projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
vfs: add file_path() helper
[deliverable/linux.git] / fs / ext4 / super.c
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/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 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
298 es->s_last_error_time = cpu_to_le32(get_seconds());
299 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
300 es->s_last_error_line = cpu_to_le32(line);
301 if (!es->s_first_error_time) {
302 es->s_first_error_time = es->s_last_error_time;
303 strncpy(es->s_first_error_func, func,
304 sizeof(es->s_first_error_func));
305 es->s_first_error_line = cpu_to_le32(line);
306 es->s_first_error_ino = es->s_last_error_ino;
307 es->s_first_error_block = es->s_last_error_block;
308 }
309 /*
310 * Start the daily error reporting function if it hasn't been
311 * started already
312 */
313 if (!es->s_error_count)
314 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
315 le32_add_cpu(&es->s_error_count, 1);
316 }
317
318 static void save_error_info(struct super_block *sb, const char *func,
319 unsigned int line)
320 {
321 __save_error_info(sb, func, line);
322 ext4_commit_super(sb, 1);
323 }
324
325 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
326 {
327 struct super_block *sb = journal->j_private;
328 struct ext4_sb_info *sbi = EXT4_SB(sb);
329 int error = is_journal_aborted(journal);
330 struct ext4_journal_cb_entry *jce;
331
332 BUG_ON(txn->t_state == T_FINISHED);
333 spin_lock(&sbi->s_md_lock);
334 while (!list_empty(&txn->t_private_list)) {
335 jce = list_entry(txn->t_private_list.next,
336 struct ext4_journal_cb_entry, jce_list);
337 list_del_init(&jce->jce_list);
338 spin_unlock(&sbi->s_md_lock);
339 jce->jce_func(sb, jce, error);
340 spin_lock(&sbi->s_md_lock);
341 }
342 spin_unlock(&sbi->s_md_lock);
343 }
344
345 /* Deal with the reporting of failure conditions on a filesystem such as
346 * inconsistencies detected or read IO failures.
347 *
348 * On ext2, we can store the error state of the filesystem in the
349 * superblock. That is not possible on ext4, because we may have other
350 * write ordering constraints on the superblock which prevent us from
351 * writing it out straight away; and given that the journal is about to
352 * be aborted, we can't rely on the current, or future, transactions to
353 * write out the superblock safely.
354 *
355 * We'll just use the jbd2_journal_abort() error code to record an error in
356 * the journal instead. On recovery, the journal will complain about
357 * that error until we've noted it down and cleared it.
358 */
359
360 static void ext4_handle_error(struct super_block *sb)
361 {
362 if (sb->s_flags & MS_RDONLY)
363 return;
364
365 if (!test_opt(sb, ERRORS_CONT)) {
366 journal_t *journal = EXT4_SB(sb)->s_journal;
367
368 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
369 if (journal)
370 jbd2_journal_abort(journal, -EIO);
371 }
372 if (test_opt(sb, ERRORS_RO)) {
373 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
374 /*
375 * Make sure updated value of ->s_mount_flags will be visible
376 * before ->s_flags update
377 */
378 smp_wmb();
379 sb->s_flags |= MS_RDONLY;
380 }
381 if (test_opt(sb, ERRORS_PANIC))
382 panic("EXT4-fs (device %s): panic forced after error\n",
383 sb->s_id);
384 }
385
386 #define ext4_error_ratelimit(sb) \
387 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
388 "EXT4-fs error")
389
390 void __ext4_error(struct super_block *sb, const char *function,
391 unsigned int line, const char *fmt, ...)
392 {
393 struct va_format vaf;
394 va_list args;
395
396 if (ext4_error_ratelimit(sb)) {
397 va_start(args, fmt);
398 vaf.fmt = fmt;
399 vaf.va = &args;
400 printk(KERN_CRIT
401 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
402 sb->s_id, function, line, current->comm, &vaf);
403 va_end(args);
404 }
405 save_error_info(sb, function, line);
406 ext4_handle_error(sb);
407 }
408
409 void __ext4_error_inode(struct inode *inode, const char *function,
410 unsigned int line, ext4_fsblk_t block,
411 const char *fmt, ...)
412 {
413 va_list args;
414 struct va_format vaf;
415 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
416
417 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
418 es->s_last_error_block = cpu_to_le64(block);
419 if (ext4_error_ratelimit(inode->i_sb)) {
420 va_start(args, fmt);
421 vaf.fmt = fmt;
422 vaf.va = &args;
423 if (block)
424 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
425 "inode #%lu: block %llu: comm %s: %pV\n",
426 inode->i_sb->s_id, function, line, inode->i_ino,
427 block, current->comm, &vaf);
428 else
429 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
430 "inode #%lu: comm %s: %pV\n",
431 inode->i_sb->s_id, function, line, inode->i_ino,
432 current->comm, &vaf);
433 va_end(args);
434 }
435 save_error_info(inode->i_sb, function, line);
436 ext4_handle_error(inode->i_sb);
437 }
438
439 void __ext4_error_file(struct file *file, const char *function,
440 unsigned int line, ext4_fsblk_t block,
441 const char *fmt, ...)
442 {
443 va_list args;
444 struct va_format vaf;
445 struct ext4_super_block *es;
446 struct inode *inode = file_inode(file);
447 char pathname[80], *path;
448
449 es = EXT4_SB(inode->i_sb)->s_es;
450 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
451 if (ext4_error_ratelimit(inode->i_sb)) {
452 path = file_path(file, pathname, sizeof(pathname));
453 if (IS_ERR(path))
454 path = "(unknown)";
455 va_start(args, fmt);
456 vaf.fmt = fmt;
457 vaf.va = &args;
458 if (block)
459 printk(KERN_CRIT
460 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
461 "block %llu: comm %s: path %s: %pV\n",
462 inode->i_sb->s_id, function, line, inode->i_ino,
463 block, current->comm, path, &vaf);
464 else
465 printk(KERN_CRIT
466 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
467 "comm %s: path %s: %pV\n",
468 inode->i_sb->s_id, function, line, inode->i_ino,
469 current->comm, path, &vaf);
470 va_end(args);
471 }
472 save_error_info(inode->i_sb, function, line);
473 ext4_handle_error(inode->i_sb);
474 }
475
476 const char *ext4_decode_error(struct super_block *sb, int errno,
477 char nbuf[16])
478 {
479 char *errstr = NULL;
480
481 switch (errno) {
482 case -EIO:
483 errstr = "IO failure";
484 break;
485 case -ENOMEM:
486 errstr = "Out of memory";
487 break;
488 case -EROFS:
489 if (!sb || (EXT4_SB(sb)->s_journal &&
490 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
491 errstr = "Journal has aborted";
492 else
493 errstr = "Readonly filesystem";
494 break;
495 default:
496 /* If the caller passed in an extra buffer for unknown
497 * errors, textualise them now. Else we just return
498 * NULL. */
499 if (nbuf) {
500 /* Check for truncated error codes... */
501 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
502 errstr = nbuf;
503 }
504 break;
505 }
506
507 return errstr;
508 }
509
510 /* __ext4_std_error decodes expected errors from journaling functions
511 * automatically and invokes the appropriate error response. */
512
513 void __ext4_std_error(struct super_block *sb, const char *function,
514 unsigned int line, int errno)
515 {
516 char nbuf[16];
517 const char *errstr;
518
519 /* Special case: if the error is EROFS, and we're not already
520 * inside a transaction, then there's really no point in logging
521 * an error. */
522 if (errno == -EROFS && journal_current_handle() == NULL &&
523 (sb->s_flags & MS_RDONLY))
524 return;
525
526 if (ext4_error_ratelimit(sb)) {
527 errstr = ext4_decode_error(sb, errno, nbuf);
528 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
529 sb->s_id, function, line, errstr);
530 }
531
532 save_error_info(sb, function, line);
533 ext4_handle_error(sb);
534 }
535
536 /*
537 * ext4_abort is a much stronger failure handler than ext4_error. The
538 * abort function may be used to deal with unrecoverable failures such
539 * as journal IO errors or ENOMEM at a critical moment in log management.
540 *
541 * We unconditionally force the filesystem into an ABORT|READONLY state,
542 * unless the error response on the fs has been set to panic in which
543 * case we take the easy way out and panic immediately.
544 */
545
546 void __ext4_abort(struct super_block *sb, const char *function,
547 unsigned int line, const char *fmt, ...)
548 {
549 va_list args;
550
551 save_error_info(sb, function, line);
552 va_start(args, fmt);
553 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
554 function, line);
555 vprintk(fmt, args);
556 printk("\n");
557 va_end(args);
558
559 if ((sb->s_flags & MS_RDONLY) == 0) {
560 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
561 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
562 /*
563 * Make sure updated value of ->s_mount_flags will be visible
564 * before ->s_flags update
565 */
566 smp_wmb();
567 sb->s_flags |= MS_RDONLY;
568 if (EXT4_SB(sb)->s_journal)
569 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
570 save_error_info(sb, function, line);
571 }
572 if (test_opt(sb, ERRORS_PANIC))
573 panic("EXT4-fs panic from previous error\n");
574 }
575
576 void __ext4_msg(struct super_block *sb,
577 const char *prefix, const char *fmt, ...)
578 {
579 struct va_format vaf;
580 va_list args;
581
582 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
583 return;
584
585 va_start(args, fmt);
586 vaf.fmt = fmt;
587 vaf.va = &args;
588 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
589 va_end(args);
590 }
591
592 void __ext4_warning(struct super_block *sb, const char *function,
593 unsigned int line, const char *fmt, ...)
594 {
595 struct va_format vaf;
596 va_list args;
597
598 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
599 "EXT4-fs warning"))
600 return;
601
602 va_start(args, fmt);
603 vaf.fmt = fmt;
604 vaf.va = &args;
605 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
606 sb->s_id, function, line, &vaf);
607 va_end(args);
608 }
609
610 void __ext4_grp_locked_error(const char *function, unsigned int line,
611 struct super_block *sb, ext4_group_t grp,
612 unsigned long ino, ext4_fsblk_t block,
613 const char *fmt, ...)
614 __releases(bitlock)
615 __acquires(bitlock)
616 {
617 struct va_format vaf;
618 va_list args;
619 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
620
621 es->s_last_error_ino = cpu_to_le32(ino);
622 es->s_last_error_block = cpu_to_le64(block);
623 __save_error_info(sb, function, line);
624
625 if (ext4_error_ratelimit(sb)) {
626 va_start(args, fmt);
627 vaf.fmt = fmt;
628 vaf.va = &args;
629 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
630 sb->s_id, function, line, grp);
631 if (ino)
632 printk(KERN_CONT "inode %lu: ", ino);
633 if (block)
634 printk(KERN_CONT "block %llu:",
635 (unsigned long long) block);
636 printk(KERN_CONT "%pV\n", &vaf);
637 va_end(args);
638 }
639
640 if (test_opt(sb, ERRORS_CONT)) {
641 ext4_commit_super(sb, 0);
642 return;
643 }
644
645 ext4_unlock_group(sb, grp);
646 ext4_handle_error(sb);
647 /*
648 * We only get here in the ERRORS_RO case; relocking the group
649 * may be dangerous, but nothing bad will happen since the
650 * filesystem will have already been marked read/only and the
651 * journal has been aborted. We return 1 as a hint to callers
652 * who might what to use the return value from
653 * ext4_grp_locked_error() to distinguish between the
654 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
655 * aggressively from the ext4 function in question, with a
656 * more appropriate error code.
657 */
658 ext4_lock_group(sb, grp);
659 return;
660 }
661
662 void ext4_update_dynamic_rev(struct super_block *sb)
663 {
664 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
665
666 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
667 return;
668
669 ext4_warning(sb,
670 "updating to rev %d because of new feature flag, "
671 "running e2fsck is recommended",
672 EXT4_DYNAMIC_REV);
673
674 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
675 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
676 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
677 /* leave es->s_feature_*compat flags alone */
678 /* es->s_uuid will be set by e2fsck if empty */
679
680 /*
681 * The rest of the superblock fields should be zero, and if not it
682 * means they are likely already in use, so leave them alone. We
683 * can leave it up to e2fsck to clean up any inconsistencies there.
684 */
685 }
686
687 /*
688 * Open the external journal device
689 */
690 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
691 {
692 struct block_device *bdev;
693 char b[BDEVNAME_SIZE];
694
695 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
696 if (IS_ERR(bdev))
697 goto fail;
698 return bdev;
699
700 fail:
701 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
702 __bdevname(dev, b), PTR_ERR(bdev));
703 return NULL;
704 }
705
706 /*
707 * Release the journal device
708 */
709 static void ext4_blkdev_put(struct block_device *bdev)
710 {
711 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
712 }
713
714 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
715 {
716 struct block_device *bdev;
717 bdev = sbi->journal_bdev;
718 if (bdev) {
719 ext4_blkdev_put(bdev);
720 sbi->journal_bdev = NULL;
721 }
722 }
723
724 static inline struct inode *orphan_list_entry(struct list_head *l)
725 {
726 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
727 }
728
729 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
730 {
731 struct list_head *l;
732
733 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
734 le32_to_cpu(sbi->s_es->s_last_orphan));
735
736 printk(KERN_ERR "sb_info orphan list:\n");
737 list_for_each(l, &sbi->s_orphan) {
738 struct inode *inode = orphan_list_entry(l);
739 printk(KERN_ERR " "
740 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
741 inode->i_sb->s_id, inode->i_ino, inode,
742 inode->i_mode, inode->i_nlink,
743 NEXT_ORPHAN(inode));
744 }
745 }
746
747 static void ext4_put_super(struct super_block *sb)
748 {
749 struct ext4_sb_info *sbi = EXT4_SB(sb);
750 struct ext4_super_block *es = sbi->s_es;
751 int i, err;
752
753 ext4_unregister_li_request(sb);
754 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
755
756 flush_workqueue(sbi->rsv_conversion_wq);
757 destroy_workqueue(sbi->rsv_conversion_wq);
758
759 if (sbi->s_journal) {
760 err = jbd2_journal_destroy(sbi->s_journal);
761 sbi->s_journal = NULL;
762 if (err < 0)
763 ext4_abort(sb, "Couldn't clean up the journal");
764 }
765
766 ext4_es_unregister_shrinker(sbi);
767 del_timer_sync(&sbi->s_err_report);
768 ext4_release_system_zone(sb);
769 ext4_mb_release(sb);
770 ext4_ext_release(sb);
771 ext4_xattr_put_super(sb);
772
773 if (!(sb->s_flags & MS_RDONLY)) {
774 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
775 es->s_state = cpu_to_le16(sbi->s_mount_state);
776 }
777 if (!(sb->s_flags & MS_RDONLY))
778 ext4_commit_super(sb, 1);
779
780 if (sbi->s_proc) {
781 remove_proc_entry("options", sbi->s_proc);
782 remove_proc_entry(sb->s_id, ext4_proc_root);
783 }
784 kobject_del(&sbi->s_kobj);
785
786 for (i = 0; i < sbi->s_gdb_count; i++)
787 brelse(sbi->s_group_desc[i]);
788 kvfree(sbi->s_group_desc);
789 kvfree(sbi->s_flex_groups);
790 percpu_counter_destroy(&sbi->s_freeclusters_counter);
791 percpu_counter_destroy(&sbi->s_freeinodes_counter);
792 percpu_counter_destroy(&sbi->s_dirs_counter);
793 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
794 brelse(sbi->s_sbh);
795 #ifdef CONFIG_QUOTA
796 for (i = 0; i < EXT4_MAXQUOTAS; i++)
797 kfree(sbi->s_qf_names[i]);
798 #endif
799
800 /* Debugging code just in case the in-memory inode orphan list
801 * isn't empty. The on-disk one can be non-empty if we've
802 * detected an error and taken the fs readonly, but the
803 * in-memory list had better be clean by this point. */
804 if (!list_empty(&sbi->s_orphan))
805 dump_orphan_list(sb, sbi);
806 J_ASSERT(list_empty(&sbi->s_orphan));
807
808 invalidate_bdev(sb->s_bdev);
809 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
810 /*
811 * Invalidate the journal device's buffers. We don't want them
812 * floating about in memory - the physical journal device may
813 * hotswapped, and it breaks the `ro-after' testing code.
814 */
815 sync_blockdev(sbi->journal_bdev);
816 invalidate_bdev(sbi->journal_bdev);
817 ext4_blkdev_remove(sbi);
818 }
819 if (sbi->s_mb_cache) {
820 ext4_xattr_destroy_cache(sbi->s_mb_cache);
821 sbi->s_mb_cache = NULL;
822 }
823 if (sbi->s_mmp_tsk)
824 kthread_stop(sbi->s_mmp_tsk);
825 sb->s_fs_info = NULL;
826 /*
827 * Now that we are completely done shutting down the
828 * superblock, we need to actually destroy the kobject.
829 */
830 kobject_put(&sbi->s_kobj);
831 wait_for_completion(&sbi->s_kobj_unregister);
832 if (sbi->s_chksum_driver)
833 crypto_free_shash(sbi->s_chksum_driver);
834 kfree(sbi->s_blockgroup_lock);
835 kfree(sbi);
836 }
837
838 static struct kmem_cache *ext4_inode_cachep;
839
840 /*
841 * Called inside transaction, so use GFP_NOFS
842 */
843 static struct inode *ext4_alloc_inode(struct super_block *sb)
844 {
845 struct ext4_inode_info *ei;
846
847 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
848 if (!ei)
849 return NULL;
850
851 ei->vfs_inode.i_version = 1;
852 spin_lock_init(&ei->i_raw_lock);
853 INIT_LIST_HEAD(&ei->i_prealloc_list);
854 spin_lock_init(&ei->i_prealloc_lock);
855 ext4_es_init_tree(&ei->i_es_tree);
856 rwlock_init(&ei->i_es_lock);
857 INIT_LIST_HEAD(&ei->i_es_list);
858 ei->i_es_all_nr = 0;
859 ei->i_es_shk_nr = 0;
860 ei->i_es_shrink_lblk = 0;
861 ei->i_reserved_data_blocks = 0;
862 ei->i_reserved_meta_blocks = 0;
863 ei->i_allocated_meta_blocks = 0;
864 ei->i_da_metadata_calc_len = 0;
865 ei->i_da_metadata_calc_last_lblock = 0;
866 spin_lock_init(&(ei->i_block_reservation_lock));
867 #ifdef CONFIG_QUOTA
868 ei->i_reserved_quota = 0;
869 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
870 #endif
871 ei->jinode = NULL;
872 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
873 spin_lock_init(&ei->i_completed_io_lock);
874 ei->i_sync_tid = 0;
875 ei->i_datasync_tid = 0;
876 atomic_set(&ei->i_ioend_count, 0);
877 atomic_set(&ei->i_unwritten, 0);
878 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
879 #ifdef CONFIG_EXT4_FS_ENCRYPTION
880 ei->i_encryption_key.mode = EXT4_ENCRYPTION_MODE_INVALID;
881 #endif
882
883 return &ei->vfs_inode;
884 }
885
886 static int ext4_drop_inode(struct inode *inode)
887 {
888 int drop = generic_drop_inode(inode);
889
890 trace_ext4_drop_inode(inode, drop);
891 return drop;
892 }
893
894 static void ext4_i_callback(struct rcu_head *head)
895 {
896 struct inode *inode = container_of(head, struct inode, i_rcu);
897 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
898 }
899
900 static void ext4_destroy_inode(struct inode *inode)
901 {
902 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
903 ext4_msg(inode->i_sb, KERN_ERR,
904 "Inode %lu (%p): orphan list check failed!",
905 inode->i_ino, EXT4_I(inode));
906 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
907 EXT4_I(inode), sizeof(struct ext4_inode_info),
908 true);
909 dump_stack();
910 }
911 call_rcu(&inode->i_rcu, ext4_i_callback);
912 }
913
914 static void init_once(void *foo)
915 {
916 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
917
918 INIT_LIST_HEAD(&ei->i_orphan);
919 init_rwsem(&ei->xattr_sem);
920 init_rwsem(&ei->i_data_sem);
921 inode_init_once(&ei->vfs_inode);
922 }
923
924 static int __init init_inodecache(void)
925 {
926 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
927 sizeof(struct ext4_inode_info),
928 0, (SLAB_RECLAIM_ACCOUNT|
929 SLAB_MEM_SPREAD),
930 init_once);
931 if (ext4_inode_cachep == NULL)
932 return -ENOMEM;
933 return 0;
934 }
935
936 static void destroy_inodecache(void)
937 {
938 /*
939 * Make sure all delayed rcu free inodes are flushed before we
940 * destroy cache.
