]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - fs/ext4/super.c
projects / mirror_ubuntu-artful-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ext4: add optimization for the lazytime mount option
[mirror_ubuntu-artful-kernel.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_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
86
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94 };
95 MODULE_ALIAS_FS("ext2");
96 MODULE_ALIAS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #else
99 #define IS_EXT2_SB(sb) (0)
100 #endif
101
102
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
106 .name = "ext3",
107 .mount = ext4_mount,
108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
110 };
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114 #else
115 #define IS_EXT3_SB(sb) (0)
116 #endif
117
118 static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
120 {
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
123 return 1;
124
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
126 }
127
128 static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
130 {
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
133 __u32 csum;
134
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136
137 return cpu_to_le32(csum);
138 }
139
140 static int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
142 {
143 if (!ext4_has_metadata_csum(sb))
144 return 1;
145
146 return es->s_checksum == ext4_superblock_csum(sb, es);
147 }
148
149 void ext4_superblock_csum_set(struct super_block *sb)
150 {
151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152
153 if (!ext4_has_metadata_csum(sb))
154 return;
155
156 es->s_checksum = ext4_superblock_csum(sb, es);
157 }
158
159 void *ext4_kvmalloc(size_t size, gfp_t flags)
160 {
161 void *ret;
162
163 ret = kmalloc(size, flags | __GFP_NOWARN);
164 if (!ret)
165 ret = __vmalloc(size, flags, PAGE_KERNEL);
166 return ret;
167 }
168
169 void *ext4_kvzalloc(size_t size, gfp_t flags)
170 {
171 void *ret;
172
173 ret = kzalloc(size, flags | __GFP_NOWARN);
174 if (!ret)
175 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
176 return ret;
177 }
178
179 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
180 struct ext4_group_desc *bg)
181 {
182 return le32_to_cpu(bg->bg_block_bitmap_lo) |
183 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
184 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
185 }
186
187 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
188 struct ext4_group_desc *bg)
189 {
190 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
191 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
192 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
193 }
194
195 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
196 struct ext4_group_desc *bg)
197 {
198 return le32_to_cpu(bg->bg_inode_table_lo) |
199 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
200 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
201 }
202
203 __u32 ext4_free_group_clusters(struct super_block *sb,
204 struct ext4_group_desc *bg)
205 {
206 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
209 }
210
211 __u32 ext4_free_inodes_count(struct super_block *sb,
212 struct ext4_group_desc *bg)
213 {
214 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
217 }
218
219 __u32 ext4_used_dirs_count(struct super_block *sb,
220 struct ext4_group_desc *bg)
221 {
222 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
225 }
226
227 __u32 ext4_itable_unused_count(struct super_block *sb,
228 struct ext4_group_desc *bg)
229 {
230 return le16_to_cpu(bg->bg_itable_unused_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
233 }
234
235 void ext4_block_bitmap_set(struct super_block *sb,
236 struct ext4_group_desc *bg, ext4_fsblk_t blk)
237 {
238 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
239 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
240 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
241 }
242
243 void ext4_inode_bitmap_set(struct super_block *sb,
244 struct ext4_group_desc *bg, ext4_fsblk_t blk)
245 {
246 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
247 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
248 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
249 }
250
251 void ext4_inode_table_set(struct super_block *sb,
252 struct ext4_group_desc *bg, ext4_fsblk_t blk)
253 {
254 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
255 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
256 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
257 }
258
259 void ext4_free_group_clusters_set(struct super_block *sb,
260 struct ext4_group_desc *bg, __u32 count)
261 {
262 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
265 }
266
267 void ext4_free_inodes_set(struct super_block *sb,
268 struct ext4_group_desc *bg, __u32 count)
269 {
270 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
273 }
274
275 void ext4_used_dirs_set(struct super_block *sb,
276 struct ext4_group_desc *bg, __u32 count)
277 {
278 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
281 }
282
283 void ext4_itable_unused_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
285 {
286 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
289 }
290
291
292 static void __save_error_info(struct super_block *sb, const char *func,
293 unsigned int line)
294 {
295 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
296
297 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
298 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
299 es->s_last_error_time = cpu_to_le32(get_seconds());
300 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
301 es->s_last_error_line = cpu_to_le32(line);
302 if (!es->s_first_error_time) {
303 es->s_first_error_time = es->s_last_error_time;
304 strncpy(es->s_first_error_func, func,
305 sizeof(es->s_first_error_func));
306 es->s_first_error_line = cpu_to_le32(line);
307 es->s_first_error_ino = es->s_last_error_ino;
308 es->s_first_error_block = es->s_last_error_block;
309 }
310 /*
311 * Start the daily error reporting function if it hasn't been
312 * started already
313 */
314 if (!es->s_error_count)
315 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
316 le32_add_cpu(&es->s_error_count, 1);
317 }
318
319 static void save_error_info(struct super_block *sb, const char *func,
320 unsigned int line)
321 {
322 __save_error_info(sb, func, line);
323 ext4_commit_super(sb, 1);
324 }
325
326 /*
327 * The del_gendisk() function uninitializes the disk-specific data
328 * structures, including the bdi structure, without telling anyone
329 * else. Once this happens, any attempt to call mark_buffer_dirty()
330 * (for example, by ext4_commit_super), will cause a kernel OOPS.
331 * This is a kludge to prevent these oops until we can put in a proper
332 * hook in del_gendisk() to inform the VFS and file system layers.
333 */
334 static int block_device_ejected(struct super_block *sb)
335 {
336 struct inode *bd_inode = sb->s_bdev->bd_inode;
337 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
338
339 return bdi->dev == NULL;
340 }
341
342 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
343 {
344 struct super_block *sb = journal->j_private;
345 struct ext4_sb_info *sbi = EXT4_SB(sb);
346 int error = is_journal_aborted(journal);
347 struct ext4_journal_cb_entry *jce;
348
349 BUG_ON(txn->t_state == T_FINISHED);
350 spin_lock(&sbi->s_md_lock);
351 while (!list_empty(&txn->t_private_list)) {
352 jce = list_entry(txn->t_private_list.next,
353 struct ext4_journal_cb_entry, jce_list);
354 list_del_init(&jce->jce_list);
355 spin_unlock(&sbi->s_md_lock);
356 jce->jce_func(sb, jce, error);
357 spin_lock(&sbi->s_md_lock);
358 }
359 spin_unlock(&sbi->s_md_lock);
360 }
361
362 /* Deal with the reporting of failure conditions on a filesystem such as
363 * inconsistencies detected or read IO failures.
364 *
365 * On ext2, we can store the error state of the filesystem in the
366 * superblock. That is not possible on ext4, because we may have other
367 * write ordering constraints on the superblock which prevent us from
368 * writing it out straight away; and given that the journal is about to
369 * be aborted, we can't rely on the current, or future, transactions to
370 * write out the superblock safely.
371 *
372 * We'll just use the jbd2_journal_abort() error code to record an error in
373 * the journal instead. On recovery, the journal will complain about
374 * that error until we've noted it down and cleared it.
375 */
376
377 static void ext4_handle_error(struct super_block *sb)
378 {
379 if (sb->s_flags & MS_RDONLY)
380 return;
381
382 if (!test_opt(sb, ERRORS_CONT)) {
383 journal_t *journal = EXT4_SB(sb)->s_journal;
384
385 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
386 if (journal)
387 jbd2_journal_abort(journal, -EIO);
388 }
389 if (test_opt(sb, ERRORS_RO)) {
390 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
391 /*
392 * Make sure updated value of ->s_mount_flags will be visible
393 * before ->s_flags update
394 */
395 smp_wmb();
396 sb->s_flags |= MS_RDONLY;
397 }
398 if (test_opt(sb, ERRORS_PANIC))
399 panic("EXT4-fs (device %s): panic forced after error\n",
400 sb->s_id);
401 }
402
403 #define ext4_error_ratelimit(sb) \
404 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
405 "EXT4-fs error")
406
407 void __ext4_error(struct super_block *sb, const char *function,
408 unsigned int line, const char *fmt, ...)
