?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
gro: Allow tunnel stacking in the case of FOU/GUE
[linux/fpc-iii.git] / fs / ext4 / super.c
blob97aa8be4017527e41c3ea02321cd6d35698e2b88
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/parser.h>
28 #include <linux/buffer_head.h>
29 #include <linux/exportfs.h>
30 #include <linux/vfs.h>
31 #include <linux/random.h>
32 #include <linux/mount.h>
33 #include <linux/namei.h>
34 #include <linux/quotaops.h>
35 #include <linux/seq_file.h>
36 #include <linux/proc_fs.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 proc_dir_entry *ext4_proc_root;
57 static struct kset *ext4_kset;
58 static struct ext4_lazy_init *ext4_li_info;
59 static struct mutex ext4_li_mtx;
60 static struct ext4_features *ext4_feat;
61 static int ext4_mballoc_ready;
62
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static int ext4_remount(struct super_block *sb, int *flags, char *data);
73 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
74 static int ext4_unfreeze(struct super_block *sb);
75 static int ext4_freeze(struct super_block *sb);
76 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
78 static inline int ext2_feature_set_ok(struct super_block *sb);
79 static inline int ext3_feature_set_ok(struct super_block *sb);
80 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
81 static void ext4_destroy_lazyinit_thread(void);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
84 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
85
86 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
87 static struct file_system_type ext2_fs_type = {
88 .owner = THIS_MODULE,
89 .name = "ext2",
90 .mount = ext4_mount,
91 .kill_sb = kill_block_super,
92 .fs_flags = FS_REQUIRES_DEV,
93 };
94 MODULE_ALIAS_FS("ext2");
95 MODULE_ALIAS("ext2");
96 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97 #else
98 #define IS_EXT2_SB(sb) (0)
99 #endif
100
101
102 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103 static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
106 .mount = ext4_mount,
107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109 };
110 MODULE_ALIAS_FS("ext3");
111 MODULE_ALIAS("ext3");
112 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
113 #else
114 #define IS_EXT3_SB(sb) (0)
115 #endif
116
117 static int ext4_verify_csum_type(struct super_block *sb,
118 struct ext4_super_block *es)
119 {
120 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
121 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
122 return 1;
123
124 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
125 }
126
127 static __le32 ext4_superblock_csum(struct super_block *sb,
128 struct ext4_super_block *es)
129 {
130 struct ext4_sb_info *sbi = EXT4_SB(sb);
131 int offset = offsetof(struct ext4_super_block, s_checksum);
132 __u32 csum;
133
134 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
135
136 return cpu_to_le32(csum);
137 }
138
139 static int ext4_superblock_csum_verify(struct super_block *sb,
140 struct ext4_super_block *es)
141 {
142 if (!ext4_has_metadata_csum(sb))
143 return 1;
144
145 return es->s_checksum == ext4_superblock_csum(sb, es);
146 }
147
148 void ext4_superblock_csum_set(struct super_block *sb)
149 {
150 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
151
152 if (!ext4_has_metadata_csum(sb))
153 return;
154
155 es->s_checksum = ext4_superblock_csum(sb, es);
156 }
157
158 void *ext4_kvmalloc(size_t size, gfp_t flags)
159 {
160 void *ret;
161
162 ret = kmalloc(size, flags | __GFP_NOWARN);
163 if (!ret)
164 ret = __vmalloc(size, flags, PAGE_KERNEL);
165 return ret;
166 }
167
168 void *ext4_kvzalloc(size_t size, gfp_t flags)
169 {
170 void *ret;
171
172 ret = kzalloc(size, flags | __GFP_NOWARN);
173 if (!ret)
174 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
175 return ret;
176 }
177
178 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
179 struct ext4_group_desc *bg)
180 {
181 return le32_to_cpu(bg->bg_block_bitmap_lo) |
182 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
183 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
184 }
185
186 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
188 {
189 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
192 }
193
194 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
195 struct ext4_group_desc *bg)
196 {
197 return le32_to_cpu(bg->bg_inode_table_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
200 }
201
202 __u32 ext4_free_group_clusters(struct super_block *sb,
203 struct ext4_group_desc *bg)
204 {
205 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
208 }
209
210 __u32 ext4_free_inodes_count(struct super_block *sb,
211 struct ext4_group_desc *bg)
212 {
213 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
216 }
217
218 __u32 ext4_used_dirs_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220 {
221 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
224 }
225
226 __u32 ext4_itable_unused_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
228 {
229 return le16_to_cpu(bg->bg_itable_unused_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
231 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
232 }
233
234 void ext4_block_bitmap_set(struct super_block *sb,
235 struct ext4_group_desc *bg, ext4_fsblk_t blk)
236 {
237 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
238 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
239 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
240 }
241
242 void ext4_inode_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
244 {
245 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
248 }
249
250 void ext4_inode_table_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
252 {
253 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
256 }
257
258 void ext4_free_group_clusters_set(struct super_block *sb,
259 struct ext4_group_desc *bg, __u32 count)
260 {
261 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
264 }
265
266 void ext4_free_inodes_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
268 {
269 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
272 }
273
274 void ext4_used_dirs_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276 {
277 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
280 }
281
282 void ext4_itable_unused_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284 {
285 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
288 }
289
290
291 static void __save_error_info(struct super_block *sb, const char *func,
292 unsigned int line)
293 {
294 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
295
296 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
297 if (bdev_read_only(sb->s_bdev))
298 return;
299 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
300 es->s_last_error_time = cpu_to_le32(get_seconds());
301 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
302 es->s_last_error_line = cpu_to_le32(line);
303 if (!es->s_first_error_time) {
304 es->s_first_error_time = es->s_last_error_time;
305 strncpy(es->s_first_error_func, func,
306 sizeof(es->s_first_error_func));
307 es->s_first_error_line = cpu_to_le32(line);
308 es->s_first_error_ino = es->s_last_error_ino;
309 es->s_first_error_block = es->s_last_error_block;
310 }
311 /*
312 * Start the daily error reporting function if it hasn't been
313 * started already
314 */
315 if (!es->s_error_count)
316 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
317 le32_add_cpu(&es->s_error_count, 1);
318 }
319
320 static void save_error_info(struct super_block *sb, const char *func,
321 unsigned int line)
322 {
323 __save_error_info(sb, func, line);
324 ext4_commit_super(sb, 1);
325 }
326
327 /*
328 * The del_gendisk() function uninitializes the disk-specific data
329 * structures, including the bdi structure, without telling anyone
330 * else. Once this happens, any attempt to call mark_buffer_dirty()
331 * (for example, by ext4_commit_super), will cause a kernel OOPS.
332 * This is a kludge to prevent these oops until we can put in a proper
333 * hook in del_gendisk() to inform the VFS and file system layers.
334 */
335 static int block_device_ejected(struct super_block *sb)
336 {
337 struct inode *bd_inode = sb->s_bdev->bd_inode;
338 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
339
340 return bdi->dev == NULL;
341 }
342
343 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
344 {
345 struct super_block *sb = journal->j_private;
346 struct ext4_sb_info *sbi = EXT4_SB(sb);
347 int error = is_journal_aborted(journal);
348 struct ext4_journal_cb_entry *jce;
349
350 BUG_ON(txn->t_state == T_FINISHED);
351 spin_lock(&sbi->s_md_lock);
352 while (!list_empty(&txn->t_private_list)) {
353 jce = list_entry(txn->t_private_list.next,
354 struct ext4_journal_cb_entry, jce_list);
355 list_del_init(&jce->jce_list);
356 spin_unlock(&sbi->s_md_lock);
357 jce->jce_func(sb, jce, error);
358 spin_lock(&sbi->s_md_lock);
359 }
360 spin_unlock(&sbi->s_md_lock);
361 }
362
363 /* Deal with the reporting of failure conditions on a filesystem such as
364 * inconsistencies detected or read IO failures.
365 *
366 * On ext2, we can store the error state of the filesystem in the
367 * superblock. That is not possible on ext4, because we may have other
368 * write ordering constraints on the superblock which prevent us from
369 * writing it out straight away; and given that the journal is about to
370 * be aborted, we can't rely on the current, or future, transactions to
371 * write out the superblock safely.
372 *
373 * We'll just use the jbd2_journal_abort() error code to record an error in
374 * the journal instead. On recovery, the journal will complain about
375 * that error until we've noted it down and cleared it.
376 */
377
378 static void ext4_handle_error(struct super_block *sb)
379 {
380 if (sb->s_flags & MS_RDONLY)
381 return;
382
383 if (!test_opt(sb, ERRORS_CONT)) {
384 journal_t *journal = EXT4_SB(sb)->s_journal;
385
386 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
387 if (journal)
388 jbd2_journal_abort(journal, -EIO);
389 }
390 if (test_opt(sb, ERRORS_RO)) {
391 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
392 /*
393 * Make sure updated value of ->s_mount_flags will be visible
394 * before ->s_flags update
395 */
396 smp_wmb();
397 sb->s_flags |= MS_RDONLY;
398 }
399 if (test_opt(sb, ERRORS_PANIC)) {
400 if (EXT4_SB(sb)->s_journal &&
401 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
402 return;
403 panic("EXT4-fs (device %s): panic forced after error\n",
404 sb->s_id);
405 }
406 }
407
408 #define ext4_error_ratelimit(sb) \
409 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
410 "EXT4-fs error")
411
412 void __ext4_error(struct super_block *sb, const char *function,
413 unsigned int line, const char *fmt, ...)
414 {
415 struct va_format vaf;
416 va_list args;
417
418 if (ext4_error_ratelimit(sb)) {
419 va_start(args, fmt);
420 vaf.fmt = fmt;
421 vaf.va = &args;
422 printk(KERN_CRIT
423 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
424 sb->s_id, function, line, current->comm, &vaf);
425 va_end(args);
426 }
427 save_error_info(sb, function, line);
428 ext4_handle_error(sb);
429 }
430
431 void __ext4_error_inode(struct inode *inode, const char *function,
432 unsigned int line, ext4_fsblk_t block,
433 const char *fmt, ...)
434 {
435 va_list args;
436 struct va_format vaf;
437 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
438
439 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
440 es->s_last_error_block = cpu_to_le64(block);
441 if (ext4_error_ratelimit(inode->i_sb)) {
442 va_start(args, fmt);
443 vaf.fmt = fmt;
444 vaf.va = &args;
445 if (block)
446 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
447 "inode #%lu: block %llu: comm %s: %pV\n",
448 inode->i_sb->s_id, function, line, inode->i_ino,
449 block, current->comm, &vaf);
450 else
451 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
452 "inode #%lu: comm %s: %pV\n",
453 inode->i_sb->s_id, function, line, inode->i_ino,
454 current->comm, &vaf);
455 va_end(args);
456 }
457 save_error_info(inode->i_sb, function, line);
458 ext4_handle_error(inode->i_sb);
459 }
460
461 void __ext4_error_file(struct file *file, const char *function,
462 unsigned int line, ext4_fsblk_t block,
463 const char *fmt, ...)
464 {
465 va_list args;
466 struct va_format vaf;
467 struct ext4_super_block *es;
468 struct inode *inode = file_inode(file);
469 char pathname[80], *path;
470
471 es = EXT4_SB(inode->i_sb)->s_es;
472 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
473 if (ext4_error_ratelimit(inode->i_sb)) {
474 path = d_path(&(file->f_path), pathname, sizeof(pathname));
475 if (IS_ERR(path))
476 path = "(unknown)";
477 va_start(args, fmt);
478 vaf.fmt = fmt;
479 vaf.va = &args;
480 if (block)
481 printk(KERN_CRIT
482 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
483 "block %llu: comm %s: path %s: %pV\n",
484 inode->i_sb->s_id, function, line, inode->i_ino,
485 block, current->comm, path, &vaf);
486 else
487 printk(KERN_CRIT
488 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
489 "comm %s: path %s: %pV\n",
490 inode->i_sb->s_id, function, line, inode->i_ino,
491 current->comm, path, &vaf);
492 va_end(args);
493 }
494 save_error_info(inode->i_sb, function, line);
495 ext4_handle_error(inode->i_sb);
496 }
497
498 const char *ext4_decode_error(struct super_block *sb, int errno,
499 char nbuf[16])
500 {
501 char *errstr = NULL;
502
503 switch (errno) {
504 case -EIO:
505 errstr = "IO failure";
506 break;
507 case -ENOMEM:
508 errstr = "Out of memory";
509 break;
510 case -EROFS:
511 if (!sb || (EXT4_SB(sb)->s_journal &&
512 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
513 errstr = "Journal has aborted";
514 else
515 errstr = "Readonly filesystem";
516 break;
517 default:
518 /* If the caller passed in an extra buffer for unknown
519 * errors, textualise them now. Else we just return
520 * NULL. */
521 if (nbuf) {
522 /* Check for truncated error codes... */
523 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
524 errstr = nbuf;
525 }
526 break;
527 }
528
529 return errstr;
530 }
531
532 /* __ext4_std_error decodes expected errors from journaling functions
533 * automatically and invokes the appropriate error response. */
534
535 void __ext4_std_error(struct super_block *sb, const char *function,
536 unsigned int line, int errno)
537 {
538 char nbuf[16];
539 const char *errstr;
540
541 /* Special case: if the error is EROFS, and we're not already
542 * inside a transaction, then there's really no point in logging
543 * an error. */
544 if (errno == -EROFS && journal_current_handle() == NULL &&
545 (sb->s_flags & MS_RDONLY))
546 return;
547
548 if (ext4_error_ratelimit(sb)) {
549 errstr = ext4_decode_error(sb, errno, nbuf);
550 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
551 sb->s_id, function, line, errstr);
552 }
553
554 save_error_info(sb, function, line);
555 ext4_handle_error(sb);
556 }
557
558 /*
559 * ext4_abort is a much stronger failure handler than ext4_error. The
560 * abort function may be used to deal with unrecoverable failures such
561 * as journal IO errors or ENOMEM at a critical moment in log management.
562 *
563 * We unconditionally force the filesystem into an ABORT|READONLY state,
564 * unless the error response on the fs has been set to panic in which
565 * case we take the easy way out and panic immediately.
566 */
567
568 void __ext4_abort(struct super_block *sb, const char *function,
569 unsigned int line, const char *fmt, ...)
570 {
571 va_list args;
572
573 save_error_info(sb, function, line);
574 va_start(args, fmt);
575 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
576 function, line);
577 vprintk(fmt, args);
578 printk("\n");
579 va_end(args);
580
581 if ((sb->s_flags & MS_RDONLY) == 0) {
582 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
583 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
584 /*
585 * Make sure updated value of ->s_mount_flags will be visible
586 * before ->s_flags update
587 */
588 smp_wmb();
589 sb->s_flags |= MS_RDONLY;
590 if (EXT4_SB(sb)->s_journal)
591 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
592 save_error_info(sb, function, line);
593 }
594 if (test_opt(sb, ERRORS_PANIC)) {
595 if (EXT4_SB(sb)->s_journal &&
596 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
597 return;
598 panic("EXT4-fs panic from previous error\n");
599 }
600 }
601
602 void __ext4_msg(struct super_block *sb,
603 const char *prefix, const char *fmt, ...)
604 {
605 struct va_format vaf;
606 va_list args;
607
608 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
609 return;
610
611 va_start(args, fmt);
612 vaf.fmt = fmt;
613 vaf.va = &args;
614 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
615 va_end(args);
616 }
617
618 void __ext4_warning(struct super_block *sb, const char *function,
619 unsigned int line, const char *fmt, ...)
620 {
621 struct va_format vaf;
622 va_list args;
623
624 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
625 "EXT4-fs warning"))
626 return;
627
628 va_start(args, fmt);
629 vaf.fmt = fmt;
630 vaf.va = &args;
631 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
632 sb->s_id, function, line, &vaf);
633 va_end(args);
634 }
635
636 void __ext4_grp_locked_error(const char *function, unsigned int line,
637 struct super_block *sb, ext4_group_t grp,
638 unsigned long ino, ext4_fsblk_t block,
639 const char *fmt, ...)
640 __releases(bitlock)
641 __acquires(bitlock)
642 {
643 struct va_format vaf;
644 va_list args;
645 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
646
647 es->s_last_error_ino = cpu_to_le32(ino);
648 es->s_last_error_block = cpu_to_le64(block);
649 __save_error_info(sb, function, line);
650
651 if (ext4_error_ratelimit(sb)) {
652 va_start(args, fmt);
653 vaf.fmt = fmt;
654 vaf.va = &args;
655 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
656 sb->s_id, function, line, grp);
657 if (ino)
658 printk(KERN_CONT "inode %lu: ", ino);
659 if (block)
660 printk(KERN_CONT "block %llu:",
661 (unsigned long long) block);
662 printk(KERN_CONT "%pV\n", &vaf);
663 va_end(args);
664 }
665
666 if (test_opt(sb, ERRORS_CONT)) {
667 ext4_commit_super(sb, 0);
668 return;
669 }
670
671 ext4_unlock_group(sb, grp);
672 ext4_handle_error(sb);
673 /*
674 * We only get here in the ERRORS_RO case; relocking the group
675 * may be dangerous, but nothing bad will happen since the
676 * filesystem will have already been marked read/only and the
677 * journal has been aborted. We return 1 as a hint to callers
678 * who might what to use the return value from
679 * ext4_grp_locked_error() to distinguish between the
680 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
681 * aggressively from the ext4 function in question, with a
682 * more appropriate error code.
683 */
684 ext4_lock_group(sb, grp);
685 return;
686 }
687
688 void ext4_update_dynamic_rev(struct super_block *sb)
689 {
690 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
691
692 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
693 return;
694
695 ext4_warning(sb,
696 "updating to rev %d because of new feature flag, "
697 "running e2fsck is recommended",
698 EXT4_DYNAMIC_REV);
699
700 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
701 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
702 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
703 /* leave es->s_feature_*compat flags alone */
704 /* es->s_uuid will be set by e2fsck if empty */
705
706 /*
707 * The rest of the superblock fields should be zero, and if not it
708 * means they are likely already in use, so leave them alone. We
709 * can leave it up to e2fsck to clean up any inconsistencies there.
710 */
711 }
712
713 /*
714 * Open the external journal device
715 */
716 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
717 {
718 struct block_device *bdev;
719 char b[BDEVNAME_SIZE];
720
721 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
722 if (IS_ERR(bdev))
723 goto fail;
724 return bdev;
725
726 fail:
727 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
728 __bdevname(dev, b), PTR_ERR(bdev));
729 return NULL;
730 }
731
732 /*
733 * Release the journal device
734 */
735 static void ext4_blkdev_put(struct block_device *bdev)
736 {
737 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
738 }
739
740 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
741 {
742 struct block_device *bdev;
743 bdev = sbi->journal_bdev;
744 if (bdev) {
745 ext4_blkdev_put(bdev);
746 sbi->journal_bdev = NULL;
747 }
748 }
749
750 static inline struct inode *orphan_list_entry(struct list_head *l)
751 {
752 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
753 }
754
755 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
756 {
757 struct list_head *l;
758
759 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
760 le32_to_cpu(sbi->s_es->s_last_orphan));
761
762 printk(KERN_ERR "sb_info orphan list:\n");
763 list_for_each(l, &sbi->s_orphan) {
764 struct inode *inode = orphan_list_entry(l);
765 printk(KERN_ERR " "
766 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
767 inode->i_sb->s_id, inode->i_ino, inode,
768 inode->i_mode, inode->i_nlink,
769 NEXT_ORPHAN(inode));
770 }
771 }
772
773 static void ext4_put_super(struct super_block *sb)
774 {
775 struct ext4_sb_info *sbi = EXT4_SB(sb);
776 struct ext4_super_block *es = sbi->s_es;
777 int i, err;
778
779 ext4_unregister_li_request(sb);
780 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
781
782 flush_workqueue(sbi->rsv_conversion_wq);
783 destroy_workqueue(sbi->rsv_conversion_wq);
784
785 if (sbi->s_journal) {
786 err = jbd2_journal_destroy(sbi->s_journal);
787 sbi->s_journal = NULL;
788 if (err < 0)
789 ext4_abort(sb, "Couldn't clean up the journal");
790 }
791
792 ext4_es_unregister_shrinker(sbi);
793 del_timer_sync(&sbi->s_err_report);
794 ext4_release_system_zone(sb);
795 ext4_mb_release(sb);
796 ext4_ext_release(sb);
797 ext4_xattr_put_super(sb);
798
799 if (!(sb->s_flags & MS_RDONLY)) {
800 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
801 es->s_state = cpu_to_le16(sbi->s_mount_state);
802 }
803 if (!(sb->s_flags & MS_RDONLY))
804 ext4_commit_super(sb, 1);
805
806 if (sbi->s_proc) {
807 remove_proc_entry("options", sbi->s_proc);
808 remove_proc_entry(sb->s_id, ext4_proc_root);
809 }
810 kobject_del(&sbi->s_kobj);
811
812 for (i = 0; i < sbi->s_gdb_count; i++)
813 brelse(sbi->s_group_desc[i]);
814 kvfree(sbi->s_group_desc);
815 kvfree(sbi->s_flex_groups);
816 percpu_counter_destroy(&sbi->s_freeclusters_counter);
817 percpu_counter_destroy(&sbi->s_freeinodes_counter);
818 percpu_counter_destroy(&sbi->s_dirs_counter);
819 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
820 brelse(sbi->s_sbh);
821 #ifdef CONFIG_QUOTA
822 for (i = 0; i < EXT4_MAXQUOTAS; i++)
823 kfree(sbi->s_qf_names[i]);
824 #endif
825
826 /* Debugging code just in case the in-memory inode orphan list
827 * isn't empty. The on-disk one can be non-empty if we've
828 * detected an error and taken the fs readonly, but the
829 * in-memory list had better be clean by this point. */
830 if (!list_empty(&sbi->s_orphan))
831 dump_orphan_list(sb, sbi);
832 J_ASSERT(list_empty(&sbi->s_orphan));
833
834 sync_blockdev(sb->s_bdev);
835 invalidate_bdev(sb->s_bdev);
836 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
837 /*
838 * Invalidate the journal device's buffers. We don't want them
839 * floating about in memory - the physical journal device may
840 * hotswapped, and it breaks the `ro-after' testing code.
