?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
HID: hiddev: Fix slab-out-of-bounds write in hiddev_ioctl_usage()
[linux/fpc-iii.git] / fs / ext4 / super.c
blobaca086a25b2ef4e08d2de31241d4619f6ac6631d
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <asm/uaccess.h>
42
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
45
46 #include "ext4.h"
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
49 #include "xattr.h"
50 #include "acl.h"
51 #include "mballoc.h"
52
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
55
56 static struct ext4_lazy_init *ext4_li_info;
57 static struct mutex ext4_li_mtx;
58 static int ext4_mballoc_ready;
59 static struct ratelimit_state ext4_mount_msg_ratelimit;
60
61 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
62 unsigned long journal_devnum);
63 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
64 static int ext4_commit_super(struct super_block *sb, int sync);
65 static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67 static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
69 static int ext4_sync_fs(struct super_block *sb, int wait);
70 static int ext4_remount(struct super_block *sb, int *flags, char *data);
71 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
72 static int ext4_unfreeze(struct super_block *sb);
73 static int ext4_freeze(struct super_block *sb);
74 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
75 const char *dev_name, void *data);
76 static inline int ext2_feature_set_ok(struct super_block *sb);
77 static inline int ext3_feature_set_ok(struct super_block *sb);
78 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
79 static void ext4_destroy_lazyinit_thread(void);
80 static void ext4_unregister_li_request(struct super_block *sb);
81 static void ext4_clear_request_list(void);
82
83 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
84 static struct file_system_type ext2_fs_type = {
85 .owner = THIS_MODULE,
86 .name = "ext2",
87 .mount = ext4_mount,
88 .kill_sb = kill_block_super,
89 .fs_flags = FS_REQUIRES_DEV,
90 };
91 MODULE_ALIAS_FS("ext2");
92 MODULE_ALIAS("ext2");
93 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
94 #else
95 #define IS_EXT2_SB(sb) (0)
96 #endif
97
98
99 static struct file_system_type ext3_fs_type = {
100 .owner = THIS_MODULE,
101 .name = "ext3",
102 .mount = ext4_mount,
103 .kill_sb = kill_block_super,
104 .fs_flags = FS_REQUIRES_DEV,
105 };
106 MODULE_ALIAS_FS("ext3");
107 MODULE_ALIAS("ext3");
108 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
109
110 static int ext4_verify_csum_type(struct super_block *sb,
111 struct ext4_super_block *es)
112 {
113 if (!ext4_has_feature_metadata_csum(sb))
114 return 1;
115
116 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
117 }
118
119 static __le32 ext4_superblock_csum(struct super_block *sb,
120 struct ext4_super_block *es)
121 {
122 struct ext4_sb_info *sbi = EXT4_SB(sb);
123 int offset = offsetof(struct ext4_super_block, s_checksum);
124 __u32 csum;
125
126 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
127
128 return cpu_to_le32(csum);
129 }
130
131 static int ext4_superblock_csum_verify(struct super_block *sb,
132 struct ext4_super_block *es)
133 {
134 if (!ext4_has_metadata_csum(sb))
135 return 1;
136
137 return es->s_checksum == ext4_superblock_csum(sb, es);
138 }
139
140 void ext4_superblock_csum_set(struct super_block *sb)
141 {
142 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
143
144 if (!ext4_has_metadata_csum(sb))
145 return;
146
147 es->s_checksum = ext4_superblock_csum(sb, es);
148 }
149
150 void *ext4_kvmalloc(size_t size, gfp_t flags)
151 {
152 void *ret;
153
154 ret = kmalloc(size, flags | __GFP_NOWARN);
155 if (!ret)
156 ret = __vmalloc(size, flags, PAGE_KERNEL);
157 return ret;
158 }
159
160 void *ext4_kvzalloc(size_t size, gfp_t flags)
161 {
162 void *ret;
163
164 ret = kzalloc(size, flags | __GFP_NOWARN);
165 if (!ret)
166 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
167 return ret;
168 }
169
170 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
171 struct ext4_group_desc *bg)
172 {
173 return le32_to_cpu(bg->bg_block_bitmap_lo) |
174 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
175 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
176 }
177
178 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
179 struct ext4_group_desc *bg)
180 {
181 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
182 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
183 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
184 }
185
186 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
187 struct ext4_group_desc *bg)
188 {
189 return le32_to_cpu(bg->bg_inode_table_lo) |
190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
192 }
193
194 __u32 ext4_free_group_clusters(struct super_block *sb,
195 struct ext4_group_desc *bg)
196 {
197 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
200 }
201
202 __u32 ext4_free_inodes_count(struct super_block *sb,
203 struct ext4_group_desc *bg)
204 {
205 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
208 }
209
210 __u32 ext4_used_dirs_count(struct super_block *sb,
211 struct ext4_group_desc *bg)
212 {
213 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
216 }
217
218 __u32 ext4_itable_unused_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220 {
221 return le16_to_cpu(bg->bg_itable_unused_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
224 }
225
226 void ext4_block_bitmap_set(struct super_block *sb,
227 struct ext4_group_desc *bg, ext4_fsblk_t blk)
228 {
229 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
230 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
231 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
232 }
233
234 void ext4_inode_bitmap_set(struct super_block *sb,
235 struct ext4_group_desc *bg, ext4_fsblk_t blk)
236 {
237 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
238 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
240 }
241
242 void ext4_inode_table_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
244 {
245 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
248 }
249
250 void ext4_free_group_clusters_set(struct super_block *sb,
251 struct ext4_group_desc *bg, __u32 count)
252 {
253 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
256 }
257
258 void ext4_free_inodes_set(struct super_block *sb,
259 struct ext4_group_desc *bg, __u32 count)
260 {
261 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
264 }
265
266 void ext4_used_dirs_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
268 {
269 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
272 }
273
274 void ext4_itable_unused_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276 {
277 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
280 }
281
282
283 static void __save_error_info(struct super_block *sb, const char *func,
284 unsigned int line)
285 {
286 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
287
288 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
289 if (bdev_read_only(sb->s_bdev))
290 return;
291 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
292 es->s_last_error_time = cpu_to_le32(get_seconds());
293 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
294 es->s_last_error_line = cpu_to_le32(line);
295 if (!es->s_first_error_time) {
296 es->s_first_error_time = es->s_last_error_time;
297 strncpy(es->s_first_error_func, func,
298 sizeof(es->s_first_error_func));
299 es->s_first_error_line = cpu_to_le32(line);
300 es->s_first_error_ino = es->s_last_error_ino;
301 es->s_first_error_block = es->s_last_error_block;
302 }
303 /*
304 * Start the daily error reporting function if it hasn't been
305 * started already
306 */
307 if (!es->s_error_count)
308 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
309 le32_add_cpu(&es->s_error_count, 1);
310 }
311
312 static void save_error_info(struct super_block *sb, const char *func,
313 unsigned int line)
314 {
315 __save_error_info(sb, func, line);
316 if (!bdev_read_only(sb->s_bdev))
317 ext4_commit_super(sb, 1);
318 }
319
320 /*
321 * The del_gendisk() function uninitializes the disk-specific data
322 * structures, including the bdi structure, without telling anyone
323 * else. Once this happens, any attempt to call mark_buffer_dirty()
324 * (for example, by ext4_commit_super), will cause a kernel OOPS.
325 * This is a kludge to prevent these oops until we can put in a proper
326 * hook in del_gendisk() to inform the VFS and file system layers.
327 */
328 static int block_device_ejected(struct super_block *sb)
329 {
330 struct inode *bd_inode = sb->s_bdev->bd_inode;
331 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
332
333 return bdi->dev == NULL;
334 }
335
336 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
337 {
338 struct super_block *sb = journal->j_private;
339 struct ext4_sb_info *sbi = EXT4_SB(sb);
340 int error = is_journal_aborted(journal);
341 struct ext4_journal_cb_entry *jce;
342
343 BUG_ON(txn->t_state == T_FINISHED);
344 spin_lock(&sbi->s_md_lock);
345 while (!list_empty(&txn->t_private_list)) {
346 jce = list_entry(txn->t_private_list.next,
347 struct ext4_journal_cb_entry, jce_list);
348 list_del_init(&jce->jce_list);
349 spin_unlock(&sbi->s_md_lock);
350 jce->jce_func(sb, jce, error);
351 spin_lock(&sbi->s_md_lock);
352 }
353 spin_unlock(&sbi->s_md_lock);
354 }
355
356 /* Deal with the reporting of failure conditions on a filesystem such as
357 * inconsistencies detected or read IO failures.
358 *
359 * On ext2, we can store the error state of the filesystem in the
360 * superblock. That is not possible on ext4, because we may have other
361 * write ordering constraints on the superblock which prevent us from
362 * writing it out straight away; and given that the journal is about to
363 * be aborted, we can't rely on the current, or future, transactions to
364 * write out the superblock safely.
365 *
366 * We'll just use the jbd2_journal_abort() error code to record an error in
367 * the journal instead. On recovery, the journal will complain about
368 * that error until we've noted it down and cleared it.
369 */
370
371 static void ext4_handle_error(struct super_block *sb)
372 {
373 if (sb->s_flags & MS_RDONLY)
374 return;
375
376 if (!test_opt(sb, ERRORS_CONT)) {
377 journal_t *journal = EXT4_SB(sb)->s_journal;
378
379 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
380 if (journal)
381 jbd2_journal_abort(journal, -EIO);
382 }
383 if (test_opt(sb, ERRORS_RO)) {
384 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
385 /*
386 * Make sure updated value of ->s_mount_flags will be visible
387 * before ->s_flags update
388 */
389 smp_wmb();
390 sb->s_flags |= MS_RDONLY;
391 }
392 if (test_opt(sb, ERRORS_PANIC)) {
393 if (EXT4_SB(sb)->s_journal &&
394 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
395 return;
396 panic("EXT4-fs (device %s): panic forced after error\n",
397 sb->s_id);
398 }
399 }
400
401 #define ext4_error_ratelimit(sb) \
402 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
403 "EXT4-fs error")
404
405 void __ext4_error(struct super_block *sb, const char *function,
406 unsigned int line, const char *fmt, ...)
407 {
408 struct va_format vaf;
409 va_list args;
410
411 if (ext4_error_ratelimit(sb)) {
412 va_start(args, fmt);
413 vaf.fmt = fmt;
414 vaf.va = &args;
415 printk(KERN_CRIT
416 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
417 sb->s_id, function, line, current->comm, &vaf);
418 va_end(args);
419 }
420 save_error_info(sb, function, line);
421 ext4_handle_error(sb);
422 }
423
424 void __ext4_error_inode(struct inode *inode, const char *function,
425 unsigned int line, ext4_fsblk_t block,
426 const char *fmt, ...)
427 {
428 va_list args;
429 struct va_format vaf;
430 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
431
432 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
433 es->s_last_error_block = cpu_to_le64(block);
434 if (ext4_error_ratelimit(inode->i_sb)) {
435 va_start(args, fmt);
436 vaf.fmt = fmt;
437 vaf.va = &args;
438 if (block)
439 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
440 "inode #%lu: block %llu: comm %s: %pV\n",
441 inode->i_sb->s_id, function, line, inode->i_ino,
442 block, current->comm, &vaf);
443 else
444 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
445 "inode #%lu: comm %s: %pV\n",
446 inode->i_sb->s_id, function, line, inode->i_ino,
447 current->comm, &vaf);
448 va_end(args);
449 }
450 save_error_info(inode->i_sb, function, line);
451 ext4_handle_error(inode->i_sb);
452 }
453
454 void __ext4_error_file(struct file *file, 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;
461 struct inode *inode = file_inode(file);
462 char pathname[80], *path;
463
464 es = EXT4_SB(inode->i_sb)->s_es;
465 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
466 if (ext4_error_ratelimit(inode->i_sb)) {
467 path = file_path(file, pathname, sizeof(pathname));
468 if (IS_ERR(path))
469 path = "(unknown)";
470 va_start(args, fmt);
471 vaf.fmt = fmt;
472 vaf.va = &args;
473 if (block)
474 printk(KERN_CRIT
475 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
476 "block %llu: comm %s: path %s: %pV\n",
477 inode->i_sb->s_id, function, line, inode->i_ino,
478 block, current->comm, path, &vaf);
479 else
480 printk(KERN_CRIT
481 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
482 "comm %s: path %s: %pV\n",
483 inode->i_sb->s_id, function, line, inode->i_ino,
484 current->comm, path, &vaf);
485 va_end(args);
486 }
487 save_error_info(inode->i_sb, function, line);
488 ext4_handle_error(inode->i_sb);
489 }
490
491 const char *ext4_decode_error(struct super_block *sb, int errno,
492 char nbuf[16])
493 {
494 char *errstr = NULL;
495
496 switch (errno) {
497 case -EFSCORRUPTED:
498 errstr = "Corrupt filesystem";
499 break;
500 case -EFSBADCRC:
501 errstr = "Filesystem failed CRC";
502 break;
503 case -EIO:
504 errstr = "IO failure";
505 break;
506 case -ENOMEM:
507 errstr = "Out of memory";
508 break;
509 case -EROFS:
510 if (!sb || (EXT4_SB(sb)->s_journal &&
511 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
512 errstr = "Journal has aborted";
513 else
514 errstr = "Readonly filesystem";
515 break;
516 default:
517 /* If the caller passed in an extra buffer for unknown
518 * errors, textualise them now. Else we just return
519 * NULL. */
520 if (nbuf) {
521 /* Check for truncated error codes... */
522 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
523 errstr = nbuf;
524 }
525 break;
526 }
527
528 return errstr;
529 }
530
531 /* __ext4_std_error decodes expected errors from journaling functions
532 * automatically and invokes the appropriate error response. */
533
534 void __ext4_std_error(struct super_block *sb, const char *function,
535 unsigned int line, int errno)
536 {
537 char nbuf[16];
538 const char *errstr;
539
540 /* Special case: if the error is EROFS, and we're not already
541 * inside a transaction, then there's really no point in logging
542 * an error. */
543 if (errno == -EROFS && journal_current_handle() == NULL &&
544 (sb->s_flags & MS_RDONLY))
545 return;
546
547 if (ext4_error_ratelimit(sb)) {
548 errstr = ext4_decode_error(sb, errno, nbuf);
549 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
550 sb->s_id, function, line, errstr);
551 }
552
553 save_error_info(sb, function, line);
554 ext4_handle_error(sb);
555 }
556
557 /*
558 * ext4_abort is a much stronger failure handler than ext4_error. The
559 * abort function may be used to deal with unrecoverable failures such
560 * as journal IO errors or ENOMEM at a critical moment in log management.
561 *
562 * We unconditionally force the filesystem into an ABORT|READONLY state,
563 * unless the error response on the fs has been set to panic in which
564 * case we take the easy way out and panic immediately.
565 */
566
567 void __ext4_abort(struct super_block *sb, const char *function,
568 unsigned int line, const char *fmt, ...)
569 {
570 va_list args;
571
572 save_error_info(sb, function, line);
573 va_start(args, fmt);
574 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
575 function, line);
576 vprintk(fmt, args);
577 printk("\n");
578 va_end(args);
579
580 if ((sb->s_flags & MS_RDONLY) == 0) {
581 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
582 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
583 /*
584 * Make sure updated value of ->s_mount_flags will be visible
585 * before ->s_flags update
586 */
587 smp_wmb();
588 sb->s_flags |= MS_RDONLY;
589 if (EXT4_SB(sb)->s_journal)
590 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
591 save_error_info(sb, function, line);
592 }
593 if (test_opt(sb, ERRORS_PANIC)) {
594 if (EXT4_SB(sb)->s_journal &&
595 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
596 return;
597 panic("EXT4-fs panic from previous error\n");
598 }
599 }
600
601 void __ext4_msg(struct super_block *sb,
602 const char *prefix, const char *fmt, ...)
603 {
604 struct va_format vaf;
605 va_list args;
606
607 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
608 return;
609
610 va_start(args, fmt);
611 vaf.fmt = fmt;
612 vaf.va = &args;
613 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
614 va_end(args);
615 }
616
617 #define ext4_warning_ratelimit(sb) \
618 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
619 "EXT4-fs warning")
620
621 void __ext4_warning(struct super_block *sb, const char *function,
622 unsigned int line, const char *fmt, ...)
623 {
624 struct va_format vaf;
625 va_list args;
626
627 if (!ext4_warning_ratelimit(sb))
628 return;
629
630 va_start(args, fmt);
631 vaf.fmt = fmt;
632 vaf.va = &args;
633 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
634 sb->s_id, function, line, &vaf);
635 va_end(args);
636 }
637
638 void __ext4_warning_inode(const struct inode *inode, const char *function,
639 unsigned int line, const char *fmt, ...)
640 {
641 struct va_format vaf;
642 va_list args;
643
644 if (!ext4_warning_ratelimit(inode->i_sb))
645 return;
646
647 va_start(args, fmt);
648 vaf.fmt = fmt;
649 vaf.va = &args;
650 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
651 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
652 function, line, inode->i_ino, current->comm, &vaf);
653 va_end(args);
654 }
655
656 void __ext4_grp_locked_error(const char *function, unsigned int line,
657 struct super_block *sb, ext4_group_t grp,
658 unsigned long ino, ext4_fsblk_t block,
659 const char *fmt, ...)
660 __releases(bitlock)
661 __acquires(bitlock)
662 {
663 struct va_format vaf;
664 va_list args;
665 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
666
667 es->s_last_error_ino = cpu_to_le32(ino);
668 es->s_last_error_block = cpu_to_le64(block);
669 __save_error_info(sb, function, line);
670
671 if (ext4_error_ratelimit(sb)) {
672 va_start(args, fmt);
673 vaf.fmt = fmt;
674 vaf.va = &args;
675 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
676 sb->s_id, function, line, grp);
677 if (ino)
678 printk(KERN_CONT "inode %lu: ", ino);
679 if (block)
680 printk(KERN_CONT "block %llu:",
681 (unsigned long long) block);
682 printk(KERN_CONT "%pV\n", &vaf);
683 va_end(args);
684 }
685
686 if (test_opt(sb, ERRORS_CONT)) {
687 ext4_commit_super(sb, 0);
688 return;
689 }
690
691 ext4_unlock_group(sb, grp);
692 ext4_commit_super(sb, 1);
693 ext4_handle_error(sb);
694 /*
695 * We only get here in the ERRORS_RO case; relocking the group
696 * may be dangerous, but nothing bad will happen since the
697 * filesystem will have already been marked read/only and the
698 * journal has been aborted. We return 1 as a hint to callers
699 * who might what to use the return value from
700 * ext4_grp_locked_error() to distinguish between the
701 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
702 * aggressively from the ext4 function in question, with a
703 * more appropriate error code.
704 */
705 ext4_lock_group(sb, grp);
706 return;
707 }
708
709 void ext4_update_dynamic_rev(struct super_block *sb)
710 {
711 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
712
713 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
714 return;
715
716 ext4_warning(sb,
717 "updating to rev %d because of new feature flag, "
718 "running e2fsck is recommended",
719 EXT4_DYNAMIC_REV);
720
721 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
722 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
723 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
724 /* leave es->s_feature_*compat flags alone */
725 /* es->s_uuid will be set by e2fsck if empty */
726
727 /*
728 * The rest of the superblock fields should be zero, and if not it
729 * means they are likely already in use, so leave them alone. We
730 * can leave it up to e2fsck to clean up any inconsistencies there.