941 */
942 rcu_barrier();
943 kmem_cache_destroy(ext4_inode_cachep);
944 }
945
946 void ext4_clear_inode(struct inode *inode)
947 {
948 invalidate_inode_buffers(inode);
949 clear_inode(inode);
950 dquot_drop(inode);
951 ext4_discard_preallocations(inode);
952 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
953 if (EXT4_I(inode)->jinode) {
954 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
955 EXT4_I(inode)->jinode);
956 jbd2_free_inode(EXT4_I(inode)->jinode);
957 EXT4_I(inode)->jinode = NULL;
958 }
959 }
960
961 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
962 u64 ino, u32 generation)
963 {
964 struct inode *inode;
965
966 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
967 return ERR_PTR(-ESTALE);
968 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
969 return ERR_PTR(-ESTALE);
970
971 /* iget isn't really right if the inode is currently unallocated!!
972 *
973 * ext4_read_inode will return a bad_inode if the inode had been
974 * deleted, so we should be safe.
975 *
976 * Currently we don't know the generation for parent directory, so
977 * a generation of 0 means "accept any"
978 */
979 inode = ext4_iget_normal(sb, ino);
980 if (IS_ERR(inode))
981 return ERR_CAST(inode);
982 if (generation && inode->i_generation != generation) {
983 iput(inode);
984 return ERR_PTR(-ESTALE);
985 }
986
987 return inode;
988 }
989
990 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
991 int fh_len, int fh_type)
992 {
993 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
994 ext4_nfs_get_inode);
995 }
996
997 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
998 int fh_len, int fh_type)
999 {
1000 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1001 ext4_nfs_get_inode);
1002 }
1003
1004 /*
1005 * Try to release metadata pages (indirect blocks, directories) which are
1006 * mapped via the block device. Since these pages could have journal heads
1007 * which would prevent try_to_free_buffers() from freeing them, we must use
1008 * jbd2 layer's try_to_free_buffers() function to release them.
1009 */
1010 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1011 gfp_t wait)
1012 {
1013 journal_t *journal = EXT4_SB(sb)->s_journal;
1014
1015 WARN_ON(PageChecked(page));
1016 if (!page_has_buffers(page))
1017 return 0;
1018 if (journal)
1019 return jbd2_journal_try_to_free_buffers(journal, page,
1020 wait & ~__GFP_WAIT);
1021 return try_to_free_buffers(page);
1022 }
1023
1024 #ifdef CONFIG_QUOTA
1025 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1026 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1027
1028 static int ext4_write_dquot(struct dquot *dquot);
1029 static int ext4_acquire_dquot(struct dquot *dquot);
1030 static int ext4_release_dquot(struct dquot *dquot);
1031 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1032 static int ext4_write_info(struct super_block *sb, int type);
1033 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1034 struct path *path);
1035 static int ext4_quota_off(struct super_block *sb, int type);
1036 static int ext4_quota_on_mount(struct super_block *sb, int type);
1037 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1038 size_t len, loff_t off);
1039 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1040 const char *data, size_t len, loff_t off);
1041 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1042 unsigned int flags);
1043 static int ext4_enable_quotas(struct super_block *sb);
1044
1045 static struct dquot **ext4_get_dquots(struct inode *inode)
1046 {
1047 return EXT4_I(inode)->i_dquot;
1048 }
1049
1050 static const struct dquot_operations ext4_quota_operations = {
1051 .get_reserved_space = ext4_get_reserved_space,
1052 .write_dquot = ext4_write_dquot,
1053 .acquire_dquot = ext4_acquire_dquot,
1054 .release_dquot = ext4_release_dquot,
1055 .mark_dirty = ext4_mark_dquot_dirty,
1056 .write_info = ext4_write_info,
1057 .alloc_dquot = dquot_alloc,
1058 .destroy_dquot = dquot_destroy,
1059 };
1060
1061 static const struct quotactl_ops ext4_qctl_operations = {
1062 .quota_on = ext4_quota_on,
1063 .quota_off = ext4_quota_off,
1064 .quota_sync = dquot_quota_sync,
1065 .get_state = dquot_get_state,
1066 .set_info = dquot_set_dqinfo,
1067 .get_dqblk = dquot_get_dqblk,
1068 .set_dqblk = dquot_set_dqblk
1069 };
1070 #endif
1071
1072 static const struct super_operations ext4_sops = {
1073 .alloc_inode = ext4_alloc_inode,
1074 .destroy_inode = ext4_destroy_inode,
1075 .write_inode = ext4_write_inode,
1076 .dirty_inode = ext4_dirty_inode,
1077 .drop_inode = ext4_drop_inode,
1078 .evict_inode = ext4_evict_inode,
1079 .put_super = ext4_put_super,
1080 .sync_fs = ext4_sync_fs,
1081 .freeze_fs = ext4_freeze,
1082 .unfreeze_fs = ext4_unfreeze,
1083 .statfs = ext4_statfs,
1084 .remount_fs = ext4_remount,
1085 .show_options = ext4_show_options,
1086 #ifdef CONFIG_QUOTA
1087 .quota_read = ext4_quota_read,
1088 .quota_write = ext4_quota_write,
1089 .get_dquots = ext4_get_dquots,
1090 #endif
1091 .bdev_try_to_free_page = bdev_try_to_free_page,
1092 };
1093
1094 static const struct export_operations ext4_export_ops = {
1095 .fh_to_dentry = ext4_fh_to_dentry,
1096 .fh_to_parent = ext4_fh_to_parent,
1097 .get_parent = ext4_get_parent,
1098 };
1099
1100 enum {
1101 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1102 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1103 Opt_nouid32, Opt_debug, Opt_removed,
1104 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1105 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1106 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1107 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1108 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1109 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1110 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1111 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1112 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1113 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1114 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1115 Opt_lazytime, Opt_nolazytime,
1116 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1117 Opt_inode_readahead_blks, Opt_journal_ioprio,
1118 Opt_dioread_nolock, Opt_dioread_lock,
1119 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1120 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1121 };
1122
1123 static const match_table_t tokens = {
1124 {Opt_bsd_df, "bsddf"},
1125 {Opt_minix_df, "minixdf"},
1126 {Opt_grpid, "grpid"},
1127 {Opt_grpid, "bsdgroups"},
1128 {Opt_nogrpid, "nogrpid"},
1129 {Opt_nogrpid, "sysvgroups"},
1130 {Opt_resgid, "resgid=%u"},
1131 {Opt_resuid, "resuid=%u"},
1132 {Opt_sb, "sb=%u"},
1133 {Opt_err_cont, "errors=continue"},
1134 {Opt_err_panic, "errors=panic"},
1135 {Opt_err_ro, "errors=remount-ro"},
1136 {Opt_nouid32, "nouid32"},
1137 {Opt_debug, "debug"},
1138 {Opt_removed, "oldalloc"},
1139 {Opt_removed, "orlov"},
1140 {Opt_user_xattr, "user_xattr"},
1141 {Opt_nouser_xattr, "nouser_xattr"},
1142 {Opt_acl, "acl"},
1143 {Opt_noacl, "noacl"},
1144 {Opt_noload, "norecovery"},
1145 {Opt_noload, "noload"},
1146 {Opt_removed, "nobh"},
1147 {Opt_removed, "bh"},
1148 {Opt_commit, "commit=%u"},
1149 {Opt_min_batch_time, "min_batch_time=%u"},
1150 {Opt_max_batch_time, "max_batch_time=%u"},
1151 {Opt_journal_dev, "journal_dev=%u"},
1152 {Opt_journal_path, "journal_path=%s"},
1153 {Opt_journal_checksum, "journal_checksum"},
1154 {Opt_nojournal_checksum, "nojournal_checksum"},
1155 {Opt_journal_async_commit, "journal_async_commit"},
1156 {Opt_abort, "abort"},
1157 {Opt_data_journal, "data=journal"},
1158 {Opt_data_ordered, "data=ordered"},
1159 {Opt_data_writeback, "data=writeback"},
1160 {Opt_data_err_abort, "data_err=abort"},
1161 {Opt_data_err_ignore, "data_err=ignore"},
1162 {Opt_offusrjquota, "usrjquota="},
1163 {Opt_usrjquota, "usrjquota=%s"},
1164 {Opt_offgrpjquota, "grpjquota="},
1165 {Opt_grpjquota, "grpjquota=%s"},
1166 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1167 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1168 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1169 {Opt_grpquota, "grpquota"},
1170 {Opt_noquota, "noquota"},
1171 {Opt_quota, "quota"},
1172 {Opt_usrquota, "usrquota"},
1173 {Opt_barrier, "barrier=%u"},
1174 {Opt_barrier, "barrier"},
1175 {Opt_nobarrier, "nobarrier"},
1176 {Opt_i_version, "i_version"},
1177 {Opt_dax, "dax"},
1178 {Opt_stripe, "stripe=%u"},
1179 {Opt_delalloc, "delalloc"},
1180 {Opt_lazytime, "lazytime"},
1181 {Opt_nolazytime, "nolazytime"},
1182 {Opt_nodelalloc, "nodelalloc"},
1183 {Opt_removed, "mblk_io_submit"},
1184 {Opt_removed, "nomblk_io_submit"},
1185 {Opt_block_validity, "block_validity"},
1186 {Opt_noblock_validity, "noblock_validity"},
1187 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1188 {Opt_journal_ioprio, "journal_ioprio=%u"},
1189 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1190 {Opt_auto_da_alloc, "auto_da_alloc"},
1191 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1192 {Opt_dioread_nolock, "dioread_nolock"},
1193 {Opt_dioread_lock, "dioread_lock"},
1194 {Opt_discard, "discard"},
1195 {Opt_nodiscard, "nodiscard"},
1196 {Opt_init_itable, "init_itable=%u"},
1197 {Opt_init_itable, "init_itable"},
1198 {Opt_noinit_itable, "noinit_itable"},
1199 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1200 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1201 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1202 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1203 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1204 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1205 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1206 {Opt_err, NULL},
1207 };
1208
1209 static ext4_fsblk_t get_sb_block(void **data)
1210 {
1211 ext4_fsblk_t sb_block;
1212 char *options = (char *) *data;
1213
1214 if (!options || strncmp(options, "sb=", 3) != 0)
1215 return 1; /* Default location */
1216
1217 options += 3;
1218 /* TODO: use simple_strtoll with >32bit ext4 */
1219 sb_block = simple_strtoul(options, &options, 0);
1220 if (*options && *options != ',') {
1221 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1222 (char *) *data);
1223 return 1;
1224 }
1225 if (*options == ',')
1226 options++;
1227 *data = (void *) options;
1228
1229 return sb_block;
1230 }
1231
1232 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1233 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1234 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1235
1236 #ifdef CONFIG_QUOTA
1237 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1238 {
1239 struct ext4_sb_info *sbi = EXT4_SB(sb);
1240 char *qname;
1241 int ret = -1;
1242
1243 if (sb_any_quota_loaded(sb) &&
1244 !sbi->s_qf_names[qtype]) {
1245 ext4_msg(sb, KERN_ERR,
1246 "Cannot change journaled "
1247 "quota options when quota turned on");
1248 return -1;
1249 }
1250 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1251 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1252 "when QUOTA feature is enabled");
1253 return -1;
1254 }
1255 qname = match_strdup(args);
1256 if (!qname) {
1257 ext4_msg(sb, KERN_ERR,
1258 "Not enough memory for storing quotafile name");
1259 return -1;
1260 }
1261 if (sbi->s_qf_names[qtype]) {
1262 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1263 ret = 1;
1264 else
1265 ext4_msg(sb, KERN_ERR,
1266 "%s quota file already specified",
1267 QTYPE2NAME(qtype));
1268 goto errout;
1269 }
1270 if (strchr(qname, '/')) {
1271 ext4_msg(sb, KERN_ERR,
1272 "quotafile must be on filesystem root");
1273 goto errout;
1274 }
1275 sbi->s_qf_names[qtype] = qname;
1276 set_opt(sb, QUOTA);
1277 return 1;
1278 errout:
1279 kfree(qname);
1280 return ret;
1281 }
1282
1283 static int clear_qf_name(struct super_block *sb, int qtype)
1284 {
1285
1286 struct ext4_sb_info *sbi = EXT4_SB(sb);
1287
1288 if (sb_any_quota_loaded(sb) &&
1289 sbi->s_qf_names[qtype]) {
1290 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1291 " when quota turned on");
1292 return -1;
1293 }
1294 kfree(sbi->s_qf_names[qtype]);
1295 sbi->s_qf_names[qtype] = NULL;
1296 return 1;
1297 }
1298 #endif
1299
1300 #define MOPT_SET 0x0001
1301 #define MOPT_CLEAR 0x0002
1302 #define MOPT_NOSUPPORT 0x0004
1303 #define MOPT_EXPLICIT 0x0008
1304 #define MOPT_CLEAR_ERR 0x0010
1305 #define MOPT_GTE0 0x0020
1306 #ifdef CONFIG_QUOTA
1307 #define MOPT_Q 0
1308 #define MOPT_QFMT 0x0040
1309 #else
1310 #define MOPT_Q MOPT_NOSUPPORT
1311 #define MOPT_QFMT MOPT_NOSUPPORT
1312 #endif
1313 #define MOPT_DATAJ 0x0080
1314 #define MOPT_NO_EXT2 0x0100
1315 #define MOPT_NO_EXT3 0x0200
1316 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1317 #define MOPT_STRING 0x0400
1318
1319 static const struct mount_opts {
1320 int token;
1321 int mount_opt;
1322 int flags;
1323 } ext4_mount_opts[] = {
1324 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1325 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1326 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1327 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1328 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1329 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1330 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1331 MOPT_EXT4_ONLY | MOPT_SET},
1332 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1333 MOPT_EXT4_ONLY | MOPT_CLEAR},
1334 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1335 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1336 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1337 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1338 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1339 MOPT_EXT4_ONLY | MOPT_CLEAR},
1340 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1341 MOPT_EXT4_ONLY | MOPT_CLEAR},
1342 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1343 MOPT_EXT4_ONLY | MOPT_SET},
1344 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1345 EXT4_MOUNT_JOURNAL_CHECKSUM),
1346 MOPT_EXT4_ONLY | MOPT_SET},
1347 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1348 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1349 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1350 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1351 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1352 MOPT_NO_EXT2 | MOPT_SET},
1353 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1354 MOPT_NO_EXT2 | MOPT_CLEAR},
1355 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1356 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1357 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1358 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1359 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1360 {Opt_commit, 0, MOPT_GTE0},
1361 {Opt_max_batch_time, 0, MOPT_GTE0},
1362 {Opt_min_batch_time, 0, MOPT_GTE0},
1363 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1364 {Opt_init_itable, 0, MOPT_GTE0},
1365 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1366 {Opt_stripe, 0, MOPT_GTE0},
1367 {Opt_resuid, 0, MOPT_GTE0},
1368 {Opt_resgid, 0, MOPT_GTE0},
1369 {Opt_journal_dev, 0, MOPT_GTE0},
1370 {Opt_journal_path, 0, MOPT_STRING},
1371 {Opt_journal_ioprio, 0, MOPT_GTE0},
1372 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1373 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1374 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1375 MOPT_NO_EXT2 | MOPT_DATAJ},
1376 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1377 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1378 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1379 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1380 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1381 #else
1382 {Opt_acl, 0, MOPT_NOSUPPORT},
1383 {Opt_noacl, 0, MOPT_NOSUPPORT},
1384 #endif
1385 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1386 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1387 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1388 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1389 MOPT_SET | MOPT_Q},
1390 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1391 MOPT_SET | MOPT_Q},
1392 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1393 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1394 {Opt_usrjquota, 0, MOPT_Q},
1395 {Opt_grpjquota, 0, MOPT_Q},
1396 {Opt_offusrjquota, 0, MOPT_Q},
1397 {Opt_offgrpjquota, 0, MOPT_Q},
1398 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1399 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1400 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1401 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1402 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1403 {Opt_err, 0, 0}
1404 };
1405
1406 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1407 substring_t *args, unsigned long *journal_devnum,
1408 unsigned int *journal_ioprio, int is_remount)
1409 {
1410 struct ext4_sb_info *sbi = EXT4_SB(sb);
1411 const struct mount_opts *m;
1412 kuid_t uid;
1413 kgid_t gid;
1414 int arg = 0;
1415
1416 #ifdef CONFIG_QUOTA
1417 if (token == Opt_usrjquota)
1418 return set_qf_name(sb, USRQUOTA, &args[0]);
1419 else if (token == Opt_grpjquota)
1420 return set_qf_name(sb, GRPQUOTA, &args[0]);
1421 else if (token == Opt_offusrjquota)
1422 return clear_qf_name(sb, USRQUOTA);
1423 else if (token == Opt_offgrpjquota)
1424 return clear_qf_name(sb, GRPQUOTA);
1425 #endif
1426 switch (token) {
1427 case Opt_noacl:
1428 case Opt_nouser_xattr:
1429 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1430 break;
1431 case Opt_sb:
1432 return 1; /* handled by get_sb_block() */
1433 case Opt_removed:
1434 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1435 return 1;
1436 case Opt_abort:
1437 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1438 return 1;
1439 case Opt_i_version:
1440 sb->s_flags |= MS_I_VERSION;
1441 return 1;
1442 case Opt_lazytime:
1443 sb->s_flags |= MS_LAZYTIME;
1444 return 1;
1445 case Opt_nolazytime:
1446 sb->s_flags &= ~MS_LAZYTIME;
1447 return 1;
1448 }
1449
1450 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1451 if (token == m->token)
1452 break;
1453
1454 if (m->token == Opt_err) {
1455 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1456 "or missing value", opt);
1457 return -1;
1458 }
1459
1460 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1461 ext4_msg(sb, KERN_ERR,
1462 "Mount option \"%s\" incompatible with ext2", opt);
1463 return -1;
1464 }
1465 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1466 ext4_msg(sb, KERN_ERR,
1467 "Mount option \"%s\" incompatible with ext3", opt);
1468 return -1;
1469 }
1470
1471 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1472 return -1;
1473 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1474 return -1;
1475 if (m->flags & MOPT_EXPLICIT)
1476 set_opt2(sb, EXPLICIT_DELALLOC);
1477 if (m->flags & MOPT_CLEAR_ERR)
1478 clear_opt(sb, ERRORS_MASK);
1479 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1480 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1481 "options when quota turned on");
1482 return -1;
1483 }
1484
1485 if (m->flags & MOPT_NOSUPPORT) {
1486 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1487 } else if (token == Opt_commit) {
1488 if (arg == 0)
1489 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1490 sbi->s_commit_interval = HZ * arg;
1491 } else if (token == Opt_max_batch_time) {
1492 sbi->s_max_batch_time = arg;
1493 } else if (token == Opt_min_batch_time) {
1494 sbi->s_min_batch_time = arg;
1495 } else if (token == Opt_inode_readahead_blks) {
1496 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1497 ext4_msg(sb, KERN_ERR,
1498 "EXT4-fs: inode_readahead_blks must be "
1499 "0 or a power of 2 smaller than 2^31");
1500 return -1;
1501 }
1502 sbi->s_inode_readahead_blks = arg;
1503 } else if (token == Opt_init_itable) {
1504 set_opt(sb, INIT_INODE_TABLE);
1505 if (!args->from)
1506 arg = EXT4_DEF_LI_WAIT_MULT;
1507 sbi->s_li_wait_mult = arg;
1508 } else if (token == Opt_max_dir_size_kb) {
1509 sbi->s_max_dir_size_kb = arg;
1510 } else if (token == Opt_stripe) {
1511 sbi->s_stripe = arg;
1512 } else if (token == Opt_resuid) {
1513 uid = make_kuid(current_user_ns(), arg);
1514 if (!uid_valid(uid)) {
1515 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1516 return -1;
1517 }
1518 sbi->s_resuid = uid;
1519 } else if (token == Opt_resgid) {
1520 gid = make_kgid(current_user_ns(), arg);
1521 if (!gid_valid(gid)) {
1522 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1523 return -1;
1524 }
1525 sbi->s_resgid = gid;
1526 } else if (token == Opt_journal_dev) {
1527 if (is_remount) {
1528 ext4_msg(sb, KERN_ERR,
1529 "Cannot specify journal on remount");
1530 return -1;
1531 }
1532 *journal_devnum = arg;
1533 } else if (token == Opt_journal_path) {
1534 char *journal_path;
1535 struct inode *journal_inode;
1536 struct path path;
1537 int error;
1538
1539 if (is_remount) {
1540 ext4_msg(sb, KERN_ERR,
1541 "Cannot specify journal on remount");
1542 return -1;
1543 }
1544 journal_path = match_strdup(&args[0]);
1545 if (!journal_path) {
1546 ext4_msg(sb, KERN_ERR, "error: could not dup "
1547 "journal device string");
1548 return -1;
1549 }
1550
1551 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1552 if (error) {
1553 ext4_msg(sb, KERN_ERR, "error: could not find "
1554 "journal device path: error %d", error);
1555 kfree(journal_path);
1556 return -1;
1557 }
1558
1559 journal_inode = d_inode(path.dentry);
1560 if (!S_ISBLK(journal_inode->i_mode)) {
1561 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1562 "is not a block device", journal_path);
1563 path_put(&path);
1564 kfree(journal_path);
1565 return -1;
1566 }
1567
1568 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1569 path_put(&path);
1570 kfree(journal_path);
1571 } else if (token == Opt_journal_ioprio) {
1572 if (arg > 7) {
1573 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1574 " (must be 0-7)");
1575 return -1;
1576 }
1577 *journal_ioprio =
1578 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1579 } else if (token == Opt_test_dummy_encryption) {
1580 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1581 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1582 ext4_msg(sb, KERN_WARNING,
1583 "Test dummy encryption mode enabled");
1584 #else
1585 ext4_msg(sb, KERN_WARNING,
1586 "Test dummy encryption mount option ignored");
1587 #endif
1588 } else if (m->flags & MOPT_DATAJ) {
1589 if (is_remount) {
1590 if (!sbi->s_journal)
1591 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1592 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1593 ext4_msg(sb, KERN_ERR,
1594 "Cannot change data mode on remount");
1595 return -1;
1596 }
1597 } else {
1598 clear_opt(sb, DATA_FLAGS);
1599 sbi->s_mount_opt |= m->mount_opt;
1600 }
1601 #ifdef CONFIG_QUOTA
1602 } else if (m->flags & MOPT_QFMT) {
1603 if (sb_any_quota_loaded(sb) &&
1604 sbi->s_jquota_fmt != m->mount_opt) {
1605 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1606 "quota options when quota turned on");
1607 return -1;
1608 }
1609 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1610 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1611 ext4_msg(sb, KERN_ERR,
1612 "Cannot set journaled quota options "
1613 "when QUOTA feature is enabled");
1614 return -1;
1615 }
1616 sbi->s_jquota_fmt = m->mount_opt;
1617 #endif
1618 #ifndef CONFIG_FS_DAX
1619 } else if (token == Opt_dax) {
1620 ext4_msg(sb, KERN_INFO, "dax option not supported");
1621 return -1;
1622 #endif
1623 } else {
1624 if (!args->from)
1625 arg = 1;
1626 if (m->flags & MOPT_CLEAR)
1627 arg = !arg;
1628 else if (unlikely(!(m->flags & MOPT_SET))) {
1629 ext4_msg(sb, KERN_WARNING,
1630 "buggy handling of option %s", opt);
1631 WARN_ON(1);
1632 return -1;
1633 }
1634 if (arg != 0)
1635 sbi->s_mount_opt |= m->mount_opt;
1636 else
1637 sbi->s_mount_opt &= ~m->mount_opt;
1638 }
1639 return 1;
1640 }
1641
1642 static int parse_options(char *options, struct super_block *sb,
1643 unsigned long *journal_devnum,
1644 unsigned int *journal_ioprio,
1645 int is_remount)
1646 {
1647 struct ext4_sb_info *sbi = EXT4_SB(sb);
1648 char *p;
1649 substring_t args[MAX_OPT_ARGS];
1650 int token;
1651
1652 if (!options)
1653 return 1;
1654
1655 while ((p = strsep(&options, ",")) != NULL) {
1656 if (!*p)
1657 continue;
1658 /*
1659 * Initialize args struct so we know whether arg was
1660 * found; some options take optional arguments.
