?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 3.12.28
[linux/fpc-iii.git] / fs / ext4 / super.c
blob9afc4ba216114527b8aa22f8c16202fcb129387b
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/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
43
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
46
47 #include "ext4.h"
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
50 #include "xattr.h"
51 #include "acl.h"
52 #include "mballoc.h"
53
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
56
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
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_sync_fs_nojournal(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
86
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94 };
95 MODULE_ALIAS_FS("ext2");
96 MODULE_ALIAS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #else
99 #define IS_EXT2_SB(sb) (0)
100 #endif
101
102
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
106 .name = "ext3",
107 .mount = ext4_mount,
108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
110 };
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114 #else
115 #define IS_EXT3_SB(sb) (0)
116 #endif
117
118 static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
120 {
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
123 return 1;
124
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
126 }
127
128 static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
130 {
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
133 __u32 csum;
134
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136
137 return cpu_to_le32(csum);
138 }
139
140 int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
142 {
143 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
144 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
145 return 1;
146
147 return es->s_checksum == ext4_superblock_csum(sb, es);
148 }
149
150 void ext4_superblock_csum_set(struct super_block *sb)
151 {
152 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
153
154 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
155 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
156 return;
157
158 es->s_checksum = ext4_superblock_csum(sb, es);
159 }
160
161 void *ext4_kvmalloc(size_t size, gfp_t flags)
162 {
163 void *ret;
164
165 ret = kmalloc(size, flags | __GFP_NOWARN);
166 if (!ret)
167 ret = __vmalloc(size, flags, PAGE_KERNEL);
168 return ret;
169 }
170
171 void *ext4_kvzalloc(size_t size, gfp_t flags)
172 {
173 void *ret;
174
175 ret = kzalloc(size, flags | __GFP_NOWARN);
176 if (!ret)
177 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
178 return ret;
179 }
180
181 void ext4_kvfree(void *ptr)
182 {
183 if (is_vmalloc_addr(ptr))
184 vfree(ptr);
185 else
186 kfree(ptr);
187
188 }
189
190 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
191 struct ext4_group_desc *bg)
192 {
193 return le32_to_cpu(bg->bg_block_bitmap_lo) |
194 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
195 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
196 }
197
198 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
199 struct ext4_group_desc *bg)
200 {
201 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
202 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
203 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
204 }
205
206 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
207 struct ext4_group_desc *bg)
208 {
209 return le32_to_cpu(bg->bg_inode_table_lo) |
210 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
211 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
212 }
213
214 __u32 ext4_free_group_clusters(struct super_block *sb,
215 struct ext4_group_desc *bg)
216 {
217 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
218 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
219 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
220 }
221
222 __u32 ext4_free_inodes_count(struct super_block *sb,
223 struct ext4_group_desc *bg)
224 {
225 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
226 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
227 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
228 }
229
230 __u32 ext4_used_dirs_count(struct super_block *sb,
231 struct ext4_group_desc *bg)
232 {
233 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
234 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
235 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
236 }
237
238 __u32 ext4_itable_unused_count(struct super_block *sb,
239 struct ext4_group_desc *bg)
240 {
241 return le16_to_cpu(bg->bg_itable_unused_lo) |
242 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
243 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
244 }
245
246 void ext4_block_bitmap_set(struct super_block *sb,
247 struct ext4_group_desc *bg, ext4_fsblk_t blk)
248 {
249 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
250 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
251 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
252 }
253
254 void ext4_inode_bitmap_set(struct super_block *sb,
255 struct ext4_group_desc *bg, ext4_fsblk_t blk)
256 {
257 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
258 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
259 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
260 }
261
262 void ext4_inode_table_set(struct super_block *sb,
263 struct ext4_group_desc *bg, ext4_fsblk_t blk)
264 {
265 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
266 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
267 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
268 }
269
270 void ext4_free_group_clusters_set(struct super_block *sb,
271 struct ext4_group_desc *bg, __u32 count)
272 {
273 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
274 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
275 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
276 }
277
278 void ext4_free_inodes_set(struct super_block *sb,
279 struct ext4_group_desc *bg, __u32 count)
280 {
281 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
282 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
283 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
284 }
285
286 void ext4_used_dirs_set(struct super_block *sb,
287 struct ext4_group_desc *bg, __u32 count)
288 {
289 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
290 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
291 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
292 }
293
294 void ext4_itable_unused_set(struct super_block *sb,
295 struct ext4_group_desc *bg, __u32 count)
296 {
297 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
298 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
299 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
300 }
301
302
303 static void __save_error_info(struct super_block *sb, const char *func,
304 unsigned int line)
305 {
306 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
307
308 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
309 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
310 es->s_last_error_time = cpu_to_le32(get_seconds());
311 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
312 es->s_last_error_line = cpu_to_le32(line);
313 if (!es->s_first_error_time) {
314 es->s_first_error_time = es->s_last_error_time;
315 strncpy(es->s_first_error_func, func,
316 sizeof(es->s_first_error_func));
317 es->s_first_error_line = cpu_to_le32(line);
318 es->s_first_error_ino = es->s_last_error_ino;
319 es->s_first_error_block = es->s_last_error_block;
320 }
321 /*
322 * Start the daily error reporting function if it hasn't been
323 * started already
324 */
325 if (!es->s_error_count)
326 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
327 le32_add_cpu(&es->s_error_count, 1);
328 }
329
330 static void save_error_info(struct super_block *sb, const char *func,
331 unsigned int line)
332 {
333 __save_error_info(sb, func, line);
334 ext4_commit_super(sb, 1);
335 }
336
337 /*
338 * The del_gendisk() function uninitializes the disk-specific data
339 * structures, including the bdi structure, without telling anyone
340 * else. Once this happens, any attempt to call mark_buffer_dirty()
341 * (for example, by ext4_commit_super), will cause a kernel OOPS.
342 * This is a kludge to prevent these oops until we can put in a proper
343 * hook in del_gendisk() to inform the VFS and file system layers.
344 */
345 static int block_device_ejected(struct super_block *sb)
346 {
347 struct inode *bd_inode = sb->s_bdev->bd_inode;
348 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
349
350 return bdi->dev == NULL;
351 }
352
353 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
354 {
355 struct super_block *sb = journal->j_private;
356 struct ext4_sb_info *sbi = EXT4_SB(sb);
357 int error = is_journal_aborted(journal);
358 struct ext4_journal_cb_entry *jce;
359
360 BUG_ON(txn->t_state == T_FINISHED);
361 spin_lock(&sbi->s_md_lock);
362 while (!list_empty(&txn->t_private_list)) {
363 jce = list_entry(txn->t_private_list.next,
364 struct ext4_journal_cb_entry, jce_list);
365 list_del_init(&jce->jce_list);
366 spin_unlock(&sbi->s_md_lock);
367 jce->jce_func(sb, jce, error);
368 spin_lock(&sbi->s_md_lock);
369 }
370 spin_unlock(&sbi->s_md_lock);
371 }
372
373 /* Deal with the reporting of failure conditions on a filesystem such as
374 * inconsistencies detected or read IO failures.
375 *
376 * On ext2, we can store the error state of the filesystem in the
377 * superblock. That is not possible on ext4, because we may have other
378 * write ordering constraints on the superblock which prevent us from
379 * writing it out straight away; and given that the journal is about to
380 * be aborted, we can't rely on the current, or future, transactions to
381 * write out the superblock safely.
382 *
383 * We'll just use the jbd2_journal_abort() error code to record an error in
384 * the journal instead. On recovery, the journal will complain about
385 * that error until we've noted it down and cleared it.
386 */
387
388 static void ext4_handle_error(struct super_block *sb)
389 {
390 if (sb->s_flags & MS_RDONLY)
391 return;
392
393 if (!test_opt(sb, ERRORS_CONT)) {
394 journal_t *journal = EXT4_SB(sb)->s_journal;
395
396 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
397 if (journal)
398 jbd2_journal_abort(journal, -EIO);
399 }
400 if (test_opt(sb, ERRORS_RO)) {
401 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
402 /*
403 * Make sure updated value of ->s_mount_flags will be visible
404 * before ->s_flags update
405 */
406 smp_wmb();
407 sb->s_flags |= MS_RDONLY;
408 }
409 if (test_opt(sb, ERRORS_PANIC))
410 panic("EXT4-fs (device %s): panic forced after error\n",
411 sb->s_id);
412 }
413
414 void __ext4_error(struct super_block *sb, const char *function,
415 unsigned int line, const char *fmt, ...)
416 {
417 struct va_format vaf;
418 va_list args;
419
420 va_start(args, fmt);
421 vaf.fmt = fmt;
422 vaf.va = &args;
423 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
424 sb->s_id, function, line, current->comm, &vaf);
425 va_end(args);
426 save_error_info(sb, function, line);
427
428 ext4_handle_error(sb);
429 }
430
431 void __ext4_error_inode(struct inode *inode, const char *function,
432 unsigned int line, ext4_fsblk_t block,
433 const char *fmt, ...)
434 {
435 va_list args;
436 struct va_format vaf;
437 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
438
439 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
440 es->s_last_error_block = cpu_to_le64(block);
441 save_error_info(inode->i_sb, function, line);
442 va_start(args, fmt);
443 vaf.fmt = fmt;
444 vaf.va = &args;
445 if (block)
446 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
447 "inode #%lu: block %llu: comm %s: %pV\n",
448 inode->i_sb->s_id, function, line, inode->i_ino,
449 block, current->comm, &vaf);
450 else
451 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
452 "inode #%lu: comm %s: %pV\n",
453 inode->i_sb->s_id, function, line, inode->i_ino,
454 current->comm, &vaf);
455 va_end(args);
456
457 ext4_handle_error(inode->i_sb);
458 }
459
460 void __ext4_error_file(struct file *file, const char *function,
461 unsigned int line, ext4_fsblk_t block,
462 const char *fmt, ...)
463 {
464 va_list args;
465 struct va_format vaf;
466 struct ext4_super_block *es;
467 struct inode *inode = file_inode(file);
468 char pathname[80], *path;
469
470 es = EXT4_SB(inode->i_sb)->s_es;
471 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
472 save_error_info(inode->i_sb, function, line);
473 path = d_path(&(file->f_path), pathname, sizeof(pathname));
474 if (IS_ERR(path))
475 path = "(unknown)";
476 va_start(args, fmt);
477 vaf.fmt = fmt;
478 vaf.va = &args;
479 if (block)
480 printk(KERN_CRIT
481 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
482 "block %llu: comm %s: path %s: %pV\n",
483 inode->i_sb->s_id, function, line, inode->i_ino,
484 block, current->comm, path, &vaf);
485 else
486 printk(KERN_CRIT
487 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
488 "comm %s: path %s: %pV\n",
489 inode->i_sb->s_id, function, line, inode->i_ino,
490 current->comm, path, &vaf);
491 va_end(args);
492
493 ext4_handle_error(inode->i_sb);
494 }
495
496 const char *ext4_decode_error(struct super_block *sb, int errno,
497 char nbuf[16])
498 {
499 char *errstr = NULL;
500
501 switch (errno) {
502 case -EIO:
503 errstr = "IO failure";
504 break;
505 case -ENOMEM:
506 errstr = "Out of memory";
507 break;
508 case -EROFS:
509 if (!sb || (EXT4_SB(sb)->s_journal &&
510 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
511 errstr = "Journal has aborted";
512 else
513 errstr = "Readonly filesystem";
514 break;
515 default:
516 /* If the caller passed in an extra buffer for unknown
517 * errors, textualise them now. Else we just return
518 * NULL. */
519 if (nbuf) {
520 /* Check for truncated error codes... */
521 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
522 errstr = nbuf;
523 }
524 break;
525 }
526
527 return errstr;
528 }
529
530 /* __ext4_std_error decodes expected errors from journaling functions
531 * automatically and invokes the appropriate error response. */
532
533 void __ext4_std_error(struct super_block *sb, const char *function,
534 unsigned int line, int errno)
535 {
536 char nbuf[16];
537 const char *errstr;
538
539 /* Special case: if the error is EROFS, and we're not already
540 * inside a transaction, then there's really no point in logging
541 * an error. */
542 if (errno == -EROFS && journal_current_handle() == NULL &&
543 (sb->s_flags & MS_RDONLY))
544 return;
545
546 errstr = ext4_decode_error(sb, errno, nbuf);
547 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
548 sb->s_id, function, line, errstr);
549 save_error_info(sb, function, line);
550
551 ext4_handle_error(sb);
552 }
553
554 /*
555 * ext4_abort is a much stronger failure handler than ext4_error. The
556 * abort function may be used to deal with unrecoverable failures such
557 * as journal IO errors or ENOMEM at a critical moment in log management.
558 *
559 * We unconditionally force the filesystem into an ABORT|READONLY state,
560 * unless the error response on the fs has been set to panic in which
561 * case we take the easy way out and panic immediately.
562 */
563
564 void __ext4_abort(struct super_block *sb, const char *function,
565 unsigned int line, const char *fmt, ...)
566 {
567 va_list args;
568
569 save_error_info(sb, function, line);
570 va_start(args, fmt);
571 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
572 function, line);
573 vprintk(fmt, args);
574 printk("\n");
575 va_end(args);
576
577 if ((sb->s_flags & MS_RDONLY) == 0) {
578 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
579 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
580 /*
581 * Make sure updated value of ->s_mount_flags will be visible
582 * before ->s_flags update
583 */
584 smp_wmb();
585 sb->s_flags |= MS_RDONLY;
586 if (EXT4_SB(sb)->s_journal)
587 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
588 save_error_info(sb, function, line);
589 }
590 if (test_opt(sb, ERRORS_PANIC))
591 panic("EXT4-fs panic from previous error\n");
592 }
593
594 void __ext4_msg(struct super_block *sb,
595 const char *prefix, const char *fmt, ...)
596 {
597 struct va_format vaf;
598 va_list args;
599
600 va_start(args, fmt);
601 vaf.fmt = fmt;
602 vaf.va = &args;
603 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
604 va_end(args);
605 }
606
607 void __ext4_warning(struct super_block *sb, const char *function,
608 unsigned int line, const char *fmt, ...)
609 {
610 struct va_format vaf;
611 va_list args;
612
613 va_start(args, fmt);
614 vaf.fmt = fmt;
615 vaf.va = &args;
616 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
617 sb->s_id, function, line, &vaf);
618 va_end(args);
619 }
620
621 void __ext4_grp_locked_error(const char *function, unsigned int line,
622 struct super_block *sb, ext4_group_t grp,
623 unsigned long ino, ext4_fsblk_t block,
624 const char *fmt, ...)
625 __releases(bitlock)
626 __acquires(bitlock)
627 {
628 struct va_format vaf;
629 va_list args;
630 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
631
632 es->s_last_error_ino = cpu_to_le32(ino);
633 es->s_last_error_block = cpu_to_le64(block);
634 __save_error_info(sb, function, line);
635
636 va_start(args, fmt);
637
638 vaf.fmt = fmt;
639 vaf.va = &args;
640 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
641 sb->s_id, function, line, grp);
642 if (ino)
643 printk(KERN_CONT "inode %lu: ", ino);
644 if (block)
645 printk(KERN_CONT "block %llu:", (unsigned long long) block);
646 printk(KERN_CONT "%pV\n", &vaf);
647 va_end(args);
648
649 if (test_opt(sb, ERRORS_CONT)) {
650 ext4_commit_super(sb, 0);
651 return;
652 }
653
654 ext4_unlock_group(sb, grp);
655 ext4_handle_error(sb);
656 /*
657 * We only get here in the ERRORS_RO case; relocking the group
658 * may be dangerous, but nothing bad will happen since the
659 * filesystem will have already been marked read/only and the
660 * journal has been aborted. We return 1 as a hint to callers
661 * who might what to use the return value from
662 * ext4_grp_locked_error() to distinguish between the
663 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
664 * aggressively from the ext4 function in question, with a
665 * more appropriate error code.
666 */
667 ext4_lock_group(sb, grp);
668 return;
669 }
670
671 void ext4_update_dynamic_rev(struct super_block *sb)
672 {
673 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
674
675 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
676 return;
677
678 ext4_warning(sb,
679 "updating to rev %d because of new feature flag, "
680 "running e2fsck is recommended",
681 EXT4_DYNAMIC_REV);
682
683 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
684 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
685 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
686 /* leave es->s_feature_*compat flags alone */
687 /* es->s_uuid will be set by e2fsck if empty */
688
689 /*
690 * The rest of the superblock fields should be zero, and if not it
691 * means they are likely already in use, so leave them alone. We
692 * can leave it up to e2fsck to clean up any inconsistencies there.
693 */
694 }
695
696 /*
697 * Open the external journal device
698 */
699 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
700 {
701 struct block_device *bdev;
702 char b[BDEVNAME_SIZE];
703
704 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
705 if (IS_ERR(bdev))
706 goto fail;
707 return bdev;
708
709 fail:
710 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
711 __bdevname(dev, b), PTR_ERR(bdev));
712 return NULL;
713 }
714
715 /*
716 * Release the journal device
717 */
718 static void ext4_blkdev_put(struct block_device *bdev)
719 {
720 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
721 }
722
723 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
724 {
725 struct block_device *bdev;
726 bdev = sbi->journal_bdev;
727 if (bdev) {
728 ext4_blkdev_put(bdev);
729 sbi->journal_bdev = NULL;
730 }
731 }
732
733 static inline struct inode *orphan_list_entry(struct list_head *l)
734 {
735 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
736 }
737
738 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
739 {
740 struct list_head *l;
741
742 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
743 le32_to_cpu(sbi->s_es->s_last_orphan));
744
745 printk(KERN_ERR "sb_info orphan list:\n");
746 list_for_each(l, &sbi->s_orphan) {
747 struct inode *inode = orphan_list_entry(l);
748 printk(KERN_ERR " "
749 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
750 inode->i_sb->s_id, inode->i_ino, inode,
751 inode->i_mode, inode->i_nlink,
752 NEXT_ORPHAN(inode));
753 }
754 }
755
756 static void ext4_put_super(struct super_block *sb)
757 {
758 struct ext4_sb_info *sbi = EXT4_SB(sb);
759 struct ext4_super_block *es = sbi->s_es;
760 int i, err;
761
762 ext4_unregister_li_request(sb);
763 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
764
765 flush_workqueue(sbi->rsv_conversion_wq);
766 destroy_workqueue(sbi->rsv_conversion_wq);
767
768 if (sbi->s_journal) {
769 err = jbd2_journal_destroy(sbi->s_journal);
770 sbi->s_journal = NULL;
771 if (err < 0)
772 ext4_abort(sb, "Couldn't clean up the journal");
773 }
774
775 ext4_es_unregister_shrinker(sbi);
776 del_timer_sync(&sbi->s_err_report);
777 ext4_release_system_zone(sb);
778 ext4_mb_release(sb);
779 ext4_ext_release(sb);
780 ext4_xattr_put_super(sb);
781
782 if (!(sb->s_flags & MS_RDONLY)) {
783 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
784 es->s_state = cpu_to_le16(sbi->s_mount_state);
785 }
786 if (!(sb->s_flags & MS_RDONLY))
787 ext4_commit_super(sb, 1);
788
789 if (sbi->s_proc) {
790 remove_proc_entry("options", sbi->s_proc);
791 remove_proc_entry(sb->s_id, ext4_proc_root);
792 }
793 kobject_del(&sbi->s_kobj);
794
795 for (i = 0; i < sbi->s_gdb_count; i++)
796 brelse(sbi->s_group_desc[i]);
797 ext4_kvfree(sbi->s_group_desc);
798 ext4_kvfree(sbi->s_flex_groups);
799 percpu_counter_destroy(&sbi->s_freeclusters_counter);
800 percpu_counter_destroy(&sbi->s_freeinodes_counter);
801 percpu_counter_destroy(&sbi->s_dirs_counter);
802 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
803 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
804 brelse(sbi->s_sbh);
805 #ifdef CONFIG_QUOTA
806 for (i = 0; i < MAXQUOTAS; i++)
807 kfree(sbi->s_qf_names[i]);
808 #endif
809
810 /* Debugging code just in case the in-memory inode orphan list
811 * isn't empty. The on-disk one can be non-empty if we've
812 * detected an error and taken the fs readonly, but the
813 * in-memory list had better be clean by this point. */
814 if (!list_empty(&sbi->s_orphan))
815 dump_orphan_list(sb, sbi);
816 J_ASSERT(list_empty(&sbi->s_orphan));
817
818 invalidate_bdev(sb->s_bdev);
819 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
820 /*
821 * Invalidate the journal device's buffers. We don't want them
822 * floating about in memory - the physical journal device may
823 * hotswapped, and it breaks the `ro-after' testing code.