731 */
732 }
733
734 /*
735 * Open the external journal device
736 */
737 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
738 {
739 struct block_device *bdev;
740 char b[BDEVNAME_SIZE];
741
742 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
743 if (IS_ERR(bdev))
744 goto fail;
745 return bdev;
746
747 fail:
748 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
749 __bdevname(dev, b), PTR_ERR(bdev));
750 return NULL;
751 }
752
753 /*
754 * Release the journal device
755 */
756 static void ext4_blkdev_put(struct block_device *bdev)
757 {
758 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
759 }
760
761 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
762 {
763 struct block_device *bdev;
764 bdev = sbi->journal_bdev;
765 if (bdev) {
766 ext4_blkdev_put(bdev);
767 sbi->journal_bdev = NULL;
768 }
769 }
770
771 static inline struct inode *orphan_list_entry(struct list_head *l)
772 {
773 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
774 }
775
776 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
777 {
778 struct list_head *l;
779
780 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
781 le32_to_cpu(sbi->s_es->s_last_orphan));
782
783 printk(KERN_ERR "sb_info orphan list:\n");
784 list_for_each(l, &sbi->s_orphan) {
785 struct inode *inode = orphan_list_entry(l);
786 printk(KERN_ERR " "
787 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
788 inode->i_sb->s_id, inode->i_ino, inode,
789 inode->i_mode, inode->i_nlink,
790 NEXT_ORPHAN(inode));
791 }
792 }
793
794 static void ext4_put_super(struct super_block *sb)
795 {
796 struct ext4_sb_info *sbi = EXT4_SB(sb);
797 struct ext4_super_block *es = sbi->s_es;
798 struct buffer_head **group_desc;
799 struct flex_groups **flex_groups;
800 int aborted = 0;
801 int i, err;
802
803 ext4_unregister_li_request(sb);
804 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
805
806 flush_workqueue(sbi->rsv_conversion_wq);
807 destroy_workqueue(sbi->rsv_conversion_wq);
808
809 if (sbi->s_journal) {
810 aborted = is_journal_aborted(sbi->s_journal);
811 err = jbd2_journal_destroy(sbi->s_journal);
812 sbi->s_journal = NULL;
813 if ((err < 0) && !aborted)
814 ext4_abort(sb, "Couldn't clean up the journal");
815 }
816
817 ext4_unregister_sysfs(sb);
818 ext4_es_unregister_shrinker(sbi);
819 del_timer_sync(&sbi->s_err_report);
820 ext4_release_system_zone(sb);
821 ext4_mb_release(sb);
822 ext4_ext_release(sb);
823 ext4_xattr_put_super(sb);
824
825 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
826 ext4_clear_feature_journal_needs_recovery(sb);
827 es->s_state = cpu_to_le16(sbi->s_mount_state);
828 }
829 if (!(sb->s_flags & MS_RDONLY))
830 ext4_commit_super(sb, 1);
831
832 rcu_read_lock();
833 group_desc = rcu_dereference(sbi->s_group_desc);
834 for (i = 0; i < sbi->s_gdb_count; i++)
835 brelse(group_desc[i]);
836 kvfree(group_desc);
837 flex_groups = rcu_dereference(sbi->s_flex_groups);
838 if (flex_groups) {
839 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
840 kvfree(flex_groups[i]);
841 kvfree(flex_groups);
842 }
843 rcu_read_unlock();
844 percpu_counter_destroy(&sbi->s_freeclusters_counter);
845 percpu_counter_destroy(&sbi->s_freeinodes_counter);
846 percpu_counter_destroy(&sbi->s_dirs_counter);
847 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
848 brelse(sbi->s_sbh);
849 #ifdef CONFIG_QUOTA
850 for (i = 0; i < EXT4_MAXQUOTAS; i++)
851 kfree(sbi->s_qf_names[i]);
852 #endif
853
854 /* Debugging code just in case the in-memory inode orphan list
855 * isn't empty. The on-disk one can be non-empty if we've
856 * detected an error and taken the fs readonly, but the
857 * in-memory list had better be clean by this point. */
858 if (!list_empty(&sbi->s_orphan))
859 dump_orphan_list(sb, sbi);
860 J_ASSERT(list_empty(&sbi->s_orphan));
861
862 sync_blockdev(sb->s_bdev);
863 invalidate_bdev(sb->s_bdev);
864 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
865 /*
866 * Invalidate the journal device's buffers. We don't want them
867 * floating about in memory - the physical journal device may
868 * hotswapped, and it breaks the `ro-after' testing code.
869 */
870 sync_blockdev(sbi->journal_bdev);
871 invalidate_bdev(sbi->journal_bdev);
872 ext4_blkdev_remove(sbi);
873 }
874 if (sbi->s_mb_cache) {
875 ext4_xattr_destroy_cache(sbi->s_mb_cache);
876 sbi->s_mb_cache = NULL;
877 }
878 if (sbi->s_mmp_tsk)
879 kthread_stop(sbi->s_mmp_tsk);
880 sb->s_fs_info = NULL;
881 /*
882 * Now that we are completely done shutting down the
883 * superblock, we need to actually destroy the kobject.
884 */
885 kobject_put(&sbi->s_kobj);
886 wait_for_completion(&sbi->s_kobj_unregister);
887 if (sbi->s_chksum_driver)
888 crypto_free_shash(sbi->s_chksum_driver);
889 kfree(sbi->s_blockgroup_lock);
890 kfree(sbi);
891 }
892
893 static struct kmem_cache *ext4_inode_cachep;
894
895 /*
896 * Called inside transaction, so use GFP_NOFS
897 */
898 static struct inode *ext4_alloc_inode(struct super_block *sb)
899 {
900 struct ext4_inode_info *ei;
901
902 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
903 if (!ei)
904 return NULL;
905
906 ei->vfs_inode.i_version = 1;
907 spin_lock_init(&ei->i_raw_lock);
908 INIT_LIST_HEAD(&ei->i_prealloc_list);
909 spin_lock_init(&ei->i_prealloc_lock);
910 ext4_es_init_tree(&ei->i_es_tree);
911 rwlock_init(&ei->i_es_lock);
912 INIT_LIST_HEAD(&ei->i_es_list);
913 ei->i_es_all_nr = 0;
914 ei->i_es_shk_nr = 0;
915 ei->i_es_shrink_lblk = 0;
916 ei->i_reserved_data_blocks = 0;
917 ei->i_reserved_meta_blocks = 0;
918 ei->i_allocated_meta_blocks = 0;
919 ei->i_da_metadata_calc_len = 0;
920 ei->i_da_metadata_calc_last_lblock = 0;
921 spin_lock_init(&(ei->i_block_reservation_lock));
922 #ifdef CONFIG_QUOTA
923 ei->i_reserved_quota = 0;
924 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
925 #endif
926 ei->jinode = NULL;
927 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
928 spin_lock_init(&ei->i_completed_io_lock);
929 ei->i_sync_tid = 0;
930 ei->i_datasync_tid = 0;
931 atomic_set(&ei->i_ioend_count, 0);
932 atomic_set(&ei->i_unwritten, 0);
933 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
934 #ifdef CONFIG_EXT4_FS_ENCRYPTION
935 ei->i_crypt_info = NULL;
936 #endif
937 return &ei->vfs_inode;
938 }
939
940 static int ext4_drop_inode(struct inode *inode)
941 {
942 int drop = generic_drop_inode(inode);
943
944 trace_ext4_drop_inode(inode, drop);
945 return drop;
946 }
947
948 static void ext4_i_callback(struct rcu_head *head)
949 {
950 struct inode *inode = container_of(head, struct inode, i_rcu);
951 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
952 }
953
954 static void ext4_destroy_inode(struct inode *inode)
955 {
956 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
957 ext4_msg(inode->i_sb, KERN_ERR,
958 "Inode %lu (%p): orphan list check failed!",
959 inode->i_ino, EXT4_I(inode));
960 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
961 EXT4_I(inode), sizeof(struct ext4_inode_info),
962 true);
963 dump_stack();
964 }
965 call_rcu(&inode->i_rcu, ext4_i_callback);
966 }
967
968 static void init_once(void *foo)
969 {
970 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
971
972 INIT_LIST_HEAD(&ei->i_orphan);
973 init_rwsem(&ei->xattr_sem);
974 init_rwsem(&ei->i_data_sem);
975 init_rwsem(&ei->i_mmap_sem);
976 inode_init_once(&ei->vfs_inode);
977 }
978
979 static int __init init_inodecache(void)
980 {
981 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
982 sizeof(struct ext4_inode_info),
983 0, (SLAB_RECLAIM_ACCOUNT|
984 SLAB_MEM_SPREAD),
985 init_once);
986 if (ext4_inode_cachep == NULL)
987 return -ENOMEM;
988 return 0;
989 }
990
991 static void destroy_inodecache(void)
992 {
993 /*
994 * Make sure all delayed rcu free inodes are flushed before we
995 * destroy cache.
996 */
997 rcu_barrier();
998 kmem_cache_destroy(ext4_inode_cachep);
999 }
1000
1001 void ext4_clear_inode(struct inode *inode)
1002 {
1003 invalidate_inode_buffers(inode);
1004 clear_inode(inode);
1005 dquot_drop(inode);
1006 ext4_discard_preallocations(inode);
1007 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1008 if (EXT4_I(inode)->jinode) {
1009 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1010 EXT4_I(inode)->jinode);
1011 jbd2_free_inode(EXT4_I(inode)->jinode);
1012 EXT4_I(inode)->jinode = NULL;
1013 }
1014 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1015 if (EXT4_I(inode)->i_crypt_info)
1016 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
1017 #endif
1018 }
1019
1020 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1021 u64 ino, u32 generation)
1022 {
1023 struct inode *inode;
1024
1025 /*
1026 * Currently we don't know the generation for parent directory, so
1027 * a generation of 0 means "accept any"
1028 */
1029 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1030 if (IS_ERR(inode))
1031 return ERR_CAST(inode);
1032 if (generation && inode->i_generation != generation) {
1033 iput(inode);
1034 return ERR_PTR(-ESTALE);
1035 }
1036
1037 return inode;
1038 }
1039
1040 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1041 int fh_len, int fh_type)
1042 {
1043 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1044 ext4_nfs_get_inode);
1045 }
1046
1047 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1048 int fh_len, int fh_type)
1049 {
1050 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1051 ext4_nfs_get_inode);
1052 }
1053
1054 static int ext4_nfs_commit_metadata(struct inode *inode)
1055 {
1056 struct writeback_control wbc = {
1057 .sync_mode = WB_SYNC_ALL
1058 };
1059
1060 trace_ext4_nfs_commit_metadata(inode);
1061 return ext4_write_inode(inode, &wbc);
1062 }
1063
1064 /*
1065 * Try to release metadata pages (indirect blocks, directories) which are
1066 * mapped via the block device. Since these pages could have journal heads
1067 * which would prevent try_to_free_buffers() from freeing them, we must use
1068 * jbd2 layer's try_to_free_buffers() function to release them.
1069 */
1070 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1071 gfp_t wait)
1072 {
1073 journal_t *journal = EXT4_SB(sb)->s_journal;
1074
1075 WARN_ON(PageChecked(page));
1076 if (!page_has_buffers(page))
1077 return 0;
1078 if (journal)
1079 return jbd2_journal_try_to_free_buffers(journal, page,
1080 wait & ~__GFP_DIRECT_RECLAIM);
1081 return try_to_free_buffers(page);
1082 }
1083
1084 #ifdef CONFIG_QUOTA
1085 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1086 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1087
1088 static int ext4_write_dquot(struct dquot *dquot);
1089 static int ext4_acquire_dquot(struct dquot *dquot);
1090 static int ext4_release_dquot(struct dquot *dquot);
1091 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1092 static int ext4_write_info(struct super_block *sb, int type);
1093 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1094 struct path *path);
1095 static int ext4_quota_off(struct super_block *sb, int type);
1096 static int ext4_quota_on_mount(struct super_block *sb, int type);
1097 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1098 size_t len, loff_t off);
1099 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1100 const char *data, size_t len, loff_t off);
1101 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1102 unsigned int flags);
1103 static int ext4_enable_quotas(struct super_block *sb);
1104
1105 static struct dquot **ext4_get_dquots(struct inode *inode)
1106 {
1107 return EXT4_I(inode)->i_dquot;
1108 }
1109
1110 static const struct dquot_operations ext4_quota_operations = {
1111 .get_reserved_space = ext4_get_reserved_space,
1112 .write_dquot = ext4_write_dquot,
1113 .acquire_dquot = ext4_acquire_dquot,
1114 .release_dquot = ext4_release_dquot,
1115 .mark_dirty = ext4_mark_dquot_dirty,
1116 .write_info = ext4_write_info,
1117 .alloc_dquot = dquot_alloc,
1118 .destroy_dquot = dquot_destroy,
1119 };
1120
1121 static const struct quotactl_ops ext4_qctl_operations = {
1122 .quota_on = ext4_quota_on,
1123 .quota_off = ext4_quota_off,
1124 .quota_sync = dquot_quota_sync,
1125 .get_state = dquot_get_state,
1126 .set_info = dquot_set_dqinfo,
1127 .get_dqblk = dquot_get_dqblk,
1128 .set_dqblk = dquot_set_dqblk
1129 };
1130 #endif
1131
1132 static const struct super_operations ext4_sops = {
1133 .alloc_inode = ext4_alloc_inode,
1134 .destroy_inode = ext4_destroy_inode,
1135 .write_inode = ext4_write_inode,
1136 .dirty_inode = ext4_dirty_inode,
1137 .drop_inode = ext4_drop_inode,
1138 .evict_inode = ext4_evict_inode,
1139 .put_super = ext4_put_super,
1140 .sync_fs = ext4_sync_fs,
1141 .freeze_fs = ext4_freeze,
1142 .unfreeze_fs = ext4_unfreeze,
1143 .statfs = ext4_statfs,
1144 .remount_fs = ext4_remount,
1145 .show_options = ext4_show_options,
1146 #ifdef CONFIG_QUOTA
1147 .quota_read = ext4_quota_read,
1148 .quota_write = ext4_quota_write,
1149 .get_dquots = ext4_get_dquots,
1150 #endif
1151 .bdev_try_to_free_page = bdev_try_to_free_page,
1152 };
1153
1154 static const struct export_operations ext4_export_ops = {
1155 .fh_to_dentry = ext4_fh_to_dentry,
1156 .fh_to_parent = ext4_fh_to_parent,
1157 .get_parent = ext4_get_parent,
1158 .commit_metadata = ext4_nfs_commit_metadata,
1159 };
1160
1161 enum {
1162 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1163 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1164 Opt_nouid32, Opt_debug, Opt_removed,
1165 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1166 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1167 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1168 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1169 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1170 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1171 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1172 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1173 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1174 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1175 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1176 Opt_lazytime, Opt_nolazytime,
1177 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1178 Opt_inode_readahead_blks, Opt_journal_ioprio,
1179 Opt_dioread_nolock, Opt_dioread_lock,
1180 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1181 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1182 };
1183
1184 static const match_table_t tokens = {
1185 {Opt_bsd_df, "bsddf"},
1186 {Opt_minix_df, "minixdf"},
1187 {Opt_grpid, "grpid"},
1188 {Opt_grpid, "bsdgroups"},
1189 {Opt_nogrpid, "nogrpid"},
1190 {Opt_nogrpid, "sysvgroups"},
1191 {Opt_resgid, "resgid=%u"},
1192 {Opt_resuid, "resuid=%u"},
1193 {Opt_sb, "sb=%u"},
1194 {Opt_err_cont, "errors=continue"},
1195 {Opt_err_panic, "errors=panic"},
1196 {Opt_err_ro, "errors=remount-ro"},
1197 {Opt_nouid32, "nouid32"},
1198 {Opt_debug, "debug"},
1199 {Opt_removed, "oldalloc"},
1200 {Opt_removed, "orlov"},
1201 {Opt_user_xattr, "user_xattr"},
1202 {Opt_nouser_xattr, "nouser_xattr"},
1203 {Opt_acl, "acl"},
1204 {Opt_noacl, "noacl"},
1205 {Opt_noload, "norecovery"},
1206 {Opt_noload, "noload"},
1207 {Opt_removed, "nobh"},
1208 {Opt_removed, "bh"},
1209 {Opt_commit, "commit=%u"},
1210 {Opt_min_batch_time, "min_batch_time=%u"},
1211 {Opt_max_batch_time, "max_batch_time=%u"},
1212 {Opt_journal_dev, "journal_dev=%u"},
1213 {Opt_journal_path, "journal_path=%s"},
1214 {Opt_journal_checksum, "journal_checksum"},
1215 {Opt_nojournal_checksum, "nojournal_checksum"},
1216 {Opt_journal_async_commit, "journal_async_commit"},
1217 {Opt_abort, "abort"},
1218 {Opt_data_journal, "data=journal"},
1219 {Opt_data_ordered, "data=ordered"},
1220 {Opt_data_writeback, "data=writeback"},
1221 {Opt_data_err_abort, "data_err=abort"},
1222 {Opt_data_err_ignore, "data_err=ignore"},
1223 {Opt_offusrjquota, "usrjquota="},
1224 {Opt_usrjquota, "usrjquota=%s"},
1225 {Opt_offgrpjquota, "grpjquota="},
1226 {Opt_grpjquota, "grpjquota=%s"},
1227 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1228 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1229 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1230 {Opt_grpquota, "grpquota"},
1231 {Opt_noquota, "noquota"},
1232 {Opt_quota, "quota"},
1233 {Opt_usrquota, "usrquota"},
1234 {Opt_barrier, "barrier=%u"},
1235 {Opt_barrier, "barrier"},
1236 {Opt_nobarrier, "nobarrier"},
1237 {Opt_i_version, "i_version"},
1238 {Opt_dax, "dax"},
1239 {Opt_stripe, "stripe=%u"},
1240 {Opt_delalloc, "delalloc"},
1241 {Opt_lazytime, "lazytime"},
1242 {Opt_nolazytime, "nolazytime"},
1243 {Opt_nodelalloc, "nodelalloc"},
1244 {Opt_removed, "mblk_io_submit"},
1245 {Opt_removed, "nomblk_io_submit"},
1246 {Opt_block_validity, "block_validity"},
1247 {Opt_noblock_validity, "noblock_validity"},
1248 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1249 {Opt_journal_ioprio, "journal_ioprio=%u"},
1250 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1251 {Opt_auto_da_alloc, "auto_da_alloc"},
1252 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1253 {Opt_dioread_nolock, "dioread_nolock"},
1254 {Opt_dioread_lock, "dioread_lock"},
1255 {Opt_discard, "discard"},
1256 {Opt_nodiscard, "nodiscard"},
1257 {Opt_init_itable, "init_itable=%u"},
1258 {Opt_init_itable, "init_itable"},
1259 {Opt_noinit_itable, "noinit_itable"},
1260 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1261 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1262 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1263 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1264 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1265 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1266 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1267 {Opt_err, NULL},
1268 };
1269
1270 static ext4_fsblk_t get_sb_block(void **data)
1271 {
1272 ext4_fsblk_t sb_block;
1273 char *options = (char *) *data;
1274
1275 if (!options || strncmp(options, "sb=", 3) != 0)
1276 return 1; /* Default location */
1277
1278 options += 3;
1279 /* TODO: use simple_strtoll with >32bit ext4 */
1280 sb_block = simple_strtoul(options, &options, 0);
1281 if (*options && *options != ',') {
1282 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1283 (char *) *data);
1284 return 1;
1285 }
1286 if (*options == ',')
1287 options++;
1288 *data = (void *) options;
1289
1290 return sb_block;
1291 }
1292
1293 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1294 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1295 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1296
1297 #ifdef CONFIG_QUOTA
1298 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1299 {
1300 struct ext4_sb_info *sbi = EXT4_SB(sb);
1301 char *qname;
1302 int ret = -1;
1303
1304 if (sb_any_quota_loaded(sb) &&
1305 !sbi->s_qf_names[qtype]) {
1306 ext4_msg(sb, KERN_ERR,
1307 "Cannot change journaled "
1308 "quota options when quota turned on");
1309 return -1;
1310 }
1311 if (ext4_has_feature_quota(sb)) {
1312 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1313 "ignored when QUOTA feature is enabled");
1314 return 1;
1315 }
1316 qname = match_strdup(args);
1317 if (!qname) {
1318 ext4_msg(sb, KERN_ERR,
1319 "Not enough memory for storing quotafile name");
1320 return -1;
1321 }
1322 if (sbi->s_qf_names[qtype]) {
1323 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1324 ret = 1;
1325 else
1326 ext4_msg(sb, KERN_ERR,
1327 "%s quota file already specified",
1328 QTYPE2NAME(qtype));
1329 goto errout;
1330 }
1331 if (strchr(qname, '/')) {
1332 ext4_msg(sb, KERN_ERR,
1333 "quotafile must be on filesystem root");
1334 goto errout;
1335 }
1336 sbi->s_qf_names[qtype] = qname;
1337 set_opt(sb, QUOTA);
1338 return 1;
1339 errout:
1340 kfree(qname);
1341 return ret;
1342 }
1343
1344 static int clear_qf_name(struct super_block *sb, int qtype)
1345 {
1346
1347 struct ext4_sb_info *sbi = EXT4_SB(sb);
1348
1349 if (sb_any_quota_loaded(sb) &&
1350 sbi->s_qf_names[qtype]) {
1351 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1352 " when quota turned on");
1353 return -1;
1354 }
1355 kfree(sbi->s_qf_names[qtype]);
1356 sbi->s_qf_names[qtype] = NULL;
1357 return 1;
1358 }
1359 #endif
1360
1361 #define MOPT_SET 0x0001
1362 #define MOPT_CLEAR 0x0002
1363 #define MOPT_NOSUPPORT 0x0004
1364 #define MOPT_EXPLICIT 0x0008
1365 #define MOPT_CLEAR_ERR 0x0010
1366 #define MOPT_GTE0 0x0020
1367 #ifdef CONFIG_QUOTA
1368 #define MOPT_Q 0
1369 #define MOPT_QFMT 0x0040
1370 #else
1371 #define MOPT_Q MOPT_NOSUPPORT
1372 #define MOPT_QFMT MOPT_NOSUPPORT
1373 #endif
1374 #define MOPT_DATAJ 0x0080
1375 #define MOPT_NO_EXT2 0x0100
1376 #define MOPT_NO_EXT3 0x0200
1377 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1378 #define MOPT_STRING 0x0400
1379
1380 static const struct mount_opts {
1381 int token;
1382 int mount_opt;
1383 int flags;
1384 } ext4_mount_opts[] = {
1385 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1386 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1387 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1388 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1389 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1390 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1391 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1392 MOPT_EXT4_ONLY | MOPT_SET},
1393 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1394 MOPT_EXT4_ONLY | MOPT_CLEAR},
1395 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1396 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1397 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1398 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1399 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1400 MOPT_EXT4_ONLY | MOPT_CLEAR},
1401 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1402 MOPT_EXT4_ONLY | MOPT_CLEAR},
1403 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1404 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1405 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1406 EXT4_MOUNT_JOURNAL_CHECKSUM),
1407 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1408 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1409 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1410 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1411 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1412 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1413 MOPT_NO_EXT2 | MOPT_SET},
1414 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1415 MOPT_NO_EXT2 | MOPT_CLEAR},
1416 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1417 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1418 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1419 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1420 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1421 {Opt_commit, 0, MOPT_GTE0},
1422 {Opt_max_batch_time, 0, MOPT_GTE0},
1423 {Opt_min_batch_time, 0, MOPT_GTE0},
1424 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1425 {Opt_init_itable, 0, MOPT_GTE0},
1426 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1427 {Opt_stripe, 0, MOPT_GTE0},
1428 {Opt_resuid, 0, MOPT_GTE0},
1429 {Opt_resgid, 0, MOPT_GTE0},
1430 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1431 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1432 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1433 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1434 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1435 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1436 MOPT_NO_EXT2 | MOPT_DATAJ},
1437 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1438 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1439 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1440 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1441 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1442 #else
1443 {Opt_acl, 0, MOPT_NOSUPPORT},
1444 {Opt_noacl, 0, MOPT_NOSUPPORT},
1445 #endif
1446 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1447 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1448 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1449 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1450 MOPT_SET | MOPT_Q},
1451 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1452 MOPT_SET | MOPT_Q},
1453 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1454 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1455 {Opt_usrjquota, 0, MOPT_Q},
1456 {Opt_grpjquota, 0, MOPT_Q},
1457 {Opt_offusrjquota, 0, MOPT_Q},
1458 {Opt_offgrpjquota, 0, MOPT_Q},
1459 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1460 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1461 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1462 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1463 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1464 {Opt_err, 0, 0}
1465 };
1466
1467 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1468 substring_t *args, unsigned long *journal_devnum,
1469 unsigned int *journal_ioprio, int is_remount)
1470 {
1471 struct ext4_sb_info *sbi = EXT4_SB(sb);
1472 const struct mount_opts *m;