1661 */
1662 args[0].to = args[0].from = NULL;
1663 token = match_token(p, tokens, args);
1664 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1665 journal_ioprio, is_remount) < 0)
1666 return 0;
1667 }
1668 #ifdef CONFIG_QUOTA
1669 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1670 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1671 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1672 "feature is enabled");
1673 return 0;
1674 }
1675 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1676 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1677 clear_opt(sb, USRQUOTA);
1678
1679 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1680 clear_opt(sb, GRPQUOTA);
1681
1682 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1683 ext4_msg(sb, KERN_ERR, "old and new quota "
1684 "format mixing");
1685 return 0;
1686 }
1687
1688 if (!sbi->s_jquota_fmt) {
1689 ext4_msg(sb, KERN_ERR, "journaled quota format "
1690 "not specified");
1691 return 0;
1692 }
1693 }
1694 #endif
1695 if (test_opt(sb, DIOREAD_NOLOCK)) {
1696 int blocksize =
1697 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1698
1699 if (blocksize < PAGE_CACHE_SIZE) {
1700 ext4_msg(sb, KERN_ERR, "can't mount with "
1701 "dioread_nolock if block size != PAGE_SIZE");
1702 return 0;
1703 }
1704 }
1705 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1706 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1707 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1708 "in data=ordered mode");
1709 return 0;
1710 }
1711 return 1;
1712 }
1713
1714 static inline void ext4_show_quota_options(struct seq_file *seq,
1715 struct super_block *sb)
1716 {
1717 #if defined(CONFIG_QUOTA)
1718 struct ext4_sb_info *sbi = EXT4_SB(sb);
1719
1720 if (sbi->s_jquota_fmt) {
1721 char *fmtname = "";
1722
1723 switch (sbi->s_jquota_fmt) {
1724 case QFMT_VFS_OLD:
1725 fmtname = "vfsold";
1726 break;
1727 case QFMT_VFS_V0:
1728 fmtname = "vfsv0";
1729 break;
1730 case QFMT_VFS_V1:
1731 fmtname = "vfsv1";
1732 break;
1733 }
1734 seq_printf(seq, ",jqfmt=%s", fmtname);
1735 }
1736
1737 if (sbi->s_qf_names[USRQUOTA])
1738 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1739
1740 if (sbi->s_qf_names[GRPQUOTA])
1741 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1742 #endif
1743 }
1744
1745 static const char *token2str(int token)
1746 {
1747 const struct match_token *t;
1748
1749 for (t = tokens; t->token != Opt_err; t++)
1750 if (t->token == token && !strchr(t->pattern, '='))
1751 break;
1752 return t->pattern;
1753 }
1754
1755 /*
1756 * Show an option if
1757 * - it's set to a non-default value OR
1758 * - if the per-sb default is different from the global default
1759 */
1760 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1761 int nodefs)
1762 {
1763 struct ext4_sb_info *sbi = EXT4_SB(sb);
1764 struct ext4_super_block *es = sbi->s_es;
1765 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1766 const struct mount_opts *m;
1767 char sep = nodefs ? '\n' : ',';
1768
1769 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1770 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1771
1772 if (sbi->s_sb_block != 1)
1773 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1774
1775 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1776 int want_set = m->flags & MOPT_SET;
1777 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1778 (m->flags & MOPT_CLEAR_ERR))
1779 continue;
1780 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1781 continue; /* skip if same as the default */
1782 if ((want_set &&
1783 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1784 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1785 continue; /* select Opt_noFoo vs Opt_Foo */
1786 SEQ_OPTS_PRINT("%s", token2str(m->token));
1787 }
1788
1789 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1790 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1791 SEQ_OPTS_PRINT("resuid=%u",
1792 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1793 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1794 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1795 SEQ_OPTS_PRINT("resgid=%u",
1796 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1797 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1798 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1799 SEQ_OPTS_PUTS("errors=remount-ro");
1800 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1801 SEQ_OPTS_PUTS("errors=continue");
1802 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1803 SEQ_OPTS_PUTS("errors=panic");
1804 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1805 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1806 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1807 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1808 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1809 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1810 if (sb->s_flags & MS_I_VERSION)
1811 SEQ_OPTS_PUTS("i_version");
1812 if (nodefs || sbi->s_stripe)
1813 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1814 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1815 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1816 SEQ_OPTS_PUTS("data=journal");
1817 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1818 SEQ_OPTS_PUTS("data=ordered");
1819 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1820 SEQ_OPTS_PUTS("data=writeback");
1821 }
1822 if (nodefs ||
1823 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1824 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1825 sbi->s_inode_readahead_blks);
1826
1827 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1828 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1829 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1830 if (nodefs || sbi->s_max_dir_size_kb)
1831 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1832
1833 ext4_show_quota_options(seq, sb);
1834 return 0;
1835 }
1836
1837 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1838 {
1839 return _ext4_show_options(seq, root->d_sb, 0);
1840 }
1841
1842 static int options_seq_show(struct seq_file *seq, void *offset)
1843 {
1844 struct super_block *sb = seq->private;
1845 int rc;
1846
1847 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1848 rc = _ext4_show_options(seq, sb, 1);
1849 seq_puts(seq, "\n");
1850 return rc;
1851 }
1852
1853 static int options_open_fs(struct inode *inode, struct file *file)
1854 {
1855 return single_open(file, options_seq_show, PDE_DATA(inode));
1856 }
1857
1858 static const struct file_operations ext4_seq_options_fops = {
1859 .owner = THIS_MODULE,
1860 .open = options_open_fs,
1861 .read = seq_read,
1862 .llseek = seq_lseek,
1863 .release = single_release,
1864 };
1865
1866 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1867 int read_only)
1868 {
1869 struct ext4_sb_info *sbi = EXT4_SB(sb);
1870 int res = 0;
1871
1872 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1873 ext4_msg(sb, KERN_ERR, "revision level too high, "
1874 "forcing read-only mode");
1875 res = MS_RDONLY;
1876 }
1877 if (read_only)
1878 goto done;
1879 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1880 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1881 "running e2fsck is recommended");
1882 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1883 ext4_msg(sb, KERN_WARNING,
1884 "warning: mounting fs with errors, "
1885 "running e2fsck is recommended");
1886 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1887 le16_to_cpu(es->s_mnt_count) >=
1888 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1889 ext4_msg(sb, KERN_WARNING,
1890 "warning: maximal mount count reached, "
1891 "running e2fsck is recommended");
1892 else if (le32_to_cpu(es->s_checkinterval) &&
1893 (le32_to_cpu(es->s_lastcheck) +
1894 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1895 ext4_msg(sb, KERN_WARNING,
1896 "warning: checktime reached, "
1897 "running e2fsck is recommended");
1898 if (!sbi->s_journal)
1899 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1900 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1901 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1902 le16_add_cpu(&es->s_mnt_count, 1);
1903 es->s_mtime = cpu_to_le32(get_seconds());
1904 ext4_update_dynamic_rev(sb);
1905 if (sbi->s_journal)
1906 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1907
1908 ext4_commit_super(sb, 1);
1909 done:
1910 if (test_opt(sb, DEBUG))
1911 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1912 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1913 sb->s_blocksize,
1914 sbi->s_groups_count,
1915 EXT4_BLOCKS_PER_GROUP(sb),
1916 EXT4_INODES_PER_GROUP(sb),
1917 sbi->s_mount_opt, sbi->s_mount_opt2);
1918
1919 cleancache_init_fs(sb);
1920 return res;
1921 }
1922
1923 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1924 {
1925 struct ext4_sb_info *sbi = EXT4_SB(sb);
1926 struct flex_groups *new_groups;
1927 int size;
1928
1929 if (!sbi->s_log_groups_per_flex)
1930 return 0;
1931
1932 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1933 if (size <= sbi->s_flex_groups_allocated)
1934 return 0;
1935
1936 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1937 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1938 if (!new_groups) {
1939 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1940 size / (int) sizeof(struct flex_groups));
1941 return -ENOMEM;
1942 }
1943
1944 if (sbi->s_flex_groups) {
1945 memcpy(new_groups, sbi->s_flex_groups,
1946 (sbi->s_flex_groups_allocated *
1947 sizeof(struct flex_groups)));
1948 kvfree(sbi->s_flex_groups);
1949 }
1950 sbi->s_flex_groups = new_groups;
1951 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1952 return 0;
1953 }
1954
1955 static int ext4_fill_flex_info(struct super_block *sb)
1956 {
1957 struct ext4_sb_info *sbi = EXT4_SB(sb);
1958 struct ext4_group_desc *gdp = NULL;
1959 ext4_group_t flex_group;
1960 int i, err;
1961
1962 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1963 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1964 sbi->s_log_groups_per_flex = 0;
1965 return 1;
1966 }
1967
1968 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1969 if (err)
1970 goto failed;
1971
1972 for (i = 0; i < sbi->s_groups_count; i++) {
1973 gdp = ext4_get_group_desc(sb, i, NULL);
1974
1975 flex_group = ext4_flex_group(sbi, i);
1976 atomic_add(ext4_free_inodes_count(sb, gdp),
1977 &sbi->s_flex_groups[flex_group].free_inodes);
1978 atomic64_add(ext4_free_group_clusters(sb, gdp),
1979 &sbi->s_flex_groups[flex_group].free_clusters);
1980 atomic_add(ext4_used_dirs_count(sb, gdp),
1981 &sbi->s_flex_groups[flex_group].used_dirs);
1982 }
1983
1984 return 1;
1985 failed:
1986 return 0;
1987 }
1988
1989 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1990 struct ext4_group_desc *gdp)
1991 {
1992 int offset;
1993 __u16 crc = 0;
1994 __le32 le_group = cpu_to_le32(block_group);
1995
1996 if (ext4_has_metadata_csum(sbi->s_sb)) {
1997 /* Use new metadata_csum algorithm */
1998 __le16 save_csum;
1999 __u32 csum32;
2000
2001 save_csum = gdp->bg_checksum;
2002 gdp->bg_checksum = 0;
2003 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2004 sizeof(le_group));
2005 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2006 sbi->s_desc_size);
2007 gdp->bg_checksum = save_csum;
2008
2009 crc = csum32 & 0xFFFF;
2010 goto out;
2011 }
2012
2013 /* old crc16 code */
2014 if (!(sbi->s_es->s_feature_ro_compat &
2015 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2016 return 0;
2017
2018 offset = offsetof(struct ext4_group_desc, bg_checksum);
2019
2020 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2021 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2022 crc = crc16(crc, (__u8 *)gdp, offset);
2023 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2024 /* for checksum of struct ext4_group_desc do the rest...*/
2025 if ((sbi->s_es->s_feature_incompat &
2026 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2027 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2028 crc = crc16(crc, (__u8 *)gdp + offset,
2029 le16_to_cpu(sbi->s_es->s_desc_size) -
2030 offset);
2031
2032 out:
2033 return cpu_to_le16(crc);
2034 }
2035
2036 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2037 struct ext4_group_desc *gdp)
2038 {
2039 if (ext4_has_group_desc_csum(sb) &&
2040 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2041 block_group, gdp)))
2042 return 0;
2043
2044 return 1;
2045 }
2046
2047 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2048 struct ext4_group_desc *gdp)
2049 {
2050 if (!ext4_has_group_desc_csum(sb))
2051 return;
2052 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2053 }
2054
2055 /* Called at mount-time, super-block is locked */
2056 static int ext4_check_descriptors(struct super_block *sb,
2057 ext4_group_t *first_not_zeroed)
2058 {
2059 struct ext4_sb_info *sbi = EXT4_SB(sb);
2060 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2061 ext4_fsblk_t last_block;
2062 ext4_fsblk_t block_bitmap;
2063 ext4_fsblk_t inode_bitmap;
2064 ext4_fsblk_t inode_table;
2065 int flexbg_flag = 0;
2066 ext4_group_t i, grp = sbi->s_groups_count;
2067
2068 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2069 flexbg_flag = 1;
2070
2071 ext4_debug("Checking group descriptors");
2072
2073 for (i = 0; i < sbi->s_groups_count; i++) {
2074 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2075
2076 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2077 last_block = ext4_blocks_count(sbi->s_es) - 1;
2078 else
2079 last_block = first_block +
2080 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2081
2082 if ((grp == sbi->s_groups_count) &&
2083 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2084 grp = i;
2085
2086 block_bitmap = ext4_block_bitmap(sb, gdp);
2087 if (block_bitmap < first_block || block_bitmap > last_block) {
2088 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2089 "Block bitmap for group %u not in group "
2090 "(block %llu)!", i, block_bitmap);
2091 return 0;
2092 }
2093 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2094 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2095 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2096 "Inode bitmap for group %u not in group "
2097 "(block %llu)!", i, inode_bitmap);
2098 return 0;
2099 }
2100 inode_table = ext4_inode_table(sb, gdp);
2101 if (inode_table < first_block ||
2102 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2103 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2104 "Inode table for group %u not in group "
2105 "(block %llu)!", i, inode_table);
2106 return 0;
2107 }
2108 ext4_lock_group(sb, i);
2109 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2110 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2111 "Checksum for group %u failed (%u!=%u)",
2112 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2113 gdp)), le16_to_cpu(gdp->bg_checksum));
2114 if (!(sb->s_flags & MS_RDONLY)) {
2115 ext4_unlock_group(sb, i);
2116 return 0;
2117 }
2118 }
2119 ext4_unlock_group(sb, i);
2120 if (!flexbg_flag)
2121 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2122 }
2123 if (NULL != first_not_zeroed)
2124 *first_not_zeroed = grp;
2125 return 1;
2126 }
2127
2128 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2129 * the superblock) which were deleted from all directories, but held open by
2130 * a process at the time of a crash. We walk the list and try to delete these
2131 * inodes at recovery time (only with a read-write filesystem).
2132 *
2133 * In order to keep the orphan inode chain consistent during traversal (in
2134 * case of crash during recovery), we link each inode into the superblock
2135 * orphan list_head and handle it the same way as an inode deletion during
2136 * normal operation (which journals the operations for us).
2137 *
2138 * We only do an iget() and an iput() on each inode, which is very safe if we
2139 * accidentally point at an in-use or already deleted inode. The worst that
2140 * can happen in this case is that we get a "bit already cleared" message from
2141 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2142 * e2fsck was run on this filesystem, and it must have already done the orphan
2143 * inode cleanup for us, so we can safely abort without any further action.
2144 */
2145 static void ext4_orphan_cleanup(struct super_block *sb,
2146 struct ext4_super_block *es)
2147 {
2148 unsigned int s_flags = sb->s_flags;
2149 int nr_orphans = 0, nr_truncates = 0;
2150 #ifdef CONFIG_QUOTA
2151 int i;
2152 #endif
2153 if (!es->s_last_orphan) {
2154 jbd_debug(4, "no orphan inodes to clean up\n");
2155 return;
2156 }
2157
2158 if (bdev_read_only(sb->s_bdev)) {
2159 ext4_msg(sb, KERN_ERR, "write access "
2160 "unavailable, skipping orphan cleanup");
2161 return;
2162 }
2163
2164 /* Check if feature set would not allow a r/w mount */
2165 if (!ext4_feature_set_ok(sb, 0)) {
2166 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2167 "unknown ROCOMPAT features");
2168 return;
2169 }
2170
2171 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2172 /* don't clear list on RO mount w/ errors */
2173 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2174 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2175 "clearing orphan list.\n");
2176 es->s_last_orphan = 0;
2177 }
2178 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2179 return;
2180 }
2181
2182 if (s_flags & MS_RDONLY) {
2183 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2184 sb->s_flags &= ~MS_RDONLY;
2185 }
2186 #ifdef CONFIG_QUOTA
2187 /* Needed for iput() to work correctly and not trash data */
2188 sb->s_flags |= MS_ACTIVE;
2189 /* Turn on quotas so that they are updated correctly */
2190 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2191 if (EXT4_SB(sb)->s_qf_names[i]) {
2192 int ret = ext4_quota_on_mount(sb, i);
2193 if (ret < 0)
2194 ext4_msg(sb, KERN_ERR,
2195 "Cannot turn on journaled "
2196 "quota: error %d", ret);
2197 }
2198 }
2199 #endif
2200
2201 while (es->s_last_orphan) {
2202 struct inode *inode;
2203
2204 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2205 if (IS_ERR(inode)) {
2206 es->s_last_orphan = 0;
2207 break;
2208 }
2209
2210 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2211 dquot_initialize(inode);
2212 if (inode->i_nlink) {
2213 if (test_opt(sb, DEBUG))
2214 ext4_msg(sb, KERN_DEBUG,
2215 "%s: truncating inode %lu to %lld bytes",
2216 __func__, inode->i_ino, inode->i_size);
2217 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2218 inode->i_ino, inode->i_size);
2219 mutex_lock(&inode->i_mutex);
2220 truncate_inode_pages(inode->i_mapping, inode->i_size);
2221 ext4_truncate(inode);
2222 mutex_unlock(&inode->i_mutex);
2223 nr_truncates++;
2224 } else {
2225 if (test_opt(sb, DEBUG))
2226 ext4_msg(sb, KERN_DEBUG,
2227 "%s: deleting unreferenced inode %lu",
2228 __func__, inode->i_ino);
2229 jbd_debug(2, "deleting unreferenced inode %lu\n",
2230 inode->i_ino);
2231 nr_orphans++;
2232 }
2233 iput(inode); /* The delete magic happens here! */
2234 }
2235
2236 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2237
2238 if (nr_orphans)
2239 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2240 PLURAL(nr_orphans));
2241 if (nr_truncates)
2242 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2243 PLURAL(nr_truncates));
2244 #ifdef CONFIG_QUOTA
2245 /* Turn quotas off */
2246 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2247 if (sb_dqopt(sb)->files[i])
2248 dquot_quota_off(sb, i);
2249 }
2250 #endif
2251 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2252 }
2253
2254 /*
2255 * Maximal extent format file size.
