projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
Merge branch 'for-linus' of git://git.infradead.org/users/vkoul/slave-dma
[deliverable/linux.git] / fs / ext4 / super.c
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/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 static int ext4_mballoc_ready;
63
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
87
88 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89 static struct file_system_type ext2_fs_type = {
90 .owner = THIS_MODULE,
91 .name = "ext2",
92 .mount = ext4_mount,
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
95 };
96 MODULE_ALIAS_FS("ext2");
97 MODULE_ALIAS("ext2");
98 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
99 #else
100 #define IS_EXT2_SB(sb) (0)
101 #endif
102
103
104 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
105 static struct file_system_type ext3_fs_type = {
106 .owner = THIS_MODULE,
107 .name = "ext3",
108 .mount = ext4_mount,
109 .kill_sb = kill_block_super,
110 .fs_flags = FS_REQUIRES_DEV,
111 };
112 MODULE_ALIAS_FS("ext3");
113 MODULE_ALIAS("ext3");
114 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
115 #else
116 #define IS_EXT3_SB(sb) (0)
117 #endif
118
119 static int ext4_verify_csum_type(struct super_block *sb,
120 struct ext4_super_block *es)
121 {
122 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
123 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
124 return 1;
125
126 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
127 }
128
129 static __le32 ext4_superblock_csum(struct super_block *sb,
130 struct ext4_super_block *es)
131 {
132 struct ext4_sb_info *sbi = EXT4_SB(sb);
133 int offset = offsetof(struct ext4_super_block, s_checksum);
134 __u32 csum;
135
136 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
137
138 return cpu_to_le32(csum);
139 }
140
141 static int ext4_superblock_csum_verify(struct super_block *sb,
142 struct ext4_super_block *es)
143 {
144 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
145 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
146 return 1;
147
148 return es->s_checksum == ext4_superblock_csum(sb, es);
149 }
150
151 void ext4_superblock_csum_set(struct super_block *sb)
152 {
153 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
154
155 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
156 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
157 return;
158
159 es->s_checksum = ext4_superblock_csum(sb, es);
160 }
161
162 void *ext4_kvmalloc(size_t size, gfp_t flags)
163 {
164 void *ret;
165
166 ret = kmalloc(size, flags | __GFP_NOWARN);
167 if (!ret)
168 ret = __vmalloc(size, flags, PAGE_KERNEL);
169 return ret;
170 }
171
172 void *ext4_kvzalloc(size_t size, gfp_t flags)
173 {
174 void *ret;
175
176 ret = kzalloc(size, flags | __GFP_NOWARN);
177 if (!ret)
178 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
179 return ret;
180 }
181
182 void ext4_kvfree(void *ptr)
183 {
184 if (is_vmalloc_addr(ptr))
185 vfree(ptr);
186 else
187 kfree(ptr);
188
189 }
190
191 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
192 struct ext4_group_desc *bg)
193 {
194 return le32_to_cpu(bg->bg_block_bitmap_lo) |
195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
196 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
197 }
198
199 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
200 struct ext4_group_desc *bg)
201 {
202 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
203 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
204 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
205 }
206
207 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
208 struct ext4_group_desc *bg)
209 {
210 return le32_to_cpu(bg->bg_inode_table_lo) |
211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
212 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
213 }
214
215 __u32 ext4_free_group_clusters(struct super_block *sb,
216 struct ext4_group_desc *bg)
217 {
218 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
220 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
221 }
222
223 __u32 ext4_free_inodes_count(struct super_block *sb,
224 struct ext4_group_desc *bg)
225 {
226 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
228 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
229 }
230
231 __u32 ext4_used_dirs_count(struct super_block *sb,
232 struct ext4_group_desc *bg)
233 {
234 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
236 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
237 }
238
239 __u32 ext4_itable_unused_count(struct super_block *sb,
240 struct ext4_group_desc *bg)
241 {
242 return le16_to_cpu(bg->bg_itable_unused_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
244 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
245 }
246
247 void ext4_block_bitmap_set(struct super_block *sb,
248 struct ext4_group_desc *bg, ext4_fsblk_t blk)
249 {
250 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
253 }
254
255 void ext4_inode_bitmap_set(struct super_block *sb,
256 struct ext4_group_desc *bg, ext4_fsblk_t blk)
257 {
258 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
259 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
260 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
261 }
262
263 void ext4_inode_table_set(struct super_block *sb,
264 struct ext4_group_desc *bg, ext4_fsblk_t blk)
265 {
266 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
269 }
270
271 void ext4_free_group_clusters_set(struct super_block *sb,
272 struct ext4_group_desc *bg, __u32 count)
273 {
274 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
277 }
278
279 void ext4_free_inodes_set(struct super_block *sb,
280 struct ext4_group_desc *bg, __u32 count)
281 {
282 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
285 }
286
287 void ext4_used_dirs_set(struct super_block *sb,
288 struct ext4_group_desc *bg, __u32 count)
289 {
290 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
293 }
294
295 void ext4_itable_unused_set(struct super_block *sb,
296 struct ext4_group_desc *bg, __u32 count)
297 {
298 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
301 }
302
303
304 static void __save_error_info(struct super_block *sb, const char *func,
305 unsigned int line)
306 {
307 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
308
309 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
310 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
311 es->s_last_error_time = cpu_to_le32(get_seconds());
312 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
313 es->s_last_error_line = cpu_to_le32(line);
314 if (!es->s_first_error_time) {
315 es->s_first_error_time = es->s_last_error_time;
316 strncpy(es->s_first_error_func, func,
317 sizeof(es->s_first_error_func));
318 es->s_first_error_line = cpu_to_le32(line);
319 es->s_first_error_ino = es->s_last_error_ino;
320 es->s_first_error_block = es->s_last_error_block;
321 }
322 /*
323 * Start the daily error reporting function if it hasn't been
324 * started already
325 */
326 if (!es->s_error_count)
327 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
328 le32_add_cpu(&es->s_error_count, 1);
329 }
330
331 static void save_error_info(struct super_block *sb, const char *func,
332 unsigned int line)
333 {
334 __save_error_info(sb, func, line);
335 ext4_commit_super(sb, 1);
336 }
337
338 /*
339 * The del_gendisk() function uninitializes the disk-specific data
340 * structures, including the bdi structure, without telling anyone
341 * else. Once this happens, any attempt to call mark_buffer_dirty()
342 * (for example, by ext4_commit_super), will cause a kernel OOPS.
343 * This is a kludge to prevent these oops until we can put in a proper
344 * hook in del_gendisk() to inform the VFS and file system layers.
345 */
346 static int block_device_ejected(struct super_block *sb)
347 {
348 struct inode *bd_inode = sb->s_bdev->bd_inode;
349 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
350
351 return bdi->dev == NULL;
352 }
353
354 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
355 {
356 struct super_block *sb = journal->j_private;
357 struct ext4_sb_info *sbi = EXT4_SB(sb);
358 int error = is_journal_aborted(journal);
359 struct ext4_journal_cb_entry *jce;
360
361 BUG_ON(txn->t_state == T_FINISHED);
362 spin_lock(&sbi->s_md_lock);
363 while (!list_empty(&txn->t_private_list)) {
364 jce = list_entry(txn->t_private_list.next,
365 struct ext4_journal_cb_entry, jce_list);
366 list_del_init(&jce->jce_list);
367 spin_unlock(&sbi->s_md_lock);
368 jce->jce_func(sb, jce, error);
369 spin_lock(&sbi->s_md_lock);
370 }
371 spin_unlock(&sbi->s_md_lock);
372 }
373
374 /* Deal with the reporting of failure conditions on a filesystem such as
375 * inconsistencies detected or read IO failures.
376 *
377 * On ext2, we can store the error state of the filesystem in the
378 * superblock. That is not possible on ext4, because we may have other
379 * write ordering constraints on the superblock which prevent us from
380 * writing it out straight away; and given that the journal is about to
381 * be aborted, we can't rely on the current, or future, transactions to
382 * write out the superblock safely.
383 *
384 * We'll just use the jbd2_journal_abort() error code to record an error in
385 * the journal instead. On recovery, the journal will complain about
386 * that error until we've noted it down and cleared it.
387 */
388
389 static void ext4_handle_error(struct super_block *sb)
390 {
391 if (sb->s_flags & MS_RDONLY)
392 return;
393
394 if (!test_opt(sb, ERRORS_CONT)) {
395 journal_t *journal = EXT4_SB(sb)->s_journal;
396
397 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
398 if (journal)
399 jbd2_journal_abort(journal, -EIO);
400 }
401 if (test_opt(sb, ERRORS_RO)) {
402 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
403 /*
404 * Make sure updated value of ->s_mount_flags will be visible
405 * before ->s_flags update
406 */
407 smp_wmb();
408 sb->s_flags |= MS_RDONLY;
409 }
410 if (test_opt(sb, ERRORS_PANIC))
411 panic("EXT4-fs (device %s): panic forced after error\n",
412 sb->s_id);
413 }
414
415 #define ext4_error_ratelimit(sb) \
416 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
417 "EXT4-fs error")
418
419 void __ext4_error(struct super_block *sb, const char *function,
420 unsigned int line, const char *fmt, ...)
421 {
422 struct va_format vaf;
423 va_list args;
424
425 if (ext4_error_ratelimit(sb)) {
426 va_start(args, fmt);
427 vaf.fmt = fmt;
428 vaf.va = &args;
429 printk(KERN_CRIT
430 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
431 sb->s_id, function, line, current->comm, &vaf);
432 va_end(args);
433 }
434 save_error_info(sb, function, line);
435 ext4_handle_error(sb);
436 }
437
438 void __ext4_error_inode(struct inode *inode, const char *function,
439 unsigned int line, ext4_fsblk_t block,
440 const char *fmt, ...)
441 {
442 va_list args;
443 struct va_format vaf;
444 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
445
446 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
447 es->s_last_error_block = cpu_to_le64(block);
448 if (ext4_error_ratelimit(inode->i_sb)) {
449 va_start(args, fmt);
450 vaf.fmt = fmt;
451 vaf.va = &args;
452 if (block)
453 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
454 "inode #%lu: block %llu: comm %s: %pV\n",
455 inode->i_sb->s_id, function, line, inode->i_ino,
456 block, current->comm, &vaf);
457 else
458 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
459 "inode #%lu: comm %s: %pV\n",
460 inode->i_sb->s_id, function, line, inode->i_ino,
461 current->comm, &vaf);
462 va_end(args);
463 }
464 save_error_info(inode->i_sb, function, line);
465 ext4_handle_error(inode->i_sb);
466 }
467
468 void __ext4_error_file(struct file *file, const char *function,
469 unsigned int line, ext4_fsblk_t block,
470 const char *fmt, ...)
471 {
472 va_list args;
473 struct va_format vaf;
474 struct ext4_super_block *es;
475 struct inode *inode = file_inode(file);
476 char pathname[80], *path;
477
478 es = EXT4_SB(inode->i_sb)->s_es;
479 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
480 if (ext4_error_ratelimit(inode->i_sb)) {
481 path = d_path(&(file->f_path), pathname, sizeof(pathname));
482 if (IS_ERR(path))
483 path = "(unknown)";
484 va_start(args, fmt);
485 vaf.fmt = fmt;
486 vaf.va = &args;
487 if (block)
488 printk(KERN_CRIT
489 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
490 "block %llu: comm %s: path %s: %pV\n",
491 inode->i_sb->s_id, function, line, inode->i_ino,
492 block, current->comm, path, &vaf);
493 else
494 printk(KERN_CRIT
495 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
496 "comm %s: path %s: %pV\n",
497 inode->i_sb->s_id, function, line, inode->i_ino,
498 current->comm, path, &vaf);
499 va_end(args);
500 }
501 save_error_info(inode->i_sb, function, line);
502 ext4_handle_error(inode->i_sb);
503 }
504
505 const char *ext4_decode_error(struct super_block *sb, int errno,
506 char nbuf[16])
507 {
508 char *errstr = NULL;
509
510 switch (errno) {
511 case -EIO:
512 errstr = "IO failure";
513 break;
514 case -ENOMEM:
515 errstr = "Out of memory";
516 break;
517 case -EROFS:
518 if (!sb || (EXT4_SB(sb)->s_journal &&
519 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
520 errstr = "Journal has aborted";
521 else
522 errstr = "Readonly filesystem";
523 break;
524 default:
525 /* If the caller passed in an extra buffer for unknown
526 * errors, textualise them now. Else we just return
527 * NULL. */
528 if (nbuf) {
529 /* Check for truncated error codes... */
530 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
531 errstr = nbuf;
532 }
533 break;
534 }
535
536 return errstr;
537 }
538
539 /* __ext4_std_error decodes expected errors from journaling functions
540 * automatically and invokes the appropriate error response. */
541
542 void __ext4_std_error(struct super_block *sb, const char *function,
543 unsigned int line, int errno)
544 {
545 char nbuf[16];
546 const char *errstr;
547
548 /* Special case: if the error is EROFS, and we're not already
549 * inside a transaction, then there's really no point in logging
550 * an error. */
551 if (errno == -EROFS && journal_current_handle() == NULL &&
552 (sb->s_flags & MS_RDONLY))
553 return;
554
555 if (ext4_error_ratelimit(sb)) {
556 errstr = ext4_decode_error(sb, errno, nbuf);
557 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
558 sb->s_id, function, line, errstr);
559 }
560
561 save_error_info(sb, function, line);
562 ext4_handle_error(sb);
563 }
564
565 /*
566 * ext4_abort is a much stronger failure handler than ext4_error. The
567 * abort function may be used to deal with unrecoverable failures such
568 * as journal IO errors or ENOMEM at a critical moment in log management.
569 *
570 * We unconditionally force the filesystem into an ABORT|READONLY state,
571 * unless the error response on the fs has been set to panic in which
572 * case we take the easy way out and panic immediately.
573 */
574
575 void __ext4_abort(struct super_block *sb, const char *function,
576 unsigned int line, const char *fmt, ...)
577 {
578 va_list args;
579
580 save_error_info(sb, function, line);
581 va_start(args, fmt);
582 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
583 function, line);
584 vprintk(fmt, args);
585 printk("\n");
586 va_end(args);
587
588 if ((sb->s_flags & MS_RDONLY) == 0) {
589 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
590 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
591 /*
592 * Make sure updated value of ->s_mount_flags will be visible
593 * before ->s_flags update
594 */
595 smp_wmb();
596 sb->s_flags |= MS_RDONLY;
597 if (EXT4_SB(sb)->s_journal)
598 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
599 save_error_info(sb, function, line);
600 }
601 if (test_opt(sb, ERRORS_PANIC))
602 panic("EXT4-fs panic from previous error\n");
603 }
604
605 void __ext4_msg(struct super_block *sb,
606 const char *prefix, const char *fmt, ...)
607 {
608 struct va_format vaf;
609 va_list args;
610
611 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
612 return;
613
614 va_start(args, fmt);
615 vaf.fmt = fmt;
616 vaf.va = &args;
617 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
618 va_end(args);
619 }
620
621 void __ext4_warning(struct super_block *sb, const char *function,
622 unsigned int line, const char *fmt, ...)
623 {
624 struct va_format vaf;
625 va_list args;
626
627 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
628 "EXT4-fs warning"))
629 return;
630
631 va_start(args, fmt);
632 vaf.fmt = fmt;
633 vaf.va = &args;
634 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
635 sb->s_id, function, line, &vaf);
636 va_end(args);
637 }
638
639 void __ext4_grp_locked_error(const char *function, unsigned int line,
640 struct super_block *sb, ext4_group_t grp,
641 unsigned long ino, ext4_fsblk_t block,
642 const char *fmt, ...)
643 __releases(bitlock)
644 __acquires(bitlock)
645 {
646 struct va_format vaf;
647 va_list args;
648 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
649
650 es->s_last_error_ino = cpu_to_le32(ino);
651 es->s_last_error_block = cpu_to_le64(block);
652 __save_error_info(sb, function, line);
653
654 if (ext4_error_ratelimit(sb)) {
655 va_start(args, fmt);
656 vaf.fmt = fmt;
657 vaf.va = &args;
658 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
659 sb->s_id, function, line, grp);
660 if (ino)
661 printk(KERN_CONT "inode %lu: ", ino);
662 if (block)
663 printk(KERN_CONT "block %llu:",
664 (unsigned long long) block);
665 printk(KERN_CONT "%pV\n", &vaf);
666 va_end(args);
667 }
668
669 if (test_opt(sb, ERRORS_CONT)) {
670 ext4_commit_super(sb, 0);
671 return;
672 }
673
674 ext4_unlock_group(sb, grp);
675 ext4_handle_error(sb);
676 /*
677 * We only get here in the ERRORS_RO case; relocking the group
678 * may be dangerous, but nothing bad will happen since the
679 * filesystem will have already been marked read/only and the
680 * journal has been aborted. We return 1 as a hint to callers
681 * who might what to use the return value from
682 * ext4_grp_locked_error() to distinguish between the
683 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
684 * aggressively from the ext4 function in question, with a
685 * more appropriate error code.
686 */
687 ext4_lock_group(sb, grp);
688 return;
689 }
690
691 void ext4_update_dynamic_rev(struct super_block *sb)
692 {
693 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
694
695 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
696 return;
697
698 ext4_warning(sb,
699 "updating to rev %d because of new feature flag, "
700 "running e2fsck is recommended",
701 EXT4_DYNAMIC_REV);
702
703 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
704 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
705 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
706 /* leave es->s_feature_*compat flags alone */
707 /* es->s_uuid will be set by e2fsck if empty */
708
709 /*
710 * The rest of the superblock fields should be zero, and if not it
711 * means they are likely already in use, so leave them alone. We
712 * can leave it up to e2fsck to clean up any inconsistencies there.
713 */
714 }
715
716 /*
717 * Open the external journal device
718 */
719 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
720 {
721 struct block_device *bdev;
722 char b[BDEVNAME_SIZE];
723
724 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
725 if (IS_ERR(bdev))
726 goto fail;
727 return bdev;
728
729 fail:
730 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
731 __bdevname(dev, b), PTR_ERR(bdev));
732 return NULL;
733 }
734
735 /*
736 * Release the journal device
737 */
738 static void ext4_blkdev_put(struct block_device *bdev)
739 {
740 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
741 }
742
743 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
744 {
745 struct block_device *bdev;
746 bdev = sbi->journal_bdev;
747 if (bdev) {
748 ext4_blkdev_put(bdev);
749 sbi->journal_bdev = NULL;
750 }
751 }
752
753 static inline struct inode *orphan_list_entry(struct list_head *l)
754 {
755 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
756 }
757
758 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
759 {
760 struct list_head *l;
761
762 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
763 le32_to_cpu(sbi->s_es->s_last_orphan));
764
765 printk(KERN_ERR "sb_info orphan list:\n");
766 list_for_each(l, &sbi->s_orphan) {
767 struct inode *inode = orphan_list_entry(l);
768 printk(KERN_ERR " "
769 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
770 inode->i_sb->s_id, inode->i_ino, inode,
771 inode->i_mode, inode->i_nlink,
772 NEXT_ORPHAN(inode));
773 }
774 }
775
776 static void ext4_put_super(struct super_block *sb)
777 {
778 struct ext4_sb_info *sbi = EXT4_SB(sb);
779 struct ext4_super_block *es = sbi->s_es;
780 int i, err;
781
782 ext4_unregister_li_request(sb);
783 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
784
785 flush_workqueue(sbi->rsv_conversion_wq);
786 destroy_workqueue(sbi->rsv_conversion_wq);
787
788 if (sbi->s_journal) {
789 err = jbd2_journal_destroy(sbi->s_journal);
790 sbi->s_journal = NULL;
791 if (err < 0)
792 ext4_abort(sb, "Couldn't clean up the journal");
793 }
794
795 ext4_es_unregister_shrinker(sbi);
796 del_timer_sync(&sbi->s_err_report);
797 ext4_release_system_zone(sb);
798 ext4_mb_release(sb);
799 ext4_ext_release(sb);
800 ext4_xattr_put_super(sb);
801
802 if (!(sb->s_flags & MS_RDONLY)) {
803 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
804 es->s_state = cpu_to_le16(sbi->s_mount_state);
805 }
806 if (!(sb->s_flags & MS_RDONLY))
807 ext4_commit_super(sb, 1);
808
809 if (sbi->s_proc) {
810 remove_proc_entry("options", sbi->s_proc);
811 remove_proc_entry(sb->s_id, ext4_proc_root);
812 }
813 kobject_del(&sbi->s_kobj);
814
815 for (i = 0; i < sbi->s_gdb_count; i++)
816 brelse(sbi->s_group_desc[i]);
817 ext4_kvfree(sbi->s_group_desc);
818 ext4_kvfree(sbi->s_flex_groups);
819 percpu_counter_destroy(&sbi->s_freeclusters_counter);
820 percpu_counter_destroy(&sbi->s_freeinodes_counter);
821 percpu_counter_destroy(&sbi->s_dirs_counter);
822 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
823 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
824 brelse(sbi->s_sbh);
825 #ifdef CONFIG_QUOTA
826 for (i = 0; i < MAXQUOTAS; i++)
827 kfree(sbi->s_qf_names[i]);
828 #endif
829
830 /* Debugging code just in case the in-memory inode orphan list
831 * isn't empty. The on-disk one can be non-empty if we've
832 * detected an error and taken the fs readonly, but the
833 * in-memory list had better be clean by this point. */
834 if (!list_empty(&sbi->s_orphan))
835 dump_orphan_list(sb, sbi);
836 J_ASSERT(list_empty(&sbi->s_orphan));
837
838 invalidate_bdev(sb->s_bdev);
839 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
840 /*
841 * Invalidate the journal device's buffers. We don't want them
842 * floating about in memory - the physical journal device may
843 * hotswapped, and it breaks the `ro-after' testing code.