1473 kuid_t uid;
1474 kgid_t gid;
1475 int arg = 0;
1476
1477 #ifdef CONFIG_QUOTA
1478 if (token == Opt_usrjquota)
1479 return set_qf_name(sb, USRQUOTA, &args[0]);
1480 else if (token == Opt_grpjquota)
1481 return set_qf_name(sb, GRPQUOTA, &args[0]);
1482 else if (token == Opt_offusrjquota)
1483 return clear_qf_name(sb, USRQUOTA);
1484 else if (token == Opt_offgrpjquota)
1485 return clear_qf_name(sb, GRPQUOTA);
1486 #endif
1487 switch (token) {
1488 case Opt_noacl:
1489 case Opt_nouser_xattr:
1490 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1491 break;
1492 case Opt_sb:
1493 return 1; /* handled by get_sb_block() */
1494 case Opt_removed:
1495 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1496 return 1;
1497 case Opt_abort:
1498 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1499 return 1;
1500 case Opt_i_version:
1501 sb->s_flags |= MS_I_VERSION;
1502 return 1;
1503 case Opt_lazytime:
1504 sb->s_flags |= MS_LAZYTIME;
1505 return 1;
1506 case Opt_nolazytime:
1507 sb->s_flags &= ~MS_LAZYTIME;
1508 return 1;
1509 }
1510
1511 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1512 if (token == m->token)
1513 break;
1514
1515 if (m->token == Opt_err) {
1516 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1517 "or missing value", opt);
1518 return -1;
1519 }
1520
1521 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1522 ext4_msg(sb, KERN_ERR,
1523 "Mount option \"%s\" incompatible with ext2", opt);
1524 return -1;
1525 }
1526 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1527 ext4_msg(sb, KERN_ERR,
1528 "Mount option \"%s\" incompatible with ext3", opt);
1529 return -1;
1530 }
1531
1532 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1533 return -1;
1534 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1535 return -1;
1536 if (m->flags & MOPT_EXPLICIT) {
1537 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1538 set_opt2(sb, EXPLICIT_DELALLOC);
1539 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1540 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1541 } else
1542 return -1;
1543 }
1544 if (m->flags & MOPT_CLEAR_ERR)
1545 clear_opt(sb, ERRORS_MASK);
1546 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1547 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1548 "options when quota turned on");
1549 return -1;
1550 }
1551
1552 if (m->flags & MOPT_NOSUPPORT) {
1553 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1554 } else if (token == Opt_commit) {
1555 if (arg == 0)
1556 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1557 sbi->s_commit_interval = HZ * arg;
1558 } else if (token == Opt_max_batch_time) {
1559 sbi->s_max_batch_time = arg;
1560 } else if (token == Opt_min_batch_time) {
1561 sbi->s_min_batch_time = arg;
1562 } else if (token == Opt_inode_readahead_blks) {
1563 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1564 ext4_msg(sb, KERN_ERR,
1565 "EXT4-fs: inode_readahead_blks must be "
1566 "0 or a power of 2 smaller than 2^31");
1567 return -1;
1568 }
1569 sbi->s_inode_readahead_blks = arg;
1570 } else if (token == Opt_init_itable) {
1571 set_opt(sb, INIT_INODE_TABLE);
1572 if (!args->from)
1573 arg = EXT4_DEF_LI_WAIT_MULT;
1574 sbi->s_li_wait_mult = arg;
1575 } else if (token == Opt_max_dir_size_kb) {
1576 sbi->s_max_dir_size_kb = arg;
1577 } else if (token == Opt_stripe) {
1578 sbi->s_stripe = arg;
1579 } else if (token == Opt_resuid) {
1580 uid = make_kuid(current_user_ns(), arg);
1581 if (!uid_valid(uid)) {
1582 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1583 return -1;
1584 }
1585 sbi->s_resuid = uid;
1586 } else if (token == Opt_resgid) {
1587 gid = make_kgid(current_user_ns(), arg);
1588 if (!gid_valid(gid)) {
1589 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1590 return -1;
1591 }
1592 sbi->s_resgid = gid;
1593 } else if (token == Opt_journal_dev) {
1594 if (is_remount) {
1595 ext4_msg(sb, KERN_ERR,
1596 "Cannot specify journal on remount");
1597 return -1;
1598 }
1599 *journal_devnum = arg;
1600 } else if (token == Opt_journal_path) {
1601 char *journal_path;
1602 struct inode *journal_inode;
1603 struct path path;
1604 int error;
1605
1606 if (is_remount) {
1607 ext4_msg(sb, KERN_ERR,
1608 "Cannot specify journal on remount");
1609 return -1;
1610 }
1611 journal_path = match_strdup(&args[0]);
1612 if (!journal_path) {
1613 ext4_msg(sb, KERN_ERR, "error: could not dup "
1614 "journal device string");
1615 return -1;
1616 }
1617
1618 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1619 if (error) {
1620 ext4_msg(sb, KERN_ERR, "error: could not find "
1621 "journal device path: error %d", error);
1622 kfree(journal_path);
1623 return -1;
1624 }
1625
1626 journal_inode = d_inode(path.dentry);
1627 if (!S_ISBLK(journal_inode->i_mode)) {
1628 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1629 "is not a block device", journal_path);
1630 path_put(&path);
1631 kfree(journal_path);
1632 return -1;
1633 }
1634
1635 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1636 path_put(&path);
1637 kfree(journal_path);
1638 } else if (token == Opt_journal_ioprio) {
1639 if (arg > 7) {
1640 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1641 " (must be 0-7)");
1642 return -1;
1643 }
1644 *journal_ioprio =
1645 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1646 } else if (token == Opt_test_dummy_encryption) {
1647 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1648 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1649 ext4_msg(sb, KERN_WARNING,
1650 "Test dummy encryption mode enabled");
1651 #else
1652 ext4_msg(sb, KERN_WARNING,
1653 "Test dummy encryption mount option ignored");
1654 #endif
1655 } else if (m->flags & MOPT_DATAJ) {
1656 if (is_remount) {
1657 if (!sbi->s_journal)
1658 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1659 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1660 ext4_msg(sb, KERN_ERR,
1661 "Cannot change data mode on remount");
1662 return -1;
1663 }
1664 } else {
1665 clear_opt(sb, DATA_FLAGS);
1666 sbi->s_mount_opt |= m->mount_opt;
1667 }
1668 #ifdef CONFIG_QUOTA
1669 } else if (m->flags & MOPT_QFMT) {
1670 if (sb_any_quota_loaded(sb) &&
1671 sbi->s_jquota_fmt != m->mount_opt) {
1672 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1673 "quota options when quota turned on");
1674 return -1;
1675 }
1676 if (ext4_has_feature_quota(sb)) {
1677 ext4_msg(sb, KERN_INFO,
1678 "Quota format mount options ignored "
1679 "when QUOTA feature is enabled");
1680 return 1;
1681 }
1682 sbi->s_jquota_fmt = m->mount_opt;
1683 #endif
1684 } else if (token == Opt_dax) {
1685 #ifdef CONFIG_FS_DAX
1686 ext4_msg(sb, KERN_WARNING,
1687 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1688 sbi->s_mount_opt |= m->mount_opt;
1689 #else
1690 ext4_msg(sb, KERN_INFO, "dax option not supported");
1691 return -1;
1692 #endif
1693 } else {
1694 if (!args->from)
1695 arg = 1;
1696 if (m->flags & MOPT_CLEAR)
1697 arg = !arg;
1698 else if (unlikely(!(m->flags & MOPT_SET))) {
1699 ext4_msg(sb, KERN_WARNING,
1700 "buggy handling of option %s", opt);
1701 WARN_ON(1);
1702 return -1;
1703 }
1704 if (arg != 0)
1705 sbi->s_mount_opt |= m->mount_opt;
1706 else
1707 sbi->s_mount_opt &= ~m->mount_opt;
1708 }
1709 return 1;
1710 }
1711
1712 static int parse_options(char *options, struct super_block *sb,
1713 unsigned long *journal_devnum,
1714 unsigned int *journal_ioprio,
1715 int is_remount)
1716 {
1717 struct ext4_sb_info *sbi = EXT4_SB(sb);
1718 char *p;
1719 substring_t args[MAX_OPT_ARGS];
1720 int token;
1721
1722 if (!options)
1723 return 1;
1724
1725 while ((p = strsep(&options, ",")) != NULL) {
1726 if (!*p)
1727 continue;
1728 /*
1729 * Initialize args struct so we know whether arg was
1730 * found; some options take optional arguments.
1731 */
1732 args[0].to = args[0].from = NULL;
1733 token = match_token(p, tokens, args);
1734 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1735 journal_ioprio, is_remount) < 0)
1736 return 0;
1737 }
1738 #ifdef CONFIG_QUOTA
1739 if (ext4_has_feature_quota(sb) &&
1740 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1741 ext4_msg(sb, KERN_INFO, "Quota feature enabled, usrquota and grpquota "
1742 "mount options ignored.");
1743 clear_opt(sb, USRQUOTA);
1744 clear_opt(sb, GRPQUOTA);
1745 } else if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1746 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1747 clear_opt(sb, USRQUOTA);
1748
1749 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1750 clear_opt(sb, GRPQUOTA);
1751
1752 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1753 ext4_msg(sb, KERN_ERR, "old and new quota "
1754 "format mixing");
1755 return 0;
1756 }
1757
1758 if (!sbi->s_jquota_fmt) {
1759 ext4_msg(sb, KERN_ERR, "journaled quota format "
1760 "not specified");
1761 return 0;
1762 }
1763 }
1764 #endif
1765 if (test_opt(sb, DIOREAD_NOLOCK)) {
1766 int blocksize =
1767 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1768
1769 if (blocksize < PAGE_CACHE_SIZE) {
1770 ext4_msg(sb, KERN_ERR, "can't mount with "
1771 "dioread_nolock if block size != PAGE_SIZE");
1772 return 0;
1773 }
1774 }
1775 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1776 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1777 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1778 "in data=ordered mode");
1779 return 0;
1780 }
1781 return 1;
1782 }
1783
1784 static inline void ext4_show_quota_options(struct seq_file *seq,
1785 struct super_block *sb)
1786 {
1787 #if defined(CONFIG_QUOTA)
1788 struct ext4_sb_info *sbi = EXT4_SB(sb);
1789
1790 if (sbi->s_jquota_fmt) {
1791 char *fmtname = "";
1792
1793 switch (sbi->s_jquota_fmt) {
1794 case QFMT_VFS_OLD:
1795 fmtname = "vfsold";
1796 break;
1797 case QFMT_VFS_V0:
1798 fmtname = "vfsv0";
1799 break;
1800 case QFMT_VFS_V1:
1801 fmtname = "vfsv1";
1802 break;
1803 }
1804 seq_printf(seq, ",jqfmt=%s", fmtname);
1805 }
1806
1807 if (sbi->s_qf_names[USRQUOTA])
1808 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1809
1810 if (sbi->s_qf_names[GRPQUOTA])
1811 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1812 #endif
1813 }
1814
1815 static const char *token2str(int token)
1816 {
1817 const struct match_token *t;
1818
1819 for (t = tokens; t->token != Opt_err; t++)
1820 if (t->token == token && !strchr(t->pattern, '='))
1821 break;
1822 return t->pattern;
1823 }
1824
1825 /*
1826 * Show an option if
1827 * - it's set to a non-default value OR
1828 * - if the per-sb default is different from the global default
1829 */
1830 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1831 int nodefs)
1832 {
1833 struct ext4_sb_info *sbi = EXT4_SB(sb);
1834 struct ext4_super_block *es = sbi->s_es;
1835 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1836 const struct mount_opts *m;
1837 char sep = nodefs ? '\n' : ',';
1838
1839 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1840 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1841
1842 if (sbi->s_sb_block != 1)
1843 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1844
1845 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1846 int want_set = m->flags & MOPT_SET;
1847 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1848 (m->flags & MOPT_CLEAR_ERR))
1849 continue;
1850 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1851 continue; /* skip if same as the default */
1852 if ((want_set &&
1853 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1854 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1855 continue; /* select Opt_noFoo vs Opt_Foo */
1856 SEQ_OPTS_PRINT("%s", token2str(m->token));
1857 }
1858
1859 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1860 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1861 SEQ_OPTS_PRINT("resuid=%u",
1862 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1863 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1864 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1865 SEQ_OPTS_PRINT("resgid=%u",
1866 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1867 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1868 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1869 SEQ_OPTS_PUTS("errors=remount-ro");
1870 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1871 SEQ_OPTS_PUTS("errors=continue");
1872 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1873 SEQ_OPTS_PUTS("errors=panic");
1874 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1875 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1876 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1877 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1878 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1879 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1880 if (sb->s_flags & MS_I_VERSION)
1881 SEQ_OPTS_PUTS("i_version");
1882 if (nodefs || sbi->s_stripe)
1883 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1884 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1885 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1886 SEQ_OPTS_PUTS("data=journal");
1887 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1888 SEQ_OPTS_PUTS("data=ordered");
1889 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1890 SEQ_OPTS_PUTS("data=writeback");
1891 }
1892 if (nodefs ||
1893 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1894 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1895 sbi->s_inode_readahead_blks);
1896
1897 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1898 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1899 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1900 if (nodefs || sbi->s_max_dir_size_kb)
1901 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1902
1903 ext4_show_quota_options(seq, sb);
1904 return 0;
1905 }
1906
1907 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1908 {
1909 return _ext4_show_options(seq, root->d_sb, 0);
1910 }
1911
1912 int ext4_seq_options_show(struct seq_file *seq, void *offset)
1913 {
1914 struct super_block *sb = seq->private;
1915 int rc;
1916
1917 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1918 rc = _ext4_show_options(seq, sb, 1);
1919 seq_puts(seq, "\n");
1920 return rc;
1921 }
1922
1923 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1924 int read_only)
1925 {
1926 struct ext4_sb_info *sbi = EXT4_SB(sb);
1927 int res = 0;
1928
1929 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1930 ext4_msg(sb, KERN_ERR, "revision level too high, "
1931 "forcing read-only mode");
1932 res = MS_RDONLY;
1933 }
1934 if (read_only)
1935 goto done;
1936 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1937 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1938 "running e2fsck is recommended");
1939 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1940 ext4_msg(sb, KERN_WARNING,
1941 "warning: mounting fs with errors, "
1942 "running e2fsck is recommended");
1943 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1944 le16_to_cpu(es->s_mnt_count) >=
1945 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1946 ext4_msg(sb, KERN_WARNING,
1947 "warning: maximal mount count reached, "
1948 "running e2fsck is recommended");
1949 else if (le32_to_cpu(es->s_checkinterval) &&
1950 (le32_to_cpu(es->s_lastcheck) +
1951 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1952 ext4_msg(sb, KERN_WARNING,
1953 "warning: checktime reached, "
1954 "running e2fsck is recommended");
1955 if (!sbi->s_journal)
1956 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1957 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1958 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1959 le16_add_cpu(&es->s_mnt_count, 1);
1960 es->s_mtime = cpu_to_le32(get_seconds());
1961 ext4_update_dynamic_rev(sb);
1962 if (sbi->s_journal)
1963 ext4_set_feature_journal_needs_recovery(sb);
1964
1965 ext4_commit_super(sb, 1);
1966 done:
1967 if (test_opt(sb, DEBUG))
1968 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1969 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1970 sb->s_blocksize,
1971 sbi->s_groups_count,
1972 EXT4_BLOCKS_PER_GROUP(sb),
1973 EXT4_INODES_PER_GROUP(sb),
1974 sbi->s_mount_opt, sbi->s_mount_opt2);
1975
1976 cleancache_init_fs(sb);
1977 return res;
1978 }
1979
1980 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1981 {
1982 struct ext4_sb_info *sbi = EXT4_SB(sb);
1983 struct flex_groups **old_groups, **new_groups;
1984 int size, i, j;
1985
1986 if (!sbi->s_log_groups_per_flex)
1987 return 0;
1988
1989 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1990 if (size <= sbi->s_flex_groups_allocated)
1991 return 0;
1992
1993 new_groups = ext4_kvzalloc(roundup_pow_of_two(size *
1994 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
1995 if (!new_groups) {
1996 ext4_msg(sb, KERN_ERR,
1997 "not enough memory for %d flex group pointers", size);
1998 return -ENOMEM;
1999 }
2000 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2001 new_groups[i] = ext4_kvzalloc(roundup_pow_of_two(
2002 sizeof(struct flex_groups)),
2003 GFP_KERNEL);
2004 if (!new_groups[i]) {
2005 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2006 kvfree(new_groups[j]);
2007 kvfree(new_groups);
2008 ext4_msg(sb, KERN_ERR,
2009 "not enough memory for %d flex groups", size);
2010 return -ENOMEM;
2011 }
2012 }
2013 rcu_read_lock();
2014 old_groups = rcu_dereference(sbi->s_flex_groups);
2015 if (old_groups)
2016 memcpy(new_groups, old_groups,
2017 (sbi->s_flex_groups_allocated *
2018 sizeof(struct flex_groups *)));
2019 rcu_read_unlock();
2020 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2021 sbi->s_flex_groups_allocated = size;
2022 if (old_groups)
2023 ext4_kvfree_array_rcu(old_groups);
2024 return 0;
2025 }
2026
2027 static int ext4_fill_flex_info(struct super_block *sb)
2028 {
2029 struct ext4_sb_info *sbi = EXT4_SB(sb);
2030 struct ext4_group_desc *gdp = NULL;
2031 struct flex_groups *fg;
2032 ext4_group_t flex_group;
2033 int i, err;
2034
2035 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2036 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2037 sbi->s_log_groups_per_flex = 0;
2038 return 1;
2039 }
2040
2041 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2042 if (err)
2043 goto failed;
2044
2045 for (i = 0; i < sbi->s_groups_count; i++) {
2046 gdp = ext4_get_group_desc(sb, i, NULL);
2047
2048 flex_group = ext4_flex_group(sbi, i);
2049 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2050 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2051 atomic64_add(ext4_free_group_clusters(sb, gdp),
2052 &fg->free_clusters);
2053 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2054 }
2055
2056 return 1;
2057 failed:
2058 return 0;
2059 }
2060
2061 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2062 struct ext4_group_desc *gdp)
2063 {
2064 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2065 __u16 crc = 0;
2066 __le32 le_group = cpu_to_le32(block_group);
2067 struct ext4_sb_info *sbi = EXT4_SB(sb);
2068
2069 if (ext4_has_metadata_csum(sbi->s_sb)) {
2070 /* Use new metadata_csum algorithm */
2071 __u32 csum32;
2072 __u16 dummy_csum = 0;
2073
2074 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2075 sizeof(le_group));
2076 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2077 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2078 sizeof(dummy_csum));
2079 offset += sizeof(dummy_csum);
2080 if (offset < sbi->s_desc_size)
2081 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2082 sbi->s_desc_size - offset);
2083
2084 crc = csum32 & 0xFFFF;
2085 goto out;
2086 }
2087
2088 /* old crc16 code */
2089 if (!ext4_has_feature_gdt_csum(sb))
2090 return 0;
2091
2092 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2093 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2094 crc = crc16(crc, (__u8 *)gdp, offset);
2095 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2096 /* for checksum of struct ext4_group_desc do the rest...*/
2097 if (ext4_has_feature_64bit(sb) &&
2098 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2099 crc = crc16(crc, (__u8 *)gdp + offset,
2100 le16_to_cpu(sbi->s_es->s_desc_size) -
2101 offset);
2102
2103 out:
2104 return cpu_to_le16(crc);
2105 }
2106
2107 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2108 struct ext4_group_desc *gdp)
2109 {
2110 if (ext4_has_group_desc_csum(sb) &&
2111 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2112 return 0;
2113
2114 return 1;
2115 }
2116
2117 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2118 struct ext4_group_desc *gdp)
2119 {
2120 if (!ext4_has_group_desc_csum(sb))
2121 return;
2122 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2123 }
2124
2125 /* Called at mount-time, super-block is locked */
2126 static int ext4_check_descriptors(struct super_block *sb,
2127 ext4_fsblk_t sb_block,
2128 ext4_group_t *first_not_zeroed)
2129 {
2130 struct ext4_sb_info *sbi = EXT4_SB(sb);
2131 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2132 ext4_fsblk_t last_block;
2133 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2134 ext4_fsblk_t block_bitmap;
2135 ext4_fsblk_t inode_bitmap;
2136 ext4_fsblk_t inode_table;
2137 int flexbg_flag = 0;
2138 ext4_group_t i, grp = sbi->s_groups_count;
2139
2140 if (ext4_has_feature_flex_bg(sb))
2141 flexbg_flag = 1;
2142
2143 ext4_debug("Checking group descriptors");
2144
2145 for (i = 0; i < sbi->s_groups_count; i++) {
2146 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2147
2148 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2149 last_block = ext4_blocks_count(sbi->s_es) - 1;
2150 else
2151 last_block = first_block +
2152 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2153
2154 if ((grp == sbi->s_groups_count) &&
2155 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2156 grp = i;
2157
2158 block_bitmap = ext4_block_bitmap(sb, gdp);
2159 if (block_bitmap == sb_block) {
2160 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2161 "Block bitmap for group %u overlaps "
2162 "superblock", i);
2163 if (!(sb->s_flags & MS_RDONLY))
2164 return 0;
2165 }
2166 if (block_bitmap >= sb_block + 1 &&
2167 block_bitmap <= last_bg_block) {
2168 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2169 "Block bitmap for group %u overlaps "
2170 "block group descriptors", i);
2171 if (!(sb->s_flags & MS_RDONLY))
2172 return 0;
2173 }
2174 if (block_bitmap < first_block || block_bitmap > last_block) {
2175 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2176 "Block bitmap for group %u not in group "
2177 "(block %llu)!", i, block_bitmap);
2178 return 0;
2179 }
2180 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2181 if (inode_bitmap == sb_block) {
2182 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2183 "Inode bitmap for group %u overlaps "
2184 "superblock", i);
2185 if (!(sb->s_flags & MS_RDONLY))
2186 return 0;
2187 }
2188 if (inode_bitmap >= sb_block + 1 &&
2189 inode_bitmap <= last_bg_block) {
2190 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2191 "Inode bitmap for group %u overlaps "
2192 "block group descriptors", i);
2193 if (!(sb->s_flags & MS_RDONLY))
2194 return 0;
2195 }
2196 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2197 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2198 "Inode bitmap for group %u not in group "
2199 "(block %llu)!", i, inode_bitmap);
2200 return 0;
2201 }
2202 inode_table = ext4_inode_table(sb, gdp);
2203 if (inode_table == sb_block) {
2204 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2205 "Inode table for group %u overlaps "
2206 "superblock", i);
2207 if (!(sb->s_flags & MS_RDONLY))
2208 return 0;
2209 }
2210 if (inode_table >= sb_block + 1 &&
2211 inode_table <= last_bg_block) {
2212 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2213 "Inode table for group %u overlaps "
2214 "block group descriptors", i);
2215 if (!(sb->s_flags & MS_RDONLY))
2216 return 0;
2217 }
2218 if (inode_table < first_block ||
2219 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2220 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2221 "Inode table for group %u not in group "
2222 "(block %llu)!", i, inode_table);
2223 return 0;
2224 }
2225 ext4_lock_group(sb, i);
2226 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2227 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2228 "Checksum for group %u failed (%u!=%u)",
2229 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2230 gdp)), le16_to_cpu(gdp->bg_checksum));
2231 if (!(sb->s_flags & MS_RDONLY)) {
2232 ext4_unlock_group(sb, i);
2233 return 0;
2234 }
2235 }
2236 ext4_unlock_group(sb, i);
2237 if (!flexbg_flag)
2238 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2239 }
2240 if (NULL != first_not_zeroed)
2241 *first_not_zeroed = grp;
2242 return 1;
2243 }
2244
2245 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2246 * the superblock) which were deleted from all directories, but held open by
2247 * a process at the time of a crash. We walk the list and try to delete these
2248 * inodes at recovery time (only with a read-write filesystem).