2256 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2257 * extent format containers, within a sector_t, and within i_blocks
2258 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2259 * so that won't be a limiting factor.
2260 *
2261 * However there is other limiting factor. We do store extents in the form
2262 * of starting block and length, hence the resulting length of the extent
2263 * covering maximum file size must fit into on-disk format containers as
2264 * well. Given that length is always by 1 unit bigger than max unit (because
2265 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2266 *
2267 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2268 */
2269 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2270 {
2271 loff_t res;
2272 loff_t upper_limit = MAX_LFS_FILESIZE;
2273
2274 /* small i_blocks in vfs inode? */
2275 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2276 /*
2277 * CONFIG_LBDAF is not enabled implies the inode
2278 * i_block represent total blocks in 512 bytes
2279 * 32 == size of vfs inode i_blocks * 8
2280 */
2281 upper_limit = (1LL << 32) - 1;
2282
2283 /* total blocks in file system block size */
2284 upper_limit >>= (blkbits - 9);
2285 upper_limit <<= blkbits;
2286 }
2287
2288 /*
2289 * 32-bit extent-start container, ee_block. We lower the maxbytes
2290 * by one fs block, so ee_len can cover the extent of maximum file
2291 * size
2292 */
2293 res = (1LL << 32) - 1;
2294 res <<= blkbits;
2295
2296 /* Sanity check against vm- & vfs- imposed limits */
2297 if (res > upper_limit)
2298 res = upper_limit;
2299
2300 return res;
2301 }
2302
2303 /*
2304 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2305 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2306 * We need to be 1 filesystem block less than the 2^48 sector limit.
2307 */
2308 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2309 {
2310 loff_t res = EXT4_NDIR_BLOCKS;
2311 int meta_blocks;
2312 loff_t upper_limit;
2313 /* This is calculated to be the largest file size for a dense, block
2314 * mapped file such that the file's total number of 512-byte sectors,
2315 * including data and all indirect blocks, does not exceed (2^48 - 1).
2316 *
2317 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2318 * number of 512-byte sectors of the file.
2319 */
2320
2321 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2322 /*
2323 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2324 * the inode i_block field represents total file blocks in
2325 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2326 */
2327 upper_limit = (1LL << 32) - 1;
2328
2329 /* total blocks in file system block size */
2330 upper_limit >>= (bits - 9);
2331
2332 } else {
2333 /*
2334 * We use 48 bit ext4_inode i_blocks
2335 * With EXT4_HUGE_FILE_FL set the i_blocks
2336 * represent total number of blocks in
2337 * file system block size
2338 */
2339 upper_limit = (1LL << 48) - 1;
2340
2341 }
2342
2343 /* indirect blocks */
2344 meta_blocks = 1;
2345 /* double indirect blocks */
2346 meta_blocks += 1 + (1LL << (bits-2));
2347 /* tripple indirect blocks */
2348 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2349
2350 upper_limit -= meta_blocks;
2351 upper_limit <<= bits;
2352
2353 res += 1LL << (bits-2);
2354 res += 1LL << (2*(bits-2));
2355 res += 1LL << (3*(bits-2));
2356 res <<= bits;
2357 if (res > upper_limit)
2358 res = upper_limit;
2359
2360 if (res > MAX_LFS_FILESIZE)
2361 res = MAX_LFS_FILESIZE;
2362
2363 return res;
2364 }
2365
2366 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2367 ext4_fsblk_t logical_sb_block, int nr)
2368 {
2369 struct ext4_sb_info *sbi = EXT4_SB(sb);
2370 ext4_group_t bg, first_meta_bg;
2371 int has_super = 0;
2372
2373 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2374
2375 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2376 nr < first_meta_bg)
2377 return logical_sb_block + nr + 1;
2378 bg = sbi->s_desc_per_block * nr;
2379 if (ext4_bg_has_super(sb, bg))
2380 has_super = 1;
2381
2382 /*
2383 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2384 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2385 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2386 * compensate.
2387 */
2388 if (sb->s_blocksize == 1024 && nr == 0 &&
2389 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2390 has_super++;
2391
2392 return (has_super + ext4_group_first_block_no(sb, bg));
2393 }
2394
2395 /**
2396 * ext4_get_stripe_size: Get the stripe size.
2397 * @sbi: In memory super block info
2398 *
2399 * If we have specified it via mount option, then
2400 * use the mount option value. If the value specified at mount time is
2401 * greater than the blocks per group use the super block value.
2402 * If the super block value is greater than blocks per group return 0.
2403 * Allocator needs it be less than blocks per group.
2404 *
2405 */
2406 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2407 {
2408 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2409 unsigned long stripe_width =
2410 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2411 int ret;
2412
2413 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2414 ret = sbi->s_stripe;
2415 else if (stripe_width <= sbi->s_blocks_per_group)
2416 ret = stripe_width;
2417 else if (stride <= sbi->s_blocks_per_group)
2418 ret = stride;
2419 else
2420 ret = 0;
2421
2422 /*
2423 * If the stripe width is 1, this makes no sense and
2424 * we set it to 0 to turn off stripe handling code.
2425 */
2426 if (ret <= 1)
2427 ret = 0;
2428
2429 return ret;
2430 }
2431
2432 /* sysfs supprt */
2433
2434 struct ext4_attr {
2435 struct attribute attr;
2436 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2437 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2438 const char *, size_t);
2439 union {
2440 int offset;
2441 int deprecated_val;
2442 } u;
2443 };
2444
2445 static int parse_strtoull(const char *buf,
2446 unsigned long long max, unsigned long long *value)
2447 {
2448 int ret;
2449
2450 ret = kstrtoull(skip_spaces(buf), 0, value);
2451 if (!ret && *value > max)
2452 ret = -EINVAL;
2453 return ret;
2454 }
2455
2456 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2457 struct ext4_sb_info *sbi,
2458 char *buf)
2459 {
2460 return snprintf(buf, PAGE_SIZE, "%llu\n",
2461 (s64) EXT4_C2B(sbi,
2462 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2463 }
2464
2465 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2466 struct ext4_sb_info *sbi, char *buf)
2467 {
2468 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2469
2470 if (!sb->s_bdev->bd_part)
2471 return snprintf(buf, PAGE_SIZE, "0\n");
2472 return snprintf(buf, PAGE_SIZE, "%lu\n",
2473 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2474 sbi->s_sectors_written_start) >> 1);
2475 }
2476
2477 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2478 struct ext4_sb_info *sbi, char *buf)
2479 {
2480 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2481
2482 if (!sb->s_bdev->bd_part)
2483 return snprintf(buf, PAGE_SIZE, "0\n");
2484 return snprintf(buf, PAGE_SIZE, "%llu\n",
2485 (unsigned long long)(sbi->s_kbytes_written +
2486 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2487 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2488 }
2489
2490 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2491 struct ext4_sb_info *sbi,
2492 const char *buf, size_t count)
2493 {
2494 unsigned long t;
2495 int ret;
2496
2497 ret = kstrtoul(skip_spaces(buf), 0, &t);
2498 if (ret)
2499 return ret;
2500
2501 if (t && (!is_power_of_2(t) || t > 0x40000000))
2502 return -EINVAL;
2503
2504 sbi->s_inode_readahead_blks = t;
2505 return count;
2506 }
2507
2508 static ssize_t sbi_ui_show(struct ext4_attr *a,
2509 struct ext4_sb_info *sbi, char *buf)
2510 {
2511 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2512
2513 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2514 }
2515
2516 static ssize_t sbi_ui_store(struct ext4_attr *a,
2517 struct ext4_sb_info *sbi,
2518 const char *buf, size_t count)
2519 {
2520 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2521 unsigned long t;
2522 int ret;
2523
2524 ret = kstrtoul(skip_spaces(buf), 0, &t);
2525 if (ret)
2526 return ret;
2527 *ui = t;
2528 return count;
2529 }
2530
2531 static ssize_t es_ui_show(struct ext4_attr *a,
2532 struct ext4_sb_info *sbi, char *buf)
2533 {
2534
2535 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2536 a->u.offset);
2537
2538 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2539 }
2540
2541 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2542 struct ext4_sb_info *sbi, char *buf)
2543 {
2544 return snprintf(buf, PAGE_SIZE, "%llu\n",
2545 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2546 }
2547
2548 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2549 struct ext4_sb_info *sbi,
2550 const char *buf, size_t count)
2551 {
2552 unsigned long long val;
2553 int ret;
2554
2555 if (parse_strtoull(buf, -1ULL, &val))
2556 return -EINVAL;
2557 ret = ext4_reserve_clusters(sbi, val);
2558
2559 return ret ? ret : count;
2560 }
2561
2562 static ssize_t trigger_test_error(struct ext4_attr *a,
2563 struct ext4_sb_info *sbi,
2564 const char *buf, size_t count)
2565 {
2566 int len = count;
2567
2568 if (!capable(CAP_SYS_ADMIN))
2569 return -EPERM;
2570
2571 if (len && buf[len-1] == '\n')
2572 len--;
2573
2574 if (len)
2575 ext4_error(sbi->s_sb, "%.*s", len, buf);
2576 return count;
2577 }
2578
2579 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2580 struct ext4_sb_info *sbi, char *buf)
2581 {
2582 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2583 }
2584
2585 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2586 static struct ext4_attr ext4_attr_##_name = { \
2587 .attr = {.name = __stringify(_name), .mode = _mode }, \
2588 .show = _show, \
2589 .store = _store, \
2590 .u = { \
2591 .offset = offsetof(struct ext4_sb_info, _elname),\
2592 }, \
2593 }
2594
2595 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2596 static struct ext4_attr ext4_attr_##_name = { \
2597 .attr = {.name = __stringify(_name), .mode = _mode }, \
2598 .show = _show, \
2599 .store = _store, \
2600 .u = { \
2601 .offset = offsetof(struct ext4_super_block, _elname), \
2602 }, \
2603 }
2604
2605 #define EXT4_ATTR(name, mode, show, store) \
2606 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2607
2608 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2609 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2610 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2611
2612 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2613 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2614 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2615 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2616
2617 #define ATTR_LIST(name) &ext4_attr_##name.attr
2618 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2619 static struct ext4_attr ext4_attr_##_name = { \
2620 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2621 .show = sbi_deprecated_show, \
2622 .u = { \
2623 .deprecated_val = _val, \
2624 }, \
2625 }
2626
2627 EXT4_RO_ATTR(delayed_allocation_blocks);
2628 EXT4_RO_ATTR(session_write_kbytes);
2629 EXT4_RO_ATTR(lifetime_write_kbytes);
2630 EXT4_RW_ATTR(reserved_clusters);
2631 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2632 inode_readahead_blks_store, s_inode_readahead_blks);
2633 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2634 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2635 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2636 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2637 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2638 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2639 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2640 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2641 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2642 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2643 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2644 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2645 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2646 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2647 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2648 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2649 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2650 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2651 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2652
2653 static struct attribute *ext4_attrs[] = {
2654 ATTR_LIST(delayed_allocation_blocks),
2655 ATTR_LIST(session_write_kbytes),
2656 ATTR_LIST(lifetime_write_kbytes),
2657 ATTR_LIST(reserved_clusters),
2658 ATTR_LIST(inode_readahead_blks),
2659 ATTR_LIST(inode_goal),
2660 ATTR_LIST(mb_stats),
2661 ATTR_LIST(mb_max_to_scan),
2662 ATTR_LIST(mb_min_to_scan),
2663 ATTR_LIST(mb_order2_req),
2664 ATTR_LIST(mb_stream_req),
2665 ATTR_LIST(mb_group_prealloc),
2666 ATTR_LIST(max_writeback_mb_bump),
2667 ATTR_LIST(extent_max_zeroout_kb),
2668 ATTR_LIST(trigger_fs_error),
2669 ATTR_LIST(err_ratelimit_interval_ms),
2670 ATTR_LIST(err_ratelimit_burst),
2671 ATTR_LIST(warning_ratelimit_interval_ms),
2672 ATTR_LIST(warning_ratelimit_burst),
2673 ATTR_LIST(msg_ratelimit_interval_ms),
2674 ATTR_LIST(msg_ratelimit_burst),
2675 ATTR_LIST(errors_count),
2676 ATTR_LIST(first_error_time),
2677 ATTR_LIST(last_error_time),
2678 NULL,
2679 };
2680
2681 /* Features this copy of ext4 supports */
2682 EXT4_INFO_ATTR(lazy_itable_init);
2683 EXT4_INFO_ATTR(batched_discard);
2684 EXT4_INFO_ATTR(meta_bg_resize);
2685 EXT4_INFO_ATTR(encryption);
2686
2687 static struct attribute *ext4_feat_attrs[] = {
2688 ATTR_LIST(lazy_itable_init),
2689 ATTR_LIST(batched_discard),
2690 ATTR_LIST(meta_bg_resize),
2691 ATTR_LIST(encryption),
2692 NULL,
2693 };
2694
2695 static ssize_t ext4_attr_show(struct kobject *kobj,
2696 struct attribute *attr, char *buf)
2697 {
2698 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2699 s_kobj);
2700 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2701
2702 return a->show ? a->show(a, sbi, buf) : 0;
2703 }
2704
2705 static ssize_t ext4_attr_store(struct kobject *kobj,
2706 struct attribute *attr,
2707 const char *buf, size_t len)
2708 {
2709 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2710 s_kobj);
2711 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2712
2713 return a->store ? a->store(a, sbi, buf, len) : 0;
2714 }
2715
2716 static void ext4_sb_release(struct kobject *kobj)
2717 {
2718 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2719 s_kobj);
2720 complete(&sbi->s_kobj_unregister);
2721 }
2722
2723 static const struct sysfs_ops ext4_attr_ops = {
2724 .show = ext4_attr_show,
2725 .store = ext4_attr_store,
2726 };
2727
2728 static struct kobj_type ext4_ktype = {
2729 .default_attrs = ext4_attrs,
2730 .sysfs_ops = &ext4_attr_ops,
2731 .release = ext4_sb_release,
2732 };
2733
2734 static void ext4_feat_release(struct kobject *kobj)
2735 {
2736 complete(&ext4_feat->f_kobj_unregister);
2737 }
2738
2739 static ssize_t ext4_feat_show(struct kobject *kobj,
2740 struct attribute *attr, char *buf)
2741 {
2742 return snprintf(buf, PAGE_SIZE, "supported\n");
2743 }
2744
2745 /*
2746 * We can not use ext4_attr_show/store because it relies on the kobject
2747 * being embedded in the ext4_sb_info structure which is definitely not
2748 * true in this case.
2749 */
2750 static const struct sysfs_ops ext4_feat_ops = {
2751 .show = ext4_feat_show,
2752 .store = NULL,
2753 };
2754
2755 static struct kobj_type ext4_feat_ktype = {
2756 .default_attrs = ext4_feat_attrs,
2757 .sysfs_ops = &ext4_feat_ops,
2758 .release = ext4_feat_release,
2759 };
2760
2761 /*
2762 * Check whether this filesystem can be mounted based on
2763 * the features present and the RDONLY/RDWR mount requested.
2764 * Returns 1 if this filesystem can be mounted as requested,
2765 * 0 if it cannot be.
2766 */
2767 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2768 {
2769 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2770 ext4_msg(sb, KERN_ERR,
2771 "Couldn't mount because of "
2772 "unsupported optional features (%x)",
2773 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2774 ~EXT4_FEATURE_INCOMPAT_SUPP));
2775 return 0;
2776 }
2777
2778 if (readonly)
2779 return 1;
2780
2781 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2782 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2783 sb->s_flags |= MS_RDONLY;
2784 return 1;
2785 }
2786
2787 /* Check that feature set is OK for a read-write mount */
2788 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2789 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2790 "unsupported optional features (%x)",
2791 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2792 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2793 return 0;
2794 }
2795 /*
2796 * Large file size enabled file system can only be mounted
2797 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2798 */
2799 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2800 if (sizeof(blkcnt_t) < sizeof(u64)) {
2801 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2802 "cannot be mounted RDWR without "
2803 "CONFIG_LBDAF");
2804 return 0;
2805 }
2806 }
2807 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2808 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2809 ext4_msg(sb, KERN_ERR,
2810 "Can't support bigalloc feature without "
2811 "extents feature\n");
2812 return 0;
2813 }
2814
2815 #ifndef CONFIG_QUOTA
2816 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2817 !readonly) {
2818 ext4_msg(sb, KERN_ERR,
2819 "Filesystem with quota feature cannot be mounted RDWR "
2820 "without CONFIG_QUOTA");
2821 return 0;
2822 }
2823 #endif /* CONFIG_QUOTA */
2824 return 1;
2825 }
2826
2827 /*
2828 * This function is called once a day if we have errors logged
2829 * on the file system
2830 */
2831 static void print_daily_error_info(unsigned long arg)
2832 {
2833 struct super_block *sb = (struct super_block *) arg;
2834 struct ext4_sb_info *sbi;
2835 struct ext4_super_block *es;
2836
2837 sbi = EXT4_SB(sb);
2838 es = sbi->s_es;
2839
2840 if (es->s_error_count)
2841 /* fsck newer than v1.41.13 is needed to clean this condition. */
2842 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2843 le32_to_cpu(es->s_error_count));
2844 if (es->s_first_error_time) {
2845 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2846 sb->s_id, le32_to_cpu(es->s_first_error_time),
2847 (int) sizeof(es->s_first_error_func),
2848 es->s_first_error_func,
2849 le32_to_cpu(es->s_first_error_line));
2850 if (es->s_first_error_ino)
2851 printk(": inode %u",
2852 le32_to_cpu(es->s_first_error_ino));
2853 if (es->s_first_error_block)
2854 printk(": block %llu", (unsigned long long)
2855 le64_to_cpu(es->s_first_error_block));
2856 printk("\n");
2857 }
2858 if (es->s_last_error_time) {
2859 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2860 sb->s_id, le32_to_cpu(es->s_last_error_time),
2861 (int) sizeof(es->s_last_error_func),
2862 es->s_last_error_func,
2863 le32_to_cpu(es->s_last_error_line));
2864 if (es->s_last_error_ino)
2865 printk(": inode %u",
2866 le32_to_cpu(es->s_last_error_ino));
2867 if (es->s_last_error_block)
2868 printk(": block %llu", (unsigned long long)
2869 le64_to_cpu(es->s_last_error_block));
2870 printk("\n");
2871 }
2872 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2873 }
2874
2875 /* Find next suitable group and run ext4_init_inode_table */
2876 static int ext4_run_li_request(struct ext4_li_request *elr)
2877 {
2878 struct ext4_group_desc *gdp = NULL;
2879 ext4_group_t group, ngroups;
2880 struct super_block *sb;
2881 unsigned long timeout = 0;
2882 int ret = 0;
2883
2884 sb = elr->lr_super;
2885 ngroups = EXT4_SB(sb)->s_groups_count;
2886
2887 sb_start_write(sb);
2888 for (group = elr->lr_next_group; group < ngroups; group++) {
2889 gdp = ext4_get_group_desc(sb, group, NULL);
2890 if (!gdp) {
2891 ret = 1;
2892 break;
2893 }
2894
2895 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2896 break;
2897 }
2898
2899 if (group >= ngroups)
2900 ret = 1;
2901
2902 if (!ret) {
2903 timeout = jiffies;
2904 ret = ext4_init_inode_table(sb, group,
2905 elr->lr_timeout ? 0 : 1);
2906 if (elr->lr_timeout == 0) {
2907 timeout = (jiffies - timeout) *
2908 elr->lr_sbi->s_li_wait_mult;
2909 elr->lr_timeout = timeout;
2910 }
2911 elr->lr_next_sched = jiffies + elr->lr_timeout;
2912 elr->lr_next_group = group + 1;
2913 }
2914 sb_end_write(sb);
2915
2916 return ret;
2917 }
2918
2919 /*
2920 * Remove lr_request from the list_request and free the
2921 * request structure. Should be called with li_list_mtx held
2922 */
2923 static void ext4_remove_li_request(struct ext4_li_request *elr)
2924 {
2925 struct ext4_sb_info *sbi;
2926
2927 if (!elr)
2928 return;
2929
2930 sbi = elr->lr_sbi;
2931
2932 list_del(&elr->lr_request);
2933 sbi->s_li_request = NULL;
2934 kfree(elr);
2935 }
2936
2937 static void ext4_unregister_li_request(struct super_block *sb)
2938 {
2939 mutex_lock(&ext4_li_mtx);
2940 if (!ext4_li_info) {
2941 mutex_unlock(&ext4_li_mtx);
2942 return;
2943 }
2944
2945 mutex_lock(&ext4_li_info->li_list_mtx);
2946 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2947 mutex_unlock(&ext4_li_info->li_list_mtx);
2948 mutex_unlock(&ext4_li_mtx);
2949 }
2950
2951 static struct task_struct *ext4_lazyinit_task;
2952
2953 /*
2954 * This is the function where ext4lazyinit thread lives. It walks
2955 * through the request list searching for next scheduled filesystem.
