condition. */
2896 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2897 le32_to_cpu(es->s_error_count));
2898 if (es->s_first_error_time) {
2899 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
2900 sb->s_id,
2901 ext4_get_tstamp(es, s_first_error_time),
2902 (int) sizeof(es->s_first_error_func),
2903 es->s_first_error_func,
2904 le32_to_cpu(es->s_first_error_line));
2905 if (es->s_first_error_ino)
2906 printk(KERN_CONT ": inode %u",
2907 le32_to_cpu(es->s_first_error_ino));
2908 if (es->s_first_error_block)
2909 printk(KERN_CONT ": block %llu", (unsigned long long)
2910 le64_to_cpu(es->s_first_error_block));
2911 printk(KERN_CONT "\n");
2912 }
2913 if (es->s_last_error_time) {
2914 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
2915 sb->s_id,
2916 ext4_get_tstamp(es, s_last_error_time),
2917 (int) sizeof(es->s_last_error_func),
2918 es->s_last_error_func,
2919 le32_to_cpu(es->s_last_error_line));
2920 if (es->s_last_error_ino)
2921 printk(KERN_CONT ": inode %u",
2922 le32_to_cpu(es->s_last_error_ino));
2923 if (es->s_last_error_block)
2924 printk(KERN_CONT ": block %llu", (unsigned long long)
2925 le64_to_cpu(es->s_last_error_block));
2926 printk(KERN_CONT "\n");
2927 }
2928 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2929 }
2931 /* Find next suitable group and run ext4_init_inode_table */
2932 static int ext4_run_li_request(struct ext4_li_request *elr)
2933 {
2934 struct ext4_group_desc *gdp = NULL;
2935 ext4_group_t group, ngroups;
2936 struct super_block *sb;
2937 unsigned long timeout = 0;
2938 int ret = 0;
2940 sb = elr->lr_super;
2941 ngroups = EXT4_SB(sb)->s_groups_count;
2943 for (group = elr->lr_next_group; group < ngroups; group++) {
2944 gdp = ext4_get_group_desc(sb, group, NULL);
2945 if (!gdp) {
2946 ret = 1;
2947 break;
2948 }
2950 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2951 break;
2952 }
2954 if (group >= ngroups)
2955 ret = 1;
2957 if (!ret) {
2958 timeout = jiffies;
2959 ret = ext4_init_inode_table(sb, group,
2960 elr->lr_timeout ? 0 : 1);
2961 if (elr->lr_timeout == 0) {
2962 timeout = (jiffies - timeout) *
2963 elr->lr_sbi->s_li_wait_mult;
2964 elr->lr_timeout = timeout;
2965 }
2966 elr->lr_next_sched = jiffies + elr->lr_timeout;
2967 elr->lr_next_group = group + 1;
2968 }
2969 return ret;
2970 }
2972 /*
2973 * Remove lr_request from the list_request and free the
2974 * request structure. Should be called with li_list_mtx held
2975 */
2976 static void ext4_remove_li_request(struct ext4_li_request *elr)
2977 {
2978 struct ext4_sb_info *sbi;
2980 if (!elr)
2981 return;
2983 sbi = elr->lr_sbi;
2985 list_del(&elr->lr_request);
2986 sbi->s_li_request = NULL;
2987 kfree(elr);
2988 }
2990 static void ext4_unregister_li_request(struct super_block *sb)
2991 {
2992 mutex_lock(&ext4_li_mtx);
2993 if (!ext4_li_info) {
2994 mutex_unlock(&ext4_li_mtx);
2995 return;
2996 }
2998 mutex_lock(&ext4_li_info->li_list_mtx);
2999 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3000 mutex_unlock(&ext4_li_info->li_list_mtx);
3001 mutex_unlock(&ext4_li_mtx);
3002 }
3004 static struct task_struct *ext4_lazyinit_task;
3006 /*
3007 * This is the function where ext4lazyinit thread lives. It walks
3008 * through the request list searching for next scheduled filesystem.
3009 * When such a fs is found, run the lazy initialization request
3010 * (ext4_rn_li_request) and keep track of the time spend in this
3011 * function. Based on that time we compute next schedule time of
3012 * the request. When walking through the list is complete, compute
3013 * next waking time and put itself into sleep.
3014 */
3015 static int ext4_lazyinit_thread(void *arg)
3016 {
3017 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3018 struct list_head *pos, *n;
3019 struct ext4_li_request *elr;
3020 unsigned long next_wakeup, cur;
3022 BUG_ON(NULL == eli);
3024 cont_thread:
3025 while (true) {
3026 next_wakeup = MAX_JIFFY_OFFSET;
3028 mutex_lock(&eli->li_list_mtx);
3029 if (list_empty(&eli->li_request_list)) {
3030 mutex_unlock(&eli->li_list_mtx);
3031 goto exit_thread;
3032 }
3033 list_for_each_safe(pos, n, &eli->li_request_list) {
3034 int err = 0;
3035 int progress = 0;
3036 elr = list_entry(pos, struct ext4_li_request,
3037 lr_request);
3039 if (time_before(jiffies, elr->lr_next_sched)) {
3040 if (time_before(elr->lr_next_sched, next_wakeup))
3041 next_wakeup = elr->lr_next_sched;
3042 continue;
3043 }
3044 if (down_read_trylock(&elr->lr_super->s_umount)) {
3045 if (sb_start_write_trylock(elr->lr_super)) {
3046 progress = 1;
3047 /*
3048 * We hold sb->s_umount, sb can not
3049 * be removed from the list, it is
3050 * now safe to drop li_list_mtx
3051 */
3052 mutex_unlock(&eli->li_list_mtx);
3053 err = ext4_run_li_request(elr);
3054 sb_end_write(elr->lr_super);
3055 mutex_lock(&eli->li_list_mtx);
3056 n = pos->next;
3057 }
3058 up_read((&elr->lr_super->s_umount));
3059 }
3060 /* error, remove the lazy_init job */
3061 if (err) {
3062 ext4_remove_li_request(elr);
3063 continue;
3064 }
3065 if (!progress) {
3066 elr->lr_next_sched = jiffies +
3067 (prandom_u32()
3068 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3069 }
3070 if (time_before(elr->lr_next_sched, next_wakeup))
3071 next_wakeup = elr->lr_next_sched;
3072 }
3073 mutex_unlock(&eli->li_list_mtx);
3075 try_to_freeze();
3077 cur = jiffies;
3078 if ((time_after_eq(cur, next_wakeup)) ||
3079 (MAX_JIFFY_OFFSET == next_wakeup)) {
3080 cond_resched();
3081 continue;
3082 }
3084 schedule_timeout_interruptible(next_wakeup - cur);
3086 if (kthread_should_stop()) {
3087 ext4_clear_request_list();
3088 goto exit_thread;
3089 }
3090 }
3092 exit_thread:
3093 /*
3094 * It looks like the request list is empty, but we need
3095 * to check it under the li_list_mtx lock, to prevent any
3096 * additions into it, and of course we should lock ext4_li_mtx
3097 * to atomically free the list and ext4_li_info, because at
3098 * this point another ext4 filesystem could be registering
3099 * new one.
3100 */
3101 mutex_lock(&ext4_li_mtx);
3102 mutex_lock(&eli->li_list_mtx);
3103 if (!list_empty(&eli->li_request_list)) {
3104 mutex_unlock(&eli->li_list_mtx);
3105 mutex_unlock(&ext4_li_mtx);
3106 goto cont_thread;
3107 }
3108 mutex_unlock(&eli->li_list_mtx);
3109 kfree(ext4_li_info);
3110 ext4_li_info = NULL;
3111 mutex_unlock(&ext4_li_mtx);
3113 return 0;
3114 }
3116 static void ext4_clear_request_list(void)
3117 {
3118 struct list_head *pos, *n;
3119 struct ext4_li_request *elr;
3121 mutex_lock(&ext4_li_info->li_list_mtx);
3122 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3123 elr = list_entry(pos, struct ext4_li_request,
3124 lr_request);
3125 ext4_remove_li_request(elr);
3126 }
3127 mutex_unlock(&ext4_li_info->li_list_mtx);
3128 }
3130 static int ext4_run_lazyinit_thread(void)
3131 {
3132 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3133 ext4_li_info, "ext4lazyinit");
3134 if (IS_ERR(ext4_lazyinit_task)) {
3135 int err = PTR_ERR(ext4_lazyinit_task);
3136 ext4_clear_request_list();
3137 kfree(ext4_li_info);
3138 ext4_li_info = NULL;
3139 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3140 "initialization thread\n",
3141 err);
3142 return err;
3143 }
3144 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3145 return 0;
3146 }
3148 /*
3149 * Check whether it make sense to run itable init. thread or not.
3150 * If there is at least one uninitialized inode table, return
3151 * corresponding group number, else the loop goes through all
3152 * groups and return total number of groups.
3153 */
3154 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3155 {
3156 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3157 struct ext4_group_desc *gdp = NULL;
3159 if (!ext4_has_group_desc_csum(sb))
3160 return ngroups;
3162 for (group = 0; group < ngroups; group++) {
3163 gdp = ext4_get_group_desc(sb, group, NULL);
3164 if (!gdp)
3165 continue;
3167 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3168 break;
3169 }
3171 return group;
3172 }
3174 static int ext4_li_info_new(void)
3175 {
3176 struct ext4_lazy_init *eli = NULL;
3178 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3179 if (!eli)
3180 return -ENOMEM;
3182 INIT_LIST_HEAD(&eli->li_request_list);
3183 mutex_init(&eli->li_list_mtx);
3185 eli->li_state |= EXT4_LAZYINIT_QUIT;
3187 ext4_li_info = eli;
3189 return 0;
3190 }
3192 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3193 ext4_group_t start)
3194 {
3195 struct ext4_sb_info *sbi = EXT4_SB(sb);
3196 struct ext4_li_request *elr;
3198 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3199 if (!elr)
3200 return NULL;
3202 elr->lr_super = sb;
3203 elr->lr_sbi = sbi;
3204 elr->lr_next_group = start;
3206 /*
3207 * Randomize first schedule time of the request to
3208 * spread the inode table initialization requests
3209 * better.
3210 */
3211 elr->lr_next_sched = jiffies + (prandom_u32() %
3212 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3213 return elr;
3214 }
3216 int ext4_register_li_request(struct super_block *sb,
3217 ext4_group_t first_not_zeroed)
3218 {
3219 struct ext4_sb_info *sbi = EXT4_SB(sb);
3220 struct ext4_li_request *elr = NULL;
3221 ext4_group_t ngroups = sbi->s_groups_count;
3222 int ret = 0;
3224 mutex_lock(&ext4_li_mtx);
3225 if (sbi->s_li_request != NULL) {
3226 /*
3227 * Reset timeout so it can be computed again, because
3228 * s_li_wait_mult might have changed.
3229 */
3230 sbi->s_li_request->lr_timeout = 0;
3231 goto out;
3232 }
3234 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3235 !test_opt(sb, INIT_INODE_TABLE))
3236 goto out;
3238 elr = ext4_li_request_new(sb, first_not_zeroed);
3239 if (!elr) {
3240 ret = -ENOMEM;
3241 goto out;
3242 }
3244 if (NULL == ext4_li_info) {
3245 ret = ext4_li_info_new();
3246 if (ret)
3247 goto out;
3248 }
3250 mutex_lock(&ext4_li_info->li_list_mtx);
3251 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3252 mutex_unlock(&ext4_li_info->li_list_mtx);
3254 sbi->s_li_request = elr;
3255 /*
3256 * set elr to NULL here since it has been inserted to
3257 * the request_list and the removal and free of it is
3258 * handled by ext4_clear_request_list from now on.
3259 */
3260 elr = NULL;
3262 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3263 ret = ext4_run_lazyinit_thread();
3264 if (ret)
3265 goto out;
3266 }
3267 out:
3268 mutex_unlock(&ext4_li_mtx);
3269 if (ret)
3270 kfree(elr);
3271 return ret;
3272 }
3274 /*
3275 * We do not need to lock anything since this is called on
3276 * module unload.