824 */
825 sync_blockdev(sbi->journal_bdev);
826 invalidate_bdev(sbi->journal_bdev);
827 ext4_blkdev_remove(sbi);
828 }
829 if (sbi->s_mmp_tsk)
830 kthread_stop(sbi->s_mmp_tsk);
831 sb->s_fs_info = NULL;
832 /*
833 * Now that we are completely done shutting down the
834 * superblock, we need to actually destroy the kobject.
835 */
836 kobject_put(&sbi->s_kobj);
837 wait_for_completion(&sbi->s_kobj_unregister);
838 if (sbi->s_chksum_driver)
839 crypto_free_shash(sbi->s_chksum_driver);
840 kfree(sbi->s_blockgroup_lock);
841 kfree(sbi);
842 }
843
844 static struct kmem_cache *ext4_inode_cachep;
845
846 /*
847 * Called inside transaction, so use GFP_NOFS
848 */
849 static struct inode *ext4_alloc_inode(struct super_block *sb)
850 {
851 struct ext4_inode_info *ei;
852
853 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
854 if (!ei)
855 return NULL;
856
857 ei->vfs_inode.i_version = 1;
858 INIT_LIST_HEAD(&ei->i_prealloc_list);
859 spin_lock_init(&ei->i_prealloc_lock);
860 ext4_es_init_tree(&ei->i_es_tree);
861 rwlock_init(&ei->i_es_lock);
862 INIT_LIST_HEAD(&ei->i_es_lru);
863 ei->i_es_lru_nr = 0;
864 ei->i_touch_when = 0;
865 ei->i_reserved_data_blocks = 0;
866 ei->i_reserved_meta_blocks = 0;
867 ei->i_allocated_meta_blocks = 0;
868 ei->i_da_metadata_calc_len = 0;
869 ei->i_da_metadata_calc_last_lblock = 0;
870 spin_lock_init(&(ei->i_block_reservation_lock));
871 #ifdef CONFIG_QUOTA
872 ei->i_reserved_quota = 0;
873 #endif
874 ei->jinode = NULL;
875 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
876 spin_lock_init(&ei->i_completed_io_lock);
877 ei->i_sync_tid = 0;
878 ei->i_datasync_tid = 0;
879 atomic_set(&ei->i_ioend_count, 0);
880 atomic_set(&ei->i_unwritten, 0);
881 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
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_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 ext4_es_lru_del(inode);
954 if (EXT4_I(inode)->jinode) {
955 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
956 EXT4_I(inode)->jinode);
957 jbd2_free_inode(EXT4_I(inode)->jinode);
958 EXT4_I(inode)->jinode = NULL;
959 }
960 }
961
962 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
963 u64 ino, u32 generation)
964 {
965 struct inode *inode;
966
967 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
968 return ERR_PTR(-ESTALE);
969 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
970 return ERR_PTR(-ESTALE);
971
972 /* iget isn't really right if the inode is currently unallocated!!
973 *
974 * ext4_read_inode will return a bad_inode if the inode had been
975 * deleted, so we should be safe.
976 *
977 * Currently we don't know the generation for parent directory, so
978 * a generation of 0 means "accept any"
979 */
980 inode = ext4_iget(sb, ino);
981 if (IS_ERR(inode))
982 return ERR_CAST(inode);
983 if (generation && inode->i_generation != generation) {
984 iput(inode);
985 return ERR_PTR(-ESTALE);
986 }
987
988 return inode;
989 }
990
991 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
992 int fh_len, int fh_type)
993 {
994 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
995 ext4_nfs_get_inode);
996 }
997
998 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
999 int fh_len, int fh_type)
1000 {
1001 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1002 ext4_nfs_get_inode);
1003 }
1004
1005 /*
1006 * Try to release metadata pages (indirect blocks, directories) which are
1007 * mapped via the block device. Since these pages could have journal heads
1008 * which would prevent try_to_free_buffers() from freeing them, we must use
1009 * jbd2 layer's try_to_free_buffers() function to release them.
1010 */
1011 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1012 gfp_t wait)
1013 {
1014 journal_t *journal = EXT4_SB(sb)->s_journal;
1015
1016 WARN_ON(PageChecked(page));
1017 if (!page_has_buffers(page))
1018 return 0;
1019 if (journal)
1020 return jbd2_journal_try_to_free_buffers(journal, page,
1021 wait & ~__GFP_WAIT);
1022 return try_to_free_buffers(page);
1023 }
1024
1025 #ifdef CONFIG_QUOTA
1026 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1027 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1028
1029 static int ext4_write_dquot(struct dquot *dquot);
1030 static int ext4_acquire_dquot(struct dquot *dquot);
1031 static int ext4_release_dquot(struct dquot *dquot);
1032 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1033 static int ext4_write_info(struct super_block *sb, int type);
1034 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1035 struct path *path);
1036 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1037 int format_id);
1038 static int ext4_quota_off(struct super_block *sb, int type);
1039 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1040 static int ext4_quota_on_mount(struct super_block *sb, int type);
1041 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1042 size_t len, loff_t off);
1043 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1044 const char *data, size_t len, loff_t off);
1045 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1046 unsigned int flags);
1047 static int ext4_enable_quotas(struct super_block *sb);
1048
1049 static const struct dquot_operations ext4_quota_operations = {
1050 .get_reserved_space = ext4_get_reserved_space,
1051 .write_dquot = ext4_write_dquot,
1052 .acquire_dquot = ext4_acquire_dquot,
1053 .release_dquot = ext4_release_dquot,
1054 .mark_dirty = ext4_mark_dquot_dirty,
1055 .write_info = ext4_write_info,
1056 .alloc_dquot = dquot_alloc,
1057 .destroy_dquot = dquot_destroy,
1058 };
1059
1060 static const struct quotactl_ops ext4_qctl_operations = {
1061 .quota_on = ext4_quota_on,
1062 .quota_off = ext4_quota_off,
1063 .quota_sync = dquot_quota_sync,
1064 .get_info = dquot_get_dqinfo,
1065 .set_info = dquot_set_dqinfo,
1066 .get_dqblk = dquot_get_dqblk,
1067 .set_dqblk = dquot_set_dqblk
1068 };
1069
1070 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1071 .quota_on_meta = ext4_quota_on_sysfile,
1072 .quota_off = ext4_quota_off_sysfile,
1073 .quota_sync = dquot_quota_sync,
1074 .get_info = dquot_get_dqinfo,
1075 .set_info = dquot_set_dqinfo,
1076 .get_dqblk = dquot_get_dqblk,
1077 .set_dqblk = dquot_set_dqblk
1078 };
1079 #endif
1080
1081 static const struct super_operations ext4_sops = {
1082 .alloc_inode = ext4_alloc_inode,
1083 .destroy_inode = ext4_destroy_inode,
1084 .write_inode = ext4_write_inode,
1085 .dirty_inode = ext4_dirty_inode,
1086 .drop_inode = ext4_drop_inode,
1087 .evict_inode = ext4_evict_inode,
1088 .put_super = ext4_put_super,
1089 .sync_fs = ext4_sync_fs,
1090 .freeze_fs = ext4_freeze,
1091 .unfreeze_fs = ext4_unfreeze,
1092 .statfs = ext4_statfs,
1093 .remount_fs = ext4_remount,
1094 .show_options = ext4_show_options,
1095 #ifdef CONFIG_QUOTA
1096 .quota_read = ext4_quota_read,
1097 .quota_write = ext4_quota_write,
1098 #endif
1099 .bdev_try_to_free_page = bdev_try_to_free_page,
1100 };
1101
1102 static const struct super_operations ext4_nojournal_sops = {
1103 .alloc_inode = ext4_alloc_inode,
1104 .destroy_inode = ext4_destroy_inode,
1105 .write_inode = ext4_write_inode,
1106 .dirty_inode = ext4_dirty_inode,
1107 .drop_inode = ext4_drop_inode,
1108 .evict_inode = ext4_evict_inode,
1109 .sync_fs = ext4_sync_fs_nojournal,
1110 .put_super = ext4_put_super,
1111 .statfs = ext4_statfs,
1112 .remount_fs = ext4_remount,
1113 .show_options = ext4_show_options,
1114 #ifdef CONFIG_QUOTA
1115 .quota_read = ext4_quota_read,
1116 .quota_write = ext4_quota_write,
1117 #endif
1118 .bdev_try_to_free_page = bdev_try_to_free_page,
1119 };
1120
1121 static const struct export_operations ext4_export_ops = {
1122 .fh_to_dentry = ext4_fh_to_dentry,
1123 .fh_to_parent = ext4_fh_to_parent,
1124 .get_parent = ext4_get_parent,
1125 };
1126
1127 enum {
1128 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1129 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1130 Opt_nouid32, Opt_debug, Opt_removed,
1131 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1132 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1133 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1134 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1135 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1136 Opt_data_err_abort, Opt_data_err_ignore,
1137 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1138 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1139 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1140 Opt_usrquota, Opt_grpquota, Opt_i_version,
1141 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1142 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1143 Opt_inode_readahead_blks, Opt_journal_ioprio,
1144 Opt_dioread_nolock, Opt_dioread_lock,
1145 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1146 Opt_max_dir_size_kb,
1147 };
1148
1149 static const match_table_t tokens = {
1150 {Opt_bsd_df, "bsddf"},
1151 {Opt_minix_df, "minixdf"},
1152 {Opt_grpid, "grpid"},
1153 {Opt_grpid, "bsdgroups"},
1154 {Opt_nogrpid, "nogrpid"},
1155 {Opt_nogrpid, "sysvgroups"},
1156 {Opt_resgid, "resgid=%u"},
1157 {Opt_resuid, "resuid=%u"},
1158 {Opt_sb, "sb=%u"},
1159 {Opt_err_cont, "errors=continue"},
1160 {Opt_err_panic, "errors=panic"},
1161 {Opt_err_ro, "errors=remount-ro"},
1162 {Opt_nouid32, "nouid32"},
1163 {Opt_debug, "debug"},
1164 {Opt_removed, "oldalloc"},
1165 {Opt_removed, "orlov"},
1166 {Opt_user_xattr, "user_xattr"},
1167 {Opt_nouser_xattr, "nouser_xattr"},
1168 {Opt_acl, "acl"},
1169 {Opt_noacl, "noacl"},
1170 {Opt_noload, "norecovery"},
1171 {Opt_noload, "noload"},
1172 {Opt_removed, "nobh"},
1173 {Opt_removed, "bh"},
1174 {Opt_commit, "commit=%u"},
1175 {Opt_min_batch_time, "min_batch_time=%u"},
1176 {Opt_max_batch_time, "max_batch_time=%u"},
1177 {Opt_journal_dev, "journal_dev=%u"},
1178 {Opt_journal_path, "journal_path=%s"},
1179 {Opt_journal_checksum, "journal_checksum"},
1180 {Opt_journal_async_commit, "journal_async_commit"},
1181 {Opt_abort, "abort"},
1182 {Opt_data_journal, "data=journal"},
1183 {Opt_data_ordered, "data=ordered"},
1184 {Opt_data_writeback, "data=writeback"},
1185 {Opt_data_err_abort, "data_err=abort"},
1186 {Opt_data_err_ignore, "data_err=ignore"},
1187 {Opt_offusrjquota, "usrjquota="},
1188 {Opt_usrjquota, "usrjquota=%s"},
1189 {Opt_offgrpjquota, "grpjquota="},
1190 {Opt_grpjquota, "grpjquota=%s"},
1191 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1192 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1193 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1194 {Opt_grpquota, "grpquota"},
1195 {Opt_noquota, "noquota"},
1196 {Opt_quota, "quota"},
1197 {Opt_usrquota, "usrquota"},
1198 {Opt_barrier, "barrier=%u"},
1199 {Opt_barrier, "barrier"},
1200 {Opt_nobarrier, "nobarrier"},
1201 {Opt_i_version, "i_version"},
1202 {Opt_stripe, "stripe=%u"},
1203 {Opt_delalloc, "delalloc"},
1204 {Opt_nodelalloc, "nodelalloc"},
1205 {Opt_removed, "mblk_io_submit"},
1206 {Opt_removed, "nomblk_io_submit"},
1207 {Opt_block_validity, "block_validity"},
1208 {Opt_noblock_validity, "noblock_validity"},
1209 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1210 {Opt_journal_ioprio, "journal_ioprio=%u"},
1211 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1212 {Opt_auto_da_alloc, "auto_da_alloc"},
1213 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1214 {Opt_dioread_nolock, "dioread_nolock"},
1215 {Opt_dioread_lock, "dioread_lock"},
1216 {Opt_discard, "discard"},
1217 {Opt_nodiscard, "nodiscard"},
1218 {Opt_init_itable, "init_itable=%u"},
1219 {Opt_init_itable, "init_itable"},
1220 {Opt_noinit_itable, "noinit_itable"},
1221 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1222 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1223 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1224 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1225 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1226 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1227 {Opt_err, NULL},
1228 };
1229
1230 static ext4_fsblk_t get_sb_block(void **data)
1231 {
1232 ext4_fsblk_t sb_block;
1233 char *options = (char *) *data;
1234
1235 if (!options || strncmp(options, "sb=", 3) != 0)
1236 return 1; /* Default location */
1237
1238 options += 3;
1239 /* TODO: use simple_strtoll with >32bit ext4 */
1240 sb_block = simple_strtoul(options, &options, 0);
1241 if (*options && *options != ',') {
1242 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1243 (char *) *data);
1244 return 1;
1245 }
1246 if (*options == ',')
1247 options++;
1248 *data = (void *) options;
1249
1250 return sb_block;
1251 }
1252
1253 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1254 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1255 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1256
1257 #ifdef CONFIG_QUOTA
1258 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1259 {
1260 struct ext4_sb_info *sbi = EXT4_SB(sb);
1261 char *qname;
1262 int ret = -1;
1263
1264 if (sb_any_quota_loaded(sb) &&
1265 !sbi->s_qf_names[qtype]) {
1266 ext4_msg(sb, KERN_ERR,
1267 "Cannot change journaled "
1268 "quota options when quota turned on");
1269 return -1;
1270 }
1271 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1272 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1273 "when QUOTA feature is enabled");
1274 return -1;
1275 }
1276 qname = match_strdup(args);
1277 if (!qname) {
1278 ext4_msg(sb, KERN_ERR,
1279 "Not enough memory for storing quotafile name");
1280 return -1;
1281 }
1282 if (sbi->s_qf_names[qtype]) {
1283 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1284 ret = 1;
1285 else
1286 ext4_msg(sb, KERN_ERR,
1287 "%s quota file already specified",
1288 QTYPE2NAME(qtype));
1289 goto errout;
1290 }
1291 if (strchr(qname, '/')) {
1292 ext4_msg(sb, KERN_ERR,
1293 "quotafile must be on filesystem root");
1294 goto errout;
1295 }
1296 sbi->s_qf_names[qtype] = qname;
1297 set_opt(sb, QUOTA);
1298 return 1;
1299 errout:
1300 kfree(qname);
1301 return ret;
1302 }
1303
1304 static int clear_qf_name(struct super_block *sb, int qtype)
1305 {
1306
1307 struct ext4_sb_info *sbi = EXT4_SB(sb);
1308
1309 if (sb_any_quota_loaded(sb) &&
1310 sbi->s_qf_names[qtype]) {
1311 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1312 " when quota turned on");
1313 return -1;
1314 }
1315 kfree(sbi->s_qf_names[qtype]);
1316 sbi->s_qf_names[qtype] = NULL;
1317 return 1;
1318 }
1319 #endif
1320
1321 #define MOPT_SET 0x0001
1322 #define MOPT_CLEAR 0x0002
1323 #define MOPT_NOSUPPORT 0x0004
1324 #define MOPT_EXPLICIT 0x0008
1325 #define MOPT_CLEAR_ERR 0x0010
1326 #define MOPT_GTE0 0x0020
1327 #ifdef CONFIG_QUOTA
1328 #define MOPT_Q 0
1329 #define MOPT_QFMT 0x0040
1330 #else
1331 #define MOPT_Q MOPT_NOSUPPORT
1332 #define MOPT_QFMT MOPT_NOSUPPORT
1333 #endif
1334 #define MOPT_DATAJ 0x0080
1335 #define MOPT_NO_EXT2 0x0100
1336 #define MOPT_NO_EXT3 0x0200
1337 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1338 #define MOPT_STRING 0x0400
1339
1340 static const struct mount_opts {
1341 int token;
1342 int mount_opt;
1343 int flags;
1344 } ext4_mount_opts[] = {
1345 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1346 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1347 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1348 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1349 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1350 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1351 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1352 MOPT_EXT4_ONLY | MOPT_SET},
1353 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1354 MOPT_EXT4_ONLY | MOPT_CLEAR},
1355 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1356 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1357 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1358 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1359 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1360 MOPT_EXT4_ONLY | MOPT_CLEAR},
1361 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1362 MOPT_EXT4_ONLY | MOPT_SET},
1363 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1364 EXT4_MOUNT_JOURNAL_CHECKSUM),
1365 MOPT_EXT4_ONLY | MOPT_SET},
1366 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1367 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1368 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1369 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1370 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1371 MOPT_NO_EXT2 | MOPT_SET},
1372 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1373 MOPT_NO_EXT2 | MOPT_CLEAR},
1374 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1375 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1376 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1377 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1378 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1379 {Opt_commit, 0, MOPT_GTE0},
1380 {Opt_max_batch_time, 0, MOPT_GTE0},
1381 {Opt_min_batch_time, 0, MOPT_GTE0},
1382 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1383 {Opt_init_itable, 0, MOPT_GTE0},
1384 {Opt_stripe, 0, MOPT_GTE0},
1385 {Opt_resuid, 0, MOPT_GTE0},
1386 {Opt_resgid, 0, MOPT_GTE0},
1387 {Opt_journal_dev, 0, MOPT_GTE0},
1388 {Opt_journal_path, 0, MOPT_STRING},
1389 {Opt_journal_ioprio, 0, MOPT_GTE0},
1390 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1391 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1392 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1393 MOPT_NO_EXT2 | MOPT_DATAJ},
1394 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1395 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1396 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1397 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1398 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1399 #else
1400 {Opt_acl, 0, MOPT_NOSUPPORT},
1401 {Opt_noacl, 0, MOPT_NOSUPPORT},
1402 #endif
1403 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1404 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1405 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1406 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1407 MOPT_SET | MOPT_Q},
1408 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1409 MOPT_SET | MOPT_Q},
1410 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1411 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1412 {Opt_usrjquota, 0, MOPT_Q},
1413 {Opt_grpjquota, 0, MOPT_Q},
1414 {Opt_offusrjquota, 0, MOPT_Q},
1415 {Opt_offgrpjquota, 0, MOPT_Q},
1416 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1417 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1418 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1419 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1420 {Opt_err, 0, 0}
1421 };
1422
1423 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1424 substring_t *args, unsigned long *journal_devnum,
1425 unsigned int *journal_ioprio, int is_remount)
1426 {
1427 struct ext4_sb_info *sbi = EXT4_SB(sb);
1428 const struct mount_opts *m;
1429 kuid_t uid;
1430 kgid_t gid;
1431 int arg = 0;
1432
1433 #ifdef CONFIG_QUOTA
1434 if (token == Opt_usrjquota)
1435 return set_qf_name(sb, USRQUOTA, &args[0]);
1436 else if (token == Opt_grpjquota)
1437 return set_qf_name(sb, GRPQUOTA, &args[0]);
1438 else if (token == Opt_offusrjquota)
1439 return clear_qf_name(sb, USRQUOTA);
1440 else if (token == Opt_offgrpjquota)
1441 return clear_qf_name(sb, GRPQUOTA);
1442 #endif
1443 switch (token) {
1444 case Opt_noacl:
1445 case Opt_nouser_xattr:
1446 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1447 break;
1448 case Opt_sb:
1449 return 1; /* handled by get_sb_block() */
1450 case Opt_removed:
1451 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1452 return 1;
1453 case Opt_abort:
1454 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1455 return 1;
1456 case Opt_i_version:
1457 sb->s_flags |= MS_I_VERSION;
1458 return 1;
1459 }
1460
1461 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1462 if (token == m->token)
1463 break;
1464
1465 if (m->token == Opt_err) {
1466 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1467 "or missing value", opt);
1468 return -1;
1469 }
1470
1471 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1472 ext4_msg(sb, KERN_ERR,
1473 "Mount option \"%s\" incompatible with ext2", opt);
1474 return -1;
1475 }
1476 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1477 ext4_msg(sb, KERN_ERR,
1478 "Mount option \"%s\" incompatible with ext3", opt);
1479 return -1;
1480 }
1481
1482 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1483 return -1;
1484 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1485 return -1;
1486 if (m->flags & MOPT_EXPLICIT)
1487 set_opt2(sb, EXPLICIT_DELALLOC);
1488 if (m->flags & MOPT_CLEAR_ERR)
1489 clear_opt(sb, ERRORS_MASK);
1490 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1491 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1492 "options when quota turned on");
1493 return -1;
1494 }
1495
1496 if (m->flags & MOPT_NOSUPPORT) {
1497 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1498 } else if (token == Opt_commit) {
