]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - fs/ext4/super.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ext4: preserve the needs_recovery flag when the journal is aborted
[mirror_ubuntu-bionic-kernel.git] / fs / ext4 / super.c
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <linux/uaccess.h>
42
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
45
46 #include "ext4.h"
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
49 #include "xattr.h"
50 #include "acl.h"
51 #include "mballoc.h"
52
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
55
56 static struct ext4_lazy_init *ext4_li_info;
57 static struct mutex ext4_li_mtx;
58 static struct ratelimit_state ext4_mount_msg_ratelimit;
59
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 unsigned long journal_devnum);
62 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
63 static int ext4_commit_super(struct super_block *sb, int sync);
64 static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66 static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
68 static int ext4_sync_fs(struct super_block *sb, int wait);
69 static int ext4_remount(struct super_block *sb, int *flags, char *data);
70 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
71 static int ext4_unfreeze(struct super_block *sb);
72 static int ext4_freeze(struct super_block *sb);
73 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74 const char *dev_name, void *data);
75 static inline int ext2_feature_set_ok(struct super_block *sb);
76 static inline int ext3_feature_set_ok(struct super_block *sb);
77 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block *sb);
80 static void ext4_clear_request_list(void);
81 static struct inode *ext4_get_journal_inode(struct super_block *sb,
82 unsigned int journal_inum);
83
84 /*
85 * Lock ordering
86 *
87 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
88 * i_mmap_rwsem (inode->i_mmap_rwsem)!
89 *
90 * page fault path:
91 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
92 * page lock -> i_data_sem (rw)
93 *
94 * buffered write path:
95 * sb_start_write -> i_mutex -> mmap_sem
96 * sb_start_write -> i_mutex -> transaction start -> page lock ->
97 * i_data_sem (rw)
98 *
99 * truncate:
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * i_mmap_rwsem (w) -> page lock
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * transaction start -> i_data_sem (rw)
104 *
105 * direct IO:
106 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
108 * transaction start -> i_data_sem (rw)
109 *
110 * writepages:
111 * transaction start -> page lock(s) -> i_data_sem (rw)
112 */
113
114 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
115 static struct file_system_type ext2_fs_type = {
116 .owner = THIS_MODULE,
117 .name = "ext2",
118 .mount = ext4_mount,
119 .kill_sb = kill_block_super,
120 .fs_flags = FS_REQUIRES_DEV,
121 };
122 MODULE_ALIAS_FS("ext2");
123 MODULE_ALIAS("ext2");
124 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
125 #else
126 #define IS_EXT2_SB(sb) (0)
127 #endif
128
129
130 static struct file_system_type ext3_fs_type = {
131 .owner = THIS_MODULE,
132 .name = "ext3",
133 .mount = ext4_mount,
134 .kill_sb = kill_block_super,
135 .fs_flags = FS_REQUIRES_DEV,
136 };
137 MODULE_ALIAS_FS("ext3");
138 MODULE_ALIAS("ext3");
139 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
140
141 static int ext4_verify_csum_type(struct super_block *sb,
142 struct ext4_super_block *es)
143 {
144 if (!ext4_has_feature_metadata_csum(sb))
145 return 1;
146
147 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
148 }
149
150 static __le32 ext4_superblock_csum(struct super_block *sb,
151 struct ext4_super_block *es)
152 {
153 struct ext4_sb_info *sbi = EXT4_SB(sb);
154 int offset = offsetof(struct ext4_super_block, s_checksum);
155 __u32 csum;
156
157 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
158
159 return cpu_to_le32(csum);
160 }
161
162 static int ext4_superblock_csum_verify(struct super_block *sb,
163 struct ext4_super_block *es)
164 {
165 if (!ext4_has_metadata_csum(sb))
166 return 1;
167
168 return es->s_checksum == ext4_superblock_csum(sb, es);
169 }
170
171 void ext4_superblock_csum_set(struct super_block *sb)
172 {
173 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
174
175 if (!ext4_has_metadata_csum(sb))
176 return;
177
178 es->s_checksum = ext4_superblock_csum(sb, es);
179 }
180
181 void *ext4_kvmalloc(size_t size, gfp_t flags)
182 {
183 void *ret;
184
185 ret = kmalloc(size, flags | __GFP_NOWARN);
186 if (!ret)
187 ret = __vmalloc(size, flags, PAGE_KERNEL);
188 return ret;
189 }
190
191 void *ext4_kvzalloc(size_t size, gfp_t flags)
192 {
193 void *ret;
194
195 ret = kzalloc(size, flags | __GFP_NOWARN);
196 if (!ret)
197 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
198 return ret;
199 }
200
201 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
202 struct ext4_group_desc *bg)
203 {
204 return le32_to_cpu(bg->bg_block_bitmap_lo) |
205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
206 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
207 }
208
209 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
210 struct ext4_group_desc *bg)
211 {
212 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
214 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
215 }
216
217 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
218 struct ext4_group_desc *bg)
219 {
220 return le32_to_cpu(bg->bg_inode_table_lo) |
221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
222 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
223 }
224
225 __u32 ext4_free_group_clusters(struct super_block *sb,
226 struct ext4_group_desc *bg)
227 {
228 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
230 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
231 }
232
233 __u32 ext4_free_inodes_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
235 {
236 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
238 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
239 }
240
241 __u32 ext4_used_dirs_count(struct super_block *sb,
242 struct ext4_group_desc *bg)
243 {
244 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
245 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
246 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
247 }
248
249 __u32 ext4_itable_unused_count(struct super_block *sb,
250 struct ext4_group_desc *bg)
251 {
252 return le16_to_cpu(bg->bg_itable_unused_lo) |
253 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
254 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
255 }
256
257 void ext4_block_bitmap_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
259 {
260 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
263 }
264
265 void ext4_inode_bitmap_set(struct super_block *sb,
266 struct ext4_group_desc *bg, ext4_fsblk_t blk)
267 {
268 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
271 }
272
273 void ext4_inode_table_set(struct super_block *sb,
274 struct ext4_group_desc *bg, ext4_fsblk_t blk)
275 {
276 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
279 }
280
281 void ext4_free_group_clusters_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
283 {
284 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
287 }
288
289 void ext4_free_inodes_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
291 {
292 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
295 }
296
297 void ext4_used_dirs_set(struct super_block *sb,
298 struct ext4_group_desc *bg, __u32 count)
299 {
300 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
301 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
302 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
303 }
304
305 void ext4_itable_unused_set(struct super_block *sb,
306 struct ext4_group_desc *bg, __u32 count)
307 {
308 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
309 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
310 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
311 }
312
313
314 static void __save_error_info(struct super_block *sb, const char *func,
315 unsigned int line)
316 {
317 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
318
319 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
320 if (bdev_read_only(sb->s_bdev))
321 return;
322 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
323 es->s_last_error_time = cpu_to_le32(get_seconds());
324 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
325 es->s_last_error_line = cpu_to_le32(line);
326 if (!es->s_first_error_time) {
327 es->s_first_error_time = es->s_last_error_time;
328 strncpy(es->s_first_error_func, func,
329 sizeof(es->s_first_error_func));
330 es->s_first_error_line = cpu_to_le32(line);
331 es->s_first_error_ino = es->s_last_error_ino;
332 es->s_first_error_block = es->s_last_error_block;
333 }
334 /*
335 * Start the daily error reporting function if it hasn't been
336 * started already
337 */
338 if (!es->s_error_count)
339 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
340 le32_add_cpu(&es->s_error_count, 1);
341 }
342
343 static void save_error_info(struct super_block *sb, const char *func,
344 unsigned int line)
345 {
346 __save_error_info(sb, func, line);
347 ext4_commit_super(sb, 1);
348 }
349
350 /*
351 * The del_gendisk() function uninitializes the disk-specific data
352 * structures, including the bdi structure, without telling anyone
353 * else. Once this happens, any attempt to call mark_buffer_dirty()
354 * (for example, by ext4_commit_super), will cause a kernel OOPS.
355 * This is a kludge to prevent these oops until we can put in a proper
356 * hook in del_gendisk() to inform the VFS and file system layers.
357 */
358 static int block_device_ejected(struct super_block *sb)
359 {
360 struct inode *bd_inode = sb->s_bdev->bd_inode;
361 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
362
363 return bdi->dev == NULL;
364 }
365
366 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
367 {
368 struct super_block *sb = journal->j_private;
369 struct ext4_sb_info *sbi = EXT4_SB(sb);
370 int error = is_journal_aborted(journal);
371 struct ext4_journal_cb_entry *jce;
372
373 BUG_ON(txn->t_state == T_FINISHED);
374 spin_lock(&sbi->s_md_lock);
375 while (!list_empty(&txn->t_private_list)) {
376 jce = list_entry(txn->t_private_list.next,
377 struct ext4_journal_cb_entry, jce_list);
378 list_del_init(&jce->jce_list);
379 spin_unlock(&sbi->s_md_lock);
380 jce->jce_func(sb, jce, error);
381 spin_lock(&sbi->s_md_lock);
382 }
383 spin_unlock(&sbi->s_md_lock);
384 }
385
386 /* Deal with the reporting of failure conditions on a filesystem such as
387 * inconsistencies detected or read IO failures.
388 *
389 * On ext2, we can store the error state of the filesystem in the
390 * superblock. That is not possible on ext4, because we may have other
391 * write ordering constraints on the superblock which prevent us from
392 * writing it out straight away; and given that the journal is about to
393 * be aborted, we can't rely on the current, or future, transactions to
394 * write out the superblock safely.
395 *
396 * We'll just use the jbd2_journal_abort() error code to record an error in
397 * the journal instead. On recovery, the journal will complain about
398 * that error until we've noted it down and cleared it.
399 */
400
401 static void ext4_handle_error(struct super_block *sb)
402 {
403 if (sb->s_flags & MS_RDONLY)
404 return;
405
406 if (!test_opt(sb, ERRORS_CONT)) {
407 journal_t *journal = EXT4_SB(sb)->s_journal;
408
409 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
410 if (journal)
411 jbd2_journal_abort(journal, -EIO);
412 }
413 if (test_opt(sb, ERRORS_RO)) {
414 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
415 /*
416 * Make sure updated value of ->s_mount_flags will be visible
417 * before ->s_flags update
418 */
419 smp_wmb();
420 sb->s_flags |= MS_RDONLY;
421 }
422 if (test_opt(sb, ERRORS_PANIC)) {
423 if (EXT4_SB(sb)->s_journal &&
424 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
425 return;
426 panic("EXT4-fs (device %s): panic forced after error\n",
427 sb->s_id);
428 }
429 }
430
431 #define ext4_error_ratelimit(sb) \
432 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
433 "EXT4-fs error")
434
435 void __ext4_error(struct super_block *sb, const char *function,
436 unsigned int line, const char *fmt, ...)
437 {
438 struct va_format vaf;
439 va_list args;
440
441 if (ext4_error_ratelimit(sb)) {
442 va_start(args, fmt);
443 vaf.fmt = fmt;
444 vaf.va = &args;
445 printk(KERN_CRIT
446 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
447 sb->s_id, function, line, current->comm, &vaf);
448 va_end(args);
449 }
450 save_error_info(sb, function, line);
451 ext4_handle_error(sb);
452 }
453
454 void __ext4_error_inode(struct inode *inode, const char *function,
455 unsigned int line, ext4_fsblk_t block,
456 const char *fmt, ...)
457 {
458 va_list args;
459 struct va_format vaf;
460 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
461
462 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
463 es->s_last_error_block = cpu_to_le64(block);
464 if (ext4_error_ratelimit(inode->i_sb)) {
465 va_start(args, fmt);
466 vaf.fmt = fmt;
467 vaf.va = &args;
468 if (block)
469 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
470 "inode #%lu: block %llu: comm %s: %pV\n",
471 inode->i_sb->s_id, function, line, inode->i_ino,
472 block, current->comm, &vaf);
473 else
474 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
475 "inode #%lu: comm %s: %pV\n",
476 inode->i_sb->s_id, function, line, inode->i_ino,
477 current->comm, &vaf);
478 va_end(args);
479 }
480 save_error_info(inode->i_sb, function, line);
481 ext4_handle_error(inode->i_sb);
482 }
483
484 void __ext4_error_file(struct file *file, const char *function,
485 unsigned int line, ext4_fsblk_t block,
486 const char *fmt, ...)
487 {
488 va_list args;
489 struct va_format vaf;
490 struct ext4_super_block *es;
491 struct inode *inode = file_inode(file);
492 char pathname[80], *path;
493
494 es = EXT4_SB(inode->i_sb)->s_es;
495 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
496 if (ext4_error_ratelimit(inode->i_sb)) {
497 path = file_path(file, pathname, sizeof(pathname));
498 if (IS_ERR(path))
499 path = "(unknown)";
500 va_start(args, fmt);
501 vaf.fmt = fmt;
502 vaf.va = &args;
503 if (block)
504 printk(KERN_CRIT
505 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
506 "block %llu: comm %s: path %s: %pV\n",
507 inode->i_sb->s_id, function, line, inode->i_ino,
508 block, current->comm, path, &vaf);
509 else
510 printk(KERN_CRIT
511 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
512 "comm %s: path %s: %pV\n",
513 inode->i_sb->s_id, function, line, inode->i_ino,
514 current->comm, path, &vaf);
515 va_end(args);
516 }
517 save_error_info(inode->i_sb, function, line);
518 ext4_handle_error(inode->i_sb);
519 }
520
521 const char *ext4_decode_error(struct super_block *sb, int errno,
522 char nbuf[16])
523 {
524 char *errstr = NULL;
525
526 switch (errno) {
527 case -EFSCORRUPTED:
528 errstr = "Corrupt filesystem";
529 break;
530 case -EFSBADCRC:
531 errstr = "Filesystem failed CRC";
532 break;
533 case -EIO:
534 errstr = "IO failure";
535 break;
536 case -ENOMEM:
537 errstr = "Out of memory";
538 break;
539 case -EROFS:
540 if (!sb || (EXT4_SB(sb)->s_journal &&
541 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
542 errstr = "Journal has aborted";
543 else
544 errstr = "Readonly filesystem";
545 break;
546 default:
547 /* If the caller passed in an extra buffer for unknown
548 * errors, textualise them now. Else we just return
549 * NULL. */
550 if (nbuf) {
551 /* Check for truncated error codes... */
552 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
553 errstr = nbuf;
554 }
555 break;
556 }
557
558 return errstr;
559 }
560
561 /* __ext4_std_error decodes expected errors from journaling functions
562 * automatically and invokes the appropriate error response. */
563
564 void __ext4_std_error(struct super_block *sb, const char *function,
565 unsigned int line, int errno)
566 {
567 char nbuf[16];
568 const char *errstr;
569
570 /* Special case: if the error is EROFS, and we're not already
571 * inside a transaction, then there's really no point in logging
572 * an error. */
573 if (errno == -EROFS && journal_current_handle() == NULL &&
574 (sb->s_flags & MS_RDONLY))
575 return;
576
577 if (ext4_error_ratelimit(sb)) {
578 errstr = ext4_decode_error(sb, errno, nbuf);
579 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
580 sb->s_id, function, line, errstr);
581 }
582
583 save_error_info(sb, function, line);
584 ext4_handle_error(sb);
585 }
586
587 /*
588 * ext4_abort is a much stronger failure handler than ext4_error. The
589 * abort function may be used to deal with unrecoverable failures such
590 * as journal IO errors or ENOMEM at a critical moment in log management.
591 *
592 * We unconditionally force the filesystem into an ABORT|READONLY state,
593 * unless the error response on the fs has been set to panic in which
594 * case we take the easy way out and panic immediately.
595 */
596
597 void __ext4_abort(struct super_block *sb, const char *function,
598 unsigned int line, const char *fmt, ...)
599 {
600 struct va_format vaf;
601 va_list args;
602
603 save_error_info(sb, function, line);
604 va_start(args, fmt);
605 vaf.fmt = fmt;
606 vaf.va = &args;
607 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
608 sb->s_id, function, line, &vaf);
609 va_end(args);
610
611 if ((sb->s_flags & MS_RDONLY) == 0) {
612 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
613 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
614 /*
615 * Make sure updated value of ->s_mount_flags will be visible
616 * before ->s_flags update
617 */
618 smp_wmb();
619 sb->s_flags |= MS_RDONLY;
620 if (EXT4_SB(sb)->s_journal)
621 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
622 save_error_info(sb, function, line);
623 }
624 if (test_opt(sb, ERRORS_PANIC)) {
625 if (EXT4_SB(sb)->s_journal &&
626 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
627 return;
628 panic("EXT4-fs panic from previous error\n");
629 }
630 }
631
632 void __ext4_msg(struct super_block *sb,
633 const char *prefix, const char *fmt, ...)
634 {
635 struct va_format vaf;
636 va_list args;
637
638 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
639 return;
640
641 va_start(args, fmt);
642 vaf.fmt = fmt;
643 vaf.va = &args;
644 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
645 va_end(args);
646 }
647
648 #define ext4_warning_ratelimit(sb) \
649 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
650 "EXT4-fs warning")
651
652 void __ext4_warning(struct super_block *sb, const char *function,
653 unsigned int line, const char *fmt, ...)
654 {
655 struct va_format vaf;
656 va_list args;
657
658 if (!ext4_warning_ratelimit(sb))
659 return;
660
661 va_start(args, fmt);
662 vaf.fmt = fmt;
663 vaf.va = &args;
664 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
665 sb->s_id, function, line, &vaf);
666 va_end(args);
667 }
668
669 void __ext4_warning_inode(const struct inode *inode, const char *function,
670 unsigned int line, const char *fmt, ...)
671 {
672 struct va_format vaf;
673 va_list args;
674
675 if (!ext4_warning_ratelimit(inode->i_sb))
676 return;
677
678 va_start(args, fmt);
679 vaf.fmt = fmt;
680 vaf.va = &args;
681 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
682 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
683 function, line, inode->i_ino, current->comm, &vaf);
684 va_end(args);
685 }
686
687 void __ext4_grp_locked_error(const char *function, unsigned int line,
688 struct super_block *sb, ext4_group_t grp,
689 unsigned long ino, ext4_fsblk_t block,
690 const char *fmt, ...)
691 __releases(bitlock)
692 __acquires(bitlock)
693 {
694 struct va_format vaf;
695 va_list args;
696 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
697
698 es->s_last_error_ino = cpu_to_le32(ino);
699 es->s_last_error_block = cpu_to_le64(block);
700 __save_error_info(sb, function, line);
701
702 if (ext4_error_ratelimit(sb)) {
703 va_start(args, fmt);
704 vaf.fmt = fmt;
705 vaf.va = &args;
706 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
707 sb->s_id, function, line, grp);
708 if (ino)
709 printk(KERN_CONT "inode %lu: ", ino);
710 if (block)
711 printk(KERN_CONT "block %llu:",
712 (unsigned long long) block);
713 printk(KERN_CONT "%pV\n", &vaf);
714 va_end(args);
715 }
716
717 if (test_opt(sb, ERRORS_CONT)) {
718 ext4_commit_super(sb, 0);
719 return;
720 }
721
722 ext4_unlock_group(sb, grp);
723 ext4_handle_error(sb);
724 /*
725 * We only get here in the ERRORS_RO case; relocking the group
726 * may be dangerous, but nothing bad will happen since the
727 * filesystem will have already been marked read/only and the
728 * journal has been aborted. We return 1 as a hint to callers
729 * who might what to use the return value from
730 * ext4_grp_locked_error() to distinguish between the
731 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
732 * aggressively from the ext4 function in question, with a
733 * more appropriate error code.
734 */
735 ext4_lock_group(sb, grp);
736 return;
737 }
738
739 void ext4_update_dynamic_rev(struct super_block *sb)
740 {
741 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
742
743 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
744 return;
745
746 ext4_warning(sb,
747 "updating to rev %d because of new feature flag, "
748 "running e2fsck is recommended",
749 EXT4_DYNAMIC_REV);
750
751 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
752 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
753 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
754 /* leave es->s_feature_*compat flags alone */
755 /* es->s_uuid will be set by e2fsck if empty */
756
757 /*
758 * The rest of the superblock fields should be zero, and if not it
759 * means they are likely already in use, so leave them alone. We
760 * can leave it up to e2fsck to clean up any inconsistencies there.