2249 *
2250 * In order to keep the orphan inode chain consistent during traversal (in
2251 * case of crash during recovery), we link each inode into the superblock
2252 * orphan list_head and handle it the same way as an inode deletion during
2253 * normal operation (which journals the operations for us).
2254 *
2255 * We only do an iget() and an iput() on each inode, which is very safe if we
2256 * accidentally point at an in-use or already deleted inode. The worst that
2257 * can happen in this case is that we get a "bit already cleared" message from
2258 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2259 * e2fsck was run on this filesystem, and it must have already done the orphan
2260 * inode cleanup for us, so we can safely abort without any further action.
2261 */
2262 static void ext4_orphan_cleanup(struct super_block *sb,
2263 struct ext4_super_block *es)
2264 {
2265 unsigned int s_flags = sb->s_flags;
2266 int nr_orphans = 0, nr_truncates = 0;
2267 #ifdef CONFIG_QUOTA
2268 int quota_update = 0;
2269 int i;
2270 #endif
2271 if (!es->s_last_orphan) {
2272 jbd_debug(4, "no orphan inodes to clean up\n");
2273 return;
2274 }
2275
2276 if (bdev_read_only(sb->s_bdev)) {
2277 ext4_msg(sb, KERN_ERR, "write access "
2278 "unavailable, skipping orphan cleanup");
2279 return;
2280 }
2281
2282 /* Check if feature set would not allow a r/w mount */
2283 if (!ext4_feature_set_ok(sb, 0)) {
2284 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2285 "unknown ROCOMPAT features");
2286 return;
2287 }
2288
2289 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2290 /* don't clear list on RO mount w/ errors */
2291 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2292 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2293 "clearing orphan list.\n");
2294 es->s_last_orphan = 0;
2295 }
2296 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2297 return;
2298 }
2299
2300 if (s_flags & MS_RDONLY) {
2301 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2302 sb->s_flags &= ~MS_RDONLY;
2303 }
2304 #ifdef CONFIG_QUOTA
2305 /* Needed for iput() to work correctly and not trash data */
2306 sb->s_flags |= MS_ACTIVE;
2307
2308 /*
2309 * Turn on quotas which were not enabled for read-only mounts if
2310 * filesystem has quota feature, so that they are updated correctly.
2311 */
2312 if (ext4_has_feature_quota(sb) && (s_flags & MS_RDONLY)) {
2313 int ret = ext4_enable_quotas(sb);
2314
2315 if (!ret)
2316 quota_update = 1;
2317 else
2318 ext4_msg(sb, KERN_ERR,
2319 "Cannot turn on quotas: error %d", ret);
2320 }
2321
2322 /* Turn on journaled quotas used for old sytle */
2323 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2324 if (EXT4_SB(sb)->s_qf_names[i]) {
2325 int ret = ext4_quota_on_mount(sb, i);
2326
2327 if (!ret)
2328 quota_update = 1;
2329 else
2330 ext4_msg(sb, KERN_ERR,
2331 "Cannot turn on journaled "
2332 "quota: type %d: error %d", i, ret);
2333 }
2334 }
2335 #endif
2336
2337 while (es->s_last_orphan) {
2338 struct inode *inode;
2339
2340 /*
2341 * We may have encountered an error during cleanup; if
2342 * so, skip the rest.
2343 */
2344 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2345 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2346 es->s_last_orphan = 0;
2347 break;
2348 }
2349
2350 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2351 if (IS_ERR(inode)) {
2352 es->s_last_orphan = 0;
2353 break;
2354 }
2355
2356 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2357 dquot_initialize(inode);
2358 if (inode->i_nlink) {
2359 if (test_opt(sb, DEBUG))
2360 ext4_msg(sb, KERN_DEBUG,
2361 "%s: truncating inode %lu to %lld bytes",
2362 __func__, inode->i_ino, inode->i_size);
2363 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2364 inode->i_ino, inode->i_size);
2365 mutex_lock(&inode->i_mutex);
2366 truncate_inode_pages(inode->i_mapping, inode->i_size);
2367 ext4_truncate(inode);
2368 mutex_unlock(&inode->i_mutex);
2369 nr_truncates++;
2370 } else {
2371 if (test_opt(sb, DEBUG))
2372 ext4_msg(sb, KERN_DEBUG,
2373 "%s: deleting unreferenced inode %lu",
2374 __func__, inode->i_ino);
2375 jbd_debug(2, "deleting unreferenced inode %lu\n",
2376 inode->i_ino);
2377 nr_orphans++;
2378 }
2379 iput(inode); /* The delete magic happens here! */
2380 }
2381
2382 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2383
2384 if (nr_orphans)
2385 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2386 PLURAL(nr_orphans));
2387 if (nr_truncates)
2388 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2389 PLURAL(nr_truncates));
2390 #ifdef CONFIG_QUOTA
2391 /* Turn off quotas if they were enabled for orphan cleanup */
2392 if (quota_update) {
2393 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2394 if (sb_dqopt(sb)->files[i])
2395 dquot_quota_off(sb, i);
2396 }
2397 }
2398 #endif
2399 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2400 }
2401
2402 /*
2403 * Maximal extent format file size.
2404 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2405 * extent format containers, within a sector_t, and within i_blocks
2406 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2407 * so that won't be a limiting factor.
2408 *
2409 * However there is other limiting factor. We do store extents in the form
2410 * of starting block and length, hence the resulting length of the extent
2411 * covering maximum file size must fit into on-disk format containers as
2412 * well. Given that length is always by 1 unit bigger than max unit (because
2413 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2414 *
2415 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2416 */
2417 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2418 {
2419 loff_t res;
2420 loff_t upper_limit = MAX_LFS_FILESIZE;
2421
2422 /* small i_blocks in vfs inode? */
2423 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2424 /*
2425 * CONFIG_LBDAF is not enabled implies the inode
2426 * i_block represent total blocks in 512 bytes
2427 * 32 == size of vfs inode i_blocks * 8
2428 */
2429 upper_limit = (1LL << 32) - 1;
2430
2431 /* total blocks in file system block size */
2432 upper_limit >>= (blkbits - 9);
2433 upper_limit <<= blkbits;
2434 }
2435
2436 /*
2437 * 32-bit extent-start container, ee_block. We lower the maxbytes
2438 * by one fs block, so ee_len can cover the extent of maximum file
2439 * size
2440 */
2441 res = (1LL << 32) - 1;
2442 res <<= blkbits;
2443
2444 /* Sanity check against vm- & vfs- imposed limits */
2445 if (res > upper_limit)
2446 res = upper_limit;
2447
2448 return res;
2449 }
2450
2451 /*
2452 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2453 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2454 * We need to be 1 filesystem block less than the 2^48 sector limit.
2455 */
2456 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2457 {
2458 loff_t res = EXT4_NDIR_BLOCKS;
2459 int meta_blocks;
2460 loff_t upper_limit;
2461 /* This is calculated to be the largest file size for a dense, block
2462 * mapped file such that the file's total number of 512-byte sectors,
2463 * including data and all indirect blocks, does not exceed (2^48 - 1).
2464 *
2465 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2466 * number of 512-byte sectors of the file.
2467 */
2468
2469 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2470 /*
2471 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2472 * the inode i_block field represents total file blocks in
2473 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2474 */
2475 upper_limit = (1LL << 32) - 1;
2476
2477 /* total blocks in file system block size */
2478 upper_limit >>= (bits - 9);
2479
2480 } else {
2481 /*
2482 * We use 48 bit ext4_inode i_blocks
2483 * With EXT4_HUGE_FILE_FL set the i_blocks
2484 * represent total number of blocks in
2485 * file system block size
2486 */
2487 upper_limit = (1LL << 48) - 1;
2488
2489 }
2490
2491 /* indirect blocks */
2492 meta_blocks = 1;
2493 /* double indirect blocks */
2494 meta_blocks += 1 + (1LL << (bits-2));
2495 /* tripple indirect blocks */
2496 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2497
2498 upper_limit -= meta_blocks;
2499 upper_limit <<= bits;
2500
2501 res += 1LL << (bits-2);
2502 res += 1LL << (2*(bits-2));
2503 res += 1LL << (3*(bits-2));
2504 res <<= bits;
2505 if (res > upper_limit)
2506 res = upper_limit;
2507
2508 if (res > MAX_LFS_FILESIZE)
2509 res = MAX_LFS_FILESIZE;
2510
2511 return res;
2512 }
2513
2514 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2515 ext4_fsblk_t logical_sb_block, int nr)
2516 {
2517 struct ext4_sb_info *sbi = EXT4_SB(sb);
2518 ext4_group_t bg, first_meta_bg;
2519 int has_super = 0;
2520
2521 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2522
2523 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2524 return logical_sb_block + nr + 1;
2525 bg = sbi->s_desc_per_block * nr;
2526 if (ext4_bg_has_super(sb, bg))
2527 has_super = 1;
2528
2529 /*
2530 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2531 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2532 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2533 * compensate.
2534 */
2535 if (sb->s_blocksize == 1024 && nr == 0 &&
2536 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2537 has_super++;
2538
2539 return (has_super + ext4_group_first_block_no(sb, bg));
2540 }
2541
2542 /**
2543 * ext4_get_stripe_size: Get the stripe size.
2544 * @sbi: In memory super block info
2545 *
2546 * If we have specified it via mount option, then
2547 * use the mount option value. If the value specified at mount time is
2548 * greater than the blocks per group use the super block value.
2549 * If the super block value is greater than blocks per group return 0.
2550 * Allocator needs it be less than blocks per group.
2551 *
2552 */
2553 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2554 {
2555 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2556 unsigned long stripe_width =
2557 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2558 int ret;
2559
2560 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2561 ret = sbi->s_stripe;
2562 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2563 ret = stripe_width;
2564 else if (stride && stride <= sbi->s_blocks_per_group)
2565 ret = stride;
2566 else
2567 ret = 0;
2568
2569 /*
2570 * If the stripe width is 1, this makes no sense and
2571 * we set it to 0 to turn off stripe handling code.
2572 */
2573 if (ret <= 1)
2574 ret = 0;
2575
2576 return ret;
2577 }
2578
2579 /*
2580 * Check whether this filesystem can be mounted based on
2581 * the features present and the RDONLY/RDWR mount requested.
2582 * Returns 1 if this filesystem can be mounted as requested,
2583 * 0 if it cannot be.
2584 */
2585 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2586 {
2587 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2588 ext4_msg(sb, KERN_ERR,
2589 "Couldn't mount because of "
2590 "unsupported optional features (%x)",
2591 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2592 ~EXT4_FEATURE_INCOMPAT_SUPP));
2593 return 0;
2594 }
2595
2596 if (readonly)
2597 return 1;
2598
2599 if (ext4_has_feature_readonly(sb)) {
2600 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2601 sb->s_flags |= MS_RDONLY;
2602 return 1;
2603 }
2604
2605 /* Check that feature set is OK for a read-write mount */
2606 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2607 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2608 "unsupported optional features (%x)",
2609 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2610 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2611 return 0;
2612 }
2613 /*
2614 * Large file size enabled file system can only be mounted
2615 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2616 */
2617 if (ext4_has_feature_huge_file(sb)) {
2618 if (sizeof(blkcnt_t) < sizeof(u64)) {
2619 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2620 "cannot be mounted RDWR without "
2621 "CONFIG_LBDAF");
2622 return 0;
2623 }
2624 }
2625 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2626 ext4_msg(sb, KERN_ERR,
2627 "Can't support bigalloc feature without "
2628 "extents feature\n");
2629 return 0;
2630 }
2631
2632 #ifndef CONFIG_QUOTA
2633 if (ext4_has_feature_quota(sb) && !readonly) {
2634 ext4_msg(sb, KERN_ERR,
2635 "Filesystem with quota feature cannot be mounted RDWR "
2636 "without CONFIG_QUOTA");
2637 return 0;
2638 }
2639 #endif /* CONFIG_QUOTA */
2640 return 1;
2641 }
2642
2643 /*
2644 * This function is called once a day if we have errors logged
2645 * on the file system
2646 */
2647 static void print_daily_error_info(unsigned long arg)
2648 {
2649 struct super_block *sb = (struct super_block *) arg;
2650 struct ext4_sb_info *sbi;
2651 struct ext4_super_block *es;
2652
2653 sbi = EXT4_SB(sb);
2654 es = sbi->s_es;
2655
2656 if (es->s_error_count)
2657 /* fsck newer than v1.41.13 is needed to clean this condition. */
2658 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2659 le32_to_cpu(es->s_error_count));
2660 if (es->s_first_error_time) {
2661 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2662 sb->s_id, le32_to_cpu(es->s_first_error_time),
2663 (int) sizeof(es->s_first_error_func),
2664 es->s_first_error_func,
2665 le32_to_cpu(es->s_first_error_line));
2666 if (es->s_first_error_ino)
2667 printk(": inode %u",
2668 le32_to_cpu(es->s_first_error_ino));
2669 if (es->s_first_error_block)
2670 printk(": block %llu", (unsigned long long)
2671 le64_to_cpu(es->s_first_error_block));
2672 printk("\n");
2673 }
2674 if (es->s_last_error_time) {
2675 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2676 sb->s_id, le32_to_cpu(es->s_last_error_time),
2677 (int) sizeof(es->s_last_error_func),
2678 es->s_last_error_func,
2679 le32_to_cpu(es->s_last_error_line));
2680 if (es->s_last_error_ino)
2681 printk(": inode %u",
2682 le32_to_cpu(es->s_last_error_ino));
2683 if (es->s_last_error_block)
2684 printk(": block %llu", (unsigned long long)
2685 le64_to_cpu(es->s_last_error_block));
2686 printk("\n");
2687 }
2688 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2689 }
2690
2691 /* Find next suitable group and run ext4_init_inode_table */
2692 static int ext4_run_li_request(struct ext4_li_request *elr)
2693 {
2694 struct ext4_group_desc *gdp = NULL;
2695 ext4_group_t group, ngroups;
2696 struct super_block *sb;
2697 unsigned long timeout = 0;
2698 int ret = 0;
2699
2700 sb = elr->lr_super;
2701 ngroups = EXT4_SB(sb)->s_groups_count;
2702
2703 sb_start_write(sb);
2704 for (group = elr->lr_next_group; group < ngroups; group++) {
2705 gdp = ext4_get_group_desc(sb, group, NULL);
2706 if (!gdp) {
2707 ret = 1;
2708 break;
2709 }
2710
2711 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2712 break;
2713 }
2714
2715 if (group >= ngroups)
2716 ret = 1;
2717
2718 if (!ret) {
2719 timeout = jiffies;
2720 ret = ext4_init_inode_table(sb, group,
2721 elr->lr_timeout ? 0 : 1);
2722 if (elr->lr_timeout == 0) {
2723 timeout = (jiffies - timeout) *
2724 elr->lr_sbi->s_li_wait_mult;
2725 elr->lr_timeout = timeout;
2726 }
2727 elr->lr_next_sched = jiffies + elr->lr_timeout;
2728 elr->lr_next_group = group + 1;
2729 }
2730 sb_end_write(sb);
2731
2732 return ret;
2733 }
2734
2735 /*
2736 * Remove lr_request from the list_request and free the
2737 * request structure. Should be called with li_list_mtx held
2738 */
2739 static void ext4_remove_li_request(struct ext4_li_request *elr)
2740 {
2741 struct ext4_sb_info *sbi;
2742
2743 if (!elr)
2744 return;
2745
2746 sbi = elr->lr_sbi;
2747
2748 list_del(&elr->lr_request);
2749 sbi->s_li_request = NULL;
2750 kfree(elr);
2751 }
2752
2753 static void ext4_unregister_li_request(struct super_block *sb)
2754 {
2755 mutex_lock(&ext4_li_mtx);
2756 if (!ext4_li_info) {
2757 mutex_unlock(&ext4_li_mtx);
2758 return;
2759 }
2760
2761 mutex_lock(&ext4_li_info->li_list_mtx);
2762 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2763 mutex_unlock(&ext4_li_info->li_list_mtx);
2764 mutex_unlock(&ext4_li_mtx);
2765 }
2766
2767 static struct task_struct *ext4_lazyinit_task;
2768
2769 /*
2770 * This is the function where ext4lazyinit thread lives. It walks
2771 * through the request list searching for next scheduled filesystem.
2772 * When such a fs is found, run the lazy initialization request
2773 * (ext4_rn_li_request) and keep track of the time spend in this
2774 * function. Based on that time we compute next schedule time of
2775 * the request. When walking through the list is complete, compute
2776 * next waking time and put itself into sleep.