3277 */
3278 static void ext4_destroy_lazyinit_thread(void)
3279 {
3280 /*
3281 * If thread exited earlier
3282 * there's nothing to be done.
3283 */
3284 if (!ext4_li_info || !ext4_lazyinit_task)
3285 return;
3287 kthread_stop(ext4_lazyinit_task);
3288 }
3290 static int set_journal_csum_feature_set(struct super_block *sb)
3291 {
3292 int ret = 1;
3293 int compat, incompat;
3294 struct ext4_sb_info *sbi = EXT4_SB(sb);
3296 if (ext4_has_metadata_csum(sb)) {
3297 /* journal checksum v3 */
3298 compat = 0;
3299 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3300 } else {
3301 /* journal checksum v1 */
3302 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3303 incompat = 0;
3304 }
3306 jbd2_journal_clear_features(sbi->s_journal,
3307 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3308 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3309 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3310 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3311 ret = jbd2_journal_set_features(sbi->s_journal,
3312 compat, 0,
3313 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3314 incompat);
3315 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3316 ret = jbd2_journal_set_features(sbi->s_journal,
3317 compat, 0,
3318 incompat);
3319 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3320 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3321 } else {
3322 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3323 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3324 }
3326 return ret;
3327 }
3329 /*
3330 * Note: calculating the overhead so we can be compatible with
3331 * historical BSD practice is quite difficult in the face of
3332 * clusters/bigalloc. This is because multiple metadata blocks from
3333 * different block group can end up in the same allocation cluster.
3334 * Calculating the exact overhead in the face of clustered allocation
3335 * requires either O(all block bitmaps) in memory or O(number of block
3336 * groups**2) in time. We will still calculate the superblock for
3337 * older file systems --- and if we come across with a bigalloc file
3338 * system with zero in s_overhead_clusters the estimate will be close to
3339 * correct especially for very large cluster sizes --- but for newer
3340 * file systems, it's better to calculate this figure once at mkfs
3341 * time, and store it in the superblock. If the superblock value is
3342 * present (even for non-bigalloc file systems), we will use it.
3343 */
3344 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3345 char *buf)
3346 {
3347 struct ext4_sb_info *sbi = EXT4_SB(sb);
3348 struct ext4_group_desc *gdp;
3349 ext4_fsblk_t first_block, last_block, b;
3350 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3351 int s, j, count = 0;
3353 if (!ext4_has_feature_bigalloc(sb))
3354 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3355 sbi->s_itb_per_group + 2);
3357 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3358 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3359 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3360 for (i = 0; i < ngroups; i++) {
3361 gdp = ext4_get_group_desc(sb, i, NULL);
3362 b = ext4_block_bitmap(sb, gdp);
3363 if (b >= first_block && b <= last_block) {
3364 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3365 count++;
3366 }
3367 b = ext4_inode_bitmap(sb, gdp);
3368 if (b >= first_block && b <= last_block) {
3369 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3370 count++;
3371 }
3372 b = ext4_inode_table(sb, gdp);
3373 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3374 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3375 int c = EXT4_B2C(sbi, b - first_block);
3376 ext4_set_bit(c, buf);
3377 count++;
3378 }
3379 if (i != grp)
3380 continue;
3381 s = 0;
3382 if (ext4_bg_has_super(sb, grp)) {
3383 ext4_set_bit(s++, buf);
3384 count++;
3385 }
3386 j = ext4_bg_num_gdb(sb, grp);
3387 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3388 ext4_error(sb, "Invalid number of block group "
3389 "descriptor blocks: %d", j);
3390 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3391 }
3392 count += j;
3393 for (; j > 0; j--)
3394 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3395 }
3396 if (!count)
3397 return 0;
3398 return EXT4_CLUSTERS_PER_GROUP(sb) -
3399 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3400 }
3402 /*
3403 * Compute the overhead and stash it in sbi->s_overhead
3404 */
3405 int ext4_calculate_overhead(struct super_block *sb)
3406 {
3407 struct ext4_sb_info *sbi = EXT4_SB(sb);
3408 struct ext4_super_block *es = sbi->s_es;
3409 struct inode *j_inode;
3410 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3411 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3412 ext4_fsblk_t overhead = 0;
3413 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3415 if (!buf)
3416 return -ENOMEM;
3418 /*
3419 * Compute the overhead (FS structures). This is constant
3420 * for a given filesystem unless the number of block groups
3421 * changes so we cache the previous value until it does.
3422 */
3424 /*
3425 * All of the blocks before first_data_block are overhead
3426 */
3427 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3429 /*
3430 * Add the overhead found in each block group
3431 */
3432 for (i = 0; i < ngroups; i++) {
3433 int blks;
3435 blks = count_overhead(sb, i, buf);
3436 overhead += blks;
3437 if (blks)
3438 memset(buf, 0, PAGE_SIZE);
3439 cond_resched();
3440 }
3442 /*
3443 * Add the internal journal blocks whether the journal has been
3444 * loaded or not
3445 */
3446 if (sbi->s_journal && !sbi->journal_bdev)
3447 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3448 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3449 j_inode = ext4_get_journal_inode(sb, j_inum);
3450 if (j_inode) {
3451 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3452 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3453 iput(j_inode);
3454 } else {
3455 ext4_msg(sb, KERN_ERR, "can't get journal size");
3456 }
3457 }
3458 sbi->s_overhead = overhead;
3459 smp_wmb();
3460 free_page((unsigned long) buf);
3461 return 0;
3462 }
3464 static void ext4_set_resv_clusters(struct super_block *sb)
3465 {
3466 ext4_fsblk_t resv_clusters;
3467 struct ext4_sb_info *sbi = EXT4_SB(sb);
3469 /*
3470 * There's no need to reserve anything when we aren't using extents.
3471 * The space estimates are exact, there are no unwritten extents,
3472 * hole punching doesn't need new metadata... This is needed especially
3473 * to keep ext2/3 backward compatibility.
3474 */
3475 if (!ext4_has_feature_extents(sb))
3476 return;
3477 /*
3478 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3479 * This should cover the situations where we can not afford to run
3480 * out of space like for example punch hole, or converting
3481 * unwritten extents in delalloc path. In most cases such
3482 * allocation would require 1, or 2 blocks, higher numbers are
3483 * very rare.
3484 */
3485 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3486 sbi->s_cluster_bits);
3488 do_div(resv_clusters, 50);
3489 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3491 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3492 }
3494 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3495 {
3496 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3497 char *orig_data = kstrdup(data, GFP_KERNEL);
3498 struct buffer_head *bh;
3499 struct ext4_super_block *es = NULL;
3500 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3501 ext4_fsblk_t block;
3502 ext4_fsblk_t sb_block = get_sb_block(&data);
3503 ext4_fsblk_t logical_sb_block;
3504 unsigned long offset = 0;
3505 unsigned long journal_devnum = 0;
3506 unsigned long def_mount_opts;
3507 struct inode *root;
3508 const char *descr;
3509 int ret = -ENOMEM;
3510 int blocksize, clustersize;
3511 unsigned int db_count;
3512 unsigned int i;
3513 int needs_recovery, has_huge_files, has_bigalloc;
3514 __u64 blocks_count;
3515 int err = 0;
3516 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3517 ext4_group_t first_not_zeroed;
3519 if ((data && !orig_data) || !sbi)
3520 goto out_free_base;
3522 sbi->s_daxdev = dax_dev;
3523 sbi->s_blockgroup_lock =
3524 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3525 if (!sbi->s_blockgroup_lock)
3526 goto out_free_base;
3528 sb->s_fs_info = sbi;
3529 sbi->s_sb = sb;
3530 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3531 sbi->s_sb_block = sb_block;
3532 if (sb->s_bdev->bd_part)
3533 sbi->s_sectors_written_start =
3534 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
3536 /* Cleanup superblock name */
3537 strreplace(sb->s_id, '/', '!');
3539 /* -EINVAL is default */
3540 ret = -EINVAL;
3541 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3542 if (!blocksize) {
3543 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3544 goto out_fail;
3545 }
3547 /*
3548 * The ext4 superblock will not be buffer aligned for other than 1kB
3549 * block sizes. We need to calculate the offset from buffer start.
3550 */
3551 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3552 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3553 offset = do_div(logical_sb_block, blocksize);
3554 } else {
3555 logical_sb_block = sb_block;
3556 }
3558 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3559 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3560 goto out_fail;
3561 }
3562 /*
3563 * Note: s_es must be initialized as soon as possible because
3564 * some ext4 macro-instructions depend on its value
3565 */
3566 es = (struct ext4_super_block *) (bh->b_data + offset);
3567 sbi->s_es = es;
3568 sb->s_magic = le16_to_cpu(es->s_magic);
3569 if (sb->s_magic != EXT4_SUPER_MAGIC)
3570 goto cantfind_ext4;
3571 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3573 /* Warn if metadata_csum and gdt_csum are both set. */
3574 if (ext4_has_feature_metadata_csum(sb) &&
3575 ext4_has_feature_gdt_csum(sb))
3576 ext4_warning(sb, "metadata_csum and uninit_bg are "
3577 "redundant flags; please run fsck.");
3579 /* Check for a known checksum algorithm */
3580 if (!ext4_verify_csum_type(sb, es)) {
3581 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3582 "unknown checksum algorithm.");
3583 silent = 1;
3584 goto cantfind_ext4;
3585 }
3587 /* Load the checksum driver */
3588 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3589 if (IS_ERR(sbi->s_chksum_driver)) {
3590 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3591 ret = PTR_ERR(sbi->s_chksum_driver);
3592 sbi->s_chksum_driver = NULL;
3593 goto failed_mount;
3594 }
3596 /* Check superblock checksum */
3597 if (!ext4_superblock_csum_verify(sb, es)) {
3598 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3599 "invalid superblock checksum. Run e2fsck?");
3600 silent = 1;
3601 ret = -EFSBADCRC;
3602 goto cantfind_ext4;
3603 }
3605 /* Precompute checksum seed for all metadata */
3606 if (ext4_has_feature_csum_seed(sb))
3607 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3608 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3609 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3610 sizeof(es->s_uuid));
3612 /* Set defaults before we parse the mount options */
3613 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3614 set_opt(sb, INIT_INODE_TABLE);
3615 if (def_mount_opts & EXT4_DEFM_DEBUG)
3616 set_opt(sb, DEBUG);
3617 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3618 set_opt(sb, GRPID);
3619 if (def_mount_opts & EXT4_DEFM_UID16)
3620 set_opt(sb, NO_UID32);
3621 /* xattr user namespace & acls are now defaulted on */
3622 set_opt(sb, XATTR_USER);
3623 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3624 set_opt(sb, POSIX_ACL);
3625 #endif
3626 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3627 if (ext4_has_metadata_csum(sb))
3628 set_opt(sb, JOURNAL_CHECKSUM);
3630 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3631 set_opt(sb, JOURNAL_DATA);
3632 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3633 set_opt(sb, ORDERED_DATA);
3634 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3635 set_opt(sb, WRITEBACK_DATA);
3637 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3638 set_opt(sb, ERRORS_PANIC);
3639 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3640 set_opt(sb, ERRORS_CONT);
3641 else
3642 set_opt(sb, ERRORS_RO);
3643 /* block_validity enabled by default; disable with noblock_validity */
3644 set_opt(sb, BLOCK_VALIDITY);
3645 if (def_mount_opts & EXT4_DEFM_DISCARD)
3646 set_opt(sb, DISCARD);
3648 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3649 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3650 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3651 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3652 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3654 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3655 set_opt(sb, BARRIER);
3657 /*
3658 * enable delayed allocation by default
3659 * Use -o nodelalloc to turn it off
3660 */
3661 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3662 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3663 set_opt(sb, DELALLOC);
3665 /*
3666 * set default s_li_wait_mult for lazyinit, for the case there is
3667 * no mount option specified.