761 */
762 }
763
764 /*
765 * Open the external journal device
766 */
767 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
768 {
769 struct block_device *bdev;
770 char b[BDEVNAME_SIZE];
771
772 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
773 if (IS_ERR(bdev))
774 goto fail;
775 return bdev;
776
777 fail:
778 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
779 __bdevname(dev, b), PTR_ERR(bdev));
780 return NULL;
781 }
782
783 /*
784 * Release the journal device
785 */
786 static void ext4_blkdev_put(struct block_device *bdev)
787 {
788 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
789 }
790
791 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
792 {
793 struct block_device *bdev;
794 bdev = sbi->journal_bdev;
795 if (bdev) {
796 ext4_blkdev_put(bdev);
797 sbi->journal_bdev = NULL;
798 }
799 }
800
801 static inline struct inode *orphan_list_entry(struct list_head *l)
802 {
803 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
804 }
805
806 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
807 {
808 struct list_head *l;
809
810 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
811 le32_to_cpu(sbi->s_es->s_last_orphan));
812
813 printk(KERN_ERR "sb_info orphan list:\n");
814 list_for_each(l, &sbi->s_orphan) {
815 struct inode *inode = orphan_list_entry(l);
816 printk(KERN_ERR " "
817 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
818 inode->i_sb->s_id, inode->i_ino, inode,
819 inode->i_mode, inode->i_nlink,
820 NEXT_ORPHAN(inode));
821 }
822 }
823
824 static void ext4_put_super(struct super_block *sb)
825 {
826 struct ext4_sb_info *sbi = EXT4_SB(sb);
827 struct ext4_super_block *es = sbi->s_es;
828 int aborted = 0;
829 int i, err;
830
831 ext4_unregister_li_request(sb);
832 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
833
834 flush_workqueue(sbi->rsv_conversion_wq);
835 destroy_workqueue(sbi->rsv_conversion_wq);
836
837 if (sbi->s_journal) {
838 aborted = is_journal_aborted(sbi->s_journal);
839 err = jbd2_journal_destroy(sbi->s_journal);
840 sbi->s_journal = NULL;
841 if ((err < 0) && !aborted)
842 ext4_abort(sb, "Couldn't clean up the journal");
843 }
844
845 ext4_unregister_sysfs(sb);
846 ext4_es_unregister_shrinker(sbi);
847 del_timer_sync(&sbi->s_err_report);
848 ext4_release_system_zone(sb);
849 ext4_mb_release(sb);
850 ext4_ext_release(sb);
851
852 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
853 ext4_clear_feature_journal_needs_recovery(sb);
854 es->s_state = cpu_to_le16(sbi->s_mount_state);
855 }
856 if (!(sb->s_flags & MS_RDONLY))
857 ext4_commit_super(sb, 1);
858
859 for (i = 0; i < sbi->s_gdb_count; i++)
860 brelse(sbi->s_group_desc[i]);
861 kvfree(sbi->s_group_desc);
862 kvfree(sbi->s_flex_groups);
863 percpu_counter_destroy(&sbi->s_freeclusters_counter);
864 percpu_counter_destroy(&sbi->s_freeinodes_counter);
865 percpu_counter_destroy(&sbi->s_dirs_counter);
866 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
867 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
868 #ifdef CONFIG_QUOTA
869 for (i = 0; i < EXT4_MAXQUOTAS; i++)
870 kfree(sbi->s_qf_names[i]);
871 #endif
872
873 /* Debugging code just in case the in-memory inode orphan list
874 * isn't empty. The on-disk one can be non-empty if we've
875 * detected an error and taken the fs readonly, but the
876 * in-memory list had better be clean by this point. */
877 if (!list_empty(&sbi->s_orphan))
878 dump_orphan_list(sb, sbi);
879 J_ASSERT(list_empty(&sbi->s_orphan));
880
881 sync_blockdev(sb->s_bdev);
882 invalidate_bdev(sb->s_bdev);
883 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
884 /*
885 * Invalidate the journal device's buffers. We don't want them
886 * floating about in memory - the physical journal device may
887 * hotswapped, and it breaks the `ro-after' testing code.
888 */
889 sync_blockdev(sbi->journal_bdev);
890 invalidate_bdev(sbi->journal_bdev);
891 ext4_blkdev_remove(sbi);
892 }
893 if (sbi->s_mb_cache) {
894 ext4_xattr_destroy_cache(sbi->s_mb_cache);
895 sbi->s_mb_cache = NULL;
896 }
897 if (sbi->s_mmp_tsk)
898 kthread_stop(sbi->s_mmp_tsk);
899 brelse(sbi->s_sbh);
900 sb->s_fs_info = NULL;
901 /*
902 * Now that we are completely done shutting down the
903 * superblock, we need to actually destroy the kobject.
904 */
905 kobject_put(&sbi->s_kobj);
906 wait_for_completion(&sbi->s_kobj_unregister);
907 if (sbi->s_chksum_driver)
908 crypto_free_shash(sbi->s_chksum_driver);
909 kfree(sbi->s_blockgroup_lock);
910 kfree(sbi);
911 }
912
913 static struct kmem_cache *ext4_inode_cachep;
914
915 /*
916 * Called inside transaction, so use GFP_NOFS
917 */
918 static struct inode *ext4_alloc_inode(struct super_block *sb)
919 {
920 struct ext4_inode_info *ei;
921
922 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
923 if (!ei)
924 return NULL;
925
926 ei->vfs_inode.i_version = 1;
927 spin_lock_init(&ei->i_raw_lock);
928 INIT_LIST_HEAD(&ei->i_prealloc_list);
929 spin_lock_init(&ei->i_prealloc_lock);
930 ext4_es_init_tree(&ei->i_es_tree);
931 rwlock_init(&ei->i_es_lock);
932 INIT_LIST_HEAD(&ei->i_es_list);
933 ei->i_es_all_nr = 0;
934 ei->i_es_shk_nr = 0;
935 ei->i_es_shrink_lblk = 0;
936 ei->i_reserved_data_blocks = 0;
937 ei->i_reserved_meta_blocks = 0;
938 ei->i_allocated_meta_blocks = 0;
939 ei->i_da_metadata_calc_len = 0;
940 ei->i_da_metadata_calc_last_lblock = 0;
941 spin_lock_init(&(ei->i_block_reservation_lock));
942 #ifdef CONFIG_QUOTA
943 ei->i_reserved_quota = 0;
944 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
945 #endif
946 ei->jinode = NULL;
947 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
948 spin_lock_init(&ei->i_completed_io_lock);
949 ei->i_sync_tid = 0;
950 ei->i_datasync_tid = 0;
951 atomic_set(&ei->i_unwritten, 0);
952 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
953 return &ei->vfs_inode;
954 }
955
956 static int ext4_drop_inode(struct inode *inode)
957 {
958 int drop = generic_drop_inode(inode);
959
960 trace_ext4_drop_inode(inode, drop);
961 return drop;
962 }
963
964 static void ext4_i_callback(struct rcu_head *head)
965 {
966 struct inode *inode = container_of(head, struct inode, i_rcu);
967 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
968 }
969
970 static void ext4_destroy_inode(struct inode *inode)
971 {
972 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
973 ext4_msg(inode->i_sb, KERN_ERR,
974 "Inode %lu (%p): orphan list check failed!",
975 inode->i_ino, EXT4_I(inode));
976 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
977 EXT4_I(inode), sizeof(struct ext4_inode_info),
978 true);
979 dump_stack();
980 }
981 call_rcu(&inode->i_rcu, ext4_i_callback);
982 }
983
984 static void init_once(void *foo)
985 {
986 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
987
988 INIT_LIST_HEAD(&ei->i_orphan);
989 init_rwsem(&ei->xattr_sem);
990 init_rwsem(&ei->i_data_sem);
991 init_rwsem(&ei->i_mmap_sem);
992 inode_init_once(&ei->vfs_inode);
993 }
994
995 static int __init init_inodecache(void)
996 {
997 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
998 sizeof(struct ext4_inode_info),
999 0, (SLAB_RECLAIM_ACCOUNT|
1000 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1001 init_once);
1002 if (ext4_inode_cachep == NULL)
1003 return -ENOMEM;
1004 return 0;
1005 }
1006
1007 static void destroy_inodecache(void)
1008 {
1009 /*
1010 * Make sure all delayed rcu free inodes are flushed before we
1011 * destroy cache.
1012 */
1013 rcu_barrier();
1014 kmem_cache_destroy(ext4_inode_cachep);
1015 }
1016
1017 void ext4_clear_inode(struct inode *inode)
1018 {
1019 invalidate_inode_buffers(inode);
1020 clear_inode(inode);
1021 dquot_drop(inode);
1022 ext4_discard_preallocations(inode);
1023 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1024 if (EXT4_I(inode)->jinode) {
1025 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1026 EXT4_I(inode)->jinode);
1027 jbd2_free_inode(EXT4_I(inode)->jinode);
1028 EXT4_I(inode)->jinode = NULL;
1029 }
1030 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1031 fscrypt_put_encryption_info(inode, NULL);
1032 #endif
1033 }
1034
1035 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1036 u64 ino, u32 generation)
1037 {
1038 struct inode *inode;
1039
1040 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1041 return ERR_PTR(-ESTALE);
1042 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1043 return ERR_PTR(-ESTALE);
1044
1045 /* iget isn't really right if the inode is currently unallocated!!
1046 *
1047 * ext4_read_inode will return a bad_inode if the inode had been
1048 * deleted, so we should be safe.
1049 *
1050 * Currently we don't know the generation for parent directory, so
1051 * a generation of 0 means "accept any"
1052 */
1053 inode = ext4_iget_normal(sb, ino);
1054 if (IS_ERR(inode))
1055 return ERR_CAST(inode);
1056 if (generation && inode->i_generation != generation) {
1057 iput(inode);
1058 return ERR_PTR(-ESTALE);
1059 }
1060
1061 return inode;
1062 }
1063
1064 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1065 int fh_len, int fh_type)
1066 {
1067 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1068 ext4_nfs_get_inode);
1069 }
1070
1071 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1072 int fh_len, int fh_type)
1073 {
1074 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1075 ext4_nfs_get_inode);
1076 }
1077
1078 /*
1079 * Try to release metadata pages (indirect blocks, directories) which are
1080 * mapped via the block device. Since these pages could have journal heads
1081 * which would prevent try_to_free_buffers() from freeing them, we must use
1082 * jbd2 layer's try_to_free_buffers() function to release them.
1083 */
1084 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1085 gfp_t wait)
1086 {
1087 journal_t *journal = EXT4_SB(sb)->s_journal;
1088
1089 WARN_ON(PageChecked(page));
1090 if (!page_has_buffers(page))
1091 return 0;
1092 if (journal)
1093 return jbd2_journal_try_to_free_buffers(journal, page,
1094 wait & ~__GFP_DIRECT_RECLAIM);
1095 return try_to_free_buffers(page);
1096 }
1097
1098 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1099 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1100 {
1101 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1102 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1103 }
1104
1105 static int ext4_prepare_context(struct inode *inode)
1106 {
1107 return ext4_convert_inline_data(inode);
1108 }
1109
1110 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1111 void *fs_data)
1112 {
1113 handle_t *handle = fs_data;
1114 int res, res2, retries = 0;
1115
1116 /*
1117 * If a journal handle was specified, then the encryption context is
1118 * being set on a new inode via inheritance and is part of a larger
1119 * transaction to create the inode. Otherwise the encryption context is
1120 * being set on an existing inode in its own transaction. Only in the
1121 * latter case should the "retry on ENOSPC" logic be used.
1122 */
1123
1124 if (handle) {
1125 res = ext4_xattr_set_handle(handle, inode,
1126 EXT4_XATTR_INDEX_ENCRYPTION,
1127 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1128 ctx, len, 0);
1129 if (!res) {
1130 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1131 ext4_clear_inode_state(inode,
1132 EXT4_STATE_MAY_INLINE_DATA);
1133 /*
1134 * Update inode->i_flags - e.g. S_DAX may get disabled
1135 */
1136 ext4_set_inode_flags(inode);
1137 }
1138 return res;
1139 }
1140
1141 retry:
1142 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1143 ext4_jbd2_credits_xattr(inode));
1144 if (IS_ERR(handle))
1145 return PTR_ERR(handle);
1146
1147 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1148 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1149 ctx, len, 0);
1150 if (!res) {
1151 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1152 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1153 ext4_set_inode_flags(inode);
1154 res = ext4_mark_inode_dirty(handle, inode);
1155 if (res)
1156 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1157 }
1158 res2 = ext4_journal_stop(handle);
1159
1160 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1161 goto retry;
1162 if (!res)
1163 res = res2;
1164 return res;
1165 }
1166
1167 static int ext4_dummy_context(struct inode *inode)
1168 {
1169 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1170 }
1171
1172 static unsigned ext4_max_namelen(struct inode *inode)
1173 {
1174 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1175 EXT4_NAME_LEN;
1176 }
1177
1178 static struct fscrypt_operations ext4_cryptops = {
1179 .key_prefix = "ext4:",
1180 .get_context = ext4_get_context,
1181 .prepare_context = ext4_prepare_context,
1182 .set_context = ext4_set_context,
1183 .dummy_context = ext4_dummy_context,
1184 .is_encrypted = ext4_encrypted_inode,
1185 .empty_dir = ext4_empty_dir,
1186 .max_namelen = ext4_max_namelen,
1187 };
1188 #else
1189 static struct fscrypt_operations ext4_cryptops = {
1190 .is_encrypted = ext4_encrypted_inode,
1191 };
1192 #endif
1193
1194 #ifdef CONFIG_QUOTA
1195 static char *quotatypes[] = INITQFNAMES;
1196 #define QTYPE2NAME(t) (quotatypes[t])
1197
1198 static int ext4_write_dquot(struct dquot *dquot);
1199 static int ext4_acquire_dquot(struct dquot *dquot);
1200 static int ext4_release_dquot(struct dquot *dquot);
1201 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1202 static int ext4_write_info(struct super_block *sb, int type);
1203 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1204 const struct path *path);
1205 static int ext4_quota_off(struct super_block *sb, int type);
1206 static int ext4_quota_on_mount(struct super_block *sb, int type);
1207 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1208 size_t len, loff_t off);
1209 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1210 const char *data, size_t len, loff_t off);
1211 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1212 unsigned int flags);
1213 static int ext4_enable_quotas(struct super_block *sb);
1214 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1215
1216 static struct dquot **ext4_get_dquots(struct inode *inode)
1217 {
1218 return EXT4_I(inode)->i_dquot;
1219 }
1220
1221 static const struct dquot_operations ext4_quota_operations = {
1222 .get_reserved_space = ext4_get_reserved_space,
1223 .write_dquot = ext4_write_dquot,
1224 .acquire_dquot = ext4_acquire_dquot,
1225 .release_dquot = ext4_release_dquot,
1226 .mark_dirty = ext4_mark_dquot_dirty,
1227 .write_info = ext4_write_info,
1228 .alloc_dquot = dquot_alloc,
1229 .destroy_dquot = dquot_destroy,
1230 .get_projid = ext4_get_projid,
1231 .get_next_id = ext4_get_next_id,
1232 };
1233
1234 static const struct quotactl_ops ext4_qctl_operations = {
1235 .quota_on = ext4_quota_on,
1236 .quota_off = ext4_quota_off,
1237 .quota_sync = dquot_quota_sync,
1238 .get_state = dquot_get_state,
1239 .set_info = dquot_set_dqinfo,
1240 .get_dqblk = dquot_get_dqblk,
1241 .set_dqblk = dquot_set_dqblk,
1242 .get_nextdqblk = dquot_get_next_dqblk,
1243 };
1244 #endif
1245
1246 static const struct super_operations ext4_sops = {
1247 .alloc_inode = ext4_alloc_inode,
1248 .destroy_inode = ext4_destroy_inode,
1249 .write_inode = ext4_write_inode,
1250 .dirty_inode = ext4_dirty_inode,
1251 .drop_inode = ext4_drop_inode,
1252 .evict_inode = ext4_evict_inode,
1253 .put_super = ext4_put_super,
1254 .sync_fs = ext4_sync_fs,
1255 .freeze_fs = ext4_freeze,
1256 .unfreeze_fs = ext4_unfreeze,
1257 .statfs = ext4_statfs,
1258 .remount_fs = ext4_remount,
1259 .show_options = ext4_show_options,
1260 #ifdef CONFIG_QUOTA
1261 .quota_read = ext4_quota_read,
1262 .quota_write = ext4_quota_write,
1263 .get_dquots = ext4_get_dquots,
1264 #endif
1265 .bdev_try_to_free_page = bdev_try_to_free_page,
1266 };
1267
1268 static const struct export_operations ext4_export_ops = {
1269 .fh_to_dentry = ext4_fh_to_dentry,
1270 .fh_to_parent = ext4_fh_to_parent,
1271 .get_parent = ext4_get_parent,
1272 };
1273
1274 enum {
1275 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1276 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1277 Opt_nouid32, Opt_debug, Opt_removed,
1278 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1279 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1280 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1281 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1282 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1283 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1284 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1285 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1286 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1287 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1288 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1289 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1290 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1291 Opt_inode_readahead_blks, Opt_journal_ioprio,
1292 Opt_dioread_nolock, Opt_dioread_lock,
1293 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1294 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1295 };
1296
1297 static const match_table_t tokens = {
1298 {Opt_bsd_df, "bsddf"},
1299 {Opt_minix_df, "minixdf"},
1300 {Opt_grpid, "grpid"},
1301 {Opt_grpid, "bsdgroups"},
1302 {Opt_nogrpid, "nogrpid"},
1303 {Opt_nogrpid, "sysvgroups"},
1304 {Opt_resgid, "resgid=%u"},
1305 {Opt_resuid, "resuid=%u"},
1306 {Opt_sb, "sb=%u"},
1307 {Opt_err_cont, "errors=continue"},
1308 {Opt_err_panic, "errors=panic"},
1309 {Opt_err_ro, "errors=remount-ro"},
1310 {Opt_nouid32, "nouid32"},
1311 {Opt_debug, "debug"},
1312 {Opt_removed, "oldalloc"},
1313 {Opt_removed, "orlov"},
1314 {Opt_user_xattr, "user_xattr"},
1315 {Opt_nouser_xattr, "nouser_xattr"},
1316 {Opt_acl, "acl"},
1317 {Opt_noacl, "noacl"},
1318 {Opt_noload, "norecovery"},
1319 {Opt_noload, "noload"},
1320 {Opt_removed, "nobh"},
1321 {Opt_removed, "bh"},
1322 {Opt_commit, "commit=%u"},
1323 {Opt_min_batch_time, "min_batch_time=%u"},
1324 {Opt_max_batch_time, "max_batch_time=%u"},
1325 {Opt_journal_dev, "journal_dev=%u"},
1326 {Opt_journal_path, "journal_path=%s"},
1327 {Opt_journal_checksum, "journal_checksum"},
1328 {Opt_nojournal_checksum, "nojournal_checksum"},
1329 {Opt_journal_async_commit, "journal_async_commit"},
1330 {Opt_abort, "abort"},
1331 {Opt_data_journal, "data=journal"},
1332 {Opt_data_ordered, "data=ordered"},
1333 {Opt_data_writeback, "data=writeback"},
1334 {Opt_data_err_abort, "data_err=abort"},
1335 {Opt_data_err_ignore, "data_err=ignore"},
1336 {Opt_offusrjquota, "usrjquota="},
1337 {Opt_usrjquota, "usrjquota=%s"},
1338 {Opt_offgrpjquota, "grpjquota="},
1339 {Opt_grpjquota, "grpjquota=%s"},
1340 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1341 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1342 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1343 {Opt_grpquota, "grpquota"},
1344 {Opt_noquota, "noquota"},
1345 {Opt_quota, "quota"},
1346 {Opt_usrquota, "usrquota"},
1347 {Opt_prjquota, "prjquota"},
1348 {Opt_barrier, "barrier=%u"},
1349 {Opt_barrier, "barrier"},
1350 {Opt_nobarrier, "nobarrier"},
1351 {Opt_i_version, "i_version"},
1352 {Opt_dax, "dax"},
1353 {Opt_stripe, "stripe=%u"},
1354 {Opt_delalloc, "delalloc"},
1355 {Opt_lazytime, "lazytime"},
1356 {Opt_nolazytime, "nolazytime"},
1357 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1358 {Opt_nodelalloc, "nodelalloc"},
1359 {Opt_removed, "mblk_io_submit"},
1360 {Opt_removed, "nomblk_io_submit"},
1361 {Opt_block_validity, "block_validity"},
1362 {Opt_noblock_validity, "noblock_validity"},
1363 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1364 {Opt_journal_ioprio, "journal_ioprio=%u"},
1365 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1366 {Opt_auto_da_alloc, "auto_da_alloc"},
1367 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1368 {Opt_dioread_nolock, "dioread_nolock"},
1369 {Opt_dioread_lock, "dioread_lock"},
1370 {Opt_discard, "discard"},
1371 {Opt_nodiscard, "nodiscard"},
1372 {Opt_init_itable, "init_itable=%u"},
1373 {Opt_init_itable, "init_itable"},
1374 {Opt_noinit_itable, "noinit_itable"},
1375 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1376 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1377 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1378 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1379 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1380 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1381 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1382 {Opt_err, NULL},
1383 };
1384
1385 static ext4_fsblk_t get_sb_block(void **data)
1386 {
1387 ext4_fsblk_t sb_block;
1388 char *options = (char *) *data;
1389
1390 if (!options || strncmp(options, "sb=", 3) != 0)
1391 return 1; /* Default location */
1392
1393 options += 3;
1394 /* TODO: use simple_strtoll with >32bit ext4 */
1395 sb_block = simple_strtoul(options, &options, 0);
1396 if (*options && *options != ',') {
1397 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1398 (char *) *data);
1399 return 1;
1400 }
1401 if (*options == ',')
1402 options++;
1403 *data = (void *) options;
1404
1405 return sb_block;
1406 }
1407
1408 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1409 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1410 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1411
1412 #ifdef CONFIG_QUOTA
1413 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1414 {
1415 struct ext4_sb_info *sbi = EXT4_SB(sb);
1416 char *qname;
1417 int ret = -1;
1418
1419 if (sb_any_quota_loaded(sb) &&
1420 !sbi->s_qf_names[qtype]) {
1421 ext4_msg(sb, KERN_ERR,
1422 "Cannot change journaled "
1423 "quota options when quota turned on");
1424 return -1;
1425 }
1426 if (ext4_has_feature_quota(sb)) {
1427 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1428 "ignored when QUOTA feature is enabled");
1429 return 1;
1430 }
1431 qname = match_strdup(args);
1432 if (!qname) {
1433 ext4_msg(sb, KERN_ERR,
1434 "Not enough memory for storing quotafile name");
1435 return -1;
1436 }
1437 if (sbi->s_qf_names[qtype]) {
1438 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1439 ret = 1;
1440 else
1441 ext4_msg(sb, KERN_ERR,
1442 "%s quota file already specified",
1443 QTYPE2NAME(qtype));
1444 goto errout;
1445 }
1446 if (strchr(qname, '/')) {
1447 ext4_msg(sb, KERN_ERR,
1448 "quotafile must be on filesystem root");
1449 goto errout;
1450 }
1451 sbi->s_qf_names[qtype] = qname;
1452 set_opt(sb, QUOTA);
1453 return 1;
1454 errout:
1455 kfree(qname);
1456 return ret;
1457 }
1458
1459 static int clear_qf_name(struct super_block *sb, int qtype)
1460 {
1461
1462 struct ext4_sb_info *sbi = EXT4_SB(sb);
1463
1464 if (sb_any_quota_loaded(sb) &&
1465 sbi->s_qf_names[qtype]) {
1466 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1467 " when quota turned on");
1468 return -1;
1469 }
1470 kfree(sbi->s_qf_names[qtype]);
1471 sbi->s_qf_names[qtype] = NULL;
1472 return 1;
1473 }
1474 #endif
1475
1476 #define MOPT_SET 0x0001
1477 #define MOPT_CLEAR 0x0002
1478 #define MOPT_NOSUPPORT 0x0004
1479 #define MOPT_EXPLICIT 0x0008
1480 #define MOPT_CLEAR_ERR 0x0010
1481 #define MOPT_GTE0 0x0020
1482 #ifdef CONFIG_QUOTA
1483 #define MOPT_Q 0
1484 #define MOPT_QFMT 0x0040
1485 #else
1486 #define MOPT_Q MOPT_NOSUPPORT
1487 #define MOPT_QFMT MOPT_NOSUPPORT
1488 #endif
1489 #define MOPT_DATAJ 0x0080
1490 #define MOPT_NO_EXT2 0x0100
1491 #define MOPT_NO_EXT3 0x0200
1492 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1493 #define MOPT_STRING 0x0400
1494
1495 static const struct mount_opts {
1496 int token;
1497 int mount_opt;
1498 int flags;
1499 } ext4_mount_opts[] = {
1500 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1501 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1502 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1503 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1504 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1505 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1506 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1507 MOPT_EXT4_ONLY | MOPT_SET},
1508 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1509 MOPT_EXT4_ONLY | MOPT_CLEAR},
1510 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1511 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1512 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1513 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1514 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1515 MOPT_EXT4_ONLY | MOPT_CLEAR},
1516 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1517 MOPT_EXT4_ONLY | MOPT_CLEAR},
1518 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1519 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1520 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1521 EXT4_MOUNT_JOURNAL_CHECKSUM),
1522 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1523 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1524 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1525 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1526 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1527 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1528 MOPT_NO_EXT2},
1529 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1530 MOPT_NO_EXT2},
1531 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1532 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1533 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1534 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1535 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1536 {Opt_commit, 0, MOPT_GTE0},
1537 {Opt_max_batch_time, 0, MOPT_GTE0},
1538 {Opt_min_batch_time, 0, MOPT_GTE0},
1539 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1540 {Opt_init_itable, 0, MOPT_GTE0},
1541 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1542 {Opt_stripe, 0, MOPT_GTE0},