1499 if (arg == 0)
1500 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1501 sbi->s_commit_interval = HZ * arg;
1502 } else if (token == Opt_max_batch_time) {
1503 sbi->s_max_batch_time = arg;
1504 } else if (token == Opt_min_batch_time) {
1505 sbi->s_min_batch_time = arg;
1506 } else if (token == Opt_inode_readahead_blks) {
1507 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1508 ext4_msg(sb, KERN_ERR,
1509 "EXT4-fs: inode_readahead_blks must be "
1510 "0 or a power of 2 smaller than 2^31");
1511 return -1;
1512 }
1513 sbi->s_inode_readahead_blks = arg;
1514 } else if (token == Opt_init_itable) {
1515 set_opt(sb, INIT_INODE_TABLE);
1516 if (!args->from)
1517 arg = EXT4_DEF_LI_WAIT_MULT;
1518 sbi->s_li_wait_mult = arg;
1519 } else if (token == Opt_max_dir_size_kb) {
1520 sbi->s_max_dir_size_kb = arg;
1521 } else if (token == Opt_stripe) {
1522 sbi->s_stripe = arg;
1523 } else if (token == Opt_resuid) {
1524 uid = make_kuid(current_user_ns(), arg);
1525 if (!uid_valid(uid)) {
1526 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1527 return -1;
1528 }
1529 sbi->s_resuid = uid;
1530 } else if (token == Opt_resgid) {
1531 gid = make_kgid(current_user_ns(), arg);
1532 if (!gid_valid(gid)) {
1533 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1534 return -1;
1535 }
1536 sbi->s_resgid = gid;
1537 } else if (token == Opt_journal_dev) {
1538 if (is_remount) {
1539 ext4_msg(sb, KERN_ERR,
1540 "Cannot specify journal on remount");
1541 return -1;
1542 }
1543 *journal_devnum = arg;
1544 } else if (token == Opt_journal_path) {
1545 char *journal_path;
1546 struct inode *journal_inode;
1547 struct path path;
1548 int error;
1549
1550 if (is_remount) {
1551 ext4_msg(sb, KERN_ERR,
1552 "Cannot specify journal on remount");
1553 return -1;
1554 }
1555 journal_path = match_strdup(&args[0]);
1556 if (!journal_path) {
1557 ext4_msg(sb, KERN_ERR, "error: could not dup "
1558 "journal device string");
1559 return -1;
1560 }
1561
1562 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1563 if (error) {
1564 ext4_msg(sb, KERN_ERR, "error: could not find "
1565 "journal device path: error %d", error);
1566 kfree(journal_path);
1567 return -1;
1568 }
1569
1570 journal_inode = path.dentry->d_inode;
1571 if (!S_ISBLK(journal_inode->i_mode)) {
1572 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1573 "is not a block device", journal_path);
1574 path_put(&path);
1575 kfree(journal_path);
1576 return -1;
1577 }
1578
1579 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1580 path_put(&path);
1581 kfree(journal_path);
1582 } else if (token == Opt_journal_ioprio) {
1583 if (arg > 7) {
1584 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1585 " (must be 0-7)");
1586 return -1;
1587 }
1588 *journal_ioprio =
1589 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1590 } else if (m->flags & MOPT_DATAJ) {
1591 if (is_remount) {
1592 if (!sbi->s_journal)
1593 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1594 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1595 ext4_msg(sb, KERN_ERR,
1596 "Cannot change data mode on remount");
1597 return -1;
1598 }
1599 } else {
1600 clear_opt(sb, DATA_FLAGS);
1601 sbi->s_mount_opt |= m->mount_opt;
1602 }
1603 #ifdef CONFIG_QUOTA
1604 } else if (m->flags & MOPT_QFMT) {
1605 if (sb_any_quota_loaded(sb) &&
1606 sbi->s_jquota_fmt != m->mount_opt) {
1607 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1608 "quota options when quota turned on");
1609 return -1;
1610 }
1611 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1612 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1613 ext4_msg(sb, KERN_ERR,
1614 "Cannot set journaled quota options "
1615 "when QUOTA feature is enabled");
1616 return -1;
1617 }
1618 sbi->s_jquota_fmt = m->mount_opt;
1619 #endif
1620 } else {
1621 if (!args->from)
1622 arg = 1;
1623 if (m->flags & MOPT_CLEAR)
1624 arg = !arg;
1625 else if (unlikely(!(m->flags & MOPT_SET))) {
1626 ext4_msg(sb, KERN_WARNING,
1627 "buggy handling of option %s", opt);
1628 WARN_ON(1);
1629 return -1;
1630 }
1631 if (arg != 0)
1632 sbi->s_mount_opt |= m->mount_opt;
1633 else
1634 sbi->s_mount_opt &= ~m->mount_opt;
1635 }
1636 return 1;
1637 }
1638
1639 static int parse_options(char *options, struct super_block *sb,
1640 unsigned long *journal_devnum,
1641 unsigned int *journal_ioprio,
1642 int is_remount)
1643 {
1644 struct ext4_sb_info *sbi = EXT4_SB(sb);
1645 char *p;
1646 substring_t args[MAX_OPT_ARGS];
1647 int token;
1648
1649 if (!options)
1650 return 1;
1651
1652 while ((p = strsep(&options, ",")) != NULL) {
1653 if (!*p)
1654 continue;
1655 /*
1656 * Initialize args struct so we know whether arg was
1657 * found; some options take optional arguments.
1658 */
1659 args[0].to = args[0].from = NULL;
1660 token = match_token(p, tokens, args);
1661 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1662 journal_ioprio, is_remount) < 0)
1663 return 0;
1664 }
1665 #ifdef CONFIG_QUOTA
1666 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1667 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1668 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1669 "feature is enabled");
1670 return 0;
1671 }
1672 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1673 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1674 clear_opt(sb, USRQUOTA);
1675
1676 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1677 clear_opt(sb, GRPQUOTA);
1678
1679 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1680 ext4_msg(sb, KERN_ERR, "old and new quota "
1681 "format mixing");
1682 return 0;
1683 }
1684
1685 if (!sbi->s_jquota_fmt) {
1686 ext4_msg(sb, KERN_ERR, "journaled quota format "
1687 "not specified");
1688 return 0;
1689 }
1690 } else {
1691 if (sbi->s_jquota_fmt) {
1692 ext4_msg(sb, KERN_ERR, "journaled quota format "
1693 "specified with no journaling "
1694 "enabled");
1695 return 0;
1696 }
1697 }
1698 #endif
1699 if (test_opt(sb, DIOREAD_NOLOCK)) {
1700 int blocksize =
1701 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1702
1703 if (blocksize < PAGE_CACHE_SIZE) {
1704 ext4_msg(sb, KERN_ERR, "can't mount with "
1705 "dioread_nolock if block size != PAGE_SIZE");
1706 return 0;
1707 }
1708 }
1709 return 1;
1710 }
1711
1712 static inline void ext4_show_quota_options(struct seq_file *seq,
1713 struct super_block *sb)
1714 {
1715 #if defined(CONFIG_QUOTA)
1716 struct ext4_sb_info *sbi = EXT4_SB(sb);
1717
1718 if (sbi->s_jquota_fmt) {
1719 char *fmtname = "";
1720
1721 switch (sbi->s_jquota_fmt) {
1722 case QFMT_VFS_OLD:
1723 fmtname = "vfsold";
1724 break;
1725 case QFMT_VFS_V0:
1726 fmtname = "vfsv0";
1727 break;
1728 case QFMT_VFS_V1:
1729 fmtname = "vfsv1";
1730 break;
1731 }
1732 seq_printf(seq, ",jqfmt=%s", fmtname);
1733 }
1734
1735 if (sbi->s_qf_names[USRQUOTA])
1736 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1737
1738 if (sbi->s_qf_names[GRPQUOTA])
1739 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1740 #endif
1741 }
1742
1743 static const char *token2str(int token)
1744 {
1745 const struct match_token *t;
1746
1747 for (t = tokens; t->token != Opt_err; t++)
1748 if (t->token == token && !strchr(t->pattern, '='))
1749 break;
1750 return t->pattern;
1751 }
1752
1753 /*
1754 * Show an option if
1755 * - it's set to a non-default value OR
1756 * - if the per-sb default is different from the global default
1757 */
1758 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1759 int nodefs)
1760 {
1761 struct ext4_sb_info *sbi = EXT4_SB(sb);
1762 struct ext4_super_block *es = sbi->s_es;
1763 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1764 const struct mount_opts *m;
1765 char sep = nodefs ? '\n' : ',';
1766
1767 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1768 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1769
1770 if (sbi->s_sb_block != 1)
1771 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1772
1773 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1774 int want_set = m->flags & MOPT_SET;
1775 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1776 (m->flags & MOPT_CLEAR_ERR))
1777 continue;
1778 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1779 continue; /* skip if same as the default */
1780 if ((want_set &&
1781 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1782 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1783 continue; /* select Opt_noFoo vs Opt_Foo */
1784 SEQ_OPTS_PRINT("%s", token2str(m->token));
1785 }
1786
1787 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1788 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1789 SEQ_OPTS_PRINT("resuid=%u",
1790 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1791 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1792 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1793 SEQ_OPTS_PRINT("resgid=%u",
1794 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1795 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1796 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1797 SEQ_OPTS_PUTS("errors=remount-ro");
1798 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1799 SEQ_OPTS_PUTS("errors=continue");
1800 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1801 SEQ_OPTS_PUTS("errors=panic");
1802 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1803 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1804 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1805 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1806 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1807 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1808 if (sb->s_flags & MS_I_VERSION)
1809 SEQ_OPTS_PUTS("i_version");
1810 if (nodefs || sbi->s_stripe)
1811 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1812 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1813 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1814 SEQ_OPTS_PUTS("data=journal");
1815 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1816 SEQ_OPTS_PUTS("data=ordered");
1817 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1818 SEQ_OPTS_PUTS("data=writeback");
1819 }
1820 if (nodefs ||
1821 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1822 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1823 sbi->s_inode_readahead_blks);
1824
1825 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1826 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1827 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1828 if (nodefs || sbi->s_max_dir_size_kb)
1829 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1830
1831 ext4_show_quota_options(seq, sb);
1832 return 0;
1833 }
1834
1835 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1836 {
1837 return _ext4_show_options(seq, root->d_sb, 0);
1838 }
1839
1840 static int options_seq_show(struct seq_file *seq, void *offset)
1841 {
1842 struct super_block *sb = seq->private;
1843 int rc;
1844
1845 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1846 rc = _ext4_show_options(seq, sb, 1);
1847 seq_puts(seq, "\n");
1848 return rc;
1849 }
1850
1851 static int options_open_fs(struct inode *inode, struct file *file)
1852 {
1853 return single_open(file, options_seq_show, PDE_DATA(inode));
1854 }
1855
1856 static const struct file_operations ext4_seq_options_fops = {
1857 .owner = THIS_MODULE,
1858 .open = options_open_fs,
1859 .read = seq_read,
1860 .llseek = seq_lseek,
1861 .release = single_release,
1862 };
1863
1864 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1865 int read_only)
1866 {
1867 struct ext4_sb_info *sbi = EXT4_SB(sb);
1868 int res = 0;
1869
1870 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1871 ext4_msg(sb, KERN_ERR, "revision level too high, "
1872 "forcing read-only mode");
1873 res = MS_RDONLY;
1874 }
1875 if (read_only)
1876 goto done;
1877 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1878 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1879 "running e2fsck is recommended");
1880 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1881 ext4_msg(sb, KERN_WARNING,
1882 "warning: mounting fs with errors, "
1883 "running e2fsck is recommended");
1884 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1885 le16_to_cpu(es->s_mnt_count) >=
1886 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1887 ext4_msg(sb, KERN_WARNING,
1888 "warning: maximal mount count reached, "
1889 "running e2fsck is recommended");
1890 else if (le32_to_cpu(es->s_checkinterval) &&
1891 (le32_to_cpu(es->s_lastcheck) +
1892 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1893 ext4_msg(sb, KERN_WARNING,
1894 "warning: checktime reached, "
1895 "running e2fsck is recommended");
1896 if (!sbi->s_journal)
1897 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1898 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1899 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1900 le16_add_cpu(&es->s_mnt_count, 1);
1901 es->s_mtime = cpu_to_le32(get_seconds());
1902 ext4_update_dynamic_rev(sb);
1903 if (sbi->s_journal)
1904 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1905
1906 ext4_commit_super(sb, 1);
1907 done:
1908 if (test_opt(sb, DEBUG))
1909 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1910 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1911 sb->s_blocksize,
1912 sbi->s_groups_count,
1913 EXT4_BLOCKS_PER_GROUP(sb),
1914 EXT4_INODES_PER_GROUP(sb),
1915 sbi->s_mount_opt, sbi->s_mount_opt2);
1916
1917 cleancache_init_fs(sb);
1918 return res;
1919 }
1920
1921 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1922 {
1923 struct ext4_sb_info *sbi = EXT4_SB(sb);
1924 struct flex_groups *new_groups;
1925 int size;
1926
1927 if (!sbi->s_log_groups_per_flex)
1928 return 0;
1929
1930 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1931 if (size <= sbi->s_flex_groups_allocated)
1932 return 0;
1933
1934 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1935 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1936 if (!new_groups) {
1937 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1938 size / (int) sizeof(struct flex_groups));
1939 return -ENOMEM;
1940 }
1941
1942 if (sbi->s_flex_groups) {
1943 memcpy(new_groups, sbi->s_flex_groups,
1944 (sbi->s_flex_groups_allocated *
1945 sizeof(struct flex_groups)));
1946 ext4_kvfree(sbi->s_flex_groups);
1947 }
1948 sbi->s_flex_groups = new_groups;
1949 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1950 return 0;
1951 }
1952
1953 static int ext4_fill_flex_info(struct super_block *sb)
1954 {
1955 struct ext4_sb_info *sbi = EXT4_SB(sb);
1956 struct ext4_group_desc *gdp = NULL;
1957 ext4_group_t flex_group;
1958 int i, err;
1959
1960 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1961 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1962 sbi->s_log_groups_per_flex = 0;
1963 return 1;
1964 }
1965
1966 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1967 if (err)
1968 goto failed;
1969
1970 for (i = 0; i < sbi->s_groups_count; i++) {
1971 gdp = ext4_get_group_desc(sb, i, NULL);
1972
1973 flex_group = ext4_flex_group(sbi, i);
1974 atomic_add(ext4_free_inodes_count(sb, gdp),
1975 &sbi->s_flex_groups[flex_group].free_inodes);
1976 atomic64_add(ext4_free_group_clusters(sb, gdp),
1977 &sbi->s_flex_groups[flex_group].free_clusters);
1978 atomic_add(ext4_used_dirs_count(sb, gdp),
1979 &sbi->s_flex_groups[flex_group].used_dirs);
1980 }
1981
1982 return 1;
1983 failed:
1984 return 0;
1985 }
1986
1987 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1988 struct ext4_group_desc *gdp)
1989 {
1990 int offset;
1991 __u16 crc = 0;
1992 __le32 le_group = cpu_to_le32(block_group);
1993
1994 if ((sbi->s_es->s_feature_ro_compat &
1995 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1996 /* Use new metadata_csum algorithm */
1997 __le16 save_csum;
1998 __u32 csum32;
1999
2000 save_csum = gdp->bg_checksum;
2001 gdp->bg_checksum = 0;
2002 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2003 sizeof(le_group));
2004 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2005 sbi->s_desc_size);
2006 gdp->bg_checksum = save_csum;
2007
2008 crc = csum32 & 0xFFFF;
2009 goto out;
2010 }
2011
2012 /* old crc16 code */
2013 offset = offsetof(struct ext4_group_desc, bg_checksum);
2014
2015 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2016 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2017 crc = crc16(crc, (__u8 *)gdp, offset);
2018 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2019 /* for checksum of struct ext4_group_desc do the rest...*/
2020 if ((sbi->s_es->s_feature_incompat &
2021 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2022 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2023 crc = crc16(crc, (__u8 *)gdp + offset,
2024 le16_to_cpu(sbi->s_es->s_desc_size) -
2025 offset);
2026
2027 out:
2028 return cpu_to_le16(crc);
2029 }
2030
2031 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2032 struct ext4_group_desc *gdp)
2033 {
2034 if (ext4_has_group_desc_csum(sb) &&
2035 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2036 block_group, gdp)))
2037 return 0;
2038
2039 return 1;
2040 }
2041
2042 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2043 struct ext4_group_desc *gdp)
2044 {
2045 if (!ext4_has_group_desc_csum(sb))
2046 return;
2047 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2048 }
2049
2050 /* Called at mount-time, super-block is locked */
2051 static int ext4_check_descriptors(struct super_block *sb,
2052 ext4_group_t *first_not_zeroed)
2053 {
2054 struct ext4_sb_info *sbi = EXT4_SB(sb);
2055 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2056 ext4_fsblk_t last_block;
2057 ext4_fsblk_t block_bitmap;
2058 ext4_fsblk_t inode_bitmap;
2059 ext4_fsblk_t inode_table;
2060 int flexbg_flag = 0;
2061 ext4_group_t i, grp = sbi->s_groups_count;
2062
2063 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2064 flexbg_flag = 1;
2065
2066 ext4_debug("Checking group descriptors");
2067
2068 for (i = 0; i < sbi->s_groups_count; i++) {
2069 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2070
2071 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2072 last_block = ext4_blocks_count(sbi->s_es) - 1;
2073 else
2074 last_block = first_block +
2075 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2076
2077 if ((grp == sbi->s_groups_count) &&
2078 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2079 grp = i;
2080
2081 block_bitmap = ext4_block_bitmap(sb, gdp);
2082 if (block_bitmap < first_block || block_bitmap > last_block) {
2083 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2084 "Block bitmap for group %u not in group "
2085 "(block %llu)!", i, block_bitmap);
2086 return 0;
2087 }
2088 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2089 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2090 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2091 "Inode bitmap for group %u not in group "
2092 "(block %llu)!", i, inode_bitmap);
2093 return 0;
2094 }
2095 inode_table = ext4_inode_table(sb, gdp);
2096 if (inode_table < first_block ||
2097 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2098 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2099 "Inode table for group %u not in group "
2100 "(block %llu)!", i, inode_table);
2101 return 0;
2102 }
2103 ext4_lock_group(sb, i);
2104 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2105 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2106 "Checksum for group %u failed (%u!=%u)",
2107 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2108 gdp)), le16_to_cpu(gdp->bg_checksum));
2109 if (!(sb->s_flags & MS_RDONLY)) {
2110 ext4_unlock_group(sb, i);
2111 return 0;
2112 }
2113 }
2114 ext4_unlock_group(sb, i);
2115 if (!flexbg_flag)
2116 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2117 }
2118 if (NULL != first_not_zeroed)
2119 *first_not_zeroed = grp;
2120
2121 ext4_free_blocks_count_set(sbi->s_es,
2122 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2123 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2124 return 1;
2125 }
2126
2127 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2128 * the superblock) which were deleted from all directories, but held open by
2129 * a process at the time of a crash. We walk the list and try to delete these
2130 * inodes at recovery time (only with a read-write filesystem).
2131 *
2132 * In order to keep the orphan inode chain consistent during traversal (in
2133 * case of crash during recovery), we link each inode into the superblock
2134 * orphan list_head and handle it the same way as an inode deletion during
2135 * normal operation (which journals the operations for us).
2136 *
2137 * We only do an iget() and an iput() on each inode, which is very safe if we
2138 * accidentally point at an in-use or already deleted inode. The worst that
2139 * can happen in this case is that we get a "bit already cleared" message from
2140 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2141 * e2fsck was run on this filesystem, and it must have already done the orphan
2142 * inode cleanup for us, so we can safely abort without any further action.