2777 */
2778 static int ext4_lazyinit_thread(void *arg)
2779 {
2780 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2781 struct list_head *pos, *n;
2782 struct ext4_li_request *elr;
2783 unsigned long next_wakeup, cur;
2784
2785 BUG_ON(NULL == eli);
2786
2787 cont_thread:
2788 while (true) {
2789 next_wakeup = MAX_JIFFY_OFFSET;
2790
2791 mutex_lock(&eli->li_list_mtx);
2792 if (list_empty(&eli->li_request_list)) {
2793 mutex_unlock(&eli->li_list_mtx);
2794 goto exit_thread;
2795 }
2796
2797 list_for_each_safe(pos, n, &eli->li_request_list) {
2798 elr = list_entry(pos, struct ext4_li_request,
2799 lr_request);
2800
2801 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2802 if (ext4_run_li_request(elr) != 0) {
2803 /* error, remove the lazy_init job */
2804 ext4_remove_li_request(elr);
2805 continue;
2806 }
2807 }
2808
2809 if (time_before(elr->lr_next_sched, next_wakeup))
2810 next_wakeup = elr->lr_next_sched;
2811 }
2812 mutex_unlock(&eli->li_list_mtx);
2813
2814 try_to_freeze();
2815
2816 cur = jiffies;
2817 if ((time_after_eq(cur, next_wakeup)) ||
2818 (MAX_JIFFY_OFFSET == next_wakeup)) {
2819 cond_resched();
2820 continue;
2821 }
2822
2823 schedule_timeout_interruptible(next_wakeup - cur);
2824
2825 if (kthread_should_stop()) {
2826 ext4_clear_request_list();
2827 goto exit_thread;
2828 }
2829 }
2830
2831 exit_thread:
2832 /*
2833 * It looks like the request list is empty, but we need
2834 * to check it under the li_list_mtx lock, to prevent any
2835 * additions into it, and of course we should lock ext4_li_mtx
2836 * to atomically free the list and ext4_li_info, because at
2837 * this point another ext4 filesystem could be registering
2838 * new one.
2839 */
2840 mutex_lock(&ext4_li_mtx);
2841 mutex_lock(&eli->li_list_mtx);
2842 if (!list_empty(&eli->li_request_list)) {
2843 mutex_unlock(&eli->li_list_mtx);
2844 mutex_unlock(&ext4_li_mtx);
2845 goto cont_thread;
2846 }
2847 mutex_unlock(&eli->li_list_mtx);
2848 kfree(ext4_li_info);
2849 ext4_li_info = NULL;
2850 mutex_unlock(&ext4_li_mtx);
2851
2852 return 0;
2853 }
2854
2855 static void ext4_clear_request_list(void)
2856 {
2857 struct list_head *pos, *n;
2858 struct ext4_li_request *elr;
2859
2860 mutex_lock(&ext4_li_info->li_list_mtx);
2861 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2862 elr = list_entry(pos, struct ext4_li_request,
2863 lr_request);
2864 ext4_remove_li_request(elr);
2865 }
2866 mutex_unlock(&ext4_li_info->li_list_mtx);
2867 }
2868
2869 static int ext4_run_lazyinit_thread(void)
2870 {
2871 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2872 ext4_li_info, "ext4lazyinit");
2873 if (IS_ERR(ext4_lazyinit_task)) {
2874 int err = PTR_ERR(ext4_lazyinit_task);
2875 ext4_clear_request_list();
2876 kfree(ext4_li_info);
2877 ext4_li_info = NULL;
2878 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2879 "initialization thread\n",
2880 err);
2881 return err;
2882 }
2883 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2884 return 0;
2885 }
2886
2887 /*
2888 * Check whether it make sense to run itable init. thread or not.
2889 * If there is at least one uninitialized inode table, return
2890 * corresponding group number, else the loop goes through all
2891 * groups and return total number of groups.
2892 */
2893 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2894 {
2895 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2896 struct ext4_group_desc *gdp = NULL;
2897
2898 if (!ext4_has_group_desc_csum(sb))
2899 return ngroups;
2900
2901 for (group = 0; group < ngroups; group++) {
2902 gdp = ext4_get_group_desc(sb, group, NULL);
2903 if (!gdp)
2904 continue;
2905
2906 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2907 break;
2908 }
2909
2910 return group;
2911 }
2912
2913 static int ext4_li_info_new(void)
2914 {
2915 struct ext4_lazy_init *eli = NULL;
2916
2917 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2918 if (!eli)
2919 return -ENOMEM;
2920
2921 INIT_LIST_HEAD(&eli->li_request_list);
2922 mutex_init(&eli->li_list_mtx);
2923
2924 eli->li_state |= EXT4_LAZYINIT_QUIT;
2925
2926 ext4_li_info = eli;
2927
2928 return 0;
2929 }
2930
2931 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2932 ext4_group_t start)
2933 {
2934 struct ext4_sb_info *sbi = EXT4_SB(sb);
2935 struct ext4_li_request *elr;
2936
2937 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2938 if (!elr)
2939 return NULL;
2940
2941 elr->lr_super = sb;
2942 elr->lr_sbi = sbi;
2943 elr->lr_next_group = start;
2944
2945 /*
2946 * Randomize first schedule time of the request to
2947 * spread the inode table initialization requests
2948 * better.
2949 */
2950 elr->lr_next_sched = jiffies + (prandom_u32() %
2951 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2952 return elr;
2953 }
2954
2955 int ext4_register_li_request(struct super_block *sb,
2956 ext4_group_t first_not_zeroed)
2957 {
2958 struct ext4_sb_info *sbi = EXT4_SB(sb);
2959 struct ext4_li_request *elr = NULL;
2960 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2961 int ret = 0;
2962
2963 mutex_lock(&ext4_li_mtx);
2964 if (sbi->s_li_request != NULL) {
2965 /*
2966 * Reset timeout so it can be computed again, because
2967 * s_li_wait_mult might have changed.
2968 */
2969 sbi->s_li_request->lr_timeout = 0;
2970 goto out;
2971 }
2972
2973 if (first_not_zeroed == ngroups ||
2974 (sb->s_flags & MS_RDONLY) ||
2975 !test_opt(sb, INIT_INODE_TABLE))
2976 goto out;
2977
2978 elr = ext4_li_request_new(sb, first_not_zeroed);
2979 if (!elr) {
2980 ret = -ENOMEM;
2981 goto out;
2982 }
2983
2984 if (NULL == ext4_li_info) {
2985 ret = ext4_li_info_new();
2986 if (ret)
2987 goto out;
2988 }
2989
2990 mutex_lock(&ext4_li_info->li_list_mtx);
2991 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2992 mutex_unlock(&ext4_li_info->li_list_mtx);
2993
2994 sbi->s_li_request = elr;
2995 /*
2996 * set elr to NULL here since it has been inserted to
2997 * the request_list and the removal and free of it is
2998 * handled by ext4_clear_request_list from now on.
2999 */
3000 elr = NULL;
3001
3002 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3003 ret = ext4_run_lazyinit_thread();
3004 if (ret)
3005 goto out;
3006 }
3007 out:
3008 mutex_unlock(&ext4_li_mtx);
3009 if (ret)
3010 kfree(elr);
3011 return ret;
3012 }
3013
3014 /*
3015 * We do not need to lock anything since this is called on
3016 * module unload.
3017 */
3018 static void ext4_destroy_lazyinit_thread(void)
3019 {
3020 /*
3021 * If thread exited earlier
3022 * there's nothing to be done.
3023 */
3024 if (!ext4_li_info || !ext4_lazyinit_task)
3025 return;
3026
3027 kthread_stop(ext4_lazyinit_task);
3028 }
3029
3030 static int set_journal_csum_feature_set(struct super_block *sb)
3031 {
3032 int ret = 1;
3033 int compat, incompat;
3034 struct ext4_sb_info *sbi = EXT4_SB(sb);
3035
3036 if (ext4_has_metadata_csum(sb)) {
3037 /* journal checksum v3 */
3038 compat = 0;
3039 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3040 } else {
3041 /* journal checksum v1 */
3042 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3043 incompat = 0;
3044 }
3045
3046 jbd2_journal_clear_features(sbi->s_journal,
3047 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3048 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3049 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3050 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3051 ret = jbd2_journal_set_features(sbi->s_journal,
3052 compat, 0,
3053 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3054 incompat);
3055 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3056 ret = jbd2_journal_set_features(sbi->s_journal,
3057 compat, 0,
3058 incompat);
3059 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3060 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3061 } else {
3062 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3063 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3064 }
3065
3066 return ret;
3067 }
3068
3069 /*
3070 * Note: calculating the overhead so we can be compatible with
3071 * historical BSD practice is quite difficult in the face of
3072 * clusters/bigalloc. This is because multiple metadata blocks from
3073 * different block group can end up in the same allocation cluster.
3074 * Calculating the exact overhead in the face of clustered allocation
3075 * requires either O(all block bitmaps) in memory or O(number of block
3076 * groups**2) in time. We will still calculate the superblock for
3077 * older file systems --- and if we come across with a bigalloc file
3078 * system with zero in s_overhead_clusters the estimate will be close to
3079 * correct especially for very large cluster sizes --- but for newer
3080 * file systems, it's better to calculate this figure once at mkfs
3081 * time, and store it in the superblock. If the superblock value is
3082 * present (even for non-bigalloc file systems), we will use it.
3083 */
3084 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3085 char *buf)
3086 {
3087 struct ext4_sb_info *sbi = EXT4_SB(sb);
3088 struct ext4_group_desc *gdp;
3089 ext4_fsblk_t first_block, last_block, b;
3090 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3091 int s, j, count = 0;
3092
3093 if (!ext4_has_feature_bigalloc(sb))
3094 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3095 sbi->s_itb_per_group + 2);
3096
3097 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3098 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3099 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3100 for (i = 0; i < ngroups; i++) {
3101 gdp = ext4_get_group_desc(sb, i, NULL);
3102 b = ext4_block_bitmap(sb, gdp);
3103 if (b >= first_block && b <= last_block) {
3104 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3105 count++;
3106 }
3107 b = ext4_inode_bitmap(sb, gdp);
3108 if (b >= first_block && b <= last_block) {
3109 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3110 count++;
3111 }
3112 b = ext4_inode_table(sb, gdp);
3113 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3114 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3115 int c = EXT4_B2C(sbi, b - first_block);
3116 ext4_set_bit(c, buf);
3117 count++;
3118 }
3119 if (i != grp)
3120 continue;
3121 s = 0;
3122 if (ext4_bg_has_super(sb, grp)) {
3123 ext4_set_bit(s++, buf);
3124 count++;
3125 }
3126 j = ext4_bg_num_gdb(sb, grp);
3127 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3128 ext4_error(sb, "Invalid number of block group "
3129 "descriptor blocks: %d", j);
3130 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3131 }
3132 count += j;
3133 for (; j > 0; j--)
3134 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3135 }
3136 if (!count)
3137 return 0;
3138 return EXT4_CLUSTERS_PER_GROUP(sb) -
3139 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3140 }
3141
3142 /*
3143 * Compute the overhead and stash it in sbi->s_overhead
3144 */
3145 int ext4_calculate_overhead(struct super_block *sb)
3146 {
3147 struct ext4_sb_info *sbi = EXT4_SB(sb);
3148 struct ext4_super_block *es = sbi->s_es;
3149 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3150 ext4_fsblk_t overhead = 0;
3151 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3152
3153 if (!buf)
3154 return -ENOMEM;
3155
3156 /*
3157 * Compute the overhead (FS structures). This is constant
3158 * for a given filesystem unless the number of block groups
3159 * changes so we cache the previous value until it does.
3160 */
3161
3162 /*
3163 * All of the blocks before first_data_block are overhead
3164 */
3165 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3166
3167 /*
3168 * Add the overhead found in each block group
3169 */
3170 for (i = 0; i < ngroups; i++) {
3171 int blks;
3172
3173 blks = count_overhead(sb, i, buf);
3174 overhead += blks;
3175 if (blks)
3176 memset(buf, 0, PAGE_SIZE);
3177 cond_resched();
3178 }
3179 /* Add the internal journal blocks as well */
3180 if (sbi->s_journal && !sbi->journal_bdev)
3181 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3182
3183 sbi->s_overhead = overhead;
3184 smp_wmb();
3185 free_page((unsigned long) buf);
3186 return 0;
3187 }
3188
3189 static void ext4_clamp_want_extra_isize(struct super_block *sb)
3190 {
3191 struct ext4_sb_info *sbi = EXT4_SB(sb);
3192 struct ext4_super_block *es = sbi->s_es;
3193 unsigned def_extra_isize = sizeof(struct ext4_inode) -
3194 EXT4_GOOD_OLD_INODE_SIZE;
3195
3196 if (sbi->s_inode_size == EXT4_GOOD_OLD_INODE_SIZE) {
3197 sbi->s_want_extra_isize = 0;
3198 return;
3199 }
3200 if (sbi->s_want_extra_isize < 4) {
3201 sbi->s_want_extra_isize = def_extra_isize;
3202 if (ext4_has_feature_extra_isize(sb)) {
3203 if (sbi->s_want_extra_isize <
3204 le16_to_cpu(es->s_want_extra_isize))
3205 sbi->s_want_extra_isize =
3206 le16_to_cpu(es->s_want_extra_isize);
3207 if (sbi->s_want_extra_isize <
3208 le16_to_cpu(es->s_min_extra_isize))
3209 sbi->s_want_extra_isize =
3210 le16_to_cpu(es->s_min_extra_isize);
3211 }
3212 }
3213 /* Check if enough inode space is available */
3214 if ((sbi->s_want_extra_isize > sbi->s_inode_size) ||
3215 (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3216 sbi->s_inode_size)) {
3217 sbi->s_want_extra_isize = def_extra_isize;
3218 ext4_msg(sb, KERN_INFO,
3219 "required extra inode space not available");
3220 }
3221 }
3222
3223 static void ext4_set_resv_clusters(struct super_block *sb)
3224 {
3225 ext4_fsblk_t resv_clusters;
3226 struct ext4_sb_info *sbi = EXT4_SB(sb);
3227
3228 /*
3229 * There's no need to reserve anything when we aren't using extents.
3230 * The space estimates are exact, there are no unwritten extents,
3231 * hole punching doesn't need new metadata... This is needed especially
3232 * to keep ext2/3 backward compatibility.
3233 */
3234 if (!ext4_has_feature_extents(sb))
3235 return;
3236 /*
3237 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3238 * This should cover the situations where we can not afford to run
3239 * out of space like for example punch hole, or converting
3240 * unwritten extents in delalloc path. In most cases such
3241 * allocation would require 1, or 2 blocks, higher numbers are
3242 * very rare.
3243 */
3244 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3245 sbi->s_cluster_bits);
3246
3247 do_div(resv_clusters, 50);
3248 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3249
3250 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3251 }
3252
3253 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3254 {
3255 char *orig_data = kstrdup(data, GFP_KERNEL);
3256 struct buffer_head *bh, **group_desc;
3257 struct ext4_super_block *es = NULL;
3258 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3259 struct flex_groups **flex_groups;
3260 ext4_fsblk_t block;
3261 ext4_fsblk_t sb_block = get_sb_block(&data);
3262 ext4_fsblk_t logical_sb_block;
3263 unsigned long offset = 0;
3264 unsigned long journal_devnum = 0;
3265 unsigned long def_mount_opts;
3266 struct inode *root;
3267 const char *descr;
3268 int ret = -ENOMEM;
3269 int blocksize, clustersize;
3270 unsigned int db_count;
3271 unsigned int i;
3272 int needs_recovery, has_huge_files, has_bigalloc;
3273 __u64 blocks_count;
3274 int err = 0;
3275 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3276 ext4_group_t first_not_zeroed;
3277
3278 if ((data && !orig_data) || !sbi)
3279 goto out_free_base;
3280
3281 sbi->s_blockgroup_lock =
3282 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3283 if (!sbi->s_blockgroup_lock)
3284 goto out_free_base;
3285
3286 sb->s_fs_info = sbi;
3287 sbi->s_sb = sb;
3288 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3289 sbi->s_sb_block = sb_block;
3290 if (sb->s_bdev->bd_part)
3291 sbi->s_sectors_written_start =
3292 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3293
3294 /* Cleanup superblock name */
3295 strreplace(sb->s_id, '/', '!');
3296
3297 /* -EINVAL is default */
3298 ret = -EINVAL;
3299 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3300 if (!blocksize) {
3301 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3302 goto out_fail;
3303 }
3304
3305 /*
3306 * The ext4 superblock will not be buffer aligned for other than 1kB
3307 * block sizes. We need to calculate the offset from buffer start.
3308 */
3309 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3310 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3311 offset = do_div(logical_sb_block, blocksize);
3312 } else {
3313 logical_sb_block = sb_block;
3314 }
3315
3316 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3317 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3318 goto out_fail;
3319 }
3320 /*
3321 * Note: s_es must be initialized as soon as possible because
3322 * some ext4 macro-instructions depend on its value
3323 */
3324 es = (struct ext4_super_block *) (bh->b_data + offset);
3325 sbi->s_es = es;
3326 sb->s_magic = le16_to_cpu(es->s_magic);
3327 if (sb->s_magic != EXT4_SUPER_MAGIC)
3328 goto cantfind_ext4;
3329 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3330
3331 /* Warn if metadata_csum and gdt_csum are both set. */
3332 if (ext4_has_feature_metadata_csum(sb) &&
3333 ext4_has_feature_gdt_csum(sb))
3334 ext4_warning(sb, "metadata_csum and uninit_bg are "
3335 "redundant flags; please run fsck.");
3336
3337 /* Check for a known checksum algorithm */
3338 if (!ext4_verify_csum_type(sb, es)) {
3339 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3340 "unknown checksum algorithm.");
3341 silent = 1;
3342 goto cantfind_ext4;
3343 }
3344
3345 /* Load the checksum driver */
3346 if (ext4_has_feature_metadata_csum(sb)) {
3347 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3348 if (IS_ERR(sbi->s_chksum_driver)) {
3349 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3350 ret = PTR_ERR(sbi->s_chksum_driver);
3351 sbi->s_chksum_driver = NULL;
3352 goto failed_mount;
3353 }
3354 }
3355
3356 /* Check superblock checksum */
3357 if (!ext4_superblock_csum_verify(sb, es)) {
3358 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3359 "invalid superblock checksum. Run e2fsck?");
3360 silent = 1;
3361 ret = -EFSBADCRC;
3362 goto cantfind_ext4;
3363 }
3364
3365 /* Precompute checksum seed for all metadata */
3366 if (ext4_has_feature_csum_seed(sb))
3367 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3368 else if (ext4_has_metadata_csum(sb))
3369 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3370 sizeof(es->s_uuid));
3371
3372 /* Set defaults before we parse the mount options */
3373 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3374 set_opt(sb, INIT_INODE_TABLE);
3375 if (def_mount_opts & EXT4_DEFM_DEBUG)
3376 set_opt(sb, DEBUG);
3377 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3378 set_opt(sb, GRPID);
3379 if (def_mount_opts & EXT4_DEFM_UID16)
3380 set_opt(sb, NO_UID32);
3381 /* xattr user namespace & acls are now defaulted on */
3382 set_opt(sb, XATTR_USER);
3383 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3384 set_opt(sb, POSIX_ACL);
3385 #endif
3386 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3387 if (ext4_has_metadata_csum(sb))
3388 set_opt(sb, JOURNAL_CHECKSUM);
3389
3390 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3391 set_opt(sb, JOURNAL_DATA);
3392 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3393 set_opt(sb, ORDERED_DATA);
3394 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3395 set_opt(sb, WRITEBACK_DATA);
3396
3397 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3398 set_opt(sb, ERRORS_PANIC);
3399 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3400 set_opt(sb, ERRORS_CONT);
3401 else
3402 set_opt(sb, ERRORS_RO);
3403 /* block_validity enabled by default; disable with noblock_validity */
3404 set_opt(sb, BLOCK_VALIDITY);
3405 if (def_mount_opts & EXT4_DEFM_DISCARD)
3406 set_opt(sb, DISCARD);
3407
3408 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3409 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3410 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3411 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3412 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3413
3414 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3415 set_opt(sb, BARRIER);
3416
3417 /*
3418 * enable delayed allocation by default
3419 * Use -o nodelalloc to turn it off
3420 */
3421 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3422 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3423 set_opt(sb, DELALLOC);
3424
3425 /*
3426 * set default s_li_wait_mult for lazyinit, for the case there is
3427 * no mount option specified.
3428 */
3429 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3430
3431 if (sbi->s_es->s_mount_opts[0]) {
3432 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3433 sizeof(sbi->s_es->s_mount_opts),
3434 GFP_KERNEL);
3435 if (!s_mount_opts)
3436 goto failed_mount;
3437 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3438 &journal_ioprio, 0)) {
3439 ext4_msg(sb, KERN_WARNING,
3440 "failed to parse options in superblock: %s",
3441 s_mount_opts);
3442 }
3443 kfree(s_mount_opts);
3444 }
3445 sbi->s_def_mount_opt = sbi->s_mount_opt;
3446 if (!parse_options((char *) data, sb, &journal_devnum,
3447 &journal_ioprio, 0))
3448 goto failed_mount;
3449
3450 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3451 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3452 "with data=journal disables delayed "
3453 "allocation and O_DIRECT support!\n");
3454 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3455 ext4_msg(sb, KERN_ERR, "can't mount with "
3456 "both data=journal and delalloc");
3457 goto failed_mount;
3458 }
3459 if (test_opt(sb, DIOREAD_NOLOCK)) {
3460 ext4_msg(sb, KERN_ERR, "can't mount with "
3461 "both data=journal and dioread_nolock");
3462 goto failed_mount;
3463 }
3464 if (test_opt(sb, DAX)) {
3465 ext4_msg(sb, KERN_ERR, "can't mount with "
3466 "both data=journal and dax");
3467 goto failed_mount;
3468 }
3469 if (ext4_has_feature_encrypt(sb)) {
3470 ext4_msg(sb, KERN_WARNING,
3471 "encrypted files will use data=ordered "
3472 "instead of data journaling mode");
3473 }
3474 if (test_opt(sb, DELALLOC))
3475 clear_opt(sb, DELALLOC);
3476 } else {
3477 sb->s_iflags |= SB_I_CGROUPWB;
3478 }
3479
3480 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3481 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3482
3483 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3484 (ext4_has_compat_features(sb) ||
3485 ext4_has_ro_compat_features(sb) ||
3486 ext4_has_incompat_features(sb)))
3487 ext4_msg(sb, KERN_WARNING,
3488 "feature flags set on rev 0 fs, "
3489 "running e2fsck is recommended");
3490
3491 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3492 set_opt2(sb, HURD_COMPAT);
3493 if (ext4_has_feature_64bit(sb)) {
3494 ext4_msg(sb, KERN_ERR,
3495 "The Hurd can't support 64-bit file systems");
3496 goto failed_mount;
3497 }
3498 }
3499
3500 if (IS_EXT2_SB(sb)) {
3501 if (ext2_feature_set_ok(sb))
3502 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3503 "using the ext4 subsystem");
3504 else {
3505 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3506 "to feature incompatibilities");
3507 goto failed_mount;
3508 }
3509 }
3510
3511 if (IS_EXT3_SB(sb)) {
3512 if (ext3_feature_set_ok(sb))
3513 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3514 "using the ext4 subsystem");
3515 else {
3516 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3517 "to feature incompatibilities");
3518 goto failed_mount;
3519 }
3520 }
3521
3522 /*
3523 * Check feature flags regardless of the revision level, since we
3524 * previously didn't change the revision level when setting the flags,
3525 * so there is a chance incompat flags are set on a rev 0 filesystem.