1543 {Opt_resuid, 0, MOPT_GTE0},
1544 {Opt_resgid, 0, MOPT_GTE0},
1545 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1546 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1547 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1548 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1549 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1550 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1551 MOPT_NO_EXT2 | MOPT_DATAJ},
1552 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1553 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1554 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1555 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1556 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1557 #else
1558 {Opt_acl, 0, MOPT_NOSUPPORT},
1559 {Opt_noacl, 0, MOPT_NOSUPPORT},
1560 #endif
1561 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1562 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1563 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1564 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1565 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1566 MOPT_SET | MOPT_Q},
1567 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1568 MOPT_SET | MOPT_Q},
1569 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1570 MOPT_SET | MOPT_Q},
1571 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1572 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1573 MOPT_CLEAR | MOPT_Q},
1574 {Opt_usrjquota, 0, MOPT_Q},
1575 {Opt_grpjquota, 0, MOPT_Q},
1576 {Opt_offusrjquota, 0, MOPT_Q},
1577 {Opt_offgrpjquota, 0, MOPT_Q},
1578 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1579 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1580 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1581 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1582 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1583 {Opt_err, 0, 0}
1584 };
1585
1586 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1587 substring_t *args, unsigned long *journal_devnum,
1588 unsigned int *journal_ioprio, int is_remount)
1589 {
1590 struct ext4_sb_info *sbi = EXT4_SB(sb);
1591 const struct mount_opts *m;
1592 kuid_t uid;
1593 kgid_t gid;
1594 int arg = 0;
1595
1596 #ifdef CONFIG_QUOTA
1597 if (token == Opt_usrjquota)
1598 return set_qf_name(sb, USRQUOTA, &args[0]);
1599 else if (token == Opt_grpjquota)
1600 return set_qf_name(sb, GRPQUOTA, &args[0]);
1601 else if (token == Opt_offusrjquota)
1602 return clear_qf_name(sb, USRQUOTA);
1603 else if (token == Opt_offgrpjquota)
1604 return clear_qf_name(sb, GRPQUOTA);
1605 #endif
1606 switch (token) {
1607 case Opt_noacl:
1608 case Opt_nouser_xattr:
1609 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1610 break;
1611 case Opt_sb:
1612 return 1; /* handled by get_sb_block() */
1613 case Opt_removed:
1614 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1615 return 1;
1616 case Opt_abort:
1617 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1618 return 1;
1619 case Opt_i_version:
1620 sb->s_flags |= MS_I_VERSION;
1621 return 1;
1622 case Opt_lazytime:
1623 sb->s_flags |= MS_LAZYTIME;
1624 return 1;
1625 case Opt_nolazytime:
1626 sb->s_flags &= ~MS_LAZYTIME;
1627 return 1;
1628 }
1629
1630 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1631 if (token == m->token)
1632 break;
1633
1634 if (m->token == Opt_err) {
1635 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1636 "or missing value", opt);
1637 return -1;
1638 }
1639
1640 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1641 ext4_msg(sb, KERN_ERR,
1642 "Mount option \"%s\" incompatible with ext2", opt);
1643 return -1;
1644 }
1645 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1646 ext4_msg(sb, KERN_ERR,
1647 "Mount option \"%s\" incompatible with ext3", opt);
1648 return -1;
1649 }
1650
1651 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1652 return -1;
1653 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1654 return -1;
1655 if (m->flags & MOPT_EXPLICIT) {
1656 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1657 set_opt2(sb, EXPLICIT_DELALLOC);
1658 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1659 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1660 } else
1661 return -1;
1662 }
1663 if (m->flags & MOPT_CLEAR_ERR)
1664 clear_opt(sb, ERRORS_MASK);
1665 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1666 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1667 "options when quota turned on");
1668 return -1;
1669 }
1670
1671 if (m->flags & MOPT_NOSUPPORT) {
1672 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1673 } else if (token == Opt_commit) {
1674 if (arg == 0)
1675 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1676 sbi->s_commit_interval = HZ * arg;
1677 } else if (token == Opt_debug_want_extra_isize) {
1678 sbi->s_want_extra_isize = arg;
1679 } else if (token == Opt_max_batch_time) {
1680 sbi->s_max_batch_time = arg;
1681 } else if (token == Opt_min_batch_time) {
1682 sbi->s_min_batch_time = arg;
1683 } else if (token == Opt_inode_readahead_blks) {
1684 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1685 ext4_msg(sb, KERN_ERR,
1686 "EXT4-fs: inode_readahead_blks must be "
1687 "0 or a power of 2 smaller than 2^31");
1688 return -1;
1689 }
1690 sbi->s_inode_readahead_blks = arg;
1691 } else if (token == Opt_init_itable) {
1692 set_opt(sb, INIT_INODE_TABLE);
1693 if (!args->from)
1694 arg = EXT4_DEF_LI_WAIT_MULT;
1695 sbi->s_li_wait_mult = arg;
1696 } else if (token == Opt_max_dir_size_kb) {
1697 sbi->s_max_dir_size_kb = arg;
1698 } else if (token == Opt_stripe) {
1699 sbi->s_stripe = arg;
1700 } else if (token == Opt_resuid) {
1701 uid = make_kuid(current_user_ns(), arg);
1702 if (!uid_valid(uid)) {
1703 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1704 return -1;
1705 }
1706 sbi->s_resuid = uid;
1707 } else if (token == Opt_resgid) {
1708 gid = make_kgid(current_user_ns(), arg);
1709 if (!gid_valid(gid)) {
1710 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1711 return -1;
1712 }
1713 sbi->s_resgid = gid;
1714 } else if (token == Opt_journal_dev) {
1715 if (is_remount) {
1716 ext4_msg(sb, KERN_ERR,
1717 "Cannot specify journal on remount");
1718 return -1;
1719 }
1720 *journal_devnum = arg;
1721 } else if (token == Opt_journal_path) {
1722 char *journal_path;
1723 struct inode *journal_inode;
1724 struct path path;
1725 int error;
1726
1727 if (is_remount) {
1728 ext4_msg(sb, KERN_ERR,
1729 "Cannot specify journal on remount");
1730 return -1;
1731 }
1732 journal_path = match_strdup(&args[0]);
1733 if (!journal_path) {
1734 ext4_msg(sb, KERN_ERR, "error: could not dup "
1735 "journal device string");
1736 return -1;
1737 }
1738
1739 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1740 if (error) {
1741 ext4_msg(sb, KERN_ERR, "error: could not find "
1742 "journal device path: error %d", error);
1743 kfree(journal_path);
1744 return -1;
1745 }
1746
1747 journal_inode = d_inode(path.dentry);
1748 if (!S_ISBLK(journal_inode->i_mode)) {
1749 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1750 "is not a block device", journal_path);
1751 path_put(&path);
1752 kfree(journal_path);
1753 return -1;
1754 }
1755
1756 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1757 path_put(&path);
1758 kfree(journal_path);
1759 } else if (token == Opt_journal_ioprio) {
1760 if (arg > 7) {
1761 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1762 " (must be 0-7)");
1763 return -1;
1764 }
1765 *journal_ioprio =
1766 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1767 } else if (token == Opt_test_dummy_encryption) {
1768 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1769 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1770 ext4_msg(sb, KERN_WARNING,
1771 "Test dummy encryption mode enabled");
1772 #else
1773 ext4_msg(sb, KERN_WARNING,
1774 "Test dummy encryption mount option ignored");
1775 #endif
1776 } else if (m->flags & MOPT_DATAJ) {
1777 if (is_remount) {
1778 if (!sbi->s_journal)
1779 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1780 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1781 ext4_msg(sb, KERN_ERR,
1782 "Cannot change data mode on remount");
1783 return -1;
1784 }
1785 } else {
1786 clear_opt(sb, DATA_FLAGS);
1787 sbi->s_mount_opt |= m->mount_opt;
1788 }
1789 #ifdef CONFIG_QUOTA
1790 } else if (m->flags & MOPT_QFMT) {
1791 if (sb_any_quota_loaded(sb) &&
1792 sbi->s_jquota_fmt != m->mount_opt) {
1793 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1794 "quota options when quota turned on");
1795 return -1;
1796 }
1797 if (ext4_has_feature_quota(sb)) {
1798 ext4_msg(sb, KERN_INFO,
1799 "Quota format mount options ignored "
1800 "when QUOTA feature is enabled");
1801 return 1;
1802 }
1803 sbi->s_jquota_fmt = m->mount_opt;
1804 #endif
1805 } else if (token == Opt_dax) {
1806 #ifdef CONFIG_FS_DAX
1807 ext4_msg(sb, KERN_WARNING,
1808 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1809 sbi->s_mount_opt |= m->mount_opt;
1810 #else
1811 ext4_msg(sb, KERN_INFO, "dax option not supported");
1812 return -1;
1813 #endif
1814 } else if (token == Opt_data_err_abort) {
1815 sbi->s_mount_opt |= m->mount_opt;
1816 } else if (token == Opt_data_err_ignore) {
1817 sbi->s_mount_opt &= ~m->mount_opt;
1818 } else {
1819 if (!args->from)
1820 arg = 1;
1821 if (m->flags & MOPT_CLEAR)
1822 arg = !arg;
1823 else if (unlikely(!(m->flags & MOPT_SET))) {
1824 ext4_msg(sb, KERN_WARNING,
1825 "buggy handling of option %s", opt);
1826 WARN_ON(1);
1827 return -1;
1828 }
1829 if (arg != 0)
1830 sbi->s_mount_opt |= m->mount_opt;
1831 else
1832 sbi->s_mount_opt &= ~m->mount_opt;
1833 }
1834 return 1;
1835 }
1836
1837 static int parse_options(char *options, struct super_block *sb,
1838 unsigned long *journal_devnum,
1839 unsigned int *journal_ioprio,
1840 int is_remount)
1841 {
1842 struct ext4_sb_info *sbi = EXT4_SB(sb);
1843 char *p;
1844 substring_t args[MAX_OPT_ARGS];
1845 int token;
1846
1847 if (!options)
1848 return 1;
1849
1850 while ((p = strsep(&options, ",")) != NULL) {
1851 if (!*p)
1852 continue;
1853 /*
1854 * Initialize args struct so we know whether arg was
1855 * found; some options take optional arguments.
1856 */
1857 args[0].to = args[0].from = NULL;
1858 token = match_token(p, tokens, args);
1859 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1860 journal_ioprio, is_remount) < 0)
1861 return 0;
1862 }
1863 #ifdef CONFIG_QUOTA
1864 /*
1865 * We do the test below only for project quotas. 'usrquota' and
1866 * 'grpquota' mount options are allowed even without quota feature
1867 * to support legacy quotas in quota files.
1868 */
1869 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1870 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1871 "Cannot enable project quota enforcement.");
1872 return 0;
1873 }
1874 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1875 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1876 clear_opt(sb, USRQUOTA);
1877
1878 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1879 clear_opt(sb, GRPQUOTA);
1880
1881 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1882 ext4_msg(sb, KERN_ERR, "old and new quota "
1883 "format mixing");
1884 return 0;
1885 }
1886
1887 if (!sbi->s_jquota_fmt) {
1888 ext4_msg(sb, KERN_ERR, "journaled quota format "
1889 "not specified");
1890 return 0;
1891 }
1892 }
1893 #endif
1894 if (test_opt(sb, DIOREAD_NOLOCK)) {
1895 int blocksize =
1896 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1897
1898 if (blocksize < PAGE_SIZE) {
1899 ext4_msg(sb, KERN_ERR, "can't mount with "
1900 "dioread_nolock if block size != PAGE_SIZE");
1901 return 0;
1902 }
1903 }
1904 return 1;
1905 }
1906
1907 static inline void ext4_show_quota_options(struct seq_file *seq,
1908 struct super_block *sb)
1909 {
1910 #if defined(CONFIG_QUOTA)
1911 struct ext4_sb_info *sbi = EXT4_SB(sb);
1912
1913 if (sbi->s_jquota_fmt) {
1914 char *fmtname = "";
1915
1916 switch (sbi->s_jquota_fmt) {
1917 case QFMT_VFS_OLD:
1918 fmtname = "vfsold";
1919 break;
1920 case QFMT_VFS_V0:
1921 fmtname = "vfsv0";
1922 break;
1923 case QFMT_VFS_V1:
1924 fmtname = "vfsv1";
1925 break;
1926 }
1927 seq_printf(seq, ",jqfmt=%s", fmtname);
1928 }
1929
1930 if (sbi->s_qf_names[USRQUOTA])
1931 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1932
1933 if (sbi->s_qf_names[GRPQUOTA])
1934 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1935 #endif
1936 }
1937
1938 static const char *token2str(int token)
1939 {
1940 const struct match_token *t;
1941
1942 for (t = tokens; t->token != Opt_err; t++)
1943 if (t->token == token && !strchr(t->pattern, '='))
1944 break;
1945 return t->pattern;
1946 }
1947
1948 /*
1949 * Show an option if
1950 * - it's set to a non-default value OR
1951 * - if the per-sb default is different from the global default
1952 */
1953 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1954 int nodefs)
1955 {
1956 struct ext4_sb_info *sbi = EXT4_SB(sb);
1957 struct ext4_super_block *es = sbi->s_es;
1958 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1959 const struct mount_opts *m;
1960 char sep = nodefs ? '\n' : ',';
1961
1962 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1963 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1964
1965 if (sbi->s_sb_block != 1)
1966 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1967
1968 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1969 int want_set = m->flags & MOPT_SET;
1970 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1971 (m->flags & MOPT_CLEAR_ERR))
1972 continue;
1973 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1974 continue; /* skip if same as the default */
1975 if ((want_set &&
1976 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1977 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1978 continue; /* select Opt_noFoo vs Opt_Foo */
1979 SEQ_OPTS_PRINT("%s", token2str(m->token));
1980 }
1981
1982 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1983 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1984 SEQ_OPTS_PRINT("resuid=%u",
1985 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1986 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1987 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1988 SEQ_OPTS_PRINT("resgid=%u",
1989 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1990 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1991 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1992 SEQ_OPTS_PUTS("errors=remount-ro");
1993 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1994 SEQ_OPTS_PUTS("errors=continue");
1995 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1996 SEQ_OPTS_PUTS("errors=panic");
1997 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1998 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1999 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2000 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2001 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2002 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2003 if (sb->s_flags & MS_I_VERSION)
2004 SEQ_OPTS_PUTS("i_version");
2005 if (nodefs || sbi->s_stripe)
2006 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2007 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2008 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2009 SEQ_OPTS_PUTS("data=journal");
2010 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2011 SEQ_OPTS_PUTS("data=ordered");
2012 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2013 SEQ_OPTS_PUTS("data=writeback");
2014 }
2015 if (nodefs ||
2016 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2017 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2018 sbi->s_inode_readahead_blks);
2019
2020 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2021 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2022 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2023 if (nodefs || sbi->s_max_dir_size_kb)
2024 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2025 if (test_opt(sb, DATA_ERR_ABORT))
2026 SEQ_OPTS_PUTS("data_err=abort");
2027
2028 ext4_show_quota_options(seq, sb);
2029 return 0;
2030 }
2031
2032 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2033 {
2034 return _ext4_show_options(seq, root->d_sb, 0);
2035 }
2036
2037 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2038 {
2039 struct super_block *sb = seq->private;
2040 int rc;
2041
2042 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2043 rc = _ext4_show_options(seq, sb, 1);
2044 seq_puts(seq, "\n");
2045 return rc;
2046 }
2047
2048 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2049 int read_only)
2050 {
2051 struct ext4_sb_info *sbi = EXT4_SB(sb);
2052 int res = 0;
2053
2054 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2055 ext4_msg(sb, KERN_ERR, "revision level too high, "
2056 "forcing read-only mode");
2057 res = MS_RDONLY;
2058 }
2059 if (read_only)
2060 goto done;
2061 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2062 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2063 "running e2fsck is recommended");
2064 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2065 ext4_msg(sb, KERN_WARNING,
2066 "warning: mounting fs with errors, "
2067 "running e2fsck is recommended");
2068 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2069 le16_to_cpu(es->s_mnt_count) >=
2070 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2071 ext4_msg(sb, KERN_WARNING,
2072 "warning: maximal mount count reached, "
2073 "running e2fsck is recommended");
2074 else if (le32_to_cpu(es->s_checkinterval) &&
2075 (le32_to_cpu(es->s_lastcheck) +
2076 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2077 ext4_msg(sb, KERN_WARNING,
2078 "warning: checktime reached, "
2079 "running e2fsck is recommended");
2080 if (!sbi->s_journal)
2081 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2082 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2083 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2084 le16_add_cpu(&es->s_mnt_count, 1);
2085 es->s_mtime = cpu_to_le32(get_seconds());
2086 ext4_update_dynamic_rev(sb);
2087 if (sbi->s_journal)
2088 ext4_set_feature_journal_needs_recovery(sb);
2089
2090 ext4_commit_super(sb, 1);
2091 done:
2092 if (test_opt(sb, DEBUG))
2093 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2094 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2095 sb->s_blocksize,
2096 sbi->s_groups_count,
2097 EXT4_BLOCKS_PER_GROUP(sb),
2098 EXT4_INODES_PER_GROUP(sb),
2099 sbi->s_mount_opt, sbi->s_mount_opt2);
2100
2101 cleancache_init_fs(sb);
2102 return res;
2103 }
2104
2105 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2106 {
2107 struct ext4_sb_info *sbi = EXT4_SB(sb);
2108 struct flex_groups *new_groups;
2109 int size;
2110
2111 if (!sbi->s_log_groups_per_flex)
2112 return 0;
2113
2114 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2115 if (size <= sbi->s_flex_groups_allocated)
2116 return 0;
2117
2118 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2119 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2120 if (!new_groups) {
2121 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2122 size / (int) sizeof(struct flex_groups));
2123 return -ENOMEM;
2124 }
2125
2126 if (sbi->s_flex_groups) {
2127 memcpy(new_groups, sbi->s_flex_groups,
2128 (sbi->s_flex_groups_allocated *
2129 sizeof(struct flex_groups)));
2130 kvfree(sbi->s_flex_groups);
2131 }
2132 sbi->s_flex_groups = new_groups;
2133 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2134 return 0;
2135 }
2136
2137 static int ext4_fill_flex_info(struct super_block *sb)
2138 {
2139 struct ext4_sb_info *sbi = EXT4_SB(sb);
2140 struct ext4_group_desc *gdp = NULL;
2141 ext4_group_t flex_group;
2142 int i, err;
2143
2144 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2145 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2146 sbi->s_log_groups_per_flex = 0;
2147 return 1;
2148 }
2149
2150 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2151 if (err)
2152 goto failed;
2153
2154 for (i = 0; i < sbi->s_groups_count; i++) {
2155 gdp = ext4_get_group_desc(sb, i, NULL);
2156
2157 flex_group = ext4_flex_group(sbi, i);
2158 atomic_add(ext4_free_inodes_count(sb, gdp),
2159 &sbi->s_flex_groups[flex_group].free_inodes);
2160 atomic64_add(ext4_free_group_clusters(sb, gdp),
2161 &sbi->s_flex_groups[flex_group].free_clusters);
2162 atomic_add(ext4_used_dirs_count(sb, gdp),
2163 &sbi->s_flex_groups[flex_group].used_dirs);
2164 }
2165
2166 return 1;
2167 failed:
2168 return 0;
2169 }
2170
2171 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2172 struct ext4_group_desc *gdp)
2173 {
2174 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2175 __u16 crc = 0;
2176 __le32 le_group = cpu_to_le32(block_group);
2177 struct ext4_sb_info *sbi = EXT4_SB(sb);
2178
2179 if (ext4_has_metadata_csum(sbi->s_sb)) {
2180 /* Use new metadata_csum algorithm */
2181 __u32 csum32;
2182 __u16 dummy_csum = 0;
2183
2184 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2185 sizeof(le_group));
2186 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2187 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2188 sizeof(dummy_csum));
2189 offset += sizeof(dummy_csum);
2190 if (offset < sbi->s_desc_size)
2191 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2192 sbi->s_desc_size - offset);
2193
2194 crc = csum32 & 0xFFFF;
2195 goto out;
2196 }
2197
2198 /* old crc16 code */
2199 if (!ext4_has_feature_gdt_csum(sb))
2200 return 0;
2201
2202 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2203 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2204 crc = crc16(crc, (__u8 *)gdp, offset);
2205 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2206 /* for checksum of struct ext4_group_desc do the rest...*/
2207 if (ext4_has_feature_64bit(sb) &&
2208 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2209 crc = crc16(crc, (__u8 *)gdp + offset,
2210 le16_to_cpu(sbi->s_es->s_desc_size) -
2211 offset);
2212
2213 out:
2214 return cpu_to_le16(crc);
2215 }
2216
2217 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2218 struct ext4_group_desc *gdp)
2219 {
2220 if (ext4_has_group_desc_csum(sb) &&
2221 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2222 return 0;
2223
2224 return 1;
2225 }
2226
2227 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2228 struct ext4_group_desc *gdp)
2229 {
2230 if (!ext4_has_group_desc_csum(sb))
2231 return;
2232 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2233 }
2234
2235 /* Called at mount-time, super-block is locked */
2236 static int ext4_check_descriptors(struct super_block *sb,
2237 ext4_fsblk_t sb_block,
2238 ext4_group_t *first_not_zeroed)
2239 {
2240 struct ext4_sb_info *sbi = EXT4_SB(sb);
2241 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2242 ext4_fsblk_t last_block;
2243 ext4_fsblk_t block_bitmap;
2244 ext4_fsblk_t inode_bitmap;
2245 ext4_fsblk_t inode_table;
2246 int flexbg_flag = 0;
2247 ext4_group_t i, grp = sbi->s_groups_count;
2248
2249 if (ext4_has_feature_flex_bg(sb))
2250 flexbg_flag = 1;
2251
2252 ext4_debug("Checking group descriptors");
2253
2254 for (i = 0; i < sbi->s_groups_count; i++) {
2255 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2256
2257 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2258 last_block = ext4_blocks_count(sbi->s_es) - 1;
2259 else
2260 last_block = first_block +
2261 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2262
2263 if ((grp == sbi->s_groups_count) &&
2264 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2265 grp = i;
2266
2267 block_bitmap = ext4_block_bitmap(sb, gdp);
2268 if (block_bitmap == sb_block) {
2269 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2270 "Block bitmap for group %u overlaps "
2271 "superblock", i);
2272 }
2273 if (block_bitmap < first_block || block_bitmap > last_block) {
2274 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2275 "Block bitmap for group %u not in group "
2276 "(block %llu)!", i, block_bitmap);
2277 return 0;
2278 }
2279 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2280 if (inode_bitmap == sb_block) {
2281 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2282 "Inode bitmap for group %u overlaps "
2283 "superblock", i);
2284 }
2285 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2286 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2287 "Inode bitmap for group %u not in group "
2288 "(block %llu)!", i, inode_bitmap);
2289 return 0;
2290 }
2291 inode_table = ext4_inode_table(sb, gdp);
2292 if (inode_table == sb_block) {
2293 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2294 "Inode table for group %u overlaps "
2295 "superblock", i);
2296 }
2297 if (inode_table < first_block ||
2298 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2299 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2300 "Inode table for group %u not in group "
2301 "(block %llu)!", i, inode_table);
2302 return 0;
2303 }
2304 ext4_lock_group(sb, i);
2305 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2306 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2307 "Checksum for group %u failed (%u!=%u)",
2308 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2309 gdp)), le16_to_cpu(gdp->bg_checksum));
2310 if (!(sb->s_flags & MS_RDONLY)) {
2311 ext4_unlock_group(sb, i);
2312 return 0;
2313 }
2314 }
2315 ext4_unlock_group(sb, i);
2316 if (!flexbg_flag)
2317 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2318 }
2319 if (NULL != first_not_zeroed)
2320 *first_not_zeroed = grp;
2321 return 1;
2322 }
2323
2324 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2325 * the superblock) which were deleted from all directories, but held open by
2326 * a process at the time of a crash. We walk the list and try to delete these
2327 * inodes at recovery time (only with a read-write filesystem).