2143 */
2144 static void ext4_orphan_cleanup(struct super_block *sb,
2145 struct ext4_super_block *es)
2146 {
2147 unsigned int s_flags = sb->s_flags;
2148 int nr_orphans = 0, nr_truncates = 0;
2149 #ifdef CONFIG_QUOTA
2150 int i;
2151 #endif
2152 if (!es->s_last_orphan) {
2153 jbd_debug(4, "no orphan inodes to clean up\n");
2154 return;
2155 }
2156
2157 if (bdev_read_only(sb->s_bdev)) {
2158 ext4_msg(sb, KERN_ERR, "write access "
2159 "unavailable, skipping orphan cleanup");
2160 return;
2161 }
2162
2163 /* Check if feature set would not allow a r/w mount */
2164 if (!ext4_feature_set_ok(sb, 0)) {
2165 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2166 "unknown ROCOMPAT features");
2167 return;
2168 }
2169
2170 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2171 /* don't clear list on RO mount w/ errors */
2172 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2173 jbd_debug(1, "Errors on filesystem, "
2174 "clearing orphan list.\n");
2175 es->s_last_orphan = 0;
2176 }
2177 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2178 return;
2179 }
2180
2181 if (s_flags & MS_RDONLY) {
2182 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2183 sb->s_flags &= ~MS_RDONLY;
2184 }
2185 #ifdef CONFIG_QUOTA
2186 /* Needed for iput() to work correctly and not trash data */
2187 sb->s_flags |= MS_ACTIVE;
2188 /* Turn on quotas so that they are updated correctly */
2189 for (i = 0; i < MAXQUOTAS; i++) {
2190 if (EXT4_SB(sb)->s_qf_names[i]) {
2191 int ret = ext4_quota_on_mount(sb, i);
2192 if (ret < 0)
2193 ext4_msg(sb, KERN_ERR,
2194 "Cannot turn on journaled "
2195 "quota: error %d", ret);
2196 }
2197 }
2198 #endif
2199
2200 while (es->s_last_orphan) {
2201 struct inode *inode;
2202
2203 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2204 if (IS_ERR(inode)) {
2205 es->s_last_orphan = 0;
2206 break;
2207 }
2208
2209 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2210 dquot_initialize(inode);
2211 if (inode->i_nlink) {
2212 if (test_opt(sb, DEBUG))
2213 ext4_msg(sb, KERN_DEBUG,
2214 "%s: truncating inode %lu to %lld bytes",
2215 __func__, inode->i_ino, inode->i_size);
2216 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2217 inode->i_ino, inode->i_size);
2218 mutex_lock(&inode->i_mutex);
2219 truncate_inode_pages(inode->i_mapping, inode->i_size);
2220 ext4_truncate(inode);
2221 mutex_unlock(&inode->i_mutex);
2222 nr_truncates++;
2223 } else {
2224 if (test_opt(sb, DEBUG))
2225 ext4_msg(sb, KERN_DEBUG,
2226 "%s: deleting unreferenced inode %lu",
2227 __func__, inode->i_ino);
2228 jbd_debug(2, "deleting unreferenced inode %lu\n",
2229 inode->i_ino);
2230 nr_orphans++;
2231 }
2232 iput(inode); /* The delete magic happens here! */
2233 }
2234
2235 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2236
2237 if (nr_orphans)
2238 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2239 PLURAL(nr_orphans));
2240 if (nr_truncates)
2241 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2242 PLURAL(nr_truncates));
2243 #ifdef CONFIG_QUOTA
2244 /* Turn quotas off */
2245 for (i = 0; i < MAXQUOTAS; i++) {
2246 if (sb_dqopt(sb)->files[i])
2247 dquot_quota_off(sb, i);
2248 }
2249 #endif
2250 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2251 }
2252
2253 /*
2254 * Maximal extent format file size.
2255 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2256 * extent format containers, within a sector_t, and within i_blocks
2257 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2258 * so that won't be a limiting factor.
2259 *
2260 * However there is other limiting factor. We do store extents in the form
2261 * of starting block and length, hence the resulting length of the extent
2262 * covering maximum file size must fit into on-disk format containers as
2263 * well. Given that length is always by 1 unit bigger than max unit (because
2264 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2265 *
2266 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2267 */
2268 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2269 {
2270 loff_t res;
2271 loff_t upper_limit = MAX_LFS_FILESIZE;
2272
2273 /* small i_blocks in vfs inode? */
2274 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2275 /*
2276 * CONFIG_LBDAF is not enabled implies the inode
2277 * i_block represent total blocks in 512 bytes
2278 * 32 == size of vfs inode i_blocks * 8
2279 */
2280 upper_limit = (1LL << 32) - 1;
2281
2282 /* total blocks in file system block size */
2283 upper_limit >>= (blkbits - 9);
2284 upper_limit <<= blkbits;
2285 }
2286
2287 /*
2288 * 32-bit extent-start container, ee_block. We lower the maxbytes
2289 * by one fs block, so ee_len can cover the extent of maximum file
2290 * size
2291 */
2292 res = (1LL << 32) - 1;
2293 res <<= blkbits;
2294
2295 /* Sanity check against vm- & vfs- imposed limits */
2296 if (res > upper_limit)
2297 res = upper_limit;
2298
2299 return res;
2300 }
2301
2302 /*
2303 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2304 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2305 * We need to be 1 filesystem block less than the 2^48 sector limit.
2306 */
2307 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2308 {
2309 loff_t res = EXT4_NDIR_BLOCKS;
2310 int meta_blocks;
2311 loff_t upper_limit;
2312 /* This is calculated to be the largest file size for a dense, block
2313 * mapped file such that the file's total number of 512-byte sectors,
2314 * including data and all indirect blocks, does not exceed (2^48 - 1).
2315 *
2316 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2317 * number of 512-byte sectors of the file.
2318 */
2319
2320 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2321 /*
2322 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2323 * the inode i_block field represents total file blocks in
2324 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2325 */
2326 upper_limit = (1LL << 32) - 1;
2327
2328 /* total blocks in file system block size */
2329 upper_limit >>= (bits - 9);
2330
2331 } else {
2332 /*
2333 * We use 48 bit ext4_inode i_blocks
2334 * With EXT4_HUGE_FILE_FL set the i_blocks
2335 * represent total number of blocks in
2336 * file system block size
2337 */
2338 upper_limit = (1LL << 48) - 1;
2339
2340 }
2341
2342 /* indirect blocks */
2343 meta_blocks = 1;
2344 /* double indirect blocks */
2345 meta_blocks += 1 + (1LL << (bits-2));
2346 /* tripple indirect blocks */
2347 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2348
2349 upper_limit -= meta_blocks;
2350 upper_limit <<= bits;
2351
2352 res += 1LL << (bits-2);
2353 res += 1LL << (2*(bits-2));
2354 res += 1LL << (3*(bits-2));
2355 res <<= bits;
2356 if (res > upper_limit)
2357 res = upper_limit;
2358
2359 if (res > MAX_LFS_FILESIZE)
2360 res = MAX_LFS_FILESIZE;
2361
2362 return res;
2363 }
2364
2365 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2366 ext4_fsblk_t logical_sb_block, int nr)
2367 {
2368 struct ext4_sb_info *sbi = EXT4_SB(sb);
2369 ext4_group_t bg, first_meta_bg;
2370 int has_super = 0;
2371
2372 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2373
2374 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2375 nr < first_meta_bg)
2376 return logical_sb_block + nr + 1;
2377 bg = sbi->s_desc_per_block * nr;
2378 if (ext4_bg_has_super(sb, bg))
2379 has_super = 1;
2380
2381 return (has_super + ext4_group_first_block_no(sb, bg));
2382 }
2383
2384 /**
2385 * ext4_get_stripe_size: Get the stripe size.
2386 * @sbi: In memory super block info
2387 *
2388 * If we have specified it via mount option, then
2389 * use the mount option value. If the value specified at mount time is
2390 * greater than the blocks per group use the super block value.
2391 * If the super block value is greater than blocks per group return 0.
2392 * Allocator needs it be less than blocks per group.
2393 *
2394 */
2395 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2396 {
2397 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2398 unsigned long stripe_width =
2399 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2400 int ret;
2401
2402 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2403 ret = sbi->s_stripe;
2404 else if (stripe_width <= sbi->s_blocks_per_group)
2405 ret = stripe_width;
2406 else if (stride <= sbi->s_blocks_per_group)
2407 ret = stride;
2408 else
2409 ret = 0;
2410
2411 /*
2412 * If the stripe width is 1, this makes no sense and
2413 * we set it to 0 to turn off stripe handling code.
2414 */
2415 if (ret <= 1)
2416 ret = 0;
2417
2418 return ret;
2419 }
2420
2421 /* sysfs supprt */
2422
2423 struct ext4_attr {
2424 struct attribute attr;
2425 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2426 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2427 const char *, size_t);
2428 union {
2429 int offset;
2430 int deprecated_val;
2431 } u;
2432 };
2433
2434 static int parse_strtoull(const char *buf,
2435 unsigned long long max, unsigned long long *value)
2436 {
2437 int ret;
2438
2439 ret = kstrtoull(skip_spaces(buf), 0, value);
2440 if (!ret && *value > max)
2441 ret = -EINVAL;
2442 return ret;
2443 }
2444
2445 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2446 struct ext4_sb_info *sbi,
2447 char *buf)
2448 {
2449 return snprintf(buf, PAGE_SIZE, "%llu\n",
2450 (s64) EXT4_C2B(sbi,
2451 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2452 }
2453
2454 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2455 struct ext4_sb_info *sbi, char *buf)
2456 {
2457 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2458
2459 if (!sb->s_bdev->bd_part)
2460 return snprintf(buf, PAGE_SIZE, "0\n");
2461 return snprintf(buf, PAGE_SIZE, "%lu\n",
2462 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2463 sbi->s_sectors_written_start) >> 1);
2464 }
2465
2466 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2467 struct ext4_sb_info *sbi, char *buf)
2468 {
2469 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2470
2471 if (!sb->s_bdev->bd_part)
2472 return snprintf(buf, PAGE_SIZE, "0\n");
2473 return snprintf(buf, PAGE_SIZE, "%llu\n",
2474 (unsigned long long)(sbi->s_kbytes_written +
2475 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2476 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2477 }
2478
2479 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2480 struct ext4_sb_info *sbi,
2481 const char *buf, size_t count)
2482 {
2483 unsigned long t;
2484 int ret;
2485
2486 ret = kstrtoul(skip_spaces(buf), 0, &t);
2487 if (ret)
2488 return ret;
2489
2490 if (t && (!is_power_of_2(t) || t > 0x40000000))
2491 return -EINVAL;
2492
2493 sbi->s_inode_readahead_blks = t;
2494 return count;
2495 }
2496
2497 static ssize_t sbi_ui_show(struct ext4_attr *a,
2498 struct ext4_sb_info *sbi, char *buf)
2499 {
2500 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2501
2502 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2503 }
2504
2505 static ssize_t sbi_ui_store(struct ext4_attr *a,
2506 struct ext4_sb_info *sbi,
2507 const char *buf, size_t count)
2508 {
2509 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2510 unsigned long t;
2511 int ret;
2512
2513 ret = kstrtoul(skip_spaces(buf), 0, &t);
2514 if (ret)
2515 return ret;
2516 *ui = t;
2517 return count;
2518 }
2519
2520 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2521 struct ext4_sb_info *sbi, char *buf)
2522 {
2523 return snprintf(buf, PAGE_SIZE, "%llu\n",
2524 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2525 }
2526
2527 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2528 struct ext4_sb_info *sbi,
2529 const char *buf, size_t count)
2530 {
2531 unsigned long long val;
2532 int ret;
2533
2534 if (parse_strtoull(buf, -1ULL, &val))
2535 return -EINVAL;
2536 ret = ext4_reserve_clusters(sbi, val);
2537
2538 return ret ? ret : count;
2539 }
2540
2541 static ssize_t trigger_test_error(struct ext4_attr *a,
2542 struct ext4_sb_info *sbi,
2543 const char *buf, size_t count)
2544 {
2545 int len = count;
2546
2547 if (!capable(CAP_SYS_ADMIN))
2548 return -EPERM;
2549
2550 if (len && buf[len-1] == '\n')
2551 len--;
2552
2553 if (len)
2554 ext4_error(sbi->s_sb, "%.*s", len, buf);
2555 return count;
2556 }
2557
2558 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2559 struct ext4_sb_info *sbi, char *buf)
2560 {
2561 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2562 }
2563
2564 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2565 static struct ext4_attr ext4_attr_##_name = { \
2566 .attr = {.name = __stringify(_name), .mode = _mode }, \
2567 .show = _show, \
2568 .store = _store, \
2569 .u = { \
2570 .offset = offsetof(struct ext4_sb_info, _elname),\
2571 }, \
2572 }
2573 #define EXT4_ATTR(name, mode, show, store) \
2574 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2575
2576 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2577 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2578 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2579 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2580 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2581 #define ATTR_LIST(name) &ext4_attr_##name.attr
2582 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2583 static struct ext4_attr ext4_attr_##_name = { \
2584 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2585 .show = sbi_deprecated_show, \
2586 .u = { \
2587 .deprecated_val = _val, \
2588 }, \
2589 }
2590
2591 EXT4_RO_ATTR(delayed_allocation_blocks);
2592 EXT4_RO_ATTR(session_write_kbytes);
2593 EXT4_RO_ATTR(lifetime_write_kbytes);
2594 EXT4_RW_ATTR(reserved_clusters);
2595 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2596 inode_readahead_blks_store, s_inode_readahead_blks);
2597 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2598 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2599 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2600 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2601 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2602 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2603 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2604 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2605 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2606 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2607
2608 static struct attribute *ext4_attrs[] = {
2609 ATTR_LIST(delayed_allocation_blocks),
2610 ATTR_LIST(session_write_kbytes),
2611 ATTR_LIST(lifetime_write_kbytes),
2612 ATTR_LIST(reserved_clusters),
2613 ATTR_LIST(inode_readahead_blks),
2614 ATTR_LIST(inode_goal),
2615 ATTR_LIST(mb_stats),
2616 ATTR_LIST(mb_max_to_scan),
2617 ATTR_LIST(mb_min_to_scan),
2618 ATTR_LIST(mb_order2_req),
2619 ATTR_LIST(mb_stream_req),
2620 ATTR_LIST(mb_group_prealloc),
2621 ATTR_LIST(max_writeback_mb_bump),
2622 ATTR_LIST(extent_max_zeroout_kb),
2623 ATTR_LIST(trigger_fs_error),
2624 NULL,
2625 };
2626
2627 /* Features this copy of ext4 supports */
2628 EXT4_INFO_ATTR(lazy_itable_init);
2629 EXT4_INFO_ATTR(batched_discard);
2630 EXT4_INFO_ATTR(meta_bg_resize);
2631
2632 static struct attribute *ext4_feat_attrs[] = {
2633 ATTR_LIST(lazy_itable_init),
2634 ATTR_LIST(batched_discard),
2635 ATTR_LIST(meta_bg_resize),
2636 NULL,
2637 };
2638
2639 static ssize_t ext4_attr_show(struct kobject *kobj,
2640 struct attribute *attr, char *buf)
2641 {
2642 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2643 s_kobj);
2644 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2645
2646 return a->show ? a->show(a, sbi, buf) : 0;
2647 }
2648
2649 static ssize_t ext4_attr_store(struct kobject *kobj,
2650 struct attribute *attr,
2651 const char *buf, size_t len)
2652 {
2653 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2654 s_kobj);
2655 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2656
2657 return a->store ? a->store(a, sbi, buf, len) : 0;
2658 }
2659
2660 static void ext4_sb_release(struct kobject *kobj)
2661 {
2662 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2663 s_kobj);
2664 complete(&sbi->s_kobj_unregister);
2665 }
2666
2667 static const struct sysfs_ops ext4_attr_ops = {
2668 .show = ext4_attr_show,
2669 .store = ext4_attr_store,
2670 };
2671
2672 static struct kobj_type ext4_ktype = {
2673 .default_attrs = ext4_attrs,
2674 .sysfs_ops = &ext4_attr_ops,
2675 .release = ext4_sb_release,
2676 };
2677
2678 static void ext4_feat_release(struct kobject *kobj)
2679 {
2680 complete(&ext4_feat->f_kobj_unregister);
2681 }
2682
2683 static struct kobj_type ext4_feat_ktype = {
2684 .default_attrs = ext4_feat_attrs,
2685 .sysfs_ops = &ext4_attr_ops,
2686 .release = ext4_feat_release,
2687 };
2688
2689 /*
2690 * Check whether this filesystem can be mounted based on
2691 * the features present and the RDONLY/RDWR mount requested.
2692 * Returns 1 if this filesystem can be mounted as requested,
2693 * 0 if it cannot be.
2694 */
2695 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2696 {
2697 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2698 ext4_msg(sb, KERN_ERR,
2699 "Couldn't mount because of "
2700 "unsupported optional features (%x)",
2701 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2702 ~EXT4_FEATURE_INCOMPAT_SUPP));
2703 return 0;
2704 }
2705
2706 if (readonly)
2707 return 1;
2708
2709 /* Check that feature set is OK for a read-write mount */
2710 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2711 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2712 "unsupported optional features (%x)",
2713 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2714 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2715 return 0;
2716 }
2717 /*
2718 * Large file size enabled file system can only be mounted
2719 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2720 */
2721 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2722 if (sizeof(blkcnt_t) < sizeof(u64)) {
2723 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2724 "cannot be mounted RDWR without "
2725 "CONFIG_LBDAF");
2726 return 0;
2727 }
2728 }
2729 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2730 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2731 ext4_msg(sb, KERN_ERR,
2732 "Can't support bigalloc feature without "
2733 "extents feature\n");
2734 return 0;
2735 }
2736
2737 #ifndef CONFIG_QUOTA
2738 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2739 !readonly) {
2740 ext4_msg(sb, KERN_ERR,
2741 "Filesystem with quota feature cannot be mounted RDWR "
2742 "without CONFIG_QUOTA");
2743 return 0;
2744 }
2745 #endif /* CONFIG_QUOTA */
2746 return 1;
2747 }
2748
2749 /*
2750 * This function is called once a day if we have errors logged
2751 * on the file system
2752 */
2753 static void print_daily_error_info(unsigned long arg)
2754 {
2755 struct super_block *sb = (struct super_block *) arg;
2756 struct ext4_sb_info *sbi;
2757 struct ext4_super_block *es;
2758
2759 sbi = EXT4_SB(sb);
2760 es = sbi->s_es;
2761
2762 if (es->s_error_count)
2763 /* fsck newer than v1.41.13 is needed to clean this condition. */
2764 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2765 le32_to_cpu(es->s_error_count));
2766 if (es->s_first_error_time) {
2767 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2768 sb->s_id, le32_to_cpu(es->s_first_error_time),
2769 (int) sizeof(es->s_first_error_func),
2770 es->s_first_error_func,
2771 le32_to_cpu(es->s_first_error_line));
2772 if (es->s_first_error_ino)
2773 printk(": inode %u",
2774 le32_to_cpu(es->s_first_error_ino));
2775 if (es->s_first_error_block)
2776 printk(": block %llu", (unsigned long long)
2777 le64_to_cpu(es->s_first_error_block));
2778 printk("\n");
2779 }
2780 if (es->s_last_error_time) {
2781 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2782 sb->s_id, le32_to_cpu(es->s_last_error_time),
2783 (int) sizeof(es->s_last_error_func),
2784 es->s_last_error_func,
2785 le32_to_cpu(es->s_last_error_line));
2786 if (es->s_last_error_ino)
2787 printk(": inode %u",
2788 le32_to_cpu(es->s_last_error_ino));
2789 if (es->s_last_error_block)
2790 printk(": block %llu", (unsigned long long)
2791 le64_to_cpu(es->s_last_error_block));
2792 printk("\n");
2793 }
2794 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2795 }
2796
2797 /* Find next suitable group and run ext4_init_inode_table */
2798 static int ext4_run_li_request(struct ext4_li_request *elr)
2799 {
2800 struct ext4_group_desc *gdp = NULL;
2801 ext4_group_t group, ngroups;
2802 struct super_block *sb;
2803 unsigned long timeout = 0;
2804 int ret = 0;
2805
2806 sb = elr->lr_super;
2807 ngroups = EXT4_SB(sb)->s_groups_count;
2808
2809 sb_start_write(sb);
2810 for (group = elr->lr_next_group; group < ngroups; group++) {
2811 gdp = ext4_get_group_desc(sb, group, NULL);
2812 if (!gdp) {
2813 ret = 1;
2814 break;
2815 }
2816
2817 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2818 break;
2819 }
2820
2821 if (group >= ngroups)
2822 ret = 1;
2823
2824 if (!ret) {
2825 timeout = jiffies;
2826 ret = ext4_init_inode_table(sb, group,
2827 elr->lr_timeout ? 0 : 1);
2828 if (elr->lr_timeout == 0) {
2829 timeout = (jiffies - timeout) *
2830 elr->lr_sbi->s_li_wait_mult;
2831 elr->lr_timeout = timeout;
2832 }
2833 elr->lr_next_sched = jiffies + elr->lr_timeout;
2834 elr->lr_next_group = group + 1;
2835 }
2836 sb_end_write(sb);
2837
2838 return ret;
2839 }
2840
2841 /*
2842 * Remove lr_request from the list_request and free the
2843 * request structure. Should be called with li_list_mtx held
2844 */
2845 static void ext4_remove_li_request(struct ext4_li_request *elr)
2846 {
2847 struct ext4_sb_info *sbi;
2848
2849 if (!elr)
2850 return;
2851
2852 sbi = elr->lr_sbi;
2853
2854 list_del(&elr->lr_request);
2855 sbi->s_li_request = NULL;
2856 kfree(elr);
2857 }
2858
2859 static void ext4_unregister_li_request(struct super_block *sb)
2860 {
2861 mutex_lock(&ext4_li_mtx);
2862 if (!ext4_li_info) {
2863 mutex_unlock(&ext4_li_mtx);
2864 return;
2865 }
2866
2867 mutex_lock(&ext4_li_info->li_list_mtx);
2868 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2869 mutex_unlock(&ext4_li_info->li_list_mtx);
2870 mutex_unlock(&ext4_li_mtx);
2871 }
2872
2873 static struct task_struct *ext4_lazyinit_task;
2874
2875 /*
2876 * This is the function where ext4lazyinit thread lives. It walks
2877 * through the request list searching for next scheduled filesystem.