841 */
842 sync_blockdev(sbi->journal_bdev);
843 invalidate_bdev(sbi->journal_bdev);
844 ext4_blkdev_remove(sbi);
845 }
846 if (sbi->s_mb_cache) {
847 ext4_xattr_destroy_cache(sbi->s_mb_cache);
848 sbi->s_mb_cache = NULL;
849 }
850 if (sbi->s_mmp_tsk)
851 kthread_stop(sbi->s_mmp_tsk);
852 sb->s_fs_info = NULL;
853 /*
854 * Now that we are completely done shutting down the
855 * superblock, we need to actually destroy the kobject.
856 */
857 kobject_put(&sbi->s_kobj);
858 wait_for_completion(&sbi->s_kobj_unregister);
859 if (sbi->s_chksum_driver)
860 crypto_free_shash(sbi->s_chksum_driver);
861 kfree(sbi->s_blockgroup_lock);
862 kfree(sbi);
863 }
864
865 static struct kmem_cache *ext4_inode_cachep;
866
867 /*
868 * Called inside transaction, so use GFP_NOFS
869 */
870 static struct inode *ext4_alloc_inode(struct super_block *sb)
871 {
872 struct ext4_inode_info *ei;
873
874 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
875 if (!ei)
876 return NULL;
877
878 ei->vfs_inode.i_version = 1;
879 spin_lock_init(&ei->i_raw_lock);
880 INIT_LIST_HEAD(&ei->i_prealloc_list);
881 spin_lock_init(&ei->i_prealloc_lock);
882 ext4_es_init_tree(&ei->i_es_tree);
883 rwlock_init(&ei->i_es_lock);
884 INIT_LIST_HEAD(&ei->i_es_list);
885 ei->i_es_all_nr = 0;
886 ei->i_es_shk_nr = 0;
887 ei->i_es_shrink_lblk = 0;
888 ei->i_reserved_data_blocks = 0;
889 ei->i_reserved_meta_blocks = 0;
890 ei->i_allocated_meta_blocks = 0;
891 ei->i_da_metadata_calc_len = 0;
892 ei->i_da_metadata_calc_last_lblock = 0;
893 spin_lock_init(&(ei->i_block_reservation_lock));
894 #ifdef CONFIG_QUOTA
895 ei->i_reserved_quota = 0;
896 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
897 #endif
898 ei->jinode = NULL;
899 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
900 spin_lock_init(&ei->i_completed_io_lock);
901 ei->i_sync_tid = 0;
902 ei->i_datasync_tid = 0;
903 atomic_set(&ei->i_ioend_count, 0);
904 atomic_set(&ei->i_unwritten, 0);
905 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
906 #ifdef CONFIG_EXT4_FS_ENCRYPTION
907 ei->i_encryption_key.mode = EXT4_ENCRYPTION_MODE_INVALID;
908 #endif
909
910 return &ei->vfs_inode;
911 }
912
913 static int ext4_drop_inode(struct inode *inode)
914 {
915 int drop = generic_drop_inode(inode);
916
917 trace_ext4_drop_inode(inode, drop);
918 return drop;
919 }
920
921 static void ext4_i_callback(struct rcu_head *head)
922 {
923 struct inode *inode = container_of(head, struct inode, i_rcu);
924 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
925 }
926
927 static void ext4_destroy_inode(struct inode *inode)
928 {
929 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
930 ext4_msg(inode->i_sb, KERN_ERR,
931 "Inode %lu (%p): orphan list check failed!",
932 inode->i_ino, EXT4_I(inode));
933 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
934 EXT4_I(inode), sizeof(struct ext4_inode_info),
935 true);
936 dump_stack();
937 }
938 call_rcu(&inode->i_rcu, ext4_i_callback);
939 }
940
941 static void init_once(void *foo)
942 {
943 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
944
945 INIT_LIST_HEAD(&ei->i_orphan);
946 init_rwsem(&ei->xattr_sem);
947 init_rwsem(&ei->i_data_sem);
948 init_rwsem(&ei->i_mmap_sem);
949 inode_init_once(&ei->vfs_inode);
950 }
951
952 static int __init init_inodecache(void)
953 {
954 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
955 sizeof(struct ext4_inode_info),
956 0, (SLAB_RECLAIM_ACCOUNT|
957 SLAB_MEM_SPREAD),
958 init_once);
959 if (ext4_inode_cachep == NULL)
960 return -ENOMEM;
961 return 0;
962 }
963
964 static void destroy_inodecache(void)
965 {
966 /*
967 * Make sure all delayed rcu free inodes are flushed before we
968 * destroy cache.
969 */
970 rcu_barrier();
971 kmem_cache_destroy(ext4_inode_cachep);
972 }
973
974 void ext4_clear_inode(struct inode *inode)
975 {
976 invalidate_inode_buffers(inode);
977 clear_inode(inode);
978 dquot_drop(inode);
979 ext4_discard_preallocations(inode);
980 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
981 if (EXT4_I(inode)->jinode) {
982 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
983 EXT4_I(inode)->jinode);
984 jbd2_free_inode(EXT4_I(inode)->jinode);
985 EXT4_I(inode)->jinode = NULL;
986 }
987 }
988
989 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
990 u64 ino, u32 generation)
991 {
992 struct inode *inode;
993
994 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
995 return ERR_PTR(-ESTALE);
996 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
997 return ERR_PTR(-ESTALE);
998
999 /* iget isn't really right if the inode is currently unallocated!!
1000 *
1001 * ext4_read_inode will return a bad_inode if the inode had been
1002 * deleted, so we should be safe.
1003 *
1004 * Currently we don't know the generation for parent directory, so
1005 * a generation of 0 means "accept any"
1006 */
1007 inode = ext4_iget_normal(sb, ino);
1008 if (IS_ERR(inode))
1009 return ERR_CAST(inode);
1010 if (generation && inode->i_generation != generation) {
1011 iput(inode);
1012 return ERR_PTR(-ESTALE);
1013 }
1014
1015 return inode;
1016 }
1017
1018 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1019 int fh_len, int fh_type)
1020 {
1021 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1022 ext4_nfs_get_inode);
1023 }
1024
1025 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1026 int fh_len, int fh_type)
1027 {
1028 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1029 ext4_nfs_get_inode);
1030 }
1031
1032 /*
1033 * Try to release metadata pages (indirect blocks, directories) which are
1034 * mapped via the block device. Since these pages could have journal heads
1035 * which would prevent try_to_free_buffers() from freeing them, we must use
1036 * jbd2 layer's try_to_free_buffers() function to release them.
1037 */
1038 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1039 gfp_t wait)
1040 {
1041 journal_t *journal = EXT4_SB(sb)->s_journal;
1042
1043 WARN_ON(PageChecked(page));
1044 if (!page_has_buffers(page))
1045 return 0;
1046 if (journal)
1047 return jbd2_journal_try_to_free_buffers(journal, page,
1048 wait & ~__GFP_WAIT);
1049 return try_to_free_buffers(page);
1050 }
1051
1052 #ifdef CONFIG_QUOTA
1053 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1054 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1055
1056 static int ext4_write_dquot(struct dquot *dquot);
1057 static int ext4_acquire_dquot(struct dquot *dquot);
1058 static int ext4_release_dquot(struct dquot *dquot);
1059 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1060 static int ext4_write_info(struct super_block *sb, int type);
1061 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1062 struct path *path);
1063 static int ext4_quota_off(struct super_block *sb, int type);
1064 static int ext4_quota_on_mount(struct super_block *sb, int type);
1065 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1066 size_t len, loff_t off);
1067 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1068 const char *data, size_t len, loff_t off);
1069 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1070 unsigned int flags);
1071 static int ext4_enable_quotas(struct super_block *sb);
1072
1073 static struct dquot **ext4_get_dquots(struct inode *inode)
1074 {
1075 return EXT4_I(inode)->i_dquot;
1076 }
1077
1078 static const struct dquot_operations ext4_quota_operations = {
1079 .get_reserved_space = ext4_get_reserved_space,
1080 .write_dquot = ext4_write_dquot,
1081 .acquire_dquot = ext4_acquire_dquot,
1082 .release_dquot = ext4_release_dquot,
1083 .mark_dirty = ext4_mark_dquot_dirty,
1084 .write_info = ext4_write_info,
1085 .alloc_dquot = dquot_alloc,
1086 .destroy_dquot = dquot_destroy,
1087 };
1088
1089 static const struct quotactl_ops ext4_qctl_operations = {
1090 .quota_on = ext4_quota_on,
1091 .quota_off = ext4_quota_off,
1092 .quota_sync = dquot_quota_sync,
1093 .get_state = dquot_get_state,
1094 .set_info = dquot_set_dqinfo,
1095 .get_dqblk = dquot_get_dqblk,
1096 .set_dqblk = dquot_set_dqblk
1097 };
1098 #endif
1099
1100 static const struct super_operations ext4_sops = {
1101 .alloc_inode = ext4_alloc_inode,
1102 .destroy_inode = ext4_destroy_inode,
1103 .write_inode = ext4_write_inode,
1104 .dirty_inode = ext4_dirty_inode,
1105 .drop_inode = ext4_drop_inode,
1106 .evict_inode = ext4_evict_inode,
1107 .put_super = ext4_put_super,
1108 .sync_fs = ext4_sync_fs,
1109 .freeze_fs = ext4_freeze,
1110 .unfreeze_fs = ext4_unfreeze,
1111 .statfs = ext4_statfs,
1112 .remount_fs = ext4_remount,
1113 .show_options = ext4_show_options,
1114 #ifdef CONFIG_QUOTA
1115 .quota_read = ext4_quota_read,
1116 .quota_write = ext4_quota_write,
1117 .get_dquots = ext4_get_dquots,
1118 #endif
1119 .bdev_try_to_free_page = bdev_try_to_free_page,
1120 };
1121
1122 static const struct export_operations ext4_export_ops = {
1123 .fh_to_dentry = ext4_fh_to_dentry,
1124 .fh_to_parent = ext4_fh_to_parent,
1125 .get_parent = ext4_get_parent,
1126 };
1127
1128 enum {
1129 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1130 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1131 Opt_nouid32, Opt_debug, Opt_removed,
1132 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1133 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1134 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1135 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1136 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1137 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1138 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1139 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1140 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1141 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1142 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1143 Opt_lazytime, Opt_nolazytime,
1144 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1145 Opt_inode_readahead_blks, Opt_journal_ioprio,
1146 Opt_dioread_nolock, Opt_dioread_lock,
1147 Opt_no_mbcache,
1148 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1149 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1150 };
1151
1152 static const match_table_t tokens = {
1153 {Opt_bsd_df, "bsddf"},
1154 {Opt_minix_df, "minixdf"},
1155 {Opt_grpid, "grpid"},
1156 {Opt_grpid, "bsdgroups"},
1157 {Opt_nogrpid, "nogrpid"},
1158 {Opt_nogrpid, "sysvgroups"},
1159 {Opt_resgid, "resgid=%u"},
1160 {Opt_resuid, "resuid=%u"},
1161 {Opt_sb, "sb=%u"},
1162 {Opt_err_cont, "errors=continue"},
1163 {Opt_err_panic, "errors=panic"},
1164 {Opt_err_ro, "errors=remount-ro"},
1165 {Opt_nouid32, "nouid32"},
1166 {Opt_debug, "debug"},
1167 {Opt_removed, "oldalloc"},
1168 {Opt_removed, "orlov"},
1169 {Opt_user_xattr, "user_xattr"},
1170 {Opt_nouser_xattr, "nouser_xattr"},
1171 {Opt_acl, "acl"},
1172 {Opt_noacl, "noacl"},
1173 {Opt_noload, "norecovery"},
1174 {Opt_noload, "noload"},
1175 {Opt_removed, "nobh"},
1176 {Opt_removed, "bh"},
1177 {Opt_commit, "commit=%u"},
1178 {Opt_min_batch_time, "min_batch_time=%u"},
1179 {Opt_max_batch_time, "max_batch_time=%u"},
1180 {Opt_journal_dev, "journal_dev=%u"},
1181 {Opt_journal_path, "journal_path=%s"},
1182 {Opt_journal_checksum, "journal_checksum"},
1183 {Opt_nojournal_checksum, "nojournal_checksum"},
1184 {Opt_journal_async_commit, "journal_async_commit"},
1185 {Opt_abort, "abort"},
1186 {Opt_data_journal, "data=journal"},
1187 {Opt_data_ordered, "data=ordered"},
1188 {Opt_data_writeback, "data=writeback"},
1189 {Opt_data_err_abort, "data_err=abort"},
1190 {Opt_data_err_ignore, "data_err=ignore"},
1191 {Opt_offusrjquota, "usrjquota="},
1192 {Opt_usrjquota, "usrjquota=%s"},
1193 {Opt_offgrpjquota, "grpjquota="},
1194 {Opt_grpjquota, "grpjquota=%s"},
1195 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1196 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1197 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1198 {Opt_grpquota, "grpquota"},
1199 {Opt_noquota, "noquota"},
1200 {Opt_quota, "quota"},
1201 {Opt_usrquota, "usrquota"},
1202 {Opt_barrier, "barrier=%u"},
1203 {Opt_barrier, "barrier"},
1204 {Opt_nobarrier, "nobarrier"},
1205 {Opt_i_version, "i_version"},
1206 {Opt_dax, "dax"},
1207 {Opt_stripe, "stripe=%u"},
1208 {Opt_delalloc, "delalloc"},
1209 {Opt_lazytime, "lazytime"},
1210 {Opt_nolazytime, "nolazytime"},
1211 {Opt_nodelalloc, "nodelalloc"},
1212 {Opt_removed, "mblk_io_submit"},
1213 {Opt_removed, "nomblk_io_submit"},
1214 {Opt_block_validity, "block_validity"},
1215 {Opt_noblock_validity, "noblock_validity"},
1216 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1217 {Opt_journal_ioprio, "journal_ioprio=%u"},
1218 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1219 {Opt_auto_da_alloc, "auto_da_alloc"},
1220 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1221 {Opt_dioread_nolock, "dioread_nolock"},
1222 {Opt_dioread_lock, "dioread_lock"},
1223 {Opt_discard, "discard"},
1224 {Opt_nodiscard, "nodiscard"},
1225 {Opt_init_itable, "init_itable=%u"},
1226 {Opt_no_mbcache, "no_mbcache"},
1227 {Opt_init_itable, "init_itable"},
1228 {Opt_noinit_itable, "noinit_itable"},
1229 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1230 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1231 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1232 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1233 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1234 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1235 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1236 {Opt_err, NULL},
1237 };
1238
1239 static ext4_fsblk_t get_sb_block(void **data)
1240 {
1241 ext4_fsblk_t sb_block;
1242 char *options = (char *) *data;
1243
1244 if (!options || strncmp(options, "sb=", 3) != 0)
1245 return 1; /* Default location */
1246
1247 options += 3;
1248 /* TODO: use simple_strtoll with >32bit ext4 */
1249 sb_block = simple_strtoul(options, &options, 0);
1250 if (*options && *options != ',') {
1251 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1252 (char *) *data);
1253 return 1;
1254 }
1255 if (*options == ',')
1256 options++;
1257 *data = (void *) options;
1258
1259 return sb_block;
1260 }
1261
1262 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1263 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1264 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1265
1266 #ifdef CONFIG_QUOTA
1267 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1268 {
1269 struct ext4_sb_info *sbi = EXT4_SB(sb);
1270 char *qname;
1271 int ret = -1;
1272
1273 if (sb_any_quota_loaded(sb) &&
1274 !sbi->s_qf_names[qtype]) {
1275 ext4_msg(sb, KERN_ERR,
1276 "Cannot change journaled "
1277 "quota options when quota turned on");
1278 return -1;
1279 }
1280 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1281 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1282 "ignored when QUOTA feature is enabled");
1283 return 1;
1284 }
1285 qname = match_strdup(args);
1286 if (!qname) {
1287 ext4_msg(sb, KERN_ERR,
1288 "Not enough memory for storing quotafile name");
1289 return -1;
1290 }
1291 if (sbi->s_qf_names[qtype]) {
1292 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1293 ret = 1;
1294 else
1295 ext4_msg(sb, KERN_ERR,
1296 "%s quota file already specified",
1297 QTYPE2NAME(qtype));
1298 goto errout;
1299 }
1300 if (strchr(qname, '/')) {
1301 ext4_msg(sb, KERN_ERR,
1302 "quotafile must be on filesystem root");
1303 goto errout;
1304 }
1305 sbi->s_qf_names[qtype] = qname;
1306 set_opt(sb, QUOTA);
1307 return 1;
1308 errout:
1309 kfree(qname);
1310 return ret;
1311 }
1312
1313 static int clear_qf_name(struct super_block *sb, int qtype)
1314 {
1315
1316 struct ext4_sb_info *sbi = EXT4_SB(sb);
1317
1318 if (sb_any_quota_loaded(sb) &&
1319 sbi->s_qf_names[qtype]) {
1320 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1321 " when quota turned on");
1322 return -1;
1323 }
1324 kfree(sbi->s_qf_names[qtype]);
1325 sbi->s_qf_names[qtype] = NULL;
1326 return 1;
1327 }
1328 #endif
1329
1330 #define MOPT_SET 0x0001
1331 #define MOPT_CLEAR 0x0002
1332 #define MOPT_NOSUPPORT 0x0004
1333 #define MOPT_EXPLICIT 0x0008
1334 #define MOPT_CLEAR_ERR 0x0010
1335 #define MOPT_GTE0 0x0020
1336 #ifdef CONFIG_QUOTA
1337 #define MOPT_Q 0
1338 #define MOPT_QFMT 0x0040
1339 #else
1340 #define MOPT_Q MOPT_NOSUPPORT
1341 #define MOPT_QFMT MOPT_NOSUPPORT
1342 #endif
1343 #define MOPT_DATAJ 0x0080
1344 #define MOPT_NO_EXT2 0x0100
1345 #define MOPT_NO_EXT3 0x0200
1346 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1347 #define MOPT_STRING 0x0400
1348
1349 static const struct mount_opts {
1350 int token;
1351 int mount_opt;
1352 int flags;
1353 } ext4_mount_opts[] = {
1354 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1355 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1356 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1357 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1358 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1359 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1360 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1361 MOPT_EXT4_ONLY | MOPT_SET},
1362 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1363 MOPT_EXT4_ONLY | MOPT_CLEAR},
1364 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1365 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1366 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1367 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1368 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1369 MOPT_EXT4_ONLY | MOPT_CLEAR},
1370 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1371 MOPT_EXT4_ONLY | MOPT_CLEAR},
1372 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1373 MOPT_EXT4_ONLY | MOPT_SET},
1374 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1375 EXT4_MOUNT_JOURNAL_CHECKSUM),
1376 MOPT_EXT4_ONLY | MOPT_SET},
1377 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1378 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1379 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1380 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1381 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1382 MOPT_NO_EXT2 | MOPT_SET},
1383 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1384 MOPT_NO_EXT2 | MOPT_CLEAR},
1385 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1386 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1387 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1388 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1389 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1390 {Opt_no_mbcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1391 {Opt_commit, 0, MOPT_GTE0},
1392 {Opt_max_batch_time, 0, MOPT_GTE0},
1393 {Opt_min_batch_time, 0, MOPT_GTE0},
1394 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1395 {Opt_init_itable, 0, MOPT_GTE0},
1396 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1397 {Opt_stripe, 0, MOPT_GTE0},
1398 {Opt_resuid, 0, MOPT_GTE0},
1399 {Opt_resgid, 0, MOPT_GTE0},
1400 {Opt_journal_dev, 0, MOPT_GTE0},
1401 {Opt_journal_path, 0, MOPT_STRING},
1402 {Opt_journal_ioprio, 0, MOPT_GTE0},
1403 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1404 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1405 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1406 MOPT_NO_EXT2 | MOPT_DATAJ},
1407 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1408 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1409 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1410 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1411 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1412 #else
1413 {Opt_acl, 0, MOPT_NOSUPPORT},
1414 {Opt_noacl, 0, MOPT_NOSUPPORT},
1415 #endif
1416 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1417 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1418 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1419 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1420 MOPT_SET | MOPT_Q},
1421 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1422 MOPT_SET | MOPT_Q},
1423 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1424 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1425 {Opt_usrjquota, 0, MOPT_Q},
1426 {Opt_grpjquota, 0, MOPT_Q},
1427 {Opt_offusrjquota, 0, MOPT_Q},
1428 {Opt_offgrpjquota, 0, MOPT_Q},
1429 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1430 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1431 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1432 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1433 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1434 {Opt_err, 0, 0}
1435 };
1436
1437 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1438 substring_t *args, unsigned long *journal_devnum,
1439 unsigned int *journal_ioprio, int is_remount)
1440 {
1441 struct ext4_sb_info *sbi = EXT4_SB(sb);
1442 const struct mount_opts *m;
1443 kuid_t uid;
1444 kgid_t gid;
1445 int arg = 0;
1446
1447 #ifdef CONFIG_QUOTA
1448 if (token == Opt_usrjquota)
1449 return set_qf_name(sb, USRQUOTA, &args[0]);
1450 else if (token == Opt_grpjquota)
1451 return set_qf_name(sb, GRPQUOTA, &args[0]);
1452 else if (token == Opt_offusrjquota)
1453 return clear_qf_name(sb, USRQUOTA);
1454 else if (token == Opt_offgrpjquota)
1455 return clear_qf_name(sb, GRPQUOTA);
1456 #endif
1457 switch (token) {
1458 case Opt_noacl:
1459 case Opt_nouser_xattr:
1460 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1461 break;
1462 case Opt_sb:
1463 return 1; /* handled by get_sb_block() */
1464 case Opt_removed:
1465 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1466 return 1;
1467 case Opt_abort:
1468 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1469 return 1;
1470 case Opt_i_version:
1471 sb->s_flags |= MS_I_VERSION;
1472 return 1;
1473 case Opt_lazytime:
1474 sb->s_flags |= MS_LAZYTIME;
1475 return 1;
1476 case Opt_nolazytime:
1477 sb->s_flags &= ~MS_LAZYTIME;
1478 return 1;
1479 }
1480
1481 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1482 if (token == m->token)
1483 break;
1484
1485 if (m->token == Opt_err) {
1486 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1487 "or missing value", opt);
1488 return -1;
1489 }
1490
1491 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1492 ext4_msg(sb, KERN_ERR,