2328 *
2329 * In order to keep the orphan inode chain consistent during traversal (in
2330 * case of crash during recovery), we link each inode into the superblock
2331 * orphan list_head and handle it the same way as an inode deletion during
2332 * normal operation (which journals the operations for us).
2333 *
2334 * We only do an iget() and an iput() on each inode, which is very safe if we
2335 * accidentally point at an in-use or already deleted inode. The worst that
2336 * can happen in this case is that we get a "bit already cleared" message from
2337 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2338 * e2fsck was run on this filesystem, and it must have already done the orphan
2339 * inode cleanup for us, so we can safely abort without any further action.
2340 */
2341 static void ext4_orphan_cleanup(struct super_block *sb,
2342 struct ext4_super_block *es)
2343 {
2344 unsigned int s_flags = sb->s_flags;
2345 int ret, nr_orphans = 0, nr_truncates = 0;
2346 #ifdef CONFIG_QUOTA
2347 int i;
2348 #endif
2349 if (!es->s_last_orphan) {
2350 jbd_debug(4, "no orphan inodes to clean up\n");
2351 return;
2352 }
2353
2354 if (bdev_read_only(sb->s_bdev)) {
2355 ext4_msg(sb, KERN_ERR, "write access "
2356 "unavailable, skipping orphan cleanup");
2357 return;
2358 }
2359
2360 /* Check if feature set would not allow a r/w mount */
2361 if (!ext4_feature_set_ok(sb, 0)) {
2362 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2363 "unknown ROCOMPAT features");
2364 return;
2365 }
2366
2367 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2368 /* don't clear list on RO mount w/ errors */
2369 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2370 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2371 "clearing orphan list.\n");
2372 es->s_last_orphan = 0;
2373 }
2374 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2375 return;
2376 }
2377
2378 if (s_flags & MS_RDONLY) {
2379 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2380 sb->s_flags &= ~MS_RDONLY;
2381 }
2382 #ifdef CONFIG_QUOTA
2383 /* Needed for iput() to work correctly and not trash data */
2384 sb->s_flags |= MS_ACTIVE;
2385 /* Turn on quotas so that they are updated correctly */
2386 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2387 if (EXT4_SB(sb)->s_qf_names[i]) {
2388 int ret = ext4_quota_on_mount(sb, i);
2389 if (ret < 0)
2390 ext4_msg(sb, KERN_ERR,
2391 "Cannot turn on journaled "
2392 "quota: error %d", ret);
2393 }
2394 }
2395 #endif
2396
2397 while (es->s_last_orphan) {
2398 struct inode *inode;
2399
2400 /*
2401 * We may have encountered an error during cleanup; if
2402 * so, skip the rest.
2403 */
2404 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2405 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2406 es->s_last_orphan = 0;
2407 break;
2408 }
2409
2410 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2411 if (IS_ERR(inode)) {
2412 es->s_last_orphan = 0;
2413 break;
2414 }
2415
2416 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2417 dquot_initialize(inode);
2418 if (inode->i_nlink) {
2419 if (test_opt(sb, DEBUG))
2420 ext4_msg(sb, KERN_DEBUG,
2421 "%s: truncating inode %lu to %lld bytes",
2422 __func__, inode->i_ino, inode->i_size);
2423 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2424 inode->i_ino, inode->i_size);
2425 inode_lock(inode);
2426 truncate_inode_pages(inode->i_mapping, inode->i_size);
2427 ret = ext4_truncate(inode);
2428 if (ret)
2429 ext4_std_error(inode->i_sb, ret);
2430 inode_unlock(inode);
2431 nr_truncates++;
2432 } else {
2433 if (test_opt(sb, DEBUG))
2434 ext4_msg(sb, KERN_DEBUG,
2435 "%s: deleting unreferenced inode %lu",
2436 __func__, inode->i_ino);
2437 jbd_debug(2, "deleting unreferenced inode %lu\n",
2438 inode->i_ino);
2439 nr_orphans++;
2440 }
2441 iput(inode); /* The delete magic happens here! */
2442 }
2443
2444 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2445
2446 if (nr_orphans)
2447 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2448 PLURAL(nr_orphans));
2449 if (nr_truncates)
2450 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2451 PLURAL(nr_truncates));
2452 #ifdef CONFIG_QUOTA
2453 /* Turn quotas off */
2454 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2455 if (sb_dqopt(sb)->files[i])
2456 dquot_quota_off(sb, i);
2457 }
2458 #endif
2459 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2460 }
2461
2462 /*
2463 * Maximal extent format file size.
2464 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2465 * extent format containers, within a sector_t, and within i_blocks
2466 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2467 * so that won't be a limiting factor.
2468 *
2469 * However there is other limiting factor. We do store extents in the form
2470 * of starting block and length, hence the resulting length of the extent
2471 * covering maximum file size must fit into on-disk format containers as
2472 * well. Given that length is always by 1 unit bigger than max unit (because
2473 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2474 *
2475 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2476 */
2477 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2478 {
2479 loff_t res;
2480 loff_t upper_limit = MAX_LFS_FILESIZE;
2481
2482 /* small i_blocks in vfs inode? */
2483 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2484 /*
2485 * CONFIG_LBDAF is not enabled implies the inode
2486 * i_block represent total blocks in 512 bytes
2487 * 32 == size of vfs inode i_blocks * 8
2488 */
2489 upper_limit = (1LL << 32) - 1;
2490
2491 /* total blocks in file system block size */
2492 upper_limit >>= (blkbits - 9);
2493 upper_limit <<= blkbits;
2494 }
2495
2496 /*
2497 * 32-bit extent-start container, ee_block. We lower the maxbytes
2498 * by one fs block, so ee_len can cover the extent of maximum file
2499 * size
2500 */
2501 res = (1LL << 32) - 1;
2502 res <<= blkbits;
2503
2504 /* Sanity check against vm- & vfs- imposed limits */
2505 if (res > upper_limit)
2506 res = upper_limit;
2507
2508 return res;
2509 }
2510
2511 /*
2512 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2513 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2514 * We need to be 1 filesystem block less than the 2^48 sector limit.
2515 */
2516 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2517 {
2518 loff_t res = EXT4_NDIR_BLOCKS;
2519 int meta_blocks;
2520 loff_t upper_limit;
2521 /* This is calculated to be the largest file size for a dense, block
2522 * mapped file such that the file's total number of 512-byte sectors,
2523 * including data and all indirect blocks, does not exceed (2^48 - 1).
2524 *
2525 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2526 * number of 512-byte sectors of the file.
2527 */
2528
2529 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2530 /*
2531 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2532 * the inode i_block field represents total file blocks in
2533 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2534 */
2535 upper_limit = (1LL << 32) - 1;
2536
2537 /* total blocks in file system block size */
2538 upper_limit >>= (bits - 9);
2539
2540 } else {
2541 /*
2542 * We use 48 bit ext4_inode i_blocks
2543 * With EXT4_HUGE_FILE_FL set the i_blocks
2544 * represent total number of blocks in
2545 * file system block size
2546 */
2547 upper_limit = (1LL << 48) - 1;
2548
2549 }
2550
2551 /* indirect blocks */
2552 meta_blocks = 1;
2553 /* double indirect blocks */
2554 meta_blocks += 1 + (1LL << (bits-2));
2555 /* tripple indirect blocks */
2556 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2557
2558 upper_limit -= meta_blocks;
2559 upper_limit <<= bits;
2560
2561 res += 1LL << (bits-2);
2562 res += 1LL << (2*(bits-2));
2563 res += 1LL << (3*(bits-2));
2564 res <<= bits;
2565 if (res > upper_limit)
2566 res = upper_limit;
2567
2568 if (res > MAX_LFS_FILESIZE)
2569 res = MAX_LFS_FILESIZE;
2570
2571 return res;
2572 }
2573
2574 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2575 ext4_fsblk_t logical_sb_block, int nr)
2576 {
2577 struct ext4_sb_info *sbi = EXT4_SB(sb);
2578 ext4_group_t bg, first_meta_bg;
2579 int has_super = 0;
2580
2581 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2582
2583 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2584 return logical_sb_block + nr + 1;
2585 bg = sbi->s_desc_per_block * nr;
2586 if (ext4_bg_has_super(sb, bg))
2587 has_super = 1;
2588
2589 /*
2590 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2591 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2592 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2593 * compensate.
2594 */
2595 if (sb->s_blocksize == 1024 && nr == 0 &&
2596 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2597 has_super++;
2598
2599 return (has_super + ext4_group_first_block_no(sb, bg));
2600 }
2601
2602 /**
2603 * ext4_get_stripe_size: Get the stripe size.
2604 * @sbi: In memory super block info
2605 *
2606 * If we have specified it via mount option, then
2607 * use the mount option value. If the value specified at mount time is
2608 * greater than the blocks per group use the super block value.
2609 * If the super block value is greater than blocks per group return 0.
2610 * Allocator needs it be less than blocks per group.
2611 *
2612 */
2613 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2614 {
2615 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2616 unsigned long stripe_width =
2617 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2618 int ret;
2619
2620 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2621 ret = sbi->s_stripe;
2622 else if (stripe_width <= sbi->s_blocks_per_group)
2623 ret = stripe_width;
2624 else if (stride <= sbi->s_blocks_per_group)
2625 ret = stride;
2626 else
2627 ret = 0;
2628
2629 /*
2630 * If the stripe width is 1, this makes no sense and
2631 * we set it to 0 to turn off stripe handling code.
2632 */
2633 if (ret <= 1)
2634 ret = 0;
2635
2636 return ret;
2637 }
2638
2639 /*
2640 * Check whether this filesystem can be mounted based on
2641 * the features present and the RDONLY/RDWR mount requested.
2642 * Returns 1 if this filesystem can be mounted as requested,
2643 * 0 if it cannot be.
2644 */
2645 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2646 {
2647 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2648 ext4_msg(sb, KERN_ERR,
2649 "Couldn't mount because of "
2650 "unsupported optional features (%x)",
2651 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2652 ~EXT4_FEATURE_INCOMPAT_SUPP));
2653 return 0;
2654 }
2655
2656 if (readonly)
2657 return 1;
2658
2659 if (ext4_has_feature_readonly(sb)) {
2660 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2661 sb->s_flags |= MS_RDONLY;
2662 return 1;
2663 }
2664
2665 /* Check that feature set is OK for a read-write mount */
2666 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2667 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2668 "unsupported optional features (%x)",
2669 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2670 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2671 return 0;
2672 }
2673 /*
2674 * Large file size enabled file system can only be mounted
2675 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2676 */
2677 if (ext4_has_feature_huge_file(sb)) {
2678 if (sizeof(blkcnt_t) < sizeof(u64)) {
2679 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2680 "cannot be mounted RDWR without "
2681 "CONFIG_LBDAF");
2682 return 0;
2683 }
2684 }
2685 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2686 ext4_msg(sb, KERN_ERR,
2687 "Can't support bigalloc feature without "
2688 "extents feature\n");
2689 return 0;
2690 }
2691
2692 #ifndef CONFIG_QUOTA
2693 if (ext4_has_feature_quota(sb) && !readonly) {
2694 ext4_msg(sb, KERN_ERR,
2695 "Filesystem with quota feature cannot be mounted RDWR "
2696 "without CONFIG_QUOTA");
2697 return 0;
2698 }
2699 if (ext4_has_feature_project(sb) && !readonly) {
2700 ext4_msg(sb, KERN_ERR,
2701 "Filesystem with project quota feature cannot be mounted RDWR "
2702 "without CONFIG_QUOTA");
2703 return 0;
2704 }
2705 #endif /* CONFIG_QUOTA */
2706 return 1;
2707 }
2708
2709 /*
2710 * This function is called once a day if we have errors logged
2711 * on the file system
2712 */
2713 static void print_daily_error_info(unsigned long arg)
2714 {
2715 struct super_block *sb = (struct super_block *) arg;
2716 struct ext4_sb_info *sbi;
2717 struct ext4_super_block *es;
2718
2719 sbi = EXT4_SB(sb);
2720 es = sbi->s_es;
2721
2722 if (es->s_error_count)
2723 /* fsck newer than v1.41.13 is needed to clean this condition. */
2724 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2725 le32_to_cpu(es->s_error_count));
2726 if (es->s_first_error_time) {
2727 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2728 sb->s_id, le32_to_cpu(es->s_first_error_time),
2729 (int) sizeof(es->s_first_error_func),
2730 es->s_first_error_func,
2731 le32_to_cpu(es->s_first_error_line));
2732 if (es->s_first_error_ino)
2733 printk(KERN_CONT ": inode %u",
2734 le32_to_cpu(es->s_first_error_ino));
2735 if (es->s_first_error_block)
2736 printk(KERN_CONT ": block %llu", (unsigned long long)
2737 le64_to_cpu(es->s_first_error_block));
2738 printk(KERN_CONT "\n");
2739 }
2740 if (es->s_last_error_time) {
2741 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2742 sb->s_id, le32_to_cpu(es->s_last_error_time),
2743 (int) sizeof(es->s_last_error_func),
2744 es->s_last_error_func,
2745 le32_to_cpu(es->s_last_error_line));
2746 if (es->s_last_error_ino)
2747 printk(KERN_CONT ": inode %u",
2748 le32_to_cpu(es->s_last_error_ino));
2749 if (es->s_last_error_block)
2750 printk(KERN_CONT ": block %llu", (unsigned long long)
2751 le64_to_cpu(es->s_last_error_block));
2752 printk(KERN_CONT "\n");
2753 }
2754 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2755 }
2756
2757 /* Find next suitable group and run ext4_init_inode_table */
2758 static int ext4_run_li_request(struct ext4_li_request *elr)
2759 {
2760 struct ext4_group_desc *gdp = NULL;
2761 ext4_group_t group, ngroups;
2762 struct super_block *sb;
2763 unsigned long timeout = 0;
2764 int ret = 0;
2765
2766 sb = elr->lr_super;
2767 ngroups = EXT4_SB(sb)->s_groups_count;
2768
2769 for (group = elr->lr_next_group; group < ngroups; group++) {
2770 gdp = ext4_get_group_desc(sb, group, NULL);
2771 if (!gdp) {
2772 ret = 1;
2773 break;
2774 }
2775
2776 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2777 break;
2778 }
2779
2780 if (group >= ngroups)
2781 ret = 1;
2782
2783 if (!ret) {
2784 timeout = jiffies;
2785 ret = ext4_init_inode_table(sb, group,
2786 elr->lr_timeout ? 0 : 1);
2787 if (elr->lr_timeout == 0) {
2788 timeout = (jiffies - timeout) *
2789 elr->lr_sbi->s_li_wait_mult;
2790 elr->lr_timeout = timeout;
2791 }
2792 elr->lr_next_sched = jiffies + elr->lr_timeout;
2793 elr->lr_next_group = group + 1;
2794 }
2795 return ret;
2796 }
2797
2798 /*
2799 * Remove lr_request from the list_request and free the
2800 * request structure. Should be called with li_list_mtx held
2801 */
2802 static void ext4_remove_li_request(struct ext4_li_request *elr)
2803 {
2804 struct ext4_sb_info *sbi;
2805
2806 if (!elr)
2807 return;
2808
2809 sbi = elr->lr_sbi;
2810
2811 list_del(&elr->lr_request);
2812 sbi->s_li_request = NULL;
2813 kfree(elr);
2814 }
2815
2816 static void ext4_unregister_li_request(struct super_block *sb)
2817 {
2818 mutex_lock(&ext4_li_mtx);
2819 if (!ext4_li_info) {
2820 mutex_unlock(&ext4_li_mtx);
2821 return;
2822 }
2823
2824 mutex_lock(&ext4_li_info->li_list_mtx);
2825 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2826 mutex_unlock(&ext4_li_info->li_list_mtx);
2827 mutex_unlock(&ext4_li_mtx);
2828 }
2829
2830 static struct task_struct *ext4_lazyinit_task;
2831
2832 /*
2833 * This is the function where ext4lazyinit thread lives. It walks
2834 * through the request list searching for next scheduled filesystem.
2835 * When such a fs is found, run the lazy initialization request
2836 * (ext4_rn_li_request) and keep track of the time spend in this
2837 * function. Based on that time we compute next schedule time of
2838 * the request. When walking through the list is complete, compute
2839 * next waking time and put itself into sleep.