2878 * When such a fs is found, run the lazy initialization request
2879 * (ext4_rn_li_request) and keep track of the time spend in this
2880 * function. Based on that time we compute next schedule time of
2881 * the request. When walking through the list is complete, compute
2882 * next waking time and put itself into sleep.
2883 */
2884 static int ext4_lazyinit_thread(void *arg)
2885 {
2886 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2887 struct list_head *pos, *n;
2888 struct ext4_li_request *elr;
2889 unsigned long next_wakeup, cur;
2890
2891 BUG_ON(NULL == eli);
2892
2893 cont_thread:
2894 while (true) {
2895 next_wakeup = MAX_JIFFY_OFFSET;
2896
2897 mutex_lock(&eli->li_list_mtx);
2898 if (list_empty(&eli->li_request_list)) {
2899 mutex_unlock(&eli->li_list_mtx);
2900 goto exit_thread;
2901 }
2902
2903 list_for_each_safe(pos, n, &eli->li_request_list) {
2904 elr = list_entry(pos, struct ext4_li_request,
2905 lr_request);
2906
2907 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2908 if (ext4_run_li_request(elr) != 0) {
2909 /* error, remove the lazy_init job */
2910 ext4_remove_li_request(elr);
2911 continue;
2912 }
2913 }
2914
2915 if (time_before(elr->lr_next_sched, next_wakeup))
2916 next_wakeup = elr->lr_next_sched;
2917 }
2918 mutex_unlock(&eli->li_list_mtx);
2919
2920 try_to_freeze();
2921
2922 cur = jiffies;
2923 if ((time_after_eq(cur, next_wakeup)) ||
2924 (MAX_JIFFY_OFFSET == next_wakeup)) {
2925 cond_resched();
2926 continue;
2927 }
2928
2929 schedule_timeout_interruptible(next_wakeup - cur);
2930
2931 if (kthread_should_stop()) {
2932 ext4_clear_request_list();
2933 goto exit_thread;
2934 }
2935 }
2936
2937 exit_thread:
2938 /*
2939 * It looks like the request list is empty, but we need
2940 * to check it under the li_list_mtx lock, to prevent any
2941 * additions into it, and of course we should lock ext4_li_mtx
2942 * to atomically free the list and ext4_li_info, because at
2943 * this point another ext4 filesystem could be registering
2944 * new one.
2945 */
2946 mutex_lock(&ext4_li_mtx);
2947 mutex_lock(&eli->li_list_mtx);
2948 if (!list_empty(&eli->li_request_list)) {
2949 mutex_unlock(&eli->li_list_mtx);
2950 mutex_unlock(&ext4_li_mtx);
2951 goto cont_thread;
2952 }
2953 mutex_unlock(&eli->li_list_mtx);
2954 kfree(ext4_li_info);
2955 ext4_li_info = NULL;
2956 mutex_unlock(&ext4_li_mtx);
2957
2958 return 0;
2959 }
2960
2961 static void ext4_clear_request_list(void)
2962 {
2963 struct list_head *pos, *n;
2964 struct ext4_li_request *elr;
2965
2966 mutex_lock(&ext4_li_info->li_list_mtx);
2967 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2968 elr = list_entry(pos, struct ext4_li_request,
2969 lr_request);
2970 ext4_remove_li_request(elr);
2971 }
2972 mutex_unlock(&ext4_li_info->li_list_mtx);
2973 }
2974
2975 static int ext4_run_lazyinit_thread(void)
2976 {
2977 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2978 ext4_li_info, "ext4lazyinit");
2979 if (IS_ERR(ext4_lazyinit_task)) {
2980 int err = PTR_ERR(ext4_lazyinit_task);
2981 ext4_clear_request_list();
2982 kfree(ext4_li_info);
2983 ext4_li_info = NULL;
2984 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2985 "initialization thread\n",
2986 err);
2987 return err;
2988 }
2989 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2990 return 0;
2991 }
2992
2993 /*
2994 * Check whether it make sense to run itable init. thread or not.
2995 * If there is at least one uninitialized inode table, return
2996 * corresponding group number, else the loop goes through all
2997 * groups and return total number of groups.
2998 */
2999 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3000 {
3001 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3002 struct ext4_group_desc *gdp = NULL;
3003
3004 for (group = 0; group < ngroups; group++) {
3005 gdp = ext4_get_group_desc(sb, group, NULL);
3006 if (!gdp)
3007 continue;
3008
3009 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3010 break;
3011 }
3012
3013 return group;
3014 }
3015
3016 static int ext4_li_info_new(void)
3017 {
3018 struct ext4_lazy_init *eli = NULL;
3019
3020 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3021 if (!eli)
3022 return -ENOMEM;
3023
3024 INIT_LIST_HEAD(&eli->li_request_list);
3025 mutex_init(&eli->li_list_mtx);
3026
3027 eli->li_state |= EXT4_LAZYINIT_QUIT;
3028
3029 ext4_li_info = eli;
3030
3031 return 0;
3032 }
3033
3034 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3035 ext4_group_t start)
3036 {
3037 struct ext4_sb_info *sbi = EXT4_SB(sb);
3038 struct ext4_li_request *elr;
3039 unsigned long rnd;
3040
3041 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3042 if (!elr)
3043 return NULL;
3044
3045 elr->lr_super = sb;
3046 elr->lr_sbi = sbi;
3047 elr->lr_next_group = start;
3048
3049 /*
3050 * Randomize first schedule time of the request to
3051 * spread the inode table initialization requests
3052 * better.
3053 */
3054 get_random_bytes(&rnd, sizeof(rnd));
3055 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3056 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3057
3058 return elr;
3059 }
3060
3061 int ext4_register_li_request(struct super_block *sb,
3062 ext4_group_t first_not_zeroed)
3063 {
3064 struct ext4_sb_info *sbi = EXT4_SB(sb);
3065 struct ext4_li_request *elr = NULL;
3066 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3067 int ret = 0;
3068
3069 mutex_lock(&ext4_li_mtx);
3070 if (sbi->s_li_request != NULL) {
3071 /*
3072 * Reset timeout so it can be computed again, because
3073 * s_li_wait_mult might have changed.
3074 */
3075 sbi->s_li_request->lr_timeout = 0;
3076 goto out;
3077 }
3078
3079 if (first_not_zeroed == ngroups ||
3080 (sb->s_flags & MS_RDONLY) ||
3081 !test_opt(sb, INIT_INODE_TABLE))
3082 goto out;
3083
3084 elr = ext4_li_request_new(sb, first_not_zeroed);
3085 if (!elr) {
3086 ret = -ENOMEM;
3087 goto out;
3088 }
3089
3090 if (NULL == ext4_li_info) {
3091 ret = ext4_li_info_new();
3092 if (ret)
3093 goto out;
3094 }
3095
3096 mutex_lock(&ext4_li_info->li_list_mtx);
3097 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3098 mutex_unlock(&ext4_li_info->li_list_mtx);
3099
3100 sbi->s_li_request = elr;
3101 /*
3102 * set elr to NULL here since it has been inserted to
3103 * the request_list and the removal and free of it is
3104 * handled by ext4_clear_request_list from now on.
3105 */
3106 elr = NULL;
3107
3108 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3109 ret = ext4_run_lazyinit_thread();
3110 if (ret)
3111 goto out;
3112 }
3113 out:
3114 mutex_unlock(&ext4_li_mtx);
3115 if (ret)
3116 kfree(elr);
3117 return ret;
3118 }
3119
3120 /*
3121 * We do not need to lock anything since this is called on
3122 * module unload.
3123 */
3124 static void ext4_destroy_lazyinit_thread(void)
3125 {
3126 /*
3127 * If thread exited earlier
3128 * there's nothing to be done.
3129 */
3130 if (!ext4_li_info || !ext4_lazyinit_task)
3131 return;
3132
3133 kthread_stop(ext4_lazyinit_task);
3134 }
3135
3136 static int set_journal_csum_feature_set(struct super_block *sb)
3137 {
3138 int ret = 1;
3139 int compat, incompat;
3140 struct ext4_sb_info *sbi = EXT4_SB(sb);
3141
3142 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3143 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3144 /* journal checksum v2 */
3145 compat = 0;
3146 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3147 } else {
3148 /* journal checksum v1 */
3149 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3150 incompat = 0;
3151 }
3152
3153 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3154 ret = jbd2_journal_set_features(sbi->s_journal,
3155 compat, 0,
3156 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3157 incompat);
3158 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3159 ret = jbd2_journal_set_features(sbi->s_journal,
3160 compat, 0,
3161 incompat);
3162 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3163 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3164 } else {
3165 jbd2_journal_clear_features(sbi->s_journal,
3166 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3167 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3168 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3169 }
3170
3171 return ret;
3172 }
3173
3174 /*
3175 * Note: calculating the overhead so we can be compatible with
3176 * historical BSD practice is quite difficult in the face of
3177 * clusters/bigalloc. This is because multiple metadata blocks from
3178 * different block group can end up in the same allocation cluster.
3179 * Calculating the exact overhead in the face of clustered allocation
3180 * requires either O(all block bitmaps) in memory or O(number of block
3181 * groups**2) in time. We will still calculate the superblock for
3182 * older file systems --- and if we come across with a bigalloc file
3183 * system with zero in s_overhead_clusters the estimate will be close to
3184 * correct especially for very large cluster sizes --- but for newer
3185 * file systems, it's better to calculate this figure once at mkfs
3186 * time, and store it in the superblock. If the superblock value is
3187 * present (even for non-bigalloc file systems), we will use it.
3188 */
3189 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3190 char *buf)
3191 {
3192 struct ext4_sb_info *sbi = EXT4_SB(sb);
3193 struct ext4_group_desc *gdp;
3194 ext4_fsblk_t first_block, last_block, b;
3195 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3196 int s, j, count = 0;
3197
3198 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3199 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3200 sbi->s_itb_per_group + 2);
3201
3202 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3203 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3204 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3205 for (i = 0; i < ngroups; i++) {
3206 gdp = ext4_get_group_desc(sb, i, NULL);
3207 b = ext4_block_bitmap(sb, gdp);
3208 if (b >= first_block && b <= last_block) {
3209 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3210 count++;
3211 }
3212 b = ext4_inode_bitmap(sb, gdp);
3213 if (b >= first_block && b <= last_block) {
3214 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3215 count++;
3216 }
3217 b = ext4_inode_table(sb, gdp);
3218 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3219 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3220 int c = EXT4_B2C(sbi, b - first_block);
3221 ext4_set_bit(c, buf);
3222 count++;
3223 }
3224 if (i != grp)
3225 continue;
3226 s = 0;
3227 if (ext4_bg_has_super(sb, grp)) {
3228 ext4_set_bit(s++, buf);
3229 count++;
3230 }
3231 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3232 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3233 count++;
3234 }
3235 }
3236 if (!count)
3237 return 0;
3238 return EXT4_CLUSTERS_PER_GROUP(sb) -
3239 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3240 }
3241
3242 /*
3243 * Compute the overhead and stash it in sbi->s_overhead
3244 */
3245 int ext4_calculate_overhead(struct super_block *sb)
3246 {
3247 struct ext4_sb_info *sbi = EXT4_SB(sb);
3248 struct ext4_super_block *es = sbi->s_es;
3249 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3250 ext4_fsblk_t overhead = 0;
3251 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3252
3253 if (!buf)
3254 return -ENOMEM;
3255
3256 /*
3257 * Compute the overhead (FS structures). This is constant
3258 * for a given filesystem unless the number of block groups
3259 * changes so we cache the previous value until it does.
3260 */
3261
3262 /*
3263 * All of the blocks before first_data_block are overhead
3264 */
3265 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3266
3267 /*
3268 * Add the overhead found in each block group
3269 */
3270 for (i = 0; i < ngroups; i++) {
3271 int blks;
3272
3273 blks = count_overhead(sb, i, buf);
3274 overhead += blks;
3275 if (blks)
3276 memset(buf, 0, PAGE_SIZE);
3277 cond_resched();
3278 }
3279 /* Add the journal blocks as well */
3280 if (sbi->s_journal)
3281 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3282
3283 sbi->s_overhead = overhead;
3284 smp_wmb();
3285 free_page((unsigned long) buf);
3286 return 0;
3287 }
3288
3289
3290 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3291 {
3292 ext4_fsblk_t resv_clusters;
3293
3294 /*
3295 * There's no need to reserve anything when we aren't using extents.
3296 * The space estimates are exact, there are no unwritten extents,
3297 * hole punching doesn't need new metadata... This is needed especially
3298 * to keep ext2/3 backward compatibility.
3299 */
3300 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3301 return 0;
3302 /*
3303 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3304 * This should cover the situations where we can not afford to run
3305 * out of space like for example punch hole, or converting
3306 * uninitialized extents in delalloc path. In most cases such
3307 * allocation would require 1, or 2 blocks, higher numbers are
3308 * very rare.
3309 */
3310 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3311 EXT4_SB(sb)->s_cluster_bits;
3312
3313 do_div(resv_clusters, 50);
3314 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3315
3316 return resv_clusters;
3317 }
3318
3319
3320 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3321 {
3322 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3323 sbi->s_cluster_bits;
3324
3325 if (count >= clusters)
3326 return -EINVAL;
3327
3328 atomic64_set(&sbi->s_resv_clusters, count);
3329 return 0;
3330 }
3331
3332 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3333 {
3334 char *orig_data = kstrdup(data, GFP_KERNEL);
3335 struct buffer_head *bh;
3336 struct ext4_super_block *es = NULL;
3337 struct ext4_sb_info *sbi;
3338 ext4_fsblk_t block;
3339 ext4_fsblk_t sb_block = get_sb_block(&data);
3340 ext4_fsblk_t logical_sb_block;
3341 unsigned long offset = 0;
3342 unsigned long journal_devnum = 0;
3343 unsigned long def_mount_opts;
3344 struct inode *root;
3345 char *cp;
3346 const char *descr;
3347 int ret = -ENOMEM;
3348 int blocksize, clustersize;
3349 unsigned int db_count;
3350 unsigned int i;
3351 int needs_recovery, has_huge_files, has_bigalloc;
3352 __u64 blocks_count;
3353 int err = 0;
3354 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3355 ext4_group_t first_not_zeroed;
3356
3357 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3358 if (!sbi)
3359 goto out_free_orig;
3360
3361 sbi->s_blockgroup_lock =
3362 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3363 if (!sbi->s_blockgroup_lock) {
3364 kfree(sbi);
3365 goto out_free_orig;
3366 }
3367 sb->s_fs_info = sbi;
3368 sbi->s_sb = sb;
3369 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3370 sbi->s_sb_block = sb_block;
3371 if (sb->s_bdev->bd_part)
3372 sbi->s_sectors_written_start =
3373 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3374
3375 /* Cleanup superblock name */
3376 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3377 *cp = '!';
3378
3379 /* -EINVAL is default */
3380 ret = -EINVAL;
3381 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3382 if (!blocksize) {
3383 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3384 goto out_fail;
3385 }
3386
3387 /*
3388 * The ext4 superblock will not be buffer aligned for other than 1kB
3389 * block sizes. We need to calculate the offset from buffer start.
3390 */
3391 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3392 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3393 offset = do_div(logical_sb_block, blocksize);
3394 } else {
3395 logical_sb_block = sb_block;
3396 }
3397
3398 if (!(bh = sb_bread(sb, logical_sb_block))) {
3399 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3400 goto out_fail;
3401 }
3402 /*
3403 * Note: s_es must be initialized as soon as possible because
3404 * some ext4 macro-instructions depend on its value
3405 */
3406 es = (struct ext4_super_block *) (bh->b_data + offset);
3407 sbi->s_es = es;
3408 sb->s_magic = le16_to_cpu(es->s_magic);
3409 if (sb->s_magic != EXT4_SUPER_MAGIC)
3410 goto cantfind_ext4;
3411 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3412
3413 /* Warn if metadata_csum and gdt_csum are both set. */
3414 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3415 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3416 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3417 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3418 "redundant flags; please run fsck.");
3419
3420 /* Check for a known checksum algorithm */
3421 if (!ext4_verify_csum_type(sb, es)) {
3422 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3423 "unknown checksum algorithm.");
3424 silent = 1;
3425 goto cantfind_ext4;
3426 }
3427
3428 /* Load the checksum driver */
3429 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3430 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3431 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3432 if (IS_ERR(sbi->s_chksum_driver)) {
3433 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3434 ret = PTR_ERR(sbi->s_chksum_driver);
3435 sbi->s_chksum_driver = NULL;
3436 goto failed_mount;
3437 }
3438 }
3439
3440 /* Check superblock checksum */
3441 if (!ext4_superblock_csum_verify(sb, es)) {
3442 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3443 "invalid superblock checksum. Run e2fsck?");
3444 silent = 1;
3445 goto cantfind_ext4;
3446 }
3447
3448 /* Precompute checksum seed for all metadata */
3449 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3450 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3451 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3452 sizeof(es->s_uuid));
3453
3454 /* Set defaults before we parse the mount options */
3455 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3456 set_opt(sb, INIT_INODE_TABLE);
3457 if (def_mount_opts & EXT4_DEFM_DEBUG)
3458 set_opt(sb, DEBUG);
3459 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3460 set_opt(sb, GRPID);
3461 if (def_mount_opts & EXT4_DEFM_UID16)
3462 set_opt(sb, NO_UID32);
3463 /* xattr user namespace & acls are now defaulted on */
3464 set_opt(sb, XATTR_USER);
3465 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3466 set_opt(sb, POSIX_ACL);
3467 #endif
3468 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3469 set_opt(sb, JOURNAL_DATA);
3470 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3471 set_opt(sb, ORDERED_DATA);
3472 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3473 set_opt(sb, WRITEBACK_DATA);
3474
3475 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3476 set_opt(sb, ERRORS_PANIC);
3477 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3478 set_opt(sb, ERRORS_CONT);
3479 else
3480 set_opt(sb, ERRORS_RO);
3481 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3482 set_opt(sb, BLOCK_VALIDITY);
3483 if (def_mount_opts & EXT4_DEFM_DISCARD)
3484 set_opt(sb, DISCARD);
3485
3486 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3487 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3488 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3489 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3490 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3491
3492 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3493 set_opt(sb, BARRIER);
3494
3495 /*
3496 * enable delayed allocation by default
3497 * Use -o nodelalloc to turn it off
3498 */
3499 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3500 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3501 set_opt(sb, DELALLOC);
3502
3503 /*
3504 * set default s_li_wait_mult for lazyinit, for the case there is
3505 * no mount option specified.