3668 */
3669 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3671 if (sbi->s_es->s_mount_opts[0]) {
3672 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3673 sizeof(sbi->s_es->s_mount_opts),
3674 GFP_KERNEL);
3675 if (!s_mount_opts)
3676 goto failed_mount;
3677 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3678 &journal_ioprio, 0)) {
3679 ext4_msg(sb, KERN_WARNING,
3680 "failed to parse options in superblock: %s",
3681 s_mount_opts);
3682 }
3683 kfree(s_mount_opts);
3684 }
3685 sbi->s_def_mount_opt = sbi->s_mount_opt;
3686 if (!parse_options((char *) data, sb, &journal_devnum,
3687 &journal_ioprio, 0))
3688 goto failed_mount;
3690 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3691 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3692 "with data=journal disables delayed "
3693 "allocation and O_DIRECT support!\n");
3694 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3695 ext4_msg(sb, KERN_ERR, "can't mount with "
3696 "both data=journal and delalloc");
3697 goto failed_mount;
3698 }
3699 if (test_opt(sb, DIOREAD_NOLOCK)) {
3700 ext4_msg(sb, KERN_ERR, "can't mount with "
3701 "both data=journal and dioread_nolock");
3702 goto failed_mount;
3703 }
3704 if (test_opt(sb, DAX)) {
3705 ext4_msg(sb, KERN_ERR, "can't mount with "
3706 "both data=journal and dax");
3707 goto failed_mount;
3708 }
3709 if (ext4_has_feature_encrypt(sb)) {
3710 ext4_msg(sb, KERN_WARNING,
3711 "encrypted files will use data=ordered "
3712 "instead of data journaling mode");
3713 }
3714 if (test_opt(sb, DELALLOC))
3715 clear_opt(sb, DELALLOC);
3716 } else {
3717 sb->s_iflags |= SB_I_CGROUPWB;
3718 }
3720 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3721 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
3723 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3724 (ext4_has_compat_features(sb) ||
3725 ext4_has_ro_compat_features(sb) ||
3726 ext4_has_incompat_features(sb)))
3727 ext4_msg(sb, KERN_WARNING,
3728 "feature flags set on rev 0 fs, "
3729 "running e2fsck is recommended");
3731 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3732 set_opt2(sb, HURD_COMPAT);
3733 if (ext4_has_feature_64bit(sb)) {
3734 ext4_msg(sb, KERN_ERR,
3735 "The Hurd can't support 64-bit file systems");
3736 goto failed_mount;
3737 }
3739 /*
3740 * ea_inode feature uses l_i_version field which is not
3741 * available in HURD_COMPAT mode.
3742 */
3743 if (ext4_has_feature_ea_inode(sb)) {
3744 ext4_msg(sb, KERN_ERR,
3745 "ea_inode feature is not supported for Hurd");
3746 goto failed_mount;
3747 }
3748 }
3750 if (IS_EXT2_SB(sb)) {
3751 if (ext2_feature_set_ok(sb))
3752 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3753 "using the ext4 subsystem");
3754 else {
3755 /*
3756 * If we're probing be silent, if this looks like
3757 * it's actually an ext[34] filesystem.
3758 */
3759 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3760 goto failed_mount;
3761 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3762 "to feature incompatibilities");
3763 goto failed_mount;
3764 }
3765 }
3767 if (IS_EXT3_SB(sb)) {
3768 if (ext3_feature_set_ok(sb))
3769 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3770 "using the ext4 subsystem");
3771 else {
3772 /*
3773 * If we're probing be silent, if this looks like
3774 * it's actually an ext4 filesystem.
3775 */
3776 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
3777 goto failed_mount;
3778 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3779 "to feature incompatibilities");
3780 goto failed_mount;
3781 }
3782 }
3784 /*
3785 * Check feature flags regardless of the revision level, since we
3786 * previously didn't change the revision level when setting the flags,
3787 * so there is a chance incompat flags are set on a rev 0 filesystem.
3788 */
3789 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
3790 goto failed_mount;
3792 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3793 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3794 blocksize > EXT4_MAX_BLOCK_SIZE) {
3795 ext4_msg(sb, KERN_ERR,
3796 "Unsupported filesystem blocksize %d (%d log_block_size)",
3797 blocksize, le32_to_cpu(es->s_log_block_size));
3798 goto failed_mount;
3799 }
3800 if (le32_to_cpu(es->s_log_block_size) >
3801 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3802 ext4_msg(sb, KERN_ERR,
3803 "Invalid log block size: %u",
3804 le32_to_cpu(es->s_log_block_size));
3805 goto failed_mount;
3806 }
3807 if (le32_to_cpu(es->s_log_cluster_size) >
3808 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3809 ext4_msg(sb, KERN_ERR,
3810 "Invalid log cluster size: %u",
3811 le32_to_cpu(es->s_log_cluster_size));
3812 goto failed_mount;
3813 }
3815 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3816 ext4_msg(sb, KERN_ERR,
3817 "Number of reserved GDT blocks insanely large: %d",
3818 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3819 goto failed_mount;
3820 }
3822 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3823 if (ext4_has_feature_inline_data(sb)) {
3824 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
3825 " that may contain inline data");
3826 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3827 }
3828 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
3829 ext4_msg(sb, KERN_ERR,
3830 "DAX unsupported by block device. Turning off DAX.");
3831 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX;
3832 }
3833 }
3835 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3836 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3837 es->s_encryption_level);
3838 goto failed_mount;
3839 }
3841 if (sb->s_blocksize != blocksize) {
3842 /* Validate the filesystem blocksize */
3843 if (!sb_set_blocksize(sb, blocksize)) {
3844 ext4_msg(sb, KERN_ERR, "bad block size %d",
3845 blocksize);
3846 goto failed_mount;
3847 }
3849 brelse(bh);
3850 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3851 offset = do_div(logical_sb_block, blocksize);
3852 bh = sb_bread_unmovable(sb, logical_sb_block);
3853 if (!bh) {
3854 ext4_msg(sb, KERN_ERR,
3855 "Can't read superblock on 2nd try");
3856 goto failed_mount;
3857 }
3858 es = (struct ext4_super_block *)(bh->b_data + offset);
3859 sbi->s_es = es;
3860 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3861 ext4_msg(sb, KERN_ERR,
3862 "Magic mismatch, very weird!");
3863 goto failed_mount;
3864 }
3865 }
3867 has_huge_files = ext4_has_feature_huge_file(sb);
3868 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3869 has_huge_files);
3870 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3872 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3873 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3874 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3875 } else {
3876 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3877 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3878 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3879 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3880 sbi->s_first_ino);
3881 goto failed_mount;
3882 }
3883 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3884 (!is_power_of_2(sbi->s_inode_size)) ||
3885 (sbi->s_inode_size > blocksize)) {
3886 ext4_msg(sb, KERN_ERR,
3887 "unsupported inode size: %d",
3888 sbi->s_inode_size);
3889 goto failed_mount;
3890 }
3891 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3892 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3893 }
3895 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3896 if (ext4_has_feature_64bit(sb)) {
3897 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3898 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3899 !is_power_of_2(sbi->s_desc_size)) {
3900 ext4_msg(sb, KERN_ERR,
3901 "unsupported descriptor size %lu",
3902 sbi->s_desc_size);
3903 goto failed_mount;
3904 }
3905 } else
3906 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3908 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3909 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3911 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3912 if (sbi->s_inodes_per_block == 0)
3913 goto cantfind_ext4;
3914 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3915 sbi->s_inodes_per_group > blocksize * 8) {
3916 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3917 sbi->s_blocks_per_group);
3918 goto failed_mount;
3919 }
3920 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3921 sbi->s_inodes_per_block;
3922 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3923 sbi->s_sbh = bh;
3924 sbi->s_mount_state = le16_to_cpu(es->s_state);
3925 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3926 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3928 for (i = 0; i < 4; i++)
3929 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3930 sbi->s_def_hash_version = es->s_def_hash_version;
3931 if (ext4_has_feature_dir_index(sb)) {
3932 i = le32_to_cpu(es->s_flags);
3933 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3934 sbi->s_hash_unsigned = 3;
3935 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3936 #ifdef __CHAR_UNSIGNED__
3937 if (!sb_rdonly(sb))
3938 es->s_flags |=
3939 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3940 sbi->s_hash_unsigned = 3;
3941 #else
3942 if (!sb_rdonly(sb))
3943 es->s_flags |=
3944 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3945 #endif
3946 }
3947 }
3949 /* Handle clustersize */
3950 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3951 has_bigalloc = ext4_has_feature_bigalloc(sb);
3952 if (has_bigalloc) {
3953 if (clustersize < blocksize) {
3954 ext4_msg(sb, KERN_ERR,
3955 "cluster size (%d) smaller than "
3956 "block size (%d)", clustersize, blocksize);
3957 goto failed_mount;
3958 }
3959 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3960 le32_to_cpu(es->s_log_block_size);
3961 sbi->s_clusters_per_group =
3962 le32_to_cpu(es->s_clusters_per_group);
3963 if (sbi->s_clusters_per_group > blocksize * 8) {
3964 ext4_msg(sb, KERN_ERR,
3965 "#clusters per group too big: %lu",
3966 sbi->s_clusters_per_group);
3967 goto failed_mount;
3968 }
3969 if (sbi->s_blocks_per_group !=
3970 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3971 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3972 "clusters per group (%lu) inconsistent",
3973 sbi->s_blocks_per_group,
3974 sbi->s_clusters_per_group);
3975 goto failed_mount;
3976 }
3977 } else {
3978 if (clustersize != blocksize) {
3979 ext4_msg(sb, KERN_ERR,
3980 "fragment/cluster size (%d) != "
3981 "block size (%d)", clustersize, blocksize);
3982 goto failed_mount;
3983 }
3984 if (sbi->s_blocks_per_group > blocksize * 8) {
3985 ext4_msg(sb, KERN_ERR,
3986 "#blocks per group too big: %lu",
3987 sbi->s_blocks_per_group);
3988 goto failed_mount;
3989 }
3990 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3991 sbi->s_cluster_bits = 0;
3992 }
3993 sbi->s_cluster_ratio = clustersize / blocksize;
3995 /* Do we have standard group size of clustersize * 8 blocks ? */
3996 if (sbi->s_blocks_per_group == clustersize << 3)
3997 set_opt2(sb, STD_GROUP_SIZE);
3999 /*
4000 * Test whether we have more sectors than will fit in sector_t,
4001 * and whether the max offset is addressable by the page cache.
4002 */
4003 err = generic_check_addressable(sb->s_blocksize_bits,
4004 ext4_blocks_count(es));
4005 if (err) {
4006 ext4_msg(sb, KERN_ERR, "filesystem"
4007 " too large to mount safely on this system");
4008 if (sizeof(sector_t) < 8)
4009 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
4010 goto failed_mount;
4011 }
4013 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4014 goto cantfind_ext4;
4016 /* check blocks count against device size */
4017 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4018 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4019 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4020 "exceeds size of device (%llu blocks)",
4021 ext4_blocks_count(es), blocks_count);
4022 goto failed_mount;
4023 }
4025 /*
4026 * It makes no sense for the first data block to be beyond the end
4027 * of the filesystem.