2840 */
2841 static int ext4_lazyinit_thread(void *arg)
2842 {
2843 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2844 struct list_head *pos, *n;
2845 struct ext4_li_request *elr;
2846 unsigned long next_wakeup, cur;
2847
2848 BUG_ON(NULL == eli);
2849
2850 cont_thread:
2851 while (true) {
2852 next_wakeup = MAX_JIFFY_OFFSET;
2853
2854 mutex_lock(&eli->li_list_mtx);
2855 if (list_empty(&eli->li_request_list)) {
2856 mutex_unlock(&eli->li_list_mtx);
2857 goto exit_thread;
2858 }
2859 list_for_each_safe(pos, n, &eli->li_request_list) {
2860 int err = 0;
2861 int progress = 0;
2862 elr = list_entry(pos, struct ext4_li_request,
2863 lr_request);
2864
2865 if (time_before(jiffies, elr->lr_next_sched)) {
2866 if (time_before(elr->lr_next_sched, next_wakeup))
2867 next_wakeup = elr->lr_next_sched;
2868 continue;
2869 }
2870 if (down_read_trylock(&elr->lr_super->s_umount)) {
2871 if (sb_start_write_trylock(elr->lr_super)) {
2872 progress = 1;
2873 /*
2874 * We hold sb->s_umount, sb can not
2875 * be removed from the list, it is
2876 * now safe to drop li_list_mtx
2877 */
2878 mutex_unlock(&eli->li_list_mtx);
2879 err = ext4_run_li_request(elr);
2880 sb_end_write(elr->lr_super);
2881 mutex_lock(&eli->li_list_mtx);
2882 n = pos->next;
2883 }
2884 up_read((&elr->lr_super->s_umount));
2885 }
2886 /* error, remove the lazy_init job */
2887 if (err) {
2888 ext4_remove_li_request(elr);
2889 continue;
2890 }
2891 if (!progress) {
2892 elr->lr_next_sched = jiffies +
2893 (prandom_u32()
2894 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2895 }
2896 if (time_before(elr->lr_next_sched, next_wakeup))
2897 next_wakeup = elr->lr_next_sched;
2898 }
2899 mutex_unlock(&eli->li_list_mtx);
2900
2901 try_to_freeze();
2902
2903 cur = jiffies;
2904 if ((time_after_eq(cur, next_wakeup)) ||
2905 (MAX_JIFFY_OFFSET == next_wakeup)) {
2906 cond_resched();
2907 continue;
2908 }
2909
2910 schedule_timeout_interruptible(next_wakeup - cur);
2911
2912 if (kthread_should_stop()) {
2913 ext4_clear_request_list();
2914 goto exit_thread;
2915 }
2916 }
2917
2918 exit_thread:
2919 /*
2920 * It looks like the request list is empty, but we need
2921 * to check it under the li_list_mtx lock, to prevent any
2922 * additions into it, and of course we should lock ext4_li_mtx
2923 * to atomically free the list and ext4_li_info, because at
2924 * this point another ext4 filesystem could be registering
2925 * new one.
2926 */
2927 mutex_lock(&ext4_li_mtx);
2928 mutex_lock(&eli->li_list_mtx);
2929 if (!list_empty(&eli->li_request_list)) {
2930 mutex_unlock(&eli->li_list_mtx);
2931 mutex_unlock(&ext4_li_mtx);
2932 goto cont_thread;
2933 }
2934 mutex_unlock(&eli->li_list_mtx);
2935 kfree(ext4_li_info);
2936 ext4_li_info = NULL;
2937 mutex_unlock(&ext4_li_mtx);
2938
2939 return 0;
2940 }
2941
2942 static void ext4_clear_request_list(void)
2943 {
2944 struct list_head *pos, *n;
2945 struct ext4_li_request *elr;
2946
2947 mutex_lock(&ext4_li_info->li_list_mtx);
2948 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2949 elr = list_entry(pos, struct ext4_li_request,
2950 lr_request);
2951 ext4_remove_li_request(elr);
2952 }
2953 mutex_unlock(&ext4_li_info->li_list_mtx);
2954 }
2955
2956 static int ext4_run_lazyinit_thread(void)
2957 {
2958 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2959 ext4_li_info, "ext4lazyinit");
2960 if (IS_ERR(ext4_lazyinit_task)) {
2961 int err = PTR_ERR(ext4_lazyinit_task);
2962 ext4_clear_request_list();
2963 kfree(ext4_li_info);
2964 ext4_li_info = NULL;
2965 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2966 "initialization thread\n",
2967 err);
2968 return err;
2969 }
2970 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2971 return 0;
2972 }
2973
2974 /*
2975 * Check whether it make sense to run itable init. thread or not.
2976 * If there is at least one uninitialized inode table, return
2977 * corresponding group number, else the loop goes through all
2978 * groups and return total number of groups.
2979 */
2980 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2981 {
2982 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2983 struct ext4_group_desc *gdp = NULL;
2984
2985 for (group = 0; group < ngroups; group++) {
2986 gdp = ext4_get_group_desc(sb, group, NULL);
2987 if (!gdp)
2988 continue;
2989
2990 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2991 break;
2992 }
2993
2994 return group;
2995 }
2996
2997 static int ext4_li_info_new(void)
2998 {
2999 struct ext4_lazy_init *eli = NULL;
3000
3001 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3002 if (!eli)
3003 return -ENOMEM;
3004
3005 INIT_LIST_HEAD(&eli->li_request_list);
3006 mutex_init(&eli->li_list_mtx);
3007
3008 eli->li_state |= EXT4_LAZYINIT_QUIT;
3009
3010 ext4_li_info = eli;
3011
3012 return 0;
3013 }
3014
3015 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3016 ext4_group_t start)
3017 {
3018 struct ext4_sb_info *sbi = EXT4_SB(sb);
3019 struct ext4_li_request *elr;
3020
3021 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3022 if (!elr)
3023 return NULL;
3024
3025 elr->lr_super = sb;
3026 elr->lr_sbi = sbi;
3027 elr->lr_next_group = start;
3028
3029 /*
3030 * Randomize first schedule time of the request to
3031 * spread the inode table initialization requests
3032 * better.
3033 */
3034 elr->lr_next_sched = jiffies + (prandom_u32() %
3035 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3036 return elr;
3037 }
3038
3039 int ext4_register_li_request(struct super_block *sb,
3040 ext4_group_t first_not_zeroed)
3041 {
3042 struct ext4_sb_info *sbi = EXT4_SB(sb);
3043 struct ext4_li_request *elr = NULL;
3044 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3045 int ret = 0;
3046
3047 mutex_lock(&ext4_li_mtx);
3048 if (sbi->s_li_request != NULL) {
3049 /*
3050 * Reset timeout so it can be computed again, because
3051 * s_li_wait_mult might have changed.
3052 */
3053 sbi->s_li_request->lr_timeout = 0;
3054 goto out;
3055 }
3056
3057 if (first_not_zeroed == ngroups ||
3058 (sb->s_flags & MS_RDONLY) ||
3059 !test_opt(sb, INIT_INODE_TABLE))
3060 goto out;
3061
3062 elr = ext4_li_request_new(sb, first_not_zeroed);
3063 if (!elr) {
3064 ret = -ENOMEM;
3065 goto out;
3066 }
3067
3068 if (NULL == ext4_li_info) {
3069 ret = ext4_li_info_new();
3070 if (ret)
3071 goto out;
3072 }
3073
3074 mutex_lock(&ext4_li_info->li_list_mtx);
3075 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3076 mutex_unlock(&ext4_li_info->li_list_mtx);
3077
3078 sbi->s_li_request = elr;
3079 /*
3080 * set elr to NULL here since it has been inserted to
3081 * the request_list and the removal and free of it is
3082 * handled by ext4_clear_request_list from now on.
3083 */
3084 elr = NULL;
3085
3086 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3087 ret = ext4_run_lazyinit_thread();
3088 if (ret)
3089 goto out;
3090 }
3091 out:
3092 mutex_unlock(&ext4_li_mtx);
3093 if (ret)
3094 kfree(elr);
3095 return ret;
3096 }
3097
3098 /*
3099 * We do not need to lock anything since this is called on
3100 * module unload.
3101 */
3102 static void ext4_destroy_lazyinit_thread(void)
3103 {
3104 /*
3105 * If thread exited earlier
3106 * there's nothing to be done.
3107 */
3108 if (!ext4_li_info || !ext4_lazyinit_task)
3109 return;
3110
3111 kthread_stop(ext4_lazyinit_task);
3112 }
3113
3114 static int set_journal_csum_feature_set(struct super_block *sb)
3115 {
3116 int ret = 1;
3117 int compat, incompat;
3118 struct ext4_sb_info *sbi = EXT4_SB(sb);
3119
3120 if (ext4_has_metadata_csum(sb)) {
3121 /* journal checksum v3 */
3122 compat = 0;
3123 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3124 } else {
3125 /* journal checksum v1 */
3126 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3127 incompat = 0;
3128 }
3129
3130 jbd2_journal_clear_features(sbi->s_journal,
3131 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3132 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3133 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3134 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3135 ret = jbd2_journal_set_features(sbi->s_journal,
3136 compat, 0,
3137 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3138 incompat);
3139 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3140 ret = jbd2_journal_set_features(sbi->s_journal,
3141 compat, 0,
3142 incompat);
3143 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3144 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3145 } else {
3146 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3147 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3148 }
3149
3150 return ret;
3151 }
3152
3153 /*
3154 * Note: calculating the overhead so we can be compatible with
3155 * historical BSD practice is quite difficult in the face of
3156 * clusters/bigalloc. This is because multiple metadata blocks from
3157 * different block group can end up in the same allocation cluster.
3158 * Calculating the exact overhead in the face of clustered allocation
3159 * requires either O(all block bitmaps) in memory or O(number of block
3160 * groups**2) in time. We will still calculate the superblock for
3161 * older file systems --- and if we come across with a bigalloc file
3162 * system with zero in s_overhead_clusters the estimate will be close to
3163 * correct especially for very large cluster sizes --- but for newer
3164 * file systems, it's better to calculate this figure once at mkfs
3165 * time, and store it in the superblock. If the superblock value is
3166 * present (even for non-bigalloc file systems), we will use it.
3167 */
3168 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3169 char *buf)
3170 {
3171 struct ext4_sb_info *sbi = EXT4_SB(sb);
3172 struct ext4_group_desc *gdp;
3173 ext4_fsblk_t first_block, last_block, b;
3174 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3175 int s, j, count = 0;
3176
3177 if (!ext4_has_feature_bigalloc(sb))
3178 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3179 sbi->s_itb_per_group + 2);
3180
3181 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3182 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3183 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3184 for (i = 0; i < ngroups; i++) {
3185 gdp = ext4_get_group_desc(sb, i, NULL);
3186 b = ext4_block_bitmap(sb, gdp);
3187 if (b >= first_block && b <= last_block) {
3188 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3189 count++;
3190 }
3191 b = ext4_inode_bitmap(sb, gdp);
3192 if (b >= first_block && b <= last_block) {
3193 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3194 count++;
3195 }
3196 b = ext4_inode_table(sb, gdp);
3197 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3198 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3199 int c = EXT4_B2C(sbi, b - first_block);
3200 ext4_set_bit(c, buf);
3201 count++;
3202 }
3203 if (i != grp)
3204 continue;
3205 s = 0;
3206 if (ext4_bg_has_super(sb, grp)) {
3207 ext4_set_bit(s++, buf);
3208 count++;
3209 }
3210 j = ext4_bg_num_gdb(sb, grp);
3211 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3212 ext4_error(sb, "Invalid number of block group "
3213 "descriptor blocks: %d", j);
3214 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3215 }
3216 count += j;
3217 for (; j > 0; j--)
3218 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3219 }
3220 if (!count)
3221 return 0;
3222 return EXT4_CLUSTERS_PER_GROUP(sb) -
3223 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3224 }
3225
3226 /*
3227 * Compute the overhead and stash it in sbi->s_overhead
3228 */
3229 int ext4_calculate_overhead(struct super_block *sb)
3230 {
3231 struct ext4_sb_info *sbi = EXT4_SB(sb);
3232 struct ext4_super_block *es = sbi->s_es;
3233 struct inode *j_inode;
3234 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3235 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3236 ext4_fsblk_t overhead = 0;
3237 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3238
3239 if (!buf)
3240 return -ENOMEM;
3241
3242 /*
3243 * Compute the overhead (FS structures). This is constant
3244 * for a given filesystem unless the number of block groups
3245 * changes so we cache the previous value until it does.
3246 */
3247
3248 /*
3249 * All of the blocks before first_data_block are overhead
3250 */
3251 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3252
3253 /*
3254 * Add the overhead found in each block group
3255 */
3256 for (i = 0; i < ngroups; i++) {
3257 int blks;
3258
3259 blks = count_overhead(sb, i, buf);
3260 overhead += blks;
3261 if (blks)
3262 memset(buf, 0, PAGE_SIZE);
3263 cond_resched();
3264 }
3265
3266 /*
3267 * Add the internal journal blocks whether the journal has been
3268 * loaded or not
3269 */
3270 if (sbi->s_journal && !sbi->journal_bdev)
3271 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3272 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3273 j_inode = ext4_get_journal_inode(sb, j_inum);
3274 if (j_inode) {
3275 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3276 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3277 iput(j_inode);
3278 } else {
3279 ext4_msg(sb, KERN_ERR, "can't get journal size");
3280 }
3281 }
3282 sbi->s_overhead = overhead;
3283 smp_wmb();
3284 free_page((unsigned long) buf);
3285 return 0;
3286 }
3287
3288 static void ext4_set_resv_clusters(struct super_block *sb)
3289 {
3290 ext4_fsblk_t resv_clusters;
3291 struct ext4_sb_info *sbi = EXT4_SB(sb);
3292
3293 /*
3294 * There's no need to reserve anything when we aren't using extents.
3295 * The space estimates are exact, there are no unwritten extents,
3296 * hole punching doesn't need new metadata... This is needed especially
3297 * to keep ext2/3 backward compatibility.
3298 */
3299 if (!ext4_has_feature_extents(sb))
3300 return;
3301 /*
3302 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3303 * This should cover the situations where we can not afford to run
3304 * out of space like for example punch hole, or converting
3305 * unwritten extents in delalloc path. In most cases such
3306 * allocation would require 1, or 2 blocks, higher numbers are
3307 * very rare.
3308 */
3309 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3310 sbi->s_cluster_bits);
3311
3312 do_div(resv_clusters, 50);
3313 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3314
3315 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3316 }
3317
3318 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3319 {
3320 char *orig_data = kstrdup(data, GFP_KERNEL);
3321 struct buffer_head *bh;
3322 struct ext4_super_block *es = NULL;
3323 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3324 ext4_fsblk_t block;
3325 ext4_fsblk_t sb_block = get_sb_block(&data);
3326 ext4_fsblk_t logical_sb_block;
3327 unsigned long offset = 0;
3328 unsigned long journal_devnum = 0;
3329 unsigned long def_mount_opts;
3330 struct inode *root;
3331 const char *descr;
3332 int ret = -ENOMEM;
3333 int blocksize, clustersize;
3334 unsigned int db_count;
3335 unsigned int i;
3336 int needs_recovery, has_huge_files, has_bigalloc;
3337 __u64 blocks_count;
3338 int err = 0;
3339 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3340 ext4_group_t first_not_zeroed;
3341
3342 if ((data && !orig_data) || !sbi)
3343 goto out_free_base;
3344
3345 sbi->s_blockgroup_lock =
3346 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3347 if (!sbi->s_blockgroup_lock)
3348 goto out_free_base;
3349
3350 sb->s_fs_info = sbi;
3351 sbi->s_sb = sb;
3352 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3353 sbi->s_sb_block = sb_block;
3354 if (sb->s_bdev->bd_part)
3355 sbi->s_sectors_written_start =
3356 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3357
3358 /* Cleanup superblock name */
3359 strreplace(sb->s_id, '/', '!');
3360
3361 /* -EINVAL is default */
3362 ret = -EINVAL;
3363 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3364 if (!blocksize) {
3365 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3366 goto out_fail;
3367 }
3368
3369 /*
3370 * The ext4 superblock will not be buffer aligned for other than 1kB
3371 * block sizes. We need to calculate the offset from buffer start.
3372 */
3373 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3374 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3375 offset = do_div(logical_sb_block, blocksize);
3376 } else {
3377 logical_sb_block = sb_block;
3378 }
3379
3380 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3381 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3382 goto out_fail;
3383 }
3384 /*
3385 * Note: s_es must be initialized as soon as possible because
3386 * some ext4 macro-instructions depend on its value
3387 */
3388 es = (struct ext4_super_block *) (bh->b_data + offset);
3389 sbi->s_es = es;
3390 sb->s_magic = le16_to_cpu(es->s_magic);
3391 if (sb->s_magic != EXT4_SUPER_MAGIC)
3392 goto cantfind_ext4;
3393 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3394
3395 /* Warn if metadata_csum and gdt_csum are both set. */
3396 if (ext4_has_feature_metadata_csum(sb) &&
3397 ext4_has_feature_gdt_csum(sb))
3398 ext4_warning(sb, "metadata_csum and uninit_bg are "
3399 "redundant flags; please run fsck.");
3400
3401 /* Check for a known checksum algorithm */
3402 if (!ext4_verify_csum_type(sb, es)) {
3403 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3404 "unknown checksum algorithm.");
3405 silent = 1;
3406 goto cantfind_ext4;
3407 }
3408
3409 /* Load the checksum driver */
3410 if (ext4_has_feature_metadata_csum(sb)) {
3411 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3412 if (IS_ERR(sbi->s_chksum_driver)) {
3413 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3414 ret = PTR_ERR(sbi->s_chksum_driver);
3415 sbi->s_chksum_driver = NULL;
3416 goto failed_mount;
3417 }
3418 }
3419
3420 /* Check superblock checksum */
3421 if (!ext4_superblock_csum_verify(sb, es)) {
3422 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3423 "invalid superblock checksum. Run e2fsck?");
3424 silent = 1;
3425 ret = -EFSBADCRC;
3426 goto cantfind_ext4;
3427 }
3428
3429 /* Precompute checksum seed for all metadata */
3430 if (ext4_has_feature_csum_seed(sb))
3431 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3432 else if (ext4_has_metadata_csum(sb))
3433 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3434 sizeof(es->s_uuid));
3435
3436 /* Set defaults before we parse the mount options */
3437 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3438 set_opt(sb, INIT_INODE_TABLE);
3439 if (def_mount_opts & EXT4_DEFM_DEBUG)
3440 set_opt(sb, DEBUG);
3441 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3442 set_opt(sb, GRPID);
3443 if (def_mount_opts & EXT4_DEFM_UID16)
3444 set_opt(sb, NO_UID32);
3445 /* xattr user namespace & acls are now defaulted on */
3446 set_opt(sb, XATTR_USER);
3447 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3448 set_opt(sb, POSIX_ACL);
3449 #endif
3450 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3451 if (ext4_has_metadata_csum(sb))
3452 set_opt(sb, JOURNAL_CHECKSUM);
3453
3454 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3455 set_opt(sb, JOURNAL_DATA);
3456 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3457 set_opt(sb, ORDERED_DATA);
3458 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3459 set_opt(sb, WRITEBACK_DATA);
3460
3461 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3462 set_opt(sb, ERRORS_PANIC);
3463 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3464 set_opt(sb, ERRORS_CONT);
3465 else
3466 set_opt(sb, ERRORS_RO);
3467 /* block_validity enabled by default; disable with noblock_validity */
3468 set_opt(sb, BLOCK_VALIDITY);
3469 if (def_mount_opts & EXT4_DEFM_DISCARD)
3470 set_opt(sb, DISCARD);
3471
3472 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3473 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3474 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3475 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3476 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3477
3478 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3479 set_opt(sb, BARRIER);
3480
3481 /*
3482 * enable delayed allocation by default
3483 * Use -o nodelalloc to turn it off
3484 */
3485 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3486 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3487 set_opt(sb, DELALLOC);
3488
3489 /*
3490 * set default s_li_wait_mult for lazyinit, for the case there is
3491 * no mount option specified.
3492 */
3493 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3494
3495 if (sbi->s_es->s_mount_opts[0]) {
3496 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3497 sizeof(sbi->s_es->s_mount_opts),
3498 GFP_KERNEL);
3499 if (!s_mount_opts)
3500 goto failed_mount;
3501 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3502 &journal_ioprio, 0)) {
3503 ext4_msg(sb, KERN_WARNING,
3504 "failed to parse options in superblock: %s",
3505 s_mount_opts);
3506 }
3507 kfree(s_mount_opts);
3508 }
3509 sbi->s_def_mount_opt = sbi->s_mount_opt;
3510 if (!parse_options((char *) data, sb, &journal_devnum,
3511 &journal_ioprio, 0))
3512 goto failed_mount;
3513
3514 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3515 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3516 "with data=journal disables delayed "
3517 "allocation and O_DIRECT support!\n");
3518 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3519 ext4_msg(sb, KERN_ERR, "can't mount with "
3520 "both data=journal and delalloc");
3521 goto failed_mount;
3522 }
3523 if (test_opt(sb, DIOREAD_NOLOCK)) {
3524 ext4_msg(sb, KERN_ERR, "can't mount with "
3525 "both data=journal and dioread_nolock");
3526 goto failed_mount;
3527 }
3528 if (test_opt(sb, DAX)) {
3529 ext4_msg(sb, KERN_ERR, "can't mount with "
3530 "both data=journal and dax");
3531 goto failed_mount;
3532 }
3533 if (ext4_has_feature_encrypt(sb)) {
3534 ext4_msg(sb, KERN_WARNING,
3535 "encrypted files will use data=ordered "
3536 "instead of data journaling mode");
3537 }
3538 if (test_opt(sb, DELALLOC))
3539 clear_opt(sb, DELALLOC);
3540 } else {
3541 sb->s_iflags |= SB_I_CGROUPWB;
3542 }
3543
3544 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3545 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3546
3547 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3548 (ext4_has_compat_features(sb) ||
3549 ext4_has_ro_compat_features(sb) ||
3550 ext4_has_incompat_features(sb)))
3551 ext4_msg(sb, KERN_WARNING,
3552 "feature flags set on rev 0 fs, "
3553 "running e2fsck is recommended");
3554
3555 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3556 set_opt2(sb, HURD_COMPAT);
3557 if (ext4_has_feature_64bit(sb)) {
3558 ext4_msg(sb, KERN_ERR,
3559 "The Hurd can't support 64-bit file systems");
3560 goto failed_mount;
3561 }
3562 }
3563
3564 if (IS_EXT2_SB(sb)) {
3565 if (ext2_feature_set_ok(sb))
3566 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3567 "using the ext4 subsystem");
3568 else {
3569 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3570 "to feature incompatibilities");
3571 goto failed_mount;
3572 }
3573 }
3574
3575 if (IS_EXT3_SB(sb)) {
3576 if (ext3_feature_set_ok(sb))
3577 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3578 "using the ext4 subsystem");
3579 else {
3580 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3581 "to feature incompatibilities");
3582 goto failed_mount;
3583 }
3584 }
3585
3586 /*
3587 * Check feature flags regardless of the revision level, since we
3588 * previously didn't change the revision level when setting the flags,
3589 * so there is a chance incompat flags are set on a rev 0 filesystem.