3506 */
3507 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3508
3509 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3510 &journal_devnum, &journal_ioprio, 0)) {
3511 ext4_msg(sb, KERN_WARNING,
3512 "failed to parse options in superblock: %s",
3513 sbi->s_es->s_mount_opts);
3514 }
3515 sbi->s_def_mount_opt = sbi->s_mount_opt;
3516 if (!parse_options((char *) data, sb, &journal_devnum,
3517 &journal_ioprio, 0))
3518 goto failed_mount;
3519
3520 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3521 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3522 "with data=journal disables delayed "
3523 "allocation and O_DIRECT support!\n");
3524 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3525 ext4_msg(sb, KERN_ERR, "can't mount with "
3526 "both data=journal and delalloc");
3527 goto failed_mount;
3528 }
3529 if (test_opt(sb, DIOREAD_NOLOCK)) {
3530 ext4_msg(sb, KERN_ERR, "can't mount with "
3531 "both data=journal and dioread_nolock");
3532 goto failed_mount;
3533 }
3534 if (test_opt(sb, DELALLOC))
3535 clear_opt(sb, DELALLOC);
3536 }
3537
3538 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3539 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3540
3541 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3542 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3543 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3544 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3545 ext4_msg(sb, KERN_WARNING,
3546 "feature flags set on rev 0 fs, "
3547 "running e2fsck is recommended");
3548
3549 if (IS_EXT2_SB(sb)) {
3550 if (ext2_feature_set_ok(sb))
3551 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3552 "using the ext4 subsystem");
3553 else {
3554 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3555 "to feature incompatibilities");
3556 goto failed_mount;
3557 }
3558 }
3559
3560 if (IS_EXT3_SB(sb)) {
3561 if (ext3_feature_set_ok(sb))
3562 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3563 "using the ext4 subsystem");
3564 else {
3565 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3566 "to feature incompatibilities");
3567 goto failed_mount;
3568 }
3569 }
3570
3571 /*
3572 * Check feature flags regardless of the revision level, since we
3573 * previously didn't change the revision level when setting the flags,
3574 * so there is a chance incompat flags are set on a rev 0 filesystem.
3575 */
3576 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3577 goto failed_mount;
3578
3579 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3580 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3581 blocksize > EXT4_MAX_BLOCK_SIZE) {
3582 ext4_msg(sb, KERN_ERR,
3583 "Unsupported filesystem blocksize %d", blocksize);
3584 goto failed_mount;
3585 }
3586
3587 if (sb->s_blocksize != blocksize) {
3588 /* Validate the filesystem blocksize */
3589 if (!sb_set_blocksize(sb, blocksize)) {
3590 ext4_msg(sb, KERN_ERR, "bad block size %d",
3591 blocksize);
3592 goto failed_mount;
3593 }
3594
3595 brelse(bh);
3596 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3597 offset = do_div(logical_sb_block, blocksize);
3598 bh = sb_bread(sb, logical_sb_block);
3599 if (!bh) {
3600 ext4_msg(sb, KERN_ERR,
3601 "Can't read superblock on 2nd try");
3602 goto failed_mount;
3603 }
3604 es = (struct ext4_super_block *)(bh->b_data + offset);
3605 sbi->s_es = es;
3606 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3607 ext4_msg(sb, KERN_ERR,
3608 "Magic mismatch, very weird!");
3609 goto failed_mount;
3610 }
3611 }
3612
3613 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3614 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3615 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3616 has_huge_files);
3617 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3618
3619 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3620 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3621 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3622 } else {
3623 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3624 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3625 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3626 (!is_power_of_2(sbi->s_inode_size)) ||
3627 (sbi->s_inode_size > blocksize)) {
3628 ext4_msg(sb, KERN_ERR,
3629 "unsupported inode size: %d",
3630 sbi->s_inode_size);
3631 goto failed_mount;
3632 }
3633 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3634 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3635 }
3636
3637 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3638 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3639 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3640 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3641 !is_power_of_2(sbi->s_desc_size)) {
3642 ext4_msg(sb, KERN_ERR,
3643 "unsupported descriptor size %lu",
3644 sbi->s_desc_size);
3645 goto failed_mount;
3646 }
3647 } else
3648 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3649
3650 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3651 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3652 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3653 goto cantfind_ext4;
3654
3655 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3656 if (sbi->s_inodes_per_block == 0)
3657 goto cantfind_ext4;
3658 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3659 sbi->s_inodes_per_block;
3660 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3661 sbi->s_sbh = bh;
3662 sbi->s_mount_state = le16_to_cpu(es->s_state);
3663 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3664 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3665
3666 for (i = 0; i < 4; i++)
3667 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3668 sbi->s_def_hash_version = es->s_def_hash_version;
3669 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3670 i = le32_to_cpu(es->s_flags);
3671 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3672 sbi->s_hash_unsigned = 3;
3673 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3674 #ifdef __CHAR_UNSIGNED__
3675 if (!(sb->s_flags & MS_RDONLY))
3676 es->s_flags |=
3677 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3678 sbi->s_hash_unsigned = 3;
3679 #else
3680 if (!(sb->s_flags & MS_RDONLY))
3681 es->s_flags |=
3682 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3683 #endif
3684 }
3685 }
3686
3687 /* Handle clustersize */
3688 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3689 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3690 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3691 if (has_bigalloc) {
3692 if (clustersize < blocksize) {
3693 ext4_msg(sb, KERN_ERR,
3694 "cluster size (%d) smaller than "
3695 "block size (%d)", clustersize, blocksize);
3696 goto failed_mount;
3697 }
3698 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3699 le32_to_cpu(es->s_log_block_size);
3700 sbi->s_clusters_per_group =
3701 le32_to_cpu(es->s_clusters_per_group);
3702 if (sbi->s_clusters_per_group > blocksize * 8) {
3703 ext4_msg(sb, KERN_ERR,
3704 "#clusters per group too big: %lu",
3705 sbi->s_clusters_per_group);
3706 goto failed_mount;
3707 }
3708 if (sbi->s_blocks_per_group !=
3709 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3710 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3711 "clusters per group (%lu) inconsistent",
3712 sbi->s_blocks_per_group,
3713 sbi->s_clusters_per_group);
3714 goto failed_mount;
3715 }
3716 } else {
3717 if (clustersize != blocksize) {
3718 ext4_warning(sb, "fragment/cluster size (%d) != "
3719 "block size (%d)", clustersize,
3720 blocksize);
3721 clustersize = blocksize;
3722 }
3723 if (sbi->s_blocks_per_group > blocksize * 8) {
3724 ext4_msg(sb, KERN_ERR,
3725 "#blocks per group too big: %lu",
3726 sbi->s_blocks_per_group);
3727 goto failed_mount;
3728 }
3729 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3730 sbi->s_cluster_bits = 0;
3731 }
3732 sbi->s_cluster_ratio = clustersize / blocksize;
3733
3734 if (sbi->s_inodes_per_group > blocksize * 8) {
3735 ext4_msg(sb, KERN_ERR,
3736 "#inodes per group too big: %lu",
3737 sbi->s_inodes_per_group);
3738 goto failed_mount;
3739 }
3740
3741 /* Do we have standard group size of clustersize * 8 blocks ? */
3742 if (sbi->s_blocks_per_group == clustersize << 3)
3743 set_opt2(sb, STD_GROUP_SIZE);
3744
3745 /*
3746 * Test whether we have more sectors than will fit in sector_t,
3747 * and whether the max offset is addressable by the page cache.
3748 */
3749 err = generic_check_addressable(sb->s_blocksize_bits,
3750 ext4_blocks_count(es));
3751 if (err) {
3752 ext4_msg(sb, KERN_ERR, "filesystem"
3753 " too large to mount safely on this system");
3754 if (sizeof(sector_t) < 8)
3755 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3756 goto failed_mount;
3757 }
3758
3759 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3760 goto cantfind_ext4;
3761
3762 /* check blocks count against device size */
3763 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3764 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3765 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3766 "exceeds size of device (%llu blocks)",
3767 ext4_blocks_count(es), blocks_count);
3768 goto failed_mount;
3769 }
3770
3771 /*
3772 * It makes no sense for the first data block to be beyond the end
3773 * of the filesystem.
3774 */
3775 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3776 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3777 "block %u is beyond end of filesystem (%llu)",
3778 le32_to_cpu(es->s_first_data_block),
3779 ext4_blocks_count(es));
3780 goto failed_mount;
3781 }
3782 blocks_count = (ext4_blocks_count(es) -
3783 le32_to_cpu(es->s_first_data_block) +
3784 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3785 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3786 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3787 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3788 "(block count %llu, first data block %u, "
3789 "blocks per group %lu)", sbi->s_groups_count,
3790 ext4_blocks_count(es),
3791 le32_to_cpu(es->s_first_data_block),
3792 EXT4_BLOCKS_PER_GROUP(sb));
3793 goto failed_mount;
3794 }
3795 sbi->s_groups_count = blocks_count;
3796 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3797 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3798 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3799 EXT4_DESC_PER_BLOCK(sb);
3800 sbi->s_group_desc = ext4_kvmalloc(db_count *
3801 sizeof(struct buffer_head *),
3802 GFP_KERNEL);
3803 if (sbi->s_group_desc == NULL) {
3804 ext4_msg(sb, KERN_ERR, "not enough memory");
3805 ret = -ENOMEM;
3806 goto failed_mount;
3807 }
3808
3809 if (ext4_proc_root)
3810 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3811
3812 if (sbi->s_proc)
3813 proc_create_data("options", S_IRUGO, sbi->s_proc,
3814 &ext4_seq_options_fops, sb);
3815
3816 bgl_lock_init(sbi->s_blockgroup_lock);
3817
3818 for (i = 0; i < db_count; i++) {
3819 block = descriptor_loc(sb, logical_sb_block, i);
3820 sbi->s_group_desc[i] = sb_bread(sb, block);
3821 if (!sbi->s_group_desc[i]) {
3822 ext4_msg(sb, KERN_ERR,
3823 "can't read group descriptor %d", i);
3824 db_count = i;
3825 goto failed_mount2;
3826 }
3827 }
3828 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3829 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3830 goto failed_mount2;
3831 }
3832 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3833 if (!ext4_fill_flex_info(sb)) {
3834 ext4_msg(sb, KERN_ERR,
3835 "unable to initialize "
3836 "flex_bg meta info!");
3837 goto failed_mount2;
3838 }
3839
3840 sbi->s_gdb_count = db_count;
3841 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3842 spin_lock_init(&sbi->s_next_gen_lock);
3843
3844 init_timer(&sbi->s_err_report);
3845 sbi->s_err_report.function = print_daily_error_info;
3846 sbi->s_err_report.data = (unsigned long) sb;
3847
3848 /* Register extent status tree shrinker */
3849 ext4_es_register_shrinker(sbi);
3850
3851 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3852 ext4_count_free_clusters(sb));
3853 if (!err) {
3854 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3855 ext4_count_free_inodes(sb));
3856 }
3857 if (!err) {
3858 err = percpu_counter_init(&sbi->s_dirs_counter,
3859 ext4_count_dirs(sb));
3860 }
3861 if (!err) {
3862 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3863 }
3864 if (!err) {
3865 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3866 }
3867 if (err) {
3868 ext4_msg(sb, KERN_ERR, "insufficient memory");
3869 goto failed_mount3;
3870 }
3871
3872 sbi->s_stripe = ext4_get_stripe_size(sbi);
3873 sbi->s_extent_max_zeroout_kb = 32;
3874
3875 /*
3876 * set up enough so that it can read an inode
3877 */
3878 if (!test_opt(sb, NOLOAD) &&
3879 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3880 sb->s_op = &ext4_sops;
3881 else
3882 sb->s_op = &ext4_nojournal_sops;
3883 sb->s_export_op = &ext4_export_ops;
3884 sb->s_xattr = ext4_xattr_handlers;
3885 #ifdef CONFIG_QUOTA
3886 sb->dq_op = &ext4_quota_operations;
3887 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3888 sb->s_qcop = &ext4_qctl_sysfile_operations;
3889 else
3890 sb->s_qcop = &ext4_qctl_operations;
3891 #endif
3892 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3893
3894 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3895 mutex_init(&sbi->s_orphan_lock);
3896
3897 sb->s_root = NULL;
3898
3899 needs_recovery = (es->s_last_orphan != 0 ||
3900 EXT4_HAS_INCOMPAT_FEATURE(sb,
3901 EXT4_FEATURE_INCOMPAT_RECOVER));
3902
3903 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3904 !(sb->s_flags & MS_RDONLY))
3905 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3906 goto failed_mount3;
3907
3908 /*
3909 * The first inode we look at is the journal inode. Don't try
3910 * root first: it may be modified in the journal!
3911 */
3912 if (!test_opt(sb, NOLOAD) &&
3913 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3914 if (ext4_load_journal(sb, es, journal_devnum))
3915 goto failed_mount3;
3916 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3917 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3918 ext4_msg(sb, KERN_ERR, "required journal recovery "
3919 "suppressed and not mounted read-only");
3920 goto failed_mount_wq;
3921 } else {
3922 clear_opt(sb, DATA_FLAGS);
3923 sbi->s_journal = NULL;
3924 needs_recovery = 0;
3925 goto no_journal;
3926 }
3927
3928 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3929 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3930 JBD2_FEATURE_INCOMPAT_64BIT)) {
3931 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3932 goto failed_mount_wq;
3933 }
3934
3935 if (!set_journal_csum_feature_set(sb)) {
3936 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3937 "feature set");
3938 goto failed_mount_wq;
3939 }
3940
3941 /* We have now updated the journal if required, so we can
3942 * validate the data journaling mode. */
3943 switch (test_opt(sb, DATA_FLAGS)) {
3944 case 0:
3945 /* No mode set, assume a default based on the journal
3946 * capabilities: ORDERED_DATA if the journal can
3947 * cope, else JOURNAL_DATA
3948 */
3949 if (jbd2_journal_check_available_features
3950 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3951 set_opt(sb, ORDERED_DATA);
3952 else
3953 set_opt(sb, JOURNAL_DATA);
3954 break;
3955
3956 case EXT4_MOUNT_ORDERED_DATA:
3957 case EXT4_MOUNT_WRITEBACK_DATA:
3958 if (!jbd2_journal_check_available_features
3959 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3960 ext4_msg(sb, KERN_ERR, "Journal does not support "
3961 "requested data journaling mode");
3962 goto failed_mount_wq;
3963 }
3964 default:
3965 break;
3966 }
3967 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3968
3969 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3970
3971 /*
3972 * The journal may have updated the bg summary counts, so we
3973 * need to update the global counters.
3974 */
3975 percpu_counter_set(&sbi->s_freeclusters_counter,
3976 ext4_count_free_clusters(sb));
3977 percpu_counter_set(&sbi->s_freeinodes_counter,
3978 ext4_count_free_inodes(sb));
3979 percpu_counter_set(&sbi->s_dirs_counter,
3980 ext4_count_dirs(sb));
3981 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3982
3983 no_journal:
3984 /*
3985 * Get the # of file system overhead blocks from the
3986 * superblock if present.
3987 */
3988 if (es->s_overhead_clusters)
3989 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3990 else {
3991 err = ext4_calculate_overhead(sb);
3992 if (err)
3993 goto failed_mount_wq;
3994 }
3995
3996 /*
3997 * The maximum number of concurrent works can be high and
3998 * concurrency isn't really necessary. Limit it to 1.
3999 */
4000 EXT4_SB(sb)->rsv_conversion_wq =
4001 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4002 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4003 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4004 ret = -ENOMEM;
4005 goto failed_mount4;
4006 }
4007
4008 /*
4009 * The jbd2_journal_load will have done any necessary log recovery,
4010 * so we can safely mount the rest of the filesystem now.
4011 */
4012
4013 root = ext4_iget(sb, EXT4_ROOT_INO);
4014 if (IS_ERR(root)) {
4015 ext4_msg(sb, KERN_ERR, "get root inode failed");
4016 ret = PTR_ERR(root);
4017 root = NULL;
4018 goto failed_mount4;
4019 }
4020 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4021 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4022 iput(root);
4023 goto failed_mount4;
4024 }
4025 sb->s_root = d_make_root(root);
4026 if (!sb->s_root) {
4027 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4028 ret = -ENOMEM;
4029 goto failed_mount4;
4030 }
4031
4032 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4033 sb->s_flags |= MS_RDONLY;
4034
4035 /* determine the minimum size of new large inodes, if present */
4036 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4037 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4038 EXT4_GOOD_OLD_INODE_SIZE;
4039 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4040 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4041 if (sbi->s_want_extra_isize <
4042 le16_to_cpu(es->s_want_extra_isize))
4043 sbi->s_want_extra_isize =
4044 le16_to_cpu(es->s_want_extra_isize);
4045 if (sbi->s_want_extra_isize <
4046 le16_to_cpu(es->s_min_extra_isize))
4047 sbi->s_want_extra_isize =
4048 le16_to_cpu(es->s_min_extra_isize);
4049 }
4050 }
4051 /* Check if enough inode space is available */
4052 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4053 sbi->s_inode_size) {
4054 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4055 EXT4_GOOD_OLD_INODE_SIZE;
4056 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4057 "available");
4058 }
4059
4060 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4061 if (err) {
4062 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4063 "reserved pool", ext4_calculate_resv_clusters(sb));
4064 goto failed_mount4a;
4065 }
4066
4067 err = ext4_setup_system_zone(sb);
4068 if (err) {
4069 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4070 "zone (%d)", err);
4071 goto failed_mount4a;
4072 }
4073
4074 ext4_ext_init(sb);
4075 err = ext4_mb_init(sb);
4076 if (err) {
4077 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4078 err);
4079 goto failed_mount5;
4080 }
4081
4082 err = ext4_register_li_request(sb, first_not_zeroed);
4083 if (err)
4084 goto failed_mount6;
4085
4086 sbi->s_kobj.kset = ext4_kset;
4087 init_completion(&sbi->s_kobj_unregister);
4088 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4089 "%s", sb->s_id);
4090 if (err)
4091 goto failed_mount7;
4092
4093 #ifdef CONFIG_QUOTA
4094 /* Enable quota usage during mount. */
4095 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4096 !(sb->s_flags & MS_RDONLY)) {
4097 err = ext4_enable_quotas(sb);
4098 if (err)
4099 goto failed_mount8;
4100 }
4101 #endif /* CONFIG_QUOTA */
4102
4103 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4104 ext4_orphan_cleanup(sb, es);
4105 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4106 if (needs_recovery) {
4107 ext4_msg(sb, KERN_INFO, "recovery complete");
4108 ext4_mark_recovery_complete(sb, es);
4109 }
4110 if (EXT4_SB(sb)->s_journal) {
4111 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4112 descr = " journalled data mode";
4113 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4114 descr = " ordered data mode";
4115 else
4116 descr = " writeback data mode";
4117 } else
4118 descr = "out journal";
4119
4120 if (test_opt(sb, DISCARD)) {
4121 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4122 if (!blk_queue_discard(q))
4123 ext4_msg(sb, KERN_WARNING,
4124 "mounting with \"discard\" option, but "
4125 "the device does not support discard");
4126 }
4127
4128 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4129 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4130 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4131
4132 if (es->s_error_count)
4133 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4134
4135 kfree(orig_data);
4136 return 0;
4137
4138 cantfind_ext4:
4139 if (!silent)
4140 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4141 goto failed_mount;
4142
4143 #ifdef CONFIG_QUOTA
4144 failed_mount8:
4145 kobject_del(&sbi->s_kobj);
4146 #endif
4147 failed_mount7:
4148 ext4_unregister_li_request(sb);
4149 failed_mount6:
4150 ext4_mb_release(sb);
4151 failed_mount5:
4152 ext4_ext_release(sb);
4153 ext4_release_system_zone(sb);
4154 failed_mount4a:
4155 dput(sb->s_root);
4156 sb->s_root = NULL;
4157 failed_mount4:
4158 ext4_msg(sb, KERN_ERR, "mount failed");
4159 if (EXT4_SB(sb)->rsv_conversion_wq)
4160 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4161 failed_mount_wq:
4162 if (sbi->s_journal) {
4163 jbd2_journal_destroy(sbi->s_journal);
4164 sbi->s_journal = NULL;
4165 }
4166 failed_mount3:
4167 ext4_es_unregister_shrinker(sbi);
4168 del_timer_sync(&sbi->s_err_report);
4169 if (sbi->s_flex_groups)
4170 ext4_kvfree(sbi->s_flex_groups);
4171 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4172 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4173 percpu_counter_destroy(&sbi->s_dirs_counter);
4174 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4175 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4176 if (sbi->s_mmp_tsk)
4177 kthread_stop(sbi->s_mmp_tsk);
4178 failed_mount2:
4179 for (i = 0; i < db_count; i++)
4180 brelse(sbi->s_group_desc[i]);
4181 ext4_kvfree(sbi->s_group_desc);
4182 failed_mount:
4183 if (sbi->s_chksum_driver)
4184 crypto_free_shash(sbi->s_chksum_driver);
4185 if (sbi->s_proc) {
4186 remove_proc_entry("options", sbi->s_proc);
4187 remove_proc_entry(sb->s_id, ext4_proc_root);
4188 }
4189 #ifdef CONFIG_QUOTA
4190 for (i = 0; i < MAXQUOTAS; i++)
4191 kfree(sbi->s_qf_names[i]);
4192 #endif
4193 ext4_blkdev_remove(sbi);
4194 brelse(bh);
4195 out_fail:
4196 sb->s_fs_info = NULL;
4197 kfree(sbi->s_blockgroup_lock);
4198 kfree(sbi);
4199 out_free_orig:
4200 kfree(orig_data);
4201 return err ? err : ret;
4202 }
4203
4204 /*
4205 * Setup any per-fs journal parameters now. We'll do this both on
4206 * initial mount, once the journal has been initialised but before we've
4207 * done any recovery; and again on any subsequent remount.