4028 */
4029 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4030 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4031 "block %u is beyond end of filesystem (%llu)",
4032 le32_to_cpu(es->s_first_data_block),
4033 ext4_blocks_count(es));
4034 goto failed_mount;
4035 }
4036 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4037 (sbi->s_cluster_ratio == 1)) {
4038 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4039 "block is 0 with a 1k block and cluster size");
4040 goto failed_mount;
4041 }
4043 blocks_count = (ext4_blocks_count(es) -
4044 le32_to_cpu(es->s_first_data_block) +
4045 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4046 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4047 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4048 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4049 "(block count %llu, first data block %u, "
4050 "blocks per group %lu)", sbi->s_groups_count,
4051 ext4_blocks_count(es),
4052 le32_to_cpu(es->s_first_data_block),
4053 EXT4_BLOCKS_PER_GROUP(sb));
4054 goto failed_mount;
4055 }
4056 sbi->s_groups_count = blocks_count;
4057 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4058 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4059 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4060 EXT4_DESC_PER_BLOCK(sb);
4061 if (ext4_has_feature_meta_bg(sb)) {
4062 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4063 ext4_msg(sb, KERN_WARNING,
4064 "first meta block group too large: %u "
4065 "(group descriptor block count %u)",
4066 le32_to_cpu(es->s_first_meta_bg), db_count);
4067 goto failed_mount;
4068 }
4069 }
4070 sbi->s_group_desc = kvmalloc_array(db_count,
4071 sizeof(struct buffer_head *),
4072 GFP_KERNEL);
4073 if (sbi->s_group_desc == NULL) {
4074 ext4_msg(sb, KERN_ERR, "not enough memory");
4075 ret = -ENOMEM;
4076 goto failed_mount;
4077 }
4078 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4079 le32_to_cpu(es->s_inodes_count)) {
4080 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4081 le32_to_cpu(es->s_inodes_count),
4082 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4083 ret = -EINVAL;
4084 goto failed_mount;
4085 }
4087 bgl_lock_init(sbi->s_blockgroup_lock);
4089 /* Pre-read the descriptors into the buffer cache */
4090 for (i = 0; i < db_count; i++) {
4091 block = descriptor_loc(sb, logical_sb_block, i);
4092 sb_breadahead(sb, block);
4093 }
4095 for (i = 0; i < db_count; i++) {
4096 block = descriptor_loc(sb, logical_sb_block, i);
4097 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
4098 if (!sbi->s_group_desc[i]) {
4099 ext4_msg(sb, KERN_ERR,
4100 "can't read group descriptor %d", i);
4101 db_count = i;
4102 goto failed_mount2;
4103 }
4104 }
4105 sbi->s_gdb_count = db_count;
4106 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4107 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4108 ret = -EFSCORRUPTED;
4109 goto failed_mount2;
4110 }
4112 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4114 /* Register extent status tree shrinker */
4115 if (ext4_es_register_shrinker(sbi))
4116 goto failed_mount3;
4118 sbi->s_stripe = ext4_get_stripe_size(sbi);
4119 sbi->s_extent_max_zeroout_kb = 32;
4121 /*
4122 * set up enough so that it can read an inode
4123 */
4124 sb->s_op = &ext4_sops;
4125 sb->s_export_op = &ext4_export_ops;
4126 sb->s_xattr = ext4_xattr_handlers;
4127 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4128 sb->s_cop = &ext4_cryptops;
4129 #endif
4130 #ifdef CONFIG_QUOTA
4131 sb->dq_op = &ext4_quota_operations;
4132 if (ext4_has_feature_quota(sb))
4133 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4134 else
4135 sb->s_qcop = &ext4_qctl_operations;
4136 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4137 #endif
4138 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4140 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4141 mutex_init(&sbi->s_orphan_lock);
4143 sb->s_root = NULL;
4145 needs_recovery = (es->s_last_orphan != 0 ||
4146 ext4_has_feature_journal_needs_recovery(sb));
4148 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4149 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4150 goto failed_mount3a;
4152 /*
4153 * The first inode we look at is the journal inode. Don't try
4154 * root first: it may be modified in the journal!
4155 */
4156 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4157 err = ext4_load_journal(sb, es, journal_devnum);
4158 if (err)
4159 goto failed_mount3a;
4160 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4161 ext4_has_feature_journal_needs_recovery(sb)) {
4162 ext4_msg(sb, KERN_ERR, "required journal recovery "
4163 "suppressed and not mounted read-only");
4164 goto failed_mount_wq;
4165 } else {
4166 /* Nojournal mode, all journal mount options are illegal */
4167 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4168 ext4_msg(sb, KERN_ERR, "can't mount with "
4169 "journal_checksum, fs mounted w/o journal");
4170 goto failed_mount_wq;
4171 }
4172 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4173 ext4_msg(sb, KERN_ERR, "can't mount with "
4174 "journal_async_commit, fs mounted w/o journal");
4175 goto failed_mount_wq;
4176 }
4177 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4178 ext4_msg(sb, KERN_ERR, "can't mount with "
4179 "commit=%lu, fs mounted w/o journal",
4180 sbi->s_commit_interval / HZ);
4181 goto failed_mount_wq;
4182 }
4183 if (EXT4_MOUNT_DATA_FLAGS &
4184 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4185 ext4_msg(sb, KERN_ERR, "can't mount with "
4186 "data=, fs mounted w/o journal");
4187 goto failed_mount_wq;
4188 }
4189 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4190 clear_opt(sb, JOURNAL_CHECKSUM);
4191 clear_opt(sb, DATA_FLAGS);
4192 sbi->s_journal = NULL;
4193 needs_recovery = 0;
4194 goto no_journal;
4195 }
4197 if (ext4_has_feature_64bit(sb) &&
4198 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4199 JBD2_FEATURE_INCOMPAT_64BIT)) {
4200 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4201 goto failed_mount_wq;
4202 }
4204 if (!set_journal_csum_feature_set(sb)) {
4205 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4206 "feature set");
4207 goto failed_mount_wq;
4208 }
4210 /* We have now updated the journal if required, so we can
4211 * validate the data journaling mode. */
4212 switch (test_opt(sb, DATA_FLAGS)) {
4213 case 0:
4214 /* No mode set, assume a default based on the journal
4215 * capabilities: ORDERED_DATA if the journal can
4216 * cope, else JOURNAL_DATA
4217 */
4218 if (jbd2_journal_check_available_features
4219 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4220 set_opt(sb, ORDERED_DATA);
4221 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4222 } else {
4223 set_opt(sb, JOURNAL_DATA);
4224 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4225 }
4226 break;
4228 case EXT4_MOUNT_ORDERED_DATA:
4229 case EXT4_MOUNT_WRITEBACK_DATA:
4230 if (!jbd2_journal_check_available_features
4231 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4232 ext4_msg(sb, KERN_ERR, "Journal does not support "
4233 "requested data journaling mode");
4234 goto failed_mount_wq;
4235 }
4236 default:
4237 break;
4238 }
4240 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4241 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4242 ext4_msg(sb, KERN_ERR, "can't mount with "
4243 "journal_async_commit in data=ordered mode");
4244 goto failed_mount_wq;
4245 }
4247 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4249 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4251 no_journal:
4252 if (!test_opt(sb, NO_MBCACHE)) {
4253 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4254 if (!sbi->s_ea_block_cache) {
4255 ext4_msg(sb, KERN_ERR,
4256 "Failed to create ea_block_cache");
4257 goto failed_mount_wq;
4258 }
4260 if (ext4_has_feature_ea_inode(sb)) {
4261 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4262 if (!sbi->s_ea_inode_cache) {
4263 ext4_msg(sb, KERN_ERR,
4264 "Failed to create ea_inode_cache");
4265 goto failed_mount_wq;
4266 }
4267 }
4268 }
4270 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4271 (blocksize != PAGE_SIZE)) {
4272 ext4_msg(sb, KERN_ERR,
4273 "Unsupported blocksize for fs encryption");
4274 goto failed_mount_wq;
4275 }
4277 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4278 !ext4_has_feature_encrypt(sb)) {
4279 ext4_set_feature_encrypt(sb);
4280 ext4_commit_super(sb, 1);
4281 }
4283 /*
4284 * Get the # of file system overhead blocks from the
4285 * superblock if present.
4286 */
4287 if (es->s_overhead_clusters)
4288 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4289 else {
4290 err = ext4_calculate_overhead(sb);
4291 if (err)
4292 goto failed_mount_wq;
4293 }
4295 /*
4296 * The maximum number of concurrent works can be high and
4297 * concurrency isn't really necessary. Limit it to 1.
4298 */
4299 EXT4_SB(sb)->rsv_conversion_wq =
4300 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4301 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4302 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4303 ret = -ENOMEM;
4304 goto failed_mount4;
4305 }
4307 /*
4308 * The jbd2_journal_load will have done any necessary log recovery,
4309 * so we can safely mount the rest of the filesystem now.