844 */
845 sync_blockdev(sbi->journal_bdev);
846 invalidate_bdev(sbi->journal_bdev);
847 ext4_blkdev_remove(sbi);
848 }
849 if (sbi->s_mb_cache) {
850 ext4_xattr_destroy_cache(sbi->s_mb_cache);
851 sbi->s_mb_cache = NULL;
852 }
853 if (sbi->s_mmp_tsk)
854 kthread_stop(sbi->s_mmp_tsk);
855 sb->s_fs_info = NULL;
856 /*
857 * Now that we are completely done shutting down the
858 * superblock, we need to actually destroy the kobject.
859 */
860 kobject_put(&sbi->s_kobj);
861 wait_for_completion(&sbi->s_kobj_unregister);
862 if (sbi->s_chksum_driver)
863 crypto_free_shash(sbi->s_chksum_driver);
864 kfree(sbi->s_blockgroup_lock);
865 kfree(sbi);
866 }
867
868 static struct kmem_cache *ext4_inode_cachep;
869
870 /*
871 * Called inside transaction, so use GFP_NOFS
872 */
873 static struct inode *ext4_alloc_inode(struct super_block *sb)
874 {
875 struct ext4_inode_info *ei;
876
877 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
878 if (!ei)
879 return NULL;
880
881 ei->vfs_inode.i_version = 1;
882 spin_lock_init(&ei->i_raw_lock);
883 INIT_LIST_HEAD(&ei->i_prealloc_list);
884 spin_lock_init(&ei->i_prealloc_lock);
885 ext4_es_init_tree(&ei->i_es_tree);
886 rwlock_init(&ei->i_es_lock);
887 INIT_LIST_HEAD(&ei->i_es_lru);
888 ei->i_es_lru_nr = 0;
889 ei->i_touch_when = 0;
890 ei->i_reserved_data_blocks = 0;
891 ei->i_reserved_meta_blocks = 0;
892 ei->i_allocated_meta_blocks = 0;
893 ei->i_da_metadata_calc_len = 0;
894 ei->i_da_metadata_calc_last_lblock = 0;
895 spin_lock_init(&(ei->i_block_reservation_lock));
896 #ifdef CONFIG_QUOTA
897 ei->i_reserved_quota = 0;
898 #endif
899 ei->jinode = NULL;
900 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
901 spin_lock_init(&ei->i_completed_io_lock);
902 ei->i_sync_tid = 0;
903 ei->i_datasync_tid = 0;
904 atomic_set(&ei->i_ioend_count, 0);
905 atomic_set(&ei->i_unwritten, 0);
906 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
907
908 return &ei->vfs_inode;
909 }
910
911 static int ext4_drop_inode(struct inode *inode)
912 {
913 int drop = generic_drop_inode(inode);
914
915 trace_ext4_drop_inode(inode, drop);
916 return drop;
917 }
918
919 static void ext4_i_callback(struct rcu_head *head)
920 {
921 struct inode *inode = container_of(head, struct inode, i_rcu);
922 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
923 }
924
925 static void ext4_destroy_inode(struct inode *inode)
926 {
927 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
928 ext4_msg(inode->i_sb, KERN_ERR,
929 "Inode %lu (%p): orphan list check failed!",
930 inode->i_ino, EXT4_I(inode));
931 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
932 EXT4_I(inode), sizeof(struct ext4_inode_info),
933 true);
934 dump_stack();
935 }
936 call_rcu(&inode->i_rcu, ext4_i_callback);
937 }
938
939 static void init_once(void *foo)
940 {
941 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
942
943 INIT_LIST_HEAD(&ei->i_orphan);
944 init_rwsem(&ei->xattr_sem);
945 init_rwsem(&ei->i_data_sem);
946 inode_init_once(&ei->vfs_inode);
947 }
948
949 static int __init init_inodecache(void)
950 {
951 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
952 sizeof(struct ext4_inode_info),
953 0, (SLAB_RECLAIM_ACCOUNT|
954 SLAB_MEM_SPREAD),
955 init_once);
956 if (ext4_inode_cachep == NULL)
957 return -ENOMEM;
958 return 0;
959 }
960
961 static void destroy_inodecache(void)
962 {
963 /*
964 * Make sure all delayed rcu free inodes are flushed before we
965 * destroy cache.
966 */
967 rcu_barrier();
968 kmem_cache_destroy(ext4_inode_cachep);
969 }
970
971 void ext4_clear_inode(struct inode *inode)
972 {
973 invalidate_inode_buffers(inode);
974 clear_inode(inode);
975 dquot_drop(inode);
976 ext4_discard_preallocations(inode);
977 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
978 ext4_es_lru_del(inode);
979 if (EXT4_I(inode)->jinode) {
980 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
981 EXT4_I(inode)->jinode);
982 jbd2_free_inode(EXT4_I(inode)->jinode);
983 EXT4_I(inode)->jinode = NULL;
984 }
985 }
986
987 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
988 u64 ino, u32 generation)
989 {
990 struct inode *inode;
991
992 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
993 return ERR_PTR(-ESTALE);
994 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
995 return ERR_PTR(-ESTALE);
996
997 /* iget isn't really right if the inode is currently unallocated!!
998 *
999 * ext4_read_inode will return a bad_inode if the inode had been
1000 * deleted, so we should be safe.
1001 *
1002 * Currently we don't know the generation for parent directory, so
1003 * a generation of 0 means "accept any"
1004 */
1005 inode = ext4_iget(sb, ino);
1006 if (IS_ERR(inode))
1007 return ERR_CAST(inode);
1008 if (generation && inode->i_generation != generation) {
1009 iput(inode);
1010 return ERR_PTR(-ESTALE);
1011 }
1012
1013 return inode;
1014 }
1015
1016 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1017 int fh_len, int fh_type)
1018 {
1019 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1020 ext4_nfs_get_inode);
1021 }
1022
1023 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1024 int fh_len, int fh_type)
1025 {
1026 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1027 ext4_nfs_get_inode);
1028 }
1029
1030 /*
1031 * Try to release metadata pages (indirect blocks, directories) which are
1032 * mapped via the block device. Since these pages could have journal heads
1033 * which would prevent try_to_free_buffers() from freeing them, we must use
1034 * jbd2 layer's try_to_free_buffers() function to release them.
1035 */
1036 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1037 gfp_t wait)
1038 {
1039 journal_t *journal = EXT4_SB(sb)->s_journal;
1040
1041 WARN_ON(PageChecked(page));
1042 if (!page_has_buffers(page))
1043 return 0;
1044 if (journal)
1045 return jbd2_journal_try_to_free_buffers(journal, page,
1046 wait & ~__GFP_WAIT);
1047 return try_to_free_buffers(page);
1048 }
1049
1050 #ifdef CONFIG_QUOTA
1051 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1052 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1053
1054 static int ext4_write_dquot(struct dquot *dquot);
1055 static int ext4_acquire_dquot(struct dquot *dquot);
1056 static int ext4_release_dquot(struct dquot *dquot);
1057 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1058 static int ext4_write_info(struct super_block *sb, int type);
1059 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1060 struct path *path);
1061 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1062 int format_id);
1063 static int ext4_quota_off(struct super_block *sb, int type);
1064 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1065 static int ext4_quota_on_mount(struct super_block *sb, int type);
1066 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1067 size_t len, loff_t off);
1068 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1069 const char *data, size_t len, loff_t off);
1070 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1071 unsigned int flags);
1072 static int ext4_enable_quotas(struct super_block *sb);
1073
1074 static const struct dquot_operations ext4_quota_operations = {
1075 .get_reserved_space = ext4_get_reserved_space,
1076 .write_dquot = ext4_write_dquot,
1077 .acquire_dquot = ext4_acquire_dquot,
1078 .release_dquot = ext4_release_dquot,
1079 .mark_dirty = ext4_mark_dquot_dirty,
1080 .write_info = ext4_write_info,
1081 .alloc_dquot = dquot_alloc,
1082 .destroy_dquot = dquot_destroy,
1083 };
1084
1085 static const struct quotactl_ops ext4_qctl_operations = {
1086 .quota_on = ext4_quota_on,
1087 .quota_off = ext4_quota_off,
1088 .quota_sync = dquot_quota_sync,
1089 .get_info = dquot_get_dqinfo,
1090 .set_info = dquot_set_dqinfo,
1091 .get_dqblk = dquot_get_dqblk,
1092 .set_dqblk = dquot_set_dqblk
1093 };
1094
1095 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1096 .quota_on_meta = ext4_quota_on_sysfile,
1097 .quota_off = ext4_quota_off_sysfile,
1098 .quota_sync = dquot_quota_sync,
1099 .get_info = dquot_get_dqinfo,
1100 .set_info = dquot_set_dqinfo,
1101 .get_dqblk = dquot_get_dqblk,
1102 .set_dqblk = dquot_set_dqblk
1103 };
1104 #endif
1105
1106 static const struct super_operations ext4_sops = {
1107 .alloc_inode = ext4_alloc_inode,
1108 .destroy_inode = ext4_destroy_inode,
1109 .write_inode = ext4_write_inode,
1110 .dirty_inode = ext4_dirty_inode,
1111 .drop_inode = ext4_drop_inode,
1112 .evict_inode = ext4_evict_inode,
1113 .put_super = ext4_put_super,
1114 .sync_fs = ext4_sync_fs,
1115 .freeze_fs = ext4_freeze,
1116 .unfreeze_fs = ext4_unfreeze,
1117 .statfs = ext4_statfs,
1118 .remount_fs = ext4_remount,
1119 .show_options = ext4_show_options,
1120 #ifdef CONFIG_QUOTA
1121 .quota_read = ext4_quota_read,
1122 .quota_write = ext4_quota_write,
1123 #endif
1124 .bdev_try_to_free_page = bdev_try_to_free_page,
1125 };
1126
1127 static const struct super_operations ext4_nojournal_sops = {
1128 .alloc_inode = ext4_alloc_inode,
1129 .destroy_inode = ext4_destroy_inode,
1130 .write_inode = ext4_write_inode,
1131 .dirty_inode = ext4_dirty_inode,
1132 .drop_inode = ext4_drop_inode,
1133 .evict_inode = ext4_evict_inode,
1134 .sync_fs = ext4_sync_fs_nojournal,
1135 .put_super = ext4_put_super,
1136 .statfs = ext4_statfs,
1137 .remount_fs = ext4_remount,
1138 .show_options = ext4_show_options,
1139 #ifdef CONFIG_QUOTA
1140 .quota_read = ext4_quota_read,
1141 .quota_write = ext4_quota_write,
1142 #endif
1143 .bdev_try_to_free_page = bdev_try_to_free_page,
1144 };
1145
1146 static const struct export_operations ext4_export_ops = {
1147 .fh_to_dentry = ext4_fh_to_dentry,
1148 .fh_to_parent = ext4_fh_to_parent,
1149 .get_parent = ext4_get_parent,
1150 };
1151
1152 enum {
1153 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1154 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1155 Opt_nouid32, Opt_debug, Opt_removed,
1156 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1157 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1158 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1159 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1160 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1161 Opt_data_err_abort, Opt_data_err_ignore,
1162 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1163 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1164 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1165 Opt_usrquota, Opt_grpquota, Opt_i_version,
1166 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1167 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1168 Opt_inode_readahead_blks, Opt_journal_ioprio,
1169 Opt_dioread_nolock, Opt_dioread_lock,
1170 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1171 Opt_max_dir_size_kb,
1172 };
1173
1174 static const match_table_t tokens = {
1175 {Opt_bsd_df, "bsddf"},
1176 {Opt_minix_df, "minixdf"},
1177 {Opt_grpid, "grpid"},
1178 {Opt_grpid, "bsdgroups"},
1179 {Opt_nogrpid, "nogrpid"},
1180 {Opt_nogrpid, "sysvgroups"},
1181 {Opt_resgid, "resgid=%u"},
1182 {Opt_resuid, "resuid=%u"},
1183 {Opt_sb, "sb=%u"},
1184 {Opt_err_cont, "errors=continue"},
1185 {Opt_err_panic, "errors=panic"},
1186 {Opt_err_ro, "errors=remount-ro"},
1187 {Opt_nouid32, "nouid32"},
1188 {Opt_debug, "debug"},
1189 {Opt_removed, "oldalloc"},
1190 {Opt_removed, "orlov"},
1191 {Opt_user_xattr, "user_xattr"},
1192 {Opt_nouser_xattr, "nouser_xattr"},
1193 {Opt_acl, "acl"},
1194 {Opt_noacl, "noacl"},
1195 {Opt_noload, "norecovery"},
1196 {Opt_noload, "noload"},
1197 {Opt_removed, "nobh"},
1198 {Opt_removed, "bh"},
1199 {Opt_commit, "commit=%u"},
1200 {Opt_min_batch_time, "min_batch_time=%u"},
1201 {Opt_max_batch_time, "max_batch_time=%u"},
1202 {Opt_journal_dev, "journal_dev=%u"},
1203 {Opt_journal_path, "journal_path=%s"},
1204 {Opt_journal_checksum, "journal_checksum"},
1205 {Opt_journal_async_commit, "journal_async_commit"},
1206 {Opt_abort, "abort"},
1207 {Opt_data_journal, "data=journal"},
1208 {Opt_data_ordered, "data=ordered"},
1209 {Opt_data_writeback, "data=writeback"},
1210 {Opt_data_err_abort, "data_err=abort"},
1211 {Opt_data_err_ignore, "data_err=ignore"},
1212 {Opt_offusrjquota, "usrjquota="},
1213 {Opt_usrjquota, "usrjquota=%s"},
1214 {Opt_offgrpjquota, "grpjquota="},
1215 {Opt_grpjquota, "grpjquota=%s"},
1216 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1217 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1218 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1219 {Opt_grpquota, "grpquota"},
1220 {Opt_noquota, "noquota"},
1221 {Opt_quota, "quota"},
1222 {Opt_usrquota, "usrquota"},
1223 {Opt_barrier, "barrier=%u"},
1224 {Opt_barrier, "barrier"},
1225 {Opt_nobarrier, "nobarrier"},
1226 {Opt_i_version, "i_version"},
1227 {Opt_stripe, "stripe=%u"},
1228 {Opt_delalloc, "delalloc"},
1229 {Opt_nodelalloc, "nodelalloc"},
1230 {Opt_removed, "mblk_io_submit"},
1231 {Opt_removed, "nomblk_io_submit"},
1232 {Opt_block_validity, "block_validity"},
1233 {Opt_noblock_validity, "noblock_validity"},
1234 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1235 {Opt_journal_ioprio, "journal_ioprio=%u"},
1236 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1237 {Opt_auto_da_alloc, "auto_da_alloc"},
1238 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1239 {Opt_dioread_nolock, "dioread_nolock"},
1240 {Opt_dioread_lock, "dioread_lock"},
1241 {Opt_discard, "discard"},
1242 {Opt_nodiscard, "nodiscard"},
1243 {Opt_init_itable, "init_itable=%u"},
1244 {Opt_init_itable, "init_itable"},
1245 {Opt_noinit_itable, "noinit_itable"},
1246 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1247 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1248 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1249 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1250 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1251 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1252 {Opt_err, NULL},
1253 };
1254
1255 static ext4_fsblk_t get_sb_block(void **data)
1256 {
1257 ext4_fsblk_t sb_block;
1258 char *options = (char *) *data;
1259
1260 if (!options || strncmp(options, "sb=", 3) != 0)
1261 return 1; /* Default location */
1262
1263 options += 3;
1264 /* TODO: use simple_strtoll with >32bit ext4 */
1265 sb_block = simple_strtoul(options, &options, 0);
1266 if (*options && *options != ',') {
1267 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1268 (char *) *data);
1269 return 1;
1270 }
1271 if (*options == ',')
1272 options++;
1273 *data = (void *) options;
1274
1275 return sb_block;
1276 }
1277
1278 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1279 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1280 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1281
1282 #ifdef CONFIG_QUOTA
1283 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1284 {
1285 struct ext4_sb_info *sbi = EXT4_SB(sb);
1286 char *qname;
1287 int ret = -1;
1288
1289 if (sb_any_quota_loaded(sb) &&
1290 !sbi->s_qf_names[qtype]) {
1291 ext4_msg(sb, KERN_ERR,
1292 "Cannot change journaled "
1293 "quota options when quota turned on");
1294 return -1;
1295 }
1296 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1297 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1298 "when QUOTA feature is enabled");
1299 return -1;
1300 }
1301 qname = match_strdup(args);
1302 if (!qname) {
1303 ext4_msg(sb, KERN_ERR,
1304 "Not enough memory for storing quotafile name");
1305 return -1;
1306 }
1307 if (sbi->s_qf_names[qtype]) {
1308 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1309 ret = 1;
1310 else
1311 ext4_msg(sb, KERN_ERR,
1312 "%s quota file already specified",
1313 QTYPE2NAME(qtype));
1314 goto errout;
1315 }
1316 if (strchr(qname, '/')) {
1317 ext4_msg(sb, KERN_ERR,
1318 "quotafile must be on filesystem root");
1319 goto errout;
1320 }
1321 sbi->s_qf_names[qtype] = qname;
1322 set_opt(sb, QUOTA);
1323 return 1;
1324 errout:
1325 kfree(qname);
1326 return ret;
1327 }
1328
1329 static int clear_qf_name(struct super_block *sb, int qtype)
1330 {
1331
1332 struct ext4_sb_info *sbi = EXT4_SB(sb);
1333
1334 if (sb_any_quota_loaded(sb) &&
1335 sbi->s_qf_names[qtype]) {
1336 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1337 " when quota turned on");
1338 return -1;
1339 }
1340 kfree(sbi->s_qf_names[qtype]);
1341 sbi->s_qf_names[qtype] = NULL;
1342 return 1;
1343 }
1344 #endif
1345
1346 #define MOPT_SET 0x0001
1347 #define MOPT_CLEAR 0x0002
1348 #define MOPT_NOSUPPORT 0x0004
1349 #define MOPT_EXPLICIT 0x0008
1350 #define MOPT_CLEAR_ERR 0x0010
1351 #define MOPT_GTE0 0x0020
1352 #ifdef CONFIG_QUOTA
1353 #define MOPT_Q 0
1354 #define MOPT_QFMT 0x0040
1355 #else
1356 #define MOPT_Q MOPT_NOSUPPORT
1357 #define MOPT_QFMT MOPT_NOSUPPORT
1358 #endif
1359 #define MOPT_DATAJ 0x0080
1360 #define MOPT_NO_EXT2 0x0100
1361 #define MOPT_NO_EXT3 0x0200
1362 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1363 #define MOPT_STRING 0x0400
1364
1365 static const struct mount_opts {
1366 int token;
1367 int mount_opt;
1368 int flags;
1369 } ext4_mount_opts[] = {
1370 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1371 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1372 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1373 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1374 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1375 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1376 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1377 MOPT_EXT4_ONLY | MOPT_SET},
1378 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1379 MOPT_EXT4_ONLY | MOPT_CLEAR},
1380 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1381 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1382 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1383 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1384 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1385 MOPT_EXT4_ONLY | MOPT_CLEAR},
1386 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1387 MOPT_EXT4_ONLY | MOPT_SET},
1388 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1389 EXT4_MOUNT_JOURNAL_CHECKSUM),
1390 MOPT_EXT4_ONLY | MOPT_SET},
1391 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1392 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1393 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1395 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1396 MOPT_NO_EXT2 | MOPT_SET},
1397 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1398 MOPT_NO_EXT2 | MOPT_CLEAR},
1399 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1400 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1401 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1402 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1403 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1404 {Opt_commit, 0, MOPT_GTE0},
1405 {Opt_max_batch_time, 0, MOPT_GTE0},
1406 {Opt_min_batch_time, 0, MOPT_GTE0},
1407 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1408 {Opt_init_itable, 0, MOPT_GTE0},
1409 {Opt_stripe, 0, MOPT_GTE0},
1410 {Opt_resuid, 0, MOPT_GTE0},
1411 {Opt_resgid, 0, MOPT_GTE0},
1412 {Opt_journal_dev, 0, MOPT_GTE0},
1413 {Opt_journal_path, 0, MOPT_STRING},
1414 {Opt_journal_ioprio, 0, MOPT_GTE0},
1415 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1416 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1417 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1418 MOPT_NO_EXT2 | MOPT_DATAJ},
1419 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1420 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1421 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1422 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1423 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1424 #else
1425 {Opt_acl, 0, MOPT_NOSUPPORT},
1426 {Opt_noacl, 0, MOPT_NOSUPPORT},
1427 #endif
1428 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1429 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1430 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1431 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1432 MOPT_SET | MOPT_Q},
1433 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1434 MOPT_SET | MOPT_Q},
1435 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1436 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1437 {Opt_usrjquota, 0, MOPT_Q},
1438 {Opt_grpjquota, 0, MOPT_Q},
1439 {Opt_offusrjquota, 0, MOPT_Q},
1440 {Opt_offgrpjquota, 0, MOPT_Q},
1441 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1442 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1443 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1444 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1445 {Opt_err, 0, 0}
1446 };
1447
1448 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1449 substring_t *args, unsigned long *journal_devnum,
1450 unsigned int *journal_ioprio, int is_remount)
1451 {
1452 struct ext4_sb_info *sbi = EXT4_SB(sb);
1453 const struct mount_opts *m;
1454 kuid_t uid;
1455 kgid_t gid;
1456 int arg = 0;
1457
1458 #ifdef CONFIG_QUOTA
1459 if (token == Opt_usrjquota)
1460 return set_qf_name(sb, USRQUOTA, &args[0]);
1461 else if (token == Opt_grpjquota)
1462 return set_qf_name(sb, GRPQUOTA, &args[0]);
1463 else if (token == Opt_offusrjquota)
1464 return clear_qf_name(sb, USRQUOTA);
1465 else if (token == Opt_offgrpjquota)
1466 return clear_qf_name(sb, GRPQUOTA);
1467 #endif
1468 switch (token) {
1469 case Opt_noacl:
1470 case Opt_nouser_xattr:
1471 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1472 break;
1473 case Opt_sb:
1474 return 1; /* handled by get_sb_block() */
1475 case Opt_removed:
1476 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1477 return 1;
1478 case Opt_abort:
1479 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1480 return 1;
1481 case Opt_i_version:
1482 sb->s_flags |= MS_I_VERSION;
1483 return 1;
1484 }
1485
1486 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1487 if (token == m->token)
1488 break;
1489
1490 if (m->token == Opt_err) {
1491 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1492 "or missing value", opt);
1493 return -1;
1494 }
1495
1496 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1497 ext4_msg(sb, KERN_ERR,
1498 "Mount option \"%s\" incompatible with ext2", opt);
1499 return -1;
1500 }
1501 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1502 ext4_msg(sb, KERN_ERR,
1503 "Mount option \"%s\" incompatible with ext3", opt);
1504 return -1;
1505 }
1506
1507 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1508 return -1;
1509 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1510 return -1;
1511 if (m->flags & MOPT_EXPLICIT)
1512 set_opt2(sb, EXPLICIT_DELALLOC);
1513 if (m->flags & MOPT_CLEAR_ERR)
1514 clear_opt(sb, ERRORS_MASK);
1515 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1516 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1517 "options when quota turned on");
1518 return -1;
1519 }
1520
1521 if (m->flags & MOPT_NOSUPPORT) {
1522 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1523 } else if (token == Opt_commit) {
1524 if (arg == 0)
1525 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1526 sbi->s_commit_interval = HZ * arg;
1527 } else if (token == Opt_max_batch_time) {
1528 sbi->s_max_batch_time = arg;
1529 } else if (token == Opt_min_batch_time) {
1530 sbi->s_min_batch_time = arg;
1531 } else if (token == Opt_inode_readahead_blks) {
1532 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1533 ext4_msg(sb, KERN_ERR,
1534 "EXT4-fs: inode_readahead_blks must be "
1535 "0 or a power of 2 smaller than 2^31");
1536 return -1;
1537 }
1538 sbi->s_inode_readahead_blks = arg;
1539 } else if (token == Opt_init_itable) {
1540 set_opt(sb, INIT_INODE_TABLE);
1541 if (!args->from)
1542 arg = EXT4_DEF_LI_WAIT_MULT;
1543 sbi->s_li_wait_mult = arg;
1544 } else if (token == Opt_max_dir_size_kb) {
1545 sbi->s_max_dir_size_kb = arg;
1546 } else if (token == Opt_stripe) {
1547 sbi->s_stripe = arg;
1548 } else if (token == Opt_resuid) {
1549 uid = make_kuid(current_user_ns(), arg);
1550 if (!uid_valid(uid)) {
1551 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1552 return -1;
1553 }
1554 sbi->s_resuid = uid;
1555 } else if (token == Opt_resgid) {
1556 gid = make_kgid(current_user_ns(), arg);
1557 if (!gid_valid(gid)) {
1558 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1559 return -1;
1560 }
1561 sbi->s_resgid = gid;
1562 } else if (token == Opt_journal_dev) {
1563 if (is_remount) {
1564 ext4_msg(sb, KERN_ERR,
1565 "Cannot specify journal on remount");
1566 return -1;
1567 }
1568 *journal_devnum = arg;
1569 } else if (token == Opt_journal_path) {
1570 char *journal_path;
1571 struct inode *journal_inode;
1572 struct path path;
1573 int error;
1574
1575 if (is_remount) {
1576 ext4_msg(sb, KERN_ERR,
1577 "Cannot specify journal on remount");
1578 return -1;
1579 }
1580 journal_path = match_strdup(&args[0]);
1581 if (!journal_path) {
1582 ext4_msg(sb, KERN_ERR, "error: could not dup "
1583 "journal device string");
1584 return -1;
1585 }
1586
1587 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1588 if (error) {
1589 ext4_msg(sb, KERN_ERR, "error: could not find "
1590 "journal device path: error %d", error);
1591 kfree(journal_path);
1592 return -1;
1593 }
1594
1595 journal_inode = path.dentry->d_inode;
1596 if (!S_ISBLK(journal_inode->i_mode)) {
1597 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1598 "is not a block device", journal_path);
1599 path_put(&path);
1600 kfree(journal_path);
1601 return -1;
1602 }
1603
1604 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1605 path_put(&path);
1606 kfree(journal_path);
1607 } else if (token == Opt_journal_ioprio) {
1608 if (arg > 7) {
1609 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1610 " (must be 0-7)");
1611 return -1;
1612 }
1613 *journal_ioprio =
1614 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1615 } else if (m->flags & MOPT_DATAJ) {
1616 if (is_remount) {
1617 if (!sbi->s_journal)
1618 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1619 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1620 ext4_msg(sb, KERN_ERR,
1621 "Cannot change data mode on remount");
1622 return -1;
1623 }
1624 } else {
1625 clear_opt(sb, DATA_FLAGS);
1626 sbi->s_mount_opt |= m->mount_opt;
1627 }
1628 #ifdef CONFIG_QUOTA
1629 } else if (m->flags & MOPT_QFMT) {
1630 if (sb_any_quota_loaded(sb) &&
1631 sbi->s_jquota_fmt != m->mount_opt) {
1632 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1633 "quota options when quota turned on");
1634 return -1;
1635 }
1636 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1637 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1638 ext4_msg(sb, KERN_ERR,
1639 "Cannot set journaled quota options "
1640 "when QUOTA feature is enabled");
1641 return -1;
1642 }
1643 sbi->s_jquota_fmt = m->mount_opt;
1644 #endif
1645 } else {
1646 if (!args->from)
1647 arg = 1;
1648 if (m->flags & MOPT_CLEAR)
1649 arg = !arg;
1650 else if (unlikely(!(m->flags & MOPT_SET))) {
1651 ext4_msg(sb, KERN_WARNING,
1652 "buggy handling of option %s", opt);
1653 WARN_ON(1);
1654 return -1;
1655 }
1656 if (arg != 0)
1657 sbi->s_mount_opt |= m->mount_opt;
1658 else
1659 sbi->s_mount_opt &= ~m->mount_opt;
1660 }
1661 return 1;
1662 }
1663
1664 static int parse_options(char *options, struct super_block *sb,
1665 unsigned long *journal_devnum,
1666 unsigned int *journal_ioprio,
1667 int is_remount)
1668 {
1669 struct ext4_sb_info *sbi = EXT4_SB(sb);
1670 char *p;
1671 substring_t args[MAX_OPT_ARGS];
1672 int token;
1673
1674 if (!options)
1675 return 1;
1676
1677 while ((p = strsep(&options, ",")) != NULL) {
1678 if (!*p)
1679 continue;
1680 /*
1681 * Initialize args struct so we know whether arg was
1682 * found; some options take optional arguments.
1683 */
1684 args[0].to = args[0].from = NULL;
1685 token = match_token(p, tokens, args);
1686 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1687 journal_ioprio, is_remount) < 0)
1688 return 0;
1689 }
1690 #ifdef CONFIG_QUOTA
1691 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1692 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1693 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1694 "feature is enabled");
1695 return 0;
1696 }
1697 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1698 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1699 clear_opt(sb, USRQUOTA);
1700
1701 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1702 clear_opt(sb, GRPQUOTA);
1703
1704 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1705 ext4_msg(sb, KERN_ERR, "old and new quota "
1706 "format mixing");
1707 return 0;
1708 }
1709
1710 if (!sbi->s_jquota_fmt) {
1711 ext4_msg(sb, KERN_ERR, "journaled quota format "
1712 "not specified");
1713 return 0;
1714 }
1715 } else {
1716 if (sbi->s_jquota_fmt) {
1717 ext4_msg(sb, KERN_ERR, "journaled quota format "
1718 "specified with no journaling "
1719 "enabled");
1720 return 0;
1721 }
1722 }
1723 #endif
1724 if (test_opt(sb, DIOREAD_NOLOCK)) {
1725 int blocksize =
1726 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1727
1728 if (blocksize < PAGE_CACHE_SIZE) {
1729 ext4_msg(sb, KERN_ERR, "can't mount with "
1730 "dioread_nolock if block size != PAGE_SIZE");
1731 return 0;
1732 }
1733 }
1734 return 1;
1735 }
1736
1737 static inline void ext4_show_quota_options(struct seq_file *seq,
1738 struct super_block *sb)
1739 {
1740 #if defined(CONFIG_QUOTA)
1741 struct ext4_sb_info *sbi = EXT4_SB(sb);
1742
1743 if (sbi->s_jquota_fmt) {
1744 char *fmtname = "";
1745
1746 switch (sbi->s_jquota_fmt) {
1747 case QFMT_VFS_OLD:
1748 fmtname = "vfsold";
1749 break;
1750 case QFMT_VFS_V0:
1751 fmtname = "vfsv0";
1752 break;
1753 case QFMT_VFS_V1:
1754 fmtname = "vfsv1";
1755 break;
1756 }
1757 seq_printf(seq, ",jqfmt=%s", fmtname);
1758 }
1759
1760 if (sbi->s_qf_names[USRQUOTA])
1761 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1762
1763 if (sbi->s_qf_names[GRPQUOTA])
1764 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1765 #endif
1766 }
1767
1768 static const char *token2str(int token)
1769 {
1770 const struct match_token *t;
1771
1772 for (t = tokens; t->token != Opt_err; t++)
1773 if (t->token == token && !strchr(t->pattern, '='))
1774 break;
1775 return t->pattern;
1776 }
1777
1778 /*
1779 * Show an option if
1780 * - it's set to a non-default value OR
1781 * - if the per-sb default is different from the global default
1782 */
1783 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1784 int nodefs)
1785 {
1786 struct ext4_sb_info *sbi = EXT4_SB(sb);
1787 struct ext4_super_block *es = sbi->s_es;
1788 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1789 const struct mount_opts *m;
1790 char sep = nodefs ? '\n' : ',';
1791
1792 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1793 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1794
1795 if (sbi->s_sb_block != 1)
1796 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1797
1798 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1799 int want_set = m->flags & MOPT_SET;
1800 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1801 (m->flags & MOPT_CLEAR_ERR))
1802 continue;
1803 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1804 continue; /* skip if same as the default */
1805 if ((want_set &&
1806 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1807 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1808 continue; /* select Opt_noFoo vs Opt_Foo */
1809 SEQ_OPTS_PRINT("%s", token2str(m->token));
1810 }
1811
1812 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1813 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1814 SEQ_OPTS_PRINT("resuid=%u",
1815 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1816 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1817 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1818 SEQ_OPTS_PRINT("resgid=%u",
1819 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1820 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1821 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1822 SEQ_OPTS_PUTS("errors=remount-ro");
1823 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1824 SEQ_OPTS_PUTS("errors=continue");
1825 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1826 SEQ_OPTS_PUTS("errors=panic");
1827 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1828 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1829 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1830 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1831 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1832 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1833 if (sb->s_flags & MS_I_VERSION)
1834 SEQ_OPTS_PUTS("i_version");
1835 if (nodefs || sbi->s_stripe)
1836 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1837 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1838 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1839 SEQ_OPTS_PUTS("data=journal");
1840 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1841 SEQ_OPTS_PUTS("data=ordered");
1842 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1843 SEQ_OPTS_PUTS("data=writeback");
1844 }
1845 if (nodefs ||
1846 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1847 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1848 sbi->s_inode_readahead_blks);
1849
1850 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1851 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1852 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1853 if (nodefs || sbi->s_max_dir_size_kb)
1854 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1855
1856 ext4_show_quota_options(seq, sb);
1857 return 0;
1858 }
1859
1860 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1861 {
1862 return _ext4_show_options(seq, root->d_sb, 0);
1863 }
1864
1865 static int options_seq_show(struct seq_file *seq, void *offset)
1866 {
1867 struct super_block *sb = seq->private;
1868 int rc;
1869
1870 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1871 rc = _ext4_show_options(seq, sb, 1);
1872 seq_puts(seq, "\n");
1873 return rc;
1874 }
1875
1876 static int options_open_fs(struct inode *inode, struct file *file)
1877 {
1878 return single_open(file, options_seq_show, PDE_DATA(inode));
1879 }
1880
1881 static const struct file_operations ext4_seq_options_fops = {
1882 .owner = THIS_MODULE,
1883 .open = options_open_fs,
1884 .read = seq_read,
1885 .llseek = seq_lseek,
1886 .release = single_release,
1887 };
1888
1889 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1890 int read_only)
1891 {
1892 struct ext4_sb_info *sbi = EXT4_SB(sb);
1893 int res = 0;
1894
1895 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1896 ext4_msg(sb, KERN_ERR, "revision level too high, "
1897 "forcing read-only mode");
1898 res = MS_RDONLY;
1899 }
1900 if (read_only)
1901 goto done;
1902 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1903 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1904 "running e2fsck is recommended");
1905 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1906 ext4_msg(sb, KERN_WARNING,
1907 "warning: mounting fs with errors, "
1908 "running e2fsck is recommended");
1909 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1910 le16_to_cpu(es->s_mnt_count) >=
1911 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1912 ext4_msg(sb, KERN_WARNING,
1913 "warning: maximal mount count reached, "
1914 "running e2fsck is recommended");
1915 else if (le32_to_cpu(es->s_checkinterval) &&
1916 (le32_to_cpu(es->s_lastcheck) +
1917 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1918 ext4_msg(sb, KERN_WARNING,
1919 "warning: checktime reached, "
1920 "running e2fsck is recommended");
1921 if (!sbi->s_journal)
1922 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1923 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1924 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1925 le16_add_cpu(&es->s_mnt_count, 1);
1926 es->s_mtime = cpu_to_le32(get_seconds());
1927 ext4_update_dynamic_rev(sb);
1928 if (sbi->s_journal)
1929 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1930
1931 ext4_commit_super(sb, 1);
1932 done:
1933 if (test_opt(sb, DEBUG))
1934 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1935 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1936 sb->s_blocksize,
1937 sbi->s_groups_count,
1938 EXT4_BLOCKS_PER_GROUP(sb),
1939 EXT4_INODES_PER_GROUP(sb),
1940 sbi->s_mount_opt, sbi->s_mount_opt2);
1941
1942 cleancache_init_fs(sb);
1943 return res;
1944 }
1945
1946 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1947 {
1948 struct ext4_sb_info *sbi = EXT4_SB(sb);
1949 struct flex_groups *new_groups;
1950 int size;
1951
1952 if (!sbi->s_log_groups_per_flex)
1953 return 0;
1954
1955 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1956 if (size <= sbi->s_flex_groups_allocated)
1957 return 0;
1958
1959 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1960 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1961 if (!new_groups) {
1962 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1963 size / (int) sizeof(struct flex_groups));
1964 return -ENOMEM;
1965 }
1966
1967 if (sbi->s_flex_groups) {
1968 memcpy(new_groups, sbi->s_flex_groups,
1969 (sbi->s_flex_groups_allocated *
1970 sizeof(struct flex_groups)));
1971 ext4_kvfree(sbi->s_flex_groups);
1972 }
1973 sbi->s_flex_groups = new_groups;
1974 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1975 return 0;
1976 }
1977
1978 static int ext4_fill_flex_info(struct super_block *sb)
1979 {
1980 struct ext4_sb_info *sbi = EXT4_SB(sb);
1981 struct ext4_group_desc *gdp = NULL;
1982 ext4_group_t flex_group;
1983 int i, err;
1984
1985 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1986 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1987 sbi->s_log_groups_per_flex = 0;
1988 return 1;
1989 }
1990
1991 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1992 if (err)
1993 goto failed;
1994
1995 for (i = 0; i < sbi->s_groups_count; i++) {
1996 gdp = ext4_get_group_desc(sb, i, NULL);
1997
1998 flex_group = ext4_flex_group(sbi, i);
1999 atomic_add(ext4_free_inodes_count(sb, gdp),
2000 &sbi->s_flex_groups[flex_group].free_inodes);
2001 atomic64_add(ext4_free_group_clusters(sb, gdp),
2002 &sbi->s_flex_groups[flex_group].free_clusters);
2003 atomic_add(ext4_used_dirs_count(sb, gdp),
2004 &sbi->s_flex_groups[flex_group].used_dirs);
2005 }
2006
2007 return 1;
2008 failed:
2009 return 0;
2010 }
2011
2012 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2013 struct ext4_group_desc *gdp)
2014 {
2015 int offset;
2016 __u16 crc = 0;
2017 __le32 le_group = cpu_to_le32(block_group);
2018
2019 if ((sbi->s_es->s_feature_ro_compat &
2020 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2021 /* Use new metadata_csum algorithm */
2022 __le16 save_csum;
2023 __u32 csum32;
2024
2025 save_csum = gdp->bg_checksum;
2026 gdp->bg_checksum = 0;
2027 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2028 sizeof(le_group));
2029 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2030 sbi->s_desc_size);
2031 gdp->bg_checksum = save_csum;
2032
2033 crc = csum32 & 0xFFFF;
2034 goto out;
2035 }
2036
2037 /* old crc16 code */
2038 offset = offsetof(struct ext4_group_desc, bg_checksum);
2039
2040 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2041 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2042 crc = crc16(crc, (__u8 *)gdp, offset);
2043 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2044 /* for checksum of struct ext4_group_desc do the rest...*/
2045 if ((sbi->s_es->s_feature_incompat &
2046 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2047 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2048 crc = crc16(crc, (__u8 *)gdp + offset,
2049 le16_to_cpu(sbi->s_es->s_desc_size) -
2050 offset);
2051
2052 out:
2053 return cpu_to_le16(crc);
2054 }
2055
2056 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2057 struct ext4_group_desc *gdp)
2058 {
2059 if (ext4_has_group_desc_csum(sb) &&
2060 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2061 block_group, gdp)))
2062 return 0;
2063
2064 return 1;
2065 }
2066
2067 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2068 struct ext4_group_desc *gdp)
2069 {
2070 if (!ext4_has_group_desc_csum(sb))
2071 return;
2072 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2073 }
2074
2075 /* Called at mount-time, super-block is locked */
2076 static int ext4_check_descriptors(struct super_block *sb,
2077 ext4_group_t *first_not_zeroed)
2078 {
2079 struct ext4_sb_info *sbi = EXT4_SB(sb);
2080 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2081 ext4_fsblk_t last_block;
2082 ext4_fsblk_t block_bitmap;
2083 ext4_fsblk_t inode_bitmap;
2084 ext4_fsblk_t inode_table;
2085 int flexbg_flag = 0;
2086 ext4_group_t i, grp = sbi->s_groups_count;
2087
2088 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2089 flexbg_flag = 1;
2090
2091 ext4_debug("Checking group descriptors");
2092
2093 for (i = 0; i < sbi->s_groups_count; i++) {
2094 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2095
2096 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2097 last_block = ext4_blocks_count(sbi->s_es) - 1;
2098 else
2099 last_block = first_block +
2100 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2101
2102 if ((grp == sbi->s_groups_count) &&
2103 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2104 grp = i;
2105
2106 block_bitmap = ext4_block_bitmap(sb, gdp);
2107 if (block_bitmap < first_block || block_bitmap > last_block) {
2108 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2109 "Block bitmap for group %u not in group "
2110 "(block %llu)!", i, block_bitmap);
2111 return 0;
2112 }
2113 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2114 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2115 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2116 "Inode bitmap for group %u not in group "
2117 "(block %llu)!", i, inode_bitmap);
2118 return 0;
2119 }
2120 inode_table = ext4_inode_table(sb, gdp);
2121 if (inode_table < first_block ||
2122 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2123 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2124 "Inode table for group %u not in group "
2125 "(block %llu)!", i, inode_table);
2126 return 0;
2127 }
2128 ext4_lock_group(sb, i);
2129 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2130 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2131 "Checksum for group %u failed (%u!=%u)",
2132 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2133 gdp)), le16_to_cpu(gdp->bg_checksum));
2134 if (!(sb->s_flags & MS_RDONLY)) {
2135 ext4_unlock_group(sb, i);
2136 return 0;
2137 }
2138 }
2139 ext4_unlock_group(sb, i);
2140 if (!flexbg_flag)
2141 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2142 }
2143 if (NULL != first_not_zeroed)
2144 *first_not_zeroed = grp;
2145 return 1;
2146 }
2147
2148 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2149 * the superblock) which were deleted from all directories, but held open by
2150 * a process at the time of a crash. We walk the list and try to delete these
2151 * inodes at recovery time (only with a read-write filesystem).