409 {
410 struct va_format vaf;
411 va_list args;
412
413 if (ext4_error_ratelimit(sb)) {
414 va_start(args, fmt);
415 vaf.fmt = fmt;
416 vaf.va = &args;
417 printk(KERN_CRIT
418 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
419 sb->s_id, function, line, current->comm, &vaf);
420 va_end(args);
421 }
422 save_error_info(sb, function, line);
423 ext4_handle_error(sb);
424 }
425
426 void __ext4_error_inode(struct inode *inode, const char *function,
427 unsigned int line, ext4_fsblk_t block,
428 const char *fmt, ...)
429 {
430 va_list args;
431 struct va_format vaf;
432 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
433
434 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
435 es->s_last_error_block = cpu_to_le64(block);
436 if (ext4_error_ratelimit(inode->i_sb)) {
437 va_start(args, fmt);
438 vaf.fmt = fmt;
439 vaf.va = &args;
440 if (block)
441 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
442 "inode #%lu: block %llu: comm %s: %pV\n",
443 inode->i_sb->s_id, function, line, inode->i_ino,
444 block, current->comm, &vaf);
445 else
446 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
447 "inode #%lu: comm %s: %pV\n",
448 inode->i_sb->s_id, function, line, inode->i_ino,
449 current->comm, &vaf);
450 va_end(args);
451 }
452 save_error_info(inode->i_sb, function, line);
453 ext4_handle_error(inode->i_sb);
454 }
455
456 void __ext4_error_file(struct file *file, const char *function,
457 unsigned int line, ext4_fsblk_t block,
458 const char *fmt, ...)
459 {
460 va_list args;
461 struct va_format vaf;
462 struct ext4_super_block *es;
463 struct inode *inode = file_inode(file);
464 char pathname[80], *path;
465
466 es = EXT4_SB(inode->i_sb)->s_es;
467 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
468 if (ext4_error_ratelimit(inode->i_sb)) {
469 path = d_path(&(file->f_path), pathname, sizeof(pathname));
470 if (IS_ERR(path))
471 path = "(unknown)";
472 va_start(args, fmt);
473 vaf.fmt = fmt;
474 vaf.va = &args;
475 if (block)
476 printk(KERN_CRIT
477 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
478 "block %llu: comm %s: path %s: %pV\n",
479 inode->i_sb->s_id, function, line, inode->i_ino,
480 block, current->comm, path, &vaf);
481 else
482 printk(KERN_CRIT
483 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
484 "comm %s: path %s: %pV\n",
485 inode->i_sb->s_id, function, line, inode->i_ino,
486 current->comm, path, &vaf);
487 va_end(args);
488 }
489 save_error_info(inode->i_sb, function, line);
490 ext4_handle_error(inode->i_sb);
491 }
492
493 const char *ext4_decode_error(struct super_block *sb, int errno,
494 char nbuf[16])
495 {
496 char *errstr = NULL;
497
498 switch (errno) {
499 case -EIO:
500 errstr = "IO failure";
501 break;
502 case -ENOMEM:
503 errstr = "Out of memory";
504 break;
505 case -EROFS:
506 if (!sb || (EXT4_SB(sb)->s_journal &&
507 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
508 errstr = "Journal has aborted";
509 else
510 errstr = "Readonly filesystem";
511 break;
512 default:
513 /* If the caller passed in an extra buffer for unknown
514 * errors, textualise them now. Else we just return
515 * NULL. */
516 if (nbuf) {
517 /* Check for truncated error codes... */
518 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
519 errstr = nbuf;
520 }
521 break;
522 }
523
524 return errstr;
525 }
526
527 /* __ext4_std_error decodes expected errors from journaling functions
528 * automatically and invokes the appropriate error response. */
529
530 void __ext4_std_error(struct super_block *sb, const char *function,
531 unsigned int line, int errno)
532 {
533 char nbuf[16];
534 const char *errstr;
535
536 /* Special case: if the error is EROFS, and we're not already
537 * inside a transaction, then there's really no point in logging
538 * an error. */
539 if (errno == -EROFS && journal_current_handle() == NULL &&
540 (sb->s_flags & MS_RDONLY))
541 return;
542
543 if (ext4_error_ratelimit(sb)) {
544 errstr = ext4_decode_error(sb, errno, nbuf);
545 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
546 sb->s_id, function, line, errstr);
547 }
548
549 save_error_info(sb, function, line);
550 ext4_handle_error(sb);
551 }
552
553 /*
554 * ext4_abort is a much stronger failure handler than ext4_error. The
555 * abort function may be used to deal with unrecoverable failures such
556 * as journal IO errors or ENOMEM at a critical moment in log management.
557 *
558 * We unconditionally force the filesystem into an ABORT|READONLY state,
559 * unless the error response on the fs has been set to panic in which
560 * case we take the easy way out and panic immediately.
561 */
562
563 void __ext4_abort(struct super_block *sb, const char *function,
564 unsigned int line, const char *fmt, ...)
565 {
566 va_list args;
567
568 save_error_info(sb, function, line);
569 va_start(args, fmt);
570 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
571 function, line);
572 vprintk(fmt, args);
573 printk("\n");
574 va_end(args);
575
576 if ((sb->s_flags & MS_RDONLY) == 0) {
577 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
578 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
579 /*
580 * Make sure updated value of ->s_mount_flags will be visible
581 * before ->s_flags update
582 */
583 smp_wmb();
584 sb->s_flags |= MS_RDONLY;
585 if (EXT4_SB(sb)->s_journal)
586 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
587 save_error_info(sb, function, line);
588 }
589 if (test_opt(sb, ERRORS_PANIC))
590 panic("EXT4-fs panic from previous error\n");
591 }
592
593 void __ext4_msg(struct super_block *sb,
594 const char *prefix, const char *fmt, ...)
595 {
596 struct va_format vaf;
597 va_list args;
598
599 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
600 return;
601
602 va_start(args, fmt);
603 vaf.fmt = fmt;
604 vaf.va = &args;
605 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
606 va_end(args);
607 }
608
609 void __ext4_warning(struct super_block *sb, const char *function,
610 unsigned int line, const char *fmt, ...)
611 {
612 struct va_format vaf;
613 va_list args;
614
615 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
616 "EXT4-fs warning"))
617 return;
618
619 va_start(args, fmt);
620 vaf.fmt = fmt;
621 vaf.va = &args;
622 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
623 sb->s_id, function, line, &vaf);
624 va_end(args);
625 }
626
627 void __ext4_grp_locked_error(const char *function, unsigned int line,
628 struct super_block *sb, ext4_group_t grp,
629 unsigned long ino, ext4_fsblk_t block,
630 const char *fmt, ...)
631 __releases(bitlock)
632 __acquires(bitlock)
633 {
634 struct va_format vaf;
635 va_list args;
636 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
637
638 es->s_last_error_ino = cpu_to_le32(ino);
639 es->s_last_error_block = cpu_to_le64(block);
640 __save_error_info(sb, function, line);
641
642 if (ext4_error_ratelimit(sb)) {
643 va_start(args, fmt);
644 vaf.fmt = fmt;
645 vaf.va = &args;
646 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
647 sb->s_id, function, line, grp);
648 if (ino)
649 printk(KERN_CONT "inode %lu: ", ino);
650 if (block)
651 printk(KERN_CONT "block %llu:",
652 (unsigned long long) block);
653 printk(KERN_CONT "%pV\n", &vaf);
654 va_end(args);
655 }
656
657 if (test_opt(sb, ERRORS_CONT)) {
658 ext4_commit_super(sb, 0);
659 return;
660 }
661
662 ext4_unlock_group(sb, grp);
663 ext4_handle_error(sb);
664 /*
665 * We only get here in the ERRORS_RO case; relocking the group
666 * may be dangerous, but nothing bad will happen since the
667 * filesystem will have already been marked read/only and the
668 * journal has been aborted. We return 1 as a hint to callers
669 * who might what to use the return value from
670 * ext4_grp_locked_error() to distinguish between the
671 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
672 * aggressively from the ext4 function in question, with a
673 * more appropriate error code.
674 */
675 ext4_lock_group(sb, grp);
676 return;
677 }
678
679 void ext4_update_dynamic_rev(struct super_block *sb)
680 {
681 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
682
683 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
684 return;
685
686 ext4_warning(sb,
687 "updating to rev %d because of new feature flag, "
688 "running e2fsck is recommended",
689 EXT4_DYNAMIC_REV);
690
691 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
692 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
693 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
694 /* leave es->s_feature_*compat flags alone */
695 /* es->s_uuid will be set by e2fsck if empty */
696
697 /*
698 * The rest of the superblock fields should be zero, and if not it
699 * means they are likely already in use, so leave them alone. We
700 * can leave it up to e2fsck to clean up any inconsistencies there.