4310 */
4312 root = ext4_iget(sb, EXT4_ROOT_INO);
4313 if (IS_ERR(root)) {
4314 ext4_msg(sb, KERN_ERR, "get root inode failed");
4315 ret = PTR_ERR(root);
4316 root = NULL;
4317 goto failed_mount4;
4318 }
4319 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4320 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4321 iput(root);
4322 goto failed_mount4;
4323 }
4324 sb->s_root = d_make_root(root);
4325 if (!sb->s_root) {
4326 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4327 ret = -ENOMEM;
4328 goto failed_mount4;
4329 }
4331 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4332 if (ret == -EROFS) {
4333 sb->s_flags |= SB_RDONLY;
4334 ret = 0;
4335 } else if (ret)
4336 goto failed_mount4a;
4338 /* determine the minimum size of new large inodes, if present */
4339 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4340 sbi->s_want_extra_isize == 0) {
4341 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4342 EXT4_GOOD_OLD_INODE_SIZE;
4343 if (ext4_has_feature_extra_isize(sb)) {
4344 if (sbi->s_want_extra_isize <
4345 le16_to_cpu(es->s_want_extra_isize))
4346 sbi->s_want_extra_isize =
4347 le16_to_cpu(es->s_want_extra_isize);
4348 if (sbi->s_want_extra_isize <
4349 le16_to_cpu(es->s_min_extra_isize))
4350 sbi->s_want_extra_isize =
4351 le16_to_cpu(es->s_min_extra_isize);
4352 }
4353 }
4354 /* Check if enough inode space is available */
4355 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4356 sbi->s_inode_size) {
4357 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4358 EXT4_GOOD_OLD_INODE_SIZE;
4359 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4360 "available");
4361 }
4363 ext4_set_resv_clusters(sb);
4365 err = ext4_setup_system_zone(sb);
4366 if (err) {
4367 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4368 "zone (%d)", err);
4369 goto failed_mount4a;
4370 }
4372 ext4_ext_init(sb);
4373 err = ext4_mb_init(sb);
4374 if (err) {
4375 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4376 err);
4377 goto failed_mount5;
4378 }
4380 block = ext4_count_free_clusters(sb);
4381 ext4_free_blocks_count_set(sbi->s_es,
4382 EXT4_C2B(sbi, block));
4383 ext4_superblock_csum_set(sb);
4384 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4385 GFP_KERNEL);
4386 if (!err) {
4387 unsigned long freei = ext4_count_free_inodes(sb);
4388 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4389 ext4_superblock_csum_set(sb);
4390 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4391 GFP_KERNEL);
4392 }
4393 if (!err)
4394 err = percpu_counter_init(&sbi->s_dirs_counter,
4395 ext4_count_dirs(sb), GFP_KERNEL);
4396 if (!err)
4397 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4398 GFP_KERNEL);
4399 if (!err)
4400 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4402 if (err) {
4403 ext4_msg(sb, KERN_ERR, "insufficient memory");
4404 goto failed_mount6;
4405 }
4407 if (ext4_has_feature_flex_bg(sb))
4408 if (!ext4_fill_flex_info(sb)) {
4409 ext4_msg(sb, KERN_ERR,
4410 "unable to initialize "
4411 "flex_bg meta info!");
4412 goto failed_mount6;
4413 }
4415 err = ext4_register_li_request(sb, first_not_zeroed);
4416 if (err)
4417 goto failed_mount6;
4419 err = ext4_register_sysfs(sb);
4420 if (err)
4421 goto failed_mount7;
4423 #ifdef CONFIG_QUOTA
4424 /* Enable quota usage during mount. */
4425 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4426 err = ext4_enable_quotas(sb);
4427 if (err)
4428 goto failed_mount8;
4429 }
4430 #endif /* CONFIG_QUOTA */
4432 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4433 ext4_orphan_cleanup(sb, es);
4434 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4435 if (needs_recovery) {
4436 ext4_msg(sb, KERN_INFO, "recovery complete");
4437 ext4_mark_recovery_complete(sb, es);
4438 }
4439 if (EXT4_SB(sb)->s_journal) {
4440 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4441 descr = " journalled data mode";
4442 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4443 descr = " ordered data mode";
4444 else
4445 descr = " writeback data mode";
4446 } else
4447 descr = "out journal";
4449 if (test_opt(sb, DISCARD)) {
4450 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4451 if (!blk_queue_discard(q))
4452 ext4_msg(sb, KERN_WARNING,
4453 "mounting with \"discard\" option, but "
4454 "the device does not support discard");
4455 }
4457 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4458 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4459 "Opts: %.*s%s%s", descr,
4460 (int) sizeof(sbi->s_es->s_mount_opts),
4461 sbi->s_es->s_mount_opts,
4462 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4464 if (es->s_error_count)
4465 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4467 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4468 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4469 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4470 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4472 kfree(orig_data);
4473 return 0;
4475 cantfind_ext4:
4476 if (!silent)
4477 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4478 goto failed_mount;
4480 #ifdef CONFIG_QUOTA
4481 failed_mount8:
4482 ext4_unregister_sysfs(sb);
4483 #endif
4484 failed_mount7:
4485 ext4_unregister_li_request(sb);
4486 failed_mount6:
4487 ext4_mb_release(sb);
4488 if (sbi->s_flex_groups)
4489 kvfree(sbi->s_flex_groups);
4490 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4491 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4492 percpu_counter_destroy(&sbi->s_dirs_counter);
4493 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4494 failed_mount5:
4495 ext4_ext_release(sb);
4496 ext4_release_system_zone(sb);
4497 failed_mount4a:
4498 dput(sb->s_root);
4499 sb->s_root = NULL;
4500 failed_mount4:
4501 ext4_msg(sb, KERN_ERR, "mount failed");
4502 if (EXT4_SB(sb)->rsv_conversion_wq)
4503 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4504 failed_mount_wq:
4505 if (sbi->s_ea_inode_cache) {
4506 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4507 sbi->s_ea_inode_cache = NULL;
4508 }
4509 if (sbi->s_ea_block_cache) {
4510 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4511 sbi->s_ea_block_cache = NULL;
4512 }
4513 if (sbi->s_journal) {
4514 jbd2_journal_destroy(sbi->s_journal);
4515 sbi->s_journal = NULL;
4516 }
4517 failed_mount3a:
4518 ext4_es_unregister_shrinker(sbi);
4519 failed_mount3:
4520 del_timer_sync(&sbi->s_err_report);
4521 if (sbi->s_mmp_tsk)
4522 kthread_stop(sbi->s_mmp_tsk);
4523 failed_mount2:
4524 for (i = 0; i < db_count; i++)
4525 brelse(sbi->s_group_desc[i]);
4526 kvfree(sbi->s_group_desc);
4527 failed_mount:
4528 if (sbi->s_chksum_driver)
4529 crypto_free_shash(sbi->s_chksum_driver);
4530 #ifdef CONFIG_QUOTA
4531 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4532 kfree(sbi->s_qf_names[i]);
4533 #endif
4534 ext4_blkdev_remove(sbi);
4535 brelse(bh);
4536 out_fail:
4537 sb->s_fs_info = NULL;
4538 kfree(sbi->s_blockgroup_lock);
4539 out_free_base:
4540 kfree(sbi);
4541 kfree(orig_data);
4542 fs_put_dax(dax_dev);
4543 return err ? err : ret;
4544 }
4546 /*
4547 * Setup any per-fs journal parameters now. We'll do this both on
4548 * initial mount, once the journal has been initialised but before we've
4549 * done any recovery; and again on any subsequent remount.
4550 */
4551 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4552 {
4553 struct ext4_sb_info *sbi = EXT4_SB(sb);
4555 journal->j_commit_interval = sbi->s_commit_interval;
4556 journal->j_min_batch_time = sbi->s_min_batch_time;
4557 journal->j_max_batch_time = sbi->s_max_batch_time;
4559 write_lock(&journal->j_state_lock);
4560 if (test_opt(sb, BARRIER))
4561 journal->j_flags |= JBD2_BARRIER;
4562 else
4563 journal->j_flags &= ~JBD2_BARRIER;
4564 if (test_opt(sb, DATA_ERR_ABORT))
4565 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4566 else
4567 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4568 write_unlock(&journal->j_state_lock);
4569 }
4571 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4572 unsigned int journal_inum)
4573 {
4574 struct inode *journal_inode;
4576 /*
4577 * Test for the existence of a valid inode on disk. Bad things
4578 * happen if we iget() an unused inode, as the subsequent iput()
4579 * will try to delete it.
4580 */
4581 journal_inode = ext4_iget(sb, journal_inum);
4582 if (IS_ERR(journal_inode)) {
4583 ext4_msg(sb, KERN_ERR, "no journal found");
4584 return NULL;
4585 }
4586 if (!journal_inode->i_nlink) {
4587 make_bad_inode(journal_inode);
4588 iput(journal_inode);
4589 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4590 return NULL;
4591 }
4593 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4594 journal_inode, journal_inode->i_size);
4595 if (!S_ISREG(journal_inode->i_mode)) {
4596 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4597 iput(journal_inode);
4598 return NULL;
4599 }
4600 return journal_inode;
4601 }
4603 static journal_t *ext4_get_journal(struct super_block *sb,
4604 unsigned int journal_inum)
4605 {
4606 struct inode *journal_inode;
4607 journal_t *journal;
4609 BUG_ON(!ext4_has_feature_journal(sb));
4611 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4612 if (!journal_inode)
4613 return NULL;
4615 journal = jbd2_journal_init_inode(journal_inode);
4616 if (!journal) {
4617 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4618 iput(journal_inode);
4619 return NULL;
4620 }
4621 journal->j_private = sb;
4622 ext4_init_journal_params(sb, journal);
4623 return journal;
4624 }
4626 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4627 dev_t j_dev)
4628 {
4629 struct buffer_head *bh;
4630 journal_t *journal;
4631 ext4_fsblk_t start;
4632 ext4_fsblk_t len;
4633 int hblock, blocksize;
4634 ext4_fsblk_t sb_block;
4635 unsigned long offset;
4636 struct ext4_super_block *es;
4637 struct block_device *bdev;
4639 BUG_ON(!ext4_has_feature_journal(sb));
4641 bdev = ext4_blkdev_get(j_dev, sb);
4642 if (bdev == NULL)
4643 return NULL;
4645 blocksize = sb->s_blocksize;
4646 hblock = bdev_logical_block_size(bdev);
4647 if (blocksize < hblock) {
4648 ext4_msg(sb, KERN_ERR,
4649 "blocksize too small for journal device");
4650 goto out_bdev;
4651 }
4653 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4654 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4655 set_blocksize(bdev, blocksize);
4656 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4657 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4658 "external journal");
4659 goto out_bdev;
4660 }
4662 es = (struct ext4_super_block *) (bh->b_data + offset);
4663 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4664 !(le32_to_cpu(es->s_feature_incompat) &
4665 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4666 ext4_msg(sb, KERN_ERR, "external journal has "
4667 "bad superblock");
4668 brelse(bh);
4669 goto out_bdev;
4670 }
4672 if ((le32_to_cpu(es->s_feature_ro_compat) &
4673 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4674 es->s_checksum != ext4_superblock_csum(sb, es)) {
4675 ext4_msg(sb, KERN_ERR, "external journal has "
4676 "corrupt superblock");
4677 brelse(bh);
4678 goto out_bdev;
4679 }
4681 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4682 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4683 brelse(bh);
4684 goto out_bdev;
4685 }
4687 len = ext4_blocks_count(es);
4688 start = sb_block + 1;
4689 brelse(bh); /* we're done with the superblock */
4691 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4692 start, len, blocksize);
4693 if (!journal) {
4694 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4695 goto out_bdev;
4696 }
4697 journal->j_private = sb;
4698 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4699 wait_on_buffer(journal->j_sb_buffer);
4700 if (!buffer_uptodate(journal->j_sb_buffer)) {
4701 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4702 goto out_journal;
4703 }
4704 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4705 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4706 "user (unsupported) - %d",
4707 be32_to_cpu(journal->j_superblock->s_nr_users));
4708 goto out_journal;
4709 }
4710 EXT4_SB(sb)->journal_bdev = bdev;
4711 ext4_init_journal_params(sb, journal);
4712 return journal;
4714 out_journal:
4715 jbd2_journal_destroy(journal);
4716 out_bdev:
4717 ext4_blkdev_put(bdev);
4718 return NULL;
4719 }
4721 static int ext4_load_journal(struct super_block *sb,
4722 struct ext4_super_block *es,
4723 unsigned long journal_devnum)
4724 {
4725 journal_t *journal;
4726 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4727 dev_t journal_dev;
4728 int err = 0;
4729 int really_read_only;
4731 BUG_ON(!ext4_has_feature_journal(sb));
4733 if (journal_devnum &&
4734 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4735 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4736 "numbers have changed");
4737 journal_dev = new_decode_dev(journal_devnum);
4738 } else
4739 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4741 really_read_only = bdev_read_only(sb->s_bdev);
4743 /*
4744 * Are we loading a blank journal or performing recovery after a
4745 * crash? For recovery, we need to check in advance whether we
4746 * can get read-write access to the device.
4747 */
4748 if (ext4_has_feature_journal_needs_recovery(sb)) {
4749 if (sb_rdonly(sb)) {
4750 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4751 "required on readonly filesystem");
4752 if (really_read_only) {
4753 ext4_msg(sb, KERN_ERR, "write access "
4754 "unavailable, cannot proceed "
4755 "(try mounting with noload)");
4756 return -EROFS;
4757 }
4758 ext4_msg(sb, KERN_INFO, "write access will "
4759 "be enabled during recovery");
4760 }
4761 }
4763 if (journal_inum && journal_dev) {
4764 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4765 "and inode journals!");
4766 return -EINVAL;
4767 }
4769 if (journal_inum) {
4770 if (!(journal = ext4_get_journal(sb, journal_inum)))
4771 return -EINVAL;
4772 } else {
4773 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4774 return -EINVAL;
4775 }
4777 if (!(journal->j_flags & JBD2_BARRIER))
4778 ext4_msg(sb, KERN_INFO, "barriers disabled");
4780 if (!ext4_has_feature_journal_needs_recovery(sb))
4781 err = jbd2_journal_wipe(journal, !really_read_only);
4782 if (!err) {
4783 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4784 if (save)
4785 memcpy(save, ((char *) es) +
4786 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4787 err = jbd2_journal_load(journal);
4788 if (save)
4789 memcpy(((char *) es) + EXT4_S_ERR_START,
4790 save, EXT4_S_ERR_LEN);
4791 kfree(save);
4792 }
4794 if (err) {
4795 ext4_msg(sb, KERN_ERR, "error loading journal");
4796 jbd2_journal_destroy(journal);
4797 return err;
4798 }
4800 EXT4_SB(sb)->s_journal = journal;
4801 ext4_clear_journal_err(sb, es);
4803 if (!really_read_only && journal_devnum &&
4804 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4805 es->s_journal_dev = cpu_to_le32(journal_devnum);
4807 /* Make sure we flush the recovery flag to disk. */
4808 ext4_commit_super(sb, 1);
4809 }
4811 return 0;
4812 }
4814 static int ext4_commit_super(struct super_block *sb, int sync)
4815 {
4816 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4817 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4818 int error = 0;
4820 if (!sbh || block_device_ejected(sb))
4821 return error;
4823 /*
4824 * The superblock bh should be mapped, but it might not be if the
4825 * device was hot-removed. Not much we can do but fail the I/O.