4208 */
4209 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4210 {
4211 struct ext4_sb_info *sbi = EXT4_SB(sb);
4212
4213 journal->j_commit_interval = sbi->s_commit_interval;
4214 journal->j_min_batch_time = sbi->s_min_batch_time;
4215 journal->j_max_batch_time = sbi->s_max_batch_time;
4216
4217 write_lock(&journal->j_state_lock);
4218 if (test_opt(sb, BARRIER))
4219 journal->j_flags |= JBD2_BARRIER;
4220 else
4221 journal->j_flags &= ~JBD2_BARRIER;
4222 if (test_opt(sb, DATA_ERR_ABORT))
4223 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4224 else
4225 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4226 write_unlock(&journal->j_state_lock);
4227 }
4228
4229 static journal_t *ext4_get_journal(struct super_block *sb,
4230 unsigned int journal_inum)
4231 {
4232 struct inode *journal_inode;
4233 journal_t *journal;
4234
4235 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4236
4237 /* First, test for the existence of a valid inode on disk. Bad
4238 * things happen if we iget() an unused inode, as the subsequent
4239 * iput() will try to delete it. */
4240
4241 journal_inode = ext4_iget(sb, journal_inum);
4242 if (IS_ERR(journal_inode)) {
4243 ext4_msg(sb, KERN_ERR, "no journal found");
4244 return NULL;
4245 }
4246 if (!journal_inode->i_nlink) {
4247 make_bad_inode(journal_inode);
4248 iput(journal_inode);
4249 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4250 return NULL;
4251 }
4252
4253 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4254 journal_inode, journal_inode->i_size);
4255 if (!S_ISREG(journal_inode->i_mode)) {
4256 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4257 iput(journal_inode);
4258 return NULL;
4259 }
4260
4261 journal = jbd2_journal_init_inode(journal_inode);
4262 if (!journal) {
4263 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4264 iput(journal_inode);
4265 return NULL;
4266 }
4267 journal->j_private = sb;
4268 ext4_init_journal_params(sb, journal);
4269 return journal;
4270 }
4271
4272 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4273 dev_t j_dev)
4274 {
4275 struct buffer_head *bh;
4276 journal_t *journal;
4277 ext4_fsblk_t start;
4278 ext4_fsblk_t len;
4279 int hblock, blocksize;
4280 ext4_fsblk_t sb_block;
4281 unsigned long offset;
4282 struct ext4_super_block *es;
4283 struct block_device *bdev;
4284
4285 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4286
4287 bdev = ext4_blkdev_get(j_dev, sb);
4288 if (bdev == NULL)
4289 return NULL;
4290
4291 blocksize = sb->s_blocksize;
4292 hblock = bdev_logical_block_size(bdev);
4293 if (blocksize < hblock) {
4294 ext4_msg(sb, KERN_ERR,
4295 "blocksize too small for journal device");
4296 goto out_bdev;
4297 }
4298
4299 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4300 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4301 set_blocksize(bdev, blocksize);
4302 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4303 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4304 "external journal");
4305 goto out_bdev;
4306 }
4307
4308 es = (struct ext4_super_block *) (bh->b_data + offset);
4309 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4310 !(le32_to_cpu(es->s_feature_incompat) &
4311 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4312 ext4_msg(sb, KERN_ERR, "external journal has "
4313 "bad superblock");
4314 brelse(bh);
4315 goto out_bdev;
4316 }
4317
4318 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4319 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4320 brelse(bh);
4321 goto out_bdev;
4322 }
4323
4324 len = ext4_blocks_count(es);
4325 start = sb_block + 1;
4326 brelse(bh); /* we're done with the superblock */
4327
4328 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4329 start, len, blocksize);
4330 if (!journal) {
4331 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4332 goto out_bdev;
4333 }
4334 journal->j_private = sb;
4335 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4336 wait_on_buffer(journal->j_sb_buffer);
4337 if (!buffer_uptodate(journal->j_sb_buffer)) {
4338 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4339 goto out_journal;
4340 }
4341 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4342 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4343 "user (unsupported) - %d",
4344 be32_to_cpu(journal->j_superblock->s_nr_users));
4345 goto out_journal;
4346 }
4347 EXT4_SB(sb)->journal_bdev = bdev;
4348 ext4_init_journal_params(sb, journal);
4349 return journal;
4350
4351 out_journal:
4352 jbd2_journal_destroy(journal);
4353 out_bdev:
4354 ext4_blkdev_put(bdev);
4355 return NULL;
4356 }
4357
4358 static int ext4_load_journal(struct super_block *sb,
4359 struct ext4_super_block *es,
4360 unsigned long journal_devnum)
4361 {
4362 journal_t *journal;
4363 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4364 dev_t journal_dev;
4365 int err = 0;
4366 int really_read_only;
4367
4368 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4369
4370 if (journal_devnum &&
4371 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4372 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4373 "numbers have changed");
4374 journal_dev = new_decode_dev(journal_devnum);
4375 } else
4376 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4377
4378 really_read_only = bdev_read_only(sb->s_bdev);
4379
4380 /*
4381 * Are we loading a blank journal or performing recovery after a
4382 * crash? For recovery, we need to check in advance whether we
4383 * can get read-write access to the device.
4384 */
4385 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4386 if (sb->s_flags & MS_RDONLY) {
4387 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4388 "required on readonly filesystem");
4389 if (really_read_only) {
4390 ext4_msg(sb, KERN_ERR, "write access "
4391 "unavailable, cannot proceed");
4392 return -EROFS;
4393 }
4394 ext4_msg(sb, KERN_INFO, "write access will "
4395 "be enabled during recovery");
4396 }
4397 }
4398
4399 if (journal_inum && journal_dev) {
4400 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4401 "and inode journals!");
4402 return -EINVAL;
4403 }
4404
4405 if (journal_inum) {
4406 if (!(journal = ext4_get_journal(sb, journal_inum)))
4407 return -EINVAL;
4408 } else {
4409 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4410 return -EINVAL;
4411 }
4412
4413 if (!(journal->j_flags & JBD2_BARRIER))
4414 ext4_msg(sb, KERN_INFO, "barriers disabled");
4415
4416 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4417 err = jbd2_journal_wipe(journal, !really_read_only);
4418 if (!err) {
4419 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4420 if (save)
4421 memcpy(save, ((char *) es) +
4422 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4423 err = jbd2_journal_load(journal);
4424 if (save)
4425 memcpy(((char *) es) + EXT4_S_ERR_START,
4426 save, EXT4_S_ERR_LEN);
4427 kfree(save);
4428 }
4429
4430 if (err) {
4431 ext4_msg(sb, KERN_ERR, "error loading journal");
4432 jbd2_journal_destroy(journal);
4433 return err;
4434 }
4435
4436 EXT4_SB(sb)->s_journal = journal;
4437 ext4_clear_journal_err(sb, es);
4438
4439 if (!really_read_only && journal_devnum &&
4440 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4441 es->s_journal_dev = cpu_to_le32(journal_devnum);
4442
4443 /* Make sure we flush the recovery flag to disk. */
4444 ext4_commit_super(sb, 1);
4445 }
4446
4447 return 0;
4448 }
4449
4450 static int ext4_commit_super(struct super_block *sb, int sync)
4451 {
4452 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4453 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4454 int error = 0;
4455
4456 if (!sbh || block_device_ejected(sb))
4457 return error;
4458 if (buffer_write_io_error(sbh)) {
4459 /*
4460 * Oh, dear. A previous attempt to write the
4461 * superblock failed. This could happen because the
4462 * USB device was yanked out. Or it could happen to
4463 * be a transient write error and maybe the block will
4464 * be remapped. Nothing we can do but to retry the
4465 * write and hope for the best.
4466 */
4467 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4468 "superblock detected");
4469 clear_buffer_write_io_error(sbh);
4470 set_buffer_uptodate(sbh);
4471 }
4472 /*
4473 * If the file system is mounted read-only, don't update the
4474 * superblock write time. This avoids updating the superblock
4475 * write time when we are mounting the root file system
4476 * read/only but we need to replay the journal; at that point,
4477 * for people who are east of GMT and who make their clock
4478 * tick in localtime for Windows bug-for-bug compatibility,
4479 * the clock is set in the future, and this will cause e2fsck
4480 * to complain and force a full file system check.
4481 */
4482 if (!(sb->s_flags & MS_RDONLY))
4483 es->s_wtime = cpu_to_le32(get_seconds());
4484 if (sb->s_bdev->bd_part)
4485 es->s_kbytes_written =
4486 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4487 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4488 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4489 else
4490 es->s_kbytes_written =
4491 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4492 ext4_free_blocks_count_set(es,
4493 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4494 &EXT4_SB(sb)->s_freeclusters_counter)));
4495 es->s_free_inodes_count =
4496 cpu_to_le32(percpu_counter_sum_positive(
4497 &EXT4_SB(sb)->s_freeinodes_counter));
4498 BUFFER_TRACE(sbh, "marking dirty");
4499 ext4_superblock_csum_set(sb);
4500 mark_buffer_dirty(sbh);
4501 if (sync) {
4502 error = sync_dirty_buffer(sbh);
4503 if (error)
4504 return error;
4505
4506 error = buffer_write_io_error(sbh);
4507 if (error) {
4508 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4509 "superblock");
4510 clear_buffer_write_io_error(sbh);
4511 set_buffer_uptodate(sbh);
4512 }
4513 }
4514 return error;
4515 }
4516
4517 /*
4518 * Have we just finished recovery? If so, and if we are mounting (or
4519 * remounting) the filesystem readonly, then we will end up with a
4520 * consistent fs on disk. Record that fact.
4521 */
4522 static void ext4_mark_recovery_complete(struct super_block *sb,
4523 struct ext4_super_block *es)
4524 {
4525 journal_t *journal = EXT4_SB(sb)->s_journal;
4526
4527 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4528 BUG_ON(journal != NULL);
4529 return;
4530 }
4531 jbd2_journal_lock_updates(journal);
4532 if (jbd2_journal_flush(journal) < 0)
4533 goto out;
4534
4535 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4536 sb->s_flags & MS_RDONLY) {
4537 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4538 ext4_commit_super(sb, 1);
4539 }
4540
4541 out:
4542 jbd2_journal_unlock_updates(journal);
4543 }
4544
4545 /*
4546 * If we are mounting (or read-write remounting) a filesystem whose journal
4547 * has recorded an error from a previous lifetime, move that error to the
4548 * main filesystem now.
4549 */
4550 static void ext4_clear_journal_err(struct super_block *sb,
4551 struct ext4_super_block *es)
4552 {
4553 journal_t *journal;
4554 int j_errno;
4555 const char *errstr;
4556
4557 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4558
4559 journal = EXT4_SB(sb)->s_journal;
4560
4561 /*
4562 * Now check for any error status which may have been recorded in the
4563 * journal by a prior ext4_error() or ext4_abort()
4564 */
4565
4566 j_errno = jbd2_journal_errno(journal);
4567 if (j_errno) {
4568 char nbuf[16];
4569
4570 errstr = ext4_decode_error(sb, j_errno, nbuf);
4571 ext4_warning(sb, "Filesystem error recorded "
4572 "from previous mount: %s", errstr);
4573 ext4_warning(sb, "Marking fs in need of filesystem check.");
4574
4575 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4576 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4577 ext4_commit_super(sb, 1);
4578
4579 jbd2_journal_clear_err(journal);
4580 jbd2_journal_update_sb_errno(journal);
4581 }
4582 }
4583
4584 /*
4585 * Force the running and committing transactions to commit,
4586 * and wait on the commit.
4587 */
4588 int ext4_force_commit(struct super_block *sb)
4589 {
4590 journal_t *journal;
4591
4592 if (sb->s_flags & MS_RDONLY)
4593 return 0;
4594
4595 journal = EXT4_SB(sb)->s_journal;
4596 return ext4_journal_force_commit(journal);
4597 }
4598
4599 static int ext4_sync_fs(struct super_block *sb, int wait)
4600 {
4601 int ret = 0;
4602 tid_t target;
4603 bool needs_barrier = false;
4604 struct ext4_sb_info *sbi = EXT4_SB(sb);
4605
4606 trace_ext4_sync_fs(sb, wait);
4607 flush_workqueue(sbi->rsv_conversion_wq);
4608 /*
4609 * Writeback quota in non-journalled quota case - journalled quota has
4610 * no dirty dquots
4611 */
4612 dquot_writeback_dquots(sb, -1);
4613 /*
4614 * Data writeback is possible w/o journal transaction, so barrier must
4615 * being sent at the end of the function. But we can skip it if
4616 * transaction_commit will do it for us.
4617 */
4618 target = jbd2_get_latest_transaction(sbi->s_journal);
4619 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4620 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4621 needs_barrier = true;
4622
4623 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4624 if (wait)
4625 ret = jbd2_log_wait_commit(sbi->s_journal, target);
4626 }
4627 if (needs_barrier) {
4628 int err;
4629 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4630 if (!ret)
4631 ret = err;
4632 }
4633
4634 return ret;
4635 }
4636
4637 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4638 {
4639 int ret = 0;
4640
4641 trace_ext4_sync_fs(sb, wait);
4642 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4643 dquot_writeback_dquots(sb, -1);
4644 if (wait && test_opt(sb, BARRIER))
4645 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4646
4647 return ret;
4648 }
4649
4650 /*
4651 * LVM calls this function before a (read-only) snapshot is created. This
4652 * gives us a chance to flush the journal completely and mark the fs clean.
4653 *
4654 * Note that only this function cannot bring a filesystem to be in a clean
4655 * state independently. It relies on upper layer to stop all data & metadata
4656 * modifications.
4657 */
4658 static int ext4_freeze(struct super_block *sb)
4659 {
4660 int error = 0;
4661 journal_t *journal;
4662
4663 if (sb->s_flags & MS_RDONLY)
4664 return 0;
4665
4666 journal = EXT4_SB(sb)->s_journal;
4667
4668 /* Now we set up the journal barrier. */
4669 jbd2_journal_lock_updates(journal);
4670
4671 /*
4672 * Don't clear the needs_recovery flag if we failed to flush
4673 * the journal.
4674 */
4675 error = jbd2_journal_flush(journal);
4676 if (error < 0)
4677 goto out;
4678
4679 /* Journal blocked and flushed, clear needs_recovery flag. */
4680 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4681 error = ext4_commit_super(sb, 1);
4682 out:
4683 /* we rely on upper layer to stop further updates */
4684 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4685 return error;
4686 }
4687
4688 /*
4689 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4690 * flag here, even though the filesystem is not technically dirty yet.
4691 */
4692 static int ext4_unfreeze(struct super_block *sb)
4693 {
4694 if (sb->s_flags & MS_RDONLY)
4695 return 0;
4696
4697 /* Reset the needs_recovery flag before the fs is unlocked. */
4698 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4699 ext4_commit_super(sb, 1);
4700 return 0;
4701 }
4702
4703 /*
4704 * Structure to save mount options for ext4_remount's benefit
4705 */
4706 struct ext4_mount_options {
4707 unsigned long s_mount_opt;
4708 unsigned long s_mount_opt2;
4709 kuid_t s_resuid;
4710 kgid_t s_resgid;
4711 unsigned long s_commit_interval;
4712 u32 s_min_batch_time, s_max_batch_time;
4713 #ifdef CONFIG_QUOTA
4714 int s_jquota_fmt;
4715 char *s_qf_names[MAXQUOTAS];
4716 #endif
4717 };
4718
4719 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4720 {
4721 struct ext4_super_block *es;
4722 struct ext4_sb_info *sbi = EXT4_SB(sb);
4723 unsigned long old_sb_flags;
4724 struct ext4_mount_options old_opts;
4725 int enable_quota = 0;
4726 ext4_group_t g;
4727 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4728 int err = 0;
4729 #ifdef CONFIG_QUOTA
4730 int i, j;
4731 #endif
4732 char *orig_data = kstrdup(data, GFP_KERNEL);
4733
4734 /* Store the original options */
4735 old_sb_flags = sb->s_flags;
4736 old_opts.s_mount_opt = sbi->s_mount_opt;
4737 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4738 old_opts.s_resuid = sbi->s_resuid;
4739 old_opts.s_resgid = sbi->s_resgid;
4740 old_opts.s_commit_interval = sbi->s_commit_interval;
4741 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4742 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4743 #ifdef CONFIG_QUOTA
4744 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4745 for (i = 0; i < MAXQUOTAS; i++)
4746 if (sbi->s_qf_names[i]) {
4747 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4748 GFP_KERNEL);
4749 if (!old_opts.s_qf_names[i]) {
4750 for (j = 0; j < i; j++)
4751 kfree(old_opts.s_qf_names[j]);
4752 kfree(orig_data);
4753 return -ENOMEM;
4754 }
4755 } else
4756 old_opts.s_qf_names[i] = NULL;
4757 #endif
4758 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4759 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4760
4761 /*
4762 * Allow the "check" option to be passed as a remount option.
4763 */
4764 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4765 err = -EINVAL;
4766 goto restore_opts;
4767 }
4768
4769 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4770 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4771 ext4_msg(sb, KERN_ERR, "can't mount with "
4772 "both data=journal and delalloc");
4773 err = -EINVAL;
4774 goto restore_opts;
4775 }
4776 if (test_opt(sb, DIOREAD_NOLOCK)) {
4777 ext4_msg(sb, KERN_ERR, "can't mount with "
4778 "both data=journal and dioread_nolock");
4779 err = -EINVAL;
4780 goto restore_opts;
4781 }
4782 }
4783
4784 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4785 ext4_abort(sb, "Abort forced by user");
4786
4787 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4788 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4789
4790 es = sbi->s_es;
4791
4792 if (sbi->s_journal) {
4793 ext4_init_journal_params(sb, sbi->s_journal);
4794 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4795 }
4796
4797 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4798 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4799 err = -EROFS;
4800 goto restore_opts;
4801 }
4802
4803 if (*flags & MS_RDONLY) {
4804 err = dquot_suspend(sb, -1);
4805 if (err < 0)
4806 goto restore_opts;
4807
4808 /*
4809 * First of all, the unconditional stuff we have to do
4810 * to disable replay of the journal when we next remount
4811 */
4812 sb->s_flags |= MS_RDONLY;
4813
4814 /*
4815 * OK, test if we are remounting a valid rw partition
4816 * readonly, and if so set the rdonly flag and then
4817 * mark the partition as valid again.
4818 */
4819 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4820 (sbi->s_mount_state & EXT4_VALID_FS))
4821 es->s_state = cpu_to_le16(sbi->s_mount_state);
4822
4823 if (sbi->s_journal)
4824 ext4_mark_recovery_complete(sb, es);
4825 } else {
4826 /* Make sure we can mount this feature set readwrite */
4827 if (!ext4_feature_set_ok(sb, 0)) {
4828 err = -EROFS;
4829 goto restore_opts;
4830 }
4831 /*
4832 * Make sure the group descriptor checksums
4833 * are sane. If they aren't, refuse to remount r/w.