3590 */
3591 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3592 goto failed_mount;
3593
3594 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3595 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3596 blocksize > EXT4_MAX_BLOCK_SIZE) {
3597 ext4_msg(sb, KERN_ERR,
3598 "Unsupported filesystem blocksize %d (%d log_block_size)",
3599 blocksize, le32_to_cpu(es->s_log_block_size));
3600 goto failed_mount;
3601 }
3602 if (le32_to_cpu(es->s_log_block_size) >
3603 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3604 ext4_msg(sb, KERN_ERR,
3605 "Invalid log block size: %u",
3606 le32_to_cpu(es->s_log_block_size));
3607 goto failed_mount;
3608 }
3609
3610 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3611 ext4_msg(sb, KERN_ERR,
3612 "Number of reserved GDT blocks insanely large: %d",
3613 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3614 goto failed_mount;
3615 }
3616
3617 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3618 err = bdev_dax_supported(sb, blocksize);
3619 if (err)
3620 goto failed_mount;
3621 }
3622
3623 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3624 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3625 es->s_encryption_level);
3626 goto failed_mount;
3627 }
3628
3629 if (sb->s_blocksize != blocksize) {
3630 /* Validate the filesystem blocksize */
3631 if (!sb_set_blocksize(sb, blocksize)) {
3632 ext4_msg(sb, KERN_ERR, "bad block size %d",
3633 blocksize);
3634 goto failed_mount;
3635 }
3636
3637 brelse(bh);
3638 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3639 offset = do_div(logical_sb_block, blocksize);
3640 bh = sb_bread_unmovable(sb, logical_sb_block);
3641 if (!bh) {
3642 ext4_msg(sb, KERN_ERR,
3643 "Can't read superblock on 2nd try");
3644 goto failed_mount;
3645 }
3646 es = (struct ext4_super_block *)(bh->b_data + offset);
3647 sbi->s_es = es;
3648 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3649 ext4_msg(sb, KERN_ERR,
3650 "Magic mismatch, very weird!");
3651 goto failed_mount;
3652 }
3653 }
3654
3655 has_huge_files = ext4_has_feature_huge_file(sb);
3656 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3657 has_huge_files);
3658 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3659
3660 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3661 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3662 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3663 } else {
3664 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3665 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3666 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3667 (!is_power_of_2(sbi->s_inode_size)) ||
3668 (sbi->s_inode_size > blocksize)) {
3669 ext4_msg(sb, KERN_ERR,
3670 "unsupported inode size: %d",
3671 sbi->s_inode_size);
3672 goto failed_mount;
3673 }
3674 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3675 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3676 }
3677
3678 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3679 if (ext4_has_feature_64bit(sb)) {
3680 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3681 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3682 !is_power_of_2(sbi->s_desc_size)) {
3683 ext4_msg(sb, KERN_ERR,
3684 "unsupported descriptor size %lu",
3685 sbi->s_desc_size);
3686 goto failed_mount;
3687 }
3688 } else
3689 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3690
3691 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3692 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3693
3694 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3695 if (sbi->s_inodes_per_block == 0)
3696 goto cantfind_ext4;
3697 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3698 sbi->s_inodes_per_group > blocksize * 8) {
3699 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3700 sbi->s_blocks_per_group);
3701 goto failed_mount;
3702 }
3703 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3704 sbi->s_inodes_per_block;
3705 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3706 sbi->s_sbh = bh;
3707 sbi->s_mount_state = le16_to_cpu(es->s_state);
3708 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3709 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3710
3711 for (i = 0; i < 4; i++)
3712 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3713 sbi->s_def_hash_version = es->s_def_hash_version;
3714 if (ext4_has_feature_dir_index(sb)) {
3715 i = le32_to_cpu(es->s_flags);
3716 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3717 sbi->s_hash_unsigned = 3;
3718 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3719 #ifdef __CHAR_UNSIGNED__
3720 if (!(sb->s_flags & MS_RDONLY))
3721 es->s_flags |=
3722 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3723 sbi->s_hash_unsigned = 3;
3724 #else
3725 if (!(sb->s_flags & MS_RDONLY))
3726 es->s_flags |=
3727 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3728 #endif
3729 }
3730 }
3731
3732 /* Handle clustersize */
3733 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3734 has_bigalloc = ext4_has_feature_bigalloc(sb);
3735 if (has_bigalloc) {
3736 if (clustersize < blocksize) {
3737 ext4_msg(sb, KERN_ERR,
3738 "cluster size (%d) smaller than "
3739 "block size (%d)", clustersize, blocksize);
3740 goto failed_mount;
3741 }
3742 if (le32_to_cpu(es->s_log_cluster_size) >
3743 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3744 ext4_msg(sb, KERN_ERR,
3745 "Invalid log cluster size: %u",
3746 le32_to_cpu(es->s_log_cluster_size));
3747 goto failed_mount;
3748 }
3749 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3750 le32_to_cpu(es->s_log_block_size);
3751 sbi->s_clusters_per_group =
3752 le32_to_cpu(es->s_clusters_per_group);
3753 if (sbi->s_clusters_per_group > blocksize * 8) {
3754 ext4_msg(sb, KERN_ERR,
3755 "#clusters per group too big: %lu",
3756 sbi->s_clusters_per_group);
3757 goto failed_mount;
3758 }
3759 if (sbi->s_blocks_per_group !=
3760 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3761 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3762 "clusters per group (%lu) inconsistent",
3763 sbi->s_blocks_per_group,
3764 sbi->s_clusters_per_group);
3765 goto failed_mount;
3766 }
3767 } else {
3768 if (clustersize != blocksize) {
3769 ext4_warning(sb, "fragment/cluster size (%d) != "
3770 "block size (%d)", clustersize,
3771 blocksize);
3772 clustersize = blocksize;
3773 }
3774 if (sbi->s_blocks_per_group > blocksize * 8) {
3775 ext4_msg(sb, KERN_ERR,
3776 "#blocks per group too big: %lu",
3777 sbi->s_blocks_per_group);
3778 goto failed_mount;
3779 }
3780 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3781 sbi->s_cluster_bits = 0;
3782 }
3783 sbi->s_cluster_ratio = clustersize / blocksize;
3784
3785 /* Do we have standard group size of clustersize * 8 blocks ? */
3786 if (sbi->s_blocks_per_group == clustersize << 3)
3787 set_opt2(sb, STD_GROUP_SIZE);
3788
3789 /*
3790 * Test whether we have more sectors than will fit in sector_t,
3791 * and whether the max offset is addressable by the page cache.
3792 */
3793 err = generic_check_addressable(sb->s_blocksize_bits,
3794 ext4_blocks_count(es));
3795 if (err) {
3796 ext4_msg(sb, KERN_ERR, "filesystem"
3797 " too large to mount safely on this system");
3798 if (sizeof(sector_t) < 8)
3799 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3800 goto failed_mount;
3801 }
3802
3803 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3804 goto cantfind_ext4;
3805
3806 /* check blocks count against device size */
3807 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3808 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3809 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3810 "exceeds size of device (%llu blocks)",
3811 ext4_blocks_count(es), blocks_count);
3812 goto failed_mount;
3813 }
3814
3815 /*
3816 * It makes no sense for the first data block to be beyond the end
3817 * of the filesystem.
3818 */
3819 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3820 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3821 "block %u is beyond end of filesystem (%llu)",
3822 le32_to_cpu(es->s_first_data_block),
3823 ext4_blocks_count(es));
3824 goto failed_mount;
3825 }
3826 blocks_count = (ext4_blocks_count(es) -
3827 le32_to_cpu(es->s_first_data_block) +
3828 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3829 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3830 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3831 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3832 "(block count %llu, first data block %u, "
3833 "blocks per group %lu)", sbi->s_groups_count,
3834 ext4_blocks_count(es),
3835 le32_to_cpu(es->s_first_data_block),
3836 EXT4_BLOCKS_PER_GROUP(sb));
3837 goto failed_mount;
3838 }
3839 sbi->s_groups_count = blocks_count;
3840 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3841 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3842 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3843 EXT4_DESC_PER_BLOCK(sb);
3844 if (ext4_has_feature_meta_bg(sb)) {
3845 if (le32_to_cpu(es->s_first_meta_bg) >= db_count) {
3846 ext4_msg(sb, KERN_WARNING,
3847 "first meta block group too large: %u "
3848 "(group descriptor block count %u)",
3849 le32_to_cpu(es->s_first_meta_bg), db_count);
3850 goto failed_mount;
3851 }
3852 }
3853 sbi->s_group_desc = ext4_kvmalloc(db_count *
3854 sizeof(struct buffer_head *),
3855 GFP_KERNEL);
3856 if (sbi->s_group_desc == NULL) {
3857 ext4_msg(sb, KERN_ERR, "not enough memory");
3858 ret = -ENOMEM;
3859 goto failed_mount;
3860 }
3861
3862 bgl_lock_init(sbi->s_blockgroup_lock);
3863
3864 for (i = 0; i < db_count; i++) {
3865 block = descriptor_loc(sb, logical_sb_block, i);
3866 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3867 if (!sbi->s_group_desc[i]) {
3868 ext4_msg(sb, KERN_ERR,
3869 "can't read group descriptor %d", i);
3870 db_count = i;
3871 goto failed_mount2;
3872 }
3873 }
3874 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3875 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3876 ret = -EFSCORRUPTED;
3877 goto failed_mount2;
3878 }
3879
3880 sbi->s_gdb_count = db_count;
3881 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3882 spin_lock_init(&sbi->s_next_gen_lock);
3883
3884 setup_timer(&sbi->s_err_report, print_daily_error_info,
3885 (unsigned long) sb);
3886
3887 /* Register extent status tree shrinker */
3888 if (ext4_es_register_shrinker(sbi))
3889 goto failed_mount3;
3890
3891 sbi->s_stripe = ext4_get_stripe_size(sbi);
3892 sbi->s_extent_max_zeroout_kb = 32;
3893
3894 /*
3895 * set up enough so that it can read an inode
3896 */
3897 sb->s_op = &ext4_sops;
3898 sb->s_export_op = &ext4_export_ops;
3899 sb->s_xattr = ext4_xattr_handlers;
3900 sb->s_cop = &ext4_cryptops;
3901 #ifdef CONFIG_QUOTA
3902 sb->dq_op = &ext4_quota_operations;
3903 if (ext4_has_feature_quota(sb))
3904 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3905 else
3906 sb->s_qcop = &ext4_qctl_operations;
3907 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3908 #endif
3909 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3910
3911 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3912 mutex_init(&sbi->s_orphan_lock);
3913
3914 sb->s_root = NULL;
3915
3916 needs_recovery = (es->s_last_orphan != 0 ||
3917 ext4_has_feature_journal_needs_recovery(sb));
3918
3919 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3920 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3921 goto failed_mount3a;
3922
3923 /*
3924 * The first inode we look at is the journal inode. Don't try
3925 * root first: it may be modified in the journal!
3926 */
3927 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3928 if (ext4_load_journal(sb, es, journal_devnum))
3929 goto failed_mount3a;
3930 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3931 ext4_has_feature_journal_needs_recovery(sb)) {
3932 ext4_msg(sb, KERN_ERR, "required journal recovery "
3933 "suppressed and not mounted read-only");
3934 goto failed_mount_wq;
3935 } else {
3936 /* Nojournal mode, all journal mount options are illegal */
3937 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3938 ext4_msg(sb, KERN_ERR, "can't mount with "
3939 "journal_checksum, fs mounted w/o journal");
3940 goto failed_mount_wq;
3941 }
3942 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3943 ext4_msg(sb, KERN_ERR, "can't mount with "
3944 "journal_async_commit, fs mounted w/o journal");
3945 goto failed_mount_wq;
3946 }
3947 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3948 ext4_msg(sb, KERN_ERR, "can't mount with "
3949 "commit=%lu, fs mounted w/o journal",
3950 sbi->s_commit_interval / HZ);
3951 goto failed_mount_wq;
3952 }
3953 if (EXT4_MOUNT_DATA_FLAGS &
3954 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3955 ext4_msg(sb, KERN_ERR, "can't mount with "
3956 "data=, fs mounted w/o journal");
3957 goto failed_mount_wq;
3958 }
3959 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3960 clear_opt(sb, JOURNAL_CHECKSUM);
3961 clear_opt(sb, DATA_FLAGS);
3962 sbi->s_journal = NULL;
3963 needs_recovery = 0;
3964 goto no_journal;
3965 }
3966
3967 if (ext4_has_feature_64bit(sb) &&
3968 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3969 JBD2_FEATURE_INCOMPAT_64BIT)) {
3970 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3971 goto failed_mount_wq;
3972 }
3973
3974 if (!set_journal_csum_feature_set(sb)) {
3975 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3976 "feature set");
3977 goto failed_mount_wq;
3978 }
3979
3980 /* We have now updated the journal if required, so we can
3981 * validate the data journaling mode. */
3982 switch (test_opt(sb, DATA_FLAGS)) {
3983 case 0:
3984 /* No mode set, assume a default based on the journal
3985 * capabilities: ORDERED_DATA if the journal can
3986 * cope, else JOURNAL_DATA
3987 */
3988 if (jbd2_journal_check_available_features
3989 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3990 set_opt(sb, ORDERED_DATA);
3991 else
3992 set_opt(sb, JOURNAL_DATA);
3993 break;
3994
3995 case EXT4_MOUNT_ORDERED_DATA:
3996 case EXT4_MOUNT_WRITEBACK_DATA:
3997 if (!jbd2_journal_check_available_features
3998 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3999 ext4_msg(sb, KERN_ERR, "Journal does not support "
4000 "requested data journaling mode");
4001 goto failed_mount_wq;
4002 }
4003 default:
4004 break;
4005 }
4006
4007 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4008 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4009 ext4_msg(sb, KERN_ERR, "can't mount with "
4010 "journal_async_commit in data=ordered mode");
4011 goto failed_mount_wq;
4012 }
4013
4014 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4015
4016 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4017
4018 no_journal:
4019 sbi->s_mb_cache = ext4_xattr_create_cache();
4020 if (!sbi->s_mb_cache) {
4021 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4022 goto failed_mount_wq;
4023 }
4024
4025 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4026 (blocksize != PAGE_SIZE)) {
4027 ext4_msg(sb, KERN_ERR,
4028 "Unsupported blocksize for fs encryption");
4029 goto failed_mount_wq;
4030 }
4031
4032 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4033 !ext4_has_feature_encrypt(sb)) {
4034 ext4_set_feature_encrypt(sb);
4035 ext4_commit_super(sb, 1);
4036 }
4037
4038 /*
4039 * Get the # of file system overhead blocks from the
4040 * superblock if present.
4041 */
4042 if (es->s_overhead_clusters)
4043 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4044 else {
4045 err = ext4_calculate_overhead(sb);
4046 if (err)
4047 goto failed_mount_wq;
4048 }
4049
4050 /*
4051 * The maximum number of concurrent works can be high and
4052 * concurrency isn't really necessary. Limit it to 1.
4053 */
4054 EXT4_SB(sb)->rsv_conversion_wq =
4055 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4056 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4057 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4058 ret = -ENOMEM;
4059 goto failed_mount4;
4060 }
4061
4062 /*
4063 * The jbd2_journal_load will have done any necessary log recovery,
4064 * so we can safely mount the rest of the filesystem now.
4065 */
4066
4067 root = ext4_iget(sb, EXT4_ROOT_INO);
4068 if (IS_ERR(root)) {
4069 ext4_msg(sb, KERN_ERR, "get root inode failed");
4070 ret = PTR_ERR(root);
4071 root = NULL;
4072 goto failed_mount4;
4073 }
4074 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4075 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4076 iput(root);
4077 goto failed_mount4;
4078 }
4079 sb->s_root = d_make_root(root);
4080 if (!sb->s_root) {
4081 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4082 ret = -ENOMEM;
4083 goto failed_mount4;
4084 }
4085
4086 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4087 sb->s_flags |= MS_RDONLY;
4088
4089 /* determine the minimum size of new large inodes, if present */
4090 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4091 sbi->s_want_extra_isize == 0) {
4092 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4093 EXT4_GOOD_OLD_INODE_SIZE;
4094 if (ext4_has_feature_extra_isize(sb)) {
4095 if (sbi->s_want_extra_isize <
4096 le16_to_cpu(es->s_want_extra_isize))
4097 sbi->s_want_extra_isize =
4098 le16_to_cpu(es->s_want_extra_isize);
4099 if (sbi->s_want_extra_isize <
4100 le16_to_cpu(es->s_min_extra_isize))
4101 sbi->s_want_extra_isize =
4102 le16_to_cpu(es->s_min_extra_isize);
4103 }
4104 }
4105 /* Check if enough inode space is available */
4106 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4107 sbi->s_inode_size) {
4108 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4109 EXT4_GOOD_OLD_INODE_SIZE;
4110 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4111 "available");
4112 }
4113
4114 ext4_set_resv_clusters(sb);
4115
4116 err = ext4_setup_system_zone(sb);
4117 if (err) {
4118 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4119 "zone (%d)", err);
4120 goto failed_mount4a;
4121 }
4122
4123 ext4_ext_init(sb);
4124 err = ext4_mb_init(sb);
4125 if (err) {
4126 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4127 err);
4128 goto failed_mount5;
4129 }
4130
4131 block = ext4_count_free_clusters(sb);
4132 ext4_free_blocks_count_set(sbi->s_es,
4133 EXT4_C2B(sbi, block));
4134 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4135 GFP_KERNEL);
4136 if (!err) {
4137 unsigned long freei = ext4_count_free_inodes(sb);
4138 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4139 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4140 GFP_KERNEL);
4141 }
4142 if (!err)
4143 err = percpu_counter_init(&sbi->s_dirs_counter,
4144 ext4_count_dirs(sb), GFP_KERNEL);
4145 if (!err)
4146 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4147 GFP_KERNEL);
4148 if (!err)
4149 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4150
4151 if (err) {
4152 ext4_msg(sb, KERN_ERR, "insufficient memory");
4153 goto failed_mount6;
4154 }
4155
4156 if (ext4_has_feature_flex_bg(sb))
4157 if (!ext4_fill_flex_info(sb)) {
4158 ext4_msg(sb, KERN_ERR,
4159 "unable to initialize "
4160 "flex_bg meta info!");
4161 goto failed_mount6;
4162 }
4163
4164 err = ext4_register_li_request(sb, first_not_zeroed);
4165 if (err)
4166 goto failed_mount6;
4167
4168 err = ext4_register_sysfs(sb);
4169 if (err)
4170 goto failed_mount7;
4171
4172 #ifdef CONFIG_QUOTA
4173 /* Enable quota usage during mount. */
4174 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4175 err = ext4_enable_quotas(sb);
4176 if (err)
4177 goto failed_mount8;
4178 }
4179 #endif /* CONFIG_QUOTA */
4180
4181 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4182 ext4_orphan_cleanup(sb, es);
4183 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4184 if (needs_recovery) {
4185 ext4_msg(sb, KERN_INFO, "recovery complete");
4186 ext4_mark_recovery_complete(sb, es);
4187 }
4188 if (EXT4_SB(sb)->s_journal) {
4189 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4190 descr = " journalled data mode";
4191 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4192 descr = " ordered data mode";
4193 else
4194 descr = " writeback data mode";
4195 } else
4196 descr = "out journal";
4197
4198 if (test_opt(sb, DISCARD)) {
4199 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4200 if (!blk_queue_discard(q))
4201 ext4_msg(sb, KERN_WARNING,
4202 "mounting with \"discard\" option, but "
4203 "the device does not support discard");
4204 }
4205
4206 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4207 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4208 "Opts: %.*s%s%s", descr,
4209 (int) sizeof(sbi->s_es->s_mount_opts),
4210 sbi->s_es->s_mount_opts,
4211 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4212
4213 if (es->s_error_count)
4214 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4215
4216 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4217 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4218 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4219 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4220
4221 kfree(orig_data);
4222 return 0;
4223
4224 cantfind_ext4:
4225 if (!silent)
4226 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4227 goto failed_mount;
4228
4229 #ifdef CONFIG_QUOTA
4230 failed_mount8:
4231 ext4_unregister_sysfs(sb);
4232 #endif
4233 failed_mount7:
4234 ext4_unregister_li_request(sb);
4235 failed_mount6:
4236 ext4_mb_release(sb);
4237 if (sbi->s_flex_groups)
4238 kvfree(sbi->s_flex_groups);
4239 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4240 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4241 percpu_counter_destroy(&sbi->s_dirs_counter);
4242 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4243 failed_mount5:
4244 ext4_ext_release(sb);
4245 ext4_release_system_zone(sb);
4246 failed_mount4a:
4247 dput(sb->s_root);
4248 sb->s_root = NULL;
4249 failed_mount4:
4250 ext4_msg(sb, KERN_ERR, "mount failed");
4251 if (EXT4_SB(sb)->rsv_conversion_wq)
4252 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4253 failed_mount_wq:
4254 if (sbi->s_mb_cache) {
4255 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4256 sbi->s_mb_cache = NULL;
4257 }
4258 if (sbi->s_journal) {
4259 jbd2_journal_destroy(sbi->s_journal);
4260 sbi->s_journal = NULL;
4261 }
4262 failed_mount3a:
4263 ext4_es_unregister_shrinker(sbi);
4264 failed_mount3:
4265 del_timer_sync(&sbi->s_err_report);
4266 if (sbi->s_mmp_tsk)
4267 kthread_stop(sbi->s_mmp_tsk);
4268 failed_mount2:
4269 for (i = 0; i < db_count; i++)
4270 brelse(sbi->s_group_desc[i]);
4271 kvfree(sbi->s_group_desc);
4272 failed_mount:
4273 if (sbi->s_chksum_driver)
4274 crypto_free_shash(sbi->s_chksum_driver);
4275 #ifdef CONFIG_QUOTA
4276 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4277 kfree(sbi->s_qf_names[i]);
4278 #endif
4279 ext4_blkdev_remove(sbi);
4280 brelse(bh);
4281 out_fail:
4282 sb->s_fs_info = NULL;
4283 kfree(sbi->s_blockgroup_lock);
4284 out_free_base:
4285 kfree(sbi);
4286 kfree(orig_data);
4287 return err ? err : ret;
4288 }
4289
4290 /*
4291 * Setup any per-fs journal parameters now. We'll do this both on
4292 * initial mount, once the journal has been initialised but before we've
4293 * done any recovery; and again on any subsequent remount.