4826 */
4827 if (!buffer_mapped(sbh))
4828 return error;
4830 /*
4831 * If the file system is mounted read-only, don't update the
4832 * superblock write time. This avoids updating the superblock
4833 * write time when we are mounting the root file system
4834 * read/only but we need to replay the journal; at that point,
4835 * for people who are east of GMT and who make their clock
4836 * tick in localtime for Windows bug-for-bug compatibility,
4837 * the clock is set in the future, and this will cause e2fsck
4838 * to complain and force a full file system check.
4839 */
4840 if (!(sb->s_flags & SB_RDONLY))
4841 ext4_update_tstamp(es, s_wtime);
4842 if (sb->s_bdev->bd_part)
4843 es->s_kbytes_written =
4844 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4845 ((part_stat_read(sb->s_bdev->bd_part,
4846 sectors[STAT_WRITE]) -
4847 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4848 else
4849 es->s_kbytes_written =
4850 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4851 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4852 ext4_free_blocks_count_set(es,
4853 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4854 &EXT4_SB(sb)->s_freeclusters_counter)));
4855 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4856 es->s_free_inodes_count =
4857 cpu_to_le32(percpu_counter_sum_positive(
4858 &EXT4_SB(sb)->s_freeinodes_counter));
4859 BUFFER_TRACE(sbh, "marking dirty");
4860 ext4_superblock_csum_set(sb);
4861 if (sync)
4862 lock_buffer(sbh);
4863 if (buffer_write_io_error(sbh)) {
4864 /*
4865 * Oh, dear. A previous attempt to write the
4866 * superblock failed. This could happen because the
4867 * USB device was yanked out. Or it could happen to
4868 * be a transient write error and maybe the block will
4869 * be remapped. Nothing we can do but to retry the
4870 * write and hope for the best.
4871 */
4872 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4873 "superblock detected");
4874 clear_buffer_write_io_error(sbh);
4875 set_buffer_uptodate(sbh);
4876 }
4877 mark_buffer_dirty(sbh);
4878 if (sync) {
4879 unlock_buffer(sbh);
4880 error = __sync_dirty_buffer(sbh,
4881 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
4882 if (buffer_write_io_error(sbh)) {
4883 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4884 "superblock");
4885 clear_buffer_write_io_error(sbh);
4886 set_buffer_uptodate(sbh);
4887 }
4888 }
4889 return error;
4890 }
4892 /*
4893 * Have we just finished recovery? If so, and if we are mounting (or
4894 * remounting) the filesystem readonly, then we will end up with a
4895 * consistent fs on disk. Record that fact.
4896 */
4897 static void ext4_mark_recovery_complete(struct super_block *sb,
4898 struct ext4_super_block *es)
4899 {
4900 journal_t *journal = EXT4_SB(sb)->s_journal;
4902 if (!ext4_has_feature_journal(sb)) {
4903 BUG_ON(journal != NULL);
4904 return;
4905 }
4906 jbd2_journal_lock_updates(journal);
4907 if (jbd2_journal_flush(journal) < 0)
4908 goto out;
4910 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
4911 ext4_clear_feature_journal_needs_recovery(sb);
4912 ext4_commit_super(sb, 1);
4913 }
4915 out:
4916 jbd2_journal_unlock_updates(journal);
4917 }
4919 /*
4920 * If we are mounting (or read-write remounting) a filesystem whose journal
4921 * has recorded an error from a previous lifetime, move that error to the
4922 * main filesystem now.
4923 */
4924 static void ext4_clear_journal_err(struct super_block *sb,
4925 struct ext4_super_block *es)
4926 {
4927 journal_t *journal;
4928 int j_errno;
4929 const char *errstr;
4931 BUG_ON(!ext4_has_feature_journal(sb));
4933 journal = EXT4_SB(sb)->s_journal;
4935 /*
4936 * Now check for any error status which may have been recorded in the
4937 * journal by a prior ext4_error() or ext4_abort()
4938 */
4940 j_errno = jbd2_journal_errno(journal);
4941 if (j_errno) {
4942 char nbuf[16];
4944 errstr = ext4_decode_error(sb, j_errno, nbuf);
4945 ext4_warning(sb, "Filesystem error recorded "
4946 "from previous mount: %s", errstr);
4947 ext4_warning(sb, "Marking fs in need of filesystem check.");
4949 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4950 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4951 ext4_commit_super(sb, 1);
4953 jbd2_journal_clear_err(journal);
4954 jbd2_journal_update_sb_errno(journal);
4955 }
4956 }
4958 /*
4959 * Force the running and committing transactions to commit,
4960 * and wait on the commit.
4961 */
4962 int ext4_force_commit(struct super_block *sb)
4963 {
4964 journal_t *journal;
4966 if (sb_rdonly(sb))
4967 return 0;
4969 journal = EXT4_SB(sb)->s_journal;
4970 return ext4_journal_force_commit(journal);
4971 }
4973 static int ext4_sync_fs(struct super_block *sb, int wait)
4974 {
4975 int ret = 0;
4976 tid_t target;
4977 bool needs_barrier = false;
4978 struct ext4_sb_info *sbi = EXT4_SB(sb);
4980 if (unlikely(ext4_forced_shutdown(sbi)))
4981 return 0;
4983 trace_ext4_sync_fs(sb, wait);
4984 flush_workqueue(sbi->rsv_conversion_wq);
4985 /*
4986 * Writeback quota in non-journalled quota case - journalled quota has
4987 * no dirty dquots
4988 */
4989 dquot_writeback_dquots(sb, -1);
4990 /*
4991 * Data writeback is possible w/o journal transaction, so barrier must
4992 * being sent at the end of the function. But we can skip it if
4993 * transaction_commit will do it for us.
4994 */
4995 if (sbi->s_journal) {
4996 target = jbd2_get_latest_transaction(sbi->s_journal);
4997 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4998 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4999 needs_barrier = true;
5001 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5002 if (wait)
5003 ret = jbd2_log_wait_commit(sbi->s_journal,
5004 target);
5005 }
5006 } else if (wait && test_opt(sb, BARRIER))
5007 needs_barrier = true;
5008 if (needs_barrier) {
5009 int err;
5010 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5011 if (!ret)
5012 ret = err;
5013 }
5015 return ret;
5016 }
5018 /*
5019 * LVM calls this function before a (read-only) snapshot is created. This
5020 * gives us a chance to flush the journal completely and mark the fs clean.
5021 *
5022 * Note that only this function cannot bring a filesystem to be in a clean
5023 * state independently. It relies on upper layer to stop all data & metadata
5024 * modifications.
5025 */
5026 static int ext4_freeze(struct super_block *sb)
5027 {
5028 int error = 0;
5029 journal_t *journal;
5031 if (sb_rdonly(sb))
5032 return 0;
5034 journal = EXT4_SB(sb)->s_journal;
5036 if (journal) {
5037 /* Now we set up the journal barrier. */
5038 jbd2_journal_lock_updates(journal);
5040 /*
5041 * Don't clear the needs_recovery flag if we failed to
5042 * flush the journal.
5043 */
5044 error = jbd2_journal_flush(journal);
5045 if (error < 0)
5046 goto out;
5048 /* Journal blocked and flushed, clear needs_recovery flag. */
5049 ext4_clear_feature_journal_needs_recovery(sb);
5050 }
5052 error = ext4_commit_super(sb, 1);
5053 out:
5054 if (journal)
5055 /* we rely on upper layer to stop further updates */
5056 jbd2_journal_unlock_updates(journal);
5057 return error;
5058 }
5060 /*
5061 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5062 * flag here, even though the filesystem is not technically dirty yet.
5063 */
5064 static int ext4_unfreeze(struct super_block *sb)
5065 {
5066 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5067 return 0;
5069 if (EXT4_SB(sb)->s_journal) {
5070 /* Reset the needs_recovery flag before the fs is unlocked. */
5071 ext4_set_feature_journal_needs_recovery(sb);
5072 }
5074 ext4_commit_super(sb, 1);
5075 return 0;
5076 }
5078 /*
5079 * Structure to save mount options for ext4_remount's benefit
5080 */
5081 struct ext4_mount_options {
5082 unsigned long s_mount_opt;
5083 unsigned long s_mount_opt2;
5084 kuid_t s_resuid;
5085 kgid_t s_resgid;
5086 unsigned long s_commit_interval;
5087 u32 s_min_batch_time, s_max_batch_time;
5088 #ifdef CONFIG_QUOTA
5089 int s_jquota_fmt;
5090 char *s_qf_names[EXT4_MAXQUOTAS];
5091 #endif
5092 };
5094 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5095 {
5096 struct ext4_super_block *es;
5097 struct ext4_sb_info *sbi = EXT4_SB(sb);
5098 unsigned long old_sb_flags;
5099 struct ext4_mount_options old_opts;
5100 int enable_quota = 0;
5101 ext4_group_t g;
5102 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5103 int err = 0;
5104 #ifdef CONFIG_QUOTA
5105 int i, j;
5106 #endif
5107 char *orig_data = kstrdup(data, GFP_KERNEL);
5109 if (data && !orig_data)
5110 return -ENOMEM;
5112 /* Store the original options */
5113 old_sb_flags = sb->s_flags;
5114 old_opts.s_mount_opt = sbi->s_mount_opt;
5115 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5116 old_opts.s_resuid = sbi->s_resuid;
5117 old_opts.s_resgid = sbi->s_resgid;
5118 old_opts.s_commit_interval = sbi->s_commit_interval;
5119 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5120 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5121 #ifdef CONFIG_QUOTA
5122 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5123 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5124 if (sbi->s_qf_names[i]) {
5125 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
5126 GFP_KERNEL);
5127 if (!old_opts.s_qf_names[i]) {
5128 for (j = 0; j < i; j++)
5129 kfree(old_opts.s_qf_names[j]);
5130 kfree(orig_data);
5131 return -ENOMEM;
5132 }
5133 } else
5134 old_opts.s_qf_names[i] = NULL;
5135 #endif
5136 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5137 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5139 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5140 err = -EINVAL;
5141 goto restore_opts;
5142 }
5144 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5145 test_opt(sb, JOURNAL_CHECKSUM)) {
5146 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5147 "during remount not supported; ignoring");
5148 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5149 }
5151 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5152 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5153 ext4_msg(sb, KERN_ERR, "can't mount with "
5154 "both data=journal and delalloc");
5155 err = -EINVAL;
5156 goto restore_opts;
5157 }
5158 if (test_opt(sb, DIOREAD_NOLOCK)) {
5159 ext4_msg(sb, KERN_ERR, "can't mount with "
5160 "both data=journal and dioread_nolock");
5161 err = -EINVAL;
5162 goto restore_opts;
5163 }
5164 if (test_opt(sb, DAX)) {
5165 ext4_msg(sb, KERN_ERR, "can't mount with "
5166 "both data=journal and dax");
5167 err = -EINVAL;
5168 goto restore_opts;
5169 }
5170 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5171 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5172 ext4_msg(sb, KERN_ERR, "can't mount with "
5173 "journal_async_commit in data=ordered mode");
5174 err = -EINVAL;
5175 goto restore_opts;
5176 }
5177 }
5179 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5180 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5181 err = -EINVAL;
5182 goto restore_opts;
5183 }
5185 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5186 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5187 "dax flag with busy inodes while remounting");
5188 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5189 }
5191 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5192 ext4_abort(sb, "Abort forced by user");
5194 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5195 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5197 es = sbi->s_es;
5199 if (sbi->s_journal) {
5200 ext4_init_journal_params(sb, sbi->s_journal);
5201 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5202 }
5204 if (*flags & SB_LAZYTIME)
5205 sb->s_flags |= SB_LAZYTIME;
5207 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5208 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5209 err = -EROFS;
5210 goto restore_opts;
5211 }
5213 if (*flags & SB_RDONLY) {
5214 err = sync_filesystem(sb);
5215 if (err < 0)
5216 goto restore_opts;
5217 err = dquot_suspend(sb, -1);
5218 if (err < 0)
5219 goto restore_opts;
5221 /*
5222 * First of all, the unconditional stuff we have to do
5223 * to disable replay of the journal when we next remount
5224 */
5225 sb->s_flags |= SB_RDONLY;
5227 /*
5228 * OK, test if we are remounting a valid rw partition
5229 * readonly, and if so set the rdonly flag and then
5230 * mark the partition as valid again.