1493 "Mount option \"%s\" incompatible with ext2", opt);
1494 return -1;
1495 }
1496 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1497 ext4_msg(sb, KERN_ERR,
1498 "Mount option \"%s\" incompatible with ext3", opt);
1499 return -1;
1500 }
1501
1502 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1503 return -1;
1504 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1505 return -1;
1506 if (m->flags & MOPT_EXPLICIT)
1507 set_opt2(sb, EXPLICIT_DELALLOC);
1508 if (m->flags & MOPT_CLEAR_ERR)
1509 clear_opt(sb, ERRORS_MASK);
1510 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1511 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1512 "options when quota turned on");
1513 return -1;
1514 }
1515
1516 if (m->flags & MOPT_NOSUPPORT) {
1517 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1518 } else if (token == Opt_commit) {
1519 if (arg == 0)
1520 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1521 sbi->s_commit_interval = HZ * arg;
1522 } else if (token == Opt_max_batch_time) {
1523 sbi->s_max_batch_time = arg;
1524 } else if (token == Opt_min_batch_time) {
1525 sbi->s_min_batch_time = arg;
1526 } else if (token == Opt_inode_readahead_blks) {
1527 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1528 ext4_msg(sb, KERN_ERR,
1529 "EXT4-fs: inode_readahead_blks must be "
1530 "0 or a power of 2 smaller than 2^31");
1531 return -1;
1532 }
1533 sbi->s_inode_readahead_blks = arg;
1534 } else if (token == Opt_init_itable) {
1535 set_opt(sb, INIT_INODE_TABLE);
1536 if (!args->from)
1537 arg = EXT4_DEF_LI_WAIT_MULT;
1538 sbi->s_li_wait_mult = arg;
1539 } else if (token == Opt_max_dir_size_kb) {
1540 sbi->s_max_dir_size_kb = arg;
1541 } else if (token == Opt_stripe) {
1542 sbi->s_stripe = arg;
1543 } else if (token == Opt_resuid) {
1544 uid = make_kuid(current_user_ns(), arg);
1545 if (!uid_valid(uid)) {
1546 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1547 return -1;
1548 }
1549 sbi->s_resuid = uid;
1550 } else if (token == Opt_resgid) {
1551 gid = make_kgid(current_user_ns(), arg);
1552 if (!gid_valid(gid)) {
1553 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1554 return -1;
1555 }
1556 sbi->s_resgid = gid;
1557 } else if (token == Opt_journal_dev) {
1558 if (is_remount) {
1559 ext4_msg(sb, KERN_ERR,
1560 "Cannot specify journal on remount");
1561 return -1;
1562 }
1563 *journal_devnum = arg;
1564 } else if (token == Opt_journal_path) {
1565 char *journal_path;
1566 struct inode *journal_inode;
1567 struct path path;
1568 int error;
1569
1570 if (is_remount) {
1571 ext4_msg(sb, KERN_ERR,
1572 "Cannot specify journal on remount");
1573 return -1;
1574 }
1575 journal_path = match_strdup(&args[0]);
1576 if (!journal_path) {
1577 ext4_msg(sb, KERN_ERR, "error: could not dup "
1578 "journal device string");
1579 return -1;
1580 }
1581
1582 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1583 if (error) {
1584 ext4_msg(sb, KERN_ERR, "error: could not find "
1585 "journal device path: error %d", error);
1586 kfree(journal_path);
1587 return -1;
1588 }
1589
1590 journal_inode = d_inode(path.dentry);
1591 if (!S_ISBLK(journal_inode->i_mode)) {
1592 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1593 "is not a block device", journal_path);
1594 path_put(&path);
1595 kfree(journal_path);
1596 return -1;
1597 }
1598
1599 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1600 path_put(&path);
1601 kfree(journal_path);
1602 } else if (token == Opt_journal_ioprio) {
1603 if (arg > 7) {
1604 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1605 " (must be 0-7)");
1606 return -1;
1607 }
1608 *journal_ioprio =
1609 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1610 } else if (token == Opt_test_dummy_encryption) {
1611 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1612 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1613 ext4_msg(sb, KERN_WARNING,
1614 "Test dummy encryption mode enabled");
1615 #else
1616 ext4_msg(sb, KERN_WARNING,
1617 "Test dummy encryption mount option ignored");
1618 #endif
1619 } else if (m->flags & MOPT_DATAJ) {
1620 if (is_remount) {
1621 if (!sbi->s_journal)
1622 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1623 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1624 ext4_msg(sb, KERN_ERR,
1625 "Cannot change data mode on remount");
1626 return -1;
1627 }
1628 } else {
1629 clear_opt(sb, DATA_FLAGS);
1630 sbi->s_mount_opt |= m->mount_opt;
1631 }
1632 #ifdef CONFIG_QUOTA
1633 } else if (m->flags & MOPT_QFMT) {
1634 if (sb_any_quota_loaded(sb) &&
1635 sbi->s_jquota_fmt != m->mount_opt) {
1636 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1637 "quota options when quota turned on");
1638 return -1;
1639 }
1640 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1641 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1642 ext4_msg(sb, KERN_INFO,
1643 "Quota format mount options ignored "
1644 "when QUOTA feature is enabled");
1645 return 1;
1646 }
1647 sbi->s_jquota_fmt = m->mount_opt;
1648 #endif
1649 #ifndef CONFIG_FS_DAX
1650 } else if (token == Opt_dax) {
1651 ext4_msg(sb, KERN_INFO, "dax option not supported");
1652 return -1;
1653 #endif
1654 } else {
1655 if (!args->from)
1656 arg = 1;
1657 if (m->flags & MOPT_CLEAR)
1658 arg = !arg;
1659 else if (unlikely(!(m->flags & MOPT_SET))) {
1660 ext4_msg(sb, KERN_WARNING,
1661 "buggy handling of option %s", opt);
1662 WARN_ON(1);
1663 return -1;
1664 }
1665 if (arg != 0)
1666 sbi->s_mount_opt |= m->mount_opt;
1667 else
1668 sbi->s_mount_opt &= ~m->mount_opt;
1669 }
1670 return 1;
1671 }
1672
1673 static int parse_options(char *options, struct super_block *sb,
1674 unsigned long *journal_devnum,
1675 unsigned int *journal_ioprio,
1676 int is_remount)
1677 {
1678 struct ext4_sb_info *sbi = EXT4_SB(sb);
1679 char *p;
1680 substring_t args[MAX_OPT_ARGS];
1681 int token;
1682
1683 if (!options)
1684 return 1;
1685
1686 while ((p = strsep(&options, ",")) != NULL) {
1687 if (!*p)
1688 continue;
1689 /*
1690 * Initialize args struct so we know whether arg was
1691 * found; some options take optional arguments.
1692 */
1693 args[0].to = args[0].from = NULL;
1694 token = match_token(p, tokens, args);
1695 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1696 journal_ioprio, is_remount) < 0)
1697 return 0;
1698 }
1699 #ifdef CONFIG_QUOTA
1700 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1701 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1702 ext4_msg(sb, KERN_INFO, "Quota feature enabled, usrquota and grpquota "
1703 "mount options ignored.");
1704 clear_opt(sb, USRQUOTA);
1705 clear_opt(sb, GRPQUOTA);
1706 } else if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1707 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1708 clear_opt(sb, USRQUOTA);
1709
1710 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1711 clear_opt(sb, GRPQUOTA);
1712
1713 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1714 ext4_msg(sb, KERN_ERR, "old and new quota "
1715 "format mixing");
1716 return 0;
1717 }
1718
1719 if (!sbi->s_jquota_fmt) {
1720 ext4_msg(sb, KERN_ERR, "journaled quota format "
1721 "not specified");
1722 return 0;
1723 }
1724 }
1725 #endif
1726 if (test_opt(sb, DIOREAD_NOLOCK)) {
1727 int blocksize =
1728 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1729
1730 if (blocksize < PAGE_CACHE_SIZE) {
1731 ext4_msg(sb, KERN_ERR, "can't mount with "
1732 "dioread_nolock if block size != PAGE_SIZE");
1733 return 0;
1734 }
1735 }
1736 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1737 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1738 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1739 "in data=ordered mode");
1740 return 0;
1741 }
1742 return 1;
1743 }
1744
1745 static inline void ext4_show_quota_options(struct seq_file *seq,
1746 struct super_block *sb)
1747 {
1748 #if defined(CONFIG_QUOTA)
1749 struct ext4_sb_info *sbi = EXT4_SB(sb);
1750
1751 if (sbi->s_jquota_fmt) {
1752 char *fmtname = "";
1753
1754 switch (sbi->s_jquota_fmt) {
1755 case QFMT_VFS_OLD:
1756 fmtname = "vfsold";
1757 break;
1758 case QFMT_VFS_V0:
1759 fmtname = "vfsv0";
1760 break;
1761 case QFMT_VFS_V1:
1762 fmtname = "vfsv1";
1763 break;
1764 }
1765 seq_printf(seq, ",jqfmt=%s", fmtname);
1766 }
1767
1768 if (sbi->s_qf_names[USRQUOTA])
1769 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1770
1771 if (sbi->s_qf_names[GRPQUOTA])
1772 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1773 #endif
1774 }
1775
1776 static const char *token2str(int token)
1777 {
1778 const struct match_token *t;
1779
1780 for (t = tokens; t->token != Opt_err; t++)
1781 if (t->token == token && !strchr(t->pattern, '='))
1782 break;
1783 return t->pattern;
1784 }
1785
1786 /*
1787 * Show an option if
1788 * - it's set to a non-default value OR
1789 * - if the per-sb default is different from the global default
1790 */
1791 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1792 int nodefs)
1793 {
1794 struct ext4_sb_info *sbi = EXT4_SB(sb);
1795 struct ext4_super_block *es = sbi->s_es;
1796 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1797 const struct mount_opts *m;
1798 char sep = nodefs ? '\n' : ',';
1799
1800 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1801 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1802
1803 if (sbi->s_sb_block != 1)
1804 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1805
1806 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1807 int want_set = m->flags & MOPT_SET;
1808 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1809 (m->flags & MOPT_CLEAR_ERR))
1810 continue;
1811 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1812 continue; /* skip if same as the default */
1813 if ((want_set &&
1814 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1815 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1816 continue; /* select Opt_noFoo vs Opt_Foo */
1817 SEQ_OPTS_PRINT("%s", token2str(m->token));
1818 }
1819
1820 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1821 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1822 SEQ_OPTS_PRINT("resuid=%u",
1823 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1824 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1825 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1826 SEQ_OPTS_PRINT("resgid=%u",
1827 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1828 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1829 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1830 SEQ_OPTS_PUTS("errors=remount-ro");
1831 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1832 SEQ_OPTS_PUTS("errors=continue");
1833 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1834 SEQ_OPTS_PUTS("errors=panic");
1835 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1836 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1837 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1838 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1839 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1840 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1841 if (sb->s_flags & MS_I_VERSION)
1842 SEQ_OPTS_PUTS("i_version");
1843 if (nodefs || sbi->s_stripe)
1844 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1845 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1846 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1847 SEQ_OPTS_PUTS("data=journal");
1848 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1849 SEQ_OPTS_PUTS("data=ordered");
1850 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1851 SEQ_OPTS_PUTS("data=writeback");
1852 }
1853 if (nodefs ||
1854 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1855 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1856 sbi->s_inode_readahead_blks);
1857
1858 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1859 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1860 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1861 if (nodefs || sbi->s_max_dir_size_kb)
1862 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1863
1864 ext4_show_quota_options(seq, sb);
1865 return 0;
1866 }
1867
1868 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1869 {
1870 return _ext4_show_options(seq, root->d_sb, 0);
1871 }
1872
1873 static int options_seq_show(struct seq_file *seq, void *offset)
1874 {
1875 struct super_block *sb = seq->private;
1876 int rc;
1877
1878 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1879 rc = _ext4_show_options(seq, sb, 1);
1880 seq_puts(seq, "\n");
1881 return rc;
1882 }
1883
1884 static int options_open_fs(struct inode *inode, struct file *file)
1885 {
1886 return single_open(file, options_seq_show, PDE_DATA(inode));
1887 }
1888
1889 static const struct file_operations ext4_seq_options_fops = {
1890 .owner = THIS_MODULE,
1891 .open = options_open_fs,
1892 .read = seq_read,
1893 .llseek = seq_lseek,
1894 .release = single_release,
1895 };
1896
1897 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1898 int read_only)
1899 {
1900 struct ext4_sb_info *sbi = EXT4_SB(sb);
1901 int res = 0;
1902
1903 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1904 ext4_msg(sb, KERN_ERR, "revision level too high, "
1905 "forcing read-only mode");
1906 res = MS_RDONLY;
1907 }
1908 if (read_only)
1909 goto done;
1910 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1911 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1912 "running e2fsck is recommended");
1913 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1914 ext4_msg(sb, KERN_WARNING,
1915 "warning: mounting fs with errors, "
1916 "running e2fsck is recommended");
1917 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1918 le16_to_cpu(es->s_mnt_count) >=
1919 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1920 ext4_msg(sb, KERN_WARNING,
1921 "warning: maximal mount count reached, "
1922 "running e2fsck is recommended");
1923 else if (le32_to_cpu(es->s_checkinterval) &&
1924 (le32_to_cpu(es->s_lastcheck) +
1925 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1926 ext4_msg(sb, KERN_WARNING,
1927 "warning: checktime reached, "
1928 "running e2fsck is recommended");
1929 if (!sbi->s_journal)
1930 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1931 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1932 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1933 le16_add_cpu(&es->s_mnt_count, 1);
1934 es->s_mtime = cpu_to_le32(get_seconds());
1935 ext4_update_dynamic_rev(sb);
1936 if (sbi->s_journal)
1937 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1938
1939 ext4_commit_super(sb, 1);
1940 done:
1941 if (test_opt(sb, DEBUG))
1942 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1943 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1944 sb->s_blocksize,
1945 sbi->s_groups_count,
1946 EXT4_BLOCKS_PER_GROUP(sb),
1947 EXT4_INODES_PER_GROUP(sb),
1948 sbi->s_mount_opt, sbi->s_mount_opt2);
1949
1950 cleancache_init_fs(sb);
1951 return res;
1952 }
1953
1954 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1955 {
1956 struct ext4_sb_info *sbi = EXT4_SB(sb);
1957 struct flex_groups *new_groups;
1958 int size;
1959
1960 if (!sbi->s_log_groups_per_flex)
1961 return 0;
1962
1963 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1964 if (size <= sbi->s_flex_groups_allocated)
1965 return 0;
1966
1967 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1968 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1969 if (!new_groups) {
1970 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1971 size / (int) sizeof(struct flex_groups));
1972 return -ENOMEM;
1973 }
1974
1975 if (sbi->s_flex_groups) {
1976 memcpy(new_groups, sbi->s_flex_groups,
1977 (sbi->s_flex_groups_allocated *
1978 sizeof(struct flex_groups)));
1979 kvfree(sbi->s_flex_groups);
1980 }
1981 sbi->s_flex_groups = new_groups;
1982 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1983 return 0;
1984 }
1985
1986 static int ext4_fill_flex_info(struct super_block *sb)
1987 {
1988 struct ext4_sb_info *sbi = EXT4_SB(sb);
1989 struct ext4_group_desc *gdp = NULL;
1990 ext4_group_t flex_group;
1991 int i, err;
1992
1993 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1994 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1995 sbi->s_log_groups_per_flex = 0;
1996 return 1;
1997 }
1998
1999 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2000 if (err)
2001 goto failed;
2002
2003 for (i = 0; i < sbi->s_groups_count; i++) {
2004 gdp = ext4_get_group_desc(sb, i, NULL);
2005
2006 flex_group = ext4_flex_group(sbi, i);
2007 atomic_add(ext4_free_inodes_count(sb, gdp),
2008 &sbi->s_flex_groups[flex_group].free_inodes);
2009 atomic64_add(ext4_free_group_clusters(sb, gdp),
2010 &sbi->s_flex_groups[flex_group].free_clusters);
2011 atomic_add(ext4_used_dirs_count(sb, gdp),
2012 &sbi->s_flex_groups[flex_group].used_dirs);
2013 }
2014
2015 return 1;
2016 failed:
2017 return 0;
2018 }
2019
2020 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2021 struct ext4_group_desc *gdp)
2022 {
2023 int offset;
2024 __u16 crc = 0;
2025 __le32 le_group = cpu_to_le32(block_group);
2026
2027 if (ext4_has_metadata_csum(sbi->s_sb)) {
2028 /* Use new metadata_csum algorithm */
2029 __le16 save_csum;
2030 __u32 csum32;
2031
2032 save_csum = gdp->bg_checksum;
2033 gdp->bg_checksum = 0;
2034 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2035 sizeof(le_group));
2036 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2037 sbi->s_desc_size);
2038 gdp->bg_checksum = save_csum;
2039
2040 crc = csum32 & 0xFFFF;
2041 goto out;
2042 }
2043
2044 /* old crc16 code */
2045 if (!(sbi->s_es->s_feature_ro_compat &
2046 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2047 return 0;
2048
2049 offset = offsetof(struct ext4_group_desc, bg_checksum);
2050
2051 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2052 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2053 crc = crc16(crc, (__u8 *)gdp, offset);
2054 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2055 /* for checksum of struct ext4_group_desc do the rest...*/
2056 if ((sbi->s_es->s_feature_incompat &
2057 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2058 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2059 crc = crc16(crc, (__u8 *)gdp + offset,
2060 le16_to_cpu(sbi->s_es->s_desc_size) -
2061 offset);
2062
2063 out:
2064 return cpu_to_le16(crc);
2065 }
2066
2067 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2068 struct ext4_group_desc *gdp)
2069 {
2070 if (ext4_has_group_desc_csum(sb) &&
2071 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2072 block_group, gdp)))
2073 return 0;
2074
2075 return 1;
2076 }
2077
2078 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2079 struct ext4_group_desc *gdp)
2080 {
2081 if (!ext4_has_group_desc_csum(sb))
2082 return;
2083 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2084 }
2085
2086 /* Called at mount-time, super-block is locked */
2087 static int ext4_check_descriptors(struct super_block *sb,
2088 ext4_fsblk_t sb_block,
2089 ext4_group_t *first_not_zeroed)
2090 {
2091 struct ext4_sb_info *sbi = EXT4_SB(sb);
2092 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2093 ext4_fsblk_t last_block;
2094 ext4_fsblk_t block_bitmap;
2095 ext4_fsblk_t inode_bitmap;
2096 ext4_fsblk_t inode_table;
2097 int flexbg_flag = 0;
2098 ext4_group_t i, grp = sbi->s_groups_count;
2099
2100 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2101 flexbg_flag = 1;
2102
2103 ext4_debug("Checking group descriptors");
2104
2105 for (i = 0; i < sbi->s_groups_count; i++) {
2106 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2107
2108 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2109 last_block = ext4_blocks_count(sbi->s_es) - 1;
2110 else
2111 last_block = first_block +
2112 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2113
2114 if ((grp == sbi->s_groups_count) &&
2115 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2116 grp = i;
2117
2118 block_bitmap = ext4_block_bitmap(sb, gdp);
2119 if (block_bitmap == sb_block) {
2120 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2121 "Block bitmap for group %u overlaps "
2122 "superblock", i);
2123 }
2124 if (block_bitmap < first_block || block_bitmap > last_block) {
2125 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2126 "Block bitmap for group %u not in group "
2127 "(block %llu)!", i, block_bitmap);
2128 return 0;
2129 }
2130 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2131 if (inode_bitmap == sb_block) {
2132 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2133 "Inode bitmap for group %u overlaps "
2134 "superblock", i);
2135 }
2136 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2137 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2138 "Inode bitmap for group %u not in group "
2139 "(block %llu)!", i, inode_bitmap);
2140 return 0;
2141 }
2142 inode_table = ext4_inode_table(sb, gdp);
2143 if (inode_table == sb_block) {
2144 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2145 "Inode table for group %u overlaps "
2146 "superblock", i);
2147 }
2148 if (inode_table < first_block ||
2149 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2150 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2151 "Inode table for group %u not in group "
2152 "(block %llu)!", i, inode_table);
2153 return 0;
2154 }
2155 ext4_lock_group(sb, i);
2156 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2157 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2158 "Checksum for group %u failed (%u!=%u)",
2159 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2160 gdp)), le16_to_cpu(gdp->bg_checksum));
2161 if (!(sb->s_flags & MS_RDONLY)) {
2162 ext4_unlock_group(sb, i);
2163 return 0;
2164 }
2165 }
2166 ext4_unlock_group(sb, i);
2167 if (!flexbg_flag)
2168 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2169 }
2170 if (NULL != first_not_zeroed)
2171 *first_not_zeroed = grp;
2172 return 1;
2173 }
2174
2175 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2176 * the superblock) which were deleted from all directories, but held open by
2177 * a process at the time of a crash. We walk the list and try to delete these
2178 * inodes at recovery time (only with a read-write filesystem).