701 */
702 }
703
704 /*
705 * Open the external journal device
706 */
707 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
708 {
709 struct block_device *bdev;
710 char b[BDEVNAME_SIZE];
711
712 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
713 if (IS_ERR(bdev))
714 goto fail;
715 return bdev;
716
717 fail:
718 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
719 __bdevname(dev, b), PTR_ERR(bdev));
720 return NULL;
721 }
722
723 /*
724 * Release the journal device
725 */
726 static void ext4_blkdev_put(struct block_device *bdev)
727 {
728 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
729 }
730
731 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
732 {
733 struct block_device *bdev;
734 bdev = sbi->journal_bdev;
735 if (bdev) {
736 ext4_blkdev_put(bdev);
737 sbi->journal_bdev = NULL;
738 }
739 }
740
741 static inline struct inode *orphan_list_entry(struct list_head *l)
742 {
743 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
744 }
745
746 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
747 {
748 struct list_head *l;
749
750 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
751 le32_to_cpu(sbi->s_es->s_last_orphan));
752
753 printk(KERN_ERR "sb_info orphan list:\n");
754 list_for_each(l, &sbi->s_orphan) {
755 struct inode *inode = orphan_list_entry(l);
756 printk(KERN_ERR " "
757 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
758 inode->i_sb->s_id, inode->i_ino, inode,
759 inode->i_mode, inode->i_nlink,
760 NEXT_ORPHAN(inode));
761 }
762 }
763
764 static void ext4_put_super(struct super_block *sb)
765 {
766 struct ext4_sb_info *sbi = EXT4_SB(sb);
767 struct ext4_super_block *es = sbi->s_es;
768 int i, err;
769
770 ext4_unregister_li_request(sb);
771 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
772
773 flush_workqueue(sbi->rsv_conversion_wq);
774 destroy_workqueue(sbi->rsv_conversion_wq);
775
776 if (sbi->s_journal) {
777 err = jbd2_journal_destroy(sbi->s_journal);
778 sbi->s_journal = NULL;
779 if (err < 0)
780 ext4_abort(sb, "Couldn't clean up the journal");
781 }
782
783 ext4_es_unregister_shrinker(sbi);
784 del_timer_sync(&sbi->s_err_report);
785 ext4_release_system_zone(sb);
786 ext4_mb_release(sb);
787 ext4_ext_release(sb);
788 ext4_xattr_put_super(sb);
789
790 if (!(sb->s_flags & MS_RDONLY)) {
791 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
792 es->s_state = cpu_to_le16(sbi->s_mount_state);
793 }
794 if (!(sb->s_flags & MS_RDONLY))
795 ext4_commit_super(sb, 1);
796
797 if (sbi->s_proc) {
798 remove_proc_entry("options", sbi->s_proc);
799 remove_proc_entry(sb->s_id, ext4_proc_root);
800 }
801 kobject_del(&sbi->s_kobj);
802
803 for (i = 0; i < sbi->s_gdb_count; i++)
804 brelse(sbi->s_group_desc[i]);
805 kvfree(sbi->s_group_desc);
806 kvfree(sbi->s_flex_groups);
807 percpu_counter_destroy(&sbi->s_freeclusters_counter);
808 percpu_counter_destroy(&sbi->s_freeinodes_counter);
809 percpu_counter_destroy(&sbi->s_dirs_counter);
810 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
811 brelse(sbi->s_sbh);
812 #ifdef CONFIG_QUOTA
813 for (i = 0; i < EXT4_MAXQUOTAS; i++)
814 kfree(sbi->s_qf_names[i]);
815 #endif
816
817 /* Debugging code just in case the in-memory inode orphan list
818 * isn't empty. The on-disk one can be non-empty if we've
819 * detected an error and taken the fs readonly, but the
820 * in-memory list had better be clean by this point. */
821 if (!list_empty(&sbi->s_orphan))
822 dump_orphan_list(sb, sbi);
823 J_ASSERT(list_empty(&sbi->s_orphan));
824
825 invalidate_bdev(sb->s_bdev);
826 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
827 /*
828 * Invalidate the journal device's buffers. We don't want them
829 * floating about in memory - the physical journal device may
830 * hotswapped, and it breaks the `ro-after' testing code.
831 */
832 sync_blockdev(sbi->journal_bdev);
833 invalidate_bdev(sbi->journal_bdev);
834 ext4_blkdev_remove(sbi);
835 }
836 if (sbi->s_mb_cache) {
837 ext4_xattr_destroy_cache(sbi->s_mb_cache);
838 sbi->s_mb_cache = NULL;
839 }
840 if (sbi->s_mmp_tsk)
841 kthread_stop(sbi->s_mmp_tsk);
842 sb->s_fs_info = NULL;
843 /*
844 * Now that we are completely done shutting down the
845 * superblock, we need to actually destroy the kobject.
846 */
847 kobject_put(&sbi->s_kobj);
848 wait_for_completion(&sbi->s_kobj_unregister);
849 if (sbi->s_chksum_driver)
850 crypto_free_shash(sbi->s_chksum_driver);
851 kfree(sbi->s_blockgroup_lock);
852 kfree(sbi);
853 }
854
855 static struct kmem_cache *ext4_inode_cachep;
856
857 /*
858 * Called inside transaction, so use GFP_NOFS
859 */
860 static struct inode *ext4_alloc_inode(struct super_block *sb)
861 {
862 struct ext4_inode_info *ei;
863
864 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
865 if (!ei)
866 return NULL;
867
868 ei->vfs_inode.i_version = 1;
869 spin_lock_init(&ei->i_raw_lock);
870 INIT_LIST_HEAD(&ei->i_prealloc_list);
871 spin_lock_init(&ei->i_prealloc_lock);
872 ext4_es_init_tree(&ei->i_es_tree);
873 rwlock_init(&ei->i_es_lock);
874 INIT_LIST_HEAD(&ei->i_es_list);
875 ei->i_es_all_nr = 0;
876 ei->i_es_shk_nr = 0;
877 ei->i_es_shrink_lblk = 0;
878 ei->i_reserved_data_blocks = 0;
879 ei->i_reserved_meta_blocks = 0;
880 ei->i_allocated_meta_blocks = 0;
881 ei->i_da_metadata_calc_len = 0;
882 ei->i_da_metadata_calc_last_lblock = 0;
883 spin_lock_init(&(ei->i_block_reservation_lock));
884 #ifdef CONFIG_QUOTA
885 ei->i_reserved_quota = 0;
886 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
887 #endif
888 ei->jinode = NULL;
889 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
890 spin_lock_init(&ei->i_completed_io_lock);
891 ei->i_sync_tid = 0;
892 ei->i_datasync_tid = 0;
893 atomic_set(&ei->i_ioend_count, 0);
894 atomic_set(&ei->i_unwritten, 0);
895 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
896
897 return &ei->vfs_inode;
898 }
899
900 static int ext4_drop_inode(struct inode *inode)
901 {
902 int drop = generic_drop_inode(inode);
903
904 trace_ext4_drop_inode(inode, drop);
905 return drop;
906 }
907
908 static void ext4_i_callback(struct rcu_head *head)
909 {
910 struct inode *inode = container_of(head, struct inode, i_rcu);
911 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
912 }
913
914 static void ext4_destroy_inode(struct inode *inode)
915 {
916 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
917 ext4_msg(inode->i_sb, KERN_ERR,
918 "Inode %lu (%p): orphan list check failed!",
919 inode->i_ino, EXT4_I(inode));
920 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
921 EXT4_I(inode), sizeof(struct ext4_inode_info),
922 true);
923 dump_stack();
924 }
925 call_rcu(&inode->i_rcu, ext4_i_callback);
926 }
927
928 static void init_once(void *foo)
929 {
930 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
931
932 INIT_LIST_HEAD(&ei->i_orphan);
933 init_rwsem(&ei->xattr_sem);
934 init_rwsem(&ei->i_data_sem);
935 inode_init_once(&ei->vfs_inode);
936 }
937
938 static int __init init_inodecache(void)
939 {
940 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
941 sizeof(struct ext4_inode_info),
942 0, (SLAB_RECLAIM_ACCOUNT|
943 SLAB_MEM_SPREAD),
944 init_once);
945 if (ext4_inode_cachep == NULL)
946 return -ENOMEM;
947 return 0;
948 }
949
950 static void destroy_inodecache(void)
951 {
952 /*
953 * Make sure all delayed rcu free inodes are flushed before we
954 * destroy cache.