5231 */
5232 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5233 (sbi->s_mount_state & EXT4_VALID_FS))
5234 es->s_state = cpu_to_le16(sbi->s_mount_state);
5236 if (sbi->s_journal)
5237 ext4_mark_recovery_complete(sb, es);
5238 if (sbi->s_mmp_tsk)
5239 kthread_stop(sbi->s_mmp_tsk);
5240 } else {
5241 /* Make sure we can mount this feature set readwrite */
5242 if (ext4_has_feature_readonly(sb) ||
5243 !ext4_feature_set_ok(sb, 0)) {
5244 err = -EROFS;
5245 goto restore_opts;
5246 }
5247 /*
5248 * Make sure the group descriptor checksums
5249 * are sane. If they aren't, refuse to remount r/w.
5250 */
5251 for (g = 0; g < sbi->s_groups_count; g++) {
5252 struct ext4_group_desc *gdp =
5253 ext4_get_group_desc(sb, g, NULL);
5255 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5256 ext4_msg(sb, KERN_ERR,
5257 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5258 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5259 le16_to_cpu(gdp->bg_checksum));
5260 err = -EFSBADCRC;
5261 goto restore_opts;
5262 }
5263 }
5265 /*
5266 * If we have an unprocessed orphan list hanging
5267 * around from a previously readonly bdev mount,
5268 * require a full umount/remount for now.
5269 */
5270 if (es->s_last_orphan) {
5271 ext4_msg(sb, KERN_WARNING, "Couldn't "
5272 "remount RDWR because of unprocessed "
5273 "orphan inode list. Please "
5274 "umount/remount instead");
5275 err = -EINVAL;
5276 goto restore_opts;
5277 }
5279 /*
5280 * Mounting a RDONLY partition read-write, so reread
5281 * and store the current valid flag. (It may have
5282 * been changed by e2fsck since we originally mounted
5283 * the partition.)
5284 */
5285 if (sbi->s_journal)
5286 ext4_clear_journal_err(sb, es);
5287 sbi->s_mount_state = le16_to_cpu(es->s_state);
5289 err = ext4_setup_super(sb, es, 0);
5290 if (err)
5291 goto restore_opts;
5293 sb->s_flags &= ~SB_RDONLY;
5294 if (ext4_has_feature_mmp(sb))
5295 if (ext4_multi_mount_protect(sb,
5296 le64_to_cpu(es->s_mmp_block))) {
5297 err = -EROFS;
5298 goto restore_opts;
5299 }
5300 enable_quota = 1;
5301 }
5302 }
5304 /*
5305 * Reinitialize lazy itable initialization thread based on
5306 * current settings
5307 */
5308 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5309 ext4_unregister_li_request(sb);
5310 else {
5311 ext4_group_t first_not_zeroed;
5312 first_not_zeroed = ext4_has_uninit_itable(sb);
5313 ext4_register_li_request(sb, first_not_zeroed);
5314 }
5316 ext4_setup_system_zone(sb);
5317 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5318 err = ext4_commit_super(sb, 1);
5319 if (err)
5320 goto restore_opts;
5321 }
5323 #ifdef CONFIG_QUOTA
5324 /* Release old quota file names */
5325 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5326 kfree(old_opts.s_qf_names[i]);
5327 if (enable_quota) {
5328 if (sb_any_quota_suspended(sb))
5329 dquot_resume(sb, -1);
5330 else if (ext4_has_feature_quota(sb)) {
5331 err = ext4_enable_quotas(sb);
5332 if (err)
5333 goto restore_opts;
5334 }
5335 }
5336 #endif
5338 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
5339 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5340 kfree(orig_data);
5341 return 0;
5343 restore_opts:
5344 sb->s_flags = old_sb_flags;
5345 sbi->s_mount_opt = old_opts.s_mount_opt;
5346 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5347 sbi->s_resuid = old_opts.s_resuid;
5348 sbi->s_resgid = old_opts.s_resgid;
5349 sbi->s_commit_interval = old_opts.s_commit_interval;
5350 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5351 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5352 #ifdef CONFIG_QUOTA
5353 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5354 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5355 kfree(sbi->s_qf_names[i]);
5356 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5357 }
5358 #endif
5359 kfree(orig_data);
5360 return err;
5361 }
5363 #ifdef CONFIG_QUOTA
5364 static int ext4_statfs_project(struct super_block *sb,
5365 kprojid_t projid, struct kstatfs *buf)
5366 {
5367 struct kqid qid;
5368 struct dquot *dquot;
5369 u64 limit;
5370 u64 curblock;
5372 qid = make_kqid_projid(projid);
5373 dquot = dqget(sb, qid);
5374 if (IS_ERR(dquot))
5375 return PTR_ERR(dquot);
5376 spin_lock(&dquot->dq_dqb_lock);
5378 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5379 dquot->dq_dqb.dqb_bsoftlimit :
5380 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5381 if (limit && buf->f_blocks > limit) {
5382 curblock = (dquot->dq_dqb.dqb_curspace +
5383 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
5384 buf->f_blocks = limit;
5385 buf->f_bfree = buf->f_bavail =
5386 (buf->f_blocks > curblock) ?
5387 (buf->f_blocks - curblock) : 0;
5388 }
5390 limit = dquot->dq_dqb.dqb_isoftlimit ?
5391 dquot->dq_dqb.dqb_isoftlimit :
5392 dquot->dq_dqb.dqb_ihardlimit;
5393 if (limit && buf->f_files > limit) {
5394 buf->f_files = limit;
5395 buf->f_ffree =
5396 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5397 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5398 }
5400 spin_unlock(&dquot->dq_dqb_lock);
5401 dqput(dquot);
5402 return 0;
5403 }
5404 #endif
5406 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5407 {
5408 struct super_block *sb = dentry->d_sb;
5409 struct ext4_sb_info *sbi = EXT4_SB(sb);
5410 struct ext4_super_block *es = sbi->s_es;
5411 ext4_fsblk_t overhead = 0, resv_blocks;
5412 u64 fsid;
5413 s64 bfree;
5414 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5416 if (!test_opt(sb, MINIX_DF))
5417 overhead = sbi->s_overhead;
5419 buf->f_type = EXT4_SUPER_MAGIC;
5420 buf->f_bsize = sb->s_blocksize;
5421 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5422 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5423 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5424 /* prevent underflow in case that few free space is available */
5425 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5426 buf->f_bavail = buf->f_bfree -
5427 (ext4_r_blocks_count(es) + resv_blocks);
5428 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5429 buf->f_bavail = 0;
5430 buf->f_files = le32_to_cpu(es->s_inodes_count);
5431 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5432 buf->f_namelen = EXT4_NAME_LEN;
5433 fsid = le64_to_cpup((void *)es->s_uuid) ^
5434 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5435 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5436 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5438 #ifdef CONFIG_QUOTA
5439 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5440 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5441 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5442 #endif
5443 return 0;
5444 }
5447 #ifdef CONFIG_QUOTA
5449 /*
5450 * Helper functions so that transaction is started before we acquire dqio_sem
5451 * to keep correct lock ordering of transaction > dqio_sem
5452 */
5453 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5454 {
5455 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5456 }
5458 static int ext4_write_dquot(struct dquot *dquot)
5459 {
5460 int ret, err;
5461 handle_t *handle;
5462 struct inode *inode;
5464 inode = dquot_to_inode(dquot);
5465 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5466 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5467 if (IS_ERR(handle))
5468 return PTR_ERR(handle);
5469 ret = dquot_commit(dquot);
5470 err = ext4_journal_stop(handle);
5471 if (!ret)
5472 ret = err;
5473 return ret;
5474 }
5476 static int ext4_acquire_dquot(struct dquot *dquot)
5477 {
5478 int ret, err;
5479 handle_t *handle;
5481 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5482 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5483 if (IS_ERR(handle))
5484 return PTR_ERR(handle);
5485 ret = dquot_acquire(dquot);
5486 err = ext4_journal_stop(handle);
5487 if (!ret)
5488 ret = err;
5489 return ret;
5490 }
5492 static int ext4_release_dquot(struct dquot *dquot)
5493 {
5494 int ret, err;
5495 handle_t *handle;
5497 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5498 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5499 if (IS_ERR(handle)) {
5500 /* Release dquot anyway to avoid endless cycle in dqput() */
5501 dquot_release(dquot);
5502 return PTR_ERR(handle);
5503 }
5504 ret = dquot_release(dquot);
5505 err = ext4_journal_stop(handle);
5506 if (!ret)
5507 ret = err;
5508 return ret;
5509 }
5511 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5512 {
5513 struct super_block *sb = dquot->dq_sb;
5514 struct ext4_sb_info *sbi = EXT4_SB(sb);
5516 /* Are we journaling quotas? */
5517 if (ext4_has_feature_quota(sb) ||
5518 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5519 dquot_mark_dquot_dirty(dquot);
5520 return ext4_write_dquot(dquot);
5521 } else {
5522 return dquot_mark_dquot_dirty(dquot);
5523 }
5524 }
5526 static int ext4_write_info(struct super_block *sb, int type)
5527 {
5528 int ret, err;
5529 handle_t *handle;
5531 /* Data block + inode block */
5532 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5533 if (IS_ERR(handle))
5534 return PTR_ERR(handle);
5535 ret = dquot_commit_info(sb, type);
5536 err = ext4_journal_stop(handle);
5537 if (!ret)
5538 ret = err;
5539 return ret;
5540 }
5542 /*
5543 * Turn on quotas during mount time - we need to find
5544 * the quota file and such...
5545 */
5546 static int ext4_quota_on_mount(struct super_block *sb, int type)
5547 {
5548 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5549 EXT4_SB(sb)->s_jquota_fmt, type);
5550 }
5552 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5553 {
5554 struct ext4_inode_info *ei = EXT4_I(inode);
5556 /* The first argument of lockdep_set_subclass has to be
5557 * *exactly* the same as the argument to init_rwsem() --- in
5558 * this case, in init_once() --- or lockdep gets unhappy
5559 * because the name of the lock is set using the
5560 * stringification of the argument to init_rwsem().
5561 */
5562 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5563 lockdep_set_subclass(&ei->i_data_sem, subclass);
5564 }
5566 /*
5567 * Standard function to be called on quota_on
5568 */
5569 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5570 const struct path *path)
5571 {
5572 int err;
5574 if (!test_opt(sb, QUOTA))
5575 return -EINVAL;
5577 /* Quotafile not on the same filesystem? */
5578 if (path->dentry->d_sb != sb)
5579 return -EXDEV;
5580 /* Journaling quota? */
5581 if (EXT4_SB(sb)->s_qf_names[type]) {
5582 /* Quotafile not in fs root? */
5583 if (path->dentry->d_parent != sb->s_root)
5584 ext4_msg(sb, KERN_WARNING,
5585 "Quota file not on filesystem root. "
5586 "Journaled quota will not work");
5587 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5588 } else {
5589 /*
5590 * Clear the flag just in case mount options changed since
5591 * last time.
5592 */
5593 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5594 }
5596 /*
5597 * When we journal data on quota file, we have to flush journal to see
5598 * all updates to the file when we bypass pagecache...