2179 *
2180 * In order to keep the orphan inode chain consistent during traversal (in
2181 * case of crash during recovery), we link each inode into the superblock
2182 * orphan list_head and handle it the same way as an inode deletion during
2183 * normal operation (which journals the operations for us).
2184 *
2185 * We only do an iget() and an iput() on each inode, which is very safe if we
2186 * accidentally point at an in-use or already deleted inode. The worst that
2187 * can happen in this case is that we get a "bit already cleared" message from
2188 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2189 * e2fsck was run on this filesystem, and it must have already done the orphan
2190 * inode cleanup for us, so we can safely abort without any further action.
2191 */
2192 static void ext4_orphan_cleanup(struct super_block *sb,
2193 struct ext4_super_block *es)
2194 {
2195 unsigned int s_flags = sb->s_flags;
2196 int nr_orphans = 0, nr_truncates = 0;
2197 #ifdef CONFIG_QUOTA
2198 int i;
2199 #endif
2200 if (!es->s_last_orphan) {
2201 jbd_debug(4, "no orphan inodes to clean up\n");
2202 return;
2203 }
2204
2205 if (bdev_read_only(sb->s_bdev)) {
2206 ext4_msg(sb, KERN_ERR, "write access "
2207 "unavailable, skipping orphan cleanup");
2208 return;
2209 }
2210
2211 /* Check if feature set would not allow a r/w mount */
2212 if (!ext4_feature_set_ok(sb, 0)) {
2213 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2214 "unknown ROCOMPAT features");
2215 return;
2216 }
2217
2218 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2219 /* don't clear list on RO mount w/ errors */
2220 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2221 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2222 "clearing orphan list.\n");
2223 es->s_last_orphan = 0;
2224 }
2225 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2226 return;
2227 }
2228
2229 if (s_flags & MS_RDONLY) {
2230 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2231 sb->s_flags &= ~MS_RDONLY;
2232 }
2233 #ifdef CONFIG_QUOTA
2234 /* Needed for iput() to work correctly and not trash data */
2235 sb->s_flags |= MS_ACTIVE;
2236 /* Turn on quotas so that they are updated correctly */
2237 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2238 if (EXT4_SB(sb)->s_qf_names[i]) {
2239 int ret = ext4_quota_on_mount(sb, i);
2240 if (ret < 0)
2241 ext4_msg(sb, KERN_ERR,
2242 "Cannot turn on journaled "
2243 "quota: error %d", ret);
2244 }
2245 }
2246 #endif
2247
2248 while (es->s_last_orphan) {
2249 struct inode *inode;
2250
2251 /*
2252 * We may have encountered an error during cleanup; if
2253 * so, skip the rest.
2254 */
2255 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2256 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2257 es->s_last_orphan = 0;
2258 break;
2259 }
2260
2261 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2262 if (IS_ERR(inode)) {
2263 es->s_last_orphan = 0;
2264 break;
2265 }
2266
2267 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2268 dquot_initialize(inode);
2269 if (inode->i_nlink) {
2270 if (test_opt(sb, DEBUG))
2271 ext4_msg(sb, KERN_DEBUG,
2272 "%s: truncating inode %lu to %lld bytes",
2273 __func__, inode->i_ino, inode->i_size);
2274 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2275 inode->i_ino, inode->i_size);
2276 mutex_lock(&inode->i_mutex);
2277 truncate_inode_pages(inode->i_mapping, inode->i_size);
2278 ext4_truncate(inode);
2279 mutex_unlock(&inode->i_mutex);
2280 nr_truncates++;
2281 } else {
2282 if (test_opt(sb, DEBUG))
2283 ext4_msg(sb, KERN_DEBUG,
2284 "%s: deleting unreferenced inode %lu",
2285 __func__, inode->i_ino);
2286 jbd_debug(2, "deleting unreferenced inode %lu\n",
2287 inode->i_ino);
2288 nr_orphans++;
2289 }
2290 iput(inode); /* The delete magic happens here! */
2291 }
2292
2293 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2294
2295 if (nr_orphans)
2296 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2297 PLURAL(nr_orphans));
2298 if (nr_truncates)
2299 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2300 PLURAL(nr_truncates));
2301 #ifdef CONFIG_QUOTA
2302 /* Turn quotas off */
2303 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2304 if (sb_dqopt(sb)->files[i])
2305 dquot_quota_off(sb, i);
2306 }
2307 #endif
2308 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2309 }
2310
2311 /*
2312 * Maximal extent format file size.
2313 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2314 * extent format containers, within a sector_t, and within i_blocks
2315 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2316 * so that won't be a limiting factor.
2317 *
2318 * However there is other limiting factor. We do store extents in the form
2319 * of starting block and length, hence the resulting length of the extent
2320 * covering maximum file size must fit into on-disk format containers as
2321 * well. Given that length is always by 1 unit bigger than max unit (because
2322 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2323 *
2324 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2325 */
2326 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2327 {
2328 loff_t res;
2329 loff_t upper_limit = MAX_LFS_FILESIZE;
2330
2331 /* small i_blocks in vfs inode? */
2332 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2333 /*
2334 * CONFIG_LBDAF is not enabled implies the inode
2335 * i_block represent total blocks in 512 bytes
2336 * 32 == size of vfs inode i_blocks * 8
2337 */
2338 upper_limit = (1LL << 32) - 1;
2339
2340 /* total blocks in file system block size */
2341 upper_limit >>= (blkbits - 9);
2342 upper_limit <<= blkbits;
2343 }
2344
2345 /*
2346 * 32-bit extent-start container, ee_block. We lower the maxbytes
2347 * by one fs block, so ee_len can cover the extent of maximum file
2348 * size
2349 */
2350 res = (1LL << 32) - 1;
2351 res <<= blkbits;
2352
2353 /* Sanity check against vm- & vfs- imposed limits */
2354 if (res > upper_limit)
2355 res = upper_limit;
2356
2357 return res;
2358 }
2359
2360 /*
2361 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2362 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2363 * We need to be 1 filesystem block less than the 2^48 sector limit.
2364 */
2365 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2366 {
2367 loff_t res = EXT4_NDIR_BLOCKS;
2368 int meta_blocks;
2369 loff_t upper_limit;
2370 /* This is calculated to be the largest file size for a dense, block
2371 * mapped file such that the file's total number of 512-byte sectors,
2372 * including data and all indirect blocks, does not exceed (2^48 - 1).
2373 *
2374 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2375 * number of 512-byte sectors of the file.
2376 */
2377
2378 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2379 /*
2380 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2381 * the inode i_block field represents total file blocks in
2382 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2383 */
2384 upper_limit = (1LL << 32) - 1;
2385
2386 /* total blocks in file system block size */
2387 upper_limit >>= (bits - 9);
2388
2389 } else {
2390 /*
2391 * We use 48 bit ext4_inode i_blocks
2392 * With EXT4_HUGE_FILE_FL set the i_blocks
2393 * represent total number of blocks in
2394 * file system block size
2395 */
2396 upper_limit = (1LL << 48) - 1;
2397
2398 }
2399
2400 /* indirect blocks */
2401 meta_blocks = 1;
2402 /* double indirect blocks */
2403 meta_blocks += 1 + (1LL << (bits-2));
2404 /* tripple indirect blocks */
2405 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2406
2407 upper_limit -= meta_blocks;
2408 upper_limit <<= bits;
2409
2410 res += 1LL << (bits-2);
2411 res += 1LL << (2*(bits-2));
2412 res += 1LL << (3*(bits-2));
2413 res <<= bits;
2414 if (res > upper_limit)
2415 res = upper_limit;
2416
2417 if (res > MAX_LFS_FILESIZE)
2418 res = MAX_LFS_FILESIZE;
2419
2420 return res;
2421 }
2422
2423 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2424 ext4_fsblk_t logical_sb_block, int nr)
2425 {
2426 struct ext4_sb_info *sbi = EXT4_SB(sb);
2427 ext4_group_t bg, first_meta_bg;
2428 int has_super = 0;
2429
2430 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2431
2432 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2433 nr < first_meta_bg)
2434 return logical_sb_block + nr + 1;
2435 bg = sbi->s_desc_per_block * nr;
2436 if (ext4_bg_has_super(sb, bg))
2437 has_super = 1;
2438
2439 /*
2440 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2441 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2442 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2443 * compensate.
2444 */
2445 if (sb->s_blocksize == 1024 && nr == 0 &&
2446 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2447 has_super++;
2448
2449 return (has_super + ext4_group_first_block_no(sb, bg));
2450 }
2451
2452 /**
2453 * ext4_get_stripe_size: Get the stripe size.
2454 * @sbi: In memory super block info
2455 *
2456 * If we have specified it via mount option, then
2457 * use the mount option value. If the value specified at mount time is
2458 * greater than the blocks per group use the super block value.
2459 * If the super block value is greater than blocks per group return 0.
2460 * Allocator needs it be less than blocks per group.
2461 *
2462 */
2463 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2464 {
2465 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2466 unsigned long stripe_width =
2467 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2468 int ret;
2469
2470 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2471 ret = sbi->s_stripe;
2472 else if (stripe_width <= sbi->s_blocks_per_group)
2473 ret = stripe_width;
2474 else if (stride <= sbi->s_blocks_per_group)
2475 ret = stride;
2476 else
2477 ret = 0;
2478
2479 /*
2480 * If the stripe width is 1, this makes no sense and
2481 * we set it to 0 to turn off stripe handling code.
2482 */
2483 if (ret <= 1)
2484 ret = 0;
2485
2486 return ret;
2487 }
2488
2489 /* sysfs supprt */
2490
2491 struct ext4_attr {
2492 struct attribute attr;
2493 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2494 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2495 const char *, size_t);
2496 union {
2497 int offset;
2498 int deprecated_val;
2499 } u;
2500 };
2501
2502 static int parse_strtoull(const char *buf,
2503 unsigned long long max, unsigned long long *value)
2504 {
2505 int ret;
2506
2507 ret = kstrtoull(skip_spaces(buf), 0, value);
2508 if (!ret && *value > max)
2509 ret = -EINVAL;
2510 return ret;
2511 }
2512
2513 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2514 struct ext4_sb_info *sbi,
2515 char *buf)
2516 {
2517 return snprintf(buf, PAGE_SIZE, "%llu\n",
2518 (s64) EXT4_C2B(sbi,
2519 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2520 }
2521
2522 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2523 struct ext4_sb_info *sbi, char *buf)
2524 {
2525 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2526
2527 if (!sb->s_bdev->bd_part)
2528 return snprintf(buf, PAGE_SIZE, "0\n");
2529 return snprintf(buf, PAGE_SIZE, "%lu\n",
2530 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2531 sbi->s_sectors_written_start) >> 1);
2532 }
2533
2534 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2535 struct ext4_sb_info *sbi, char *buf)
2536 {
2537 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2538
2539 if (!sb->s_bdev->bd_part)
2540 return snprintf(buf, PAGE_SIZE, "0\n");
2541 return snprintf(buf, PAGE_SIZE, "%llu\n",
2542 (unsigned long long)(sbi->s_kbytes_written +
2543 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2544 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2545 }
2546
2547 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2548 struct ext4_sb_info *sbi,
2549 const char *buf, size_t count)
2550 {
2551 unsigned long t;
2552 int ret;
2553
2554 ret = kstrtoul(skip_spaces(buf), 0, &t);
2555 if (ret)
2556 return ret;
2557
2558 if (t && (!is_power_of_2(t) || t > 0x40000000))
2559 return -EINVAL;
2560
2561 sbi->s_inode_readahead_blks = t;
2562 return count;
2563 }
2564
2565 static ssize_t sbi_ui_show(struct ext4_attr *a,
2566 struct ext4_sb_info *sbi, char *buf)
2567 {
2568 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2569
2570 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2571 }
2572
2573 static ssize_t sbi_ui_store(struct ext4_attr *a,
2574 struct ext4_sb_info *sbi,
2575 const char *buf, size_t count)
2576 {
2577 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2578 unsigned long t;
2579 int ret;
2580
2581 ret = kstrtoul(skip_spaces(buf), 0, &t);
2582 if (ret)
2583 return ret;
2584 *ui = t;
2585 return count;
2586 }
2587
2588 static ssize_t es_ui_show(struct ext4_attr *a,
2589 struct ext4_sb_info *sbi, char *buf)
2590 {
2591
2592 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2593 a->u.offset);
2594
2595 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2596 }
2597
2598 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2599 struct ext4_sb_info *sbi, char *buf)
2600 {
2601 return snprintf(buf, PAGE_SIZE, "%llu\n",
2602 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2603 }
2604
2605 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2606 struct ext4_sb_info *sbi,
2607 const char *buf, size_t count)
2608 {
2609 unsigned long long val;
2610 int ret;
2611
2612 if (parse_strtoull(buf, -1ULL, &val))
2613 return -EINVAL;
2614 ret = ext4_reserve_clusters(sbi, val);
2615
2616 return ret ? ret : count;
2617 }
2618
2619 static ssize_t trigger_test_error(struct ext4_attr *a,
2620 struct ext4_sb_info *sbi,
2621 const char *buf, size_t count)
2622 {
2623 int len = count;
2624
2625 if (!capable(CAP_SYS_ADMIN))
2626 return -EPERM;
2627
2628 if (len && buf[len-1] == '\n')
2629 len--;
2630
2631 if (len)
2632 ext4_error(sbi->s_sb, "%.*s", len, buf);
2633 return count;
2634 }
2635
2636 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2637 struct ext4_sb_info *sbi, char *buf)
2638 {
2639 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2640 }
2641
2642 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2643 static struct ext4_attr ext4_attr_##_name = { \
2644 .attr = {.name = __stringify(_name), .mode = _mode }, \
2645 .show = _show, \
2646 .store = _store, \
2647 .u = { \
2648 .offset = offsetof(struct ext4_sb_info, _elname),\
2649 }, \
2650 }
2651
2652 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2653 static struct ext4_attr ext4_attr_##_name = { \
2654 .attr = {.name = __stringify(_name), .mode = _mode }, \
2655 .show = _show, \
2656 .store = _store, \
2657 .u = { \
2658 .offset = offsetof(struct ext4_super_block, _elname), \
2659 }, \
2660 }
2661
2662 #define EXT4_ATTR(name, mode, show, store) \
2663 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2664
2665 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2666 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2667 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2668
2669 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2670 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2671 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2672 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2673
2674 #define ATTR_LIST(name) &ext4_attr_##name.attr
2675 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2676 static struct ext4_attr ext4_attr_##_name = { \
2677 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2678 .show = sbi_deprecated_show, \
2679 .u = { \
2680 .deprecated_val = _val, \
2681 }, \
2682 }
2683
2684 EXT4_RO_ATTR(delayed_allocation_blocks);
2685 EXT4_RO_ATTR(session_write_kbytes);
2686 EXT4_RO_ATTR(lifetime_write_kbytes);
2687 EXT4_RW_ATTR(reserved_clusters);
2688 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2689 inode_readahead_blks_store, s_inode_readahead_blks);
2690 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2691 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2692 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2693 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2694 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2695 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2696 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2697 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2698 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2699 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2700 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2701 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2702 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2703 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2704 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2705 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2706 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2707 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2708 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2709
2710 static struct attribute *ext4_attrs[] = {
2711 ATTR_LIST(delayed_allocation_blocks),
2712 ATTR_LIST(session_write_kbytes),
2713 ATTR_LIST(lifetime_write_kbytes),
2714 ATTR_LIST(reserved_clusters),
2715 ATTR_LIST(inode_readahead_blks),
2716 ATTR_LIST(inode_goal),
2717 ATTR_LIST(mb_stats),
2718 ATTR_LIST(mb_max_to_scan),
2719 ATTR_LIST(mb_min_to_scan),
2720 ATTR_LIST(mb_order2_req),
2721 ATTR_LIST(mb_stream_req),
2722 ATTR_LIST(mb_group_prealloc),
2723 ATTR_LIST(max_writeback_mb_bump),
2724 ATTR_LIST(extent_max_zeroout_kb),
2725 ATTR_LIST(trigger_fs_error),
2726 ATTR_LIST(err_ratelimit_interval_ms),
2727 ATTR_LIST(err_ratelimit_burst),
2728 ATTR_LIST(warning_ratelimit_interval_ms),
2729 ATTR_LIST(warning_ratelimit_burst),
2730 ATTR_LIST(msg_ratelimit_interval_ms),
2731 ATTR_LIST(msg_ratelimit_burst),
2732 ATTR_LIST(errors_count),
2733 ATTR_LIST(first_error_time),
2734 ATTR_LIST(last_error_time),
2735 NULL,
2736 };
2737
2738 /* Features this copy of ext4 supports */
2739 EXT4_INFO_ATTR(lazy_itable_init);
2740 EXT4_INFO_ATTR(batched_discard);
2741 EXT4_INFO_ATTR(meta_bg_resize);
2742 EXT4_INFO_ATTR(encryption);
2743
2744 static struct attribute *ext4_feat_attrs[] = {
2745 ATTR_LIST(lazy_itable_init),
2746 ATTR_LIST(batched_discard),
2747 ATTR_LIST(meta_bg_resize),
2748 ATTR_LIST(encryption),
2749 NULL,
2750 };
2751
2752 static ssize_t ext4_attr_show(struct kobject *kobj,
2753 struct attribute *attr, char *buf)
2754 {
2755 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2756 s_kobj);
2757 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2758
2759 return a->show ? a->show(a, sbi, buf) : 0;
2760 }
2761
2762 static ssize_t ext4_attr_store(struct kobject *kobj,
2763 struct attribute *attr,
2764 const char *buf, size_t len)
2765 {
2766 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2767 s_kobj);
2768 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2769
2770 return a->store ? a->store(a, sbi, buf, len) : 0;
2771 }
2772
2773 static void ext4_sb_release(struct kobject *kobj)
2774 {
2775 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2776 s_kobj);
2777 complete(&sbi->s_kobj_unregister);
2778 }
2779
2780 static const struct sysfs_ops ext4_attr_ops = {
2781 .show = ext4_attr_show,
2782 .store = ext4_attr_store,
2783 };
2784
2785 static struct kobj_type ext4_ktype = {
2786 .default_attrs = ext4_attrs,
2787 .sysfs_ops = &ext4_attr_ops,
2788 .release = ext4_sb_release,
2789 };
2790
2791 static void ext4_feat_release(struct kobject *kobj)
2792 {
2793 complete(&ext4_feat->f_kobj_unregister);
2794 }
2795
2796 static ssize_t ext4_feat_show(struct kobject *kobj,
2797 struct attribute *attr, char *buf)
2798 {
2799 return snprintf(buf, PAGE_SIZE, "supported\n");
2800 }
2801
2802 /*
2803 * We can not use ext4_attr_show/store because it relies on the kobject
2804 * being embedded in the ext4_sb_info structure which is definitely not
2805 * true in this case.
2806 */
2807 static const struct sysfs_ops ext4_feat_ops = {
2808 .show = ext4_feat_show,
2809 .store = NULL,
2810 };
2811
2812 static struct kobj_type ext4_feat_ktype = {
2813 .default_attrs = ext4_feat_attrs,
2814 .sysfs_ops = &ext4_feat_ops,
2815 .release = ext4_feat_release,
2816 };
2817
2818 /*
2819 * Check whether this filesystem can be mounted based on
2820 * the features present and the RDONLY/RDWR mount requested.
2821 * Returns 1 if this filesystem can be mounted as requested,
2822 * 0 if it cannot be.