955 */
956 rcu_barrier();
957 kmem_cache_destroy(ext4_inode_cachep);
958 }
959
960 void ext4_clear_inode(struct inode *inode)
961 {
962 invalidate_inode_buffers(inode);
963 clear_inode(inode);
964 dquot_drop(inode);
965 ext4_discard_preallocations(inode);
966 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
967 if (EXT4_I(inode)->jinode) {
968 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
969 EXT4_I(inode)->jinode);
970 jbd2_free_inode(EXT4_I(inode)->jinode);
971 EXT4_I(inode)->jinode = NULL;
972 }
973 }
974
975 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
976 u64 ino, u32 generation)
977 {
978 struct inode *inode;
979
980 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
981 return ERR_PTR(-ESTALE);
982 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
983 return ERR_PTR(-ESTALE);
984
985 /* iget isn't really right if the inode is currently unallocated!!
986 *
987 * ext4_read_inode will return a bad_inode if the inode had been
988 * deleted, so we should be safe.
989 *
990 * Currently we don't know the generation for parent directory, so
991 * a generation of 0 means "accept any"
992 */
993 inode = ext4_iget_normal(sb, ino);
994 if (IS_ERR(inode))
995 return ERR_CAST(inode);
996 if (generation && inode->i_generation != generation) {
997 iput(inode);
998 return ERR_PTR(-ESTALE);
999 }
1000
1001 return inode;
1002 }
1003
1004 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1005 int fh_len, int fh_type)
1006 {
1007 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1008 ext4_nfs_get_inode);
1009 }
1010
1011 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1012 int fh_len, int fh_type)
1013 {
1014 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1015 ext4_nfs_get_inode);
1016 }
1017
1018 /*
1019 * Try to release metadata pages (indirect blocks, directories) which are
1020 * mapped via the block device. Since these pages could have journal heads
1021 * which would prevent try_to_free_buffers() from freeing them, we must use
1022 * jbd2 layer's try_to_free_buffers() function to release them.
1023 */
1024 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1025 gfp_t wait)
1026 {
1027 journal_t *journal = EXT4_SB(sb)->s_journal;
1028
1029 WARN_ON(PageChecked(page));
1030 if (!page_has_buffers(page))
1031 return 0;
1032 if (journal)
1033 return jbd2_journal_try_to_free_buffers(journal, page,
1034 wait & ~__GFP_WAIT);
1035 return try_to_free_buffers(page);
1036 }
1037
1038 #ifdef CONFIG_QUOTA
1039 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1040 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1041
1042 static int ext4_write_dquot(struct dquot *dquot);
1043 static int ext4_acquire_dquot(struct dquot *dquot);
1044 static int ext4_release_dquot(struct dquot *dquot);
1045 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1046 static int ext4_write_info(struct super_block *sb, int type);
1047 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1048 struct path *path);
1049 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1050 int format_id);
1051 static int ext4_quota_off(struct super_block *sb, int type);
1052 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1053 static int ext4_quota_on_mount(struct super_block *sb, int type);
1054 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1055 size_t len, loff_t off);
1056 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1057 const char *data, size_t len, loff_t off);
1058 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1059 unsigned int flags);
1060 static int ext4_enable_quotas(struct super_block *sb);
1061
1062 static struct dquot **ext4_get_dquots(struct inode *inode)
1063 {
1064 return EXT4_I(inode)->i_dquot;
1065 }
1066
1067 static const struct dquot_operations ext4_quota_operations = {
1068 .get_reserved_space = ext4_get_reserved_space,
1069 .write_dquot = ext4_write_dquot,
1070 .acquire_dquot = ext4_acquire_dquot,
1071 .release_dquot = ext4_release_dquot,
1072 .mark_dirty = ext4_mark_dquot_dirty,
1073 .write_info = ext4_write_info,
1074 .alloc_dquot = dquot_alloc,
1075 .destroy_dquot = dquot_destroy,
1076 };
1077
1078 static const struct quotactl_ops ext4_qctl_operations = {
1079 .quota_on = ext4_quota_on,
1080 .quota_off = ext4_quota_off,
1081 .quota_sync = dquot_quota_sync,
1082 .get_info = dquot_get_dqinfo,
1083 .set_info = dquot_set_dqinfo,
1084 .get_dqblk = dquot_get_dqblk,
1085 .set_dqblk = dquot_set_dqblk
1086 };
1087
1088 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1089 .quota_on_meta = ext4_quota_on_sysfile,
1090 .quota_off = ext4_quota_off_sysfile,
1091 .quota_sync = dquot_quota_sync,
1092 .get_info = dquot_get_dqinfo,
1093 .set_info = dquot_set_dqinfo,
1094 .get_dqblk = dquot_get_dqblk,
1095 .set_dqblk = dquot_set_dqblk
1096 };
1097 #endif
1098
1099 static const struct super_operations ext4_sops = {
1100 .alloc_inode = ext4_alloc_inode,
1101 .destroy_inode = ext4_destroy_inode,
1102 .write_inode = ext4_write_inode,
1103 .dirty_inode = ext4_dirty_inode,
1104 .drop_inode = ext4_drop_inode,
1105 .evict_inode = ext4_evict_inode,
1106 .put_super = ext4_put_super,
1107 .sync_fs = ext4_sync_fs,
1108 .freeze_fs = ext4_freeze,
1109 .unfreeze_fs = ext4_unfreeze,
1110 .statfs = ext4_statfs,
1111 .remount_fs = ext4_remount,
1112 .show_options = ext4_show_options,
1113 #ifdef CONFIG_QUOTA
1114 .quota_read = ext4_quota_read,
1115 .quota_write = ext4_quota_write,
1116 .get_dquots = ext4_get_dquots,
1117 #endif
1118 .bdev_try_to_free_page = bdev_try_to_free_page,
1119 };
1120
1121 static const struct export_operations ext4_export_ops = {
1122 .fh_to_dentry = ext4_fh_to_dentry,
1123 .fh_to_parent = ext4_fh_to_parent,
1124 .get_parent = ext4_get_parent,
1125 };
1126
1127 enum {
1128 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1129 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1130 Opt_nouid32, Opt_debug, Opt_removed,
1131 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1132 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1133 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1134 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1135 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1136 Opt_data_err_abort, Opt_data_err_ignore,
1137 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1138 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1139 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1140 Opt_usrquota, Opt_grpquota, Opt_i_version,
1141 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1142 Opt_lazytime, Opt_nolazytime,
1143 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1144 Opt_inode_readahead_blks, Opt_journal_ioprio,
1145 Opt_dioread_nolock, Opt_dioread_lock,
1146 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1147 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1148 };
1149
1150 static const match_table_t tokens = {
1151 {Opt_bsd_df, "bsddf"},
1152 {Opt_minix_df, "minixdf"},
1153 {Opt_grpid, "grpid"},
1154 {Opt_grpid, "bsdgroups"},
1155 {Opt_nogrpid, "nogrpid"},
1156 {Opt_nogrpid, "sysvgroups"},
1157 {Opt_resgid, "resgid=%u"},
1158 {Opt_resuid, "resuid=%u"},
1159 {Opt_sb, "sb=%u"},
1160 {Opt_err_cont, "errors=continue"},
1161 {Opt_err_panic, "errors=panic"},
1162 {Opt_err_ro, "errors=remount-ro"},
1163 {Opt_nouid32, "nouid32"},
1164 {Opt_debug, "debug"},
1165 {Opt_removed, "oldalloc"},
1166 {Opt_removed, "orlov"},
1167 {Opt_user_xattr, "user_xattr"},
1168 {Opt_nouser_xattr, "nouser_xattr"},
1169 {Opt_acl, "acl"},
1170 {Opt_noacl, "noacl"},
1171 {Opt_noload, "norecovery"},
1172 {Opt_noload, "noload"},
1173 {Opt_removed, "nobh"},
1174 {Opt_removed, "bh"},
1175 {Opt_commit, "commit=%u"},
1176 {Opt_min_batch_time, "min_batch_time=%u"},
1177 {Opt_max_batch_time, "max_batch_time=%u"},
1178 {Opt_journal_dev, "journal_dev=%u"},
1179 {Opt_journal_path, "journal_path=%s"},
1180 {Opt_journal_checksum, "journal_checksum"},
1181 {Opt_nojournal_checksum, "nojournal_checksum"},
1182 {Opt_journal_async_commit, "journal_async_commit"},
1183 {Opt_abort, "abort"},
1184 {Opt_data_journal, "data=journal"},