2152 *
2153 * In order to keep the orphan inode chain consistent during traversal (in
2154 * case of crash during recovery), we link each inode into the superblock
2155 * orphan list_head and handle it the same way as an inode deletion during
2156 * normal operation (which journals the operations for us).
2157 *
2158 * We only do an iget() and an iput() on each inode, which is very safe if we
2159 * accidentally point at an in-use or already deleted inode. The worst that
2160 * can happen in this case is that we get a "bit already cleared" message from
2161 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2162 * e2fsck was run on this filesystem, and it must have already done the orphan
2163 * inode cleanup for us, so we can safely abort without any further action.
2164 */
2165 static void ext4_orphan_cleanup(struct super_block *sb,
2166 struct ext4_super_block *es)
2167 {
2168 unsigned int s_flags = sb->s_flags;
2169 int nr_orphans = 0, nr_truncates = 0;
2170 #ifdef CONFIG_QUOTA
2171 int i;
2172 #endif
2173 if (!es->s_last_orphan) {
2174 jbd_debug(4, "no orphan inodes to clean up\n");
2175 return;
2176 }
2177
2178 if (bdev_read_only(sb->s_bdev)) {
2179 ext4_msg(sb, KERN_ERR, "write access "
2180 "unavailable, skipping orphan cleanup");
2181 return;
2182 }
2183
2184 /* Check if feature set would not allow a r/w mount */
2185 if (!ext4_feature_set_ok(sb, 0)) {
2186 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2187 "unknown ROCOMPAT features");
2188 return;
2189 }
2190
2191 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2192 /* don't clear list on RO mount w/ errors */
2193 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2194 jbd_debug(1, "Errors on filesystem, "
2195 "clearing orphan list.\n");
2196 es->s_last_orphan = 0;
2197 }
2198 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2199 return;
2200 }
2201
2202 if (s_flags & MS_RDONLY) {
2203 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2204 sb->s_flags &= ~MS_RDONLY;
2205 }
2206 #ifdef CONFIG_QUOTA
2207 /* Needed for iput() to work correctly and not trash data */
2208 sb->s_flags |= MS_ACTIVE;
2209 /* Turn on quotas so that they are updated correctly */
2210 for (i = 0; i < MAXQUOTAS; i++) {
2211 if (EXT4_SB(sb)->s_qf_names[i]) {
2212 int ret = ext4_quota_on_mount(sb, i);
2213 if (ret < 0)
2214 ext4_msg(sb, KERN_ERR,
2215 "Cannot turn on journaled "
2216 "quota: error %d", ret);
2217 }
2218 }
2219 #endif
2220
2221 while (es->s_last_orphan) {
2222 struct inode *inode;
2223
2224 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2225 if (IS_ERR(inode)) {
2226 es->s_last_orphan = 0;
2227 break;
2228 }
2229
2230 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2231 dquot_initialize(inode);
2232 if (inode->i_nlink) {
2233 if (test_opt(sb, DEBUG))
2234 ext4_msg(sb, KERN_DEBUG,
2235 "%s: truncating inode %lu to %lld bytes",
2236 __func__, inode->i_ino, inode->i_size);
2237 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2238 inode->i_ino, inode->i_size);
2239 mutex_lock(&inode->i_mutex);
2240 truncate_inode_pages(inode->i_mapping, inode->i_size);
2241 ext4_truncate(inode);
2242 mutex_unlock(&inode->i_mutex);
2243 nr_truncates++;
2244 } else {
2245 if (test_opt(sb, DEBUG))
2246 ext4_msg(sb, KERN_DEBUG,
2247 "%s: deleting unreferenced inode %lu",
2248 __func__, inode->i_ino);
2249 jbd_debug(2, "deleting unreferenced inode %lu\n",
2250 inode->i_ino);
2251 nr_orphans++;
2252 }
2253 iput(inode); /* The delete magic happens here! */
2254 }
2255
2256 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2257
2258 if (nr_orphans)
2259 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2260 PLURAL(nr_orphans));
2261 if (nr_truncates)
2262 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2263 PLURAL(nr_truncates));
2264 #ifdef CONFIG_QUOTA
2265 /* Turn quotas off */
2266 for (i = 0; i < MAXQUOTAS; i++) {
2267 if (sb_dqopt(sb)->files[i])
2268 dquot_quota_off(sb, i);
2269 }
2270 #endif
2271 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2272 }
2273
2274 /*
2275 * Maximal extent format file size.
2276 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2277 * extent format containers, within a sector_t, and within i_blocks
2278 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2279 * so that won't be a limiting factor.
2280 *
2281 * However there is other limiting factor. We do store extents in the form
2282 * of starting block and length, hence the resulting length of the extent
2283 * covering maximum file size must fit into on-disk format containers as
2284 * well. Given that length is always by 1 unit bigger than max unit (because
2285 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2286 *
2287 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2288 */
2289 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2290 {
2291 loff_t res;
2292 loff_t upper_limit = MAX_LFS_FILESIZE;
2293
2294 /* small i_blocks in vfs inode? */
2295 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2296 /*
2297 * CONFIG_LBDAF is not enabled implies the inode
2298 * i_block represent total blocks in 512 bytes
2299 * 32 == size of vfs inode i_blocks * 8
2300 */
2301 upper_limit = (1LL << 32) - 1;
2302
2303 /* total blocks in file system block size */
2304 upper_limit >>= (blkbits - 9);
2305 upper_limit <<= blkbits;
2306 }
2307
2308 /*
2309 * 32-bit extent-start container, ee_block. We lower the maxbytes
2310 * by one fs block, so ee_len can cover the extent of maximum file
2311 * size
2312 */
2313 res = (1LL << 32) - 1;
2314 res <<= blkbits;
2315
2316 /* Sanity check against vm- & vfs- imposed limits */
2317 if (res > upper_limit)
2318 res = upper_limit;
2319
2320 return res;
2321 }
2322
2323 /*
2324 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2325 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2326 * We need to be 1 filesystem block less than the 2^48 sector limit.
2327 */
2328 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2329 {
2330 loff_t res = EXT4_NDIR_BLOCKS;
2331 int meta_blocks;
2332 loff_t upper_limit;
2333 /* This is calculated to be the largest file size for a dense, block
2334 * mapped file such that the file's total number of 512-byte sectors,
2335 * including data and all indirect blocks, does not exceed (2^48 - 1).
2336 *
2337 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2338 * number of 512-byte sectors of the file.
2339 */
2340
2341 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2342 /*
2343 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2344 * the inode i_block field represents total file blocks in
2345 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2346 */
2347 upper_limit = (1LL << 32) - 1;
2348
2349 /* total blocks in file system block size */
2350 upper_limit >>= (bits - 9);
2351
2352 } else {
2353 /*
2354 * We use 48 bit ext4_inode i_blocks
2355 * With EXT4_HUGE_FILE_FL set the i_blocks
2356 * represent total number of blocks in
2357 * file system block size
2358 */
2359 upper_limit = (1LL << 48) - 1;
2360
2361 }
2362
2363 /* indirect blocks */
2364 meta_blocks = 1;
2365 /* double indirect blocks */
2366 meta_blocks += 1 + (1LL << (bits-2));
2367 /* tripple indirect blocks */
2368 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2369
2370 upper_limit -= meta_blocks;
2371 upper_limit <<= bits;
2372
2373 res += 1LL << (bits-2);
2374 res += 1LL << (2*(bits-2));
2375 res += 1LL << (3*(bits-2));
2376 res <<= bits;
2377 if (res > upper_limit)
2378 res = upper_limit;
2379
2380 if (res > MAX_LFS_FILESIZE)
2381 res = MAX_LFS_FILESIZE;
2382
2383 return res;
2384 }
2385
2386 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2387 ext4_fsblk_t logical_sb_block, int nr)
2388 {
2389 struct ext4_sb_info *sbi = EXT4_SB(sb);
2390 ext4_group_t bg, first_meta_bg;
2391 int has_super = 0;
2392
2393 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2394
2395 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2396 nr < first_meta_bg)
2397 return logical_sb_block + nr + 1;
2398 bg = sbi->s_desc_per_block * nr;
2399 if (ext4_bg_has_super(sb, bg))
2400 has_super = 1;
2401
2402 /*
2403 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2404 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2405 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2406 * compensate.
2407 */
2408 if (sb->s_blocksize == 1024 && nr == 0 &&
2409 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2410 has_super++;
2411
2412 return (has_super + ext4_group_first_block_no(sb, bg));
2413 }
2414
2415 /**
2416 * ext4_get_stripe_size: Get the stripe size.
2417 * @sbi: In memory super block info
2418 *
2419 * If we have specified it via mount option, then
2420 * use the mount option value. If the value specified at mount time is
2421 * greater than the blocks per group use the super block value.
2422 * If the super block value is greater than blocks per group return 0.
2423 * Allocator needs it be less than blocks per group.
2424 *
2425 */
2426 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2427 {
2428 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2429 unsigned long stripe_width =
2430 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2431 int ret;
2432
2433 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2434 ret = sbi->s_stripe;
2435 else if (stripe_width <= sbi->s_blocks_per_group)
2436 ret = stripe_width;
2437 else if (stride <= sbi->s_blocks_per_group)
2438 ret = stride;
2439 else
2440 ret = 0;
2441
2442 /*
2443 * If the stripe width is 1, this makes no sense and
2444 * we set it to 0 to turn off stripe handling code.
2445 */
2446 if (ret <= 1)
2447 ret = 0;
2448
2449 return ret;
2450 }
2451
2452 /* sysfs supprt */
2453
2454 struct ext4_attr {
2455 struct attribute attr;
2456 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2457 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2458 const char *, size_t);
2459 union {
2460 int offset;
2461 int deprecated_val;
2462 } u;
2463 };
2464
2465 static int parse_strtoull(const char *buf,
2466 unsigned long long max, unsigned long long *value)
2467 {
2468 int ret;
2469
2470 ret = kstrtoull(skip_spaces(buf), 0, value);
2471 if (!ret && *value > max)
2472 ret = -EINVAL;
2473 return ret;
2474 }
2475
2476 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2477 struct ext4_sb_info *sbi,
2478 char *buf)
2479 {
2480 return snprintf(buf, PAGE_SIZE, "%llu\n",
2481 (s64) EXT4_C2B(sbi,
2482 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2483 }
2484
2485 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2486 struct ext4_sb_info *sbi, char *buf)
2487 {
2488 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2489
2490 if (!sb->s_bdev->bd_part)
2491 return snprintf(buf, PAGE_SIZE, "0\n");
2492 return snprintf(buf, PAGE_SIZE, "%lu\n",
2493 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2494 sbi->s_sectors_written_start) >> 1);
2495 }
2496
2497 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2498 struct ext4_sb_info *sbi, char *buf)
2499 {
2500 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2501
2502 if (!sb->s_bdev->bd_part)
2503 return snprintf(buf, PAGE_SIZE, "0\n");
2504 return snprintf(buf, PAGE_SIZE, "%llu\n",
2505 (unsigned long long)(sbi->s_kbytes_written +
2506 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2507 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2508 }
2509
2510 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2511 struct ext4_sb_info *sbi,
2512 const char *buf, size_t count)
2513 {
2514 unsigned long t;
2515 int ret;
2516
2517 ret = kstrtoul(skip_spaces(buf), 0, &t);
2518 if (ret)
2519 return ret;
2520
2521 if (t && (!is_power_of_2(t) || t > 0x40000000))
2522 return -EINVAL;
2523
2524 sbi->s_inode_readahead_blks = t;
2525 return count;
2526 }
2527
2528 static ssize_t sbi_ui_show(struct ext4_attr *a,
2529 struct ext4_sb_info *sbi, char *buf)
2530 {
2531 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2532
2533 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2534 }
2535
2536 static ssize_t sbi_ui_store(struct ext4_attr *a,
2537 struct ext4_sb_info *sbi,
2538 const char *buf, size_t count)
2539 {
2540 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2541 unsigned long t;
2542 int ret;
2543
2544 ret = kstrtoul(skip_spaces(buf), 0, &t);
2545 if (ret)
2546 return ret;
2547 *ui = t;
2548 return count;
2549 }
2550
2551 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2552 struct ext4_sb_info *sbi, char *buf)
2553 {
2554 return snprintf(buf, PAGE_SIZE, "%llu\n",
2555 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2556 }
2557
2558 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2559 struct ext4_sb_info *sbi,
2560 const char *buf, size_t count)
2561 {
2562 unsigned long long val;
2563 int ret;
2564
2565 if (parse_strtoull(buf, -1ULL, &val))
2566 return -EINVAL;
2567 ret = ext4_reserve_clusters(sbi, val);
2568
2569 return ret ? ret : count;
2570 }
2571
2572 static ssize_t trigger_test_error(struct ext4_attr *a,
2573 struct ext4_sb_info *sbi,
2574 const char *buf, size_t count)
2575 {
2576 int len = count;
2577
2578 if (!capable(CAP_SYS_ADMIN))
2579 return -EPERM;
2580
2581 if (len && buf[len-1] == '\n')
2582 len--;
2583
2584 if (len)
2585 ext4_error(sbi->s_sb, "%.*s", len, buf);
2586 return count;
2587 }
2588
2589 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2590 struct ext4_sb_info *sbi, char *buf)
2591 {
2592 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2593 }
2594
2595 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2596 static struct ext4_attr ext4_attr_##_name = { \
2597 .attr = {.name = __stringify(_name), .mode = _mode }, \
2598 .show = _show, \
2599 .store = _store, \
2600 .u = { \
2601 .offset = offsetof(struct ext4_sb_info, _elname),\
2602 }, \
2603 }
2604 #define EXT4_ATTR(name, mode, show, store) \
2605 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2606
2607 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2608 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2609 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2610 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2611 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2612 #define ATTR_LIST(name) &ext4_attr_##name.attr
2613 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2614 static struct ext4_attr ext4_attr_##_name = { \
2615 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2616 .show = sbi_deprecated_show, \
2617 .u = { \
2618 .deprecated_val = _val, \
2619 }, \
2620 }
2621
2622 EXT4_RO_ATTR(delayed_allocation_blocks);
2623 EXT4_RO_ATTR(session_write_kbytes);
2624 EXT4_RO_ATTR(lifetime_write_kbytes);
2625 EXT4_RW_ATTR(reserved_clusters);
2626 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2627 inode_readahead_blks_store, s_inode_readahead_blks);
2628 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2629 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2630 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2631 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2632 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2633 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2634 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2635 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2636 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2637 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2638 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2639 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2640 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2641 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2642 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2643 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2644
2645 static struct attribute *ext4_attrs[] = {
2646 ATTR_LIST(delayed_allocation_blocks),
2647 ATTR_LIST(session_write_kbytes),
2648 ATTR_LIST(lifetime_write_kbytes),
2649 ATTR_LIST(reserved_clusters),
2650 ATTR_LIST(inode_readahead_blks),
2651 ATTR_LIST(inode_goal),
2652 ATTR_LIST(mb_stats),
2653 ATTR_LIST(mb_max_to_scan),
2654 ATTR_LIST(mb_min_to_scan),
2655 ATTR_LIST(mb_order2_req),
2656 ATTR_LIST(mb_stream_req),
2657 ATTR_LIST(mb_group_prealloc),
2658 ATTR_LIST(max_writeback_mb_bump),
2659 ATTR_LIST(extent_max_zeroout_kb),
2660 ATTR_LIST(trigger_fs_error),
2661 ATTR_LIST(err_ratelimit_interval_ms),
2662 ATTR_LIST(err_ratelimit_burst),
2663 ATTR_LIST(warning_ratelimit_interval_ms),
2664 ATTR_LIST(warning_ratelimit_burst),
2665 ATTR_LIST(msg_ratelimit_interval_ms),
2666 ATTR_LIST(msg_ratelimit_burst),
2667 NULL,
2668 };
2669
2670 /* Features this copy of ext4 supports */
2671 EXT4_INFO_ATTR(lazy_itable_init);
2672 EXT4_INFO_ATTR(batched_discard);
2673 EXT4_INFO_ATTR(meta_bg_resize);
2674
2675 static struct attribute *ext4_feat_attrs[] = {
2676 ATTR_LIST(lazy_itable_init),
2677 ATTR_LIST(batched_discard),
2678 ATTR_LIST(meta_bg_resize),
2679 NULL,
2680 };
2681
2682 static ssize_t ext4_attr_show(struct kobject *kobj,
2683 struct attribute *attr, char *buf)
2684 {
2685 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2686 s_kobj);
2687 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2688
2689 return a->show ? a->show(a, sbi, buf) : 0;
2690 }
2691
2692 static ssize_t ext4_attr_store(struct kobject *kobj,
2693 struct attribute *attr,
2694 const char *buf, size_t len)
2695 {
2696 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2697 s_kobj);
2698 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2699
2700 return a->store ? a->store(a, sbi, buf, len) : 0;
2701 }
2702
2703 static void ext4_sb_release(struct kobject *kobj)
2704 {
2705 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2706 s_kobj);
2707 complete(&sbi->s_kobj_unregister);
2708 }
2709
2710 static const struct sysfs_ops ext4_attr_ops = {
2711 .show = ext4_attr_show,
2712 .store = ext4_attr_store,
2713 };
2714
2715 static struct kobj_type ext4_ktype = {
2716 .default_attrs = ext4_attrs,
2717 .sysfs_ops = &ext4_attr_ops,
2718 .release = ext4_sb_release,
2719 };
2720
2721 static void ext4_feat_release(struct kobject *kobj)
2722 {
2723 complete(&ext4_feat->f_kobj_unregister);
2724 }
2725
2726 static struct kobj_type ext4_feat_ktype = {
2727 .default_attrs = ext4_feat_attrs,
2728 .sysfs_ops = &ext4_attr_ops,
2729 .release = ext4_feat_release,
2730 };
2731
2732 /*
2733 * Check whether this filesystem can be mounted based on
2734 * the features present and the RDONLY/RDWR mount requested.
2735 * Returns 1 if this filesystem can be mounted as requested,
2736 * 0 if it cannot be.