2823 */
2824 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2825 {
2826 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2827 ext4_msg(sb, KERN_ERR,
2828 "Couldn't mount because of "
2829 "unsupported optional features (%x)",
2830 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2831 ~EXT4_FEATURE_INCOMPAT_SUPP));
2832 return 0;
2833 }
2834
2835 if (readonly)
2836 return 1;
2837
2838 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2839 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2840 sb->s_flags |= MS_RDONLY;
2841 return 1;
2842 }
2843
2844 /* Check that feature set is OK for a read-write mount */
2845 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2846 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2847 "unsupported optional features (%x)",
2848 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2849 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2850 return 0;
2851 }
2852 /*
2853 * Large file size enabled file system can only be mounted
2854 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2855 */
2856 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2857 if (sizeof(blkcnt_t) < sizeof(u64)) {
2858 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2859 "cannot be mounted RDWR without "
2860 "CONFIG_LBDAF");
2861 return 0;
2862 }
2863 }
2864 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2865 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2866 ext4_msg(sb, KERN_ERR,
2867 "Can't support bigalloc feature without "
2868 "extents feature\n");
2869 return 0;
2870 }
2871
2872 #ifndef CONFIG_QUOTA
2873 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2874 !readonly) {
2875 ext4_msg(sb, KERN_ERR,
2876 "Filesystem with quota feature cannot be mounted RDWR "
2877 "without CONFIG_QUOTA");
2878 return 0;
2879 }
2880 #endif /* CONFIG_QUOTA */
2881 return 1;
2882 }
2883
2884 /*
2885 * This function is called once a day if we have errors logged
2886 * on the file system
2887 */
2888 static void print_daily_error_info(unsigned long arg)
2889 {
2890 struct super_block *sb = (struct super_block *) arg;
2891 struct ext4_sb_info *sbi;
2892 struct ext4_super_block *es;
2893
2894 sbi = EXT4_SB(sb);
2895 es = sbi->s_es;
2896
2897 if (es->s_error_count)
2898 /* fsck newer than v1.41.13 is needed to clean this condition. */
2899 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2900 le32_to_cpu(es->s_error_count));
2901 if (es->s_first_error_time) {
2902 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2903 sb->s_id, le32_to_cpu(es->s_first_error_time),
2904 (int) sizeof(es->s_first_error_func),
2905 es->s_first_error_func,
2906 le32_to_cpu(es->s_first_error_line));
2907 if (es->s_first_error_ino)
2908 printk(": inode %u",
2909 le32_to_cpu(es->s_first_error_ino));
2910 if (es->s_first_error_block)
2911 printk(": block %llu", (unsigned long long)
2912 le64_to_cpu(es->s_first_error_block));
2913 printk("\n");
2914 }
2915 if (es->s_last_error_time) {
2916 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2917 sb->s_id, le32_to_cpu(es->s_last_error_time),
2918 (int) sizeof(es->s_last_error_func),
2919 es->s_last_error_func,
2920 le32_to_cpu(es->s_last_error_line));
2921 if (es->s_last_error_ino)
2922 printk(": inode %u",
2923 le32_to_cpu(es->s_last_error_ino));
2924 if (es->s_last_error_block)
2925 printk(": block %llu", (unsigned long long)
2926 le64_to_cpu(es->s_last_error_block));
2927 printk("\n");
2928 }
2929 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2930 }
2931
2932 /* Find next suitable group and run ext4_init_inode_table */
2933 static int ext4_run_li_request(struct ext4_li_request *elr)
2934 {
2935 struct ext4_group_desc *gdp = NULL;
2936 ext4_group_t group, ngroups;
2937 struct super_block *sb;
2938 unsigned long timeout = 0;
2939 int ret = 0;
2940
2941 sb = elr->lr_super;
2942 ngroups = EXT4_SB(sb)->s_groups_count;
2943
2944 sb_start_write(sb);
2945 for (group = elr->lr_next_group; group < ngroups; group++) {
2946 gdp = ext4_get_group_desc(sb, group, NULL);
2947 if (!gdp) {
2948 ret = 1;
2949 break;
2950 }
2951
2952 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2953 break;
2954 }
2955
2956 if (group >= ngroups)
2957 ret = 1;
2958
2959 if (!ret) {
2960 timeout = jiffies;
2961 ret = ext4_init_inode_table(sb, group,
2962 elr->lr_timeout ? 0 : 1);
2963 if (elr->lr_timeout == 0) {
2964 timeout = (jiffies - timeout) *
2965 elr->lr_sbi->s_li_wait_mult;
2966 elr->lr_timeout = timeout;
2967 }
2968 elr->lr_next_sched = jiffies + elr->lr_timeout;
2969 elr->lr_next_group = group + 1;
2970 }
2971 sb_end_write(sb);
2972
2973 return ret;
2974 }
2975
2976 /*
2977 * Remove lr_request from the list_request and free the
2978 * request structure. Should be called with li_list_mtx held
2979 */
2980 static void ext4_remove_li_request(struct ext4_li_request *elr)
2981 {
2982 struct ext4_sb_info *sbi;
2983
2984 if (!elr)
2985 return;
2986
2987 sbi = elr->lr_sbi;
2988
2989 list_del(&elr->lr_request);
2990 sbi->s_li_request = NULL;
2991 kfree(elr);
2992 }
2993
2994 static void ext4_unregister_li_request(struct super_block *sb)
2995 {
2996 mutex_lock(&ext4_li_mtx);
2997 if (!ext4_li_info) {
2998 mutex_unlock(&ext4_li_mtx);
2999 return;
3000 }
3001
3002 mutex_lock(&ext4_li_info->li_list_mtx);
3003 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3004 mutex_unlock(&ext4_li_info->li_list_mtx);
3005 mutex_unlock(&ext4_li_mtx);
3006 }
3007
3008 static struct task_struct *ext4_lazyinit_task;
3009
3010 /*
3011 * This is the function where ext4lazyinit thread lives. It walks
3012 * through the request list searching for next scheduled filesystem.
3013 * When such a fs is found, run the lazy initialization request
3014 * (ext4_rn_li_request) and keep track of the time spend in this
3015 * function. Based on that time we compute next schedule time of
3016 * the request. When walking through the list is complete, compute
3017 * next waking time and put itself into sleep.
3018 */
3019 static int ext4_lazyinit_thread(void *arg)
3020 {
3021 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3022 struct list_head *pos, *n;
3023 struct ext4_li_request *elr;
3024 unsigned long next_wakeup, cur;
3025
3026 BUG_ON(NULL == eli);
3027
3028 cont_thread:
3029 while (true) {
3030 next_wakeup = MAX_JIFFY_OFFSET;
3031
3032 mutex_lock(&eli->li_list_mtx);
3033 if (list_empty(&eli->li_request_list)) {
3034 mutex_unlock(&eli->li_list_mtx);
3035 goto exit_thread;
3036 }
3037
3038 list_for_each_safe(pos, n, &eli->li_request_list) {
3039 elr = list_entry(pos, struct ext4_li_request,
3040 lr_request);
3041
3042 if (time_after_eq(jiffies, elr->lr_next_sched)) {
3043 if (ext4_run_li_request(elr) != 0) {
3044 /* error, remove the lazy_init job */
3045 ext4_remove_li_request(elr);
3046 continue;
3047 }
3048 }
3049
3050 if (time_before(elr->lr_next_sched, next_wakeup))
3051 next_wakeup = elr->lr_next_sched;
3052 }
3053 mutex_unlock(&eli->li_list_mtx);
3054
3055 try_to_freeze();
3056
3057 cur = jiffies;
3058 if ((time_after_eq(cur, next_wakeup)) ||
3059 (MAX_JIFFY_OFFSET == next_wakeup)) {
3060 cond_resched();
3061 continue;
3062 }
3063
3064 schedule_timeout_interruptible(next_wakeup - cur);
3065
3066 if (kthread_should_stop()) {
3067 ext4_clear_request_list();
3068 goto exit_thread;
3069 }
3070 }
3071
3072 exit_thread:
3073 /*
3074 * It looks like the request list is empty, but we need
3075 * to check it under the li_list_mtx lock, to prevent any
3076 * additions into it, and of course we should lock ext4_li_mtx
3077 * to atomically free the list and ext4_li_info, because at
3078 * this point another ext4 filesystem could be registering
3079 * new one.
3080 */
3081 mutex_lock(&ext4_li_mtx);
3082 mutex_lock(&eli->li_list_mtx);
3083 if (!list_empty(&eli->li_request_list)) {
3084 mutex_unlock(&eli->li_list_mtx);
3085 mutex_unlock(&ext4_li_mtx);
3086 goto cont_thread;
3087 }
3088 mutex_unlock(&eli->li_list_mtx);
3089 kfree(ext4_li_info);
3090 ext4_li_info = NULL;
3091 mutex_unlock(&ext4_li_mtx);
3092
3093 return 0;
3094 }
3095
3096 static void ext4_clear_request_list(void)
3097 {
3098 struct list_head *pos, *n;
3099 struct ext4_li_request *elr;
3100
3101 mutex_lock(&ext4_li_info->li_list_mtx);
3102 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3103 elr = list_entry(pos, struct ext4_li_request,
3104 lr_request);
3105 ext4_remove_li_request(elr);
3106 }
3107 mutex_unlock(&ext4_li_info->li_list_mtx);
3108 }
3109
3110 static int ext4_run_lazyinit_thread(void)
3111 {
3112 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3113 ext4_li_info, "ext4lazyinit");
3114 if (IS_ERR(ext4_lazyinit_task)) {
3115 int err = PTR_ERR(ext4_lazyinit_task);
3116 ext4_clear_request_list();
3117 kfree(ext4_li_info);
3118 ext4_li_info = NULL;
3119 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3120 "initialization thread\n",
3121 err);
3122 return err;
3123 }
3124 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3125 return 0;
3126 }
3127
3128 /*
3129 * Check whether it make sense to run itable init. thread or not.
3130 * If there is at least one uninitialized inode table, return
3131 * corresponding group number, else the loop goes through all
3132 * groups and return total number of groups.
3133 */
3134 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3135 {
3136 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3137 struct ext4_group_desc *gdp = NULL;
3138
3139 for (group = 0; group < ngroups; group++) {
3140 gdp = ext4_get_group_desc(sb, group, NULL);
3141 if (!gdp)
3142 continue;
3143
3144 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3145 break;
3146 }
3147
3148 return group;
3149 }
3150
3151 static int ext4_li_info_new(void)
3152 {
3153 struct ext4_lazy_init *eli = NULL;
3154
3155 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3156 if (!eli)
3157 return -ENOMEM;
3158
3159 INIT_LIST_HEAD(&eli->li_request_list);
3160 mutex_init(&eli->li_list_mtx);
3161
3162 eli->li_state |= EXT4_LAZYINIT_QUIT;
3163
3164 ext4_li_info = eli;
3165
3166 return 0;
3167 }
3168
3169 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3170 ext4_group_t start)
3171 {
3172 struct ext4_sb_info *sbi = EXT4_SB(sb);
3173 struct ext4_li_request *elr;
3174
3175 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3176 if (!elr)
3177 return NULL;
3178
3179 elr->lr_super = sb;
3180 elr->lr_sbi = sbi;
3181 elr->lr_next_group = start;
3182
3183 /*
3184 * Randomize first schedule time of the request to
3185 * spread the inode table initialization requests
3186 * better.
3187 */
3188 elr->lr_next_sched = jiffies + (prandom_u32() %
3189 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3190 return elr;
3191 }
3192
3193 int ext4_register_li_request(struct super_block *sb,
3194 ext4_group_t first_not_zeroed)
3195 {
3196 struct ext4_sb_info *sbi = EXT4_SB(sb);
3197 struct ext4_li_request *elr = NULL;
3198 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3199 int ret = 0;
3200
3201 mutex_lock(&ext4_li_mtx);
3202 if (sbi->s_li_request != NULL) {
3203 /*
3204 * Reset timeout so it can be computed again, because
3205 * s_li_wait_mult might have changed.
3206 */
3207 sbi->s_li_request->lr_timeout = 0;
3208 goto out;
3209 }
3210
3211 if (first_not_zeroed == ngroups ||
3212 (sb->s_flags & MS_RDONLY) ||
3213 !test_opt(sb, INIT_INODE_TABLE))
3214 goto out;
3215
3216 elr = ext4_li_request_new(sb, first_not_zeroed);
3217 if (!elr) {
3218 ret = -ENOMEM;
3219 goto out;
3220 }
3221
3222 if (NULL == ext4_li_info) {
3223 ret = ext4_li_info_new();
3224 if (ret)
3225 goto out;
3226 }
3227
3228 mutex_lock(&ext4_li_info->li_list_mtx);
3229 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3230 mutex_unlock(&ext4_li_info->li_list_mtx);
3231
3232 sbi->s_li_request = elr;
3233 /*
3234 * set elr to NULL here since it has been inserted to
3235 * the request_list and the removal and free of it is
3236 * handled by ext4_clear_request_list from now on.
3237 */
3238 elr = NULL;
3239
3240 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3241 ret = ext4_run_lazyinit_thread();
3242 if (ret)
3243 goto out;
3244 }
3245 out:
3246 mutex_unlock(&ext4_li_mtx);
3247 if (ret)
3248 kfree(elr);
3249 return ret;
3250 }
3251
3252 /*
3253 * We do not need to lock anything since this is called on
3254 * module unload.
3255 */
3256 static void ext4_destroy_lazyinit_thread(void)
3257 {
3258 /*
3259 * If thread exited earlier
3260 * there's nothing to be done.
3261 */
3262 if (!ext4_li_info || !ext4_lazyinit_task)
3263 return;
3264
3265 kthread_stop(ext4_lazyinit_task);
3266 }
3267
3268 static int set_journal_csum_feature_set(struct super_block *sb)
3269 {
3270 int ret = 1;
3271 int compat, incompat;
3272 struct ext4_sb_info *sbi = EXT4_SB(sb);
3273
3274 if (ext4_has_metadata_csum(sb)) {
3275 /* journal checksum v3 */
3276 compat = 0;
3277 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3278 } else {
3279 /* journal checksum v1 */
3280 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3281 incompat = 0;
3282 }
3283
3284 jbd2_journal_clear_features(sbi->s_journal,
3285 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3286 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3287 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3288 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3289 ret = jbd2_journal_set_features(sbi->s_journal,
3290 compat, 0,
3291 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3292 incompat);
3293 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3294 ret = jbd2_journal_set_features(sbi->s_journal,
3295 compat, 0,
3296 incompat);
3297 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3298 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3299 } else {
3300 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3301 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3302 }
3303
3304 return ret;
3305 }
3306
3307 /*
3308 * Note: calculating the overhead so we can be compatible with
3309 * historical BSD practice is quite difficult in the face of
3310 * clusters/bigalloc. This is because multiple metadata blocks from
3311 * different block group can end up in the same allocation cluster.
3312 * Calculating the exact overhead in the face of clustered allocation
3313 * requires either O(all block bitmaps) in memory or O(number of block
3314 * groups**2) in time. We will still calculate the superblock for
3315 * older file systems --- and if we come across with a bigalloc file
3316 * system with zero in s_overhead_clusters the estimate will be close to
3317 * correct especially for very large cluster sizes --- but for newer
3318 * file systems, it's better to calculate this figure once at mkfs
3319 * time, and store it in the superblock. If the superblock value is
3320 * present (even for non-bigalloc file systems), we will use it.
3321 */
3322 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3323 char *buf)
3324 {
3325 struct ext4_sb_info *sbi = EXT4_SB(sb);
3326 struct ext4_group_desc *gdp;
3327 ext4_fsblk_t first_block, last_block, b;
3328 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3329 int s, j, count = 0;
3330
3331 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3332 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3333 sbi->s_itb_per_group + 2);
3334
3335 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3336 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3337 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3338 for (i = 0; i < ngroups; i++) {
3339 gdp = ext4_get_group_desc(sb, i, NULL);
3340 b = ext4_block_bitmap(sb, gdp);
3341 if (b >= first_block && b <= last_block) {
3342 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3343 count++;
3344 }
3345 b = ext4_inode_bitmap(sb, gdp);
3346 if (b >= first_block && b <= last_block) {
3347 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3348 count++;
3349 }
3350 b = ext4_inode_table(sb, gdp);
3351 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3352 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3353 int c = EXT4_B2C(sbi, b - first_block);
3354 ext4_set_bit(c, buf);
3355 count++;
3356 }
3357 if (i != grp)
3358 continue;
3359 s = 0;
3360 if (ext4_bg_has_super(sb, grp)) {
3361 ext4_set_bit(s++, buf);
3362 count++;
3363 }
3364 j = ext4_bg_num_gdb(sb, grp);
3365 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3366 ext4_error(sb, "Invalid number of block group "
3367 "descriptor blocks: %d", j);
3368 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3369 }
3370 count += j;
3371 for (; j > 0; j--)
3372 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3373 }
3374 if (!count)
3375 return 0;
3376 return EXT4_CLUSTERS_PER_GROUP(sb) -
3377 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3378 }
3379
3380 /*
3381 * Compute the overhead and stash it in sbi->s_overhead
3382 */
3383 int ext4_calculate_overhead(struct super_block *sb)
3384 {
3385 struct ext4_sb_info *sbi = EXT4_SB(sb);
3386 struct ext4_super_block *es = sbi->s_es;
3387 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3388 ext4_fsblk_t overhead = 0;
3389 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3390
3391 if (!buf)
3392 return -ENOMEM;
3393
3394 /*
3395 * Compute the overhead (FS structures). This is constant
3396 * for a given filesystem unless the number of block groups
3397 * changes so we cache the previous value until it does.
3398 */
3399
3400 /*
3401 * All of the blocks before first_data_block are overhead
3402 */
3403 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3404
3405 /*
3406 * Add the overhead found in each block group
3407 */
3408 for (i = 0; i < ngroups; i++) {
3409 int blks;
3410
3411 blks = count_overhead(sb, i, buf);
3412 overhead += blks;
3413 if (blks)
3414 memset(buf, 0, PAGE_SIZE);
3415 cond_resched();
3416 }
3417 /* Add the internal journal blocks as well */
3418 if (sbi->s_journal && !sbi->journal_bdev)
3419 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3420
3421 sbi->s_overhead = overhead;
3422 smp_wmb();
3423 free_page((unsigned long) buf);
3424 return 0;
3425 }
3426
3427
3428 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3429 {
3430 ext4_fsblk_t resv_clusters;
3431
3432 /*
3433 * There's no need to reserve anything when we aren't using extents.
3434 * The space estimates are exact, there are no unwritten extents,
3435 * hole punching doesn't need new metadata... This is needed especially
3436 * to keep ext2/3 backward compatibility.
3437 */
3438 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3439 return 0;
3440 /*
3441 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3442 * This should cover the situations where we can not afford to run
3443 * out of space like for example punch hole, or converting
3444 * unwritten extents in delalloc path. In most cases such
3445 * allocation would require 1, or 2 blocks, higher numbers are
3446 * very rare.
3447 */
3448 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3449 EXT4_SB(sb)->s_cluster_bits;
3450
3451 do_div(resv_clusters, 50);
3452 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3453
3454 return resv_clusters;
3455 }
3456
3457
3458 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3459 {
3460 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3461 sbi->s_cluster_bits;
3462
3463 if (count >= clusters)
3464 return -EINVAL;
3465
3466 atomic64_set(&sbi->s_resv_clusters, count);
3467 return 0;
3468 }
3469
3470 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3471 {
3472 char *orig_data = kstrdup(data, GFP_KERNEL);
3473 struct buffer_head *bh;
3474 struct ext4_super_block *es = NULL;
3475 struct ext4_sb_info *sbi;
3476 ext4_fsblk_t block;
3477 ext4_fsblk_t sb_block = get_sb_block(&data);
3478 ext4_fsblk_t logical_sb_block;
3479 unsigned long offset = 0;
3480 unsigned long journal_devnum = 0;
3481 unsigned long def_mount_opts;
3482 struct inode *root;
3483 char *cp;
3484 const char *descr;
3485 int ret = -ENOMEM;
3486 int blocksize, clustersize;
3487 unsigned int db_count;
3488 unsigned int i;
3489 int needs_recovery, has_huge_files, has_bigalloc;
3490 __u64 blocks_count;
3491 int err = 0;
3492 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3493 ext4_group_t first_not_zeroed;
3494
3495 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3496 if (!sbi)
3497 goto out_free_orig;
3498
3499 sbi->s_blockgroup_lock =
3500 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3501 if (!sbi->s_blockgroup_lock) {
3502 kfree(sbi);
3503 goto out_free_orig;
3504 }
3505 sb->s_fs_info = sbi;
3506 sbi->s_sb = sb;
3507 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3508 sbi->s_sb_block = sb_block;
3509 if (sb->s_bdev->bd_part)
3510 sbi->s_sectors_written_start =
3511 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3512 #ifdef CONFIG_EXT4_FS_ENCRYPTION
3513 /* Modes of operations for file and directory encryption. */
3514 sbi->s_file_encryption_mode = EXT4_ENCRYPTION_MODE_AES_256_XTS;
3515 sbi->s_dir_encryption_mode = EXT4_ENCRYPTION_MODE_INVALID;
3516 #endif
3517
3518 /* Cleanup superblock name */
3519 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3520 *cp = '!';
3521
3522 /* -EINVAL is default */
3523 ret = -EINVAL;
3524 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3525 if (!blocksize) {
3526 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3527 goto out_fail;
3528 }
3529
3530 /*
3531 * The ext4 superblock will not be buffer aligned for other than 1kB
3532 * block sizes. We need to calculate the offset from buffer start.
3533 */
3534 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3535 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3536 offset = do_div(logical_sb_block, blocksize);
3537 } else {
3538 logical_sb_block = sb_block;
3539 }
3540
3541 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3542 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3543 goto out_fail;
3544 }
3545 /*
3546 * Note: s_es must be initialized as soon as possible because
3547 * some ext4 macro-instructions depend on its value
3548 */
3549 es = (struct ext4_super_block *) (bh->b_data + offset);
3550 sbi->s_es = es;
3551 sb->s_magic = le16_to_cpu(es->s_magic);
3552 if (sb->s_magic != EXT4_SUPER_MAGIC)
3553 goto cantfind_ext4;
3554 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3555
3556 /* Warn if metadata_csum and gdt_csum are both set. */
3557 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3558 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3559 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3560 ext4_warning(sb, "metadata_csum and uninit_bg are "
3561 "redundant flags; please run fsck.");
3562
3563 /* Check for a known checksum algorithm */
3564 if (!ext4_verify_csum_type(sb, es)) {
3565 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3566 "unknown checksum algorithm.");
3567 silent = 1;
3568 goto cantfind_ext4;
3569 }
3570
3571 /* Load the checksum driver */
3572 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3573 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3574 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3575 if (IS_ERR(sbi->s_chksum_driver)) {
3576 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3577 ret = PTR_ERR(sbi->s_chksum_driver);
3578 sbi->s_chksum_driver = NULL;
3579 goto failed_mount;
3580 }
3581 }
3582
3583 /* Check superblock checksum */
3584 if (!ext4_superblock_csum_verify(sb, es)) {
3585 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3586 "invalid superblock checksum. Run e2fsck?");
3587 silent = 1;
3588 goto cantfind_ext4;
3589 }
3590
3591 /* Precompute checksum seed for all metadata */
3592 if (ext4_has_metadata_csum(sb))
3593 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3594 sizeof(es->s_uuid));
3595
3596 /* Set defaults before we parse the mount options */
3597 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3598 set_opt(sb, INIT_INODE_TABLE);
3599 if (def_mount_opts & EXT4_DEFM_DEBUG)
3600 set_opt(sb, DEBUG);
3601 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3602 set_opt(sb, GRPID);
3603 if (def_mount_opts & EXT4_DEFM_UID16)
3604 set_opt(sb, NO_UID32);
3605 /* xattr user namespace & acls are now defaulted on */
3606 set_opt(sb, XATTR_USER);
3607 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3608 set_opt(sb, POSIX_ACL);
3609 #endif
3610 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3611 if (ext4_has_metadata_csum(sb))
3612 set_opt(sb, JOURNAL_CHECKSUM);
3613
3614 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3615 set_opt(sb, JOURNAL_DATA);
3616 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3617 set_opt(sb, ORDERED_DATA);
3618 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3619 set_opt(sb, WRITEBACK_DATA);
3620
3621 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3622 set_opt(sb, ERRORS_PANIC);
3623 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3624 set_opt(sb, ERRORS_CONT);
3625 else
3626 set_opt(sb, ERRORS_RO);
3627 /* block_validity enabled by default; disable with noblock_validity */
3628 set_opt(sb, BLOCK_VALIDITY);
3629 if (def_mount_opts & EXT4_DEFM_DISCARD)
3630 set_opt(sb, DISCARD);
3631
3632 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3633 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3634 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3635 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3636 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3637
3638 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3639 set_opt(sb, BARRIER);
3640
3641 /*
3642 * enable delayed allocation by default
3643 * Use -o nodelalloc to turn it off
3644 */
3645 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3646 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3647 set_opt(sb, DELALLOC);
3648
3649 /*
3650 * set default s_li_wait_mult for lazyinit, for the case there is
3651 * no mount option specified.