3526 */
3527 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3528 goto failed_mount;
3529
3530 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3531 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3532 blocksize > EXT4_MAX_BLOCK_SIZE) {
3533 ext4_msg(sb, KERN_ERR,
3534 "Unsupported filesystem blocksize %d (%d log_block_size)",
3535 blocksize, le32_to_cpu(es->s_log_block_size));
3536 goto failed_mount;
3537 }
3538 if (le32_to_cpu(es->s_log_block_size) >
3539 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3540 ext4_msg(sb, KERN_ERR,
3541 "Invalid log block size: %u",
3542 le32_to_cpu(es->s_log_block_size));
3543 goto failed_mount;
3544 }
3545 if (le32_to_cpu(es->s_log_cluster_size) >
3546 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3547 ext4_msg(sb, KERN_ERR,
3548 "Invalid log cluster size: %u",
3549 le32_to_cpu(es->s_log_cluster_size));
3550 goto failed_mount;
3551 }
3552
3553 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3554 ext4_msg(sb, KERN_ERR,
3555 "Number of reserved GDT blocks insanely large: %d",
3556 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3557 goto failed_mount;
3558 }
3559
3560 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3561 if (blocksize != PAGE_SIZE) {
3562 ext4_msg(sb, KERN_ERR,
3563 "error: unsupported blocksize for dax");
3564 goto failed_mount;
3565 }
3566 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3567 ext4_msg(sb, KERN_ERR,
3568 "error: device does not support dax");
3569 goto failed_mount;
3570 }
3571 }
3572
3573 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3574 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3575 es->s_encryption_level);
3576 goto failed_mount;
3577 }
3578
3579 if (sb->s_blocksize != blocksize) {
3580 /* Validate the filesystem blocksize */
3581 if (!sb_set_blocksize(sb, blocksize)) {
3582 ext4_msg(sb, KERN_ERR, "bad block size %d",
3583 blocksize);
3584 goto failed_mount;
3585 }
3586
3587 brelse(bh);
3588 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3589 offset = do_div(logical_sb_block, blocksize);
3590 bh = sb_bread_unmovable(sb, logical_sb_block);
3591 if (!bh) {
3592 ext4_msg(sb, KERN_ERR,
3593 "Can't read superblock on 2nd try");
3594 goto failed_mount;
3595 }
3596 es = (struct ext4_super_block *)(bh->b_data + offset);
3597 sbi->s_es = es;
3598 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3599 ext4_msg(sb, KERN_ERR,
3600 "Magic mismatch, very weird!");
3601 goto failed_mount;
3602 }
3603 }
3604
3605 has_huge_files = ext4_has_feature_huge_file(sb);
3606 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3607 has_huge_files);
3608 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3609
3610 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3611 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3612 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3613 } else {
3614 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3615 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3616 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3617 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3618 sbi->s_first_ino);
3619 goto failed_mount;
3620 }
3621 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3622 (!is_power_of_2(sbi->s_inode_size)) ||
3623 (sbi->s_inode_size > blocksize)) {
3624 ext4_msg(sb, KERN_ERR,
3625 "unsupported inode size: %d",
3626 sbi->s_inode_size);
3627 goto failed_mount;
3628 }
3629 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3630 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3631 }
3632
3633 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3634 if (ext4_has_feature_64bit(sb)) {
3635 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3636 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3637 !is_power_of_2(sbi->s_desc_size)) {
3638 ext4_msg(sb, KERN_ERR,
3639 "unsupported descriptor size %lu",
3640 sbi->s_desc_size);
3641 goto failed_mount;
3642 }
3643 } else
3644 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3645
3646 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3647 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3648
3649 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3650 if (sbi->s_inodes_per_block == 0)
3651 goto cantfind_ext4;
3652 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3653 sbi->s_inodes_per_group > blocksize * 8) {
3654 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3655 sbi->s_inodes_per_group);
3656 goto failed_mount;
3657 }
3658 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3659 sbi->s_inodes_per_block;
3660 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3661 sbi->s_sbh = bh;
3662 sbi->s_mount_state = le16_to_cpu(es->s_state);
3663 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3664 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3665
3666 for (i = 0; i < 4; i++)
3667 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3668 sbi->s_def_hash_version = es->s_def_hash_version;
3669 if (ext4_has_feature_dir_index(sb)) {
3670 i = le32_to_cpu(es->s_flags);
3671 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3672 sbi->s_hash_unsigned = 3;
3673 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3674 #ifdef __CHAR_UNSIGNED__
3675 if (!(sb->s_flags & MS_RDONLY))
3676 es->s_flags |=
3677 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3678 sbi->s_hash_unsigned = 3;
3679 #else
3680 if (!(sb->s_flags & MS_RDONLY))
3681 es->s_flags |=
3682 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3683 #endif
3684 }
3685 }
3686
3687 /* Handle clustersize */
3688 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3689 has_bigalloc = ext4_has_feature_bigalloc(sb);
3690 if (has_bigalloc) {
3691 if (clustersize < blocksize) {
3692 ext4_msg(sb, KERN_ERR,
3693 "cluster size (%d) smaller than "
3694 "block size (%d)", clustersize, blocksize);
3695 goto failed_mount;
3696 }
3697 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3698 le32_to_cpu(es->s_log_block_size);
3699 sbi->s_clusters_per_group =
3700 le32_to_cpu(es->s_clusters_per_group);
3701 if (sbi->s_clusters_per_group > blocksize * 8) {
3702 ext4_msg(sb, KERN_ERR,
3703 "#clusters per group too big: %lu",
3704 sbi->s_clusters_per_group);
3705 goto failed_mount;
3706 }
3707 if (sbi->s_blocks_per_group !=
3708 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3709 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3710 "clusters per group (%lu) inconsistent",
3711 sbi->s_blocks_per_group,
3712 sbi->s_clusters_per_group);
3713 goto failed_mount;
3714 }
3715 } else {
3716 if (clustersize != blocksize) {
3717 ext4_msg(sb, KERN_ERR,
3718 "fragment/cluster size (%d) != "
3719 "block size (%d)", clustersize, blocksize);
3720 goto failed_mount;
3721 }
3722 if (sbi->s_blocks_per_group > blocksize * 8) {
3723 ext4_msg(sb, KERN_ERR,
3724 "#blocks per group too big: %lu",
3725 sbi->s_blocks_per_group);
3726 goto failed_mount;
3727 }
3728 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3729 sbi->s_cluster_bits = 0;
3730 }
3731 sbi->s_cluster_ratio = clustersize / blocksize;
3732
3733 /* Do we have standard group size of clustersize * 8 blocks ? */
3734 if (sbi->s_blocks_per_group == clustersize << 3)
3735 set_opt2(sb, STD_GROUP_SIZE);
3736
3737 /*
3738 * Test whether we have more sectors than will fit in sector_t,
3739 * and whether the max offset is addressable by the page cache.
3740 */
3741 err = generic_check_addressable(sb->s_blocksize_bits,
3742 ext4_blocks_count(es));
3743 if (err) {
3744 ext4_msg(sb, KERN_ERR, "filesystem"
3745 " too large to mount safely on this system");
3746 if (sizeof(sector_t) < 8)
3747 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3748 goto failed_mount;
3749 }
3750
3751 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3752 goto cantfind_ext4;
3753
3754 /* check blocks count against device size */
3755 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3756 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3757 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3758 "exceeds size of device (%llu blocks)",
3759 ext4_blocks_count(es), blocks_count);
3760 goto failed_mount;
3761 }
3762
3763 /*
3764 * It makes no sense for the first data block to be beyond the end
3765 * of the filesystem.
3766 */
3767 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3768 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3769 "block %u is beyond end of filesystem (%llu)",
3770 le32_to_cpu(es->s_first_data_block),
3771 ext4_blocks_count(es));
3772 goto failed_mount;
3773 }
3774 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
3775 (sbi->s_cluster_ratio == 1)) {
3776 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3777 "block is 0 with a 1k block and cluster size");
3778 goto failed_mount;
3779 }
3780
3781 blocks_count = (ext4_blocks_count(es) -
3782 le32_to_cpu(es->s_first_data_block) +
3783 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3784 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3785 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3786 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
3787 "(block count %llu, first data block %u, "
3788 "blocks per group %lu)", blocks_count,
3789 ext4_blocks_count(es),
3790 le32_to_cpu(es->s_first_data_block),
3791 EXT4_BLOCKS_PER_GROUP(sb));
3792 goto failed_mount;
3793 }
3794 sbi->s_groups_count = blocks_count;
3795 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3796 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3797 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
3798 le32_to_cpu(es->s_inodes_count)) {
3799 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
3800 le32_to_cpu(es->s_inodes_count),
3801 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
3802 ret = -EINVAL;
3803 goto failed_mount;
3804 }
3805 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3806 EXT4_DESC_PER_BLOCK(sb);
3807 if (ext4_has_feature_meta_bg(sb)) {
3808 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3809 ext4_msg(sb, KERN_WARNING,
3810 "first meta block group too large: %u "
3811 "(group descriptor block count %u)",
3812 le32_to_cpu(es->s_first_meta_bg), db_count);
3813 goto failed_mount;
3814 }
3815 }
3816 rcu_assign_pointer(sbi->s_group_desc,
3817 ext4_kvmalloc(db_count *
3818 sizeof(struct buffer_head *),
3819 GFP_KERNEL));
3820 if (sbi->s_group_desc == NULL) {
3821 ext4_msg(sb, KERN_ERR, "not enough memory");
3822 ret = -ENOMEM;
3823 goto failed_mount;
3824 }
3825
3826 bgl_lock_init(sbi->s_blockgroup_lock);
3827
3828 for (i = 0; i < db_count; i++) {
3829 struct buffer_head *bh;
3830
3831 block = descriptor_loc(sb, logical_sb_block, i);
3832 bh = sb_bread_unmovable(sb, block);
3833 if (!bh) {
3834 ext4_msg(sb, KERN_ERR,
3835 "can't read group descriptor %d", i);
3836 db_count = i;
3837 goto failed_mount2;
3838 }
3839 rcu_read_lock();
3840 rcu_dereference(sbi->s_group_desc)[i] = bh;
3841 rcu_read_unlock();
3842 }
3843 sbi->s_gdb_count = db_count;
3844 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3845 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3846 ret = -EFSCORRUPTED;
3847 goto failed_mount2;
3848 }
3849
3850 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3851 spin_lock_init(&sbi->s_next_gen_lock);
3852
3853 setup_timer(&sbi->s_err_report, print_daily_error_info,
3854 (unsigned long) sb);
3855
3856 /* Register extent status tree shrinker */
3857 if (ext4_es_register_shrinker(sbi))
3858 goto failed_mount3;
3859
3860 sbi->s_stripe = ext4_get_stripe_size(sbi);
3861 sbi->s_extent_max_zeroout_kb = 32;
3862
3863 /*
3864 * set up enough so that it can read an inode
3865 */
3866 sb->s_op = &ext4_sops;
3867 sb->s_export_op = &ext4_export_ops;
3868 sb->s_xattr = ext4_xattr_handlers;
3869 #ifdef CONFIG_QUOTA
3870 sb->dq_op = &ext4_quota_operations;
3871 if (ext4_has_feature_quota(sb))
3872 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3873 else
3874 sb->s_qcop = &ext4_qctl_operations;
3875 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3876 #endif
3877 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3878
3879 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3880 mutex_init(&sbi->s_orphan_lock);
3881
3882 sb->s_root = NULL;
3883
3884 needs_recovery = (es->s_last_orphan != 0 ||
3885 ext4_has_feature_journal_needs_recovery(sb));
3886
3887 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3888 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3889 goto failed_mount3a;
3890
3891 /*
3892 * The first inode we look at is the journal inode. Don't try
3893 * root first: it may be modified in the journal!
3894 */
3895 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3896 err = ext4_load_journal(sb, es, journal_devnum);
3897 if (err)
3898 goto failed_mount3a;
3899 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3900 ext4_has_feature_journal_needs_recovery(sb)) {
3901 ext4_msg(sb, KERN_ERR, "required journal recovery "
3902 "suppressed and not mounted read-only");
3903 goto failed_mount_wq;
3904 } else {
3905 /* Nojournal mode, all journal mount options are illegal */
3906 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3907 ext4_msg(sb, KERN_ERR, "can't mount with "
3908 "journal_checksum, fs mounted w/o journal");
3909 goto failed_mount_wq;
3910 }
3911 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3912 ext4_msg(sb, KERN_ERR, "can't mount with "
3913 "journal_async_commit, fs mounted w/o journal");
3914 goto failed_mount_wq;
3915 }
3916 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3917 ext4_msg(sb, KERN_ERR, "can't mount with "
3918 "commit=%lu, fs mounted w/o journal",
3919 sbi->s_commit_interval / HZ);
3920 goto failed_mount_wq;
3921 }
3922 if (EXT4_MOUNT_DATA_FLAGS &
3923 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3924 ext4_msg(sb, KERN_ERR, "can't mount with "
3925 "data=, fs mounted w/o journal");
3926 goto failed_mount_wq;
3927 }
3928 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
3929 clear_opt(sb, JOURNAL_CHECKSUM);
3930 clear_opt(sb, DATA_FLAGS);
3931 sbi->s_journal = NULL;
3932 needs_recovery = 0;
3933 goto no_journal;
3934 }
3935
3936 if (ext4_has_feature_64bit(sb) &&
3937 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3938 JBD2_FEATURE_INCOMPAT_64BIT)) {
3939 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3940 goto failed_mount_wq;
3941 }
3942
3943 if (!set_journal_csum_feature_set(sb)) {
3944 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3945 "feature set");
3946 goto failed_mount_wq;
3947 }
3948
3949 /* We have now updated the journal if required, so we can
3950 * validate the data journaling mode. */
3951 switch (test_opt(sb, DATA_FLAGS)) {
3952 case 0:
3953 /* No mode set, assume a default based on the journal
3954 * capabilities: ORDERED_DATA if the journal can
3955 * cope, else JOURNAL_DATA
3956 */
3957 if (jbd2_journal_check_available_features
3958 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3959 set_opt(sb, ORDERED_DATA);
3960 else
3961 set_opt(sb, JOURNAL_DATA);
3962 break;
3963
3964 case EXT4_MOUNT_ORDERED_DATA:
3965 case EXT4_MOUNT_WRITEBACK_DATA:
3966 if (!jbd2_journal_check_available_features
3967 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3968 ext4_msg(sb, KERN_ERR, "Journal does not support "
3969 "requested data journaling mode");
3970 goto failed_mount_wq;
3971 }
3972 default:
3973 break;
3974 }
3975 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3976
3977 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3978
3979 no_journal:
3980 if (ext4_mballoc_ready) {
3981 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
3982 if (!sbi->s_mb_cache) {
3983 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
3984 goto failed_mount_wq;
3985 }
3986 }
3987
3988 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
3989 (blocksize != PAGE_CACHE_SIZE)) {
3990 ext4_msg(sb, KERN_ERR,
3991 "Unsupported blocksize for fs encryption");
3992 goto failed_mount_wq;
3993 }
3994
3995 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
3996 !ext4_has_feature_encrypt(sb)) {
3997 ext4_set_feature_encrypt(sb);
3998 ext4_commit_super(sb, 1);
3999 }
4000
4001 /*
4002 * Get the # of file system overhead blocks from the
4003 * superblock if present.
4004 */
4005 if (es->s_overhead_clusters)
4006 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4007 else {
4008 err = ext4_calculate_overhead(sb);
4009 if (err)
4010 goto failed_mount_wq;
4011 }
4012
4013 /*
4014 * The maximum number of concurrent works can be high and
4015 * concurrency isn't really necessary. Limit it to 1.
4016 */
4017 EXT4_SB(sb)->rsv_conversion_wq =
4018 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4019 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4020 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4021 ret = -ENOMEM;
4022 goto failed_mount4;
4023 }
4024
4025 /*
4026 * The jbd2_journal_load will have done any necessary log recovery,
4027 * so we can safely mount the rest of the filesystem now.
4028 */
4029
4030 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4031 if (IS_ERR(root)) {
4032 ext4_msg(sb, KERN_ERR, "get root inode failed");
4033 ret = PTR_ERR(root);
4034 root = NULL;
4035 goto failed_mount4;
4036 }
4037 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4038 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4039 iput(root);
4040 goto failed_mount4;
4041 }
4042 sb->s_root = d_make_root(root);
4043 if (!sb->s_root) {
4044 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4045 ret = -ENOMEM;
4046 goto failed_mount4;
4047 }
4048
4049 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4050 sb->s_flags |= MS_RDONLY;
4051
4052 ext4_clamp_want_extra_isize(sb);
4053
4054 ext4_set_resv_clusters(sb);
4055
4056 err = ext4_setup_system_zone(sb);
4057 if (err) {
4058 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4059 "zone (%d)", err);
4060 goto failed_mount4a;
4061 }
4062
4063 ext4_ext_init(sb);
4064 err = ext4_mb_init(sb);
4065 if (err) {
4066 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4067 err);
4068 goto failed_mount5;
4069 }
4070
4071 block = ext4_count_free_clusters(sb);
4072 ext4_free_blocks_count_set(sbi->s_es,
4073 EXT4_C2B(sbi, block));
4074 ext4_superblock_csum_set(sb);
4075 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4076 GFP_KERNEL);
4077 if (!err) {
4078 unsigned long freei = ext4_count_free_inodes(sb);
4079 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4080 ext4_superblock_csum_set(sb);
4081 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4082 GFP_KERNEL);
4083 }
4084 if (!err)
4085 err = percpu_counter_init(&sbi->s_dirs_counter,
4086 ext4_count_dirs(sb), GFP_KERNEL);
4087 if (!err)
4088 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4089 GFP_KERNEL);
4090 if (err) {
4091 ext4_msg(sb, KERN_ERR, "insufficient memory");
4092 goto failed_mount6;
4093 }
4094
4095 if (ext4_has_feature_flex_bg(sb))
4096 if (!ext4_fill_flex_info(sb)) {
4097 ext4_msg(sb, KERN_ERR,
4098 "unable to initialize "
4099 "flex_bg meta info!");
4100 goto failed_mount6;
4101 }
4102
4103 err = ext4_register_li_request(sb, first_not_zeroed);
4104 if (err)
4105 goto failed_mount6;
4106
4107 err = ext4_register_sysfs(sb);
4108 if (err)
4109 goto failed_mount7;
4110
4111 #ifdef CONFIG_QUOTA
4112 /* Enable quota usage during mount. */
4113 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4114 err = ext4_enable_quotas(sb);
4115 if (err)
4116 goto failed_mount8;
4117 }
4118 #endif /* CONFIG_QUOTA */
4119
4120 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4121 ext4_orphan_cleanup(sb, es);
4122 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4123 if (needs_recovery) {
4124 ext4_msg(sb, KERN_INFO, "recovery complete");
4125 ext4_mark_recovery_complete(sb, es);
4126 }
4127 if (EXT4_SB(sb)->s_journal) {
4128 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4129 descr = " journalled data mode";
4130 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4131 descr = " ordered data mode";
4132 else
4133 descr = " writeback data mode";
4134 } else
4135 descr = "out journal";
4136
4137 if (test_opt(sb, DISCARD)) {
4138 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4139 if (!blk_queue_discard(q))
4140 ext4_msg(sb, KERN_WARNING,
4141 "mounting with \"discard\" option, but "
4142 "the device does not support discard");
4143 }
4144
4145 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4146 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4147 "Opts: %.*s%s%s", descr,
4148 (int) sizeof(sbi->s_es->s_mount_opts),
4149 sbi->s_es->s_mount_opts,
4150 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4151
4152 if (es->s_error_count)
4153 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4154
4155 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4156 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4157 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4158 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4159
4160 kfree(orig_data);
4161 return 0;
4162
4163 cantfind_ext4:
4164 if (!silent)
4165 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4166 goto failed_mount;
4167
4168 #ifdef CONFIG_QUOTA
4169 failed_mount8:
4170 ext4_unregister_sysfs(sb);
4171 #endif
4172 failed_mount7:
4173 ext4_unregister_li_request(sb);
4174 failed_mount6:
4175 ext4_mb_release(sb);
4176 rcu_read_lock();
4177 flex_groups = rcu_dereference(sbi->s_flex_groups);
4178 if (flex_groups) {
4179 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4180 kvfree(flex_groups[i]);
4181 kvfree(flex_groups);
4182 }
4183 rcu_read_unlock();
4184 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4185 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4186 percpu_counter_destroy(&sbi->s_dirs_counter);
4187 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4188 failed_mount5:
4189 ext4_ext_release(sb);
4190 ext4_release_system_zone(sb);
4191 failed_mount4a:
4192 dput(sb->s_root);
4193 sb->s_root = NULL;
4194 failed_mount4:
4195 ext4_msg(sb, KERN_ERR, "mount failed");
4196 if (EXT4_SB(sb)->rsv_conversion_wq)
4197 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4198 failed_mount_wq:
4199 if (sbi->s_journal) {
4200 jbd2_journal_destroy(sbi->s_journal);
4201 sbi->s_journal = NULL;
4202 }
4203 failed_mount3a:
4204 ext4_es_unregister_shrinker(sbi);
4205 failed_mount3:
4206 del_timer_sync(&sbi->s_err_report);
4207 if (sbi->s_mmp_tsk)
4208 kthread_stop(sbi->s_mmp_tsk);
4209 failed_mount2:
4210 rcu_read_lock();
4211 group_desc = rcu_dereference(sbi->s_group_desc);
4212 for (i = 0; i < db_count; i++)
4213 brelse(group_desc[i]);
4214 kvfree(group_desc);
4215 rcu_read_unlock();
4216 failed_mount:
4217 if (sbi->s_chksum_driver)
4218 crypto_free_shash(sbi->s_chksum_driver);
4219 #ifdef CONFIG_QUOTA
4220 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4221 kfree(sbi->s_qf_names[i]);
4222 #endif
4223 ext4_blkdev_remove(sbi);
4224 brelse(bh);
4225 out_fail:
4226 sb->s_fs_info = NULL;
4227 kfree(sbi->s_blockgroup_lock);
4228 out_free_base:
4229 kfree(sbi);
4230 kfree(orig_data);
4231 return err ? err : ret;
4232 }
4233
4234 /*
4235 * Setup any per-fs journal parameters now. We'll do this both on
4236 * initial mount, once the journal has been initialised but before we've
4237 * done any recovery; and again on any subsequent remount.