2737 */
2738 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2739 {
2740 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2741 ext4_msg(sb, KERN_ERR,
2742 "Couldn't mount because of "
2743 "unsupported optional features (%x)",
2744 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2745 ~EXT4_FEATURE_INCOMPAT_SUPP));
2746 return 0;
2747 }
2748
2749 if (readonly)
2750 return 1;
2751
2752 /* Check that feature set is OK for a read-write mount */
2753 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2754 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2755 "unsupported optional features (%x)",
2756 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2757 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2758 return 0;
2759 }
2760 /*
2761 * Large file size enabled file system can only be mounted
2762 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2763 */
2764 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2765 if (sizeof(blkcnt_t) < sizeof(u64)) {
2766 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2767 "cannot be mounted RDWR without "
2768 "CONFIG_LBDAF");
2769 return 0;
2770 }
2771 }
2772 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2773 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2774 ext4_msg(sb, KERN_ERR,
2775 "Can't support bigalloc feature without "
2776 "extents feature\n");
2777 return 0;
2778 }
2779
2780 #ifndef CONFIG_QUOTA
2781 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2782 !readonly) {
2783 ext4_msg(sb, KERN_ERR,
2784 "Filesystem with quota feature cannot be mounted RDWR "
2785 "without CONFIG_QUOTA");
2786 return 0;
2787 }
2788 #endif /* CONFIG_QUOTA */
2789 return 1;
2790 }
2791
2792 /*
2793 * This function is called once a day if we have errors logged
2794 * on the file system
2795 */
2796 static void print_daily_error_info(unsigned long arg)
2797 {
2798 struct super_block *sb = (struct super_block *) arg;
2799 struct ext4_sb_info *sbi;
2800 struct ext4_super_block *es;
2801
2802 sbi = EXT4_SB(sb);
2803 es = sbi->s_es;
2804
2805 if (es->s_error_count)
2806 /* fsck newer than v1.41.13 is needed to clean this condition. */
2807 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2808 le32_to_cpu(es->s_error_count));
2809 if (es->s_first_error_time) {
2810 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2811 sb->s_id, le32_to_cpu(es->s_first_error_time),
2812 (int) sizeof(es->s_first_error_func),
2813 es->s_first_error_func,
2814 le32_to_cpu(es->s_first_error_line));
2815 if (es->s_first_error_ino)
2816 printk(": inode %u",
2817 le32_to_cpu(es->s_first_error_ino));
2818 if (es->s_first_error_block)
2819 printk(": block %llu", (unsigned long long)
2820 le64_to_cpu(es->s_first_error_block));
2821 printk("\n");
2822 }
2823 if (es->s_last_error_time) {
2824 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2825 sb->s_id, le32_to_cpu(es->s_last_error_time),
2826 (int) sizeof(es->s_last_error_func),
2827 es->s_last_error_func,
2828 le32_to_cpu(es->s_last_error_line));
2829 if (es->s_last_error_ino)
2830 printk(": inode %u",
2831 le32_to_cpu(es->s_last_error_ino));
2832 if (es->s_last_error_block)
2833 printk(": block %llu", (unsigned long long)
2834 le64_to_cpu(es->s_last_error_block));
2835 printk("\n");
2836 }
2837 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2838 }
2839
2840 /* Find next suitable group and run ext4_init_inode_table */
2841 static int ext4_run_li_request(struct ext4_li_request *elr)
2842 {
2843 struct ext4_group_desc *gdp = NULL;
2844 ext4_group_t group, ngroups;
2845 struct super_block *sb;
2846 unsigned long timeout = 0;
2847 int ret = 0;
2848
2849 sb = elr->lr_super;
2850 ngroups = EXT4_SB(sb)->s_groups_count;
2851
2852 sb_start_write(sb);
2853 for (group = elr->lr_next_group; group < ngroups; group++) {
2854 gdp = ext4_get_group_desc(sb, group, NULL);
2855 if (!gdp) {
2856 ret = 1;
2857 break;
2858 }
2859
2860 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2861 break;
2862 }
2863
2864 if (group >= ngroups)
2865 ret = 1;
2866
2867 if (!ret) {
2868 timeout = jiffies;
2869 ret = ext4_init_inode_table(sb, group,
2870 elr->lr_timeout ? 0 : 1);
2871 if (elr->lr_timeout == 0) {
2872 timeout = (jiffies - timeout) *
2873 elr->lr_sbi->s_li_wait_mult;
2874 elr->lr_timeout = timeout;
2875 }
2876 elr->lr_next_sched = jiffies + elr->lr_timeout;
2877 elr->lr_next_group = group + 1;
2878 }
2879 sb_end_write(sb);
2880
2881 return ret;
2882 }
2883
2884 /*
2885 * Remove lr_request from the list_request and free the
2886 * request structure. Should be called with li_list_mtx held
2887 */
2888 static void ext4_remove_li_request(struct ext4_li_request *elr)
2889 {
2890 struct ext4_sb_info *sbi;
2891
2892 if (!elr)
2893 return;
2894
2895 sbi = elr->lr_sbi;
2896
2897 list_del(&elr->lr_request);
2898 sbi->s_li_request = NULL;
2899 kfree(elr);
2900 }
2901
2902 static void ext4_unregister_li_request(struct super_block *sb)
2903 {
2904 mutex_lock(&ext4_li_mtx);
2905 if (!ext4_li_info) {
2906 mutex_unlock(&ext4_li_mtx);
2907 return;
2908 }
2909
2910 mutex_lock(&ext4_li_info->li_list_mtx);
2911 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2912 mutex_unlock(&ext4_li_info->li_list_mtx);
2913 mutex_unlock(&ext4_li_mtx);
2914 }
2915
2916 static struct task_struct *ext4_lazyinit_task;
2917
2918 /*
2919 * This is the function where ext4lazyinit thread lives. It walks
2920 * through the request list searching for next scheduled filesystem.
2921 * When such a fs is found, run the lazy initialization request
2922 * (ext4_rn_li_request) and keep track of the time spend in this
2923 * function. Based on that time we compute next schedule time of
2924 * the request. When walking through the list is complete, compute
2925 * next waking time and put itself into sleep.
2926 */
2927 static int ext4_lazyinit_thread(void *arg)
2928 {
2929 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2930 struct list_head *pos, *n;
2931 struct ext4_li_request *elr;
2932 unsigned long next_wakeup, cur;
2933
2934 BUG_ON(NULL == eli);
2935
2936 cont_thread:
2937 while (true) {
2938 next_wakeup = MAX_JIFFY_OFFSET;
2939
2940 mutex_lock(&eli->li_list_mtx);
2941 if (list_empty(&eli->li_request_list)) {
2942 mutex_unlock(&eli->li_list_mtx);
2943 goto exit_thread;
2944 }
2945
2946 list_for_each_safe(pos, n, &eli->li_request_list) {
2947 elr = list_entry(pos, struct ext4_li_request,
2948 lr_request);
2949
2950 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2951 if (ext4_run_li_request(elr) != 0) {
2952 /* error, remove the lazy_init job */
2953 ext4_remove_li_request(elr);
2954 continue;
2955 }
2956 }
2957
2958 if (time_before(elr->lr_next_sched, next_wakeup))
2959 next_wakeup = elr->lr_next_sched;
2960 }
2961 mutex_unlock(&eli->li_list_mtx);
2962
2963 try_to_freeze();
2964
2965 cur = jiffies;
2966 if ((time_after_eq(cur, next_wakeup)) ||
2967 (MAX_JIFFY_OFFSET == next_wakeup)) {
2968 cond_resched();
2969 continue;
2970 }
2971
2972 schedule_timeout_interruptible(next_wakeup - cur);
2973
2974 if (kthread_should_stop()) {
2975 ext4_clear_request_list();
2976 goto exit_thread;
2977 }
2978 }
2979
2980 exit_thread:
2981 /*
2982 * It looks like the request list is empty, but we need
2983 * to check it under the li_list_mtx lock, to prevent any
2984 * additions into it, and of course we should lock ext4_li_mtx
2985 * to atomically free the list and ext4_li_info, because at
2986 * this point another ext4 filesystem could be registering
2987 * new one.
2988 */
2989 mutex_lock(&ext4_li_mtx);
2990 mutex_lock(&eli->li_list_mtx);
2991 if (!list_empty(&eli->li_request_list)) {
2992 mutex_unlock(&eli->li_list_mtx);
2993 mutex_unlock(&ext4_li_mtx);
2994 goto cont_thread;
2995 }
2996 mutex_unlock(&eli->li_list_mtx);
2997 kfree(ext4_li_info);
2998 ext4_li_info = NULL;
2999 mutex_unlock(&ext4_li_mtx);
3000
3001 return 0;
3002 }
3003
3004 static void ext4_clear_request_list(void)
3005 {
3006 struct list_head *pos, *n;
3007 struct ext4_li_request *elr;
3008
3009 mutex_lock(&ext4_li_info->li_list_mtx);
3010 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3011 elr = list_entry(pos, struct ext4_li_request,
3012 lr_request);
3013 ext4_remove_li_request(elr);
3014 }
3015 mutex_unlock(&ext4_li_info->li_list_mtx);
3016 }
3017
3018 static int ext4_run_lazyinit_thread(void)
3019 {
3020 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3021 ext4_li_info, "ext4lazyinit");
3022 if (IS_ERR(ext4_lazyinit_task)) {
3023 int err = PTR_ERR(ext4_lazyinit_task);
3024 ext4_clear_request_list();
3025 kfree(ext4_li_info);
3026 ext4_li_info = NULL;
3027 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3028 "initialization thread\n",
3029 err);
3030 return err;
3031 }
3032 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3033 return 0;
3034 }
3035
3036 /*
3037 * Check whether it make sense to run itable init. thread or not.
3038 * If there is at least one uninitialized inode table, return
3039 * corresponding group number, else the loop goes through all
3040 * groups and return total number of groups.
3041 */
3042 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3043 {
3044 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3045 struct ext4_group_desc *gdp = NULL;
3046
3047 for (group = 0; group < ngroups; group++) {
3048 gdp = ext4_get_group_desc(sb, group, NULL);
3049 if (!gdp)
3050 continue;
3051
3052 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3053 break;
3054 }
3055
3056 return group;
3057 }
3058
3059 static int ext4_li_info_new(void)
3060 {
3061 struct ext4_lazy_init *eli = NULL;
3062
3063 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3064 if (!eli)
3065 return -ENOMEM;
3066
3067 INIT_LIST_HEAD(&eli->li_request_list);
3068 mutex_init(&eli->li_list_mtx);
3069
3070 eli->li_state |= EXT4_LAZYINIT_QUIT;
3071
3072 ext4_li_info = eli;
3073
3074 return 0;
3075 }
3076
3077 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3078 ext4_group_t start)
3079 {
3080 struct ext4_sb_info *sbi = EXT4_SB(sb);
3081 struct ext4_li_request *elr;
3082
3083 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3084 if (!elr)
3085 return NULL;
3086
3087 elr->lr_super = sb;
3088 elr->lr_sbi = sbi;
3089 elr->lr_next_group = start;
3090
3091 /*
3092 * Randomize first schedule time of the request to
3093 * spread the inode table initialization requests
3094 * better.
3095 */
3096 elr->lr_next_sched = jiffies + (prandom_u32() %
3097 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3098 return elr;
3099 }
3100
3101 int ext4_register_li_request(struct super_block *sb,
3102 ext4_group_t first_not_zeroed)
3103 {
3104 struct ext4_sb_info *sbi = EXT4_SB(sb);
3105 struct ext4_li_request *elr = NULL;
3106 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3107 int ret = 0;
3108
3109 mutex_lock(&ext4_li_mtx);
3110 if (sbi->s_li_request != NULL) {
3111 /*
3112 * Reset timeout so it can be computed again, because
3113 * s_li_wait_mult might have changed.
3114 */
3115 sbi->s_li_request->lr_timeout = 0;
3116 goto out;
3117 }
3118
3119 if (first_not_zeroed == ngroups ||
3120 (sb->s_flags & MS_RDONLY) ||
3121 !test_opt(sb, INIT_INODE_TABLE))
3122 goto out;
3123
3124 elr = ext4_li_request_new(sb, first_not_zeroed);
3125 if (!elr) {
3126 ret = -ENOMEM;
3127 goto out;
3128 }
3129
3130 if (NULL == ext4_li_info) {
3131 ret = ext4_li_info_new();
3132 if (ret)
3133 goto out;
3134 }
3135
3136 mutex_lock(&ext4_li_info->li_list_mtx);
3137 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3138 mutex_unlock(&ext4_li_info->li_list_mtx);
3139
3140 sbi->s_li_request = elr;
3141 /*
3142 * set elr to NULL here since it has been inserted to
3143 * the request_list and the removal and free of it is
3144 * handled by ext4_clear_request_list from now on.
3145 */
3146 elr = NULL;
3147
3148 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3149 ret = ext4_run_lazyinit_thread();
3150 if (ret)
3151 goto out;
3152 }
3153 out:
3154 mutex_unlock(&ext4_li_mtx);
3155 if (ret)
3156 kfree(elr);
3157 return ret;
3158 }
3159
3160 /*
3161 * We do not need to lock anything since this is called on
3162 * module unload.
3163 */
3164 static void ext4_destroy_lazyinit_thread(void)
3165 {
3166 /*
3167 * If thread exited earlier
3168 * there's nothing to be done.
3169 */
3170 if (!ext4_li_info || !ext4_lazyinit_task)
3171 return;
3172
3173 kthread_stop(ext4_lazyinit_task);
3174 }
3175
3176 static int set_journal_csum_feature_set(struct super_block *sb)
3177 {
3178 int ret = 1;
3179 int compat, incompat;
3180 struct ext4_sb_info *sbi = EXT4_SB(sb);
3181
3182 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3183 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3184 /* journal checksum v2 */
3185 compat = 0;
3186 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3187 } else {
3188 /* journal checksum v1 */
3189 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3190 incompat = 0;
3191 }
3192
3193 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3194 ret = jbd2_journal_set_features(sbi->s_journal,
3195 compat, 0,
3196 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3197 incompat);
3198 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3199 ret = jbd2_journal_set_features(sbi->s_journal,
3200 compat, 0,
3201 incompat);
3202 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3203 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3204 } else {
3205 jbd2_journal_clear_features(sbi->s_journal,
3206 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3207 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3208 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3209 }
3210
3211 return ret;
3212 }
3213
3214 /*
3215 * Note: calculating the overhead so we can be compatible with
3216 * historical BSD practice is quite difficult in the face of
3217 * clusters/bigalloc. This is because multiple metadata blocks from
3218 * different block group can end up in the same allocation cluster.
3219 * Calculating the exact overhead in the face of clustered allocation
3220 * requires either O(all block bitmaps) in memory or O(number of block
3221 * groups**2) in time. We will still calculate the superblock for
3222 * older file systems --- and if we come across with a bigalloc file
3223 * system with zero in s_overhead_clusters the estimate will be close to
3224 * correct especially for very large cluster sizes --- but for newer
3225 * file systems, it's better to calculate this figure once at mkfs
3226 * time, and store it in the superblock. If the superblock value is
3227 * present (even for non-bigalloc file systems), we will use it.
3228 */
3229 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3230 char *buf)
3231 {
3232 struct ext4_sb_info *sbi = EXT4_SB(sb);
3233 struct ext4_group_desc *gdp;
3234 ext4_fsblk_t first_block, last_block, b;
3235 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3236 int s, j, count = 0;
3237
3238 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3239 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3240 sbi->s_itb_per_group + 2);
3241
3242 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3243 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3244 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3245 for (i = 0; i < ngroups; i++) {
3246 gdp = ext4_get_group_desc(sb, i, NULL);
3247 b = ext4_block_bitmap(sb, gdp);
3248 if (b >= first_block && b <= last_block) {
3249 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3250 count++;
3251 }
3252 b = ext4_inode_bitmap(sb, gdp);
3253 if (b >= first_block && b <= last_block) {
3254 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3255 count++;
3256 }
3257 b = ext4_inode_table(sb, gdp);
3258 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3259 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3260 int c = EXT4_B2C(sbi, b - first_block);
3261 ext4_set_bit(c, buf);
3262 count++;
3263 }
3264 if (i != grp)
3265 continue;
3266 s = 0;
3267 if (ext4_bg_has_super(sb, grp)) {
3268 ext4_set_bit(s++, buf);
3269 count++;
3270 }
3271 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3272 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3273 count++;
3274 }
3275 }
3276 if (!count)
3277 return 0;
3278 return EXT4_CLUSTERS_PER_GROUP(sb) -
3279 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3280 }
3281
3282 /*
3283 * Compute the overhead and stash it in sbi->s_overhead
3284 */
3285 int ext4_calculate_overhead(struct super_block *sb)
3286 {
3287 struct ext4_sb_info *sbi = EXT4_SB(sb);
3288 struct ext4_super_block *es = sbi->s_es;
3289 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3290 ext4_fsblk_t overhead = 0;
3291 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3292
3293 if (!buf)
3294 return -ENOMEM;
3295
3296 /*
3297 * Compute the overhead (FS structures). This is constant
3298 * for a given filesystem unless the number of block groups
3299 * changes so we cache the previous value until it does.
3300 */
3301
3302 /*
3303 * All of the blocks before first_data_block are overhead
3304 */
3305 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3306
3307 /*
3308 * Add the overhead found in each block group
3309 */
3310 for (i = 0; i < ngroups; i++) {
3311 int blks;
3312
3313 blks = count_overhead(sb, i, buf);
3314 overhead += blks;
3315 if (blks)
3316 memset(buf, 0, PAGE_SIZE);
3317 cond_resched();
3318 }
3319 /* Add the journal blocks as well */
3320 if (sbi->s_journal)
3321 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3322
3323 sbi->s_overhead = overhead;
3324 smp_wmb();
3325 free_page((unsigned long) buf);
3326 return 0;
3327 }
3328
3329
3330 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3331 {
3332 ext4_fsblk_t resv_clusters;
3333
3334 /*
3335 * There's no need to reserve anything when we aren't using extents.
3336 * The space estimates are exact, there are no unwritten extents,
3337 * hole punching doesn't need new metadata... This is needed especially
3338 * to keep ext2/3 backward compatibility.
3339 */
3340 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3341 return 0;
3342 /*
3343 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3344 * This should cover the situations where we can not afford to run
3345 * out of space like for example punch hole, or converting
3346 * unwritten extents in delalloc path. In most cases such
3347 * allocation would require 1, or 2 blocks, higher numbers are
3348 * very rare.
3349 */
3350 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3351 EXT4_SB(sb)->s_cluster_bits;
3352
3353 do_div(resv_clusters, 50);
3354 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3355
3356 return resv_clusters;
3357 }
3358
3359
3360 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3361 {
3362 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3363 sbi->s_cluster_bits;
3364
3365 if (count >= clusters)
3366 return -EINVAL;
3367
3368 atomic64_set(&sbi->s_resv_clusters, count);
3369 return 0;
3370 }
3371
3372 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3373 {
3374 char *orig_data = kstrdup(data, GFP_KERNEL);
3375 struct buffer_head *bh;
3376 struct ext4_super_block *es = NULL;
3377 struct ext4_sb_info *sbi;
3378 ext4_fsblk_t block;
3379 ext4_fsblk_t sb_block = get_sb_block(&data);
3380 ext4_fsblk_t logical_sb_block;
3381 unsigned long offset = 0;
3382 unsigned long journal_devnum = 0;
3383 unsigned long def_mount_opts;
3384 struct inode *root;
3385 char *cp;
3386 const char *descr;
3387 int ret = -ENOMEM;
3388 int blocksize, clustersize;
3389 unsigned int db_count;
3390 unsigned int i;
3391 int needs_recovery, has_huge_files, has_bigalloc;
3392 __u64 blocks_count;
3393 int err = 0;
3394 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3395 ext4_group_t first_not_zeroed;
3396
3397 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3398 if (!sbi)
3399 goto out_free_orig;
3400
3401 sbi->s_blockgroup_lock =
3402 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3403 if (!sbi->s_blockgroup_lock) {
3404 kfree(sbi);
3405 goto out_free_orig;
3406 }
3407 sb->s_fs_info = sbi;
3408 sbi->s_sb = sb;
3409 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3410 sbi->s_sb_block = sb_block;
3411 if (sb->s_bdev->bd_part)
3412 sbi->s_sectors_written_start =
3413 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3414
3415 /* Cleanup superblock name */
3416 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3417 *cp = '!';
3418
3419 /* -EINVAL is default */
3420 ret = -EINVAL;
3421 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3422 if (!blocksize) {
3423 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3424 goto out_fail;
3425 }
3426
3427 /*
3428 * The ext4 superblock will not be buffer aligned for other than 1kB
3429 * block sizes. We need to calculate the offset from buffer start.
3430 */
3431 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3432 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3433 offset = do_div(logical_sb_block, blocksize);
3434 } else {
3435 logical_sb_block = sb_block;
3436 }
3437
3438 if (!(bh = sb_bread(sb, logical_sb_block))) {
3439 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3440 goto out_fail;
3441 }
3442 /*
3443 * Note: s_es must be initialized as soon as possible because
3444 * some ext4 macro-instructions depend on its value
3445 */
3446 es = (struct ext4_super_block *) (bh->b_data + offset);
3447 sbi->s_es = es;
3448 sb->s_magic = le16_to_cpu(es->s_magic);
3449 if (sb->s_magic != EXT4_SUPER_MAGIC)
3450 goto cantfind_ext4;
3451 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3452
3453 /* Warn if metadata_csum and gdt_csum are both set. */
3454 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3455 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3456 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3457 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3458 "redundant flags; please run fsck.");
3459
3460 /* Check for a known checksum algorithm */
3461 if (!ext4_verify_csum_type(sb, es)) {
3462 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3463 "unknown checksum algorithm.");
3464 silent = 1;
3465 goto cantfind_ext4;
3466 }
3467
3468 /* Load the checksum driver */
3469 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3470 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3471 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3472 if (IS_ERR(sbi->s_chksum_driver)) {
3473 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3474 ret = PTR_ERR(sbi->s_chksum_driver);
3475 sbi->s_chksum_driver = NULL;
3476 goto failed_mount;
3477 }
3478 }
3479
3480 /* Check superblock checksum */
3481 if (!ext4_superblock_csum_verify(sb, es)) {
3482 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3483 "invalid superblock checksum. Run e2fsck?");
3484 silent = 1;
3485 goto cantfind_ext4;
3486 }
3487
3488 /* Precompute checksum seed for all metadata */
3489 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3490 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3491 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3492 sizeof(es->s_uuid));
3493
3494 /* Set defaults before we parse the mount options */
3495 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3496 set_opt(sb, INIT_INODE_TABLE);
3497 if (def_mount_opts & EXT4_DEFM_DEBUG)
3498 set_opt(sb, DEBUG);
3499 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3500 set_opt(sb, GRPID);
3501 if (def_mount_opts & EXT4_DEFM_UID16)
3502 set_opt(sb, NO_UID32);
3503 /* xattr user namespace & acls are now defaulted on */
3504 set_opt(sb, XATTR_USER);
3505 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3506 set_opt(sb, POSIX_ACL);
3507 #endif
3508 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3509 set_opt(sb, JOURNAL_DATA);
3510 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3511 set_opt(sb, ORDERED_DATA);
3512 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3513 set_opt(sb, WRITEBACK_DATA);
3514
3515 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3516 set_opt(sb, ERRORS_PANIC);
3517 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3518 set_opt(sb, ERRORS_CONT);
3519 else
3520 set_opt(sb, ERRORS_RO);
3521 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3522 set_opt(sb, BLOCK_VALIDITY);
3523 if (def_mount_opts & EXT4_DEFM_DISCARD)
3524 set_opt(sb, DISCARD);
3525
3526 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3527 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3528 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3529 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3530 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3531
3532 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3533 set_opt(sb, BARRIER);
3534
3535 /*
3536 * enable delayed allocation by default
3537 * Use -o nodelalloc to turn it off
3538 */
3539 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3540 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3541 set_opt(sb, DELALLOC);
3542
3543 /*
3544 * set default s_li_wait_mult for lazyinit, for the case there is
3545 * no mount option specified.