1185 {Opt_data_ordered, "data=ordered"},
1186 {Opt_data_writeback, "data=writeback"},
1187 {Opt_data_err_abort, "data_err=abort"},
1188 {Opt_data_err_ignore, "data_err=ignore"},
1189 {Opt_offusrjquota, "usrjquota="},
1190 {Opt_usrjquota, "usrjquota=%s"},
1191 {Opt_offgrpjquota, "grpjquota="},
1192 {Opt_grpjquota, "grpjquota=%s"},
1193 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1194 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1195 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1196 {Opt_grpquota, "grpquota"},
1197 {Opt_noquota, "noquota"},
1198 {Opt_quota, "quota"},
1199 {Opt_usrquota, "usrquota"},
1200 {Opt_barrier, "barrier=%u"},
1201 {Opt_barrier, "barrier"},
1202 {Opt_nobarrier, "nobarrier"},
1203 {Opt_i_version, "i_version"},
1204 {Opt_stripe, "stripe=%u"},
1205 {Opt_delalloc, "delalloc"},
1206 {Opt_lazytime, "lazytime"},
1207 {Opt_nolazytime, "nolazytime"},
1208 {Opt_nodelalloc, "nodelalloc"},
1209 {Opt_removed, "mblk_io_submit"},
1210 {Opt_removed, "nomblk_io_submit"},
1211 {Opt_block_validity, "block_validity"},
1212 {Opt_noblock_validity, "noblock_validity"},
1213 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1214 {Opt_journal_ioprio, "journal_ioprio=%u"},
1215 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1216 {Opt_auto_da_alloc, "auto_da_alloc"},
1217 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1218 {Opt_dioread_nolock, "dioread_nolock"},
1219 {Opt_dioread_lock, "dioread_lock"},
1220 {Opt_discard, "discard"},
1221 {Opt_nodiscard, "nodiscard"},
1222 {Opt_init_itable, "init_itable=%u"},
1223 {Opt_init_itable, "init_itable"},
1224 {Opt_noinit_itable, "noinit_itable"},
1225 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1226 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1227 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1228 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1229 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1230 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1231 {Opt_err, NULL},
1232 };
1233
1234 static ext4_fsblk_t get_sb_block(void **data)
1235 {
1236 ext4_fsblk_t sb_block;
1237 char *options = (char *) *data;
1238
1239 if (!options || strncmp(options, "sb=", 3) != 0)
1240 return 1; /* Default location */
1241
1242 options += 3;
1243 /* TODO: use simple_strtoll with >32bit ext4 */
1244 sb_block = simple_strtoul(options, &options, 0);
1245 if (*options && *options != ',') {
1246 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1247 (char *) *data);
1248 return 1;
1249 }
1250 if (*options == ',')
1251 options++;
1252 *data = (void *) options;
1253
1254 return sb_block;
1255 }
1256
1257 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1258 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1259 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1260
1261 #ifdef CONFIG_QUOTA
1262 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1263 {
1264 struct ext4_sb_info *sbi = EXT4_SB(sb);
1265 char *qname;
1266 int ret = -1;
1267
1268 if (sb_any_quota_loaded(sb) &&
1269 !sbi->s_qf_names[qtype]) {
1270 ext4_msg(sb, KERN_ERR,
1271 "Cannot change journaled "
1272 "quota options when quota turned on");
1273 return -1;
1274 }
1275 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1276 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1277 "when QUOTA feature is enabled");
1278 return -1;
1279 }
1280 qname = match_strdup(args);
1281 if (!qname) {
1282 ext4_msg(sb, KERN_ERR,
1283 "Not enough memory for storing quotafile name");
1284 return -1;
1285 }
1286 if (sbi->s_qf_names[qtype]) {
1287 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1288 ret = 1;
1289 else
1290 ext4_msg(sb, KERN_ERR,
1291 "%s quota file already specified",
1292 QTYPE2NAME(qtype));
1293 goto errout;
1294 }
1295 if (strchr(qname, '/')) {
1296 ext4_msg(sb, KERN_ERR,
1297 "quotafile must be on filesystem root");
1298 goto errout;
1299 }
1300 sbi->s_qf_names[qtype] = qname;
1301 set_opt(sb, QUOTA);
1302 return 1;
1303 errout:
1304 kfree(qname);
1305 return ret;
1306 }
1307
1308 static int clear_qf_name(struct super_block *sb, int qtype)
1309 {
1310
1311 struct ext4_sb_info *sbi = EXT4_SB(sb);
1312
1313 if (sb_any_quota_loaded(sb) &&
1314 sbi->s_qf_names[qtype]) {
1315 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1316 " when quota turned on");
1317 return -1;
1318 }
1319 kfree(sbi->s_qf_names[qtype]);
1320 sbi->s_qf_names[qtype] = NULL;
1321 return 1;
1322 }
1323 #endif
1324
1325 #define MOPT_SET 0x0001
1326 #define MOPT_CLEAR 0x0002
1327 #define MOPT_NOSUPPORT 0x0004
1328 #define MOPT_EXPLICIT 0x0008
1329 #define MOPT_CLEAR_ERR 0x0010
1330 #define MOPT_GTE0 0x0020
1331 #ifdef CONFIG_QUOTA
1332 #define MOPT_Q 0
1333 #define MOPT_QFMT 0x0040
1334 #else
1335 #define MOPT_Q MOPT_NOSUPPORT
1336 #define MOPT_QFMT MOPT_NOSUPPORT
1337 #endif
1338 #define MOPT_DATAJ 0x0080
1339 #define MOPT_NO_EXT2 0x0100
1340 #define MOPT_NO_EXT3 0x0200
1341 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1342 #define MOPT_STRING 0x0400
1343
1344 static const struct mount_opts {
1345 int token;
1346 int mount_opt;
1347 int flags;
1348 } ext4_mount_opts[] = {
1349 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1350 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1351 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1352 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1353 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1354 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1355 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1356 MOPT_EXT4_ONLY | MOPT_SET},
1357 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1358 MOPT_EXT4_ONLY | MOPT_CLEAR},
1359 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1360 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1361 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1362 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1363 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1364 MOPT_EXT4_ONLY | MOPT_CLEAR},
1365 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1366 MOPT_EXT4_ONLY | MOPT_CLEAR},
1367 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1368 MOPT_EXT4_ONLY | MOPT_SET},
1369 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1370 EXT4_MOUNT_JOURNAL_CHECKSUM),
1371 MOPT_EXT4_ONLY | MOPT_SET},
1372 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1373 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1374 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1375 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1376 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1377 MOPT_NO_EXT2 | MOPT_SET},
1378 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1379 MOPT_NO_EXT2 | MOPT_CLEAR},
1380 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1381 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1382 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1383 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1384 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1385 {Opt_commit, 0, MOPT_GTE0},
1386 {Opt_max_batch_time, 0, MOPT_GTE0},
1387 {Opt_min_batch_time, 0, MOPT_GTE0},
1388 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1389 {Opt_init_itable, 0, MOPT_GTE0},
1390 {Opt_stripe, 0, MOPT_GTE0},
1391 {Opt_resuid, 0, MOPT_GTE0},
1392 {Opt_resgid, 0, MOPT_GTE0},
1393 {Opt_journal_dev, 0, MOPT_GTE0},
1394 {Opt_journal_path, 0, MOPT_STRING},
1395 {Opt_journal_ioprio, 0, MOPT_GTE0},
1396 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1397 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1398 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1399 MOPT_NO_EXT2 | MOPT_DATAJ},
1400 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1401 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1402 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1403 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1404 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1405 #else
1406 {Opt_acl, 0, MOPT_NOSUPPORT},
1407 {Opt_noacl, 0, MOPT_NOSUPPORT},
1408 #endif