3652 */
3653 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3654
3655 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3656 &journal_devnum, &journal_ioprio, 0)) {
3657 ext4_msg(sb, KERN_WARNING,
3658 "failed to parse options in superblock: %s",
3659 sbi->s_es->s_mount_opts);
3660 }
3661 sbi->s_def_mount_opt = sbi->s_mount_opt;
3662 if (!parse_options((char *) data, sb, &journal_devnum,
3663 &journal_ioprio, 0))
3664 goto failed_mount;
3665
3666 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3667 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3668 "with data=journal disables delayed "
3669 "allocation and O_DIRECT support!\n");
3670 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3671 ext4_msg(sb, KERN_ERR, "can't mount with "
3672 "both data=journal and delalloc");
3673 goto failed_mount;
3674 }
3675 if (test_opt(sb, DIOREAD_NOLOCK)) {
3676 ext4_msg(sb, KERN_ERR, "can't mount with "
3677 "both data=journal and dioread_nolock");
3678 goto failed_mount;
3679 }
3680 if (test_opt(sb, DAX)) {
3681 ext4_msg(sb, KERN_ERR, "can't mount with "
3682 "both data=journal and dax");
3683 goto failed_mount;
3684 }
3685 if (test_opt(sb, DELALLOC))
3686 clear_opt(sb, DELALLOC);
3687 }
3688
3689 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3690 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3691
3692 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3693 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3694 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3695 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3696 ext4_msg(sb, KERN_WARNING,
3697 "feature flags set on rev 0 fs, "
3698 "running e2fsck is recommended");
3699
3700 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3701 set_opt2(sb, HURD_COMPAT);
3702 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3703 EXT4_FEATURE_INCOMPAT_64BIT)) {
3704 ext4_msg(sb, KERN_ERR,
3705 "The Hurd can't support 64-bit file systems");
3706 goto failed_mount;
3707 }
3708 }
3709
3710 if (IS_EXT2_SB(sb)) {
3711 if (ext2_feature_set_ok(sb))
3712 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3713 "using the ext4 subsystem");
3714 else {
3715 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3716 "to feature incompatibilities");
3717 goto failed_mount;
3718 }
3719 }
3720
3721 if (IS_EXT3_SB(sb)) {
3722 if (ext3_feature_set_ok(sb))
3723 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3724 "using the ext4 subsystem");
3725 else {
3726 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3727 "to feature incompatibilities");
3728 goto failed_mount;
3729 }
3730 }
3731
3732 /*
3733 * Check feature flags regardless of the revision level, since we
3734 * previously didn't change the revision level when setting the flags,
3735 * so there is a chance incompat flags are set on a rev 0 filesystem.
3736 */
3737 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3738 goto failed_mount;
3739
3740 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3741 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3742 blocksize > EXT4_MAX_BLOCK_SIZE) {
3743 ext4_msg(sb, KERN_ERR,
3744 "Unsupported filesystem blocksize %d (%d log_block_size)",
3745 blocksize, le32_to_cpu(es->s_log_block_size));
3746 goto failed_mount;
3747 }
3748 if (le32_to_cpu(es->s_log_block_size) >
3749 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3750 ext4_msg(sb, KERN_ERR,
3751 "Invalid log block size: %u",
3752 le32_to_cpu(es->s_log_block_size));
3753 goto failed_mount;
3754 }
3755
3756 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3757 ext4_msg(sb, KERN_ERR,
3758 "Number of reserved GDT blocks insanely large: %d",
3759 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3760 goto failed_mount;
3761 }
3762
3763 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3764 if (blocksize != PAGE_SIZE) {
3765 ext4_msg(sb, KERN_ERR,
3766 "error: unsupported blocksize for dax");
3767 goto failed_mount;
3768 }
3769 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3770 ext4_msg(sb, KERN_ERR,
3771 "error: device does not support dax");
3772 goto failed_mount;
3773 }
3774 }
3775
3776 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3777 es->s_encryption_level) {
3778 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3779 es->s_encryption_level);
3780 goto failed_mount;
3781 }
3782
3783 if (sb->s_blocksize != blocksize) {
3784 /* Validate the filesystem blocksize */
3785 if (!sb_set_blocksize(sb, blocksize)) {
3786 ext4_msg(sb, KERN_ERR, "bad block size %d",
3787 blocksize);
3788 goto failed_mount;
3789 }
3790
3791 brelse(bh);
3792 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3793 offset = do_div(logical_sb_block, blocksize);
3794 bh = sb_bread_unmovable(sb, logical_sb_block);
3795 if (!bh) {
3796 ext4_msg(sb, KERN_ERR,
3797 "Can't read superblock on 2nd try");
3798 goto failed_mount;
3799 }
3800 es = (struct ext4_super_block *)(bh->b_data + offset);
3801 sbi->s_es = es;
3802 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3803 ext4_msg(sb, KERN_ERR,
3804 "Magic mismatch, very weird!");
3805 goto failed_mount;
3806 }
3807 }
3808
3809 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3810 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3811 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3812 has_huge_files);
3813 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3814
3815 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3816 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3817 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3818 } else {
3819 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3820 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3821 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3822 (!is_power_of_2(sbi->s_inode_size)) ||
3823 (sbi->s_inode_size > blocksize)) {
3824 ext4_msg(sb, KERN_ERR,
3825 "unsupported inode size: %d",
3826 sbi->s_inode_size);
3827 goto failed_mount;
3828 }
3829 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3830 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3831 }
3832
3833 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3834 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3835 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3836 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3837 !is_power_of_2(sbi->s_desc_size)) {
3838 ext4_msg(sb, KERN_ERR,
3839 "unsupported descriptor size %lu",
3840 sbi->s_desc_size);
3841 goto failed_mount;
3842 }
3843 } else
3844 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3845
3846 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3847 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3848
3849 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3850 if (sbi->s_inodes_per_block == 0)
3851 goto cantfind_ext4;
3852 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3853 sbi->s_inodes_per_group > blocksize * 8) {
3854 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3855 sbi->s_blocks_per_group);
3856 goto failed_mount;
3857 }
3858 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3859 sbi->s_inodes_per_block;
3860 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3861 sbi->s_sbh = bh;
3862 sbi->s_mount_state = le16_to_cpu(es->s_state);
3863 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3864 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3865
3866 for (i = 0; i < 4; i++)
3867 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3868 sbi->s_def_hash_version = es->s_def_hash_version;
3869 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3870 i = le32_to_cpu(es->s_flags);
3871 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3872 sbi->s_hash_unsigned = 3;
3873 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3874 #ifdef __CHAR_UNSIGNED__
3875 if (!(sb->s_flags & MS_RDONLY))
3876 es->s_flags |=
3877 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3878 sbi->s_hash_unsigned = 3;
3879 #else
3880 if (!(sb->s_flags & MS_RDONLY))
3881 es->s_flags |=
3882 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3883 #endif
3884 }
3885 }
3886
3887 /* Handle clustersize */
3888 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3889 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3890 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3891 if (has_bigalloc) {
3892 if (clustersize < blocksize) {
3893 ext4_msg(sb, KERN_ERR,
3894 "cluster size (%d) smaller than "
3895 "block size (%d)", clustersize, blocksize);
3896 goto failed_mount;
3897 }
3898 if (le32_to_cpu(es->s_log_cluster_size) >
3899 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3900 ext4_msg(sb, KERN_ERR,
3901 "Invalid log cluster size: %u",
3902 le32_to_cpu(es->s_log_cluster_size));
3903 goto failed_mount;
3904 }
3905 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3906 le32_to_cpu(es->s_log_block_size);
3907 sbi->s_clusters_per_group =
3908 le32_to_cpu(es->s_clusters_per_group);
3909 if (sbi->s_clusters_per_group > blocksize * 8) {
3910 ext4_msg(sb, KERN_ERR,
3911 "#clusters per group too big: %lu",
3912 sbi->s_clusters_per_group);
3913 goto failed_mount;
3914 }
3915 if (sbi->s_blocks_per_group !=
3916 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3917 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3918 "clusters per group (%lu) inconsistent",
3919 sbi->s_blocks_per_group,
3920 sbi->s_clusters_per_group);
3921 goto failed_mount;
3922 }
3923 } else {
3924 if (clustersize != blocksize) {
3925 ext4_warning(sb, "fragment/cluster size (%d) != "
3926 "block size (%d)", clustersize,
3927 blocksize);
3928 clustersize = blocksize;
3929 }
3930 if (sbi->s_blocks_per_group > blocksize * 8) {
3931 ext4_msg(sb, KERN_ERR,
3932 "#blocks per group too big: %lu",
3933 sbi->s_blocks_per_group);
3934 goto failed_mount;
3935 }
3936 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3937 sbi->s_cluster_bits = 0;
3938 }
3939 sbi->s_cluster_ratio = clustersize / blocksize;
3940
3941 /* Do we have standard group size of clustersize * 8 blocks ? */
3942 if (sbi->s_blocks_per_group == clustersize << 3)
3943 set_opt2(sb, STD_GROUP_SIZE);
3944
3945 /*
3946 * Test whether we have more sectors than will fit in sector_t,
3947 * and whether the max offset is addressable by the page cache.
3948 */
3949 err = generic_check_addressable(sb->s_blocksize_bits,
3950 ext4_blocks_count(es));
3951 if (err) {
3952 ext4_msg(sb, KERN_ERR, "filesystem"
3953 " too large to mount safely on this system");
3954 if (sizeof(sector_t) < 8)
3955 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3956 goto failed_mount;
3957 }
3958
3959 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3960 goto cantfind_ext4;
3961
3962 /* check blocks count against device size */
3963 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3964 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3965 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3966 "exceeds size of device (%llu blocks)",
3967 ext4_blocks_count(es), blocks_count);
3968 goto failed_mount;
3969 }
3970
3971 /*
3972 * It makes no sense for the first data block to be beyond the end
3973 * of the filesystem.
3974 */
3975 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3976 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3977 "block %u is beyond end of filesystem (%llu)",
3978 le32_to_cpu(es->s_first_data_block),
3979 ext4_blocks_count(es));
3980 goto failed_mount;
3981 }
3982 blocks_count = (ext4_blocks_count(es) -
3983 le32_to_cpu(es->s_first_data_block) +
3984 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3985 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3986 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3987 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3988 "(block count %llu, first data block %u, "
3989 "blocks per group %lu)", sbi->s_groups_count,
3990 ext4_blocks_count(es),
3991 le32_to_cpu(es->s_first_data_block),
3992 EXT4_BLOCKS_PER_GROUP(sb));
3993 goto failed_mount;
3994 }
3995 sbi->s_groups_count = blocks_count;
3996 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3997 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3998 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3999 EXT4_DESC_PER_BLOCK(sb);
4000 sbi->s_group_desc = ext4_kvmalloc(db_count *
4001 sizeof(struct buffer_head *),
4002 GFP_KERNEL);
4003 if (sbi->s_group_desc == NULL) {
4004 ext4_msg(sb, KERN_ERR, "not enough memory");
4005 ret = -ENOMEM;
4006 goto failed_mount;
4007 }
4008
4009 if (ext4_proc_root)
4010 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
4011
4012 if (sbi->s_proc)
4013 proc_create_data("options", S_IRUGO, sbi->s_proc,
4014 &ext4_seq_options_fops, sb);
4015
4016 bgl_lock_init(sbi->s_blockgroup_lock);
4017
4018 for (i = 0; i < db_count; i++) {
4019 block = descriptor_loc(sb, logical_sb_block, i);
4020 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
4021 if (!sbi->s_group_desc[i]) {
4022 ext4_msg(sb, KERN_ERR,
4023 "can't read group descriptor %d", i);
4024 db_count = i;
4025 goto failed_mount2;
4026 }
4027 }
4028 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4029 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4030 goto failed_mount2;
4031 }
4032
4033 sbi->s_gdb_count = db_count;
4034 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
4035 spin_lock_init(&sbi->s_next_gen_lock);
4036
4037 setup_timer(&sbi->s_err_report, print_daily_error_info,
4038 (unsigned long) sb);
4039
4040 /* Register extent status tree shrinker */
4041 if (ext4_es_register_shrinker(sbi))
4042 goto failed_mount3;
4043
4044 sbi->s_stripe = ext4_get_stripe_size(sbi);
4045 sbi->s_extent_max_zeroout_kb = 32;
4046
4047 /*
4048 * set up enough so that it can read an inode
4049 */
4050 sb->s_op = &ext4_sops;
4051 sb->s_export_op = &ext4_export_ops;
4052 sb->s_xattr = ext4_xattr_handlers;
4053 #ifdef CONFIG_QUOTA
4054 sb->dq_op = &ext4_quota_operations;
4055 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4056 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4057 else
4058 sb->s_qcop = &ext4_qctl_operations;
4059 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
4060 #endif
4061 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4062
4063 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4064 mutex_init(&sbi->s_orphan_lock);
4065
4066 sb->s_root = NULL;
4067
4068 needs_recovery = (es->s_last_orphan != 0 ||
4069 EXT4_HAS_INCOMPAT_FEATURE(sb,
4070 EXT4_FEATURE_INCOMPAT_RECOVER));
4071
4072 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
4073 !(sb->s_flags & MS_RDONLY))
4074 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4075 goto failed_mount3a;
4076
4077 /*
4078 * The first inode we look at is the journal inode. Don't try
4079 * root first: it may be modified in the journal!
4080 */
4081 if (!test_opt(sb, NOLOAD) &&
4082 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4083 if (ext4_load_journal(sb, es, journal_devnum))
4084 goto failed_mount3a;
4085 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4086 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4087 ext4_msg(sb, KERN_ERR, "required journal recovery "
4088 "suppressed and not mounted read-only");
4089 goto failed_mount_wq;
4090 } else {
4091 clear_opt(sb, DATA_FLAGS);
4092 sbi->s_journal = NULL;
4093 needs_recovery = 0;
4094 goto no_journal;
4095 }
4096
4097 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
4098 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4099 JBD2_FEATURE_INCOMPAT_64BIT)) {
4100 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4101 goto failed_mount_wq;
4102 }
4103
4104 if (!set_journal_csum_feature_set(sb)) {
4105 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4106 "feature set");
4107 goto failed_mount_wq;
4108 }
4109
4110 /* We have now updated the journal if required, so we can
4111 * validate the data journaling mode. */
4112 switch (test_opt(sb, DATA_FLAGS)) {
4113 case 0:
4114 /* No mode set, assume a default based on the journal
4115 * capabilities: ORDERED_DATA if the journal can
4116 * cope, else JOURNAL_DATA
4117 */
4118 if (jbd2_journal_check_available_features
4119 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4120 set_opt(sb, ORDERED_DATA);
4121 else
4122 set_opt(sb, JOURNAL_DATA);
4123 break;
4124
4125 case EXT4_MOUNT_ORDERED_DATA:
4126 case EXT4_MOUNT_WRITEBACK_DATA:
4127 if (!jbd2_journal_check_available_features
4128 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4129 ext4_msg(sb, KERN_ERR, "Journal does not support "
4130 "requested data journaling mode");
4131 goto failed_mount_wq;
4132 }
4133 default:
4134 break;
4135 }
4136 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4137
4138 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4139
4140 no_journal:
4141 if (ext4_mballoc_ready) {
4142 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4143 if (!sbi->s_mb_cache) {
4144 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4145 goto failed_mount_wq;
4146 }
4147 }
4148
4149 if (unlikely(sbi->s_mount_flags & EXT4_MF_TEST_DUMMY_ENCRYPTION) &&
4150 !(sb->s_flags & MS_RDONLY) &&
4151 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
4152 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
4153 ext4_commit_super(sb, 1);
4154 }
4155
4156 /*
4157 * Get the # of file system overhead blocks from the
4158 * superblock if present.
4159 */
4160 if (es->s_overhead_clusters)
4161 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4162 else {
4163 err = ext4_calculate_overhead(sb);
4164 if (err)
4165 goto failed_mount_wq;
4166 }
4167
4168 /*
4169 * The maximum number of concurrent works can be high and
4170 * concurrency isn't really necessary. Limit it to 1.
4171 */
4172 EXT4_SB(sb)->rsv_conversion_wq =
4173 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4174 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4175 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4176 ret = -ENOMEM;
4177 goto failed_mount4;
4178 }
4179
4180 /*
4181 * The jbd2_journal_load will have done any necessary log recovery,
4182 * so we can safely mount the rest of the filesystem now.
4183 */
4184
4185 root = ext4_iget(sb, EXT4_ROOT_INO);
4186 if (IS_ERR(root)) {
4187 ext4_msg(sb, KERN_ERR, "get root inode failed");
4188 ret = PTR_ERR(root);
4189 root = NULL;
4190 goto failed_mount4;
4191 }
4192 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4193 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4194 iput(root);
4195 goto failed_mount4;
4196 }
4197 sb->s_root = d_make_root(root);
4198 if (!sb->s_root) {
4199 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4200 ret = -ENOMEM;
4201 goto failed_mount4;
4202 }
4203
4204 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4205 sb->s_flags |= MS_RDONLY;
4206
4207 /* determine the minimum size of new large inodes, if present */
4208 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4209 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4210 EXT4_GOOD_OLD_INODE_SIZE;
4211 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4212 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4213 if (sbi->s_want_extra_isize <
4214 le16_to_cpu(es->s_want_extra_isize))
4215 sbi->s_want_extra_isize =
4216 le16_to_cpu(es->s_want_extra_isize);
4217 if (sbi->s_want_extra_isize <
4218 le16_to_cpu(es->s_min_extra_isize))
4219 sbi->s_want_extra_isize =
4220 le16_to_cpu(es->s_min_extra_isize);
4221 }
4222 }
4223 /* Check if enough inode space is available */
4224 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4225 sbi->s_inode_size) {
4226 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4227 EXT4_GOOD_OLD_INODE_SIZE;
4228 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4229 "available");
4230 }
4231
4232 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4233 if (err) {
4234 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4235 "reserved pool", ext4_calculate_resv_clusters(sb));
4236 goto failed_mount4a;
4237 }
4238
4239 err = ext4_setup_system_zone(sb);
4240 if (err) {
4241 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4242 "zone (%d)", err);
4243 goto failed_mount4a;
4244 }
4245
4246 ext4_ext_init(sb);
4247 err = ext4_mb_init(sb);
4248 if (err) {
4249 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4250 err);
4251 goto failed_mount5;
4252 }
4253
4254 block = ext4_count_free_clusters(sb);
4255 ext4_free_blocks_count_set(sbi->s_es,
4256 EXT4_C2B(sbi, block));
4257 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4258 GFP_KERNEL);
4259 if (!err) {
4260 unsigned long freei = ext4_count_free_inodes(sb);
4261 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4262 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4263 GFP_KERNEL);
4264 }
4265 if (!err)
4266 err = percpu_counter_init(&sbi->s_dirs_counter,
4267 ext4_count_dirs(sb), GFP_KERNEL);
4268 if (!err)
4269 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4270 GFP_KERNEL);
4271 if (err) {
4272 ext4_msg(sb, KERN_ERR, "insufficient memory");
4273 goto failed_mount6;
4274 }
4275
4276 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4277 if (!ext4_fill_flex_info(sb)) {
4278 ext4_msg(sb, KERN_ERR,
4279 "unable to initialize "
4280 "flex_bg meta info!");
4281 goto failed_mount6;
4282 }
4283
4284 err = ext4_register_li_request(sb, first_not_zeroed);
4285 if (err)
4286 goto failed_mount6;
4287
4288 sbi->s_kobj.kset = ext4_kset;
4289 init_completion(&sbi->s_kobj_unregister);
4290 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4291 "%s", sb->s_id);
4292 if (err)
4293 goto failed_mount7;
4294
4295 #ifdef CONFIG_QUOTA
4296 /* Enable quota usage during mount. */
4297 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4298 !(sb->s_flags & MS_RDONLY)) {
4299 err = ext4_enable_quotas(sb);
4300 if (err)
4301 goto failed_mount8;
4302 }
4303 #endif /* CONFIG_QUOTA */
4304
4305 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4306 ext4_orphan_cleanup(sb, es);
4307 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4308 if (needs_recovery) {
4309 ext4_msg(sb, KERN_INFO, "recovery complete");
4310 ext4_mark_recovery_complete(sb, es);
4311 }
4312 if (EXT4_SB(sb)->s_journal) {
4313 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4314 descr = " journalled data mode";
4315 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4316 descr = " ordered data mode";
4317 else
4318 descr = " writeback data mode";
4319 } else
4320 descr = "out journal";
4321
4322 if (test_opt(sb, DISCARD)) {
4323 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4324 if (!blk_queue_discard(q))
4325 ext4_msg(sb, KERN_WARNING,
4326 "mounting with \"discard\" option, but "
4327 "the device does not support discard");
4328 }
4329
4330 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4331 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4332 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4333
4334 if (es->s_error_count)
4335 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4336
4337 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4338 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4339 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4340 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4341
4342 kfree(orig_data);
4343 return 0;
4344
4345 cantfind_ext4:
4346 if (!silent)
4347 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4348 goto failed_mount;
4349
4350 #ifdef CONFIG_QUOTA
4351 failed_mount8:
4352 kobject_del(&sbi->s_kobj);
4353 #endif
4354 failed_mount7:
4355 ext4_unregister_li_request(sb);
4356 failed_mount6:
4357 ext4_mb_release(sb);
4358 if (sbi->s_flex_groups)
4359 kvfree(sbi->s_flex_groups);
4360 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4361 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4362 percpu_counter_destroy(&sbi->s_dirs_counter);
4363 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4364 failed_mount5:
4365 ext4_ext_release(sb);
4366 ext4_release_system_zone(sb);
4367 failed_mount4a:
4368 dput(sb->s_root);
4369 sb->s_root = NULL;
4370 failed_mount4:
4371 ext4_msg(sb, KERN_ERR, "mount failed");
4372 if (EXT4_SB(sb)->rsv_conversion_wq)
4373 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4374 failed_mount_wq:
4375 if (sbi->s_journal) {
4376 jbd2_journal_destroy(sbi->s_journal);
4377 sbi->s_journal = NULL;
4378 }
4379 failed_mount3a:
4380 ext4_es_unregister_shrinker(sbi);
4381 failed_mount3:
4382 del_timer_sync(&sbi->s_err_report);
4383 if (sbi->s_mmp_tsk)
4384 kthread_stop(sbi->s_mmp_tsk);
4385 failed_mount2:
4386 for (i = 0; i < db_count; i++)
4387 brelse(sbi->s_group_desc[i]);
4388 kvfree(sbi->s_group_desc);
4389 failed_mount:
4390 if (sbi->s_chksum_driver)
4391 crypto_free_shash(sbi->s_chksum_driver);
4392 if (sbi->s_proc) {
4393 remove_proc_entry("options", sbi->s_proc);
4394 remove_proc_entry(sb->s_id, ext4_proc_root);
4395 }
4396 #ifdef CONFIG_QUOTA
4397 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4398 kfree(sbi->s_qf_names[i]);
4399 #endif
4400 ext4_blkdev_remove(sbi);
4401 brelse(bh);
4402 out_fail:
4403 sb->s_fs_info = NULL;
4404 kfree(sbi->s_blockgroup_lock);
4405 kfree(sbi);
4406 out_free_orig:
4407 kfree(orig_data);
4408 return err ? err : ret;
4409 }
4410
4411 /*
4412 * Setup any per-fs journal parameters now. We'll do this both on
4413 * initial mount, once the journal has been initialised but before we've
4414 * done any recovery; and again on any subsequent remount.