4294 */
4295 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4296 {
4297 struct ext4_sb_info *sbi = EXT4_SB(sb);
4298
4299 journal->j_commit_interval = sbi->s_commit_interval;
4300 journal->j_min_batch_time = sbi->s_min_batch_time;
4301 journal->j_max_batch_time = sbi->s_max_batch_time;
4302
4303 write_lock(&journal->j_state_lock);
4304 if (test_opt(sb, BARRIER))
4305 journal->j_flags |= JBD2_BARRIER;
4306 else
4307 journal->j_flags &= ~JBD2_BARRIER;
4308 if (test_opt(sb, DATA_ERR_ABORT))
4309 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4310 else
4311 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4312 write_unlock(&journal->j_state_lock);
4313 }
4314
4315 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4316 unsigned int journal_inum)
4317 {
4318 struct inode *journal_inode;
4319
4320 /*
4321 * Test for the existence of a valid inode on disk. Bad things
4322 * happen if we iget() an unused inode, as the subsequent iput()
4323 * will try to delete it.
4324 */
4325 journal_inode = ext4_iget(sb, journal_inum);
4326 if (IS_ERR(journal_inode)) {
4327 ext4_msg(sb, KERN_ERR, "no journal found");
4328 return NULL;
4329 }
4330 if (!journal_inode->i_nlink) {
4331 make_bad_inode(journal_inode);
4332 iput(journal_inode);
4333 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4334 return NULL;
4335 }
4336
4337 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4338 journal_inode, journal_inode->i_size);
4339 if (!S_ISREG(journal_inode->i_mode)) {
4340 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4341 iput(journal_inode);
4342 return NULL;
4343 }
4344 return journal_inode;
4345 }
4346
4347 static journal_t *ext4_get_journal(struct super_block *sb,
4348 unsigned int journal_inum)
4349 {
4350 struct inode *journal_inode;
4351 journal_t *journal;
4352
4353 BUG_ON(!ext4_has_feature_journal(sb));
4354
4355 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4356 if (!journal_inode)
4357 return NULL;
4358
4359 journal = jbd2_journal_init_inode(journal_inode);
4360 if (!journal) {
4361 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4362 iput(journal_inode);
4363 return NULL;
4364 }
4365 journal->j_private = sb;
4366 ext4_init_journal_params(sb, journal);
4367 return journal;
4368 }
4369
4370 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4371 dev_t j_dev)
4372 {
4373 struct buffer_head *bh;
4374 journal_t *journal;
4375 ext4_fsblk_t start;
4376 ext4_fsblk_t len;
4377 int hblock, blocksize;
4378 ext4_fsblk_t sb_block;
4379 unsigned long offset;
4380 struct ext4_super_block *es;
4381 struct block_device *bdev;
4382
4383 BUG_ON(!ext4_has_feature_journal(sb));
4384
4385 bdev = ext4_blkdev_get(j_dev, sb);
4386 if (bdev == NULL)
4387 return NULL;
4388
4389 blocksize = sb->s_blocksize;
4390 hblock = bdev_logical_block_size(bdev);
4391 if (blocksize < hblock) {
4392 ext4_msg(sb, KERN_ERR,
4393 "blocksize too small for journal device");
4394 goto out_bdev;
4395 }
4396
4397 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4398 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4399 set_blocksize(bdev, blocksize);
4400 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4401 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4402 "external journal");
4403 goto out_bdev;
4404 }
4405
4406 es = (struct ext4_super_block *) (bh->b_data + offset);
4407 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4408 !(le32_to_cpu(es->s_feature_incompat) &
4409 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4410 ext4_msg(sb, KERN_ERR, "external journal has "
4411 "bad superblock");
4412 brelse(bh);
4413 goto out_bdev;
4414 }
4415
4416 if ((le32_to_cpu(es->s_feature_ro_compat) &
4417 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4418 es->s_checksum != ext4_superblock_csum(sb, es)) {
4419 ext4_msg(sb, KERN_ERR, "external journal has "
4420 "corrupt superblock");
4421 brelse(bh);
4422 goto out_bdev;
4423 }
4424
4425 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4426 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4427 brelse(bh);
4428 goto out_bdev;
4429 }
4430
4431 len = ext4_blocks_count(es);
4432 start = sb_block + 1;
4433 brelse(bh); /* we're done with the superblock */
4434
4435 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4436 start, len, blocksize);
4437 if (!journal) {
4438 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4439 goto out_bdev;
4440 }
4441 journal->j_private = sb;
4442 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4443 wait_on_buffer(journal->j_sb_buffer);
4444 if (!buffer_uptodate(journal->j_sb_buffer)) {
4445 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4446 goto out_journal;
4447 }
4448 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4449 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4450 "user (unsupported) - %d",
4451 be32_to_cpu(journal->j_superblock->s_nr_users));
4452 goto out_journal;
4453 }
4454 EXT4_SB(sb)->journal_bdev = bdev;
4455 ext4_init_journal_params(sb, journal);
4456 return journal;
4457
4458 out_journal:
4459 jbd2_journal_destroy(journal);
4460 out_bdev:
4461 ext4_blkdev_put(bdev);
4462 return NULL;
4463 }
4464
4465 static int ext4_load_journal(struct super_block *sb,
4466 struct ext4_super_block *es,
4467 unsigned long journal_devnum)
4468 {
4469 journal_t *journal;
4470 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4471 dev_t journal_dev;
4472 int err = 0;
4473 int really_read_only;
4474
4475 BUG_ON(!ext4_has_feature_journal(sb));
4476
4477 if (journal_devnum &&
4478 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4479 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4480 "numbers have changed");
4481 journal_dev = new_decode_dev(journal_devnum);
4482 } else
4483 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4484
4485 really_read_only = bdev_read_only(sb->s_bdev);
4486
4487 /*
4488 * Are we loading a blank journal or performing recovery after a
4489 * crash? For recovery, we need to check in advance whether we
4490 * can get read-write access to the device.
4491 */
4492 if (ext4_has_feature_journal_needs_recovery(sb)) {
4493 if (sb->s_flags & MS_RDONLY) {
4494 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4495 "required on readonly filesystem");
4496 if (really_read_only) {
4497 ext4_msg(sb, KERN_ERR, "write access "
4498 "unavailable, cannot proceed");
4499 return -EROFS;
4500 }
4501 ext4_msg(sb, KERN_INFO, "write access will "
4502 "be enabled during recovery");
4503 }
4504 }
4505
4506 if (journal_inum && journal_dev) {
4507 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4508 "and inode journals!");
4509 return -EINVAL;
4510 }
4511
4512 if (journal_inum) {
4513 if (!(journal = ext4_get_journal(sb, journal_inum)))
4514 return -EINVAL;
4515 } else {
4516 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4517 return -EINVAL;
4518 }
4519
4520 if (!(journal->j_flags & JBD2_BARRIER))
4521 ext4_msg(sb, KERN_INFO, "barriers disabled");
4522
4523 if (!ext4_has_feature_journal_needs_recovery(sb))
4524 err = jbd2_journal_wipe(journal, !really_read_only);
4525 if (!err) {
4526 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4527 if (save)
4528 memcpy(save, ((char *) es) +
4529 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4530 err = jbd2_journal_load(journal);
4531 if (save)
4532 memcpy(((char *) es) + EXT4_S_ERR_START,
4533 save, EXT4_S_ERR_LEN);
4534 kfree(save);
4535 }
4536
4537 if (err) {
4538 ext4_msg(sb, KERN_ERR, "error loading journal");
4539 jbd2_journal_destroy(journal);
4540 return err;
4541 }
4542
4543 EXT4_SB(sb)->s_journal = journal;
4544 ext4_clear_journal_err(sb, es);
4545
4546 if (!really_read_only && journal_devnum &&
4547 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4548 es->s_journal_dev = cpu_to_le32(journal_devnum);
4549
4550 /* Make sure we flush the recovery flag to disk. */
4551 ext4_commit_super(sb, 1);
4552 }
4553
4554 return 0;
4555 }
4556
4557 static int ext4_commit_super(struct super_block *sb, int sync)
4558 {
4559 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4560 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4561 int error = 0;
4562
4563 if (!sbh || block_device_ejected(sb))
4564 return error;
4565 /*
4566 * If the file system is mounted read-only, don't update the
4567 * superblock write time. This avoids updating the superblock
4568 * write time when we are mounting the root file system
4569 * read/only but we need to replay the journal; at that point,
4570 * for people who are east of GMT and who make their clock
4571 * tick in localtime for Windows bug-for-bug compatibility,
4572 * the clock is set in the future, and this will cause e2fsck
4573 * to complain and force a full file system check.
4574 */
4575 if (!(sb->s_flags & MS_RDONLY))
4576 es->s_wtime = cpu_to_le32(get_seconds());
4577 if (sb->s_bdev->bd_part)
4578 es->s_kbytes_written =
4579 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4580 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4581 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4582 else
4583 es->s_kbytes_written =
4584 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4585 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4586 ext4_free_blocks_count_set(es,
4587 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4588 &EXT4_SB(sb)->s_freeclusters_counter)));
4589 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4590 es->s_free_inodes_count =
4591 cpu_to_le32(percpu_counter_sum_positive(
4592 &EXT4_SB(sb)->s_freeinodes_counter));
4593 BUFFER_TRACE(sbh, "marking dirty");
4594 ext4_superblock_csum_set(sb);
4595 if (sync)
4596 lock_buffer(sbh);
4597 if (buffer_write_io_error(sbh)) {
4598 /*
4599 * Oh, dear. A previous attempt to write the
4600 * superblock failed. This could happen because the
4601 * USB device was yanked out. Or it could happen to
4602 * be a transient write error and maybe the block will
4603 * be remapped. Nothing we can do but to retry the
4604 * write and hope for the best.
4605 */
4606 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4607 "superblock detected");
4608 clear_buffer_write_io_error(sbh);
4609 set_buffer_uptodate(sbh);
4610 }
4611 mark_buffer_dirty(sbh);
4612 if (sync) {
4613 unlock_buffer(sbh);
4614 error = __sync_dirty_buffer(sbh,
4615 test_opt(sb, BARRIER) ? REQ_FUA : REQ_SYNC);
4616 if (error)
4617 return error;
4618
4619 error = buffer_write_io_error(sbh);
4620 if (error) {
4621 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4622 "superblock");
4623 clear_buffer_write_io_error(sbh);
4624 set_buffer_uptodate(sbh);
4625 }
4626 }
4627 return error;
4628 }
4629
4630 /*
4631 * Have we just finished recovery? If so, and if we are mounting (or
4632 * remounting) the filesystem readonly, then we will end up with a
4633 * consistent fs on disk. Record that fact.
4634 */
4635 static void ext4_mark_recovery_complete(struct super_block *sb,
4636 struct ext4_super_block *es)
4637 {
4638 journal_t *journal = EXT4_SB(sb)->s_journal;
4639
4640 if (!ext4_has_feature_journal(sb)) {
4641 BUG_ON(journal != NULL);
4642 return;
4643 }
4644 jbd2_journal_lock_updates(journal);
4645 if (jbd2_journal_flush(journal) < 0)
4646 goto out;
4647
4648 if (ext4_has_feature_journal_needs_recovery(sb) &&
4649 sb->s_flags & MS_RDONLY) {
4650 ext4_clear_feature_journal_needs_recovery(sb);
4651 ext4_commit_super(sb, 1);
4652 }
4653
4654 out:
4655 jbd2_journal_unlock_updates(journal);
4656 }
4657
4658 /*
4659 * If we are mounting (or read-write remounting) a filesystem whose journal
4660 * has recorded an error from a previous lifetime, move that error to the
4661 * main filesystem now.
4662 */
4663 static void ext4_clear_journal_err(struct super_block *sb,
4664 struct ext4_super_block *es)
4665 {
4666 journal_t *journal;
4667 int j_errno;
4668 const char *errstr;
4669
4670 BUG_ON(!ext4_has_feature_journal(sb));
4671
4672 journal = EXT4_SB(sb)->s_journal;
4673
4674 /*
4675 * Now check for any error status which may have been recorded in the
4676 * journal by a prior ext4_error() or ext4_abort()
4677 */
4678
4679 j_errno = jbd2_journal_errno(journal);
4680 if (j_errno) {
4681 char nbuf[16];
4682
4683 errstr = ext4_decode_error(sb, j_errno, nbuf);
4684 ext4_warning(sb, "Filesystem error recorded "
4685 "from previous mount: %s", errstr);
4686 ext4_warning(sb, "Marking fs in need of filesystem check.");
4687
4688 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4689 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4690 ext4_commit_super(sb, 1);
4691
4692 jbd2_journal_clear_err(journal);
4693 jbd2_journal_update_sb_errno(journal);
4694 }
4695 }
4696
4697 /*
4698 * Force the running and committing transactions to commit,
4699 * and wait on the commit.
4700 */
4701 int ext4_force_commit(struct super_block *sb)
4702 {
4703 journal_t *journal;
4704
4705 if (sb->s_flags & MS_RDONLY)
4706 return 0;
4707
4708 journal = EXT4_SB(sb)->s_journal;
4709 return ext4_journal_force_commit(journal);
4710 }
4711
4712 static int ext4_sync_fs(struct super_block *sb, int wait)
4713 {
4714 int ret = 0;
4715 tid_t target;
4716 bool needs_barrier = false;
4717 struct ext4_sb_info *sbi = EXT4_SB(sb);
4718
4719 trace_ext4_sync_fs(sb, wait);
4720 flush_workqueue(sbi->rsv_conversion_wq);
4721 /*
4722 * Writeback quota in non-journalled quota case - journalled quota has
4723 * no dirty dquots
4724 */
4725 dquot_writeback_dquots(sb, -1);
4726 /*
4727 * Data writeback is possible w/o journal transaction, so barrier must
4728 * being sent at the end of the function. But we can skip it if
4729 * transaction_commit will do it for us.
4730 */
4731 if (sbi->s_journal) {
4732 target = jbd2_get_latest_transaction(sbi->s_journal);
4733 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4734 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4735 needs_barrier = true;
4736
4737 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4738 if (wait)
4739 ret = jbd2_log_wait_commit(sbi->s_journal,
4740 target);
4741 }
4742 } else if (wait && test_opt(sb, BARRIER))
4743 needs_barrier = true;
4744 if (needs_barrier) {
4745 int err;
4746 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4747 if (!ret)
4748 ret = err;
4749 }
4750
4751 return ret;
4752 }
4753
4754 /*
4755 * LVM calls this function before a (read-only) snapshot is created. This
4756 * gives us a chance to flush the journal completely and mark the fs clean.
4757 *
4758 * Note that only this function cannot bring a filesystem to be in a clean
4759 * state independently. It relies on upper layer to stop all data & metadata
4760 * modifications.
4761 */
4762 static int ext4_freeze(struct super_block *sb)
4763 {
4764 int error = 0;
4765 journal_t *journal;
4766
4767 if (sb->s_flags & MS_RDONLY)
4768 return 0;
4769
4770 journal = EXT4_SB(sb)->s_journal;
4771
4772 if (journal) {
4773 /* Now we set up the journal barrier. */
4774 jbd2_journal_lock_updates(journal);
4775
4776 /*
4777 * Don't clear the needs_recovery flag if we failed to
4778 * flush the journal.
4779 */
4780 error = jbd2_journal_flush(journal);
4781 if (error < 0)
4782 goto out;
4783
4784 /* Journal blocked and flushed, clear needs_recovery flag. */
4785 ext4_clear_feature_journal_needs_recovery(sb);
4786 }
4787
4788 error = ext4_commit_super(sb, 1);
4789 out:
4790 if (journal)
4791 /* we rely on upper layer to stop further updates */
4792 jbd2_journal_unlock_updates(journal);
4793 return error;
4794 }
4795
4796 /*
4797 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4798 * flag here, even though the filesystem is not technically dirty yet.
4799 */
4800 static int ext4_unfreeze(struct super_block *sb)
4801 {
4802 if (sb->s_flags & MS_RDONLY)
4803 return 0;
4804
4805 if (EXT4_SB(sb)->s_journal) {
4806 /* Reset the needs_recovery flag before the fs is unlocked. */
4807 ext4_set_feature_journal_needs_recovery(sb);
4808 }
4809
4810 ext4_commit_super(sb, 1);
4811 return 0;
4812 }
4813
4814 /*
4815 * Structure to save mount options for ext4_remount's benefit
4816 */
4817 struct ext4_mount_options {
4818 unsigned long s_mount_opt;
4819 unsigned long s_mount_opt2;
4820 kuid_t s_resuid;
4821 kgid_t s_resgid;
4822 unsigned long s_commit_interval;
4823 u32 s_min_batch_time, s_max_batch_time;
4824 #ifdef CONFIG_QUOTA
4825 int s_jquota_fmt;
4826 char *s_qf_names[EXT4_MAXQUOTAS];
4827 #endif
4828 };
4829
4830 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4831 {
4832 struct ext4_super_block *es;
4833 struct ext4_sb_info *sbi = EXT4_SB(sb);
4834 unsigned long old_sb_flags;
4835 struct ext4_mount_options old_opts;
4836 int enable_quota = 0;
4837 ext4_group_t g;
4838 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4839 int err = 0;
4840 #ifdef CONFIG_QUOTA
4841 int i, j;
4842 #endif
4843 char *orig_data = kstrdup(data, GFP_KERNEL);
4844
4845 /* Store the original options */
4846 old_sb_flags = sb->s_flags;
4847 old_opts.s_mount_opt = sbi->s_mount_opt;
4848 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4849 old_opts.s_resuid = sbi->s_resuid;
4850 old_opts.s_resgid = sbi->s_resgid;
4851 old_opts.s_commit_interval = sbi->s_commit_interval;
4852 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4853 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4854 #ifdef CONFIG_QUOTA
4855 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4856 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4857 if (sbi->s_qf_names[i]) {
4858 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4859 GFP_KERNEL);
4860 if (!old_opts.s_qf_names[i]) {
4861 for (j = 0; j < i; j++)
4862 kfree(old_opts.s_qf_names[j]);
4863 kfree(orig_data);
4864 return -ENOMEM;
4865 }
4866 } else
4867 old_opts.s_qf_names[i] = NULL;
4868 #endif
4869 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4870 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4871
4872 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4873 err = -EINVAL;
4874 goto restore_opts;
4875 }
4876
4877 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4878 test_opt(sb, JOURNAL_CHECKSUM)) {
4879 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4880 "during remount not supported; ignoring");
4881 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4882 }
4883
4884 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4885 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4886 ext4_msg(sb, KERN_ERR, "can't mount with "
4887 "both data=journal and delalloc");
4888 err = -EINVAL;
4889 goto restore_opts;
4890 }
4891 if (test_opt(sb, DIOREAD_NOLOCK)) {
4892 ext4_msg(sb, KERN_ERR, "can't mount with "
4893 "both data=journal and dioread_nolock");
4894 err = -EINVAL;
4895 goto restore_opts;
4896 }
4897 if (test_opt(sb, DAX)) {
4898 ext4_msg(sb, KERN_ERR, "can't mount with "
4899 "both data=journal and dax");
4900 err = -EINVAL;
4901 goto restore_opts;
4902 }
4903 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
4904 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4905 ext4_msg(sb, KERN_ERR, "can't mount with "
4906 "journal_async_commit in data=ordered mode");
4907 err = -EINVAL;
4908 goto restore_opts;
4909 }
4910 }
4911
4912 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4913 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4914 "dax flag with busy inodes while remounting");
4915 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4916 }
4917
4918 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4919 ext4_abort(sb, "Abort forced by user");
4920
4921 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4922 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4923
4924 es = sbi->s_es;
4925
4926 if (sbi->s_journal) {
4927 ext4_init_journal_params(sb, sbi->s_journal);
4928 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4929 }
4930
4931 if (*flags & MS_LAZYTIME)
4932 sb->s_flags |= MS_LAZYTIME;
4933
4934 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4935 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4936 err = -EROFS;
4937 goto restore_opts;
4938 }
4939
4940 if (*flags & MS_RDONLY) {
4941 err = sync_filesystem(sb);
4942 if (err < 0)
4943 goto restore_opts;
4944 err = dquot_suspend(sb, -1);
4945 if (err < 0)
4946 goto restore_opts;
4947
4948 /*
4949 * First of all, the unconditional stuff we have to do
4950 * to disable replay of the journal when we next remount
4951 */
4952 sb->s_flags |= MS_RDONLY;
4953
4954 /*
4955 * OK, test if we are remounting a valid rw partition
4956 * readonly, and if so set the rdonly flag and then
4957 * mark the partition as valid again.