1409 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1410 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1411 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1412 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1413 MOPT_SET | MOPT_Q},
1414 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1415 MOPT_SET | MOPT_Q},
1416 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1417 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1418 {Opt_usrjquota, 0, MOPT_Q},
1419 {Opt_grpjquota, 0, MOPT_Q},
1420 {Opt_offusrjquota, 0, MOPT_Q},
1421 {Opt_offgrpjquota, 0, MOPT_Q},
1422 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1423 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1424 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1425 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1426 {Opt_err, 0, 0}
1427 };
1428
1429 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1430 substring_t *args, unsigned long *journal_devnum,
1431 unsigned int *journal_ioprio, int is_remount)
1432 {
1433 struct ext4_sb_info *sbi = EXT4_SB(sb);
1434 const struct mount_opts *m;
1435 kuid_t uid;
1436 kgid_t gid;
1437 int arg = 0;
1438
1439 #ifdef CONFIG_QUOTA
1440 if (token == Opt_usrjquota)
1441 return set_qf_name(sb, USRQUOTA, &args[0]);
1442 else if (token == Opt_grpjquota)
1443 return set_qf_name(sb, GRPQUOTA, &args[0]);
1444 else if (token == Opt_offusrjquota)
1445 return clear_qf_name(sb, USRQUOTA);
1446 else if (token == Opt_offgrpjquota)
1447 return clear_qf_name(sb, GRPQUOTA);
1448 #endif
1449 switch (token) {
1450 case Opt_noacl:
1451 case Opt_nouser_xattr:
1452 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1453 break;
1454 case Opt_sb:
1455 return 1; /* handled by get_sb_block() */
1456 case Opt_removed:
1457 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1458 return 1;
1459 case Opt_abort:
1460 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1461 return 1;
1462 case Opt_i_version:
1463 sb->s_flags |= MS_I_VERSION;
1464 return 1;
1465 case Opt_lazytime:
1466 sb->s_flags |= MS_LAZYTIME;
1467 return 1;
1468 case Opt_nolazytime:
1469 sb->s_flags &= ~MS_LAZYTIME;
1470 return 1;
1471 }
1472
1473 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1474 if (token == m->token)
1475 break;
1476
1477 if (m->token == Opt_err) {
1478 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1479 "or missing value", opt);
1480 return -1;
1481 }
1482
1483 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1484 ext4_msg(sb, KERN_ERR,
1485 "Mount option \"%s\" incompatible with ext2", opt);
1486 return -1;
1487 }
1488 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1489 ext4_msg(sb, KERN_ERR,
1490 "Mount option \"%s\" incompatible with ext3", opt);
1491 return -1;
1492 }
1493
1494 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1495 return -1;
1496 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1497 return -1;
1498 if (m->flags & MOPT_EXPLICIT)
1499 set_opt2(sb, EXPLICIT_DELALLOC);
1500 if (m->flags & MOPT_CLEAR_ERR)
1501 clear_opt(sb, ERRORS_MASK);
1502 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1503 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1504 "options when quota turned on");
1505 return -1;
1506 }
1507
1508 if (m->flags & MOPT_NOSUPPORT) {
1509 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1510 } else if (token == Opt_commit) {
1511 if (arg == 0)
1512 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1513 sbi->s_commit_interval = HZ * arg;
1514 } else if (token == Opt_max_batch_time) {
1515 sbi->s_max_batch_time = arg;
1516 } else if (token == Opt_min_batch_time) {
1517 sbi->s_min_batch_time = arg;
1518 } else if (token == Opt_inode_readahead_blks) {
1519 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1520 ext4_msg(sb, KERN_ERR,
1521 "EXT4-fs: inode_readahead_blks must be "
1522 "0 or a power of 2 smaller than 2^31");
1523 return -1;
1524 }
1525 sbi->s_inode_readahead_blks = arg;
1526 } else if (token == Opt_init_itable) {
1527 set_opt(sb, INIT_INODE_TABLE);
1528 if (!args->from)
1529 arg = EXT4_DEF_LI_WAIT_MULT;
1530 sbi->s_li_wait_mult = arg;
1531 } else if (token == Opt_max_dir_size_kb) {
1532 sbi->s_max_dir_size_kb = arg;
1533 } else if (token == Opt_stripe) {
1534 sbi->s_stripe = arg;
1535 } else if (token == Opt_resuid) {
1536 uid = make_kuid(current_user_ns(), arg);
1537 if (!uid_valid(uid)) {
1538 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1539 return -1;
1540 }
1541 sbi->s_resuid = uid;
1542 } else if (token == Opt_resgid) {
1543 gid = make_kgid(current_user_ns(), arg);
1544 if (!gid_valid(gid)) {
1545 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1546 return -1;
1547 }
1548 sbi->s_resgid = gid;
1549 } else if (token == Opt_journal_dev) {
1550 if (is_remount) {
1551 ext4_msg(sb, KERN_ERR,
1552 "Cannot specify journal on remount");
1553 return -1;
1554 }
1555 *journal_devnum = arg;
1556 } else if (token == Opt_journal_path) {
1557 char *journal_path;
1558 struct inode *journal_inode;
1559 struct path path;
1560 int error;
1561
1562 if (is_remount) {
1563 ext4_msg(sb, KERN_ERR,
1564 "Cannot specify journal on remount");
1565 return -1;
1566 }
1567 journal_path = match_strdup(&args[0]);
1568 if (!journal_path) {
1569 ext4_msg(sb, KERN_ERR, "error: could not dup "
1570 "journal device string");
1571 return -1;
1572 }
1573
1574 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1575 if (error) {
1576 ext4_msg(sb, KERN_ERR, "error: could not find "
1577 "journal device path: error %d", error);
1578 kfree(journal_path);
1579 return -1;
1580 }
1581
1582 journal_inode = path.dentry->d_inode;
1583 if (!S_ISBLK(journal_inode->i_mode)) {
1584 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1585 "is not a block device", journal_path);
1586 path_put(&path);
1587 kfree(journal_path);
1588 return -1;
1589 }
1590
1591 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1592 path_put(&path);
1593 kfree(journal_path);
1594 } else if (token == Opt_journal_ioprio) {
1595 if (arg > 7) {
1596 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1597 " (must be 0-7)");
1598 return -1;
1599 }
1600 *journal_ioprio =
1601 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1602 } else if (m->flags & MOPT_DATAJ) {
1603 if (is_remount) {
1604 if (!sbi->s_journal)
1605 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1606 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1607 ext4_msg(sb, KERN_ERR,
1608 "Cannot change data mode on remount");
1609 return -1;
1610 }
1611 } else {
1612 clear_opt(sb, DATA_FLAGS);
1613 sbi->s_mount_opt |= m->mount_opt;
1614 }
1615 #ifdef CONFIG_QUOTA
1616 } else if (m->flags & MOPT_QFMT) {
1617 if (sb_any_quota_loaded(sb) &&
1618 sbi->s_jquota_fmt != m->mount_opt) {
1619 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1620 "quota options when quota turned on");
1621 return -1;
1622 }
1623 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1624 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1625 ext4_msg(sb, KERN_ERR,
1626 "Cannot set journaled quota options "
1627 "when QUOTA feature is enabled");
1628 return -1;
1629 }
1630 sbi->s_jquota_fmt = m->mount_opt;
1631 #endif
1632 } else {
1633 if (!args->from)
1634 arg = 1;
1635 if (m->flags & MOPT_CLEAR)
1636 arg = !arg;
1637 else if (unlikely(!(m->flags & MOPT_SET))) {
1638 ext4_msg(sb, KERN_WARNING,
1639 "buggy handling of option %s", opt);
1640 WARN_ON(1);
1641 return -1;
1642 }
1643 if (arg != 0)
1644 sbi->s_mount_opt |= m->mount_opt;
1645 else
1646 sbi->s_mount_opt &= ~m->mount_opt;
1647 }
1648 return 1;
1649 }
1650
1651 static int parse_options(char *options, struct super_block *sb,
1652 unsigned long *journal_devnum,
1653 unsigned int *journal_ioprio,
1654 int is_remount)
1655 {
1656 struct ext4_sb_info *sbi = EXT4_SB(sb);
1657 char *p;
1658 substring_t args[MAX_OPT_ARGS];
1659 int token;
1660
1661 if (!options)
1662 return 1;
1663
1664 while ((p = strsep(&options, ",")) != NULL) {
1665 if (!*p)
1666 continue;
1667 /*
1668 * Initialize args struct so we know whether arg was
1669 * found; some options take optional arguments.