4238 */
4239 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4240 {
4241 struct ext4_sb_info *sbi = EXT4_SB(sb);
4242
4243 journal->j_commit_interval = sbi->s_commit_interval;
4244 journal->j_min_batch_time = sbi->s_min_batch_time;
4245 journal->j_max_batch_time = sbi->s_max_batch_time;
4246
4247 write_lock(&journal->j_state_lock);
4248 if (test_opt(sb, BARRIER))
4249 journal->j_flags |= JBD2_BARRIER;
4250 else
4251 journal->j_flags &= ~JBD2_BARRIER;
4252 if (test_opt(sb, DATA_ERR_ABORT))
4253 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4254 else
4255 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4256 write_unlock(&journal->j_state_lock);
4257 }
4258
4259 static journal_t *ext4_get_journal(struct super_block *sb,
4260 unsigned int journal_inum)
4261 {
4262 struct inode *journal_inode;
4263 journal_t *journal;
4264
4265 BUG_ON(!ext4_has_feature_journal(sb));
4266
4267 /*
4268 * Test for the existence of a valid inode on disk. Bad things
4269 * happen if we iget() an unused inode, as the subsequent iput()
4270 * will try to delete it.
4271 */
4272 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
4273 if (IS_ERR(journal_inode)) {
4274 ext4_msg(sb, KERN_ERR, "no journal found");
4275 return NULL;
4276 }
4277 if (!journal_inode->i_nlink) {
4278 make_bad_inode(journal_inode);
4279 iput(journal_inode);
4280 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4281 return NULL;
4282 }
4283
4284 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4285 journal_inode, journal_inode->i_size);
4286 if (!S_ISREG(journal_inode->i_mode)) {
4287 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4288 iput(journal_inode);
4289 return NULL;
4290 }
4291
4292 journal = jbd2_journal_init_inode(journal_inode);
4293 if (!journal) {
4294 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4295 iput(journal_inode);
4296 return NULL;
4297 }
4298 journal->j_private = sb;
4299 ext4_init_journal_params(sb, journal);
4300 return journal;
4301 }
4302
4303 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4304 dev_t j_dev)
4305 {
4306 struct buffer_head *bh;
4307 journal_t *journal;
4308 ext4_fsblk_t start;
4309 ext4_fsblk_t len;
4310 int hblock, blocksize;
4311 ext4_fsblk_t sb_block;
4312 unsigned long offset;
4313 struct ext4_super_block *es;
4314 struct block_device *bdev;
4315
4316 BUG_ON(!ext4_has_feature_journal(sb));
4317
4318 bdev = ext4_blkdev_get(j_dev, sb);
4319 if (bdev == NULL)
4320 return NULL;
4321
4322 blocksize = sb->s_blocksize;
4323 hblock = bdev_logical_block_size(bdev);
4324 if (blocksize < hblock) {
4325 ext4_msg(sb, KERN_ERR,
4326 "blocksize too small for journal device");
4327 goto out_bdev;
4328 }
4329
4330 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4331 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4332 set_blocksize(bdev, blocksize);
4333 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4334 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4335 "external journal");
4336 goto out_bdev;
4337 }
4338
4339 es = (struct ext4_super_block *) (bh->b_data + offset);
4340 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4341 !(le32_to_cpu(es->s_feature_incompat) &
4342 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4343 ext4_msg(sb, KERN_ERR, "external journal has "
4344 "bad superblock");
4345 brelse(bh);
4346 goto out_bdev;
4347 }
4348
4349 if ((le32_to_cpu(es->s_feature_ro_compat) &
4350 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4351 es->s_checksum != ext4_superblock_csum(sb, es)) {
4352 ext4_msg(sb, KERN_ERR, "external journal has "
4353 "corrupt superblock");
4354 brelse(bh);
4355 goto out_bdev;
4356 }
4357
4358 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4359 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4360 brelse(bh);
4361 goto out_bdev;
4362 }
4363
4364 len = ext4_blocks_count(es);
4365 start = sb_block + 1;
4366 brelse(bh); /* we're done with the superblock */
4367
4368 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4369 start, len, blocksize);
4370 if (!journal) {
4371 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4372 goto out_bdev;
4373 }
4374 journal->j_private = sb;
4375 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4376 wait_on_buffer(journal->j_sb_buffer);
4377 if (!buffer_uptodate(journal->j_sb_buffer)) {
4378 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4379 goto out_journal;
4380 }
4381 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4382 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4383 "user (unsupported) - %d",
4384 be32_to_cpu(journal->j_superblock->s_nr_users));
4385 goto out_journal;
4386 }
4387 EXT4_SB(sb)->journal_bdev = bdev;
4388 ext4_init_journal_params(sb, journal);
4389 return journal;
4390
4391 out_journal:
4392 jbd2_journal_destroy(journal);
4393 out_bdev:
4394 ext4_blkdev_put(bdev);
4395 return NULL;
4396 }
4397
4398 static int ext4_load_journal(struct super_block *sb,
4399 struct ext4_super_block *es,
4400 unsigned long journal_devnum)
4401 {
4402 journal_t *journal;
4403 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4404 dev_t journal_dev;
4405 int err = 0;
4406 int really_read_only;
4407
4408 BUG_ON(!ext4_has_feature_journal(sb));
4409
4410 if (journal_devnum &&
4411 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4412 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4413 "numbers have changed");
4414 journal_dev = new_decode_dev(journal_devnum);
4415 } else
4416 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4417
4418 really_read_only = bdev_read_only(sb->s_bdev);
4419
4420 /*
4421 * Are we loading a blank journal or performing recovery after a
4422 * crash? For recovery, we need to check in advance whether we
4423 * can get read-write access to the device.
4424 */
4425 if (ext4_has_feature_journal_needs_recovery(sb)) {
4426 if (sb->s_flags & MS_RDONLY) {
4427 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4428 "required on readonly filesystem");
4429 if (really_read_only) {
4430 ext4_msg(sb, KERN_ERR, "write access "
4431 "unavailable, cannot proceed");
4432 return -EROFS;
4433 }
4434 ext4_msg(sb, KERN_INFO, "write access will "
4435 "be enabled during recovery");
4436 }
4437 }
4438
4439 if (journal_inum && journal_dev) {
4440 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4441 "and inode journals!");
4442 return -EINVAL;
4443 }
4444
4445 if (journal_inum) {
4446 if (!(journal = ext4_get_journal(sb, journal_inum)))
4447 return -EINVAL;
4448 } else {
4449 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4450 return -EINVAL;
4451 }
4452
4453 if (!(journal->j_flags & JBD2_BARRIER))
4454 ext4_msg(sb, KERN_INFO, "barriers disabled");
4455
4456 if (!ext4_has_feature_journal_needs_recovery(sb))
4457 err = jbd2_journal_wipe(journal, !really_read_only);
4458 if (!err) {
4459 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4460 if (save)
4461 memcpy(save, ((char *) es) +
4462 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4463 err = jbd2_journal_load(journal);
4464 if (save)
4465 memcpy(((char *) es) + EXT4_S_ERR_START,
4466 save, EXT4_S_ERR_LEN);
4467 kfree(save);
4468 }
4469
4470 if (err) {
4471 ext4_msg(sb, KERN_ERR, "error loading journal");
4472 jbd2_journal_destroy(journal);
4473 return err;
4474 }
4475
4476 EXT4_SB(sb)->s_journal = journal;
4477 ext4_clear_journal_err(sb, es);
4478
4479 if (!really_read_only && journal_devnum &&
4480 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4481 es->s_journal_dev = cpu_to_le32(journal_devnum);
4482
4483 /* Make sure we flush the recovery flag to disk. */
4484 ext4_commit_super(sb, 1);
4485 }
4486
4487 return 0;
4488 }
4489
4490 static int ext4_commit_super(struct super_block *sb, int sync)
4491 {
4492 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4493 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4494 int error = 0;
4495
4496 if (!sbh || block_device_ejected(sb))
4497 return error;
4498
4499 /*
4500 * The superblock bh should be mapped, but it might not be if the
4501 * device was hot-removed. Not much we can do but fail the I/O.
4502 */
4503 if (!buffer_mapped(sbh))
4504 return error;
4505
4506 if (buffer_write_io_error(sbh)) {
4507 /*
4508 * Oh, dear. A previous attempt to write the
4509 * superblock failed. This could happen because the
4510 * USB device was yanked out. Or it could happen to
4511 * be a transient write error and maybe the block will
4512 * be remapped. Nothing we can do but to retry the
4513 * write and hope for the best.
4514 */
4515 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4516 "superblock detected");
4517 clear_buffer_write_io_error(sbh);
4518 set_buffer_uptodate(sbh);
4519 }
4520 /*
4521 * If the file system is mounted read-only, don't update the
4522 * superblock write time. This avoids updating the superblock
4523 * write time when we are mounting the root file system
4524 * read/only but we need to replay the journal; at that point,
4525 * for people who are east of GMT and who make their clock
4526 * tick in localtime for Windows bug-for-bug compatibility,
4527 * the clock is set in the future, and this will cause e2fsck
4528 * to complain and force a full file system check.
4529 */
4530 if (!(sb->s_flags & MS_RDONLY))
4531 es->s_wtime = cpu_to_le32(get_seconds());
4532 if (sb->s_bdev->bd_part)
4533 es->s_kbytes_written =
4534 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4535 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4536 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4537 else
4538 es->s_kbytes_written =
4539 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4540 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4541 ext4_free_blocks_count_set(es,
4542 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4543 &EXT4_SB(sb)->s_freeclusters_counter)));
4544 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4545 es->s_free_inodes_count =
4546 cpu_to_le32(percpu_counter_sum_positive(
4547 &EXT4_SB(sb)->s_freeinodes_counter));
4548 BUFFER_TRACE(sbh, "marking dirty");
4549 ext4_superblock_csum_set(sb);
4550 mark_buffer_dirty(sbh);
4551 if (sync) {
4552 error = __sync_dirty_buffer(sbh,
4553 test_opt(sb, BARRIER) ? WRITE_FUA : WRITE_SYNC);
4554 if (error)
4555 return error;
4556
4557 error = buffer_write_io_error(sbh);
4558 if (error) {
4559 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4560 "superblock");
4561 clear_buffer_write_io_error(sbh);
4562 set_buffer_uptodate(sbh);
4563 }
4564 }
4565 return error;
4566 }
4567
4568 /*
4569 * Have we just finished recovery? If so, and if we are mounting (or
4570 * remounting) the filesystem readonly, then we will end up with a
4571 * consistent fs on disk. Record that fact.
4572 */
4573 static void ext4_mark_recovery_complete(struct super_block *sb,
4574 struct ext4_super_block *es)
4575 {
4576 journal_t *journal = EXT4_SB(sb)->s_journal;
4577
4578 if (!ext4_has_feature_journal(sb)) {
4579 BUG_ON(journal != NULL);
4580 return;
4581 }
4582 jbd2_journal_lock_updates(journal);
4583 if (jbd2_journal_flush(journal) < 0)
4584 goto out;
4585
4586 if (ext4_has_feature_journal_needs_recovery(sb) &&
4587 sb->s_flags & MS_RDONLY) {
4588 ext4_clear_feature_journal_needs_recovery(sb);
4589 ext4_commit_super(sb, 1);
4590 }
4591
4592 out:
4593 jbd2_journal_unlock_updates(journal);
4594 }
4595
4596 /*
4597 * If we are mounting (or read-write remounting) a filesystem whose journal
4598 * has recorded an error from a previous lifetime, move that error to the
4599 * main filesystem now.
4600 */
4601 static void ext4_clear_journal_err(struct super_block *sb,
4602 struct ext4_super_block *es)
4603 {
4604 journal_t *journal;
4605 int j_errno;
4606 const char *errstr;
4607
4608 BUG_ON(!ext4_has_feature_journal(sb));
4609
4610 journal = EXT4_SB(sb)->s_journal;
4611
4612 /*
4613 * Now check for any error status which may have been recorded in the
4614 * journal by a prior ext4_error() or ext4_abort()
4615 */
4616
4617 j_errno = jbd2_journal_errno(journal);
4618 if (j_errno) {
4619 char nbuf[16];
4620
4621 errstr = ext4_decode_error(sb, j_errno, nbuf);
4622 ext4_warning(sb, "Filesystem error recorded "
4623 "from previous mount: %s", errstr);
4624 ext4_warning(sb, "Marking fs in need of filesystem check.");
4625
4626 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4627 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4628 ext4_commit_super(sb, 1);
4629
4630 jbd2_journal_clear_err(journal);
4631 jbd2_journal_update_sb_errno(journal);
4632 }
4633 }
4634
4635 /*
4636 * Force the running and committing transactions to commit,
4637 * and wait on the commit.
4638 */
4639 int ext4_force_commit(struct super_block *sb)
4640 {
4641 journal_t *journal;
4642
4643 if (sb->s_flags & MS_RDONLY)
4644 return 0;
4645
4646 journal = EXT4_SB(sb)->s_journal;
4647 return ext4_journal_force_commit(journal);
4648 }
4649
4650 static int ext4_sync_fs(struct super_block *sb, int wait)
4651 {
4652 int ret = 0;
4653 tid_t target;
4654 bool needs_barrier = false;
4655 struct ext4_sb_info *sbi = EXT4_SB(sb);
4656
4657 trace_ext4_sync_fs(sb, wait);
4658 flush_workqueue(sbi->rsv_conversion_wq);
4659 /*
4660 * Writeback quota in non-journalled quota case - journalled quota has
4661 * no dirty dquots
4662 */
4663 dquot_writeback_dquots(sb, -1);
4664 /*
4665 * Data writeback is possible w/o journal transaction, so barrier must
4666 * being sent at the end of the function. But we can skip it if
4667 * transaction_commit will do it for us.
4668 */
4669 if (sbi->s_journal) {
4670 target = jbd2_get_latest_transaction(sbi->s_journal);
4671 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4672 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4673 needs_barrier = true;
4674
4675 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4676 if (wait)
4677 ret = jbd2_log_wait_commit(sbi->s_journal,
4678 target);
4679 }
4680 } else if (wait && test_opt(sb, BARRIER))
4681 needs_barrier = true;
4682 if (needs_barrier) {
4683 int err;
4684 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4685 if (!ret)
4686 ret = err;
4687 }
4688
4689 return ret;
4690 }
4691
4692 /*
4693 * LVM calls this function before a (read-only) snapshot is created. This
4694 * gives us a chance to flush the journal completely and mark the fs clean.
4695 *
4696 * Note that only this function cannot bring a filesystem to be in a clean
4697 * state independently. It relies on upper layer to stop all data & metadata
4698 * modifications.
4699 */
4700 static int ext4_freeze(struct super_block *sb)
4701 {
4702 int error = 0;
4703 journal_t *journal;
4704
4705 if (sb->s_flags & MS_RDONLY)
4706 return 0;
4707
4708 journal = EXT4_SB(sb)->s_journal;
4709
4710 if (journal) {
4711 /* Now we set up the journal barrier. */
4712 jbd2_journal_lock_updates(journal);
4713
4714 /*
4715 * Don't clear the needs_recovery flag if we failed to
4716 * flush the journal.
4717 */
4718 error = jbd2_journal_flush(journal);
4719 if (error < 0)
4720 goto out;
4721
4722 /* Journal blocked and flushed, clear needs_recovery flag. */
4723 ext4_clear_feature_journal_needs_recovery(sb);
4724 }
4725
4726 error = ext4_commit_super(sb, 1);
4727 out:
4728 if (journal)
4729 /* we rely on upper layer to stop further updates */
4730 jbd2_journal_unlock_updates(journal);
4731 return error;
4732 }
4733
4734 /*
4735 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4736 * flag here, even though the filesystem is not technically dirty yet.
4737 */
4738 static int ext4_unfreeze(struct super_block *sb)
4739 {
4740 if (sb->s_flags & MS_RDONLY)
4741 return 0;
4742
4743 if (EXT4_SB(sb)->s_journal) {
4744 /* Reset the needs_recovery flag before the fs is unlocked. */
4745 ext4_set_feature_journal_needs_recovery(sb);
4746 }
4747
4748 ext4_commit_super(sb, 1);
4749 return 0;
4750 }
4751
4752 /*
4753 * Structure to save mount options for ext4_remount's benefit
4754 */
4755 struct ext4_mount_options {
4756 unsigned long s_mount_opt;
4757 unsigned long s_mount_opt2;
4758 kuid_t s_resuid;
4759 kgid_t s_resgid;
4760 unsigned long s_commit_interval;
4761 u32 s_min_batch_time, s_max_batch_time;
4762 #ifdef CONFIG_QUOTA
4763 int s_jquota_fmt;
4764 char *s_qf_names[EXT4_MAXQUOTAS];
4765 #endif
4766 };
4767
4768 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4769 {
4770 struct ext4_super_block *es;
4771 struct ext4_sb_info *sbi = EXT4_SB(sb);
4772 unsigned long old_sb_flags;
4773 struct ext4_mount_options old_opts;
4774 int enable_quota = 0;
4775 ext4_group_t g;
4776 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4777 int err = 0;
4778 #ifdef CONFIG_QUOTA
4779 int i, j;
4780 #endif
4781 char *orig_data = kstrdup(data, GFP_KERNEL);
4782
4783 /* Store the original options */
4784 old_sb_flags = sb->s_flags;
4785 old_opts.s_mount_opt = sbi->s_mount_opt;
4786 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4787 old_opts.s_resuid = sbi->s_resuid;
4788 old_opts.s_resgid = sbi->s_resgid;
4789 old_opts.s_commit_interval = sbi->s_commit_interval;
4790 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4791 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4792 #ifdef CONFIG_QUOTA
4793 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4794 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4795 if (sbi->s_qf_names[i]) {
4796 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4797 GFP_KERNEL);
4798 if (!old_opts.s_qf_names[i]) {
4799 for (j = 0; j < i; j++)
4800 kfree(old_opts.s_qf_names[j]);
4801 kfree(orig_data);
4802 return -ENOMEM;
4803 }
4804 } else
4805 old_opts.s_qf_names[i] = NULL;
4806 #endif
4807 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4808 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4809
4810 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4811 err = -EINVAL;
4812 goto restore_opts;
4813 }
4814
4815 ext4_clamp_want_extra_isize(sb);
4816
4817 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4818 test_opt(sb, JOURNAL_CHECKSUM)) {
4819 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4820 "during remount not supported; ignoring");
4821 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4822 }
4823
4824 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4825 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4826 ext4_msg(sb, KERN_ERR, "can't mount with "
4827 "both data=journal and delalloc");
4828 err = -EINVAL;
4829 goto restore_opts;
4830 }
4831 if (test_opt(sb, DIOREAD_NOLOCK)) {
4832 ext4_msg(sb, KERN_ERR, "can't mount with "
4833 "both data=journal and dioread_nolock");
4834 err = -EINVAL;
4835 goto restore_opts;
4836 }
4837 if (test_opt(sb, DAX)) {
4838 ext4_msg(sb, KERN_ERR, "can't mount with "
4839 "both data=journal and dax");
4840 err = -EINVAL;
4841 goto restore_opts;
4842 }
4843 }
4844
4845 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4846 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4847 "dax flag with busy inodes while remounting");
4848 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4849 }
4850
4851 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4852 ext4_abort(sb, "Abort forced by user");
4853
4854 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4855 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4856
4857 es = sbi->s_es;
4858
4859 if (sbi->s_journal) {
4860 ext4_init_journal_params(sb, sbi->s_journal);
4861 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4862 }
4863
4864 if (*flags & MS_LAZYTIME)
4865 sb->s_flags |= MS_LAZYTIME;
4866
4867 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4868 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4869 err = -EROFS;
4870 goto restore_opts;
4871 }
4872
4873 if (*flags & MS_RDONLY) {
4874 err = sync_filesystem(sb);
4875 if (err < 0)
4876 goto restore_opts;
4877 err = dquot_suspend(sb, -1);
4878 if (err < 0)
4879 goto restore_opts;
4880
4881 /*
4882 * First of all, the unconditional stuff we have to do
4883 * to disable replay of the journal when we next remount
4884 */
4885 sb->s_flags |= MS_RDONLY;
4886
4887 /*
4888 * OK, test if we are remounting a valid rw partition
4889 * readonly, and if so set the rdonly flag and then
4890 * mark the partition as valid again.