4415 */
4416 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4417 {
4418 struct ext4_sb_info *sbi = EXT4_SB(sb);
4419
4420 journal->j_commit_interval = sbi->s_commit_interval;
4421 journal->j_min_batch_time = sbi->s_min_batch_time;
4422 journal->j_max_batch_time = sbi->s_max_batch_time;
4423
4424 write_lock(&journal->j_state_lock);
4425 if (test_opt(sb, BARRIER))
4426 journal->j_flags |= JBD2_BARRIER;
4427 else
4428 journal->j_flags &= ~JBD2_BARRIER;
4429 if (test_opt(sb, DATA_ERR_ABORT))
4430 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4431 else
4432 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4433 write_unlock(&journal->j_state_lock);
4434 }
4435
4436 static journal_t *ext4_get_journal(struct super_block *sb,
4437 unsigned int journal_inum)
4438 {
4439 struct inode *journal_inode;
4440 journal_t *journal;
4441
4442 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4443
4444 /* First, test for the existence of a valid inode on disk. Bad
4445 * things happen if we iget() an unused inode, as the subsequent
4446 * iput() will try to delete it. */
4447
4448 journal_inode = ext4_iget(sb, journal_inum);
4449 if (IS_ERR(journal_inode)) {
4450 ext4_msg(sb, KERN_ERR, "no journal found");
4451 return NULL;
4452 }
4453 if (!journal_inode->i_nlink) {
4454 make_bad_inode(journal_inode);
4455 iput(journal_inode);
4456 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4457 return NULL;
4458 }
4459
4460 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4461 journal_inode, journal_inode->i_size);
4462 if (!S_ISREG(journal_inode->i_mode)) {
4463 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4464 iput(journal_inode);
4465 return NULL;
4466 }
4467
4468 journal = jbd2_journal_init_inode(journal_inode);
4469 if (!journal) {
4470 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4471 iput(journal_inode);
4472 return NULL;
4473 }
4474 journal->j_private = sb;
4475 ext4_init_journal_params(sb, journal);
4476 return journal;
4477 }
4478
4479 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4480 dev_t j_dev)
4481 {
4482 struct buffer_head *bh;
4483 journal_t *journal;
4484 ext4_fsblk_t start;
4485 ext4_fsblk_t len;
4486 int hblock, blocksize;
4487 ext4_fsblk_t sb_block;
4488 unsigned long offset;
4489 struct ext4_super_block *es;
4490 struct block_device *bdev;
4491
4492 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4493
4494 bdev = ext4_blkdev_get(j_dev, sb);
4495 if (bdev == NULL)
4496 return NULL;
4497
4498 blocksize = sb->s_blocksize;
4499 hblock = bdev_logical_block_size(bdev);
4500 if (blocksize < hblock) {
4501 ext4_msg(sb, KERN_ERR,
4502 "blocksize too small for journal device");
4503 goto out_bdev;
4504 }
4505
4506 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4507 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4508 set_blocksize(bdev, blocksize);
4509 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4510 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4511 "external journal");
4512 goto out_bdev;
4513 }
4514
4515 es = (struct ext4_super_block *) (bh->b_data + offset);
4516 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4517 !(le32_to_cpu(es->s_feature_incompat) &
4518 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4519 ext4_msg(sb, KERN_ERR, "external journal has "
4520 "bad superblock");
4521 brelse(bh);
4522 goto out_bdev;
4523 }
4524
4525 if ((le32_to_cpu(es->s_feature_ro_compat) &
4526 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4527 es->s_checksum != ext4_superblock_csum(sb, es)) {
4528 ext4_msg(sb, KERN_ERR, "external journal has "
4529 "corrupt superblock");
4530 brelse(bh);
4531 goto out_bdev;
4532 }
4533
4534 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4535 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4536 brelse(bh);
4537 goto out_bdev;
4538 }
4539
4540 len = ext4_blocks_count(es);
4541 start = sb_block + 1;
4542 brelse(bh); /* we're done with the superblock */
4543
4544 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4545 start, len, blocksize);
4546 if (!journal) {
4547 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4548 goto out_bdev;
4549 }
4550 journal->j_private = sb;
4551 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4552 wait_on_buffer(journal->j_sb_buffer);
4553 if (!buffer_uptodate(journal->j_sb_buffer)) {
4554 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4555 goto out_journal;
4556 }
4557 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4558 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4559 "user (unsupported) - %d",
4560 be32_to_cpu(journal->j_superblock->s_nr_users));
4561 goto out_journal;
4562 }
4563 EXT4_SB(sb)->journal_bdev = bdev;
4564 ext4_init_journal_params(sb, journal);
4565 return journal;
4566
4567 out_journal:
4568 jbd2_journal_destroy(journal);
4569 out_bdev:
4570 ext4_blkdev_put(bdev);
4571 return NULL;
4572 }
4573
4574 static int ext4_load_journal(struct super_block *sb,
4575 struct ext4_super_block *es,
4576 unsigned long journal_devnum)
4577 {
4578 journal_t *journal;
4579 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4580 dev_t journal_dev;
4581 int err = 0;
4582 int really_read_only;
4583
4584 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4585
4586 if (journal_devnum &&
4587 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4588 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4589 "numbers have changed");
4590 journal_dev = new_decode_dev(journal_devnum);
4591 } else
4592 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4593
4594 really_read_only = bdev_read_only(sb->s_bdev);
4595
4596 /*
4597 * Are we loading a blank journal or performing recovery after a
4598 * crash? For recovery, we need to check in advance whether we
4599 * can get read-write access to the device.
4600 */
4601 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4602 if (sb->s_flags & MS_RDONLY) {
4603 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4604 "required on readonly filesystem");
4605 if (really_read_only) {
4606 ext4_msg(sb, KERN_ERR, "write access "
4607 "unavailable, cannot proceed");
4608 return -EROFS;
4609 }
4610 ext4_msg(sb, KERN_INFO, "write access will "
4611 "be enabled during recovery");
4612 }
4613 }
4614
4615 if (journal_inum && journal_dev) {
4616 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4617 "and inode journals!");
4618 return -EINVAL;
4619 }
4620
4621 if (journal_inum) {
4622 if (!(journal = ext4_get_journal(sb, journal_inum)))
4623 return -EINVAL;
4624 } else {
4625 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4626 return -EINVAL;
4627 }
4628
4629 if (!(journal->j_flags & JBD2_BARRIER))
4630 ext4_msg(sb, KERN_INFO, "barriers disabled");
4631
4632 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4633 err = jbd2_journal_wipe(journal, !really_read_only);
4634 if (!err) {
4635 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4636 if (save)
4637 memcpy(save, ((char *) es) +
4638 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4639 err = jbd2_journal_load(journal);
4640 if (save)
4641 memcpy(((char *) es) + EXT4_S_ERR_START,
4642 save, EXT4_S_ERR_LEN);
4643 kfree(save);
4644 }
4645
4646 if (err) {
4647 ext4_msg(sb, KERN_ERR, "error loading journal");
4648 jbd2_journal_destroy(journal);
4649 return err;
4650 }
4651
4652 EXT4_SB(sb)->s_journal = journal;
4653 ext4_clear_journal_err(sb, es);
4654
4655 if (!really_read_only && journal_devnum &&
4656 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4657 es->s_journal_dev = cpu_to_le32(journal_devnum);
4658
4659 /* Make sure we flush the recovery flag to disk. */
4660 ext4_commit_super(sb, 1);
4661 }
4662
4663 return 0;
4664 }
4665
4666 static int ext4_commit_super(struct super_block *sb, int sync)
4667 {
4668 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4669 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4670 int error = 0;
4671
4672 if (!sbh || block_device_ejected(sb))
4673 return error;
4674 if (buffer_write_io_error(sbh)) {
4675 /*
4676 * Oh, dear. A previous attempt to write the
4677 * superblock failed. This could happen because the
4678 * USB device was yanked out. Or it could happen to
4679 * be a transient write error and maybe the block will
4680 * be remapped. Nothing we can do but to retry the
4681 * write and hope for the best.
4682 */
4683 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4684 "superblock detected");
4685 clear_buffer_write_io_error(sbh);
4686 set_buffer_uptodate(sbh);
4687 }
4688 /*
4689 * If the file system is mounted read-only, don't update the
4690 * superblock write time. This avoids updating the superblock
4691 * write time when we are mounting the root file system
4692 * read/only but we need to replay the journal; at that point,
4693 * for people who are east of GMT and who make their clock
4694 * tick in localtime for Windows bug-for-bug compatibility,
4695 * the clock is set in the future, and this will cause e2fsck
4696 * to complain and force a full file system check.
4697 */
4698 if (!(sb->s_flags & MS_RDONLY))
4699 es->s_wtime = cpu_to_le32(get_seconds());
4700 if (sb->s_bdev->bd_part)
4701 es->s_kbytes_written =
4702 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4703 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4704 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4705 else
4706 es->s_kbytes_written =
4707 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4708 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4709 ext4_free_blocks_count_set(es,
4710 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4711 &EXT4_SB(sb)->s_freeclusters_counter)));
4712 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4713 es->s_free_inodes_count =
4714 cpu_to_le32(percpu_counter_sum_positive(
4715 &EXT4_SB(sb)->s_freeinodes_counter));
4716 BUFFER_TRACE(sbh, "marking dirty");
4717 ext4_superblock_csum_set(sb);
4718 mark_buffer_dirty(sbh);
4719 if (sync) {
4720 error = sync_dirty_buffer(sbh);
4721 if (error)
4722 return error;
4723
4724 error = buffer_write_io_error(sbh);
4725 if (error) {
4726 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4727 "superblock");
4728 clear_buffer_write_io_error(sbh);
4729 set_buffer_uptodate(sbh);
4730 }
4731 }
4732 return error;
4733 }
4734
4735 /*
4736 * Have we just finished recovery? If so, and if we are mounting (or
4737 * remounting) the filesystem readonly, then we will end up with a
4738 * consistent fs on disk. Record that fact.
4739 */
4740 static void ext4_mark_recovery_complete(struct super_block *sb,
4741 struct ext4_super_block *es)
4742 {
4743 journal_t *journal = EXT4_SB(sb)->s_journal;
4744
4745 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4746 BUG_ON(journal != NULL);
4747 return;
4748 }
4749 jbd2_journal_lock_updates(journal);
4750 if (jbd2_journal_flush(journal) < 0)
4751 goto out;
4752
4753 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4754 sb->s_flags & MS_RDONLY) {
4755 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4756 ext4_commit_super(sb, 1);
4757 }
4758
4759 out:
4760 jbd2_journal_unlock_updates(journal);
4761 }
4762
4763 /*
4764 * If we are mounting (or read-write remounting) a filesystem whose journal
4765 * has recorded an error from a previous lifetime, move that error to the
4766 * main filesystem now.
4767 */
4768 static void ext4_clear_journal_err(struct super_block *sb,
4769 struct ext4_super_block *es)
4770 {
4771 journal_t *journal;
4772 int j_errno;
4773 const char *errstr;
4774
4775 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4776
4777 journal = EXT4_SB(sb)->s_journal;
4778
4779 /*
4780 * Now check for any error status which may have been recorded in the
4781 * journal by a prior ext4_error() or ext4_abort()
4782 */
4783
4784 j_errno = jbd2_journal_errno(journal);
4785 if (j_errno) {
4786 char nbuf[16];
4787
4788 errstr = ext4_decode_error(sb, j_errno, nbuf);
4789 ext4_warning(sb, "Filesystem error recorded "
4790 "from previous mount: %s", errstr);
4791 ext4_warning(sb, "Marking fs in need of filesystem check.");
4792
4793 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4794 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4795 ext4_commit_super(sb, 1);
4796
4797 jbd2_journal_clear_err(journal);
4798 jbd2_journal_update_sb_errno(journal);
4799 }
4800 }
4801
4802 /*
4803 * Force the running and committing transactions to commit,
4804 * and wait on the commit.
4805 */
4806 int ext4_force_commit(struct super_block *sb)
4807 {
4808 journal_t *journal;
4809
4810 if (sb->s_flags & MS_RDONLY)
4811 return 0;
4812
4813 journal = EXT4_SB(sb)->s_journal;
4814 return ext4_journal_force_commit(journal);
4815 }
4816
4817 static int ext4_sync_fs(struct super_block *sb, int wait)
4818 {
4819 int ret = 0;
4820 tid_t target;
4821 bool needs_barrier = false;
4822 struct ext4_sb_info *sbi = EXT4_SB(sb);
4823
4824 trace_ext4_sync_fs(sb, wait);
4825 flush_workqueue(sbi->rsv_conversion_wq);
4826 /*
4827 * Writeback quota in non-journalled quota case - journalled quota has
4828 * no dirty dquots
4829 */
4830 dquot_writeback_dquots(sb, -1);
4831 /*
4832 * Data writeback is possible w/o journal transaction, so barrier must
4833 * being sent at the end of the function. But we can skip it if
4834 * transaction_commit will do it for us.
4835 */
4836 if (sbi->s_journal) {
4837 target = jbd2_get_latest_transaction(sbi->s_journal);
4838 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4839 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4840 needs_barrier = true;
4841
4842 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4843 if (wait)
4844 ret = jbd2_log_wait_commit(sbi->s_journal,
4845 target);
4846 }
4847 } else if (wait && test_opt(sb, BARRIER))
4848 needs_barrier = true;
4849 if (needs_barrier) {
4850 int err;
4851 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4852 if (!ret)
4853 ret = err;
4854 }
4855
4856 return ret;
4857 }
4858
4859 /*
4860 * LVM calls this function before a (read-only) snapshot is created. This
4861 * gives us a chance to flush the journal completely and mark the fs clean.
4862 *
4863 * Note that only this function cannot bring a filesystem to be in a clean
4864 * state independently. It relies on upper layer to stop all data & metadata
4865 * modifications.
4866 */
4867 static int ext4_freeze(struct super_block *sb)
4868 {
4869 int error = 0;
4870 journal_t *journal;
4871
4872 if (sb->s_flags & MS_RDONLY)
4873 return 0;
4874
4875 journal = EXT4_SB(sb)->s_journal;
4876
4877 if (journal) {
4878 /* Now we set up the journal barrier. */
4879 jbd2_journal_lock_updates(journal);
4880
4881 /*
4882 * Don't clear the needs_recovery flag if we failed to
4883 * flush the journal.
4884 */
4885 error = jbd2_journal_flush(journal);
4886 if (error < 0)
4887 goto out;
4888
4889 /* Journal blocked and flushed, clear needs_recovery flag. */
4890 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4891 }
4892
4893 error = ext4_commit_super(sb, 1);
4894 out:
4895 if (journal)
4896 /* we rely on upper layer to stop further updates */
4897 jbd2_journal_unlock_updates(journal);
4898 return error;
4899 }
4900
4901 /*
4902 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4903 * flag here, even though the filesystem is not technically dirty yet.
4904 */
4905 static int ext4_unfreeze(struct super_block *sb)
4906 {
4907 if (sb->s_flags & MS_RDONLY)
4908 return 0;
4909
4910 if (EXT4_SB(sb)->s_journal) {
4911 /* Reset the needs_recovery flag before the fs is unlocked. */
4912 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4913 }
4914
4915 ext4_commit_super(sb, 1);
4916 return 0;
4917 }
4918
4919 /*
4920 * Structure to save mount options for ext4_remount's benefit
4921 */
4922 struct ext4_mount_options {
4923 unsigned long s_mount_opt;
4924 unsigned long s_mount_opt2;
4925 kuid_t s_resuid;
4926 kgid_t s_resgid;
4927 unsigned long s_commit_interval;
4928 u32 s_min_batch_time, s_max_batch_time;
4929 #ifdef CONFIG_QUOTA
4930 int s_jquota_fmt;
4931 char *s_qf_names[EXT4_MAXQUOTAS];
4932 #endif
4933 };
4934
4935 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4936 {
4937 struct ext4_super_block *es;
4938 struct ext4_sb_info *sbi = EXT4_SB(sb);
4939 unsigned long old_sb_flags;
4940 struct ext4_mount_options old_opts;
4941 int enable_quota = 0;
4942 ext4_group_t g;
4943 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4944 int err = 0;
4945 #ifdef CONFIG_QUOTA
4946 int i, j;
4947 #endif
4948 char *orig_data = kstrdup(data, GFP_KERNEL);
4949
4950 /* Store the original options */
4951 old_sb_flags = sb->s_flags;
4952 old_opts.s_mount_opt = sbi->s_mount_opt;
4953 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4954 old_opts.s_resuid = sbi->s_resuid;
4955 old_opts.s_resgid = sbi->s_resgid;
4956 old_opts.s_commit_interval = sbi->s_commit_interval;
4957 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4958 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4959 #ifdef CONFIG_QUOTA
4960 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4961 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4962 if (sbi->s_qf_names[i]) {
4963 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4964 GFP_KERNEL);
4965 if (!old_opts.s_qf_names[i]) {
4966 for (j = 0; j < i; j++)
4967 kfree(old_opts.s_qf_names[j]);
4968 kfree(orig_data);
4969 return -ENOMEM;
4970 }
4971 } else
4972 old_opts.s_qf_names[i] = NULL;
4973 #endif
4974 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4975 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4976
4977 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4978 err = -EINVAL;
4979 goto restore_opts;
4980 }
4981
4982 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4983 test_opt(sb, JOURNAL_CHECKSUM)) {
4984 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4985 "during remount not supported; ignoring");
4986 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4987 }
4988
4989 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4990 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4991 ext4_msg(sb, KERN_ERR, "can't mount with "
4992 "both data=journal and delalloc");
4993 err = -EINVAL;
4994 goto restore_opts;
4995 }
4996 if (test_opt(sb, DIOREAD_NOLOCK)) {
4997 ext4_msg(sb, KERN_ERR, "can't mount with "
4998 "both data=journal and dioread_nolock");
4999 err = -EINVAL;
5000 goto restore_opts;
5001 }
5002 if (test_opt(sb, DAX)) {
5003 ext4_msg(sb, KERN_ERR, "can't mount with "
5004 "both data=journal and dax");
5005 err = -EINVAL;
5006 goto restore_opts;
5007 }
5008 }
5009
5010 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5011 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5012 "dax flag with busy inodes while remounting");
5013 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5014 }
5015
5016 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5017 ext4_abort(sb, "Abort forced by user");
5018
5019 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
5020 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
5021
5022 es = sbi->s_es;
5023
5024 if (sbi->s_journal) {
5025 ext4_init_journal_params(sb, sbi->s_journal);
5026 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5027 }
5028
5029 if (*flags & MS_LAZYTIME)
5030 sb->s_flags |= MS_LAZYTIME;
5031
5032 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
5033 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5034 err = -EROFS;
5035 goto restore_opts;
5036 }
5037
5038 if (*flags & MS_RDONLY) {
5039 err = sync_filesystem(sb);
5040 if (err < 0)
5041 goto restore_opts;
5042 err = dquot_suspend(sb, -1);
5043 if (err < 0)
5044 goto restore_opts;
5045
5046 /*
5047 * First of all, the unconditional stuff we have to do
5048 * to disable replay of the journal when we next remount
5049 */
5050 sb->s_flags |= MS_RDONLY;
5051
5052 /*
5053 * OK, test if we are remounting a valid rw partition
5054 * readonly, and if so set the rdonly flag and then
5055 * mark the partition as valid again.