1670 */
1671 args[0].to = args[0].from = NULL;
1672 token = match_token(p, tokens, args);
1673 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1674 journal_ioprio, is_remount) < 0)
1675 return 0;
1676 }
1677 #ifdef CONFIG_QUOTA
1678 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1679 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1680 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1681 "feature is enabled");
1682 return 0;
1683 }
1684 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1685 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1686 clear_opt(sb, USRQUOTA);
1687
1688 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1689 clear_opt(sb, GRPQUOTA);
1690
1691 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1692 ext4_msg(sb, KERN_ERR, "old and new quota "
1693 "format mixing");
1694 return 0;
1695 }
1696
1697 if (!sbi->s_jquota_fmt) {
1698 ext4_msg(sb, KERN_ERR, "journaled quota format "
1699 "not specified");
1700 return 0;
1701 }
1702 }
1703 #endif
1704 if (test_opt(sb, DIOREAD_NOLOCK)) {
1705 int blocksize =
1706 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1707
1708 if (blocksize < PAGE_CACHE_SIZE) {
1709 ext4_msg(sb, KERN_ERR, "can't mount with "
1710 "dioread_nolock if block size != PAGE_SIZE");
1711 return 0;
1712 }
1713 }
1714 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1715 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1716 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1717 "in data=ordered mode");
1718 return 0;
1719 }
1720 return 1;
1721 }
1722
1723 static inline void ext4_show_quota_options(struct seq_file *seq,
1724 struct super_block *sb)
1725 {
1726 #if defined(CONFIG_QUOTA)
1727 struct ext4_sb_info *sbi = EXT4_SB(sb);
1728
1729 if (sbi->s_jquota_fmt) {
1730 char *fmtname = "";
1731
1732 switch (sbi->s_jquota_fmt) {
1733 case QFMT_VFS_OLD:
1734 fmtname = "vfsold";
1735 break;
1736 case QFMT_VFS_V0:
1737 fmtname = "vfsv0";
1738 break;
1739 case QFMT_VFS_V1:
1740 fmtname = "vfsv1";
1741 break;
1742 }
1743 seq_printf(seq, ",jqfmt=%s", fmtname);
1744 }
1745
1746 if (sbi->s_qf_names[USRQUOTA])
1747 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1748
1749 if (sbi->s_qf_names[GRPQUOTA])
1750 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1751 #endif
1752 }
1753
1754 static const char *token2str(int token)
1755 {
1756 const struct match_token *t;
1757
1758 for (t = tokens; t->token != Opt_err; t++)
1759 if (t->token == token && !strchr(t->pattern, '='))
1760 break;
1761 return t->pattern;
1762 }
1763
1764 /*
1765 * Show an option if
1766 * - it's set to a non-default value OR
1767 * - if the per-sb default is different from the global default
1768 */
1769 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1770 int nodefs)
1771 {
1772 struct ext4_sb_info *sbi = EXT4_SB(sb);
1773 struct ext4_super_block *es = sbi->s_es;
1774 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1775 const struct mount_opts *m;
1776 char sep = nodefs ? '\n' : ',';
1777
1778 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1779 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1780
1781 if (sbi->s_sb_block != 1)
1782 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1783
1784 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1785 int want_set = m->flags & MOPT_SET;
1786 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1787 (m->flags & MOPT_CLEAR_ERR))
1788 continue;
1789 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1790 continue; /* skip if same as the default */
1791 if ((want_set &&
1792 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1793 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1794 continue; /* select Opt_noFoo vs Opt_Foo */
1795 SEQ_OPTS_PRINT("%s", token2str(m->token));
1796 }
1797
1798 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1799 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1800 SEQ_OPTS_PRINT("resuid=%u",
1801 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1802 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1803 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1804 SEQ_OPTS_PRINT("resgid=%u",
1805 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1806 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1807 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1808 SEQ_OPTS_PUTS("errors=remount-ro");
1809 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1810 SEQ_OPTS_PUTS("errors=continue");
1811 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1812 SEQ_OPTS_PUTS("errors=panic");
1813 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1814 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1815 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1816 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1817 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1818 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1819 if (sb->s_flags & MS_I_VERSION)
1820 SEQ_OPTS_PUTS("i_version");
1821 if (nodefs || sbi->s_stripe)
1822 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1823 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1824 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1825 SEQ_OPTS_PUTS("data=journal");
1826 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1827 SEQ_OPTS_PUTS("data=ordered");
1828 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1829 SEQ_OPTS_PUTS("data=writeback");
1830 }
1831 if (nodefs ||
1832 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1833 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1834 sbi->s_inode_readahead_blks);
1835
1836 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1837 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1838 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1839 if (nodefs || sbi->s_max_dir_size_kb)
1840 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1841
1842 ext4_show_quota_options(seq, sb);
1843 return 0;
1844 }
1845
1846 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1847 {
1848 return _ext4_show_options(seq, root->d_sb, 0);
1849 }
1850
1851 static int options_seq_show(struct seq_file *seq, void *offset)
1852 {
1853 struct super_block *sb = seq->private;
1854 int rc;
1855
1856 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1857 rc = _ext4_show_options(seq, sb, 1);
1858 seq_puts(seq, "\n");
1859 return rc;
1860 }
1861
1862 static int options_open_fs(struct inode *inode, struct file *file)
1863 {
1864 return single_open(file, options_seq_show, PDE_DATA(inode));
1865 }
1866
1867 static const struct file_operations ext4_seq_options_fops = {
1868 .owner = THIS_MODULE,
1869 .open = options_open_fs,
1870 .read = seq_read,
1871 .llseek = seq_lseek,
1872 .release = single_release,
1873 };
1874
1875 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1876 int read_only)
1877 {
1878 struct ext4_sb_info *sbi = EXT4_SB(sb);
1879 int res = 0;
1880
1881 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1882 ext4_msg(sb, KERN_ERR, "revision level too high, "
1883 "forcing read-only mode");
1884 res = MS_RDONLY;
1885 }
1886 if (read_only)
1887 goto done;
1888 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1889 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1890 "running e2fsck is recommended");
1891 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1892 ext4_msg(sb, KERN_WARNING,
1893 "warning: mounting fs with errors, "
1894 "running e2fsck is recommended");
1895 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1896 le16_to_cpu(es->s_mnt_count) >=
1897 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1898 ext4_msg(sb, KERN_WARNING,
1899 "warning: maximal mount count reached, "
1900 "running e2fsck is recommended");
1901 else if (le32_to_cpu(es->s_checkinterval) &&
1902 (le32_to_cpu(es->s_lastcheck) +
1903 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1904 ext4_msg(sb, KERN_WARNING,
1905 "warning: checktime reached, "
1906 "running e2fsck is recommended");
1907 if (!sbi->s_journal)
1908 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1909 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1910 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1911 le16_add_cpu(&es->s_mnt_count, 1);
1912 es->s_mtime = cpu_to_le32(get_seconds());
1913 ext4_update_dynamic_rev(sb);
1914 if (sbi->s_journal)
1915 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1916
1917 ext4_commit_super(sb, 1);
1918 done:
1919 if (test_opt(sb, DEBUG))
1920 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1921 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1922 sb->s_blocksize,
1923 sbi->s_groups_count,
1924 EXT4_BLOCKS_PER_GROUP(sb),
1925 EXT4_INODES_PER_GROUP(sb),
1926 sbi->s_mount_opt, sbi->s_mount_opt2);
1927
1928 cleancache_init_fs(sb);
1929 return res;
1930 }
1931
1932 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1933 {
1934 struct ext4_sb_info *sbi = EXT4_SB(sb);
1935 struct flex_groups *new_groups;
1936 int size;
1937
1938 if (!sbi->s_log_groups_per_flex)