2956 * When such a fs is found, run the lazy initialization request
2957 * (ext4_rn_li_request) and keep track of the time spend in this
2958 * function. Based on that time we compute next schedule time of
2959 * the request. When walking through the list is complete, compute
2960 * next waking time and put itself into sleep.
2961 */
2962 static int ext4_lazyinit_thread(void *arg)
2963 {
2964 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2965 struct list_head *pos, *n;
2966 struct ext4_li_request *elr;
2967 unsigned long next_wakeup, cur;
2968
2969 BUG_ON(NULL == eli);
2970
2971 cont_thread:
2972 while (true) {
2973 next_wakeup = MAX_JIFFY_OFFSET;
2974
2975 mutex_lock(&eli->li_list_mtx);
2976 if (list_empty(&eli->li_request_list)) {
2977 mutex_unlock(&eli->li_list_mtx);
2978 goto exit_thread;
2979 }
2980
2981 list_for_each_safe(pos, n, &eli->li_request_list) {
2982 elr = list_entry(pos, struct ext4_li_request,
2983 lr_request);
2984
2985 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2986 if (ext4_run_li_request(elr) != 0) {
2987 /* error, remove the lazy_init job */
2988 ext4_remove_li_request(elr);
2989 continue;
2990 }
2991 }
2992
2993 if (time_before(elr->lr_next_sched, next_wakeup))
2994 next_wakeup = elr->lr_next_sched;
2995 }
2996 mutex_unlock(&eli->li_list_mtx);
2997
2998 try_to_freeze();
2999
3000 cur = jiffies;
3001 if ((time_after_eq(cur, next_wakeup)) ||
3002 (MAX_JIFFY_OFFSET == next_wakeup)) {
3003 cond_resched();
3004 continue;
3005 }
3006
3007 schedule_timeout_interruptible(next_wakeup - cur);
3008
3009 if (kthread_should_stop()) {
3010 ext4_clear_request_list();
3011 goto exit_thread;
3012 }
3013 }
3014
3015 exit_thread:
3016 /*
3017 * It looks like the request list is empty, but we need
3018 * to check it under the li_list_mtx lock, to prevent any
3019 * additions into it, and of course we should lock ext4_li_mtx
3020 * to atomically free the list and ext4_li_info, because at
3021 * this point another ext4 filesystem could be registering
3022 * new one.
3023 */
3024 mutex_lock(&ext4_li_mtx);
3025 mutex_lock(&eli->li_list_mtx);
3026 if (!list_empty(&eli->li_request_list)) {
3027 mutex_unlock(&eli->li_list_mtx);
3028 mutex_unlock(&ext4_li_mtx);
3029 goto cont_thread;
3030 }
3031 mutex_unlock(&eli->li_list_mtx);
3032 kfree(ext4_li_info);
3033 ext4_li_info = NULL;
3034 mutex_unlock(&ext4_li_mtx);
3035
3036 return 0;
3037 }
3038
3039 static void ext4_clear_request_list(void)
3040 {
3041 struct list_head *pos, *n;
3042 struct ext4_li_request *elr;
3043
3044 mutex_lock(&ext4_li_info->li_list_mtx);
3045 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3046 elr = list_entry(pos, struct ext4_li_request,
3047 lr_request);
3048 ext4_remove_li_request(elr);
3049 }
3050 mutex_unlock(&ext4_li_info->li_list_mtx);
3051 }
3052
3053 static int ext4_run_lazyinit_thread(void)
3054 {
3055 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3056 ext4_li_info, "ext4lazyinit");
3057 if (IS_ERR(ext4_lazyinit_task)) {
3058 int err = PTR_ERR(ext4_lazyinit_task);
3059 ext4_clear_request_list();
3060 kfree(ext4_li_info);
3061 ext4_li_info = NULL;
3062 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3063 "initialization thread\n",
3064 err);
3065 return err;
3066 }
3067 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3068 return 0;
3069 }
3070
3071 /*
3072 * Check whether it make sense to run itable init. thread or not.
3073 * If there is at least one uninitialized inode table, return
3074 * corresponding group number, else the loop goes through all
3075 * groups and return total number of groups.
3076 */
3077 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3078 {
3079 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3080 struct ext4_group_desc *gdp = NULL;
3081
3082 for (group = 0; group < ngroups; group++) {
3083 gdp = ext4_get_group_desc(sb, group, NULL);
3084 if (!gdp)
3085 continue;
3086
3087 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3088 break;
3089 }
3090
3091 return group;
3092 }
3093
3094 static int ext4_li_info_new(void)
3095 {
3096 struct ext4_lazy_init *eli = NULL;
3097
3098 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3099 if (!eli)
3100 return -ENOMEM;
3101
3102 INIT_LIST_HEAD(&eli->li_request_list);
3103 mutex_init(&eli->li_list_mtx);
3104
3105 eli->li_state |= EXT4_LAZYINIT_QUIT;
3106
3107 ext4_li_info = eli;
3108
3109 return 0;
3110 }
3111
3112 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3113 ext4_group_t start)
3114 {
3115 struct ext4_sb_info *sbi = EXT4_SB(sb);
3116 struct ext4_li_request *elr;
3117
3118 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3119 if (!elr)
3120 return NULL;
3121
3122 elr->lr_super = sb;
3123 elr->lr_sbi = sbi;
3124 elr->lr_next_group = start;
3125
3126 /*
3127 * Randomize first schedule time of the request to
3128 * spread the inode table initialization requests
3129 * better.
3130 */
3131 elr->lr_next_sched = jiffies + (prandom_u32() %
3132 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3133 return elr;
3134 }
3135
3136 int ext4_register_li_request(struct super_block *sb,
3137 ext4_group_t first_not_zeroed)
3138 {
3139 struct ext4_sb_info *sbi = EXT4_SB(sb);
3140 struct ext4_li_request *elr = NULL;
3141 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3142 int ret = 0;
3143
3144 mutex_lock(&ext4_li_mtx);
3145 if (sbi->s_li_request != NULL) {
3146 /*
3147 * Reset timeout so it can be computed again, because
3148 * s_li_wait_mult might have changed.
3149 */
3150 sbi->s_li_request->lr_timeout = 0;
3151 goto out;
3152 }
3153
3154 if (first_not_zeroed == ngroups ||
3155 (sb->s_flags & MS_RDONLY) ||
3156 !test_opt(sb, INIT_INODE_TABLE))
3157 goto out;
3158
3159 elr = ext4_li_request_new(sb, first_not_zeroed);
3160 if (!elr) {
3161 ret = -ENOMEM;
3162 goto out;
3163 }
3164
3165 if (NULL == ext4_li_info) {
3166 ret = ext4_li_info_new();
3167 if (ret)
3168 goto out;
3169 }
3170
3171 mutex_lock(&ext4_li_info->li_list_mtx);
3172 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3173 mutex_unlock(&ext4_li_info->li_list_mtx);
3174
3175 sbi->s_li_request = elr;
3176 /*
3177 * set elr to NULL here since it has been inserted to
3178 * the request_list and the removal and free of it is
3179 * handled by ext4_clear_request_list from now on.
3180 */
3181 elr = NULL;
3182
3183 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3184 ret = ext4_run_lazyinit_thread();
3185 if (ret)
3186 goto out;
3187 }
3188 out:
3189 mutex_unlock(&ext4_li_mtx);
3190 if (ret)
3191 kfree(elr);
3192 return ret;
3193 }
3194
3195 /*
3196 * We do not need to lock anything since this is called on
3197 * module unload.
3198 */
3199 static void ext4_destroy_lazyinit_thread(void)
3200 {
3201 /*
3202 * If thread exited earlier
3203 * there's nothing to be done.
3204 */
3205 if (!ext4_li_info || !ext4_lazyinit_task)
3206 return;
3207
3208 kthread_stop(ext4_lazyinit_task);
3209 }
3210
3211 static int set_journal_csum_feature_set(struct super_block *sb)
3212 {
3213 int ret = 1;
3214 int compat, incompat;
3215 struct ext4_sb_info *sbi = EXT4_SB(sb);
3216
3217 if (ext4_has_metadata_csum(sb)) {
3218 /* journal checksum v3 */
3219 compat = 0;
3220 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3221 } else {
3222 /* journal checksum v1 */
3223 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3224 incompat = 0;
3225 }
3226
3227 jbd2_journal_clear_features(sbi->s_journal,
3228 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3229 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3230 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3231 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3232 ret = jbd2_journal_set_features(sbi->s_journal,
3233 compat, 0,
3234 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3235 incompat);
3236 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3237 ret = jbd2_journal_set_features(sbi->s_journal,
3238 compat, 0,
3239 incompat);
3240 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3241 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3242 } else {
3243 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3244 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3245 }
3246
3247 return ret;
3248 }
3249
3250 /*
3251 * Note: calculating the overhead so we can be compatible with
3252 * historical BSD practice is quite difficult in the face of
3253 * clusters/bigalloc. This is because multiple metadata blocks from
3254 * different block group can end up in the same allocation cluster.
3255 * Calculating the exact overhead in the face of clustered allocation
3256 * requires either O(all block bitmaps) in memory or O(number of block
3257 * groups**2) in time. We will still calculate the superblock for
3258 * older file systems --- and if we come across with a bigalloc file
3259 * system with zero in s_overhead_clusters the estimate will be close to
3260 * correct especially for very large cluster sizes --- but for newer
3261 * file systems, it's better to calculate this figure once at mkfs
3262 * time, and store it in the superblock. If the superblock value is
3263 * present (even for non-bigalloc file systems), we will use it.
3264 */
3265 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3266 char *buf)
3267 {
3268 struct ext4_sb_info *sbi = EXT4_SB(sb);
3269 struct ext4_group_desc *gdp;
3270 ext4_fsblk_t first_block, last_block, b;
3271 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3272 int s, j, count = 0;
3273
3274 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3275 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3276 sbi->s_itb_per_group + 2);
3277
3278 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3279 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3280 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3281 for (i = 0; i < ngroups; i++) {
3282 gdp = ext4_get_group_desc(sb, i, NULL);
3283 b = ext4_block_bitmap(sb, gdp);
3284 if (b >= first_block && b <= last_block) {
3285 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3286 count++;
3287 }
3288 b = ext4_inode_bitmap(sb, gdp);
3289 if (b >= first_block && b <= last_block) {
3290 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3291 count++;
3292 }
3293 b = ext4_inode_table(sb, gdp);
3294 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3295 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3296 int c = EXT4_B2C(sbi, b - first_block);
3297 ext4_set_bit(c, buf);
3298 count++;
3299 }
3300 if (i != grp)
3301 continue;
3302 s = 0;
3303 if (ext4_bg_has_super(sb, grp)) {
3304 ext4_set_bit(s++, buf);
3305 count++;
3306 }
3307 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3308 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3309 count++;
3310 }
3311 }
3312 if (!count)
3313 return 0;
3314 return EXT4_CLUSTERS_PER_GROUP(sb) -
3315 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3316 }
3317
3318 /*
3319 * Compute the overhead and stash it in sbi->s_overhead
3320 */
3321 int ext4_calculate_overhead(struct super_block *sb)
3322 {
3323 struct ext4_sb_info *sbi = EXT4_SB(sb);
3324 struct ext4_super_block *es = sbi->s_es;
3325 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3326 ext4_fsblk_t overhead = 0;
3327 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3328
3329 if (!buf)
3330 return -ENOMEM;
3331
3332 /*
3333 * Compute the overhead (FS structures). This is constant
3334 * for a given filesystem unless the number of block groups
3335 * changes so we cache the previous value until it does.
3336 */
3337
3338 /*
3339 * All of the blocks before first_data_block are overhead
3340 */
3341 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3342
3343 /*
3344 * Add the overhead found in each block group
3345 */
3346 for (i = 0; i < ngroups; i++) {
3347 int blks;
3348
3349 blks = count_overhead(sb, i, buf);
3350 overhead += blks;
3351 if (blks)
3352 memset(buf, 0, PAGE_SIZE);
3353 cond_resched();
3354 }
3355 /* Add the internal journal blocks as well */
3356 if (sbi->s_journal && !sbi->journal_bdev)
3357 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3358
3359 sbi->s_overhead = overhead;
3360 smp_wmb();
3361 free_page((unsigned long) buf);
3362 return 0;
3363 }
3364
3365
3366 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3367 {
3368 ext4_fsblk_t resv_clusters;
3369
3370 /*
3371 * There's no need to reserve anything when we aren't using extents.
3372 * The space estimates are exact, there are no unwritten extents,
3373 * hole punching doesn't need new metadata... This is needed especially
3374 * to keep ext2/3 backward compatibility.
3375 */
3376 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3377 return 0;
3378 /*
3379 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3380 * This should cover the situations where we can not afford to run
3381 * out of space like for example punch hole, or converting
3382 * unwritten extents in delalloc path. In most cases such
3383 * allocation would require 1, or 2 blocks, higher numbers are
3384 * very rare.
3385 */
3386 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3387 EXT4_SB(sb)->s_cluster_bits;
3388
3389 do_div(resv_clusters, 50);
3390 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3391
3392 return resv_clusters;
3393 }
3394
3395
3396 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3397 {
3398 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3399 sbi->s_cluster_bits;
3400
3401 if (count >= clusters)
3402 return -EINVAL;
3403
3404 atomic64_set(&sbi->s_resv_clusters, count);
3405 return 0;
3406 }
3407
3408 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3409 {
3410 char *orig_data = kstrdup(data, GFP_KERNEL);
3411 struct buffer_head *bh;
3412 struct ext4_super_block *es = NULL;
3413 struct ext4_sb_info *sbi;
3414 ext4_fsblk_t block;
3415 ext4_fsblk_t sb_block = get_sb_block(&data);
3416 ext4_fsblk_t logical_sb_block;
3417 unsigned long offset = 0;
3418 unsigned long journal_devnum = 0;
3419 unsigned long def_mount_opts;
3420 struct inode *root;
3421 char *cp;
3422 const char *descr;
3423 int ret = -ENOMEM;
3424 int blocksize, clustersize;
3425 unsigned int db_count;
3426 unsigned int i;
3427 int needs_recovery, has_huge_files, has_bigalloc;
3428 __u64 blocks_count;
3429 int err = 0;
3430 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3431 ext4_group_t first_not_zeroed;
3432
3433 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3434 if (!sbi)
3435 goto out_free_orig;
3436
3437 sbi->s_blockgroup_lock =
3438 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3439 if (!sbi->s_blockgroup_lock) {
3440 kfree(sbi);
3441 goto out_free_orig;
3442 }
3443 sb->s_fs_info = sbi;
3444 sbi->s_sb = sb;
3445 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3446 sbi->s_sb_block = sb_block;
3447 if (sb->s_bdev->bd_part)
3448 sbi->s_sectors_written_start =
3449 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3450 #ifdef CONFIG_EXT4_FS_ENCRYPTION
3451 /* Modes of operations for file and directory encryption. */
3452 sbi->s_file_encryption_mode = EXT4_ENCRYPTION_MODE_AES_256_XTS;
3453 sbi->s_dir_encryption_mode = EXT4_ENCRYPTION_MODE_INVALID;
3454 #endif
3455
3456 /* Cleanup superblock name */
3457 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3458 *cp = '!';
3459
3460 /* -EINVAL is default */
3461 ret = -EINVAL;
3462 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3463 if (!blocksize) {
3464 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3465 goto out_fail;
3466 }
3467
3468 /*
3469 * The ext4 superblock will not be buffer aligned for other than 1kB
3470 * block sizes. We need to calculate the offset from buffer start.
3471 */
3472 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3473 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3474 offset = do_div(logical_sb_block, blocksize);
3475 } else {
3476 logical_sb_block = sb_block;
3477 }
3478
3479 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3480 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3481 goto out_fail;
3482 }
3483 /*
3484 * Note: s_es must be initialized as soon as possible because
3485 * some ext4 macro-instructions depend on its value
3486 */
3487 es = (struct ext4_super_block *) (bh->b_data + offset);
3488 sbi->s_es = es;
3489 sb->s_magic = le16_to_cpu(es->s_magic);
3490 if (sb->s_magic != EXT4_SUPER_MAGIC)
3491 goto cantfind_ext4;
3492 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3493
3494 /* Warn if metadata_csum and gdt_csum are both set. */
3495 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3496 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3497 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3498 ext4_warning(sb, "metadata_csum and uninit_bg are "
3499 "redundant flags; please run fsck.");
3500
3501 /* Check for a known checksum algorithm */
3502 if (!ext4_verify_csum_type(sb, es)) {
3503 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3504 "unknown checksum algorithm.");
3505 silent = 1;
3506 goto cantfind_ext4;
3507 }
3508
3509 /* Load the checksum driver */
3510 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3511 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3512 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3513 if (IS_ERR(sbi->s_chksum_driver)) {
3514 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3515 ret = PTR_ERR(sbi->s_chksum_driver);
3516 sbi->s_chksum_driver = NULL;
3517 goto failed_mount;
3518 }
3519 }
3520
3521 /* Check superblock checksum */
3522 if (!ext4_superblock_csum_verify(sb, es)) {
3523 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3524 "invalid superblock checksum. Run e2fsck?");
3525 silent = 1;
3526 goto cantfind_ext4;
3527 }
3528
3529 /* Precompute checksum seed for all metadata */
3530 if (ext4_has_metadata_csum(sb))
3531 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3532 sizeof(es->s_uuid));
3533
3534 /* Set defaults before we parse the mount options */
3535 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3536 set_opt(sb, INIT_INODE_TABLE);
3537 if (def_mount_opts & EXT4_DEFM_DEBUG)
3538 set_opt(sb, DEBUG);
3539 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3540 set_opt(sb, GRPID);
3541 if (def_mount_opts & EXT4_DEFM_UID16)
3542 set_opt(sb, NO_UID32);
3543 /* xattr user namespace & acls are now defaulted on */
3544 set_opt(sb, XATTR_USER);
3545 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3546 set_opt(sb, POSIX_ACL);
3547 #endif
3548 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3549 if (ext4_has_metadata_csum(sb))
3550 set_opt(sb, JOURNAL_CHECKSUM);
3551
3552 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3553 set_opt(sb, JOURNAL_DATA);
3554 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3555 set_opt(sb, ORDERED_DATA);
3556 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3557 set_opt(sb, WRITEBACK_DATA);
3558
3559 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3560 set_opt(sb, ERRORS_PANIC);
3561 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3562 set_opt(sb, ERRORS_CONT);
3563 else
3564 set_opt(sb, ERRORS_RO);
3565 /* block_validity enabled by default; disable with noblock_validity */
3566 set_opt(sb, BLOCK_VALIDITY);
3567 if (def_mount_opts & EXT4_DEFM_DISCARD)
3568 set_opt(sb, DISCARD);
3569
3570 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3571 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3572 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3573 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3574 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3575
3576 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3577 set_opt(sb, BARRIER);
3578
3579 /*
3580 * enable delayed allocation by default
3581 * Use -o nodelalloc to turn it off
3582 */
3583 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3584 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3585 set_opt(sb, DELALLOC);
3586
3587 /*
3588 * set default s_li_wait_mult for lazyinit, for the case there is
3589 * no mount option specified.