5056 */
5057 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5058 (sbi->s_mount_state & EXT4_VALID_FS))
5059 es->s_state = cpu_to_le16(sbi->s_mount_state);
5060
5061 if (sbi->s_journal)
5062 ext4_mark_recovery_complete(sb, es);
5063 } else {
5064 /* Make sure we can mount this feature set readwrite */
5065 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5066 EXT4_FEATURE_RO_COMPAT_READONLY) ||
5067 !ext4_feature_set_ok(sb, 0)) {
5068 err = -EROFS;
5069 goto restore_opts;
5070 }
5071 /*
5072 * Make sure the group descriptor checksums
5073 * are sane. If they aren't, refuse to remount r/w.
5074 */
5075 for (g = 0; g < sbi->s_groups_count; g++) {
5076 struct ext4_group_desc *gdp =
5077 ext4_get_group_desc(sb, g, NULL);
5078
5079 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5080 ext4_msg(sb, KERN_ERR,
5081 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5082 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
5083 le16_to_cpu(gdp->bg_checksum));
5084 err = -EINVAL;
5085 goto restore_opts;
5086 }
5087 }
5088
5089 /*
5090 * If we have an unprocessed orphan list hanging
5091 * around from a previously readonly bdev mount,
5092 * require a full umount/remount for now.
5093 */
5094 if (es->s_last_orphan) {
5095 ext4_msg(sb, KERN_WARNING, "Couldn't "
5096 "remount RDWR because of unprocessed "
5097 "orphan inode list. Please "
5098 "umount/remount instead");
5099 err = -EINVAL;
5100 goto restore_opts;
5101 }
5102
5103 /*
5104 * Mounting a RDONLY partition read-write, so reread
5105 * and store the current valid flag. (It may have
5106 * been changed by e2fsck since we originally mounted
5107 * the partition.)
5108 */
5109 if (sbi->s_journal)
5110 ext4_clear_journal_err(sb, es);
5111 sbi->s_mount_state = le16_to_cpu(es->s_state);
5112 if (!ext4_setup_super(sb, es, 0))
5113 sb->s_flags &= ~MS_RDONLY;
5114 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5115 EXT4_FEATURE_INCOMPAT_MMP))
5116 if (ext4_multi_mount_protect(sb,
5117 le64_to_cpu(es->s_mmp_block))) {
5118 err = -EROFS;
5119 goto restore_opts;
5120 }
5121 enable_quota = 1;
5122 }
5123 }
5124
5125 /*
5126 * Reinitialize lazy itable initialization thread based on
5127 * current settings
5128 */
5129 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5130 ext4_unregister_li_request(sb);
5131 else {
5132 ext4_group_t first_not_zeroed;
5133 first_not_zeroed = ext4_has_uninit_itable(sb);
5134 ext4_register_li_request(sb, first_not_zeroed);
5135 }
5136
5137 ext4_setup_system_zone(sb);
5138 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5139 ext4_commit_super(sb, 1);
5140
5141 #ifdef CONFIG_QUOTA
5142 /* Release old quota file names */
5143 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5144 kfree(old_opts.s_qf_names[i]);
5145 if (enable_quota) {
5146 if (sb_any_quota_suspended(sb))
5147 dquot_resume(sb, -1);
5148 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5149 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5150 err = ext4_enable_quotas(sb);
5151 if (err)
5152 goto restore_opts;
5153 }
5154 }
5155 #endif
5156
5157 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5158 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5159 kfree(orig_data);
5160 return 0;
5161
5162 restore_opts:
5163 sb->s_flags = old_sb_flags;
5164 sbi->s_mount_opt = old_opts.s_mount_opt;
5165 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5166 sbi->s_resuid = old_opts.s_resuid;
5167 sbi->s_resgid = old_opts.s_resgid;
5168 sbi->s_commit_interval = old_opts.s_commit_interval;
5169 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5170 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5171 #ifdef CONFIG_QUOTA
5172 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5173 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5174 kfree(sbi->s_qf_names[i]);
5175 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5176 }
5177 #endif
5178 kfree(orig_data);
5179 return err;
5180 }
5181
5182 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5183 {
5184 struct super_block *sb = dentry->d_sb;
5185 struct ext4_sb_info *sbi = EXT4_SB(sb);
5186 struct ext4_super_block *es = sbi->s_es;
5187 ext4_fsblk_t overhead = 0, resv_blocks;
5188 u64 fsid;
5189 s64 bfree;
5190 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5191
5192 if (!test_opt(sb, MINIX_DF))
5193 overhead = sbi->s_overhead;
5194
5195 buf->f_type = EXT4_SUPER_MAGIC;
5196 buf->f_bsize = sb->s_blocksize;
5197 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5198 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5199 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5200 /* prevent underflow in case that few free space is available */
5201 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5202 buf->f_bavail = buf->f_bfree -
5203 (ext4_r_blocks_count(es) + resv_blocks);
5204 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5205 buf->f_bavail = 0;
5206 buf->f_files = le32_to_cpu(es->s_inodes_count);
5207 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5208 buf->f_namelen = EXT4_NAME_LEN;
5209 fsid = le64_to_cpup((void *)es->s_uuid) ^
5210 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5211 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5212 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5213
5214 return 0;
5215 }
5216
5217 /* Helper function for writing quotas on sync - we need to start transaction
5218 * before quota file is locked for write. Otherwise the are possible deadlocks:
5219 * Process 1 Process 2
5220 * ext4_create() quota_sync()
5221 * jbd2_journal_start() write_dquot()
5222 * dquot_initialize() down(dqio_mutex)
5223 * down(dqio_mutex) jbd2_journal_start()
5224 *
5225 */
5226
5227 #ifdef CONFIG_QUOTA
5228
5229 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5230 {
5231 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5232 }
5233
5234 static int ext4_write_dquot(struct dquot *dquot)
5235 {
5236 int ret, err;
5237 handle_t *handle;
5238 struct inode *inode;
5239
5240 inode = dquot_to_inode(dquot);
5241 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5242 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5243 if (IS_ERR(handle))
5244 return PTR_ERR(handle);
5245 ret = dquot_commit(dquot);
5246 err = ext4_journal_stop(handle);
5247 if (!ret)
5248 ret = err;
5249 return ret;
5250 }
5251
5252 static int ext4_acquire_dquot(struct dquot *dquot)
5253 {
5254 int ret, err;
5255 handle_t *handle;
5256
5257 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5258 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5259 if (IS_ERR(handle))
5260 return PTR_ERR(handle);
5261 ret = dquot_acquire(dquot);
5262 err = ext4_journal_stop(handle);
5263 if (!ret)
5264 ret = err;
5265 return ret;
5266 }
5267
5268 static int ext4_release_dquot(struct dquot *dquot)
5269 {
5270 int ret, err;
5271 handle_t *handle;
5272
5273 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5274 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5275 if (IS_ERR(handle)) {
5276 /* Release dquot anyway to avoid endless cycle in dqput() */
5277 dquot_release(dquot);
5278 return PTR_ERR(handle);
5279 }
5280 ret = dquot_release(dquot);
5281 err = ext4_journal_stop(handle);
5282 if (!ret)
5283 ret = err;
5284 return ret;
5285 }
5286
5287 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5288 {
5289 struct super_block *sb = dquot->dq_sb;
5290 struct ext4_sb_info *sbi = EXT4_SB(sb);
5291
5292 /* Are we journaling quotas? */
5293 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5294 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5295 dquot_mark_dquot_dirty(dquot);
5296 return ext4_write_dquot(dquot);
5297 } else {
5298 return dquot_mark_dquot_dirty(dquot);
5299 }
5300 }
5301
5302 static int ext4_write_info(struct super_block *sb, int type)
5303 {
5304 int ret, err;
5305 handle_t *handle;
5306
5307 /* Data block + inode block */
5308 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5309 if (IS_ERR(handle))
5310 return PTR_ERR(handle);
5311 ret = dquot_commit_info(sb, type);
5312 err = ext4_journal_stop(handle);
5313 if (!ret)
5314 ret = err;
5315 return ret;
5316 }
5317
5318 /*
5319 * Turn on quotas during mount time - we need to find
5320 * the quota file and such...
5321 */
5322 static int ext4_quota_on_mount(struct super_block *sb, int type)
5323 {
5324 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5325 EXT4_SB(sb)->s_jquota_fmt, type);
5326 }
5327
5328 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5329 {
5330 struct ext4_inode_info *ei = EXT4_I(inode);
5331
5332 /* The first argument of lockdep_set_subclass has to be
5333 * *exactly* the same as the argument to init_rwsem() --- in
5334 * this case, in init_once() --- or lockdep gets unhappy
5335 * because the name of the lock is set using the
5336 * stringification of the argument to init_rwsem().
5337 */
5338 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5339 lockdep_set_subclass(&ei->i_data_sem, subclass);
5340 }
5341
5342 /*
5343 * Standard function to be called on quota_on
5344 */
5345 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5346 struct path *path)
5347 {
5348 int err;
5349
5350 if (!test_opt(sb, QUOTA))
5351 return -EINVAL;
5352
5353 /* Quotafile not on the same filesystem? */
5354 if (path->dentry->d_sb != sb)
5355 return -EXDEV;
5356 /* Journaling quota? */
5357 if (EXT4_SB(sb)->s_qf_names[type]) {
5358 /* Quotafile not in fs root? */
5359 if (path->dentry->d_parent != sb->s_root)
5360 ext4_msg(sb, KERN_WARNING,
5361 "Quota file not on filesystem root. "
5362 "Journaled quota will not work");
5363 }
5364
5365 /*
5366 * When we journal data on quota file, we have to flush journal to see
5367 * all updates to the file when we bypass pagecache...
5368 */
5369 if (EXT4_SB(sb)->s_journal &&
5370 ext4_should_journal_data(d_inode(path->dentry))) {
5371 /*
5372 * We don't need to lock updates but journal_flush() could
5373 * otherwise be livelocked...
5374 */
5375 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5376 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5377 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5378 if (err)
5379 return err;
5380 }
5381 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5382 err = dquot_quota_on(sb, type, format_id, path);
5383 if (err)
5384 lockdep_set_quota_inode(path->dentry->d_inode,
5385 I_DATA_SEM_NORMAL);
5386 return err;
5387 }
5388
5389 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5390 unsigned int flags)
5391 {
5392 int err;
5393 struct inode *qf_inode;
5394 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5395 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5396 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5397 };
5398
5399 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5400
5401 if (!qf_inums[type])
5402 return -EPERM;
5403
5404 qf_inode = ext4_iget(sb, qf_inums[type]);
5405 if (IS_ERR(qf_inode)) {
5406 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5407 return PTR_ERR(qf_inode);
5408 }
5409
5410 /* Don't account quota for quota files to avoid recursion */
5411 qf_inode->i_flags |= S_NOQUOTA;
5412 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5413 err = dquot_enable(qf_inode, type, format_id, flags);
5414 iput(qf_inode);
5415 if (err)
5416 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5417
5418 return err;
5419 }
5420
5421 /* Enable usage tracking for all quota types. */
5422 static int ext4_enable_quotas(struct super_block *sb)
5423 {
5424 int type, err = 0;
5425 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5426 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5427 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5428 };
5429
5430 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5431 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5432 if (qf_inums[type]) {
5433 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5434 DQUOT_USAGE_ENABLED);
5435 if (err) {
5436 ext4_warning(sb,
5437 "Failed to enable quota tracking "
5438 "(type=%d, err=%d). Please run "
5439 "e2fsck to fix.", type, err);
5440 return err;
5441 }
5442 }
5443 }
5444 return 0;
5445 }
5446
5447 static int ext4_quota_off(struct super_block *sb, int type)
5448 {
5449 struct inode *inode = sb_dqopt(sb)->files[type];
5450 handle_t *handle;
5451
5452 /* Force all delayed allocation blocks to be allocated.
5453 * Caller already holds s_umount sem */
5454 if (test_opt(sb, DELALLOC))
5455 sync_filesystem(sb);
5456
5457 if (!inode)
5458 goto out;
5459
5460 /* Update modification times of quota files when userspace can
5461 * start looking at them */
5462 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5463 if (IS_ERR(handle))
5464 goto out;
5465 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5466 ext4_mark_inode_dirty(handle, inode);
5467 ext4_journal_stop(handle);
5468
5469 out:
5470 return dquot_quota_off(sb, type);
5471 }
5472
5473 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5474 * acquiring the locks... As quota files are never truncated and quota code
5475 * itself serializes the operations (and no one else should touch the files)
5476 * we don't have to be afraid of races */
5477 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5478 size_t len, loff_t off)
5479 {
5480 struct inode *inode = sb_dqopt(sb)->files[type];
5481 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5482 int offset = off & (sb->s_blocksize - 1);
5483 int tocopy;
5484 size_t toread;
5485 struct buffer_head *bh;
5486 loff_t i_size = i_size_read(inode);
5487
5488 if (off > i_size)
5489 return 0;
5490 if (off+len > i_size)
5491 len = i_size-off;
5492 toread = len;
5493 while (toread > 0) {
5494 tocopy = sb->s_blocksize - offset < toread ?
5495 sb->s_blocksize - offset : toread;
5496 bh = ext4_bread(NULL, inode, blk, 0);
5497 if (IS_ERR(bh))
5498 return PTR_ERR(bh);
5499 if (!bh) /* A hole? */
5500 memset(data, 0, tocopy);
5501 else
5502 memcpy(data, bh->b_data+offset, tocopy);
5503 brelse(bh);
5504 offset = 0;
5505 toread -= tocopy;
5506 data += tocopy;
5507 blk++;
5508 }
5509 return len;
5510 }
5511
5512 /* Write to quotafile (we know the transaction is already started and has
5513 * enough credits) */
5514 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5515 const char *data, size_t len, loff_t off)
5516 {
5517 struct inode *inode = sb_dqopt(sb)->files[type];
5518 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5519 int err, offset = off & (sb->s_blocksize - 1);
5520 struct buffer_head *bh;
5521 handle_t *handle = journal_current_handle();
5522
5523 if (EXT4_SB(sb)->s_journal && !handle) {
5524 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5525 " cancelled because transaction is not started",
5526 (unsigned long long)off, (unsigned long long)len);
5527 return -EIO;
5528 }
5529 /*
5530 * Since we account only one data block in transaction credits,
5531 * then it is impossible to cross a block boundary.
5532 */
5533 if (sb->s_blocksize - offset < len) {
5534 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5535 " cancelled because not block aligned",
5536 (unsigned long long)off, (unsigned long long)len);
5537 return -EIO;
5538 }
5539
5540 bh = ext4_bread(handle, inode, blk, 1);
5541 if (IS_ERR(bh))
5542 return PTR_ERR(bh);
5543 if (!bh)
5544 goto out;
5545 BUFFER_TRACE(bh, "get write access");
5546 err = ext4_journal_get_write_access(handle, bh);
5547 if (err) {
5548 brelse(bh);
5549 return err;
5550 }
5551 lock_buffer(bh);
5552 memcpy(bh->b_data+offset, data, len);
5553 flush_dcache_page(bh->b_page);
5554 unlock_buffer(bh);
5555 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5556 brelse(bh);
5557 out:
5558 if (inode->i_size < off + len) {
5559 i_size_write(inode, off + len);
5560 EXT4_I(inode)->i_disksize = inode->i_size;
5561 ext4_mark_inode_dirty(handle, inode);
5562 }
5563 return len;
5564 }
5565
5566 #endif
5567
5568 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5569 const char *dev_name, void *data)
5570 {
5571 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5572 }
5573
5574 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5575 static inline void register_as_ext2(void)
5576 {
5577 int err = register_filesystem(&ext2_fs_type);
5578 if (err)
5579 printk(KERN_WARNING
5580 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5581 }
5582
5583 static inline void unregister_as_ext2(void)
5584 {
5585 unregister_filesystem(&ext2_fs_type);
5586 }
5587
5588 static inline int ext2_feature_set_ok(struct super_block *sb)
5589 {
5590 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5591 return 0;
5592 if (sb->s_flags & MS_RDONLY)
5593 return 1;
5594 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5595 return 0;
5596 return 1;
5597 }
5598 #else
5599 static inline void register_as_ext2(void) { }
5600 static inline void unregister_as_ext2(void) { }
5601 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5602 #endif
5603
5604 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5605 static inline void register_as_ext3(void)
5606 {
5607 int err = register_filesystem(&ext3_fs_type);
5608 if (err)
5609 printk(KERN_WARNING
5610 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5611 }
5612
5613 static inline void unregister_as_ext3(void)
5614 {
5615 unregister_filesystem(&ext3_fs_type);
5616 }
5617
5618 static inline int ext3_feature_set_ok(struct super_block *sb)
5619 {
5620 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5621 return 0;
5622 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5623 return 0;
5624 if (sb->s_flags & MS_RDONLY)
5625 return 1;
5626 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5627 return 0;
5628 return 1;
5629 }
5630 #else
5631 static inline void register_as_ext3(void) { }
5632 static inline void unregister_as_ext3(void) { }
5633 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5634 #endif
5635
5636 static struct file_system_type ext4_fs_type = {
5637 .owner = THIS_MODULE,
5638 .name = "ext4",
5639 .mount = ext4_mount,
5640 .kill_sb = kill_block_super,
5641 .fs_flags = FS_REQUIRES_DEV,
5642 };
5643 MODULE_ALIAS_FS("ext4");
5644
5645 static int __init ext4_init_feat_adverts(void)
5646 {
5647 struct ext4_features *ef;
5648 int ret = -ENOMEM;
5649
5650 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5651 if (!ef)
5652 goto out;
5653
5654 ef->f_kobj.kset = ext4_kset;
5655 init_completion(&ef->f_kobj_unregister);
5656 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5657 "features");
5658 if (ret) {
5659 kfree(ef);
5660 goto out;
5661 }
5662
5663 ext4_feat = ef;
5664 ret = 0;
5665 out:
5666 return ret;
5667 }
5668
5669 static void ext4_exit_feat_adverts(void)
5670 {
5671 kobject_put(&ext4_feat->f_kobj);
5672 wait_for_completion(&ext4_feat->f_kobj_unregister);
5673 kfree(ext4_feat);
5674 }
5675
5676 /* Shared across all ext4 file systems */
5677 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5678 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5679
5680 static int __init ext4_init_fs(void)
5681 {
5682 int i, err;
5683
5684 ext4_li_info = NULL;
5685 mutex_init(&ext4_li_mtx);
5686
5687 /* Build-time check for flags consistency */
5688 ext4_check_flag_values();
5689
5690 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5691 mutex_init(&ext4__aio_mutex[i]);
5692 init_waitqueue_head(&ext4__ioend_wq[i]);
5693 }
5694
5695 err = ext4_init_es();
5696 if (err)
5697 return err;
5698
5699 err = ext4_init_pageio();
5700 if (err)
5701 goto out7;
5702
5703 err = ext4_init_system_zone();
5704 if (err)
5705 goto out6;
5706 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5707 if (!ext4_kset) {
5708 err = -ENOMEM;
5709 goto out5;
5710 }
5711 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5712
5713 err = ext4_init_feat_adverts();
5714 if (err)
5715 goto out4;
5716
5717 err = ext4_init_mballoc();
5718 if (err)
5719 goto out2;
5720 else
5721 ext4_mballoc_ready = 1;
5722 err = init_inodecache();
5723 if (err)
5724 goto out1;
5725 register_as_ext3();
5726 register_as_ext2();
5727 err = register_filesystem(&ext4_fs_type);
5728 if (err)
5729 goto out;
5730
5731 return 0;
5732 out:
5733 unregister_as_ext2();
5734 unregister_as_ext3();
5735 destroy_inodecache();
5736 out1:
5737 ext4_mballoc_ready = 0;
5738 ext4_exit_mballoc();
5739 out2:
5740 ext4_exit_feat_adverts();
5741 out4:
5742 if (ext4_proc_root)
5743 remove_proc_entry("fs/ext4", NULL);
5744 kset_unregister(ext4_kset);
5745 out5:
5746 ext4_exit_system_zone();
5747 out6:
5748 ext4_exit_pageio();
5749 out7:
5750 ext4_exit_es();
5751
5752 return err;
5753 }
5754
5755 static void __exit ext4_exit_fs(void)
5756 {
5757 ext4_destroy_lazyinit_thread();
5758 unregister_as_ext2();
5759 unregister_as_ext3();
5760 unregister_filesystem(&ext4_fs_type);
5761 destroy_inodecache();
5762 ext4_exit_mballoc();
5763 ext4_exit_feat_adverts();
5764 remove_proc_entry("fs/ext4", NULL);
5765 kset_unregister(ext4_kset);
5766 ext4_exit_system_zone();
5767 ext4_exit_pageio();
5768 ext4_exit_es();
5769 }
5770
5771 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5772 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5773 MODULE_LICENSE("GPL");
5774 module_init(ext4_init_fs)
5775 module_exit(ext4_exit_fs)
Linux with added FPC-III support