3546 */
3547 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3548
3549 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3550 &journal_devnum, &journal_ioprio, 0)) {
3551 ext4_msg(sb, KERN_WARNING,
3552 "failed to parse options in superblock: %s",
3553 sbi->s_es->s_mount_opts);
3554 }
3555 sbi->s_def_mount_opt = sbi->s_mount_opt;
3556 if (!parse_options((char *) data, sb, &journal_devnum,
3557 &journal_ioprio, 0))
3558 goto failed_mount;
3559
3560 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3561 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3562 "with data=journal disables delayed "
3563 "allocation and O_DIRECT support!\n");
3564 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3565 ext4_msg(sb, KERN_ERR, "can't mount with "
3566 "both data=journal and delalloc");
3567 goto failed_mount;
3568 }
3569 if (test_opt(sb, DIOREAD_NOLOCK)) {
3570 ext4_msg(sb, KERN_ERR, "can't mount with "
3571 "both data=journal and dioread_nolock");
3572 goto failed_mount;
3573 }
3574 if (test_opt(sb, DELALLOC))
3575 clear_opt(sb, DELALLOC);
3576 }
3577
3578 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3579 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3580
3581 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3582 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3583 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3584 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3585 ext4_msg(sb, KERN_WARNING,
3586 "feature flags set on rev 0 fs, "
3587 "running e2fsck is recommended");
3588
3589 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3590 set_opt2(sb, HURD_COMPAT);
3591 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3592 EXT4_FEATURE_INCOMPAT_64BIT)) {
3593 ext4_msg(sb, KERN_ERR,
3594 "The Hurd can't support 64-bit file systems");
3595 goto failed_mount;
3596 }
3597 }
3598
3599 if (IS_EXT2_SB(sb)) {
3600 if (ext2_feature_set_ok(sb))
3601 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3602 "using the ext4 subsystem");
3603 else {
3604 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3605 "to feature incompatibilities");
3606 goto failed_mount;
3607 }
3608 }
3609
3610 if (IS_EXT3_SB(sb)) {
3611 if (ext3_feature_set_ok(sb))
3612 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3613 "using the ext4 subsystem");
3614 else {
3615 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3616 "to feature incompatibilities");
3617 goto failed_mount;
3618 }
3619 }
3620
3621 /*
3622 * Check feature flags regardless of the revision level, since we
3623 * previously didn't change the revision level when setting the flags,
3624 * so there is a chance incompat flags are set on a rev 0 filesystem.
3625 */
3626 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3627 goto failed_mount;
3628
3629 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3630 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3631 blocksize > EXT4_MAX_BLOCK_SIZE) {
3632 ext4_msg(sb, KERN_ERR,
3633 "Unsupported filesystem blocksize %d", blocksize);
3634 goto failed_mount;
3635 }
3636
3637 if (sb->s_blocksize != blocksize) {
3638 /* Validate the filesystem blocksize */
3639 if (!sb_set_blocksize(sb, blocksize)) {
3640 ext4_msg(sb, KERN_ERR, "bad block size %d",
3641 blocksize);
3642 goto failed_mount;
3643 }
3644
3645 brelse(bh);
3646 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3647 offset = do_div(logical_sb_block, blocksize);
3648 bh = sb_bread(sb, logical_sb_block);
3649 if (!bh) {
3650 ext4_msg(sb, KERN_ERR,
3651 "Can't read superblock on 2nd try");
3652 goto failed_mount;
3653 }
3654 es = (struct ext4_super_block *)(bh->b_data + offset);
3655 sbi->s_es = es;
3656 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3657 ext4_msg(sb, KERN_ERR,
3658 "Magic mismatch, very weird!");
3659 goto failed_mount;
3660 }
3661 }
3662
3663 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3664 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3665 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3666 has_huge_files);
3667 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3668
3669 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3670 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3671 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3672 } else {
3673 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3674 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3675 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3676 (!is_power_of_2(sbi->s_inode_size)) ||
3677 (sbi->s_inode_size > blocksize)) {
3678 ext4_msg(sb, KERN_ERR,
3679 "unsupported inode size: %d",
3680 sbi->s_inode_size);
3681 goto failed_mount;
3682 }
3683 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3684 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3685 }
3686
3687 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3688 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3689 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3690 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3691 !is_power_of_2(sbi->s_desc_size)) {
3692 ext4_msg(sb, KERN_ERR,
3693 "unsupported descriptor size %lu",
3694 sbi->s_desc_size);
3695 goto failed_mount;
3696 }
3697 } else
3698 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3699
3700 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3701 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3702 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3703 goto cantfind_ext4;
3704
3705 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3706 if (sbi->s_inodes_per_block == 0)
3707 goto cantfind_ext4;
3708 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3709 sbi->s_inodes_per_block;
3710 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3711 sbi->s_sbh = bh;
3712 sbi->s_mount_state = le16_to_cpu(es->s_state);
3713 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3714 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3715
3716 for (i = 0; i < 4; i++)
3717 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3718 sbi->s_def_hash_version = es->s_def_hash_version;
3719 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3720 i = le32_to_cpu(es->s_flags);
3721 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3722 sbi->s_hash_unsigned = 3;
3723 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3724 #ifdef __CHAR_UNSIGNED__
3725 if (!(sb->s_flags & MS_RDONLY))
3726 es->s_flags |=
3727 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3728 sbi->s_hash_unsigned = 3;
3729 #else
3730 if (!(sb->s_flags & MS_RDONLY))
3731 es->s_flags |=
3732 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3733 #endif
3734 }
3735 }
3736
3737 /* Handle clustersize */
3738 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3739 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3740 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3741 if (has_bigalloc) {
3742 if (clustersize < blocksize) {
3743 ext4_msg(sb, KERN_ERR,
3744 "cluster size (%d) smaller than "
3745 "block size (%d)", clustersize, blocksize);
3746 goto failed_mount;
3747 }
3748 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3749 le32_to_cpu(es->s_log_block_size);
3750 sbi->s_clusters_per_group =
3751 le32_to_cpu(es->s_clusters_per_group);
3752 if (sbi->s_clusters_per_group > blocksize * 8) {
3753 ext4_msg(sb, KERN_ERR,
3754 "#clusters per group too big: %lu",
3755 sbi->s_clusters_per_group);
3756 goto failed_mount;
3757 }
3758 if (sbi->s_blocks_per_group !=
3759 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3760 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3761 "clusters per group (%lu) inconsistent",
3762 sbi->s_blocks_per_group,
3763 sbi->s_clusters_per_group);
3764 goto failed_mount;
3765 }
3766 } else {
3767 if (clustersize != blocksize) {
3768 ext4_warning(sb, "fragment/cluster size (%d) != "
3769 "block size (%d)", clustersize,
3770 blocksize);
3771 clustersize = blocksize;
3772 }
3773 if (sbi->s_blocks_per_group > blocksize * 8) {
3774 ext4_msg(sb, KERN_ERR,
3775 "#blocks per group too big: %lu",
3776 sbi->s_blocks_per_group);
3777 goto failed_mount;
3778 }
3779 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3780 sbi->s_cluster_bits = 0;
3781 }
3782 sbi->s_cluster_ratio = clustersize / blocksize;
3783
3784 if (sbi->s_inodes_per_group > blocksize * 8) {
3785 ext4_msg(sb, KERN_ERR,
3786 "#inodes per group too big: %lu",
3787 sbi->s_inodes_per_group);
3788 goto failed_mount;
3789 }
3790
3791 /* Do we have standard group size of clustersize * 8 blocks ? */
3792 if (sbi->s_blocks_per_group == clustersize << 3)
3793 set_opt2(sb, STD_GROUP_SIZE);
3794
3795 /*
3796 * Test whether we have more sectors than will fit in sector_t,
3797 * and whether the max offset is addressable by the page cache.
3798 */
3799 err = generic_check_addressable(sb->s_blocksize_bits,
3800 ext4_blocks_count(es));
3801 if (err) {
3802 ext4_msg(sb, KERN_ERR, "filesystem"
3803 " too large to mount safely on this system");
3804 if (sizeof(sector_t) < 8)
3805 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3806 goto failed_mount;
3807 }
3808
3809 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3810 goto cantfind_ext4;
3811
3812 /* check blocks count against device size */
3813 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3814 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3815 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3816 "exceeds size of device (%llu blocks)",
3817 ext4_blocks_count(es), blocks_count);
3818 goto failed_mount;
3819 }
3820
3821 /*
3822 * It makes no sense for the first data block to be beyond the end
3823 * of the filesystem.
3824 */
3825 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3826 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3827 "block %u is beyond end of filesystem (%llu)",
3828 le32_to_cpu(es->s_first_data_block),
3829 ext4_blocks_count(es));
3830 goto failed_mount;
3831 }
3832 blocks_count = (ext4_blocks_count(es) -
3833 le32_to_cpu(es->s_first_data_block) +
3834 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3835 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3836 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3837 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3838 "(block count %llu, first data block %u, "
3839 "blocks per group %lu)", sbi->s_groups_count,
3840 ext4_blocks_count(es),
3841 le32_to_cpu(es->s_first_data_block),
3842 EXT4_BLOCKS_PER_GROUP(sb));
3843 goto failed_mount;
3844 }
3845 sbi->s_groups_count = blocks_count;
3846 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3847 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3848 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3849 EXT4_DESC_PER_BLOCK(sb);
3850 sbi->s_group_desc = ext4_kvmalloc(db_count *
3851 sizeof(struct buffer_head *),
3852 GFP_KERNEL);
3853 if (sbi->s_group_desc == NULL) {
3854 ext4_msg(sb, KERN_ERR, "not enough memory");
3855 ret = -ENOMEM;
3856 goto failed_mount;
3857 }
3858
3859 if (ext4_proc_root)
3860 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3861
3862 if (sbi->s_proc)
3863 proc_create_data("options", S_IRUGO, sbi->s_proc,
3864 &ext4_seq_options_fops, sb);
3865
3866 bgl_lock_init(sbi->s_blockgroup_lock);
3867
3868 for (i = 0; i < db_count; i++) {
3869 block = descriptor_loc(sb, logical_sb_block, i);
3870 sbi->s_group_desc[i] = sb_bread(sb, block);
3871 if (!sbi->s_group_desc[i]) {
3872 ext4_msg(sb, KERN_ERR,
3873 "can't read group descriptor %d", i);
3874 db_count = i;
3875 goto failed_mount2;
3876 }
3877 }
3878 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3879 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3880 goto failed_mount2;
3881 }
3882
3883 sbi->s_gdb_count = db_count;
3884 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3885 spin_lock_init(&sbi->s_next_gen_lock);
3886
3887 init_timer(&sbi->s_err_report);
3888 sbi->s_err_report.function = print_daily_error_info;
3889 sbi->s_err_report.data = (unsigned long) sb;
3890
3891 /* Register extent status tree shrinker */
3892 ext4_es_register_shrinker(sbi);
3893
3894 if ((err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0)) != 0) {
3895 ext4_msg(sb, KERN_ERR, "insufficient memory");
3896 goto failed_mount3;
3897 }
3898
3899 sbi->s_stripe = ext4_get_stripe_size(sbi);
3900 sbi->s_extent_max_zeroout_kb = 32;
3901
3902 /*
3903 * set up enough so that it can read an inode
3904 */
3905 if (!test_opt(sb, NOLOAD) &&
3906 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3907 sb->s_op = &ext4_sops;
3908 else
3909 sb->s_op = &ext4_nojournal_sops;
3910 sb->s_export_op = &ext4_export_ops;
3911 sb->s_xattr = ext4_xattr_handlers;
3912 #ifdef CONFIG_QUOTA
3913 sb->dq_op = &ext4_quota_operations;
3914 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3915 sb->s_qcop = &ext4_qctl_sysfile_operations;
3916 else
3917 sb->s_qcop = &ext4_qctl_operations;
3918 #endif
3919 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3920
3921 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3922 mutex_init(&sbi->s_orphan_lock);
3923
3924 sb->s_root = NULL;
3925
3926 needs_recovery = (es->s_last_orphan != 0 ||
3927 EXT4_HAS_INCOMPAT_FEATURE(sb,
3928 EXT4_FEATURE_INCOMPAT_RECOVER));
3929
3930 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3931 !(sb->s_flags & MS_RDONLY))
3932 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3933 goto failed_mount3;
3934
3935 /*
3936 * The first inode we look at is the journal inode. Don't try
3937 * root first: it may be modified in the journal!
3938 */
3939 if (!test_opt(sb, NOLOAD) &&
3940 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3941 if (ext4_load_journal(sb, es, journal_devnum))
3942 goto failed_mount3;
3943 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3944 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3945 ext4_msg(sb, KERN_ERR, "required journal recovery "
3946 "suppressed and not mounted read-only");
3947 goto failed_mount_wq;
3948 } else {
3949 clear_opt(sb, DATA_FLAGS);
3950 sbi->s_journal = NULL;
3951 needs_recovery = 0;
3952 goto no_journal;
3953 }
3954
3955 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3956 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3957 JBD2_FEATURE_INCOMPAT_64BIT)) {
3958 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3959 goto failed_mount_wq;
3960 }
3961
3962 if (!set_journal_csum_feature_set(sb)) {
3963 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3964 "feature set");
3965 goto failed_mount_wq;
3966 }
3967
3968 /* We have now updated the journal if required, so we can
3969 * validate the data journaling mode. */
3970 switch (test_opt(sb, DATA_FLAGS)) {
3971 case 0:
3972 /* No mode set, assume a default based on the journal
3973 * capabilities: ORDERED_DATA if the journal can
3974 * cope, else JOURNAL_DATA
3975 */
3976 if (jbd2_journal_check_available_features
3977 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3978 set_opt(sb, ORDERED_DATA);
3979 else
3980 set_opt(sb, JOURNAL_DATA);
3981 break;
3982
3983 case EXT4_MOUNT_ORDERED_DATA:
3984 case EXT4_MOUNT_WRITEBACK_DATA:
3985 if (!jbd2_journal_check_available_features
3986 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3987 ext4_msg(sb, KERN_ERR, "Journal does not support "
3988 "requested data journaling mode");
3989 goto failed_mount_wq;
3990 }
3991 default:
3992 break;
3993 }
3994 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3995
3996 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3997
3998 no_journal:
3999 if (ext4_mballoc_ready) {
4000 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4001 if (!sbi->s_mb_cache) {
4002 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4003 goto failed_mount_wq;
4004 }
4005 }
4006
4007 /*
4008 * Get the # of file system overhead blocks from the
4009 * superblock if present.
4010 */
4011 if (es->s_overhead_clusters)
4012 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4013 else {
4014 err = ext4_calculate_overhead(sb);
4015 if (err)
4016 goto failed_mount_wq;
4017 }
4018
4019 /*
4020 * The maximum number of concurrent works can be high and
4021 * concurrency isn't really necessary. Limit it to 1.
4022 */
4023 EXT4_SB(sb)->rsv_conversion_wq =
4024 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4025 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4026 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4027 ret = -ENOMEM;
4028 goto failed_mount4;
4029 }
4030
4031 /*
4032 * The jbd2_journal_load will have done any necessary log recovery,
4033 * so we can safely mount the rest of the filesystem now.
4034 */
4035
4036 root = ext4_iget(sb, EXT4_ROOT_INO);
4037 if (IS_ERR(root)) {
4038 ext4_msg(sb, KERN_ERR, "get root inode failed");
4039 ret = PTR_ERR(root);
4040 root = NULL;
4041 goto failed_mount4;
4042 }
4043 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4044 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4045 iput(root);
4046 goto failed_mount4;
4047 }
4048 sb->s_root = d_make_root(root);
4049 if (!sb->s_root) {
4050 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4051 ret = -ENOMEM;
4052 goto failed_mount4;
4053 }
4054
4055 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4056 sb->s_flags |= MS_RDONLY;
4057
4058 /* determine the minimum size of new large inodes, if present */
4059 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4060 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4061 EXT4_GOOD_OLD_INODE_SIZE;
4062 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4063 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4064 if (sbi->s_want_extra_isize <
4065 le16_to_cpu(es->s_want_extra_isize))
4066 sbi->s_want_extra_isize =
4067 le16_to_cpu(es->s_want_extra_isize);
4068 if (sbi->s_want_extra_isize <
4069 le16_to_cpu(es->s_min_extra_isize))
4070 sbi->s_want_extra_isize =
4071 le16_to_cpu(es->s_min_extra_isize);
4072 }
4073 }
4074 /* Check if enough inode space is available */
4075 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4076 sbi->s_inode_size) {
4077 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4078 EXT4_GOOD_OLD_INODE_SIZE;
4079 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4080 "available");
4081 }
4082
4083 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4084 if (err) {
4085 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4086 "reserved pool", ext4_calculate_resv_clusters(sb));
4087 goto failed_mount4a;
4088 }
4089
4090 err = ext4_setup_system_zone(sb);
4091 if (err) {
4092 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4093 "zone (%d)", err);
4094 goto failed_mount4a;
4095 }
4096
4097 ext4_ext_init(sb);
4098 err = ext4_mb_init(sb);
4099 if (err) {
4100 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4101 err);
4102 goto failed_mount5;
4103 }
4104
4105 block = ext4_count_free_clusters(sb);
4106 ext4_free_blocks_count_set(sbi->s_es,
4107 EXT4_C2B(sbi, block));
4108 err = percpu_counter_init(&sbi->s_freeclusters_counter, block);
4109 if (!err) {
4110 unsigned long freei = ext4_count_free_inodes(sb);
4111 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4112 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei);
4113 }
4114 if (!err)
4115 err = percpu_counter_init(&sbi->s_dirs_counter,
4116 ext4_count_dirs(sb));
4117 if (!err)
4118 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
4119 if (err) {
4120 ext4_msg(sb, KERN_ERR, "insufficient memory");
4121 goto failed_mount6;
4122 }
4123
4124 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4125 if (!ext4_fill_flex_info(sb)) {
4126 ext4_msg(sb, KERN_ERR,
4127 "unable to initialize "
4128 "flex_bg meta info!");
4129 goto failed_mount6;
4130 }
4131
4132 err = ext4_register_li_request(sb, first_not_zeroed);
4133 if (err)
4134 goto failed_mount6;
4135
4136 sbi->s_kobj.kset = ext4_kset;
4137 init_completion(&sbi->s_kobj_unregister);
4138 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4139 "%s", sb->s_id);
4140 if (err)
4141 goto failed_mount7;
4142
4143 #ifdef CONFIG_QUOTA
4144 /* Enable quota usage during mount. */
4145 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4146 !(sb->s_flags & MS_RDONLY)) {
4147 err = ext4_enable_quotas(sb);
4148 if (err)
4149 goto failed_mount8;
4150 }
4151 #endif /* CONFIG_QUOTA */
4152
4153 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4154 ext4_orphan_cleanup(sb, es);
4155 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4156 if (needs_recovery) {
4157 ext4_msg(sb, KERN_INFO, "recovery complete");
4158 ext4_mark_recovery_complete(sb, es);
4159 }
4160 if (EXT4_SB(sb)->s_journal) {
4161 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4162 descr = " journalled data mode";
4163 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4164 descr = " ordered data mode";
4165 else
4166 descr = " writeback data mode";
4167 } else
4168 descr = "out journal";
4169
4170 if (test_opt(sb, DISCARD)) {
4171 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4172 if (!blk_queue_discard(q))
4173 ext4_msg(sb, KERN_WARNING,
4174 "mounting with \"discard\" option, but "
4175 "the device does not support discard");
4176 }
4177
4178 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4179 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4180 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4181
4182 if (es->s_error_count)
4183 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4184
4185 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4186 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4187 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4188 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4189
4190 kfree(orig_data);
4191 return 0;
4192
4193 cantfind_ext4:
4194 if (!silent)
4195 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4196 goto failed_mount;
4197
4198 #ifdef CONFIG_QUOTA
4199 failed_mount8:
4200 kobject_del(&sbi->s_kobj);
4201 #endif
4202 failed_mount7:
4203 ext4_unregister_li_request(sb);
4204 failed_mount6:
4205 ext4_mb_release(sb);
4206 if (sbi->s_flex_groups)
4207 ext4_kvfree(sbi->s_flex_groups);
4208 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4209 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4210 percpu_counter_destroy(&sbi->s_dirs_counter);
4211 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4212 failed_mount5:
4213 ext4_ext_release(sb);
4214 ext4_release_system_zone(sb);
4215 failed_mount4a:
4216 dput(sb->s_root);
4217 sb->s_root = NULL;
4218 failed_mount4:
4219 ext4_msg(sb, KERN_ERR, "mount failed");
4220 if (EXT4_SB(sb)->rsv_conversion_wq)
4221 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4222 failed_mount_wq:
4223 if (sbi->s_journal) {
4224 jbd2_journal_destroy(sbi->s_journal);
4225 sbi->s_journal = NULL;
4226 }
4227 failed_mount3:
4228 ext4_es_unregister_shrinker(sbi);
4229 del_timer_sync(&sbi->s_err_report);
4230 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4231 if (sbi->s_mmp_tsk)
4232 kthread_stop(sbi->s_mmp_tsk);
4233 failed_mount2:
4234 for (i = 0; i < db_count; i++)
4235 brelse(sbi->s_group_desc[i]);
4236 ext4_kvfree(sbi->s_group_desc);
4237 failed_mount:
4238 if (sbi->s_chksum_driver)
4239 crypto_free_shash(sbi->s_chksum_driver);
4240 if (sbi->s_proc) {
4241 remove_proc_entry("options", sbi->s_proc);
4242 remove_proc_entry(sb->s_id, ext4_proc_root);
4243 }
4244 #ifdef CONFIG_QUOTA
4245 for (i = 0; i < MAXQUOTAS; i++)
4246 kfree(sbi->s_qf_names[i]);
4247 #endif
4248 ext4_blkdev_remove(sbi);
4249 brelse(bh);
4250 out_fail:
4251 sb->s_fs_info = NULL;
4252 kfree(sbi->s_blockgroup_lock);
4253 kfree(sbi);
4254 out_free_orig:
4255 kfree(orig_data);
4256 return err ? err : ret;
4257 }
4258
4259 /*
4260 * Setup any per-fs journal parameters now. We'll do this both on
4261 * initial mount, once the journal has been initialised but before we've
4262 * done any recovery; and again on any subsequent remount.