3590 */
3591 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3592
3593 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3594 &journal_devnum, &journal_ioprio, 0)) {
3595 ext4_msg(sb, KERN_WARNING,
3596 "failed to parse options in superblock: %s",
3597 sbi->s_es->s_mount_opts);
3598 }
3599 sbi->s_def_mount_opt = sbi->s_mount_opt;
3600 if (!parse_options((char *) data, sb, &journal_devnum,
3601 &journal_ioprio, 0))
3602 goto failed_mount;
3603
3604 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3605 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3606 "with data=journal disables delayed "
3607 "allocation and O_DIRECT support!\n");
3608 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3609 ext4_msg(sb, KERN_ERR, "can't mount with "
3610 "both data=journal and delalloc");
3611 goto failed_mount;
3612 }
3613 if (test_opt(sb, DIOREAD_NOLOCK)) {
3614 ext4_msg(sb, KERN_ERR, "can't mount with "
3615 "both data=journal and dioread_nolock");
3616 goto failed_mount;
3617 }
3618 if (test_opt(sb, DAX)) {
3619 ext4_msg(sb, KERN_ERR, "can't mount with "
3620 "both data=journal and dax");
3621 goto failed_mount;
3622 }
3623 if (test_opt(sb, DELALLOC))
3624 clear_opt(sb, DELALLOC);
3625 }
3626
3627 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3628 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3629
3630 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3631 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3632 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3633 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3634 ext4_msg(sb, KERN_WARNING,
3635 "feature flags set on rev 0 fs, "
3636 "running e2fsck is recommended");
3637
3638 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3639 set_opt2(sb, HURD_COMPAT);
3640 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3641 EXT4_FEATURE_INCOMPAT_64BIT)) {
3642 ext4_msg(sb, KERN_ERR,
3643 "The Hurd can't support 64-bit file systems");
3644 goto failed_mount;
3645 }
3646 }
3647
3648 if (IS_EXT2_SB(sb)) {
3649 if (ext2_feature_set_ok(sb))
3650 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3651 "using the ext4 subsystem");
3652 else {
3653 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3654 "to feature incompatibilities");
3655 goto failed_mount;
3656 }
3657 }
3658
3659 if (IS_EXT3_SB(sb)) {
3660 if (ext3_feature_set_ok(sb))
3661 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3662 "using the ext4 subsystem");
3663 else {
3664 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3665 "to feature incompatibilities");
3666 goto failed_mount;
3667 }
3668 }
3669
3670 /*
3671 * Check feature flags regardless of the revision level, since we
3672 * previously didn't change the revision level when setting the flags,
3673 * so there is a chance incompat flags are set on a rev 0 filesystem.
3674 */
3675 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3676 goto failed_mount;
3677
3678 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3679 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3680 blocksize > EXT4_MAX_BLOCK_SIZE) {
3681 ext4_msg(sb, KERN_ERR,
3682 "Unsupported filesystem blocksize %d", blocksize);
3683 goto failed_mount;
3684 }
3685
3686 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3687 if (blocksize != PAGE_SIZE) {
3688 ext4_msg(sb, KERN_ERR,
3689 "error: unsupported blocksize for dax");
3690 goto failed_mount;
3691 }
3692 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3693 ext4_msg(sb, KERN_ERR,
3694 "error: device does not support dax");
3695 goto failed_mount;
3696 }
3697 }
3698
3699 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3700 es->s_encryption_level) {
3701 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3702 es->s_encryption_level);
3703 goto failed_mount;
3704 }
3705
3706 if (sb->s_blocksize != blocksize) {
3707 /* Validate the filesystem blocksize */
3708 if (!sb_set_blocksize(sb, blocksize)) {
3709 ext4_msg(sb, KERN_ERR, "bad block size %d",
3710 blocksize);
3711 goto failed_mount;
3712 }
3713
3714 brelse(bh);
3715 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3716 offset = do_div(logical_sb_block, blocksize);
3717 bh = sb_bread_unmovable(sb, logical_sb_block);
3718 if (!bh) {
3719 ext4_msg(sb, KERN_ERR,
3720 "Can't read superblock on 2nd try");
3721 goto failed_mount;
3722 }
3723 es = (struct ext4_super_block *)(bh->b_data + offset);
3724 sbi->s_es = es;
3725 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3726 ext4_msg(sb, KERN_ERR,
3727 "Magic mismatch, very weird!");
3728 goto failed_mount;
3729 }
3730 }
3731
3732 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3733 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3734 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3735 has_huge_files);
3736 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3737
3738 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3739 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3740 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3741 } else {
3742 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3743 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3744 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3745 (!is_power_of_2(sbi->s_inode_size)) ||
3746 (sbi->s_inode_size > blocksize)) {
3747 ext4_msg(sb, KERN_ERR,
3748 "unsupported inode size: %d",
3749 sbi->s_inode_size);
3750 goto failed_mount;
3751 }
3752 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3753 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3754 }
3755
3756 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3757 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3758 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3759 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3760 !is_power_of_2(sbi->s_desc_size)) {
3761 ext4_msg(sb, KERN_ERR,
3762 "unsupported descriptor size %lu",
3763 sbi->s_desc_size);
3764 goto failed_mount;
3765 }
3766 } else
3767 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3768
3769 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3770 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3771 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3772 goto cantfind_ext4;
3773
3774 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3775 if (sbi->s_inodes_per_block == 0)
3776 goto cantfind_ext4;
3777 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3778 sbi->s_inodes_per_block;
3779 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3780 sbi->s_sbh = bh;
3781 sbi->s_mount_state = le16_to_cpu(es->s_state);
3782 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3783 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3784
3785 for (i = 0; i < 4; i++)
3786 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3787 sbi->s_def_hash_version = es->s_def_hash_version;
3788 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3789 i = le32_to_cpu(es->s_flags);
3790 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3791 sbi->s_hash_unsigned = 3;
3792 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3793 #ifdef __CHAR_UNSIGNED__
3794 if (!(sb->s_flags & MS_RDONLY))
3795 es->s_flags |=
3796 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3797 sbi->s_hash_unsigned = 3;
3798 #else
3799 if (!(sb->s_flags & MS_RDONLY))
3800 es->s_flags |=
3801 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3802 #endif
3803 }
3804 }
3805
3806 /* Handle clustersize */
3807 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3808 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3809 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3810 if (has_bigalloc) {
3811 if (clustersize < blocksize) {
3812 ext4_msg(sb, KERN_ERR,
3813 "cluster size (%d) smaller than "
3814 "block size (%d)", clustersize, blocksize);
3815 goto failed_mount;
3816 }
3817 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3818 le32_to_cpu(es->s_log_block_size);
3819 sbi->s_clusters_per_group =
3820 le32_to_cpu(es->s_clusters_per_group);
3821 if (sbi->s_clusters_per_group > blocksize * 8) {
3822 ext4_msg(sb, KERN_ERR,
3823 "#clusters per group too big: %lu",
3824 sbi->s_clusters_per_group);
3825 goto failed_mount;
3826 }
3827 if (sbi->s_blocks_per_group !=
3828 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3829 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3830 "clusters per group (%lu) inconsistent",
3831 sbi->s_blocks_per_group,
3832 sbi->s_clusters_per_group);
3833 goto failed_mount;
3834 }
3835 } else {
3836 if (clustersize != blocksize) {
3837 ext4_warning(sb, "fragment/cluster size (%d) != "
3838 "block size (%d)", clustersize,
3839 blocksize);
3840 clustersize = blocksize;
3841 }
3842 if (sbi->s_blocks_per_group > blocksize * 8) {
3843 ext4_msg(sb, KERN_ERR,
3844 "#blocks per group too big: %lu",
3845 sbi->s_blocks_per_group);
3846 goto failed_mount;
3847 }
3848 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3849 sbi->s_cluster_bits = 0;
3850 }
3851 sbi->s_cluster_ratio = clustersize / blocksize;
3852
3853 if (sbi->s_inodes_per_group > blocksize * 8) {
3854 ext4_msg(sb, KERN_ERR,
3855 "#inodes per group too big: %lu",
3856 sbi->s_inodes_per_group);
3857 goto failed_mount;
3858 }
3859
3860 /* Do we have standard group size of clustersize * 8 blocks ? */
3861 if (sbi->s_blocks_per_group == clustersize << 3)
3862 set_opt2(sb, STD_GROUP_SIZE);
3863
3864 /*
3865 * Test whether we have more sectors than will fit in sector_t,
3866 * and whether the max offset is addressable by the page cache.
3867 */
3868 err = generic_check_addressable(sb->s_blocksize_bits,
3869 ext4_blocks_count(es));
3870 if (err) {
3871 ext4_msg(sb, KERN_ERR, "filesystem"
3872 " too large to mount safely on this system");
3873 if (sizeof(sector_t) < 8)
3874 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3875 goto failed_mount;
3876 }
3877
3878 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3879 goto cantfind_ext4;
3880
3881 /* check blocks count against device size */
3882 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3883 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3884 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3885 "exceeds size of device (%llu blocks)",
3886 ext4_blocks_count(es), blocks_count);
3887 goto failed_mount;
3888 }
3889
3890 /*
3891 * It makes no sense for the first data block to be beyond the end
3892 * of the filesystem.
3893 */
3894 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3895 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3896 "block %u is beyond end of filesystem (%llu)",
3897 le32_to_cpu(es->s_first_data_block),
3898 ext4_blocks_count(es));
3899 goto failed_mount;
3900 }
3901 blocks_count = (ext4_blocks_count(es) -
3902 le32_to_cpu(es->s_first_data_block) +
3903 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3904 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3905 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3906 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3907 "(block count %llu, first data block %u, "
3908 "blocks per group %lu)", sbi->s_groups_count,
3909 ext4_blocks_count(es),
3910 le32_to_cpu(es->s_first_data_block),
3911 EXT4_BLOCKS_PER_GROUP(sb));
3912 goto failed_mount;
3913 }
3914 sbi->s_groups_count = blocks_count;
3915 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3916 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3917 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3918 EXT4_DESC_PER_BLOCK(sb);
3919 sbi->s_group_desc = ext4_kvmalloc(db_count *
3920 sizeof(struct buffer_head *),
3921 GFP_KERNEL);
3922 if (sbi->s_group_desc == NULL) {
3923 ext4_msg(sb, KERN_ERR, "not enough memory");
3924 ret = -ENOMEM;
3925 goto failed_mount;
3926 }
3927
3928 if (ext4_proc_root)
3929 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3930
3931 if (sbi->s_proc)
3932 proc_create_data("options", S_IRUGO, sbi->s_proc,
3933 &ext4_seq_options_fops, sb);
3934
3935 bgl_lock_init(sbi->s_blockgroup_lock);
3936
3937 for (i = 0; i < db_count; i++) {
3938 block = descriptor_loc(sb, logical_sb_block, i);
3939 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3940 if (!sbi->s_group_desc[i]) {
3941 ext4_msg(sb, KERN_ERR,
3942 "can't read group descriptor %d", i);
3943 db_count = i;
3944 goto failed_mount2;
3945 }
3946 }
3947 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3948 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3949 goto failed_mount2;
3950 }
3951
3952 sbi->s_gdb_count = db_count;
3953 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3954 spin_lock_init(&sbi->s_next_gen_lock);
3955
3956 setup_timer(&sbi->s_err_report, print_daily_error_info,
3957 (unsigned long) sb);
3958
3959 /* Register extent status tree shrinker */
3960 if (ext4_es_register_shrinker(sbi))
3961 goto failed_mount3;
3962
3963 sbi->s_stripe = ext4_get_stripe_size(sbi);
3964 sbi->s_extent_max_zeroout_kb = 32;
3965
3966 /*
3967 * set up enough so that it can read an inode
3968 */
3969 sb->s_op = &ext4_sops;
3970 sb->s_export_op = &ext4_export_ops;
3971 sb->s_xattr = ext4_xattr_handlers;
3972 #ifdef CONFIG_QUOTA
3973 sb->dq_op = &ext4_quota_operations;
3974 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3975 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3976 else
3977 sb->s_qcop = &ext4_qctl_operations;
3978 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3979 #endif
3980 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3981
3982 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3983 mutex_init(&sbi->s_orphan_lock);
3984
3985 sb->s_root = NULL;
3986
3987 needs_recovery = (es->s_last_orphan != 0 ||
3988 EXT4_HAS_INCOMPAT_FEATURE(sb,
3989 EXT4_FEATURE_INCOMPAT_RECOVER));
3990
3991 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3992 !(sb->s_flags & MS_RDONLY))
3993 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3994 goto failed_mount3a;
3995
3996 /*
3997 * The first inode we look at is the journal inode. Don't try
3998 * root first: it may be modified in the journal!
3999 */
4000 if (!test_opt(sb, NOLOAD) &&
4001 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4002 if (ext4_load_journal(sb, es, journal_devnum))
4003 goto failed_mount3a;
4004 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4005 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4006 ext4_msg(sb, KERN_ERR, "required journal recovery "
4007 "suppressed and not mounted read-only");
4008 goto failed_mount_wq;
4009 } else {
4010 clear_opt(sb, DATA_FLAGS);
4011 sbi->s_journal = NULL;
4012 needs_recovery = 0;
4013 goto no_journal;
4014 }
4015
4016 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
4017 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4018 JBD2_FEATURE_INCOMPAT_64BIT)) {
4019 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4020 goto failed_mount_wq;
4021 }
4022
4023 if (!set_journal_csum_feature_set(sb)) {
4024 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4025 "feature set");
4026 goto failed_mount_wq;
4027 }
4028
4029 /* We have now updated the journal if required, so we can
4030 * validate the data journaling mode. */
4031 switch (test_opt(sb, DATA_FLAGS)) {
4032 case 0:
4033 /* No mode set, assume a default based on the journal
4034 * capabilities: ORDERED_DATA if the journal can
4035 * cope, else JOURNAL_DATA
4036 */
4037 if (jbd2_journal_check_available_features
4038 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4039 set_opt(sb, ORDERED_DATA);
4040 else
4041 set_opt(sb, JOURNAL_DATA);
4042 break;
4043
4044 case EXT4_MOUNT_ORDERED_DATA:
4045 case EXT4_MOUNT_WRITEBACK_DATA:
4046 if (!jbd2_journal_check_available_features
4047 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4048 ext4_msg(sb, KERN_ERR, "Journal does not support "
4049 "requested data journaling mode");
4050 goto failed_mount_wq;
4051 }
4052 default:
4053 break;
4054 }
4055 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4056
4057 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4058
4059 no_journal:
4060 if (ext4_mballoc_ready) {
4061 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4062 if (!sbi->s_mb_cache) {
4063 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4064 goto failed_mount_wq;
4065 }
4066 }
4067
4068 if (unlikely(sbi->s_mount_flags & EXT4_MF_TEST_DUMMY_ENCRYPTION) &&
4069 !(sb->s_flags & MS_RDONLY) &&
4070 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
4071 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
4072 ext4_commit_super(sb, 1);
4073 }
4074
4075 /*
4076 * Get the # of file system overhead blocks from the
4077 * superblock if present.
4078 */
4079 if (es->s_overhead_clusters)
4080 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4081 else {
4082 err = ext4_calculate_overhead(sb);
4083 if (err)
4084 goto failed_mount_wq;
4085 }
4086
4087 /*
4088 * The maximum number of concurrent works can be high and
4089 * concurrency isn't really necessary. Limit it to 1.
4090 */
4091 EXT4_SB(sb)->rsv_conversion_wq =
4092 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4093 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4094 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4095 ret = -ENOMEM;
4096 goto failed_mount4;
4097 }
4098
4099 /*
4100 * The jbd2_journal_load will have done any necessary log recovery,
4101 * so we can safely mount the rest of the filesystem now.
4102 */
4103
4104 root = ext4_iget(sb, EXT4_ROOT_INO);
4105 if (IS_ERR(root)) {
4106 ext4_msg(sb, KERN_ERR, "get root inode failed");
4107 ret = PTR_ERR(root);
4108 root = NULL;
4109 goto failed_mount4;
4110 }
4111 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4112 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4113 iput(root);
4114 goto failed_mount4;
4115 }
4116 sb->s_root = d_make_root(root);
4117 if (!sb->s_root) {
4118 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4119 ret = -ENOMEM;
4120 goto failed_mount4;
4121 }
4122
4123 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4124 sb->s_flags |= MS_RDONLY;
4125
4126 /* determine the minimum size of new large inodes, if present */
4127 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4128 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4129 EXT4_GOOD_OLD_INODE_SIZE;
4130 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4131 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4132 if (sbi->s_want_extra_isize <
4133 le16_to_cpu(es->s_want_extra_isize))
4134 sbi->s_want_extra_isize =
4135 le16_to_cpu(es->s_want_extra_isize);
4136 if (sbi->s_want_extra_isize <
4137 le16_to_cpu(es->s_min_extra_isize))
4138 sbi->s_want_extra_isize =
4139 le16_to_cpu(es->s_min_extra_isize);
4140 }
4141 }
4142 /* Check if enough inode space is available */
4143 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4144 sbi->s_inode_size) {
4145 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4146 EXT4_GOOD_OLD_INODE_SIZE;
4147 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4148 "available");
4149 }
4150
4151 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4152 if (err) {
4153 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4154 "reserved pool", ext4_calculate_resv_clusters(sb));
4155 goto failed_mount4a;
4156 }
4157
4158 err = ext4_setup_system_zone(sb);
4159 if (err) {
4160 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4161 "zone (%d)", err);
4162 goto failed_mount4a;
4163 }
4164
4165 ext4_ext_init(sb);
4166 err = ext4_mb_init(sb);
4167 if (err) {
4168 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4169 err);
4170 goto failed_mount5;
4171 }
4172
4173 block = ext4_count_free_clusters(sb);
4174 ext4_free_blocks_count_set(sbi->s_es,
4175 EXT4_C2B(sbi, block));
4176 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4177 GFP_KERNEL);
4178 if (!err) {
4179 unsigned long freei = ext4_count_free_inodes(sb);
4180 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4181 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4182 GFP_KERNEL);
4183 }
4184 if (!err)
4185 err = percpu_counter_init(&sbi->s_dirs_counter,
4186 ext4_count_dirs(sb), GFP_KERNEL);
4187 if (!err)
4188 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4189 GFP_KERNEL);
4190 if (err) {
4191 ext4_msg(sb, KERN_ERR, "insufficient memory");
4192 goto failed_mount6;
4193 }
4194
4195 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4196 if (!ext4_fill_flex_info(sb)) {
4197 ext4_msg(sb, KERN_ERR,
4198 "unable to initialize "
4199 "flex_bg meta info!");
4200 goto failed_mount6;
4201 }
4202
4203 err = ext4_register_li_request(sb, first_not_zeroed);
4204 if (err)
4205 goto failed_mount6;
4206
4207 sbi->s_kobj.kset = ext4_kset;
4208 init_completion(&sbi->s_kobj_unregister);
4209 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4210 "%s", sb->s_id);
4211 if (err)
4212 goto failed_mount7;
4213
4214 #ifdef CONFIG_QUOTA
4215 /* Enable quota usage during mount. */
4216 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4217 !(sb->s_flags & MS_RDONLY)) {
4218 err = ext4_enable_quotas(sb);
4219 if (err)
4220 goto failed_mount8;
4221 }
4222 #endif /* CONFIG_QUOTA */
4223
4224 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4225 ext4_orphan_cleanup(sb, es);
4226 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4227 if (needs_recovery) {
4228 ext4_msg(sb, KERN_INFO, "recovery complete");
4229 ext4_mark_recovery_complete(sb, es);
4230 }
4231 if (EXT4_SB(sb)->s_journal) {
4232 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4233 descr = " journalled data mode";
4234 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4235 descr = " ordered data mode";
4236 else
4237 descr = " writeback data mode";
4238 } else
4239 descr = "out journal";
4240
4241 if (test_opt(sb, DISCARD)) {
4242 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4243 if (!blk_queue_discard(q))
4244 ext4_msg(sb, KERN_WARNING,
4245 "mounting with \"discard\" option, but "
4246 "the device does not support discard");
4247 }
4248
4249 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4250 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4251 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4252
4253 if (es->s_error_count)
4254 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4255
4256 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4257 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4258 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4259 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4260
4261 kfree(orig_data);
4262 return 0;
4263
4264 cantfind_ext4:
4265 if (!silent)
4266 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4267 goto failed_mount;
4268
4269 #ifdef CONFIG_QUOTA
4270 failed_mount8:
4271 kobject_del(&sbi->s_kobj);
4272 #endif
4273 failed_mount7:
4274 ext4_unregister_li_request(sb);
4275 failed_mount6:
4276 ext4_mb_release(sb);
4277 if (sbi->s_flex_groups)
4278 kvfree(sbi->s_flex_groups);
4279 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4280 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4281 percpu_counter_destroy(&sbi->s_dirs_counter);
4282 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4283 failed_mount5:
4284 ext4_ext_release(sb);
4285 ext4_release_system_zone(sb);
4286 failed_mount4a:
4287 dput(sb->s_root);
4288 sb->s_root = NULL;
4289 failed_mount4:
4290 ext4_msg(sb, KERN_ERR, "mount failed");
4291 if (EXT4_SB(sb)->rsv_conversion_wq)
4292 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4293 failed_mount_wq:
4294 if (sbi->s_journal) {
4295 jbd2_journal_destroy(sbi->s_journal);
4296 sbi->s_journal = NULL;
4297 }
4298 failed_mount3a:
4299 ext4_es_unregister_shrinker(sbi);
4300 failed_mount3:
4301 del_timer_sync(&sbi->s_err_report);
4302 if (sbi->s_mmp_tsk)
4303 kthread_stop(sbi->s_mmp_tsk);
4304 failed_mount2:
4305 for (i = 0; i < db_count; i++)
4306 brelse(sbi->s_group_desc[i]);
4307 kvfree(sbi->s_group_desc);
4308 failed_mount:
4309 if (sbi->s_chksum_driver)
4310 crypto_free_shash(sbi->s_chksum_driver);
4311 if (sbi->s_proc) {
4312 remove_proc_entry("options", sbi->s_proc);
4313 remove_proc_entry(sb->s_id, ext4_proc_root);
4314 }
4315 #ifdef CONFIG_QUOTA
4316 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4317 kfree(sbi->s_qf_names[i]);
4318 #endif
4319 ext4_blkdev_remove(sbi);
4320 brelse(bh);
4321 out_fail:
4322 sb->s_fs_info = NULL;
4323 kfree(sbi->s_blockgroup_lock);
4324 kfree(sbi);
4325 out_free_orig:
4326 kfree(orig_data);
4327 return err ? err : ret;
4328 }
4329
4330 /*
4331 * Setup any per-fs journal parameters now. We'll do this both on
4332 * initial mount, once the journal has been initialised but before we've
4333 * done any recovery; and again on any subsequent remount.