4891 */
4892 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4893 (sbi->s_mount_state & EXT4_VALID_FS))
4894 es->s_state = cpu_to_le16(sbi->s_mount_state);
4895
4896 if (sbi->s_journal)
4897 ext4_mark_recovery_complete(sb, es);
4898 } else {
4899 /* Make sure we can mount this feature set readwrite */
4900 if (ext4_has_feature_readonly(sb) ||
4901 !ext4_feature_set_ok(sb, 0)) {
4902 err = -EROFS;
4903 goto restore_opts;
4904 }
4905 /*
4906 * Make sure the group descriptor checksums
4907 * are sane. If they aren't, refuse to remount r/w.
4908 */
4909 for (g = 0; g < sbi->s_groups_count; g++) {
4910 struct ext4_group_desc *gdp =
4911 ext4_get_group_desc(sb, g, NULL);
4912
4913 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4914 ext4_msg(sb, KERN_ERR,
4915 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4916 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
4917 le16_to_cpu(gdp->bg_checksum));
4918 err = -EFSBADCRC;
4919 goto restore_opts;
4920 }
4921 }
4922
4923 /*
4924 * If we have an unprocessed orphan list hanging
4925 * around from a previously readonly bdev mount,
4926 * require a full umount/remount for now.
4927 */
4928 if (es->s_last_orphan) {
4929 ext4_msg(sb, KERN_WARNING, "Couldn't "
4930 "remount RDWR because of unprocessed "
4931 "orphan inode list. Please "
4932 "umount/remount instead");
4933 err = -EINVAL;
4934 goto restore_opts;
4935 }
4936
4937 /*
4938 * Mounting a RDONLY partition read-write, so reread
4939 * and store the current valid flag. (It may have
4940 * been changed by e2fsck since we originally mounted
4941 * the partition.)
4942 */
4943 if (sbi->s_journal)
4944 ext4_clear_journal_err(sb, es);
4945 sbi->s_mount_state = le16_to_cpu(es->s_state);
4946 if (!ext4_setup_super(sb, es, 0))
4947 sb->s_flags &= ~MS_RDONLY;
4948 if (ext4_has_feature_mmp(sb))
4949 if (ext4_multi_mount_protect(sb,
4950 le64_to_cpu(es->s_mmp_block))) {
4951 err = -EROFS;
4952 goto restore_opts;
4953 }
4954 enable_quota = 1;
4955 }
4956 }
4957
4958 /*
4959 * Reinitialize lazy itable initialization thread based on
4960 * current settings
4961 */
4962 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4963 ext4_unregister_li_request(sb);
4964 else {
4965 ext4_group_t first_not_zeroed;
4966 first_not_zeroed = ext4_has_uninit_itable(sb);
4967 ext4_register_li_request(sb, first_not_zeroed);
4968 }
4969
4970 ext4_setup_system_zone(sb);
4971 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4972 ext4_commit_super(sb, 1);
4973
4974 #ifdef CONFIG_QUOTA
4975 /* Release old quota file names */
4976 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4977 kfree(old_opts.s_qf_names[i]);
4978 if (enable_quota) {
4979 if (sb_any_quota_suspended(sb))
4980 dquot_resume(sb, -1);
4981 else if (ext4_has_feature_quota(sb)) {
4982 err = ext4_enable_quotas(sb);
4983 if (err)
4984 goto restore_opts;
4985 }
4986 }
4987 #endif
4988
4989 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
4990 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4991 kfree(orig_data);
4992 return 0;
4993
4994 restore_opts:
4995 sb->s_flags = old_sb_flags;
4996 sbi->s_mount_opt = old_opts.s_mount_opt;
4997 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4998 sbi->s_resuid = old_opts.s_resuid;
4999 sbi->s_resgid = old_opts.s_resgid;
5000 sbi->s_commit_interval = old_opts.s_commit_interval;
5001 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5002 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5003 #ifdef CONFIG_QUOTA
5004 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5005 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5006 kfree(sbi->s_qf_names[i]);
5007 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5008 }
5009 #endif
5010 kfree(orig_data);
5011 return err;
5012 }
5013
5014 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5015 {
5016 struct super_block *sb = dentry->d_sb;
5017 struct ext4_sb_info *sbi = EXT4_SB(sb);
5018 struct ext4_super_block *es = sbi->s_es;
5019 ext4_fsblk_t overhead = 0, resv_blocks;
5020 u64 fsid;
5021 s64 bfree;
5022 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5023
5024 if (!test_opt(sb, MINIX_DF))
5025 overhead = sbi->s_overhead;
5026
5027 buf->f_type = EXT4_SUPER_MAGIC;
5028 buf->f_bsize = sb->s_blocksize;
5029 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5030 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5031 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5032 /* prevent underflow in case that few free space is available */
5033 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5034 buf->f_bavail = buf->f_bfree -
5035 (ext4_r_blocks_count(es) + resv_blocks);
5036 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5037 buf->f_bavail = 0;
5038 buf->f_files = le32_to_cpu(es->s_inodes_count);
5039 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5040 buf->f_namelen = EXT4_NAME_LEN;
5041 fsid = le64_to_cpup((void *)es->s_uuid) ^
5042 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5043 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5044 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5045
5046 return 0;
5047 }
5048
5049 /* Helper function for writing quotas on sync - we need to start transaction
5050 * before quota file is locked for write. Otherwise the are possible deadlocks:
5051 * Process 1 Process 2
5052 * ext4_create() quota_sync()
5053 * jbd2_journal_start() write_dquot()
5054 * dquot_initialize() down(dqio_mutex)
5055 * down(dqio_mutex) jbd2_journal_start()
5056 *
5057 */
5058
5059 #ifdef CONFIG_QUOTA
5060
5061 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5062 {
5063 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5064 }
5065
5066 static int ext4_write_dquot(struct dquot *dquot)
5067 {
5068 int ret, err;
5069 handle_t *handle;
5070 struct inode *inode;
5071
5072 inode = dquot_to_inode(dquot);
5073 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5074 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5075 if (IS_ERR(handle))
5076 return PTR_ERR(handle);
5077 ret = dquot_commit(dquot);
5078 err = ext4_journal_stop(handle);
5079 if (!ret)
5080 ret = err;
5081 return ret;
5082 }
5083
5084 static int ext4_acquire_dquot(struct dquot *dquot)
5085 {
5086 int ret, err;
5087 handle_t *handle;
5088
5089 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5090 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5091 if (IS_ERR(handle))
5092 return PTR_ERR(handle);
5093 ret = dquot_acquire(dquot);
5094 err = ext4_journal_stop(handle);
5095 if (!ret)
5096 ret = err;
5097 return ret;
5098 }
5099
5100 static int ext4_release_dquot(struct dquot *dquot)
5101 {
5102 int ret, err;
5103 handle_t *handle;
5104
5105 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5106 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5107 if (IS_ERR(handle)) {
5108 /* Release dquot anyway to avoid endless cycle in dqput() */
5109 dquot_release(dquot);
5110 return PTR_ERR(handle);
5111 }
5112 ret = dquot_release(dquot);
5113 err = ext4_journal_stop(handle);
5114 if (!ret)
5115 ret = err;
5116 return ret;
5117 }
5118
5119 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5120 {
5121 struct super_block *sb = dquot->dq_sb;
5122 struct ext4_sb_info *sbi = EXT4_SB(sb);
5123
5124 /* Are we journaling quotas? */
5125 if (ext4_has_feature_quota(sb) ||
5126 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5127 dquot_mark_dquot_dirty(dquot);
5128 return ext4_write_dquot(dquot);
5129 } else {
5130 return dquot_mark_dquot_dirty(dquot);
5131 }
5132 }
5133
5134 static int ext4_write_info(struct super_block *sb, int type)
5135 {
5136 int ret, err;
5137 handle_t *handle;
5138
5139 /* Data block + inode block */
5140 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5141 if (IS_ERR(handle))
5142 return PTR_ERR(handle);
5143 ret = dquot_commit_info(sb, type);
5144 err = ext4_journal_stop(handle);
5145 if (!ret)
5146 ret = err;
5147 return ret;
5148 }
5149
5150 /*
5151 * Turn on quotas during mount time - we need to find
5152 * the quota file and such...
5153 */
5154 static int ext4_quota_on_mount(struct super_block *sb, int type)
5155 {
5156 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5157 EXT4_SB(sb)->s_jquota_fmt, type);
5158 }
5159
5160 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5161 {
5162 struct ext4_inode_info *ei = EXT4_I(inode);
5163
5164 /* The first argument of lockdep_set_subclass has to be
5165 * *exactly* the same as the argument to init_rwsem() --- in
5166 * this case, in init_once() --- or lockdep gets unhappy
5167 * because the name of the lock is set using the
5168 * stringification of the argument to init_rwsem().
5169 */
5170 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5171 lockdep_set_subclass(&ei->i_data_sem, subclass);
5172 }
5173
5174 /*
5175 * Standard function to be called on quota_on
5176 */
5177 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5178 struct path *path)
5179 {
5180 int err;
5181
5182 if (!test_opt(sb, QUOTA))
5183 return -EINVAL;
5184
5185 /* Quotafile not on the same filesystem? */
5186 if (path->dentry->d_sb != sb)
5187 return -EXDEV;
5188 /* Journaling quota? */
5189 if (EXT4_SB(sb)->s_qf_names[type]) {
5190 /* Quotafile not in fs root? */
5191 if (path->dentry->d_parent != sb->s_root)
5192 ext4_msg(sb, KERN_WARNING,
5193 "Quota file not on filesystem root. "
5194 "Journaled quota will not work");
5195 }
5196
5197 /*
5198 * When we journal data on quota file, we have to flush journal to see
5199 * all updates to the file when we bypass pagecache...
5200 */
5201 if (EXT4_SB(sb)->s_journal &&
5202 ext4_should_journal_data(d_inode(path->dentry))) {
5203 /*
5204 * We don't need to lock updates but journal_flush() could
5205 * otherwise be livelocked...
5206 */
5207 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5208 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5209 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5210 if (err)
5211 return err;
5212 }
5213 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5214 err = dquot_quota_on(sb, type, format_id, path);
5215 if (err)
5216 lockdep_set_quota_inode(path->dentry->d_inode,
5217 I_DATA_SEM_NORMAL);
5218 return err;
5219 }
5220
5221 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5222 unsigned int flags)
5223 {
5224 int err;
5225 struct inode *qf_inode;
5226 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5227 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5228 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5229 };
5230
5231 BUG_ON(!ext4_has_feature_quota(sb));
5232
5233 if (!qf_inums[type])
5234 return -EPERM;
5235
5236 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
5237 if (IS_ERR(qf_inode)) {
5238 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5239 return PTR_ERR(qf_inode);
5240 }
5241
5242 /* Don't account quota for quota files to avoid recursion */
5243 qf_inode->i_flags |= S_NOQUOTA;
5244 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5245 err = dquot_enable(qf_inode, type, format_id, flags);
5246 if (err)
5247 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5248 iput(qf_inode);
5249
5250 return err;
5251 }
5252
5253 /* Enable usage tracking for all quota types. */
5254 static int ext4_enable_quotas(struct super_block *sb)
5255 {
5256 int type, err = 0;
5257 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5258 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5259 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5260 };
5261
5262 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5263 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5264 if (qf_inums[type]) {
5265 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5266 DQUOT_USAGE_ENABLED);
5267 if (err) {
5268 for (type--; type >= 0; type--)
5269 dquot_quota_off(sb, type);
5270
5271 ext4_warning(sb,
5272 "Failed to enable quota tracking "
5273 "(type=%d, err=%d). Please run "
5274 "e2fsck to fix.", type, err);
5275 return err;
5276 }
5277 }
5278 }
5279 return 0;
5280 }
5281
5282 static int ext4_quota_off(struct super_block *sb, int type)
5283 {
5284 struct inode *inode = sb_dqopt(sb)->files[type];
5285 handle_t *handle;
5286
5287 /* Force all delayed allocation blocks to be allocated.
5288 * Caller already holds s_umount sem */
5289 if (test_opt(sb, DELALLOC))
5290 sync_filesystem(sb);
5291
5292 if (!inode)
5293 goto out;
5294
5295 /* Update modification times of quota files when userspace can
5296 * start looking at them */
5297 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5298 if (IS_ERR(handle))
5299 goto out;
5300 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5301 ext4_mark_inode_dirty(handle, inode);
5302 ext4_journal_stop(handle);
5303
5304 out:
5305 return dquot_quota_off(sb, type);
5306 }
5307
5308 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5309 * acquiring the locks... As quota files are never truncated and quota code
5310 * itself serializes the operations (and no one else should touch the files)
5311 * we don't have to be afraid of races */
5312 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5313 size_t len, loff_t off)
5314 {
5315 struct inode *inode = sb_dqopt(sb)->files[type];
5316 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5317 int offset = off & (sb->s_blocksize - 1);
5318 int tocopy;
5319 size_t toread;
5320 struct buffer_head *bh;
5321 loff_t i_size = i_size_read(inode);
5322
5323 if (off > i_size)
5324 return 0;
5325 if (off+len > i_size)
5326 len = i_size-off;
5327 toread = len;
5328 while (toread > 0) {
5329 tocopy = sb->s_blocksize - offset < toread ?
5330 sb->s_blocksize - offset : toread;
5331 bh = ext4_bread(NULL, inode, blk, 0);
5332 if (IS_ERR(bh))
5333 return PTR_ERR(bh);
5334 if (!bh) /* A hole? */
5335 memset(data, 0, tocopy);
5336 else
5337 memcpy(data, bh->b_data+offset, tocopy);
5338 brelse(bh);
5339 offset = 0;
5340 toread -= tocopy;
5341 data += tocopy;
5342 blk++;
5343 }
5344 return len;
5345 }
5346
5347 /* Write to quotafile (we know the transaction is already started and has
5348 * enough credits) */
5349 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5350 const char *data, size_t len, loff_t off)
5351 {
5352 struct inode *inode = sb_dqopt(sb)->files[type];
5353 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5354 int err, offset = off & (sb->s_blocksize - 1);
5355 int retries = 0;
5356 struct buffer_head *bh;
5357 handle_t *handle = journal_current_handle();
5358
5359 if (EXT4_SB(sb)->s_journal && !handle) {
5360 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5361 " cancelled because transaction is not started",
5362 (unsigned long long)off, (unsigned long long)len);
5363 return -EIO;
5364 }
5365 /*
5366 * Since we account only one data block in transaction credits,
5367 * then it is impossible to cross a block boundary.
5368 */
5369 if (sb->s_blocksize - offset < len) {
5370 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5371 " cancelled because not block aligned",
5372 (unsigned long long)off, (unsigned long long)len);
5373 return -EIO;
5374 }
5375
5376 do {
5377 bh = ext4_bread(handle, inode, blk,
5378 EXT4_GET_BLOCKS_CREATE |
5379 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5380 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5381 ext4_should_retry_alloc(inode->i_sb, &retries));
5382 if (IS_ERR(bh))
5383 return PTR_ERR(bh);
5384 if (!bh)
5385 goto out;
5386 BUFFER_TRACE(bh, "get write access");
5387 err = ext4_journal_get_write_access(handle, bh);
5388 if (err) {
5389 brelse(bh);
5390 return err;
5391 }
5392 lock_buffer(bh);
5393 memcpy(bh->b_data+offset, data, len);
5394 flush_dcache_page(bh->b_page);
5395 unlock_buffer(bh);
5396 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5397 brelse(bh);
5398 out:
5399 if (inode->i_size < off + len) {
5400 i_size_write(inode, off + len);
5401 EXT4_I(inode)->i_disksize = inode->i_size;
5402 ext4_mark_inode_dirty(handle, inode);
5403 }
5404 return len;
5405 }
5406
5407 #endif
5408
5409 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5410 const char *dev_name, void *data)
5411 {
5412 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5413 }
5414
5415 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5416 static inline void register_as_ext2(void)
5417 {
5418 int err = register_filesystem(&ext2_fs_type);
5419 if (err)
5420 printk(KERN_WARNING
5421 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5422 }
5423
5424 static inline void unregister_as_ext2(void)
5425 {
5426 unregister_filesystem(&ext2_fs_type);
5427 }
5428
5429 static inline int ext2_feature_set_ok(struct super_block *sb)
5430 {
5431 if (ext4_has_unknown_ext2_incompat_features(sb))
5432 return 0;
5433 if (sb->s_flags & MS_RDONLY)
5434 return 1;
5435 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5436 return 0;
5437 return 1;
5438 }
5439 #else
5440 static inline void register_as_ext2(void) { }
5441 static inline void unregister_as_ext2(void) { }
5442 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5443 #endif
5444
5445 static inline void register_as_ext3(void)
5446 {
5447 int err = register_filesystem(&ext3_fs_type);
5448 if (err)
5449 printk(KERN_WARNING
5450 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5451 }
5452
5453 static inline void unregister_as_ext3(void)
5454 {
5455 unregister_filesystem(&ext3_fs_type);
5456 }
5457
5458 static inline int ext3_feature_set_ok(struct super_block *sb)
5459 {
5460 if (ext4_has_unknown_ext3_incompat_features(sb))
5461 return 0;
5462 if (!ext4_has_feature_journal(sb))
5463 return 0;
5464 if (sb->s_flags & MS_RDONLY)
5465 return 1;
5466 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5467 return 0;
5468 return 1;
5469 }
5470
5471 static struct file_system_type ext4_fs_type = {
5472 .owner = THIS_MODULE,
5473 .name = "ext4",
5474 .mount = ext4_mount,
5475 .kill_sb = kill_block_super,
5476 .fs_flags = FS_REQUIRES_DEV,
5477 };
5478 MODULE_ALIAS_FS("ext4");
5479
5480 /* Shared across all ext4 file systems */
5481 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5482 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5483
5484 static int __init ext4_init_fs(void)
5485 {
5486 int i, err;
5487
5488 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5489 ext4_li_info = NULL;
5490 mutex_init(&ext4_li_mtx);
5491
5492 /* Build-time check for flags consistency */
5493 ext4_check_flag_values();
5494
5495 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5496 mutex_init(&ext4__aio_mutex[i]);
5497 init_waitqueue_head(&ext4__ioend_wq[i]);
5498 }
5499
5500 err = ext4_init_es();
5501 if (err)
5502 return err;
5503
5504 err = ext4_init_pageio();
5505 if (err)
5506 goto out5;
5507
5508 err = ext4_init_system_zone();
5509 if (err)
5510 goto out4;
5511
5512 err = ext4_init_sysfs();
5513 if (err)
5514 goto out3;
5515
5516 err = ext4_init_mballoc();
5517 if (err)
5518 goto out2;
5519 else
5520 ext4_mballoc_ready = 1;
5521 err = init_inodecache();
5522 if (err)
5523 goto out1;
5524 register_as_ext3();
5525 register_as_ext2();
5526 err = register_filesystem(&ext4_fs_type);
5527 if (err)
5528 goto out;
5529
5530 return 0;
5531 out:
5532 unregister_as_ext2();
5533 unregister_as_ext3();
5534 destroy_inodecache();
5535 out1:
5536 ext4_mballoc_ready = 0;
5537 ext4_exit_mballoc();
5538 out2:
5539 ext4_exit_sysfs();
5540 out3:
5541 ext4_exit_system_zone();
5542 out4:
5543 ext4_exit_pageio();
5544 out5:
5545 ext4_exit_es();
5546
5547 return err;
5548 }
5549
5550 static void __exit ext4_exit_fs(void)
5551 {
5552 ext4_exit_crypto();
5553 ext4_destroy_lazyinit_thread();
5554 unregister_as_ext2();
5555 unregister_as_ext3();
5556 unregister_filesystem(&ext4_fs_type);
5557 destroy_inodecache();
5558 ext4_exit_mballoc();
5559 ext4_exit_sysfs();
5560 ext4_exit_system_zone();
5561 ext4_exit_pageio();
5562 ext4_exit_es();
5563 }
5564
5565 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5566 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5567 MODULE_LICENSE("GPL");
5568 module_init(ext4_init_fs)
5569 module_exit(ext4_exit_fs)
Linux with added FPC-III support