4263 */
4264 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4265 {
4266 struct ext4_sb_info *sbi = EXT4_SB(sb);
4267
4268 journal->j_commit_interval = sbi->s_commit_interval;
4269 journal->j_min_batch_time = sbi->s_min_batch_time;
4270 journal->j_max_batch_time = sbi->s_max_batch_time;
4271
4272 write_lock(&journal->j_state_lock);
4273 if (test_opt(sb, BARRIER))
4274 journal->j_flags |= JBD2_BARRIER;
4275 else
4276 journal->j_flags &= ~JBD2_BARRIER;
4277 if (test_opt(sb, DATA_ERR_ABORT))
4278 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4279 else
4280 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4281 write_unlock(&journal->j_state_lock);
4282 }
4283
4284 static journal_t *ext4_get_journal(struct super_block *sb,
4285 unsigned int journal_inum)
4286 {
4287 struct inode *journal_inode;
4288 journal_t *journal;
4289
4290 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4291
4292 /* First, test for the existence of a valid inode on disk. Bad
4293 * things happen if we iget() an unused inode, as the subsequent
4294 * iput() will try to delete it. */
4295
4296 journal_inode = ext4_iget(sb, journal_inum);
4297 if (IS_ERR(journal_inode)) {
4298 ext4_msg(sb, KERN_ERR, "no journal found");
4299 return NULL;
4300 }
4301 if (!journal_inode->i_nlink) {
4302 make_bad_inode(journal_inode);
4303 iput(journal_inode);
4304 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4305 return NULL;
4306 }
4307
4308 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4309 journal_inode, journal_inode->i_size);
4310 if (!S_ISREG(journal_inode->i_mode)) {
4311 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4312 iput(journal_inode);
4313 return NULL;
4314 }
4315
4316 journal = jbd2_journal_init_inode(journal_inode);
4317 if (!journal) {
4318 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4319 iput(journal_inode);
4320 return NULL;
4321 }
4322 journal->j_private = sb;
4323 ext4_init_journal_params(sb, journal);
4324 return journal;
4325 }
4326
4327 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4328 dev_t j_dev)
4329 {
4330 struct buffer_head *bh;
4331 journal_t *journal;
4332 ext4_fsblk_t start;
4333 ext4_fsblk_t len;
4334 int hblock, blocksize;
4335 ext4_fsblk_t sb_block;
4336 unsigned long offset;
4337 struct ext4_super_block *es;
4338 struct block_device *bdev;
4339
4340 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4341
4342 bdev = ext4_blkdev_get(j_dev, sb);
4343 if (bdev == NULL)
4344 return NULL;
4345
4346 blocksize = sb->s_blocksize;
4347 hblock = bdev_logical_block_size(bdev);
4348 if (blocksize < hblock) {
4349 ext4_msg(sb, KERN_ERR,
4350 "blocksize too small for journal device");
4351 goto out_bdev;
4352 }
4353
4354 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4355 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4356 set_blocksize(bdev, blocksize);
4357 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4358 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4359 "external journal");
4360 goto out_bdev;
4361 }
4362
4363 es = (struct ext4_super_block *) (bh->b_data + offset);
4364 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4365 !(le32_to_cpu(es->s_feature_incompat) &
4366 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4367 ext4_msg(sb, KERN_ERR, "external journal has "
4368 "bad superblock");
4369 brelse(bh);
4370 goto out_bdev;
4371 }
4372
4373 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4374 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4375 brelse(bh);
4376 goto out_bdev;
4377 }
4378
4379 len = ext4_blocks_count(es);
4380 start = sb_block + 1;
4381 brelse(bh); /* we're done with the superblock */
4382
4383 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4384 start, len, blocksize);
4385 if (!journal) {
4386 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4387 goto out_bdev;
4388 }
4389 journal->j_private = sb;
4390 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4391 wait_on_buffer(journal->j_sb_buffer);
4392 if (!buffer_uptodate(journal->j_sb_buffer)) {
4393 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4394 goto out_journal;
4395 }
4396 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4397 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4398 "user (unsupported) - %d",
4399 be32_to_cpu(journal->j_superblock->s_nr_users));
4400 goto out_journal;
4401 }
4402 EXT4_SB(sb)->journal_bdev = bdev;
4403 ext4_init_journal_params(sb, journal);
4404 return journal;
4405
4406 out_journal:
4407 jbd2_journal_destroy(journal);
4408 out_bdev:
4409 ext4_blkdev_put(bdev);
4410 return NULL;
4411 }
4412
4413 static int ext4_load_journal(struct super_block *sb,
4414 struct ext4_super_block *es,
4415 unsigned long journal_devnum)
4416 {
4417 journal_t *journal;
4418 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4419 dev_t journal_dev;
4420 int err = 0;
4421 int really_read_only;
4422
4423 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4424
4425 if (journal_devnum &&
4426 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4427 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4428 "numbers have changed");
4429 journal_dev = new_decode_dev(journal_devnum);
4430 } else
4431 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4432
4433 really_read_only = bdev_read_only(sb->s_bdev);
4434
4435 /*
4436 * Are we loading a blank journal or performing recovery after a
4437 * crash? For recovery, we need to check in advance whether we
4438 * can get read-write access to the device.
4439 */
4440 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4441 if (sb->s_flags & MS_RDONLY) {
4442 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4443 "required on readonly filesystem");
4444 if (really_read_only) {
4445 ext4_msg(sb, KERN_ERR, "write access "
4446 "unavailable, cannot proceed");
4447 return -EROFS;
4448 }
4449 ext4_msg(sb, KERN_INFO, "write access will "
4450 "be enabled during recovery");
4451 }
4452 }
4453
4454 if (journal_inum && journal_dev) {
4455 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4456 "and inode journals!");
4457 return -EINVAL;
4458 }
4459
4460 if (journal_inum) {
4461 if (!(journal = ext4_get_journal(sb, journal_inum)))
4462 return -EINVAL;
4463 } else {
4464 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4465 return -EINVAL;
4466 }
4467
4468 if (!(journal->j_flags & JBD2_BARRIER))
4469 ext4_msg(sb, KERN_INFO, "barriers disabled");
4470
4471 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4472 err = jbd2_journal_wipe(journal, !really_read_only);
4473 if (!err) {
4474 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4475 if (save)
4476 memcpy(save, ((char *) es) +
4477 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4478 err = jbd2_journal_load(journal);
4479 if (save)
4480 memcpy(((char *) es) + EXT4_S_ERR_START,
4481 save, EXT4_S_ERR_LEN);
4482 kfree(save);
4483 }
4484
4485 if (err) {
4486 ext4_msg(sb, KERN_ERR, "error loading journal");
4487 jbd2_journal_destroy(journal);
4488 return err;
4489 }
4490
4491 EXT4_SB(sb)->s_journal = journal;
4492 ext4_clear_journal_err(sb, es);
4493
4494 if (!really_read_only && journal_devnum &&
4495 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4496 es->s_journal_dev = cpu_to_le32(journal_devnum);
4497
4498 /* Make sure we flush the recovery flag to disk. */
4499 ext4_commit_super(sb, 1);
4500 }
4501
4502 return 0;
4503 }
4504
4505 static int ext4_commit_super(struct super_block *sb, int sync)
4506 {
4507 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4508 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4509 int error = 0;
4510
4511 if (!sbh || block_device_ejected(sb))
4512 return error;
4513 if (buffer_write_io_error(sbh)) {
4514 /*
4515 * Oh, dear. A previous attempt to write the
4516 * superblock failed. This could happen because the
4517 * USB device was yanked out. Or it could happen to
4518 * be a transient write error and maybe the block will
4519 * be remapped. Nothing we can do but to retry the
4520 * write and hope for the best.
4521 */
4522 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4523 "superblock detected");
4524 clear_buffer_write_io_error(sbh);
4525 set_buffer_uptodate(sbh);
4526 }
4527 /*
4528 * If the file system is mounted read-only, don't update the
4529 * superblock write time. This avoids updating the superblock
4530 * write time when we are mounting the root file system
4531 * read/only but we need to replay the journal; at that point,
4532 * for people who are east of GMT and who make their clock
4533 * tick in localtime for Windows bug-for-bug compatibility,
4534 * the clock is set in the future, and this will cause e2fsck
4535 * to complain and force a full file system check.
4536 */
4537 if (!(sb->s_flags & MS_RDONLY))
4538 es->s_wtime = cpu_to_le32(get_seconds());
4539 if (sb->s_bdev->bd_part)
4540 es->s_kbytes_written =
4541 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4542 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4543 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4544 else
4545 es->s_kbytes_written =
4546 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4547 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4548 ext4_free_blocks_count_set(es,
4549 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4550 &EXT4_SB(sb)->s_freeclusters_counter)));
4551 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4552 es->s_free_inodes_count =
4553 cpu_to_le32(percpu_counter_sum_positive(
4554 &EXT4_SB(sb)->s_freeinodes_counter));
4555 BUFFER_TRACE(sbh, "marking dirty");
4556 ext4_superblock_csum_set(sb);
4557 mark_buffer_dirty(sbh);
4558 if (sync) {
4559 error = sync_dirty_buffer(sbh);
4560 if (error)
4561 return error;
4562
4563 error = buffer_write_io_error(sbh);
4564 if (error) {
4565 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4566 "superblock");
4567 clear_buffer_write_io_error(sbh);
4568 set_buffer_uptodate(sbh);
4569 }
4570 }
4571 return error;
4572 }
4573
4574 /*
4575 * Have we just finished recovery? If so, and if we are mounting (or
4576 * remounting) the filesystem readonly, then we will end up with a
4577 * consistent fs on disk. Record that fact.
4578 */
4579 static void ext4_mark_recovery_complete(struct super_block *sb,
4580 struct ext4_super_block *es)
4581 {
4582 journal_t *journal = EXT4_SB(sb)->s_journal;
4583
4584 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4585 BUG_ON(journal != NULL);
4586 return;
4587 }
4588 jbd2_journal_lock_updates(journal);
4589 if (jbd2_journal_flush(journal) < 0)
4590 goto out;
4591
4592 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4593 sb->s_flags & MS_RDONLY) {
4594 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4595 ext4_commit_super(sb, 1);
4596 }
4597
4598 out:
4599 jbd2_journal_unlock_updates(journal);
4600 }
4601
4602 /*
4603 * If we are mounting (or read-write remounting) a filesystem whose journal
4604 * has recorded an error from a previous lifetime, move that error to the
4605 * main filesystem now.
4606 */
4607 static void ext4_clear_journal_err(struct super_block *sb,
4608 struct ext4_super_block *es)
4609 {
4610 journal_t *journal;
4611 int j_errno;
4612 const char *errstr;
4613
4614 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4615
4616 journal = EXT4_SB(sb)->s_journal;
4617
4618 /*
4619 * Now check for any error status which may have been recorded in the
4620 * journal by a prior ext4_error() or ext4_abort()
4621 */
4622
4623 j_errno = jbd2_journal_errno(journal);
4624 if (j_errno) {
4625 char nbuf[16];
4626
4627 errstr = ext4_decode_error(sb, j_errno, nbuf);
4628 ext4_warning(sb, "Filesystem error recorded "
4629 "from previous mount: %s", errstr);
4630 ext4_warning(sb, "Marking fs in need of filesystem check.");
4631
4632 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4633 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4634 ext4_commit_super(sb, 1);
4635
4636 jbd2_journal_clear_err(journal);
4637 jbd2_journal_update_sb_errno(journal);
4638 }
4639 }
4640
4641 /*
4642 * Force the running and committing transactions to commit,
4643 * and wait on the commit.
4644 */
4645 int ext4_force_commit(struct super_block *sb)
4646 {
4647 journal_t *journal;
4648
4649 if (sb->s_flags & MS_RDONLY)
4650 return 0;
4651
4652 journal = EXT4_SB(sb)->s_journal;
4653 return ext4_journal_force_commit(journal);
4654 }
4655
4656 static int ext4_sync_fs(struct super_block *sb, int wait)
4657 {
4658 int ret = 0;
4659 tid_t target;
4660 bool needs_barrier = false;
4661 struct ext4_sb_info *sbi = EXT4_SB(sb);
4662
4663 trace_ext4_sync_fs(sb, wait);
4664 flush_workqueue(sbi->rsv_conversion_wq);
4665 /*
4666 * Writeback quota in non-journalled quota case - journalled quota has
4667 * no dirty dquots
4668 */
4669 dquot_writeback_dquots(sb, -1);
4670 /*
4671 * Data writeback is possible w/o journal transaction, so barrier must
4672 * being sent at the end of the function. But we can skip it if
4673 * transaction_commit will do it for us.
4674 */
4675 target = jbd2_get_latest_transaction(sbi->s_journal);
4676 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4677 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4678 needs_barrier = true;
4679
4680 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4681 if (wait)
4682 ret = jbd2_log_wait_commit(sbi->s_journal, target);
4683 }
4684 if (needs_barrier) {
4685 int err;
4686 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4687 if (!ret)
4688 ret = err;
4689 }
4690
4691 return ret;
4692 }
4693
4694 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4695 {
4696 int ret = 0;
4697
4698 trace_ext4_sync_fs(sb, wait);
4699 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4700 dquot_writeback_dquots(sb, -1);
4701 if (wait && test_opt(sb, BARRIER))
4702 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4703
4704 return ret;
4705 }
4706
4707 /*
4708 * LVM calls this function before a (read-only) snapshot is created. This
4709 * gives us a chance to flush the journal completely and mark the fs clean.
4710 *
4711 * Note that only this function cannot bring a filesystem to be in a clean
4712 * state independently. It relies on upper layer to stop all data & metadata
4713 * modifications.
4714 */
4715 static int ext4_freeze(struct super_block *sb)
4716 {
4717 int error = 0;
4718 journal_t *journal;
4719
4720 if (sb->s_flags & MS_RDONLY)
4721 return 0;
4722
4723 journal = EXT4_SB(sb)->s_journal;
4724
4725 /* Now we set up the journal barrier. */
4726 jbd2_journal_lock_updates(journal);
4727
4728 /*
4729 * Don't clear the needs_recovery flag if we failed to flush
4730 * the journal.
4731 */
4732 error = jbd2_journal_flush(journal);
4733 if (error < 0)
4734 goto out;
4735
4736 /* Journal blocked and flushed, clear needs_recovery flag. */
4737 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4738 error = ext4_commit_super(sb, 1);
4739 out:
4740 /* we rely on upper layer to stop further updates */
4741 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4742 return error;
4743 }
4744
4745 /*
4746 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4747 * flag here, even though the filesystem is not technically dirty yet.
4748 */
4749 static int ext4_unfreeze(struct super_block *sb)
4750 {
4751 if (sb->s_flags & MS_RDONLY)
4752 return 0;
4753
4754 /* Reset the needs_recovery flag before the fs is unlocked. */
4755 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4756 ext4_commit_super(sb, 1);
4757 return 0;
4758 }
4759
4760 /*
4761 * Structure to save mount options for ext4_remount's benefit
4762 */
4763 struct ext4_mount_options {
4764 unsigned long s_mount_opt;
4765 unsigned long s_mount_opt2;
4766 kuid_t s_resuid;
4767 kgid_t s_resgid;
4768 unsigned long s_commit_interval;
4769 u32 s_min_batch_time, s_max_batch_time;
4770 #ifdef CONFIG_QUOTA
4771 int s_jquota_fmt;
4772 char *s_qf_names[MAXQUOTAS];
4773 #endif
4774 };
4775
4776 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4777 {
4778 struct ext4_super_block *es;
4779 struct ext4_sb_info *sbi = EXT4_SB(sb);
4780 unsigned long old_sb_flags;
4781 struct ext4_mount_options old_opts;
4782 int enable_quota = 0;
4783 ext4_group_t g;
4784 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4785 int err = 0;
4786 #ifdef CONFIG_QUOTA
4787 int i, j;
4788 #endif
4789 char *orig_data = kstrdup(data, GFP_KERNEL);
4790
4791 /* Store the original options */
4792 old_sb_flags = sb->s_flags;
4793 old_opts.s_mount_opt = sbi->s_mount_opt;
4794 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4795 old_opts.s_resuid = sbi->s_resuid;
4796 old_opts.s_resgid = sbi->s_resgid;
4797 old_opts.s_commit_interval = sbi->s_commit_interval;
4798 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4799 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4800 #ifdef CONFIG_QUOTA
4801 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4802 for (i = 0; i < MAXQUOTAS; i++)
4803 if (sbi->s_qf_names[i]) {
4804 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4805 GFP_KERNEL);
4806 if (!old_opts.s_qf_names[i]) {
4807 for (j = 0; j < i; j++)
4808 kfree(old_opts.s_qf_names[j]);
4809 kfree(orig_data);
4810 return -ENOMEM;
4811 }
4812 } else
4813 old_opts.s_qf_names[i] = NULL;
4814 #endif
4815 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4816 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4817
4818 /*
4819 * Allow the "check" option to be passed as a remount option.
4820 */
4821 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4822 err = -EINVAL;
4823 goto restore_opts;
4824 }
4825
4826 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4827 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4828 ext4_msg(sb, KERN_ERR, "can't mount with "
4829 "both data=journal and delalloc");
4830 err = -EINVAL;
4831 goto restore_opts;
4832 }
4833 if (test_opt(sb, DIOREAD_NOLOCK)) {
4834 ext4_msg(sb, KERN_ERR, "can't mount with "
4835 "both data=journal and dioread_nolock");
4836 err = -EINVAL;
4837 goto restore_opts;
4838 }
4839 }
4840
4841 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4842 ext4_abort(sb, "Abort forced by user");
4843
4844 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4845 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4846
4847 es = sbi->s_es;
4848
4849 if (sbi->s_journal) {
4850 ext4_init_journal_params(sb, sbi->s_journal);
4851 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4852 }
4853
4854 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4855 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4856 err = -EROFS;
4857 goto restore_opts;
4858 }
4859
4860 if (*flags & MS_RDONLY) {
4861 err = sync_filesystem(sb);
4862 if (err < 0)
4863 goto restore_opts;
4864 err = dquot_suspend(sb, -1);
4865 if (err < 0)
4866 goto restore_opts;
4867
4868 /*
4869 * First of all, the unconditional stuff we have to do
4870 * to disable replay of the journal when we next remount
4871 */
4872 sb->s_flags |= MS_RDONLY;
4873
4874 /*
4875 * OK, test if we are remounting a valid rw partition
4876 * readonly, and if so set the rdonly flag and then
4877 * mark the partition as valid again.