4334 */
4335 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4336 {
4337 struct ext4_sb_info *sbi = EXT4_SB(sb);
4338
4339 journal->j_commit_interval = sbi->s_commit_interval;
4340 journal->j_min_batch_time = sbi->s_min_batch_time;
4341 journal->j_max_batch_time = sbi->s_max_batch_time;
4342
4343 write_lock(&journal->j_state_lock);
4344 if (test_opt(sb, BARRIER))
4345 journal->j_flags |= JBD2_BARRIER;
4346 else
4347 journal->j_flags &= ~JBD2_BARRIER;
4348 if (test_opt(sb, DATA_ERR_ABORT))
4349 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4350 else
4351 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4352 write_unlock(&journal->j_state_lock);
4353 }
4354
4355 static journal_t *ext4_get_journal(struct super_block *sb,
4356 unsigned int journal_inum)
4357 {
4358 struct inode *journal_inode;
4359 journal_t *journal;
4360
4361 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4362
4363 /* First, test for the existence of a valid inode on disk. Bad
4364 * things happen if we iget() an unused inode, as the subsequent
4365 * iput() will try to delete it. */
4366
4367 journal_inode = ext4_iget(sb, journal_inum);
4368 if (IS_ERR(journal_inode)) {
4369 ext4_msg(sb, KERN_ERR, "no journal found");
4370 return NULL;
4371 }
4372 if (!journal_inode->i_nlink) {
4373 make_bad_inode(journal_inode);
4374 iput(journal_inode);
4375 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4376 return NULL;
4377 }
4378
4379 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4380 journal_inode, journal_inode->i_size);
4381 if (!S_ISREG(journal_inode->i_mode)) {
4382 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4383 iput(journal_inode);
4384 return NULL;
4385 }
4386
4387 journal = jbd2_journal_init_inode(journal_inode);
4388 if (!journal) {
4389 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4390 iput(journal_inode);
4391 return NULL;
4392 }
4393 journal->j_private = sb;
4394 ext4_init_journal_params(sb, journal);
4395 return journal;
4396 }
4397
4398 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4399 dev_t j_dev)
4400 {
4401 struct buffer_head *bh;
4402 journal_t *journal;
4403 ext4_fsblk_t start;
4404 ext4_fsblk_t len;
4405 int hblock, blocksize;
4406 ext4_fsblk_t sb_block;
4407 unsigned long offset;
4408 struct ext4_super_block *es;
4409 struct block_device *bdev;
4410
4411 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4412
4413 bdev = ext4_blkdev_get(j_dev, sb);
4414 if (bdev == NULL)
4415 return NULL;
4416
4417 blocksize = sb->s_blocksize;
4418 hblock = bdev_logical_block_size(bdev);
4419 if (blocksize < hblock) {
4420 ext4_msg(sb, KERN_ERR,
4421 "blocksize too small for journal device");
4422 goto out_bdev;
4423 }
4424
4425 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4426 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4427 set_blocksize(bdev, blocksize);
4428 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4429 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4430 "external journal");
4431 goto out_bdev;
4432 }
4433
4434 es = (struct ext4_super_block *) (bh->b_data + offset);
4435 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4436 !(le32_to_cpu(es->s_feature_incompat) &
4437 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4438 ext4_msg(sb, KERN_ERR, "external journal has "
4439 "bad superblock");
4440 brelse(bh);
4441 goto out_bdev;
4442 }
4443
4444 if ((le32_to_cpu(es->s_feature_ro_compat) &
4445 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4446 es->s_checksum != ext4_superblock_csum(sb, es)) {
4447 ext4_msg(sb, KERN_ERR, "external journal has "
4448 "corrupt superblock");
4449 brelse(bh);
4450 goto out_bdev;
4451 }
4452
4453 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4454 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4455 brelse(bh);
4456 goto out_bdev;
4457 }
4458
4459 len = ext4_blocks_count(es);
4460 start = sb_block + 1;
4461 brelse(bh); /* we're done with the superblock */
4462
4463 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4464 start, len, blocksize);
4465 if (!journal) {
4466 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4467 goto out_bdev;
4468 }
4469 journal->j_private = sb;
4470 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4471 wait_on_buffer(journal->j_sb_buffer);
4472 if (!buffer_uptodate(journal->j_sb_buffer)) {
4473 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4474 goto out_journal;
4475 }
4476 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4477 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4478 "user (unsupported) - %d",
4479 be32_to_cpu(journal->j_superblock->s_nr_users));
4480 goto out_journal;
4481 }
4482 EXT4_SB(sb)->journal_bdev = bdev;
4483 ext4_init_journal_params(sb, journal);
4484 return journal;
4485
4486 out_journal:
4487 jbd2_journal_destroy(journal);
4488 out_bdev:
4489 ext4_blkdev_put(bdev);
4490 return NULL;
4491 }
4492
4493 static int ext4_load_journal(struct super_block *sb,
4494 struct ext4_super_block *es,
4495 unsigned long journal_devnum)
4496 {
4497 journal_t *journal;
4498 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4499 dev_t journal_dev;
4500 int err = 0;
4501 int really_read_only;
4502
4503 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4504
4505 if (journal_devnum &&
4506 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4507 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4508 "numbers have changed");
4509 journal_dev = new_decode_dev(journal_devnum);
4510 } else
4511 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4512
4513 really_read_only = bdev_read_only(sb->s_bdev);
4514
4515 /*
4516 * Are we loading a blank journal or performing recovery after a
4517 * crash? For recovery, we need to check in advance whether we
4518 * can get read-write access to the device.
4519 */
4520 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4521 if (sb->s_flags & MS_RDONLY) {
4522 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4523 "required on readonly filesystem");
4524 if (really_read_only) {
4525 ext4_msg(sb, KERN_ERR, "write access "
4526 "unavailable, cannot proceed");
4527 return -EROFS;
4528 }
4529 ext4_msg(sb, KERN_INFO, "write access will "
4530 "be enabled during recovery");
4531 }
4532 }
4533
4534 if (journal_inum && journal_dev) {
4535 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4536 "and inode journals!");
4537 return -EINVAL;
4538 }
4539
4540 if (journal_inum) {
4541 if (!(journal = ext4_get_journal(sb, journal_inum)))
4542 return -EINVAL;
4543 } else {
4544 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4545 return -EINVAL;
4546 }
4547
4548 if (!(journal->j_flags & JBD2_BARRIER))
4549 ext4_msg(sb, KERN_INFO, "barriers disabled");
4550
4551 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4552 err = jbd2_journal_wipe(journal, !really_read_only);
4553 if (!err) {
4554 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4555 if (save)
4556 memcpy(save, ((char *) es) +
4557 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4558 err = jbd2_journal_load(journal);
4559 if (save)
4560 memcpy(((char *) es) + EXT4_S_ERR_START,
4561 save, EXT4_S_ERR_LEN);
4562 kfree(save);
4563 }
4564
4565 if (err) {
4566 ext4_msg(sb, KERN_ERR, "error loading journal");
4567 jbd2_journal_destroy(journal);
4568 return err;
4569 }
4570
4571 EXT4_SB(sb)->s_journal = journal;
4572 ext4_clear_journal_err(sb, es);
4573
4574 if (!really_read_only && journal_devnum &&
4575 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4576 es->s_journal_dev = cpu_to_le32(journal_devnum);
4577
4578 /* Make sure we flush the recovery flag to disk. */
4579 ext4_commit_super(sb, 1);
4580 }
4581
4582 return 0;
4583 }
4584
4585 static int ext4_commit_super(struct super_block *sb, int sync)
4586 {
4587 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4588 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4589 int error = 0;
4590
4591 if (!sbh)
4592 return error;
4593 if (buffer_write_io_error(sbh)) {
4594 /*
4595 * Oh, dear. A previous attempt to write the
4596 * superblock failed. This could happen because the
4597 * USB device was yanked out. Or it could happen to
4598 * be a transient write error and maybe the block will
4599 * be remapped. Nothing we can do but to retry the
4600 * write and hope for the best.
4601 */
4602 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4603 "superblock detected");
4604 clear_buffer_write_io_error(sbh);
4605 set_buffer_uptodate(sbh);
4606 }
4607 /*
4608 * If the file system is mounted read-only, don't update the
4609 * superblock write time. This avoids updating the superblock
4610 * write time when we are mounting the root file system
4611 * read/only but we need to replay the journal; at that point,
4612 * for people who are east of GMT and who make their clock
4613 * tick in localtime for Windows bug-for-bug compatibility,
4614 * the clock is set in the future, and this will cause e2fsck
4615 * to complain and force a full file system check.
4616 */
4617 if (!(sb->s_flags & MS_RDONLY))
4618 es->s_wtime = cpu_to_le32(get_seconds());
4619 if (sb->s_bdev->bd_part)
4620 es->s_kbytes_written =
4621 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4622 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4623 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4624 else
4625 es->s_kbytes_written =
4626 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4627 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4628 ext4_free_blocks_count_set(es,
4629 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4630 &EXT4_SB(sb)->s_freeclusters_counter)));
4631 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4632 es->s_free_inodes_count =
4633 cpu_to_le32(percpu_counter_sum_positive(
4634 &EXT4_SB(sb)->s_freeinodes_counter));
4635 BUFFER_TRACE(sbh, "marking dirty");
4636 ext4_superblock_csum_set(sb);
4637 mark_buffer_dirty(sbh);
4638 if (sync) {
4639 error = sync_dirty_buffer(sbh);
4640 if (error)
4641 return error;
4642
4643 error = buffer_write_io_error(sbh);
4644 if (error) {
4645 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4646 "superblock");
4647 clear_buffer_write_io_error(sbh);
4648 set_buffer_uptodate(sbh);
4649 }
4650 }
4651 return error;
4652 }
4653
4654 /*
4655 * Have we just finished recovery? If so, and if we are mounting (or
4656 * remounting) the filesystem readonly, then we will end up with a
4657 * consistent fs on disk. Record that fact.
4658 */
4659 static void ext4_mark_recovery_complete(struct super_block *sb,
4660 struct ext4_super_block *es)
4661 {
4662 journal_t *journal = EXT4_SB(sb)->s_journal;
4663
4664 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4665 BUG_ON(journal != NULL);
4666 return;
4667 }
4668 jbd2_journal_lock_updates(journal);
4669 if (jbd2_journal_flush(journal) < 0)
4670 goto out;
4671
4672 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4673 sb->s_flags & MS_RDONLY) {
4674 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4675 ext4_commit_super(sb, 1);
4676 }
4677
4678 out:
4679 jbd2_journal_unlock_updates(journal);
4680 }
4681
4682 /*
4683 * If we are mounting (or read-write remounting) a filesystem whose journal
4684 * has recorded an error from a previous lifetime, move that error to the
4685 * main filesystem now.
4686 */
4687 static void ext4_clear_journal_err(struct super_block *sb,
4688 struct ext4_super_block *es)
4689 {
4690 journal_t *journal;
4691 int j_errno;
4692 const char *errstr;
4693
4694 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4695
4696 journal = EXT4_SB(sb)->s_journal;
4697
4698 /*
4699 * Now check for any error status which may have been recorded in the
4700 * journal by a prior ext4_error() or ext4_abort()
4701 */
4702
4703 j_errno = jbd2_journal_errno(journal);
4704 if (j_errno) {
4705 char nbuf[16];
4706
4707 errstr = ext4_decode_error(sb, j_errno, nbuf);
4708 ext4_warning(sb, "Filesystem error recorded "
4709 "from previous mount: %s", errstr);
4710 ext4_warning(sb, "Marking fs in need of filesystem check.");
4711
4712 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4713 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4714 ext4_commit_super(sb, 1);
4715
4716 jbd2_journal_clear_err(journal);
4717 jbd2_journal_update_sb_errno(journal);
4718 }
4719 }
4720
4721 /*
4722 * Force the running and committing transactions to commit,
4723 * and wait on the commit.
4724 */
4725 int ext4_force_commit(struct super_block *sb)
4726 {
4727 journal_t *journal;
4728
4729 if (sb->s_flags & MS_RDONLY)
4730 return 0;
4731
4732 journal = EXT4_SB(sb)->s_journal;
4733 return ext4_journal_force_commit(journal);
4734 }
4735
4736 static int ext4_sync_fs(struct super_block *sb, int wait)
4737 {
4738 int ret = 0;
4739 tid_t target;
4740 bool needs_barrier = false;
4741 struct ext4_sb_info *sbi = EXT4_SB(sb);
4742
4743 trace_ext4_sync_fs(sb, wait);
4744 flush_workqueue(sbi->rsv_conversion_wq);
4745 /*
4746 * Writeback quota in non-journalled quota case - journalled quota has
4747 * no dirty dquots
4748 */
4749 dquot_writeback_dquots(sb, -1);
4750 /*
4751 * Data writeback is possible w/o journal transaction, so barrier must
4752 * being sent at the end of the function. But we can skip it if
4753 * transaction_commit will do it for us.
4754 */
4755 if (sbi->s_journal) {
4756 target = jbd2_get_latest_transaction(sbi->s_journal);
4757 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4758 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4759 needs_barrier = true;
4760
4761 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4762 if (wait)
4763 ret = jbd2_log_wait_commit(sbi->s_journal,
4764 target);
4765 }
4766 } else if (wait && test_opt(sb, BARRIER))
4767 needs_barrier = true;
4768 if (needs_barrier) {
4769 int err;
4770 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4771 if (!ret)
4772 ret = err;
4773 }
4774
4775 return ret;
4776 }
4777
4778 /*
4779 * LVM calls this function before a (read-only) snapshot is created. This
4780 * gives us a chance to flush the journal completely and mark the fs clean.
4781 *
4782 * Note that only this function cannot bring a filesystem to be in a clean
4783 * state independently. It relies on upper layer to stop all data & metadata
4784 * modifications.
4785 */
4786 static int ext4_freeze(struct super_block *sb)
4787 {
4788 int error = 0;
4789 journal_t *journal;
4790
4791 if (sb->s_flags & MS_RDONLY)
4792 return 0;
4793
4794 journal = EXT4_SB(sb)->s_journal;
4795
4796 if (journal) {
4797 /* Now we set up the journal barrier. */
4798 jbd2_journal_lock_updates(journal);
4799
4800 /*
4801 * Don't clear the needs_recovery flag if we failed to
4802 * flush the journal.
4803 */
4804 error = jbd2_journal_flush(journal);
4805 if (error < 0)
4806 goto out;
4807 }
4808
4809 /* Journal blocked and flushed, clear needs_recovery flag. */
4810 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4811 error = ext4_commit_super(sb, 1);
4812 out:
4813 if (journal)
4814 /* we rely on upper layer to stop further updates */
4815 jbd2_journal_unlock_updates(journal);
4816 return error;
4817 }
4818
4819 /*
4820 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4821 * flag here, even though the filesystem is not technically dirty yet.
4822 */
4823 static int ext4_unfreeze(struct super_block *sb)
4824 {
4825 if (sb->s_flags & MS_RDONLY)
4826 return 0;
4827
4828 /* Reset the needs_recovery flag before the fs is unlocked. */
4829 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4830 ext4_commit_super(sb, 1);
4831 return 0;
4832 }
4833
4834 /*
4835 * Structure to save mount options for ext4_remount's benefit
4836 */
4837 struct ext4_mount_options {
4838 unsigned long s_mount_opt;
4839 unsigned long s_mount_opt2;
4840 kuid_t s_resuid;
4841 kgid_t s_resgid;
4842 unsigned long s_commit_interval;
4843 u32 s_min_batch_time, s_max_batch_time;
4844 #ifdef CONFIG_QUOTA
4845 int s_jquota_fmt;
4846 char *s_qf_names[EXT4_MAXQUOTAS];
4847 #endif
4848 };
4849
4850 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4851 {
4852 struct ext4_super_block *es;
4853 struct ext4_sb_info *sbi = EXT4_SB(sb);
4854 unsigned long old_sb_flags;
4855 struct ext4_mount_options old_opts;
4856 int enable_quota = 0;
4857 ext4_group_t g;
4858 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4859 int err = 0;
4860 #ifdef CONFIG_QUOTA
4861 int i, j;
4862 #endif
4863 char *orig_data = kstrdup(data, GFP_KERNEL);
4864
4865 /* Store the original options */
4866 old_sb_flags = sb->s_flags;
4867 old_opts.s_mount_opt = sbi->s_mount_opt;
4868 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4869 old_opts.s_resuid = sbi->s_resuid;
4870 old_opts.s_resgid = sbi->s_resgid;
4871 old_opts.s_commit_interval = sbi->s_commit_interval;
4872 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4873 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4874 #ifdef CONFIG_QUOTA
4875 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4876 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4877 if (sbi->s_qf_names[i]) {
4878 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4879 GFP_KERNEL);
4880 if (!old_opts.s_qf_names[i]) {
4881 for (j = 0; j < i; j++)
4882 kfree(old_opts.s_qf_names[j]);
4883 kfree(orig_data);
4884 return -ENOMEM;
4885 }
4886 } else
4887 old_opts.s_qf_names[i] = NULL;
4888 #endif
4889 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4890 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4891
4892 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4893 err = -EINVAL;
4894 goto restore_opts;
4895 }
4896
4897 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4898 test_opt(sb, JOURNAL_CHECKSUM)) {
4899 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4900 "during remount not supported; ignoring");
4901 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4902 }
4903
4904 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4905 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4906 ext4_msg(sb, KERN_ERR, "can't mount with "
4907 "both data=journal and delalloc");
4908 err = -EINVAL;
4909 goto restore_opts;
4910 }
4911 if (test_opt(sb, DIOREAD_NOLOCK)) {
4912 ext4_msg(sb, KERN_ERR, "can't mount with "
4913 "both data=journal and dioread_nolock");
4914 err = -EINVAL;
4915 goto restore_opts;
4916 }
4917 if (test_opt(sb, DAX)) {
4918 ext4_msg(sb, KERN_ERR, "can't mount with "
4919 "both data=journal and dax");
4920 err = -EINVAL;
4921 goto restore_opts;
4922 }
4923 }
4924
4925 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4926 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4927 "dax flag with busy inodes while remounting");
4928 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4929 }
4930
4931 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4932 ext4_abort(sb, "Abort forced by user");
4933
4934 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4935 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4936
4937 es = sbi->s_es;
4938
4939 if (sbi->s_journal) {
4940 ext4_init_journal_params(sb, sbi->s_journal);
4941 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4942 }
4943
4944 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4945 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4946 err = -EROFS;
4947 goto restore_opts;
4948 }
4949
4950 if (*flags & MS_RDONLY) {
4951 err = sync_filesystem(sb);
4952 if (err < 0)
4953 goto restore_opts;
4954 err = dquot_suspend(sb, -1);
4955 if (err < 0)
4956 goto restore_opts;
4957
4958 /*
4959 * First of all, the unconditional stuff we have to do
4960 * to disable replay of the journal when we next remount
4961 */
4962 sb->s_flags |= MS_RDONLY;
4963
4964 /*
4965 * OK, test if we are remounting a valid rw partition
4966 * readonly, and if so set the rdonly flag and then
4967 * mark the partition as valid again.