5599 */
5600 if (EXT4_SB(sb)->s_journal &&
5601 ext4_should_journal_data(d_inode(path->dentry))) {
5602 /*
5603 * We don't need to lock updates but journal_flush() could
5604 * otherwise be livelocked...
5605 */
5606 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5607 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5608 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5609 if (err)
5610 return err;
5611 }
5613 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5614 err = dquot_quota_on(sb, type, format_id, path);
5615 if (err) {
5616 lockdep_set_quota_inode(path->dentry->d_inode,
5617 I_DATA_SEM_NORMAL);
5618 } else {
5619 struct inode *inode = d_inode(path->dentry);
5620 handle_t *handle;
5622 /*
5623 * Set inode flags to prevent userspace from messing with quota
5624 * files. If this fails, we return success anyway since quotas
5625 * are already enabled and this is not a hard failure.
5626 */
5627 inode_lock(inode);
5628 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5629 if (IS_ERR(handle))
5630 goto unlock_inode;
5631 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5632 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5633 S_NOATIME | S_IMMUTABLE);
5634 ext4_mark_inode_dirty(handle, inode);
5635 ext4_journal_stop(handle);
5636 unlock_inode:
5637 inode_unlock(inode);
5638 }
5639 return err;
5640 }
5642 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5643 unsigned int flags)
5644 {
5645 int err;
5646 struct inode *qf_inode;
5647 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5648 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5649 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5650 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5651 };
5653 BUG_ON(!ext4_has_feature_quota(sb));
5655 if (!qf_inums[type])
5656 return -EPERM;
5658 qf_inode = ext4_iget(sb, qf_inums[type]);
5659 if (IS_ERR(qf_inode)) {
5660 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5661 return PTR_ERR(qf_inode);
5662 }
5664 /* Don't account quota for quota files to avoid recursion */
5665 qf_inode->i_flags |= S_NOQUOTA;
5666 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5667 err = dquot_enable(qf_inode, type, format_id, flags);
5668 iput(qf_inode);
5669 if (err)
5670 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5672 return err;
5673 }
5675 /* Enable usage tracking for all quota types. */
5676 static int ext4_enable_quotas(struct super_block *sb)
5677 {
5678 int type, err = 0;
5679 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5680 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5681 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5682 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5683 };
5684 bool quota_mopt[EXT4_MAXQUOTAS] = {
5685 test_opt(sb, USRQUOTA),
5686 test_opt(sb, GRPQUOTA),
5687 test_opt(sb, PRJQUOTA),
5688 };
5690 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5691 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5692 if (qf_inums[type]) {
5693 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5694 DQUOT_USAGE_ENABLED |
5695 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5696 if (err) {
5697 ext4_warning(sb,
5698 "Failed to enable quota tracking "
5699 "(type=%d, err=%d). Please run "
5700 "e2fsck to fix.", type, err);
5701 for (type--; type >= 0; type--)
5702 dquot_quota_off(sb, type);
5704 return err;
5705 }
5706 }
5707 }
5708 return 0;
5709 }
5711 static int ext4_quota_off(struct super_block *sb, int type)
5712 {
5713 struct inode *inode = sb_dqopt(sb)->files[type];
5714 handle_t *handle;
5715 int err;
5717 /* Force all delayed allocation blocks to be allocated.
5718 * Caller already holds s_umount sem */
5719 if (test_opt(sb, DELALLOC))
5720 sync_filesystem(sb);
5722 if (!inode || !igrab(inode))
5723 goto out;
5725 err = dquot_quota_off(sb, type);
5726 if (err || ext4_has_feature_quota(sb))
5727 goto out_put;
5729 inode_lock(inode);
5730 /*
5731 * Update modification times of quota files when userspace can
5732 * start looking at them. If we fail, we return success anyway since
5733 * this is not a hard failure and quotas are already disabled.
5734 */
5735 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5736 if (IS_ERR(handle))
5737 goto out_unlock;
5738 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5739 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5740 inode->i_mtime = inode->i_ctime = current_time(inode);
5741 ext4_mark_inode_dirty(handle, inode);
5742 ext4_journal_stop(handle);
5743 out_unlock:
5744 inode_unlock(inode);
5745 out_put:
5746 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5747 iput(inode);
5748 return err;
5749 out:
5750 return dquot_quota_off(sb, type);
5751 }
5753 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5754 * acquiring the locks... As quota files are never truncated and quota code
5755 * itself serializes the operations (and no one else should touch the files)
5756 * we don't have to be afraid of races */
5757 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5758 size_t len, loff_t off)
5759 {
5760 struct inode *inode = sb_dqopt(sb)->files[type];
5761 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5762 int offset = off & (sb->s_blocksize - 1);
5763 int tocopy;
5764 size_t toread;
5765 struct buffer_head *bh;
5766 loff_t i_size = i_size_read(inode);
5768 if (off > i_size)
5769 return 0;
5770 if (off+len > i_size)
5771 len = i_size-off;
5772 toread = len;
5773 while (toread > 0) {
5774 tocopy = sb->s_blocksize - offset < toread ?
5775 sb->s_blocksize - offset : toread;
5776 bh = ext4_bread(NULL, inode, blk, 0);
5777 if (IS_ERR(bh))
5778 return PTR_ERR(bh);
5779 if (!bh) /* A hole? */
5780 memset(data, 0, tocopy);
5781 else
5782 memcpy(data, bh->b_data+offset, tocopy);
5783 brelse(bh);
5784 offset = 0;
5785 toread -= tocopy;
5786 data += tocopy;
5787 blk++;
5788 }
5789 return len;
5790 }
5792 /* Write to quotafile (we know the transaction is already started and has
5793 * enough credits) */
5794 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5795 const char *data, size_t len, loff_t off)
5796 {
5797 struct inode *inode = sb_dqopt(sb)->files[type];
5798 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5799 int err, offset = off & (sb->s_blocksize - 1);
5800 int retries = 0;
5801 struct buffer_head *bh;
5802 handle_t *handle = journal_current_handle();
5804 if (EXT4_SB(sb)->s_journal && !handle) {
5805 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5806 " cancelled because transaction is not started",
5807 (unsigned long long)off, (unsigned long long)len);
5808 return -EIO;
5809 }
5810 /*
5811 * Since we account only one data block in transaction credits,
5812 * then it is impossible to cross a block boundary.
5813 */
5814 if (sb->s_blocksize - offset < len) {
5815 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5816 " cancelled because not block aligned",
5817 (unsigned long long)off, (unsigned long long)len);
5818 return -EIO;
5819 }
5821 do {
5822 bh = ext4_bread(handle, inode, blk,
5823 EXT4_GET_BLOCKS_CREATE |
5824 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5825 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5826 ext4_should_retry_alloc(inode->i_sb, &retries));
5827 if (IS_ERR(bh))
5828 return PTR_ERR(bh);
5829 if (!bh)
5830 goto out;
5831 BUFFER_TRACE(bh, "get write access");
5832 err = ext4_journal_get_write_access(handle, bh);
5833 if (err) {
5834 brelse(bh);
5835 return err;
5836 }
5837 lock_buffer(bh);
5838 memcpy(bh->b_data+offset, data, len);
5839 flush_dcache_page(bh->b_page);
5840 unlock_buffer(bh);
5841 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5842 brelse(bh);
5843 out:
5844 if (inode->i_size < off + len) {
5845 i_size_write(inode, off + len);
5846 EXT4_I(inode)->i_disksize = inode->i_size;
5847 ext4_mark_inode_dirty(handle, inode);
5848 }
5849 return len;
5850 }
5852 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5853 {
5854 const struct quota_format_ops *ops;
5856 if (!sb_has_quota_loaded(sb, qid->type))
5857 return -ESRCH;
5858 ops = sb_dqopt(sb)->ops[qid->type];
5859 if (!ops || !ops->get_next_id)
5860 return -ENOSYS;
5861 return dquot_get_next_id(sb, qid);
5862 }
5863 #endif
5865 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5866 const char *dev_name, void *data)
5867 {
5868 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5869 }
5871 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5872 static inline void register_as_ext2(void)
5873 {
5874 int err = register_filesystem(&ext2_fs_type);
5875 if (err)
5876 printk(KERN_WARNING
5877 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5878 }
5880 static inline void unregister_as_ext2(void)
5881 {
5882 unregister_filesystem(&ext2_fs_type);
5883 }
5885 static inline int ext2_feature_set_ok(struct super_block *sb)
5886 {
5887 if (ext4_has_unknown_ext2_incompat_features(sb))
5888 return 0;
5889 if (sb_rdonly(sb))
5890 return 1;
5891 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5892 return 0;
5893 return 1;
5894 }
5895 #else
5896 static inline void register_as_ext2(void) { }
5897 static inline void unregister_as_ext2(void) { }
5898 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5899 #endif
5901 static inline void register_as_ext3(void)
5902 {
5903 int err = register_filesystem(&ext3_fs_type);
5904 if (err)
5905 printk(KERN_WARNING
5906 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5907 }
5909 static inline void unregister_as_ext3(void)
5910 {
5911 unregister_filesystem(&ext3_fs_type);
5912 }
5914 static inline int ext3_feature_set_ok(struct super_block *sb)
5915 {
5916 if (ext4_has_unknown_ext3_incompat_features(sb))
5917 return 0;
5918 if (!ext4_has_feature_journal(sb))
5919 return 0;
5920 if (sb_rdonly(sb))
5921 return 1;
5922 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5923 return 0;
5924 return 1;
5925 }
5927 static struct file_system_type ext4_fs_type = {
5928 .owner = THIS_MODULE,
5929 .name = "ext4",
5930 .mount = ext4_mount,
5931 .kill_sb = kill_block_super,
5932 .fs_flags = FS_REQUIRES_DEV,
5933 };
5934 MODULE_ALIAS_FS("ext4");
5936 /* Shared across all ext4 file systems */
5937 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5939 static int __init ext4_init_fs(void)
5940 {
5941 int i, err;
5943 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5944 ext4_li_info = NULL;
5945 mutex_init(&ext4_li_mtx);
5947 /* Build-time check for flags consistency */
5948 ext4_check_flag_values();
5950 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5951 init_waitqueue_head(&ext4__ioend_wq[i]);
5953 err = ext4_init_es();
5954 if (err)
5955 return err;
5957 err = ext4_init_pageio();
5958 if (err)
5959 goto out5;
5961 err = ext4_init_system_zone();
5962 if (err)
5963 goto out4;
5965 err = ext4_init_sysfs();
5966 if (err)
5967 goto out3;
5969 err = ext4_init_mballoc();
5970 if (err)
5971 goto out2;
5972 err = init_inodecache();
5973 if (err)
5974 goto out1;
5975 register_as_ext3();
5976 register_as_ext2();
5977 err = register_filesystem(&ext4_fs_type);
5978 if (err)
5979 goto out;
5981 return 0;
5982 out:
5983 unregister_as_ext2();
5984 unregister_as_ext3();
5985 destroy_inodecache();
5986 out1:
5987 ext4_exit_mballoc();
5988 out2:
5989 ext4_exit_sysfs();
5990 out3:
5991 ext4_exit_system_zone();
5992 out4:
5993 ext4_exit_pageio();
5994 out5:
5995 ext4_exit_es();
5997 return err;
5998 }
6000 static void __exit ext4_exit_fs(void)
6001 {
6002 ext4_destroy_lazyinit_thread();
6003 unregister_as_ext2();
6004 unregister_as_ext3();
6005 unregister_filesystem(&ext4_fs_type);
6006 destroy_inodecache();
6007 ext4_exit_mballoc();
6008 ext4_exit_sysfs();
6009 ext4_exit_system_zone();
6010 ext4_exit_pageio();
6011 ext4_exit_es();
6012 }
6014 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6015 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6016 MODULE_LICENSE("GPL");
6017 MODULE_SOFTDEP("pre: crc32c");
6018 module_init(ext4_init_fs)
6019 module_exit(ext4_exit_fs)