4878 */
4879 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4880 (sbi->s_mount_state & EXT4_VALID_FS))
4881 es->s_state = cpu_to_le16(sbi->s_mount_state);
4882
4883 if (sbi->s_journal)
4884 ext4_mark_recovery_complete(sb, es);
4885 } else {
4886 /* Make sure we can mount this feature set readwrite */
4887 if (!ext4_feature_set_ok(sb, 0)) {
4888 err = -EROFS;
4889 goto restore_opts;
4890 }
4891 /*
4892 * Make sure the group descriptor checksums
4893 * are sane. If they aren't, refuse to remount r/w.
4894 */
4895 for (g = 0; g < sbi->s_groups_count; g++) {
4896 struct ext4_group_desc *gdp =
4897 ext4_get_group_desc(sb, g, NULL);
4898
4899 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4900 ext4_msg(sb, KERN_ERR,
4901 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4902 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4903 le16_to_cpu(gdp->bg_checksum));
4904 err = -EINVAL;
4905 goto restore_opts;
4906 }
4907 }
4908
4909 /*
4910 * If we have an unprocessed orphan list hanging
4911 * around from a previously readonly bdev mount,
4912 * require a full umount/remount for now.
4913 */
4914 if (es->s_last_orphan) {
4915 ext4_msg(sb, KERN_WARNING, "Couldn't "
4916 "remount RDWR because of unprocessed "
4917 "orphan inode list. Please "
4918 "umount/remount instead");
4919 err = -EINVAL;
4920 goto restore_opts;
4921 }
4922
4923 /*
4924 * Mounting a RDONLY partition read-write, so reread
4925 * and store the current valid flag. (It may have
4926 * been changed by e2fsck since we originally mounted
4927 * the partition.)
4928 */
4929 if (sbi->s_journal)
4930 ext4_clear_journal_err(sb, es);
4931 sbi->s_mount_state = le16_to_cpu(es->s_state);
4932 if (!ext4_setup_super(sb, es, 0))
4933 sb->s_flags &= ~MS_RDONLY;
4934 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4935 EXT4_FEATURE_INCOMPAT_MMP))
4936 if (ext4_multi_mount_protect(sb,
4937 le64_to_cpu(es->s_mmp_block))) {
4938 err = -EROFS;
4939 goto restore_opts;
4940 }
4941 enable_quota = 1;
4942 }
4943 }
4944
4945 /*
4946 * Reinitialize lazy itable initialization thread based on
4947 * current settings
4948 */
4949 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4950 ext4_unregister_li_request(sb);
4951 else {
4952 ext4_group_t first_not_zeroed;
4953 first_not_zeroed = ext4_has_uninit_itable(sb);
4954 ext4_register_li_request(sb, first_not_zeroed);
4955 }
4956
4957 ext4_setup_system_zone(sb);
4958 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4959 ext4_commit_super(sb, 1);
4960
4961 #ifdef CONFIG_QUOTA
4962 /* Release old quota file names */
4963 for (i = 0; i < MAXQUOTAS; i++)
4964 kfree(old_opts.s_qf_names[i]);
4965 if (enable_quota) {
4966 if (sb_any_quota_suspended(sb))
4967 dquot_resume(sb, -1);
4968 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4969 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4970 err = ext4_enable_quotas(sb);
4971 if (err)
4972 goto restore_opts;
4973 }
4974 }
4975 #endif
4976
4977 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4978 kfree(orig_data);
4979 return 0;
4980
4981 restore_opts:
4982 sb->s_flags = old_sb_flags;
4983 sbi->s_mount_opt = old_opts.s_mount_opt;
4984 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4985 sbi->s_resuid = old_opts.s_resuid;
4986 sbi->s_resgid = old_opts.s_resgid;
4987 sbi->s_commit_interval = old_opts.s_commit_interval;
4988 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4989 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4990 #ifdef CONFIG_QUOTA
4991 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4992 for (i = 0; i < MAXQUOTAS; i++) {
4993 kfree(sbi->s_qf_names[i]);
4994 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4995 }
4996 #endif
4997 kfree(orig_data);
4998 return err;
4999 }
5000
5001 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5002 {
5003 struct super_block *sb = dentry->d_sb;
5004 struct ext4_sb_info *sbi = EXT4_SB(sb);
5005 struct ext4_super_block *es = sbi->s_es;
5006 ext4_fsblk_t overhead = 0, resv_blocks;
5007 u64 fsid;
5008 s64 bfree;
5009 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5010
5011 if (!test_opt(sb, MINIX_DF))
5012 overhead = sbi->s_overhead;
5013
5014 buf->f_type = EXT4_SUPER_MAGIC;
5015 buf->f_bsize = sb->s_blocksize;
5016 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5017 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5018 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5019 /* prevent underflow in case that few free space is available */
5020 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5021 buf->f_bavail = buf->f_bfree -
5022 (ext4_r_blocks_count(es) + resv_blocks);
5023 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5024 buf->f_bavail = 0;
5025 buf->f_files = le32_to_cpu(es->s_inodes_count);
5026 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5027 buf->f_namelen = EXT4_NAME_LEN;
5028 fsid = le64_to_cpup((void *)es->s_uuid) ^
5029 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5030 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5031 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5032
5033 return 0;
5034 }
5035
5036 /* Helper function for writing quotas on sync - we need to start transaction
5037 * before quota file is locked for write. Otherwise the are possible deadlocks:
5038 * Process 1 Process 2
5039 * ext4_create() quota_sync()
5040 * jbd2_journal_start() write_dquot()
5041 * dquot_initialize() down(dqio_mutex)
5042 * down(dqio_mutex) jbd2_journal_start()
5043 *
5044 */
5045
5046 #ifdef CONFIG_QUOTA
5047
5048 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5049 {
5050 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5051 }
5052
5053 static int ext4_write_dquot(struct dquot *dquot)
5054 {
5055 int ret, err;
5056 handle_t *handle;
5057 struct inode *inode;
5058
5059 inode = dquot_to_inode(dquot);
5060 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5061 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5062 if (IS_ERR(handle))
5063 return PTR_ERR(handle);
5064 ret = dquot_commit(dquot);
5065 err = ext4_journal_stop(handle);
5066 if (!ret)
5067 ret = err;
5068 return ret;
5069 }
5070
5071 static int ext4_acquire_dquot(struct dquot *dquot)
5072 {
5073 int ret, err;
5074 handle_t *handle;
5075
5076 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5077 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5078 if (IS_ERR(handle))
5079 return PTR_ERR(handle);
5080 ret = dquot_acquire(dquot);
5081 err = ext4_journal_stop(handle);
5082 if (!ret)
5083 ret = err;
5084 return ret;
5085 }
5086
5087 static int ext4_release_dquot(struct dquot *dquot)
5088 {
5089 int ret, err;
5090 handle_t *handle;
5091
5092 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5093 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5094 if (IS_ERR(handle)) {
5095 /* Release dquot anyway to avoid endless cycle in dqput() */
5096 dquot_release(dquot);
5097 return PTR_ERR(handle);
5098 }
5099 ret = dquot_release(dquot);
5100 err = ext4_journal_stop(handle);
5101 if (!ret)
5102 ret = err;
5103 return ret;
5104 }
5105
5106 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5107 {
5108 struct super_block *sb = dquot->dq_sb;
5109 struct ext4_sb_info *sbi = EXT4_SB(sb);
5110
5111 /* Are we journaling quotas? */
5112 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5113 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5114 dquot_mark_dquot_dirty(dquot);
5115 return ext4_write_dquot(dquot);
5116 } else {
5117 return dquot_mark_dquot_dirty(dquot);
5118 }
5119 }
5120
5121 static int ext4_write_info(struct super_block *sb, int type)
5122 {
5123 int ret, err;
5124 handle_t *handle;
5125
5126 /* Data block + inode block */
5127 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5128 if (IS_ERR(handle))
5129 return PTR_ERR(handle);
5130 ret = dquot_commit_info(sb, type);
5131 err = ext4_journal_stop(handle);
5132 if (!ret)
5133 ret = err;
5134 return ret;
5135 }
5136
5137 /*
5138 * Turn on quotas during mount time - we need to find
5139 * the quota file and such...
5140 */
5141 static int ext4_quota_on_mount(struct super_block *sb, int type)
5142 {
5143 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5144 EXT4_SB(sb)->s_jquota_fmt, type);
5145 }
5146
5147 /*
5148 * Standard function to be called on quota_on
5149 */
5150 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5151 struct path *path)
5152 {
5153 int err;
5154
5155 if (!test_opt(sb, QUOTA))
5156 return -EINVAL;
5157
5158 /* Quotafile not on the same filesystem? */
5159 if (path->dentry->d_sb != sb)
5160 return -EXDEV;
5161 /* Journaling quota? */
5162 if (EXT4_SB(sb)->s_qf_names[type]) {
5163 /* Quotafile not in fs root? */
5164 if (path->dentry->d_parent != sb->s_root)
5165 ext4_msg(sb, KERN_WARNING,
5166 "Quota file not on filesystem root. "
5167 "Journaled quota will not work");
5168 }
5169
5170 /*
5171 * When we journal data on quota file, we have to flush journal to see
5172 * all updates to the file when we bypass pagecache...
5173 */
5174 if (EXT4_SB(sb)->s_journal &&
5175 ext4_should_journal_data(path->dentry->d_inode)) {
5176 /*
5177 * We don't need to lock updates but journal_flush() could
5178 * otherwise be livelocked...
5179 */
5180 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5181 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5182 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5183 if (err)
5184 return err;
5185 }
5186
5187 return dquot_quota_on(sb, type, format_id, path);
5188 }
5189
5190 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5191 unsigned int flags)
5192 {
5193 int err;
5194 struct inode *qf_inode;
5195 unsigned long qf_inums[MAXQUOTAS] = {
5196 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5197 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5198 };
5199
5200 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5201
5202 if (!qf_inums[type])
5203 return -EPERM;
5204
5205 qf_inode = ext4_iget(sb, qf_inums[type]);
5206 if (IS_ERR(qf_inode)) {
5207 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5208 return PTR_ERR(qf_inode);
5209 }
5210
5211 /* Don't account quota for quota files to avoid recursion */
5212 qf_inode->i_flags |= S_NOQUOTA;
5213 err = dquot_enable(qf_inode, type, format_id, flags);
5214 iput(qf_inode);
5215
5216 return err;
5217 }
5218
5219 /* Enable usage tracking for all quota types. */
5220 static int ext4_enable_quotas(struct super_block *sb)
5221 {
5222 int type, err = 0;
5223 unsigned long qf_inums[MAXQUOTAS] = {
5224 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5225 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5226 };
5227
5228 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5229 for (type = 0; type < MAXQUOTAS; type++) {
5230 if (qf_inums[type]) {
5231 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5232 DQUOT_USAGE_ENABLED);
5233 if (err) {
5234 ext4_warning(sb,
5235 "Failed to enable quota tracking "
5236 "(type=%d, err=%d). Please run "
5237 "e2fsck to fix.", type, err);
5238 return err;
5239 }
5240 }
5241 }
5242 return 0;
5243 }
5244
5245 /*
5246 * quota_on function that is used when QUOTA feature is set.
5247 */
5248 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5249 int format_id)
5250 {
5251 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5252 return -EINVAL;
5253
5254 /*
5255 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5256 */
5257 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5258 }
5259
5260 static int ext4_quota_off(struct super_block *sb, int type)
5261 {
5262 struct inode *inode = sb_dqopt(sb)->files[type];
5263 handle_t *handle;
5264
5265 /* Force all delayed allocation blocks to be allocated.
5266 * Caller already holds s_umount sem */
5267 if (test_opt(sb, DELALLOC))
5268 sync_filesystem(sb);
5269
5270 if (!inode)
5271 goto out;
5272
5273 /* Update modification times of quota files when userspace can
5274 * start looking at them */
5275 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5276 if (IS_ERR(handle))
5277 goto out;
5278 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5279 ext4_mark_inode_dirty(handle, inode);
5280 ext4_journal_stop(handle);
5281
5282 out:
5283 return dquot_quota_off(sb, type);
5284 }
5285
5286 /*
5287 * quota_off function that is used when QUOTA feature is set.
5288 */
5289 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5290 {
5291 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5292 return -EINVAL;
5293
5294 /* Disable only the limits. */
5295 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5296 }
5297
5298 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5299 * acquiring the locks... As quota files are never truncated and quota code
5300 * itself serializes the operations (and no one else should touch the files)
5301 * we don't have to be afraid of races */
5302 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5303 size_t len, loff_t off)
5304 {
5305 struct inode *inode = sb_dqopt(sb)->files[type];
5306 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5307 int err = 0;
5308 int offset = off & (sb->s_blocksize - 1);
5309 int tocopy;
5310 size_t toread;
5311 struct buffer_head *bh;
5312 loff_t i_size = i_size_read(inode);
5313
5314 if (off > i_size)
5315 return 0;
5316 if (off+len > i_size)
5317 len = i_size-off;
5318 toread = len;
5319 while (toread > 0) {
5320 tocopy = sb->s_blocksize - offset < toread ?
5321 sb->s_blocksize - offset : toread;
5322 bh = ext4_bread(NULL, inode, blk, 0, &err);
5323 if (err)
5324 return err;
5325 if (!bh) /* A hole? */
5326 memset(data, 0, tocopy);
5327 else
5328 memcpy(data, bh->b_data+offset, tocopy);
5329 brelse(bh);
5330 offset = 0;
5331 toread -= tocopy;
5332 data += tocopy;
5333 blk++;
5334 }
5335 return len;
5336 }
5337
5338 /* Write to quotafile (we know the transaction is already started and has
5339 * enough credits) */
5340 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5341 const char *data, size_t len, loff_t off)
5342 {
5343 struct inode *inode = sb_dqopt(sb)->files[type];
5344 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5345 int err = 0;
5346 int offset = off & (sb->s_blocksize - 1);
5347 struct buffer_head *bh;
5348 handle_t *handle = journal_current_handle();
5349
5350 if (EXT4_SB(sb)->s_journal && !handle) {
5351 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5352 " cancelled because transaction is not started",
5353 (unsigned long long)off, (unsigned long long)len);
5354 return -EIO;
5355 }
5356 /*
5357 * Since we account only one data block in transaction credits,
5358 * then it is impossible to cross a block boundary.
5359 */
5360 if (sb->s_blocksize - offset < len) {
5361 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5362 " cancelled because not block aligned",
5363 (unsigned long long)off, (unsigned long long)len);
5364 return -EIO;
5365 }
5366
5367 bh = ext4_bread(handle, inode, blk, 1, &err);
5368 if (!bh)
5369 goto out;
5370 BUFFER_TRACE(bh, "get write access");
5371 err = ext4_journal_get_write_access(handle, bh);
5372 if (err) {
5373 brelse(bh);
5374 goto out;
5375 }
5376 lock_buffer(bh);
5377 memcpy(bh->b_data+offset, data, len);
5378 flush_dcache_page(bh->b_page);
5379 unlock_buffer(bh);
5380 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5381 brelse(bh);
5382 out:
5383 if (err)
5384 return err;
5385 if (inode->i_size < off + len) {
5386 i_size_write(inode, off + len);
5387 EXT4_I(inode)->i_disksize = inode->i_size;
5388 ext4_mark_inode_dirty(handle, inode);
5389 }
5390 return len;
5391 }
5392
5393 #endif
5394
5395 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5396 const char *dev_name, void *data)
5397 {
5398 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5399 }
5400
5401 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5402 static inline void register_as_ext2(void)
5403 {
5404 int err = register_filesystem(&ext2_fs_type);
5405 if (err)
5406 printk(KERN_WARNING
5407 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5408 }
5409
5410 static inline void unregister_as_ext2(void)
5411 {
5412 unregister_filesystem(&ext2_fs_type);
5413 }
5414
5415 static inline int ext2_feature_set_ok(struct super_block *sb)
5416 {
5417 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5418 return 0;
5419 if (sb->s_flags & MS_RDONLY)
5420 return 1;
5421 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5422 return 0;
5423 return 1;
5424 }
5425 #else
5426 static inline void register_as_ext2(void) { }
5427 static inline void unregister_as_ext2(void) { }
5428 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5429 #endif
5430
5431 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5432 static inline void register_as_ext3(void)
5433 {
5434 int err = register_filesystem(&ext3_fs_type);
5435 if (err)
5436 printk(KERN_WARNING
5437 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5438 }
5439
5440 static inline void unregister_as_ext3(void)
5441 {
5442 unregister_filesystem(&ext3_fs_type);
5443 }
5444
5445 static inline int ext3_feature_set_ok(struct super_block *sb)
5446 {
5447 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5448 return 0;
5449 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5450 return 0;
5451 if (sb->s_flags & MS_RDONLY)
5452 return 1;
5453 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5454 return 0;
5455 return 1;
5456 }
5457 #else
5458 static inline void register_as_ext3(void) { }
5459 static inline void unregister_as_ext3(void) { }
5460 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5461 #endif
5462
5463 static struct file_system_type ext4_fs_type = {
5464 .owner = THIS_MODULE,
5465 .name = "ext4",
5466 .mount = ext4_mount,
5467 .kill_sb = kill_block_super,
5468 .fs_flags = FS_REQUIRES_DEV,
5469 };
5470 MODULE_ALIAS_FS("ext4");
5471
5472 static int __init ext4_init_feat_adverts(void)
5473 {
5474 struct ext4_features *ef;
5475 int ret = -ENOMEM;
5476
5477 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5478 if (!ef)
5479 goto out;
5480
5481 ef->f_kobj.kset = ext4_kset;
5482 init_completion(&ef->f_kobj_unregister);
5483 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5484 "features");
5485 if (ret) {
5486 kfree(ef);
5487 goto out;
5488 }
5489
5490 ext4_feat = ef;
5491 ret = 0;
5492 out:
5493 return ret;
5494 }
5495
5496 static void ext4_exit_feat_adverts(void)
5497 {
5498 kobject_put(&ext4_feat->f_kobj);
5499 wait_for_completion(&ext4_feat->f_kobj_unregister);
5500 kfree(ext4_feat);
5501 }
5502
5503 /* Shared across all ext4 file systems */
5504 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5505 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5506
5507 static int __init ext4_init_fs(void)
5508 {
5509 int i, err;
5510
5511 ext4_li_info = NULL;
5512 mutex_init(&ext4_li_mtx);
5513
5514 /* Build-time check for flags consistency */
5515 ext4_check_flag_values();
5516
5517 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5518 mutex_init(&ext4__aio_mutex[i]);
5519 init_waitqueue_head(&ext4__ioend_wq[i]);
5520 }
5521
5522 err = ext4_init_es();
5523 if (err)
5524 return err;
5525
5526 err = ext4_init_pageio();
5527 if (err)
5528 goto out7;
5529
5530 err = ext4_init_system_zone();
5531 if (err)
5532 goto out6;
5533 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5534 if (!ext4_kset) {
5535 err = -ENOMEM;
5536 goto out5;
5537 }
5538 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5539
5540 err = ext4_init_feat_adverts();
5541 if (err)
5542 goto out4;
5543
5544 err = ext4_init_mballoc();
5545 if (err)
5546 goto out2;
5547 else
5548 ext4_mballoc_ready = 1;
5549 err = init_inodecache();
5550 if (err)
5551 goto out1;
5552 register_as_ext3();
5553 register_as_ext2();
5554 err = register_filesystem(&ext4_fs_type);
5555 if (err)
5556 goto out;
5557
5558 return 0;
5559 out:
5560 unregister_as_ext2();
5561 unregister_as_ext3();
5562 destroy_inodecache();
5563 out1:
5564 ext4_mballoc_ready = 0;
5565 ext4_exit_mballoc();
5566 out2:
5567 ext4_exit_feat_adverts();
5568 out4:
5569 if (ext4_proc_root)
5570 remove_proc_entry("fs/ext4", NULL);
5571 kset_unregister(ext4_kset);
5572 out5:
5573 ext4_exit_system_zone();
5574 out6:
5575 ext4_exit_pageio();
5576 out7:
5577 ext4_exit_es();
5578
5579 return err;
5580 }
5581
5582 static void __exit ext4_exit_fs(void)
5583 {
5584 ext4_destroy_lazyinit_thread();
5585 unregister_as_ext2();
5586 unregister_as_ext3();
5587 unregister_filesystem(&ext4_fs_type);
5588 destroy_inodecache();
5589 ext4_exit_mballoc();
5590 ext4_exit_feat_adverts();
5591 remove_proc_entry("fs/ext4", NULL);
5592 kset_unregister(ext4_kset);
5593 ext4_exit_system_zone();
5594 ext4_exit_pageio();
5595 ext4_exit_es();
5596 }
5597
5598 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5599 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5600 MODULE_LICENSE("GPL");
5601 module_init(ext4_init_fs)
5602 module_exit(ext4_exit_fs)
Linux kernel - Deliverables
This page took 0.149272 seconds and 5 git commands to generate.