4834 */
4835 for (g = 0; g < sbi->s_groups_count; g++) {
4836 struct ext4_group_desc *gdp =
4837 ext4_get_group_desc(sb, g, NULL);
4838
4839 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4840 ext4_msg(sb, KERN_ERR,
4841 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4842 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4843 le16_to_cpu(gdp->bg_checksum));
4844 err = -EINVAL;
4845 goto restore_opts;
4846 }
4847 }
4848
4849 /*
4850 * If we have an unprocessed orphan list hanging
4851 * around from a previously readonly bdev mount,
4852 * require a full umount/remount for now.
4853 */
4854 if (es->s_last_orphan) {
4855 ext4_msg(sb, KERN_WARNING, "Couldn't "
4856 "remount RDWR because of unprocessed "
4857 "orphan inode list. Please "
4858 "umount/remount instead");
4859 err = -EINVAL;
4860 goto restore_opts;
4861 }
4862
4863 /*
4864 * Mounting a RDONLY partition read-write, so reread
4865 * and store the current valid flag. (It may have
4866 * been changed by e2fsck since we originally mounted
4867 * the partition.)
4868 */
4869 if (sbi->s_journal)
4870 ext4_clear_journal_err(sb, es);
4871 sbi->s_mount_state = le16_to_cpu(es->s_state);
4872 if (!ext4_setup_super(sb, es, 0))
4873 sb->s_flags &= ~MS_RDONLY;
4874 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4875 EXT4_FEATURE_INCOMPAT_MMP))
4876 if (ext4_multi_mount_protect(sb,
4877 le64_to_cpu(es->s_mmp_block))) {
4878 err = -EROFS;
4879 goto restore_opts;
4880 }
4881 enable_quota = 1;
4882 }
4883 }
4884
4885 /*
4886 * Reinitialize lazy itable initialization thread based on
4887 * current settings
4888 */
4889 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4890 ext4_unregister_li_request(sb);
4891 else {
4892 ext4_group_t first_not_zeroed;
4893 first_not_zeroed = ext4_has_uninit_itable(sb);
4894 ext4_register_li_request(sb, first_not_zeroed);
4895 }
4896
4897 ext4_setup_system_zone(sb);
4898 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4899 ext4_commit_super(sb, 1);
4900
4901 #ifdef CONFIG_QUOTA
4902 /* Release old quota file names */
4903 for (i = 0; i < MAXQUOTAS; i++)
4904 kfree(old_opts.s_qf_names[i]);
4905 if (enable_quota) {
4906 if (sb_any_quota_suspended(sb))
4907 dquot_resume(sb, -1);
4908 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4909 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4910 err = ext4_enable_quotas(sb);
4911 if (err)
4912 goto restore_opts;
4913 }
4914 }
4915 #endif
4916
4917 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4918 kfree(orig_data);
4919 return 0;
4920
4921 restore_opts:
4922 sb->s_flags = old_sb_flags;
4923 sbi->s_mount_opt = old_opts.s_mount_opt;
4924 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4925 sbi->s_resuid = old_opts.s_resuid;
4926 sbi->s_resgid = old_opts.s_resgid;
4927 sbi->s_commit_interval = old_opts.s_commit_interval;
4928 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4929 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4930 #ifdef CONFIG_QUOTA
4931 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4932 for (i = 0; i < MAXQUOTAS; i++) {
4933 kfree(sbi->s_qf_names[i]);
4934 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4935 }
4936 #endif
4937 kfree(orig_data);
4938 return err;
4939 }
4940
4941 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4942 {
4943 struct super_block *sb = dentry->d_sb;
4944 struct ext4_sb_info *sbi = EXT4_SB(sb);
4945 struct ext4_super_block *es = sbi->s_es;
4946 ext4_fsblk_t overhead = 0, resv_blocks;
4947 u64 fsid;
4948 s64 bfree;
4949 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4950
4951 if (!test_opt(sb, MINIX_DF))
4952 overhead = sbi->s_overhead;
4953
4954 buf->f_type = EXT4_SUPER_MAGIC;
4955 buf->f_bsize = sb->s_blocksize;
4956 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4957 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4958 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4959 /* prevent underflow in case that few free space is available */
4960 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4961 buf->f_bavail = buf->f_bfree -
4962 (ext4_r_blocks_count(es) + resv_blocks);
4963 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4964 buf->f_bavail = 0;
4965 buf->f_files = le32_to_cpu(es->s_inodes_count);
4966 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4967 buf->f_namelen = EXT4_NAME_LEN;
4968 fsid = le64_to_cpup((void *)es->s_uuid) ^
4969 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4970 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4971 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4972
4973 return 0;
4974 }
4975
4976 /* Helper function for writing quotas on sync - we need to start transaction
4977 * before quota file is locked for write. Otherwise the are possible deadlocks:
4978 * Process 1 Process 2
4979 * ext4_create() quota_sync()
4980 * jbd2_journal_start() write_dquot()
4981 * dquot_initialize() down(dqio_mutex)
4982 * down(dqio_mutex) jbd2_journal_start()
4983 *
4984 */
4985
4986 #ifdef CONFIG_QUOTA
4987
4988 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4989 {
4990 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4991 }
4992
4993 static int ext4_write_dquot(struct dquot *dquot)
4994 {
4995 int ret, err;
4996 handle_t *handle;
4997 struct inode *inode;
4998
4999 inode = dquot_to_inode(dquot);
5000 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5001 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5002 if (IS_ERR(handle))
5003 return PTR_ERR(handle);
5004 ret = dquot_commit(dquot);
5005 err = ext4_journal_stop(handle);
5006 if (!ret)
5007 ret = err;
5008 return ret;
5009 }
5010
5011 static int ext4_acquire_dquot(struct dquot *dquot)
5012 {
5013 int ret, err;
5014 handle_t *handle;
5015
5016 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5017 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5018 if (IS_ERR(handle))
5019 return PTR_ERR(handle);
5020 ret = dquot_acquire(dquot);
5021 err = ext4_journal_stop(handle);
5022 if (!ret)
5023 ret = err;
5024 return ret;
5025 }
5026
5027 static int ext4_release_dquot(struct dquot *dquot)
5028 {
5029 int ret, err;
5030 handle_t *handle;
5031
5032 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5033 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5034 if (IS_ERR(handle)) {
5035 /* Release dquot anyway to avoid endless cycle in dqput() */
5036 dquot_release(dquot);
5037 return PTR_ERR(handle);
5038 }
5039 ret = dquot_release(dquot);
5040 err = ext4_journal_stop(handle);
5041 if (!ret)
5042 ret = err;
5043 return ret;
5044 }
5045
5046 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5047 {
5048 struct super_block *sb = dquot->dq_sb;
5049 struct ext4_sb_info *sbi = EXT4_SB(sb);
5050
5051 /* Are we journaling quotas? */
5052 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5053 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5054 dquot_mark_dquot_dirty(dquot);
5055 return ext4_write_dquot(dquot);
5056 } else {
5057 return dquot_mark_dquot_dirty(dquot);
5058 }
5059 }
5060
5061 static int ext4_write_info(struct super_block *sb, int type)
5062 {
5063 int ret, err;
5064 handle_t *handle;
5065
5066 /* Data block + inode block */
5067 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5068 if (IS_ERR(handle))
5069 return PTR_ERR(handle);
5070 ret = dquot_commit_info(sb, type);
5071 err = ext4_journal_stop(handle);
5072 if (!ret)
5073 ret = err;
5074 return ret;
5075 }
5076
5077 /*
5078 * Turn on quotas during mount time - we need to find
5079 * the quota file and such...
5080 */
5081 static int ext4_quota_on_mount(struct super_block *sb, int type)
5082 {
5083 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5084 EXT4_SB(sb)->s_jquota_fmt, type);
5085 }
5086
5087 /*
5088 * Standard function to be called on quota_on
5089 */
5090 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5091 struct path *path)
5092 {
5093 int err;
5094
5095 if (!test_opt(sb, QUOTA))
5096 return -EINVAL;
5097
5098 /* Quotafile not on the same filesystem? */
5099 if (path->dentry->d_sb != sb)
5100 return -EXDEV;
5101 /* Journaling quota? */
5102 if (EXT4_SB(sb)->s_qf_names[type]) {
5103 /* Quotafile not in fs root? */
5104 if (path->dentry->d_parent != sb->s_root)
5105 ext4_msg(sb, KERN_WARNING,
5106 "Quota file not on filesystem root. "
5107 "Journaled quota will not work");
5108 }
5109
5110 /*
5111 * When we journal data on quota file, we have to flush journal to see
5112 * all updates to the file when we bypass pagecache...
5113 */
5114 if (EXT4_SB(sb)->s_journal &&
5115 ext4_should_journal_data(path->dentry->d_inode)) {
5116 /*
5117 * We don't need to lock updates but journal_flush() could
5118 * otherwise be livelocked...
5119 */
5120 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5121 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5122 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5123 if (err)
5124 return err;
5125 }
5126
5127 return dquot_quota_on(sb, type, format_id, path);
5128 }
5129
5130 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5131 unsigned int flags)
5132 {
5133 int err;
5134 struct inode *qf_inode;
5135 unsigned long qf_inums[MAXQUOTAS] = {
5136 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5137 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5138 };
5139
5140 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5141
5142 if (!qf_inums[type])
5143 return -EPERM;
5144
5145 qf_inode = ext4_iget(sb, qf_inums[type]);
5146 if (IS_ERR(qf_inode)) {
5147 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5148 return PTR_ERR(qf_inode);
5149 }
5150
5151 /* Don't account quota for quota files to avoid recursion */
5152 qf_inode->i_flags |= S_NOQUOTA;
5153 err = dquot_enable(qf_inode, type, format_id, flags);
5154 iput(qf_inode);
5155
5156 return err;
5157 }
5158
5159 /* Enable usage tracking for all quota types. */
5160 static int ext4_enable_quotas(struct super_block *sb)
5161 {
5162 int type, err = 0;
5163 unsigned long qf_inums[MAXQUOTAS] = {
5164 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5165 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5166 };
5167
5168 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5169 for (type = 0; type < MAXQUOTAS; type++) {
5170 if (qf_inums[type]) {
5171 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5172 DQUOT_USAGE_ENABLED);
5173 if (err) {
5174 ext4_warning(sb,
5175 "Failed to enable quota tracking "
5176 "(type=%d, err=%d). Please run "
5177 "e2fsck to fix.", type, err);
5178 return err;
5179 }
5180 }
5181 }
5182 return 0;
5183 }
5184
5185 /*
5186 * quota_on function that is used when QUOTA feature is set.
5187 */
5188 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5189 int format_id)
5190 {
5191 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5192 return -EINVAL;
5193
5194 /*
5195 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5196 */
5197 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5198 }
5199
5200 static int ext4_quota_off(struct super_block *sb, int type)
5201 {
5202 struct inode *inode = sb_dqopt(sb)->files[type];
5203 handle_t *handle;
5204
5205 /* Force all delayed allocation blocks to be allocated.
5206 * Caller already holds s_umount sem */
5207 if (test_opt(sb, DELALLOC))
5208 sync_filesystem(sb);
5209
5210 if (!inode)
5211 goto out;
5212
5213 /* Update modification times of quota files when userspace can
5214 * start looking at them */
5215 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5216 if (IS_ERR(handle))
5217 goto out;
5218 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5219 ext4_mark_inode_dirty(handle, inode);
5220 ext4_journal_stop(handle);
5221
5222 out:
5223 return dquot_quota_off(sb, type);
5224 }
5225
5226 /*
5227 * quota_off function that is used when QUOTA feature is set.
5228 */
5229 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5230 {
5231 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5232 return -EINVAL;
5233
5234 /* Disable only the limits. */
5235 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5236 }
5237
5238 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5239 * acquiring the locks... As quota files are never truncated and quota code
5240 * itself serializes the operations (and no one else should touch the files)
5241 * we don't have to be afraid of races */
5242 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5243 size_t len, loff_t off)
5244 {
5245 struct inode *inode = sb_dqopt(sb)->files[type];
5246 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5247 int err = 0;
5248 int offset = off & (sb->s_blocksize - 1);
5249 int tocopy;
5250 size_t toread;
5251 struct buffer_head *bh;
5252 loff_t i_size = i_size_read(inode);
5253
5254 if (off > i_size)
5255 return 0;
5256 if (off+len > i_size)
5257 len = i_size-off;
5258 toread = len;
5259 while (toread > 0) {
5260 tocopy = sb->s_blocksize - offset < toread ?
5261 sb->s_blocksize - offset : toread;
5262 bh = ext4_bread(NULL, inode, blk, 0, &err);
5263 if (err)
5264 return err;
5265 if (!bh) /* A hole? */
5266 memset(data, 0, tocopy);
5267 else
5268 memcpy(data, bh->b_data+offset, tocopy);
5269 brelse(bh);
5270 offset = 0;
5271 toread -= tocopy;
5272 data += tocopy;
5273 blk++;
5274 }
5275 return len;
5276 }
5277
5278 /* Write to quotafile (we know the transaction is already started and has
5279 * enough credits) */
5280 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5281 const char *data, size_t len, loff_t off)
5282 {
5283 struct inode *inode = sb_dqopt(sb)->files[type];
5284 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5285 int err = 0;
5286 int offset = off & (sb->s_blocksize - 1);
5287 struct buffer_head *bh;
5288 handle_t *handle = journal_current_handle();
5289
5290 if (EXT4_SB(sb)->s_journal && !handle) {
5291 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5292 " cancelled because transaction is not started",
5293 (unsigned long long)off, (unsigned long long)len);
5294 return -EIO;
5295 }
5296 /*
5297 * Since we account only one data block in transaction credits,
5298 * then it is impossible to cross a block boundary.
5299 */
5300 if (sb->s_blocksize - offset < len) {
5301 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5302 " cancelled because not block aligned",
5303 (unsigned long long)off, (unsigned long long)len);
5304 return -EIO;
5305 }
5306
5307 bh = ext4_bread(handle, inode, blk, 1, &err);
5308 if (!bh)
5309 goto out;
5310 err = ext4_journal_get_write_access(handle, bh);
5311 if (err) {
5312 brelse(bh);
5313 goto out;
5314 }
5315 lock_buffer(bh);
5316 memcpy(bh->b_data+offset, data, len);
5317 flush_dcache_page(bh->b_page);
5318 unlock_buffer(bh);
5319 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5320 brelse(bh);
5321 out:
5322 if (err)
5323 return err;
5324 if (inode->i_size < off + len) {
5325 i_size_write(inode, off + len);
5326 EXT4_I(inode)->i_disksize = inode->i_size;
5327 ext4_mark_inode_dirty(handle, inode);
5328 }
5329 return len;
5330 }
5331
5332 #endif
5333
5334 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5335 const char *dev_name, void *data)
5336 {
5337 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5338 }
5339
5340 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5341 static inline void register_as_ext2(void)
5342 {
5343 int err = register_filesystem(&ext2_fs_type);
5344 if (err)
5345 printk(KERN_WARNING
5346 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5347 }
5348
5349 static inline void unregister_as_ext2(void)
5350 {
5351 unregister_filesystem(&ext2_fs_type);
5352 }
5353
5354 static inline int ext2_feature_set_ok(struct super_block *sb)
5355 {
5356 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5357 return 0;
5358 if (sb->s_flags & MS_RDONLY)
5359 return 1;
5360 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5361 return 0;
5362 return 1;
5363 }
5364 #else
5365 static inline void register_as_ext2(void) { }
5366 static inline void unregister_as_ext2(void) { }
5367 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5368 #endif
5369
5370 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5371 static inline void register_as_ext3(void)
5372 {
5373 int err = register_filesystem(&ext3_fs_type);
5374 if (err)
5375 printk(KERN_WARNING
5376 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5377 }
5378
5379 static inline void unregister_as_ext3(void)
5380 {
5381 unregister_filesystem(&ext3_fs_type);
5382 }
5383
5384 static inline int ext3_feature_set_ok(struct super_block *sb)
5385 {
5386 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5387 return 0;
5388 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5389 return 0;
5390 if (sb->s_flags & MS_RDONLY)
5391 return 1;
5392 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5393 return 0;
5394 return 1;
5395 }
5396 #else
5397 static inline void register_as_ext3(void) { }
5398 static inline void unregister_as_ext3(void) { }
5399 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5400 #endif
5401
5402 static struct file_system_type ext4_fs_type = {
5403 .owner = THIS_MODULE,
5404 .name = "ext4",
5405 .mount = ext4_mount,
5406 .kill_sb = kill_block_super,
5407 .fs_flags = FS_REQUIRES_DEV,
5408 };
5409 MODULE_ALIAS_FS("ext4");
5410
5411 static int __init ext4_init_feat_adverts(void)
5412 {
5413 struct ext4_features *ef;
5414 int ret = -ENOMEM;
5415
5416 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5417 if (!ef)
5418 goto out;
5419
5420 ef->f_kobj.kset = ext4_kset;
5421 init_completion(&ef->f_kobj_unregister);
5422 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5423 "features");
5424 if (ret) {
5425 kfree(ef);
5426 goto out;
5427 }
5428
5429 ext4_feat = ef;
5430 ret = 0;
5431 out:
5432 return ret;
5433 }
5434
5435 static void ext4_exit_feat_adverts(void)
5436 {
5437 kobject_put(&ext4_feat->f_kobj);
5438 wait_for_completion(&ext4_feat->f_kobj_unregister);
5439 kfree(ext4_feat);
5440 }
5441
5442 /* Shared across all ext4 file systems */
5443 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5444 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5445
5446 static int __init ext4_init_fs(void)
5447 {
5448 int i, err;
5449
5450 ext4_li_info = NULL;
5451 mutex_init(&ext4_li_mtx);
5452
5453 /* Build-time check for flags consistency */
5454 ext4_check_flag_values();
5455
5456 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5457 mutex_init(&ext4__aio_mutex[i]);
5458 init_waitqueue_head(&ext4__ioend_wq[i]);
5459 }
5460
5461 err = ext4_init_es();
5462 if (err)
5463 return err;
5464
5465 err = ext4_init_pageio();
5466 if (err)
5467 goto out7;
5468
5469 err = ext4_init_system_zone();
5470 if (err)
5471 goto out6;
5472 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5473 if (!ext4_kset) {
5474 err = -ENOMEM;
5475 goto out5;
5476 }
5477 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5478
5479 err = ext4_init_feat_adverts();
5480 if (err)
5481 goto out4;
5482
5483 err = ext4_init_mballoc();
5484 if (err)
5485 goto out3;
5486
5487 err = ext4_init_xattr();
5488 if (err)
5489 goto out2;
5490 err = init_inodecache();
5491 if (err)
5492 goto out1;
5493 register_as_ext3();
5494 register_as_ext2();
5495 err = register_filesystem(&ext4_fs_type);
5496 if (err)
5497 goto out;
5498
5499 return 0;
5500 out:
5501 unregister_as_ext2();
5502 unregister_as_ext3();
5503 destroy_inodecache();
5504 out1:
5505 ext4_exit_xattr();
5506 out2:
5507 ext4_exit_mballoc();
5508 out3:
5509 ext4_exit_feat_adverts();
5510 out4:
5511 if (ext4_proc_root)
5512 remove_proc_entry("fs/ext4", NULL);
5513 kset_unregister(ext4_kset);
5514 out5:
5515 ext4_exit_system_zone();
5516 out6:
5517 ext4_exit_pageio();
5518 out7:
5519 ext4_exit_es();
5520
5521 return err;
5522 }
5523
5524 static void __exit ext4_exit_fs(void)
5525 {
5526 ext4_destroy_lazyinit_thread();
5527 unregister_as_ext2();
5528 unregister_as_ext3();
5529 unregister_filesystem(&ext4_fs_type);
5530 destroy_inodecache();
5531 ext4_exit_xattr();
5532 ext4_exit_mballoc();
5533 ext4_exit_feat_adverts();
5534 remove_proc_entry("fs/ext4", NULL);
5535 kset_unregister(ext4_kset);
5536 ext4_exit_system_zone();
5537 ext4_exit_pageio();
5538 ext4_exit_es();
5539 }
5540
5541 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5542 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5543 MODULE_LICENSE("GPL");
5544 module_init(ext4_init_fs)
5545 module_exit(ext4_exit_fs)
Linux with added FPC-III support