4958 */
4959 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4960 (sbi->s_mount_state & EXT4_VALID_FS))
4961 es->s_state = cpu_to_le16(sbi->s_mount_state);
4962
4963 if (sbi->s_journal)
4964 ext4_mark_recovery_complete(sb, es);
4965 } else {
4966 /* Make sure we can mount this feature set readwrite */
4967 if (ext4_has_feature_readonly(sb) ||
4968 !ext4_feature_set_ok(sb, 0)) {
4969 err = -EROFS;
4970 goto restore_opts;
4971 }
4972 /*
4973 * Make sure the group descriptor checksums
4974 * are sane. If they aren't, refuse to remount r/w.
4975 */
4976 for (g = 0; g < sbi->s_groups_count; g++) {
4977 struct ext4_group_desc *gdp =
4978 ext4_get_group_desc(sb, g, NULL);
4979
4980 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4981 ext4_msg(sb, KERN_ERR,
4982 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4983 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4984 le16_to_cpu(gdp->bg_checksum));
4985 err = -EFSBADCRC;
4986 goto restore_opts;
4987 }
4988 }
4989
4990 /*
4991 * If we have an unprocessed orphan list hanging
4992 * around from a previously readonly bdev mount,
4993 * require a full umount/remount for now.
4994 */
4995 if (es->s_last_orphan) {
4996 ext4_msg(sb, KERN_WARNING, "Couldn't "
4997 "remount RDWR because of unprocessed "
4998 "orphan inode list. Please "
4999 "umount/remount instead");
5000 err = -EINVAL;
5001 goto restore_opts;
5002 }
5003
5004 /*
5005 * Mounting a RDONLY partition read-write, so reread
5006 * and store the current valid flag. (It may have
5007 * been changed by e2fsck since we originally mounted
5008 * the partition.)
5009 */
5010 if (sbi->s_journal)
5011 ext4_clear_journal_err(sb, es);
5012 sbi->s_mount_state = le16_to_cpu(es->s_state);
5013 if (!ext4_setup_super(sb, es, 0))
5014 sb->s_flags &= ~MS_RDONLY;
5015 if (ext4_has_feature_mmp(sb))
5016 if (ext4_multi_mount_protect(sb,
5017 le64_to_cpu(es->s_mmp_block))) {
5018 err = -EROFS;
5019 goto restore_opts;
5020 }
5021 enable_quota = 1;
5022 }
5023 }
5024
5025 /*
5026 * Reinitialize lazy itable initialization thread based on
5027 * current settings
5028 */
5029 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5030 ext4_unregister_li_request(sb);
5031 else {
5032 ext4_group_t first_not_zeroed;
5033 first_not_zeroed = ext4_has_uninit_itable(sb);
5034 ext4_register_li_request(sb, first_not_zeroed);
5035 }
5036
5037 ext4_setup_system_zone(sb);
5038 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5039 ext4_commit_super(sb, 1);
5040
5041 #ifdef CONFIG_QUOTA
5042 /* Release old quota file names */
5043 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5044 kfree(old_opts.s_qf_names[i]);
5045 if (enable_quota) {
5046 if (sb_any_quota_suspended(sb))
5047 dquot_resume(sb, -1);
5048 else if (ext4_has_feature_quota(sb)) {
5049 err = ext4_enable_quotas(sb);
5050 if (err)
5051 goto restore_opts;
5052 }
5053 }
5054 #endif
5055
5056 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5057 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5058 kfree(orig_data);
5059 return 0;
5060
5061 restore_opts:
5062 sb->s_flags = old_sb_flags;
5063 sbi->s_mount_opt = old_opts.s_mount_opt;
5064 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5065 sbi->s_resuid = old_opts.s_resuid;
5066 sbi->s_resgid = old_opts.s_resgid;
5067 sbi->s_commit_interval = old_opts.s_commit_interval;
5068 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5069 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5070 #ifdef CONFIG_QUOTA
5071 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5072 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5073 kfree(sbi->s_qf_names[i]);
5074 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5075 }
5076 #endif
5077 kfree(orig_data);
5078 return err;
5079 }
5080
5081 #ifdef CONFIG_QUOTA
5082 static int ext4_statfs_project(struct super_block *sb,
5083 kprojid_t projid, struct kstatfs *buf)
5084 {
5085 struct kqid qid;
5086 struct dquot *dquot;
5087 u64 limit;
5088 u64 curblock;
5089
5090 qid = make_kqid_projid(projid);
5091 dquot = dqget(sb, qid);
5092 if (IS_ERR(dquot))
5093 return PTR_ERR(dquot);
5094 spin_lock(&dq_data_lock);
5095
5096 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5097 dquot->dq_dqb.dqb_bsoftlimit :
5098 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5099 if (limit && buf->f_blocks > limit) {
5100 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5101 buf->f_blocks = limit;
5102 buf->f_bfree = buf->f_bavail =
5103 (buf->f_blocks > curblock) ?
5104 (buf->f_blocks - curblock) : 0;
5105 }
5106
5107 limit = dquot->dq_dqb.dqb_isoftlimit ?
5108 dquot->dq_dqb.dqb_isoftlimit :
5109 dquot->dq_dqb.dqb_ihardlimit;
5110 if (limit && buf->f_files > limit) {
5111 buf->f_files = limit;
5112 buf->f_ffree =
5113 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5114 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5115 }
5116
5117 spin_unlock(&dq_data_lock);
5118 dqput(dquot);
5119 return 0;
5120 }
5121 #endif
5122
5123 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5124 {
5125 struct super_block *sb = dentry->d_sb;
5126 struct ext4_sb_info *sbi = EXT4_SB(sb);
5127 struct ext4_super_block *es = sbi->s_es;
5128 ext4_fsblk_t overhead = 0, resv_blocks;
5129 u64 fsid;
5130 s64 bfree;
5131 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5132
5133 if (!test_opt(sb, MINIX_DF))
5134 overhead = sbi->s_overhead;
5135
5136 buf->f_type = EXT4_SUPER_MAGIC;
5137 buf->f_bsize = sb->s_blocksize;
5138 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5139 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5140 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5141 /* prevent underflow in case that few free space is available */
5142 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5143 buf->f_bavail = buf->f_bfree -
5144 (ext4_r_blocks_count(es) + resv_blocks);
5145 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5146 buf->f_bavail = 0;
5147 buf->f_files = le32_to_cpu(es->s_inodes_count);
5148 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5149 buf->f_namelen = EXT4_NAME_LEN;
5150 fsid = le64_to_cpup((void *)es->s_uuid) ^
5151 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5152 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5153 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5154
5155 #ifdef CONFIG_QUOTA
5156 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5157 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5158 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5159 #endif
5160 return 0;
5161 }
5162
5163 /* Helper function for writing quotas on sync - we need to start transaction
5164 * before quota file is locked for write. Otherwise the are possible deadlocks:
5165 * Process 1 Process 2
5166 * ext4_create() quota_sync()
5167 * jbd2_journal_start() write_dquot()
5168 * dquot_initialize() down(dqio_mutex)
5169 * down(dqio_mutex) jbd2_journal_start()
5170 *
5171 */
5172
5173 #ifdef CONFIG_QUOTA
5174
5175 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5176 {
5177 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5178 }
5179
5180 static int ext4_write_dquot(struct dquot *dquot)
5181 {
5182 int ret, err;
5183 handle_t *handle;
5184 struct inode *inode;
5185
5186 inode = dquot_to_inode(dquot);
5187 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5188 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5189 if (IS_ERR(handle))
5190 return PTR_ERR(handle);
5191 ret = dquot_commit(dquot);
5192 err = ext4_journal_stop(handle);
5193 if (!ret)
5194 ret = err;
5195 return ret;
5196 }
5197
5198 static int ext4_acquire_dquot(struct dquot *dquot)
5199 {
5200 int ret, err;
5201 handle_t *handle;
5202
5203 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5204 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5205 if (IS_ERR(handle))
5206 return PTR_ERR(handle);
5207 ret = dquot_acquire(dquot);
5208 err = ext4_journal_stop(handle);
5209 if (!ret)
5210 ret = err;
5211 return ret;
5212 }
5213
5214 static int ext4_release_dquot(struct dquot *dquot)
5215 {
5216 int ret, err;
5217 handle_t *handle;
5218
5219 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5220 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5221 if (IS_ERR(handle)) {
5222 /* Release dquot anyway to avoid endless cycle in dqput() */
5223 dquot_release(dquot);
5224 return PTR_ERR(handle);
5225 }
5226 ret = dquot_release(dquot);
5227 err = ext4_journal_stop(handle);
5228 if (!ret)
5229 ret = err;
5230 return ret;
5231 }
5232
5233 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5234 {
5235 struct super_block *sb = dquot->dq_sb;
5236 struct ext4_sb_info *sbi = EXT4_SB(sb);
5237
5238 /* Are we journaling quotas? */
5239 if (ext4_has_feature_quota(sb) ||
5240 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5241 dquot_mark_dquot_dirty(dquot);
5242 return ext4_write_dquot(dquot);
5243 } else {
5244 return dquot_mark_dquot_dirty(dquot);
5245 }
5246 }
5247
5248 static int ext4_write_info(struct super_block *sb, int type)
5249 {
5250 int ret, err;
5251 handle_t *handle;
5252
5253 /* Data block + inode block */
5254 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5255 if (IS_ERR(handle))
5256 return PTR_ERR(handle);
5257 ret = dquot_commit_info(sb, type);
5258 err = ext4_journal_stop(handle);
5259 if (!ret)
5260 ret = err;
5261 return ret;
5262 }
5263
5264 /*
5265 * Turn on quotas during mount time - we need to find
5266 * the quota file and such...
5267 */
5268 static int ext4_quota_on_mount(struct super_block *sb, int type)
5269 {
5270 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5271 EXT4_SB(sb)->s_jquota_fmt, type);
5272 }
5273
5274 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5275 {
5276 struct ext4_inode_info *ei = EXT4_I(inode);
5277
5278 /* The first argument of lockdep_set_subclass has to be
5279 * *exactly* the same as the argument to init_rwsem() --- in
5280 * this case, in init_once() --- or lockdep gets unhappy
5281 * because the name of the lock is set using the
5282 * stringification of the argument to init_rwsem().
5283 */
5284 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5285 lockdep_set_subclass(&ei->i_data_sem, subclass);
5286 }
5287
5288 /*
5289 * Standard function to be called on quota_on
5290 */
5291 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5292 const struct path *path)
5293 {
5294 int err;
5295
5296 if (!test_opt(sb, QUOTA))
5297 return -EINVAL;
5298
5299 /* Quotafile not on the same filesystem? */
5300 if (path->dentry->d_sb != sb)
5301 return -EXDEV;
5302 /* Journaling quota? */
5303 if (EXT4_SB(sb)->s_qf_names[type]) {
5304 /* Quotafile not in fs root? */
5305 if (path->dentry->d_parent != sb->s_root)
5306 ext4_msg(sb, KERN_WARNING,
5307 "Quota file not on filesystem root. "
5308 "Journaled quota will not work");
5309 }
5310
5311 /*
5312 * When we journal data on quota file, we have to flush journal to see
5313 * all updates to the file when we bypass pagecache...
5314 */
5315 if (EXT4_SB(sb)->s_journal &&
5316 ext4_should_journal_data(d_inode(path->dentry))) {
5317 /*
5318 * We don't need to lock updates but journal_flush() could
5319 * otherwise be livelocked...
5320 */
5321 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5322 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5323 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5324 if (err)
5325 return err;
5326 }
5327 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5328 err = dquot_quota_on(sb, type, format_id, path);
5329 if (err)
5330 lockdep_set_quota_inode(path->dentry->d_inode,
5331 I_DATA_SEM_NORMAL);
5332 return err;
5333 }
5334
5335 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5336 unsigned int flags)
5337 {
5338 int err;
5339 struct inode *qf_inode;
5340 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5341 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5342 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5343 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5344 };
5345
5346 BUG_ON(!ext4_has_feature_quota(sb));
5347
5348 if (!qf_inums[type])
5349 return -EPERM;
5350
5351 qf_inode = ext4_iget(sb, qf_inums[type]);
5352 if (IS_ERR(qf_inode)) {
5353 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5354 return PTR_ERR(qf_inode);
5355 }
5356
5357 /* Don't account quota for quota files to avoid recursion */
5358 qf_inode->i_flags |= S_NOQUOTA;
5359 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5360 err = dquot_enable(qf_inode, type, format_id, flags);
5361 iput(qf_inode);
5362 if (err)
5363 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5364
5365 return err;
5366 }
5367
5368 /* Enable usage tracking for all quota types. */
5369 static int ext4_enable_quotas(struct super_block *sb)
5370 {
5371 int type, err = 0;
5372 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5373 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5374 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5375 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5376 };
5377 bool quota_mopt[EXT4_MAXQUOTAS] = {
5378 test_opt(sb, USRQUOTA),
5379 test_opt(sb, GRPQUOTA),
5380 test_opt(sb, PRJQUOTA),
5381 };
5382
5383 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5384 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5385 if (qf_inums[type]) {
5386 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5387 DQUOT_USAGE_ENABLED |
5388 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5389 if (err) {
5390 ext4_warning(sb,
5391 "Failed to enable quota tracking "
5392 "(type=%d, err=%d). Please run "
5393 "e2fsck to fix.", type, err);
5394 return err;
5395 }
5396 }
5397 }
5398 return 0;
5399 }
5400
5401 static int ext4_quota_off(struct super_block *sb, int type)
5402 {
5403 struct inode *inode = sb_dqopt(sb)->files[type];
5404 handle_t *handle;
5405
5406 /* Force all delayed allocation blocks to be allocated.
5407 * Caller already holds s_umount sem */
5408 if (test_opt(sb, DELALLOC))
5409 sync_filesystem(sb);
5410
5411 if (!inode)
5412 goto out;
5413
5414 /* Update modification times of quota files when userspace can
5415 * start looking at them */
5416 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5417 if (IS_ERR(handle))
5418 goto out;
5419 inode->i_mtime = inode->i_ctime = current_time(inode);
5420 ext4_mark_inode_dirty(handle, inode);
5421 ext4_journal_stop(handle);
5422
5423 out:
5424 return dquot_quota_off(sb, type);
5425 }
5426
5427 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5428 * acquiring the locks... As quota files are never truncated and quota code
5429 * itself serializes the operations (and no one else should touch the files)
5430 * we don't have to be afraid of races */
5431 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5432 size_t len, loff_t off)
5433 {
5434 struct inode *inode = sb_dqopt(sb)->files[type];
5435 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5436 int offset = off & (sb->s_blocksize - 1);
5437 int tocopy;
5438 size_t toread;
5439 struct buffer_head *bh;
5440 loff_t i_size = i_size_read(inode);
5441
5442 if (off > i_size)
5443 return 0;
5444 if (off+len > i_size)
5445 len = i_size-off;
5446 toread = len;
5447 while (toread > 0) {
5448 tocopy = sb->s_blocksize - offset < toread ?
5449 sb->s_blocksize - offset : toread;
5450 bh = ext4_bread(NULL, inode, blk, 0);
5451 if (IS_ERR(bh))
5452 return PTR_ERR(bh);
5453 if (!bh) /* A hole? */
5454 memset(data, 0, tocopy);
5455 else
5456 memcpy(data, bh->b_data+offset, tocopy);
5457 brelse(bh);
5458 offset = 0;
5459 toread -= tocopy;
5460 data += tocopy;
5461 blk++;
5462 }
5463 return len;
5464 }
5465
5466 /* Write to quotafile (we know the transaction is already started and has
5467 * enough credits) */
5468 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5469 const char *data, size_t len, loff_t off)
5470 {
5471 struct inode *inode = sb_dqopt(sb)->files[type];
5472 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5473 int err, offset = off & (sb->s_blocksize - 1);
5474 int retries = 0;
5475 struct buffer_head *bh;
5476 handle_t *handle = journal_current_handle();
5477
5478 if (EXT4_SB(sb)->s_journal && !handle) {
5479 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5480 " cancelled because transaction is not started",
5481 (unsigned long long)off, (unsigned long long)len);
5482 return -EIO;
5483 }
5484 /*
5485 * Since we account only one data block in transaction credits,
5486 * then it is impossible to cross a block boundary.
5487 */
5488 if (sb->s_blocksize - offset < len) {
5489 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5490 " cancelled because not block aligned",
5491 (unsigned long long)off, (unsigned long long)len);
5492 return -EIO;
5493 }
5494
5495 do {
5496 bh = ext4_bread(handle, inode, blk,
5497 EXT4_GET_BLOCKS_CREATE |
5498 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5499 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5500 ext4_should_retry_alloc(inode->i_sb, &retries));
5501 if (IS_ERR(bh))
5502 return PTR_ERR(bh);
5503 if (!bh)
5504 goto out;
5505 BUFFER_TRACE(bh, "get write access");
5506 err = ext4_journal_get_write_access(handle, bh);
5507 if (err) {
5508 brelse(bh);
5509 return err;
5510 }
5511 lock_buffer(bh);
5512 memcpy(bh->b_data+offset, data, len);
5513 flush_dcache_page(bh->b_page);
5514 unlock_buffer(bh);
5515 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5516 brelse(bh);
5517 out:
5518 if (inode->i_size < off + len) {
5519 i_size_write(inode, off + len);
5520 EXT4_I(inode)->i_disksize = inode->i_size;
5521 ext4_mark_inode_dirty(handle, inode);
5522 }
5523 return len;
5524 }
5525
5526 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5527 {
5528 const struct quota_format_ops *ops;
5529
5530 if (!sb_has_quota_loaded(sb, qid->type))
5531 return -ESRCH;
5532 ops = sb_dqopt(sb)->ops[qid->type];
5533 if (!ops || !ops->get_next_id)
5534 return -ENOSYS;
5535 return dquot_get_next_id(sb, qid);
5536 }
5537 #endif
5538
5539 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5540 const char *dev_name, void *data)
5541 {
5542 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5543 }
5544
5545 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5546 static inline void register_as_ext2(void)
5547 {
5548 int err = register_filesystem(&ext2_fs_type);
5549 if (err)
5550 printk(KERN_WARNING
5551 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5552 }
5553
5554 static inline void unregister_as_ext2(void)
5555 {
5556 unregister_filesystem(&ext2_fs_type);
5557 }
5558
5559 static inline int ext2_feature_set_ok(struct super_block *sb)
5560 {
5561 if (ext4_has_unknown_ext2_incompat_features(sb))
5562 return 0;
5563 if (sb->s_flags & MS_RDONLY)
5564 return 1;
5565 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5566 return 0;
5567 return 1;
5568 }
5569 #else
5570 static inline void register_as_ext2(void) { }
5571 static inline void unregister_as_ext2(void) { }
5572 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5573 #endif
5574
5575 static inline void register_as_ext3(void)
5576 {
5577 int err = register_filesystem(&ext3_fs_type);
5578 if (err)
5579 printk(KERN_WARNING
5580 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5581 }
5582
5583 static inline void unregister_as_ext3(void)
5584 {
5585 unregister_filesystem(&ext3_fs_type);
5586 }
5587
5588 static inline int ext3_feature_set_ok(struct super_block *sb)
5589 {
5590 if (ext4_has_unknown_ext3_incompat_features(sb))
5591 return 0;
5592 if (!ext4_has_feature_journal(sb))
5593 return 0;
5594 if (sb->s_flags & MS_RDONLY)
5595 return 1;
5596 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5597 return 0;
5598 return 1;
5599 }
5600
5601 static struct file_system_type ext4_fs_type = {
5602 .owner = THIS_MODULE,
5603 .name = "ext4",
5604 .mount = ext4_mount,
5605 .kill_sb = kill_block_super,
5606 .fs_flags = FS_REQUIRES_DEV,
5607 };
5608 MODULE_ALIAS_FS("ext4");
5609
5610 /* Shared across all ext4 file systems */
5611 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5612
5613 static int __init ext4_init_fs(void)
5614 {
5615 int i, err;
5616
5617 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5618 ext4_li_info = NULL;
5619 mutex_init(&ext4_li_mtx);
5620
5621 /* Build-time check for flags consistency */
5622 ext4_check_flag_values();
5623
5624 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5625 init_waitqueue_head(&ext4__ioend_wq[i]);
5626
5627 err = ext4_init_es();
5628 if (err)
5629 return err;
5630
5631 err = ext4_init_pageio();
5632 if (err)
5633 goto out5;
5634
5635 err = ext4_init_system_zone();
5636 if (err)
5637 goto out4;
5638
5639 err = ext4_init_sysfs();
5640 if (err)
5641 goto out3;
5642
5643 err = ext4_init_mballoc();
5644 if (err)
5645 goto out2;
5646 err = init_inodecache();
5647 if (err)
5648 goto out1;
5649 register_as_ext3();
5650 register_as_ext2();
5651 err = register_filesystem(&ext4_fs_type);
5652 if (err)
5653 goto out;
5654
5655 return 0;
5656 out:
5657 unregister_as_ext2();
5658 unregister_as_ext3();
5659 destroy_inodecache();
5660 out1:
5661 ext4_exit_mballoc();
5662 out2:
5663 ext4_exit_sysfs();
5664 out3:
5665 ext4_exit_system_zone();
5666 out4:
5667 ext4_exit_pageio();
5668 out5:
5669 ext4_exit_es();
5670
5671 return err;
5672 }
5673
5674 static void __exit ext4_exit_fs(void)
5675 {
5676 ext4_destroy_lazyinit_thread();
5677 unregister_as_ext2();
5678 unregister_as_ext3();
5679 unregister_filesystem(&ext4_fs_type);
5680 destroy_inodecache();
5681 ext4_exit_mballoc();
5682 ext4_exit_sysfs();
5683 ext4_exit_system_zone();
5684 ext4_exit_pageio();
5685 ext4_exit_es();
5686 }
5687
5688 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5689 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5690 MODULE_LICENSE("GPL");
5691 module_init(ext4_init_fs)
5692 module_exit(ext4_exit_fs)
ubuntu-bionic-kernel mirror