4968 */
4969 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4970 (sbi->s_mount_state & EXT4_VALID_FS))
4971 es->s_state = cpu_to_le16(sbi->s_mount_state);
4972
4973 if (sbi->s_journal)
4974 ext4_mark_recovery_complete(sb, es);
4975 } else {
4976 /* Make sure we can mount this feature set readwrite */
4977 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4978 EXT4_FEATURE_RO_COMPAT_READONLY) ||
4979 !ext4_feature_set_ok(sb, 0)) {
4980 err = -EROFS;
4981 goto restore_opts;
4982 }
4983 /*
4984 * Make sure the group descriptor checksums
4985 * are sane. If they aren't, refuse to remount r/w.
4986 */
4987 for (g = 0; g < sbi->s_groups_count; g++) {
4988 struct ext4_group_desc *gdp =
4989 ext4_get_group_desc(sb, g, NULL);
4990
4991 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4992 ext4_msg(sb, KERN_ERR,
4993 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4994 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4995 le16_to_cpu(gdp->bg_checksum));
4996 err = -EINVAL;
4997 goto restore_opts;
4998 }
4999 }
5000
5001 /*
5002 * If we have an unprocessed orphan list hanging
5003 * around from a previously readonly bdev mount,
5004 * require a full umount/remount for now.
5005 */
5006 if (es->s_last_orphan) {
5007 ext4_msg(sb, KERN_WARNING, "Couldn't "
5008 "remount RDWR because of unprocessed "
5009 "orphan inode list. Please "
5010 "umount/remount instead");
5011 err = -EINVAL;
5012 goto restore_opts;
5013 }
5014
5015 /*
5016 * Mounting a RDONLY partition read-write, so reread
5017 * and store the current valid flag. (It may have
5018 * been changed by e2fsck since we originally mounted
5019 * the partition.)
5020 */
5021 if (sbi->s_journal)
5022 ext4_clear_journal_err(sb, es);
5023 sbi->s_mount_state = le16_to_cpu(es->s_state);
5024 if (!ext4_setup_super(sb, es, 0))
5025 sb->s_flags &= ~MS_RDONLY;
5026 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5027 EXT4_FEATURE_INCOMPAT_MMP))
5028 if (ext4_multi_mount_protect(sb,
5029 le64_to_cpu(es->s_mmp_block))) {
5030 err = -EROFS;
5031 goto restore_opts;
5032 }
5033 enable_quota = 1;
5034 }
5035 }
5036
5037 /*
5038 * Reinitialize lazy itable initialization thread based on
5039 * current settings
5040 */
5041 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5042 ext4_unregister_li_request(sb);
5043 else {
5044 ext4_group_t first_not_zeroed;
5045 first_not_zeroed = ext4_has_uninit_itable(sb);
5046 ext4_register_li_request(sb, first_not_zeroed);
5047 }
5048
5049 ext4_setup_system_zone(sb);
5050 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5051 ext4_commit_super(sb, 1);
5052
5053 #ifdef CONFIG_QUOTA
5054 /* Release old quota file names */
5055 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5056 kfree(old_opts.s_qf_names[i]);
5057 if (enable_quota) {
5058 if (sb_any_quota_suspended(sb))
5059 dquot_resume(sb, -1);
5060 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5061 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5062 err = ext4_enable_quotas(sb);
5063 if (err)
5064 goto restore_opts;
5065 }
5066 }
5067 #endif
5068
5069 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5070 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5071 kfree(orig_data);
5072 return 0;
5073
5074 restore_opts:
5075 sb->s_flags = old_sb_flags;
5076 sbi->s_mount_opt = old_opts.s_mount_opt;
5077 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5078 sbi->s_resuid = old_opts.s_resuid;
5079 sbi->s_resgid = old_opts.s_resgid;
5080 sbi->s_commit_interval = old_opts.s_commit_interval;
5081 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5082 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5083 #ifdef CONFIG_QUOTA
5084 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5085 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5086 kfree(sbi->s_qf_names[i]);
5087 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5088 }
5089 #endif
5090 kfree(orig_data);
5091 return err;
5092 }
5093
5094 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5095 {
5096 struct super_block *sb = dentry->d_sb;
5097 struct ext4_sb_info *sbi = EXT4_SB(sb);
5098 struct ext4_super_block *es = sbi->s_es;
5099 ext4_fsblk_t overhead = 0, resv_blocks;
5100 u64 fsid;
5101 s64 bfree;
5102 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5103
5104 if (!test_opt(sb, MINIX_DF))
5105 overhead = sbi->s_overhead;
5106
5107 buf->f_type = EXT4_SUPER_MAGIC;
5108 buf->f_bsize = sb->s_blocksize;
5109 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5110 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5111 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5112 /* prevent underflow in case that few free space is available */
5113 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5114 buf->f_bavail = buf->f_bfree -
5115 (ext4_r_blocks_count(es) + resv_blocks);
5116 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5117 buf->f_bavail = 0;
5118 buf->f_files = le32_to_cpu(es->s_inodes_count);
5119 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5120 buf->f_namelen = EXT4_NAME_LEN;
5121 fsid = le64_to_cpup((void *)es->s_uuid) ^
5122 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5123 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5124 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5125
5126 return 0;
5127 }
5128
5129 /* Helper function for writing quotas on sync - we need to start transaction
5130 * before quota file is locked for write. Otherwise the are possible deadlocks:
5131 * Process 1 Process 2
5132 * ext4_create() quota_sync()
5133 * jbd2_journal_start() write_dquot()
5134 * dquot_initialize() down(dqio_mutex)
5135 * down(dqio_mutex) jbd2_journal_start()
5136 *
5137 */
5138
5139 #ifdef CONFIG_QUOTA
5140
5141 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5142 {
5143 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5144 }
5145
5146 static int ext4_write_dquot(struct dquot *dquot)
5147 {
5148 int ret, err;
5149 handle_t *handle;
5150 struct inode *inode;
5151
5152 inode = dquot_to_inode(dquot);
5153 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5154 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5155 if (IS_ERR(handle))
5156 return PTR_ERR(handle);
5157 ret = dquot_commit(dquot);
5158 err = ext4_journal_stop(handle);
5159 if (!ret)
5160 ret = err;
5161 return ret;
5162 }
5163
5164 static int ext4_acquire_dquot(struct dquot *dquot)
5165 {
5166 int ret, err;
5167 handle_t *handle;
5168
5169 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5170 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5171 if (IS_ERR(handle))
5172 return PTR_ERR(handle);
5173 ret = dquot_acquire(dquot);
5174 err = ext4_journal_stop(handle);
5175 if (!ret)
5176 ret = err;
5177 return ret;
5178 }
5179
5180 static int ext4_release_dquot(struct dquot *dquot)
5181 {
5182 int ret, err;
5183 handle_t *handle;
5184
5185 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5186 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5187 if (IS_ERR(handle)) {
5188 /* Release dquot anyway to avoid endless cycle in dqput() */
5189 dquot_release(dquot);
5190 return PTR_ERR(handle);
5191 }
5192 ret = dquot_release(dquot);
5193 err = ext4_journal_stop(handle);
5194 if (!ret)
5195 ret = err;
5196 return ret;
5197 }
5198
5199 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5200 {
5201 struct super_block *sb = dquot->dq_sb;
5202 struct ext4_sb_info *sbi = EXT4_SB(sb);
5203
5204 /* Are we journaling quotas? */
5205 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5206 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5207 dquot_mark_dquot_dirty(dquot);
5208 return ext4_write_dquot(dquot);
5209 } else {
5210 return dquot_mark_dquot_dirty(dquot);
5211 }
5212 }
5213
5214 static int ext4_write_info(struct super_block *sb, int type)
5215 {
5216 int ret, err;
5217 handle_t *handle;
5218
5219 /* Data block + inode block */
5220 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5221 if (IS_ERR(handle))
5222 return PTR_ERR(handle);
5223 ret = dquot_commit_info(sb, type);
5224 err = ext4_journal_stop(handle);
5225 if (!ret)
5226 ret = err;
5227 return ret;
5228 }
5229
5230 /*
5231 * Turn on quotas during mount time - we need to find
5232 * the quota file and such...
5233 */
5234 static int ext4_quota_on_mount(struct super_block *sb, int type)
5235 {
5236 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5237 EXT4_SB(sb)->s_jquota_fmt, type);
5238 }
5239
5240 /*
5241 * Standard function to be called on quota_on
5242 */
5243 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5244 struct path *path)
5245 {
5246 int err;
5247
5248 if (!test_opt(sb, QUOTA))
5249 return -EINVAL;
5250
5251 /* Quotafile not on the same filesystem? */
5252 if (path->dentry->d_sb != sb)
5253 return -EXDEV;
5254 /* Journaling quota? */
5255 if (EXT4_SB(sb)->s_qf_names[type]) {
5256 /* Quotafile not in fs root? */
5257 if (path->dentry->d_parent != sb->s_root)
5258 ext4_msg(sb, KERN_WARNING,
5259 "Quota file not on filesystem root. "
5260 "Journaled quota will not work");
5261 }
5262
5263 /*
5264 * When we journal data on quota file, we have to flush journal to see
5265 * all updates to the file when we bypass pagecache...
5266 */
5267 if (EXT4_SB(sb)->s_journal &&
5268 ext4_should_journal_data(d_inode(path->dentry))) {
5269 /*
5270 * We don't need to lock updates but journal_flush() could
5271 * otherwise be livelocked...
5272 */
5273 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5274 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5275 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5276 if (err)
5277 return err;
5278 }
5279
5280 return dquot_quota_on(sb, type, format_id, path);
5281 }
5282
5283 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5284 unsigned int flags)
5285 {
5286 int err;
5287 struct inode *qf_inode;
5288 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5289 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5290 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5291 };
5292
5293 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5294
5295 if (!qf_inums[type])
5296 return -EPERM;
5297
5298 qf_inode = ext4_iget(sb, qf_inums[type]);
5299 if (IS_ERR(qf_inode)) {
5300 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5301 return PTR_ERR(qf_inode);
5302 }
5303
5304 /* Don't account quota for quota files to avoid recursion */
5305 qf_inode->i_flags |= S_NOQUOTA;
5306 err = dquot_enable(qf_inode, type, format_id, flags);
5307 iput(qf_inode);
5308
5309 return err;
5310 }
5311
5312 /* Enable usage tracking for all quota types. */
5313 static int ext4_enable_quotas(struct super_block *sb)
5314 {
5315 int type, err = 0;
5316 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5317 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5318 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5319 };
5320
5321 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5322 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5323 if (qf_inums[type]) {
5324 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5325 DQUOT_USAGE_ENABLED);
5326 if (err) {
5327 ext4_warning(sb,
5328 "Failed to enable quota tracking "
5329 "(type=%d, err=%d). Please run "
5330 "e2fsck to fix.", type, err);
5331 return err;
5332 }
5333 }
5334 }
5335 return 0;
5336 }
5337
5338 static int ext4_quota_off(struct super_block *sb, int type)
5339 {
5340 struct inode *inode = sb_dqopt(sb)->files[type];
5341 handle_t *handle;
5342
5343 /* Force all delayed allocation blocks to be allocated.
5344 * Caller already holds s_umount sem */
5345 if (test_opt(sb, DELALLOC))
5346 sync_filesystem(sb);
5347
5348 if (!inode)
5349 goto out;
5350
5351 /* Update modification times of quota files when userspace can
5352 * start looking at them */
5353 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5354 if (IS_ERR(handle))
5355 goto out;
5356 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5357 ext4_mark_inode_dirty(handle, inode);
5358 ext4_journal_stop(handle);
5359
5360 out:
5361 return dquot_quota_off(sb, type);
5362 }
5363
5364 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5365 * acquiring the locks... As quota files are never truncated and quota code
5366 * itself serializes the operations (and no one else should touch the files)
5367 * we don't have to be afraid of races */
5368 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5369 size_t len, loff_t off)
5370 {
5371 struct inode *inode = sb_dqopt(sb)->files[type];
5372 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5373 int offset = off & (sb->s_blocksize - 1);
5374 int tocopy;
5375 size_t toread;
5376 struct buffer_head *bh;
5377 loff_t i_size = i_size_read(inode);
5378
5379 if (off > i_size)
5380 return 0;
5381 if (off+len > i_size)
5382 len = i_size-off;
5383 toread = len;
5384 while (toread > 0) {
5385 tocopy = sb->s_blocksize - offset < toread ?
5386 sb->s_blocksize - offset : toread;
5387 bh = ext4_bread(NULL, inode, blk, 0);
5388 if (IS_ERR(bh))
5389 return PTR_ERR(bh);
5390 if (!bh) /* A hole? */
5391 memset(data, 0, tocopy);
5392 else
5393 memcpy(data, bh->b_data+offset, tocopy);
5394 brelse(bh);
5395 offset = 0;
5396 toread -= tocopy;
5397 data += tocopy;
5398 blk++;
5399 }
5400 return len;
5401 }
5402
5403 /* Write to quotafile (we know the transaction is already started and has
5404 * enough credits) */
5405 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5406 const char *data, size_t len, loff_t off)
5407 {
5408 struct inode *inode = sb_dqopt(sb)->files[type];
5409 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5410 int err, offset = off & (sb->s_blocksize - 1);
5411 struct buffer_head *bh;
5412 handle_t *handle = journal_current_handle();
5413
5414 if (EXT4_SB(sb)->s_journal && !handle) {
5415 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5416 " cancelled because transaction is not started",
5417 (unsigned long long)off, (unsigned long long)len);
5418 return -EIO;
5419 }
5420 /*
5421 * Since we account only one data block in transaction credits,
5422 * then it is impossible to cross a block boundary.
5423 */
5424 if (sb->s_blocksize - offset < len) {
5425 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5426 " cancelled because not block aligned",
5427 (unsigned long long)off, (unsigned long long)len);
5428 return -EIO;
5429 }
5430
5431 bh = ext4_bread(handle, inode, blk, 1);
5432 if (IS_ERR(bh))
5433 return PTR_ERR(bh);
5434 if (!bh)
5435 goto out;
5436 BUFFER_TRACE(bh, "get write access");
5437 err = ext4_journal_get_write_access(handle, bh);
5438 if (err) {
5439 brelse(bh);
5440 return err;
5441 }
5442 lock_buffer(bh);
5443 memcpy(bh->b_data+offset, data, len);
5444 flush_dcache_page(bh->b_page);
5445 unlock_buffer(bh);
5446 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5447 brelse(bh);
5448 out:
5449 if (inode->i_size < off + len) {
5450 i_size_write(inode, off + len);
5451 EXT4_I(inode)->i_disksize = inode->i_size;
5452 ext4_mark_inode_dirty(handle, inode);
5453 }
5454 return len;
5455 }
5456
5457 #endif
5458
5459 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5460 const char *dev_name, void *data)
5461 {
5462 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5463 }
5464
5465 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5466 static inline void register_as_ext2(void)
5467 {
5468 int err = register_filesystem(&ext2_fs_type);
5469 if (err)
5470 printk(KERN_WARNING
5471 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5472 }
5473
5474 static inline void unregister_as_ext2(void)
5475 {
5476 unregister_filesystem(&ext2_fs_type);
5477 }
5478
5479 static inline int ext2_feature_set_ok(struct super_block *sb)
5480 {
5481 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5482 return 0;
5483 if (sb->s_flags & MS_RDONLY)
5484 return 1;
5485 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5486 return 0;
5487 return 1;
5488 }
5489 #else
5490 static inline void register_as_ext2(void) { }
5491 static inline void unregister_as_ext2(void) { }
5492 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5493 #endif
5494
5495 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5496 static inline void register_as_ext3(void)
5497 {
5498 int err = register_filesystem(&ext3_fs_type);
5499 if (err)
5500 printk(KERN_WARNING
5501 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5502 }
5503
5504 static inline void unregister_as_ext3(void)
5505 {
5506 unregister_filesystem(&ext3_fs_type);
5507 }
5508
5509 static inline int ext3_feature_set_ok(struct super_block *sb)
5510 {
5511 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5512 return 0;
5513 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5514 return 0;
5515 if (sb->s_flags & MS_RDONLY)
5516 return 1;
5517 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5518 return 0;
5519 return 1;
5520 }
5521 #else
5522 static inline void register_as_ext3(void) { }
5523 static inline void unregister_as_ext3(void) { }
5524 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5525 #endif
5526
5527 static struct file_system_type ext4_fs_type = {
5528 .owner = THIS_MODULE,
5529 .name = "ext4",
5530 .mount = ext4_mount,
5531 .kill_sb = kill_block_super,
5532 .fs_flags = FS_REQUIRES_DEV,
5533 };
5534 MODULE_ALIAS_FS("ext4");
5535
5536 static int __init ext4_init_feat_adverts(void)
5537 {
5538 struct ext4_features *ef;
5539 int ret = -ENOMEM;
5540
5541 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5542 if (!ef)
5543 goto out;
5544
5545 ef->f_kobj.kset = ext4_kset;
5546 init_completion(&ef->f_kobj_unregister);
5547 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5548 "features");
5549 if (ret) {
5550 kfree(ef);
5551 goto out;
5552 }
5553
5554 ext4_feat = ef;
5555 ret = 0;
5556 out:
5557 return ret;
5558 }
5559
5560 static void ext4_exit_feat_adverts(void)
5561 {
5562 kobject_put(&ext4_feat->f_kobj);
5563 wait_for_completion(&ext4_feat->f_kobj_unregister);
5564 kfree(ext4_feat);
5565 }
5566
5567 /* Shared across all ext4 file systems */
5568 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5569 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5570
5571 static int __init ext4_init_fs(void)
5572 {
5573 int i, err;
5574
5575 ext4_li_info = NULL;
5576 mutex_init(&ext4_li_mtx);
5577
5578 /* Build-time check for flags consistency */
5579 ext4_check_flag_values();
5580
5581 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5582 mutex_init(&ext4__aio_mutex[i]);
5583 init_waitqueue_head(&ext4__ioend_wq[i]);
5584 }
5585
5586 err = ext4_init_es();
5587 if (err)
5588 return err;
5589
5590 err = ext4_init_pageio();
5591 if (err)
5592 goto out7;
5593
5594 err = ext4_init_system_zone();
5595 if (err)
5596 goto out6;
5597 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5598 if (!ext4_kset) {
5599 err = -ENOMEM;
5600 goto out5;
5601 }
5602 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5603
5604 err = ext4_init_feat_adverts();
5605 if (err)
5606 goto out4;
5607
5608 err = ext4_init_mballoc();
5609 if (err)
5610 goto out2;
5611 else
5612 ext4_mballoc_ready = 1;
5613 err = init_inodecache();
5614 if (err)
5615 goto out1;
5616 register_as_ext3();
5617 register_as_ext2();
5618 err = register_filesystem(&ext4_fs_type);
5619 if (err)
5620 goto out;
5621
5622 return 0;
5623 out:
5624 unregister_as_ext2();
5625 unregister_as_ext3();
5626 destroy_inodecache();
5627 out1:
5628 ext4_mballoc_ready = 0;
5629 ext4_exit_mballoc();
5630 out2:
5631 ext4_exit_feat_adverts();
5632 out4:
5633 if (ext4_proc_root)
5634 remove_proc_entry("fs/ext4", NULL);
5635 kset_unregister(ext4_kset);
5636 out5:
5637 ext4_exit_system_zone();
5638 out6:
5639 ext4_exit_pageio();
5640 out7:
5641 ext4_exit_es();
5642
5643 return err;
5644 }
5645
5646 static void __exit ext4_exit_fs(void)
5647 {
5648 ext4_destroy_lazyinit_thread();
5649 unregister_as_ext2();
5650 unregister_as_ext3();
5651 unregister_filesystem(&ext4_fs_type);
5652 destroy_inodecache();
5653 ext4_exit_mballoc();
5654 ext4_exit_feat_adverts();
5655 remove_proc_entry("fs/ext4", NULL);
5656 kset_unregister(ext4_kset);
5657 ext4_exit_system_zone();
5658 ext4_exit_pageio();
5659 ext4_exit_es();
5660 }
5661
5662 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5663 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5664 MODULE_LICENSE("GPL");
5665 module_init(ext4_init_fs)
5666 module_exit(ext4_exit_fs)
Linux kernel - Deliverables
This page took 0.154732 seconds and 5 git commands to generate.