1939 return 0;
1940
1941 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1942 if (size <= sbi->s_flex_groups_allocated)
1943 return 0;
1944
1945 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1946 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1947 if (!new_groups) {
1948 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1949 size / (int) sizeof(struct flex_groups));
1950 return -ENOMEM;
1951 }
1952
1953 if (sbi->s_flex_groups) {
1954 memcpy(new_groups, sbi->s_flex_groups,
1955 (sbi->s_flex_groups_allocated *
1956 sizeof(struct flex_groups)));
1957 kvfree(sbi->s_flex_groups);
1958 }
1959 sbi->s_flex_groups = new_groups;
1960 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1961 return 0;
1962 }
1963
1964 static int ext4_fill_flex_info(struct super_block *sb)
1965 {
1966 struct ext4_sb_info *sbi = EXT4_SB(sb);
1967 struct ext4_group_desc *gdp = NULL;
1968 ext4_group_t flex_group;
1969 int i, err;
1970
1971 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1972 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1973 sbi->s_log_groups_per_flex = 0;
1974 return 1;
1975 }
1976
1977 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1978 if (err)
1979 goto failed;
1980
1981 for (i = 0; i < sbi->s_groups_count; i++) {
1982 gdp = ext4_get_group_desc(sb, i, NULL);
1983
1984 flex_group = ext4_flex_group(sbi, i);
1985 atomic_add(ext4_free_inodes_count(sb, gdp),
1986 &sbi->s_flex_groups[flex_group].free_inodes);
1987 atomic64_add(ext4_free_group_clusters(sb, gdp),
1988 &sbi->s_flex_groups[flex_group].free_clusters);
1989 atomic_add(ext4_used_dirs_count(sb, gdp),
1990 &sbi->s_flex_groups[flex_group].used_dirs);
1991 }
1992
1993 return 1;
1994 failed:
1995 return 0;
1996 }
1997
1998 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1999 struct ext4_group_desc *gdp)
2000 {
2001 int offset;
2002 __u16 crc = 0;
2003 __le32 le_group = cpu_to_le32(block_group);
2004
2005 if (ext4_has_metadata_csum(sbi->s_sb)) {
2006 /* Use new metadata_csum algorithm */
2007 __le16 save_csum;
2008 __u32 csum32;
2009
2010 save_csum = gdp->bg_checksum;
2011 gdp->bg_checksum = 0;
2012 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2013 sizeof(le_group));
2014 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2015 sbi->s_desc_size);
2016 gdp->bg_checksum = save_csum;
2017
2018 crc = csum32 & 0xFFFF;
2019 goto out;
2020 }
2021
2022 /* old crc16 code */
2023 if (!(sbi->s_es->s_feature_ro_compat &
2024 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2025 return 0;
2026
2027 offset = offsetof(struct ext4_group_desc, bg_checksum);
2028
2029 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2030 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2031 crc = crc16(crc, (__u8 *)gdp, offset);
2032 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2033 /* for checksum of struct ext4_group_desc do the rest...*/
2034 if ((sbi->s_es->s_feature_incompat &
2035 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2036 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2037 crc = crc16(crc, (__u8 *)gdp + offset,
2038 le16_to_cpu(sbi->s_es->s_desc_size) -
2039 offset);
2040
2041 out:
2042 return cpu_to_le16(crc);
2043 }
2044
2045 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2046 struct ext4_group_desc *gdp)
2047 {
2048 if (ext4_has_group_desc_csum(sb) &&
2049 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2050 block_group, gdp)))
2051 return 0;
2052
2053 return 1;
2054 }
2055
2056 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2057 struct ext4_group_desc *gdp)
2058 {
2059 if (!ext4_has_group_desc_csum(sb))
2060 return;
2061 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2062 }
2063
2064 /* Called at mount-time, super-block is locked */
2065 static int ext4_check_descriptors(struct super_block *sb,
2066 ext4_group_t *first_not_zeroed)
2067 {
2068 struct ext4_sb_info *sbi = EXT4_SB(sb);
2069 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2070 ext4_fsblk_t last_block;
2071 ext4_fsblk_t block_bitmap;
2072 ext4_fsblk_t inode_bitmap;
2073 ext4_fsblk_t inode_table;
2074 int flexbg_flag = 0;
2075 ext4_group_t i, grp = sbi->s_groups_count;
2076
2077 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2078 flexbg_flag = 1;
2079
2080 ext4_debug("Checking group descriptors");
2081
2082 for (i = 0; i < sbi->s_groups_count; i++) {
2083 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2084
2085 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2086 last_block = ext4_blocks_count(sbi->s_es) - 1;
2087 else
2088 last_block = first_block +
2089 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2090
2091 if ((grp == sbi->s_groups_count) &&
2092 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2093 grp = i;
2094
2095 block_bitmap = ext4_block_bitmap(sb, gdp);
2096 if (block_bitmap < first_block || block_bitmap > last_block) {
2097 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2098 "Block bitmap for group %u not in group "
2099 "(block %llu)!", i, block_bitmap);
2100 return 0;
2101 }
2102 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2103 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2104 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2105 "Inode bitmap for group %u not in group "
2106 "(block %llu)!", i, inode_bitmap);
2107 return 0;
2108 }
2109 inode_table = ext4_inode_table(sb, gdp);
2110 if (inode_table < first_block ||
2111 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2112 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2113 "Inode table for group %u not in group "
2114 "(block %llu)!", i, inode_table);
2115 return 0;
2116 }
2117 ext4_lock_group(sb, i);
2118 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2119 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2120 "Checksum for group %u failed (%u!=%u)",
2121 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2122 gdp)), le16_to_cpu(gdp->bg_checksum));
2123 if (!(sb->s_flags & MS_RDONLY)) {
2124 ext4_unlock_group(sb, i);
2125 return 0;
2126 }
2127 }
2128 ext4_unlock_group(sb, i);
2129 if (!flexbg_flag)
2130 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2131 }
2132 if (NULL != first_not_zeroed)
2133 *first_not_zeroed = grp;
2134 return 1;
2135 }
2136
2137 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2138 * the superblock) which were deleted from all directories, but held open by
2139 * a process at the time of a crash. We walk the list and try to delete these
2140 * inodes at recovery time (only with a read-write filesystem).
2141 *
2142 * In order to keep the orphan inode chain consistent during traversal (in
2143 * case of crash during recovery), we link each inode into the superblock
2144 * orphan list_head and handle it the same way as an inode deletion during
2145 * normal operation (which journals the operations for us).
2146 *
2147 * We only do an iget() and an iput() on each inode, which is very safe if we
2148 * accidentally point at an in-use or already deleted inode. The worst that
2149 * can happen in this case is that we get a "bit already cleared" message from
2150 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2151 * e2fsck was run on this filesystem, and it must have already done the orphan
2152 * inode cleanup for us, so we can safely abort without any further action.
2153 */
2154 static void ext4_orphan_cleanup(struct super_block *sb,
2155 struct ext4_super_block *es)
2156 {
2157 unsigned int s_flags = sb->s_flags;
2158 int nr_orphans = 0, nr_truncates = 0;
2159 #ifdef CONFIG_QUOTA
2160 int i;
2161 #endif
2162 if (!es->s_last_orphan) {
2163 jbd_debug(4, "no orphan inodes to clean up\n");
2164 return;
2165 }
2166
2167 if (bdev_read_only(sb->s_bdev)) {
2168 ext4_msg(sb, KERN_ERR, "write access "
2169 "unavailable, skipping orphan cleanup");
2170 return;
2171 }
2172
2173 /* Check if feature set would not allow a r/w mount */
2174 if (!ext4_feature_set_ok(sb, 0)) {
2175 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2176 "unknown ROCOMPAT features");
2177 return;
2178 }
2179
2180 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2181 /* don't clear list on RO mount w/ errors */
2182 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2183 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2184 "clearing orphan list.\n");
2185 es->s_last_orphan = 0;
2186 }
2187 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2188 return;
2189 }
2190
2191 if (s_flags & MS_RDONLY) {
2192 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2193 sb->s_flags &= ~MS_RDONLY;
2194 }
2195 #ifdef CONFIG_QUOTA
2196 /* Needed for iput() to work correctly and not trash data */
2197 sb->s_flags |= MS_ACTIVE;
2198 /* Turn on quotas so that they are updated correctly */
2199 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2200 if (EXT4_SB(sb)->s_qf_names[i]) {
2201 int ret = ext4_quota_on_mount(sb, i);
2202 if (ret < 0)
2203 ext4_msg(sb, KERN_ERR,
2204 "Cannot turn on journaled "
2205 "quota: error %d", ret);
2206 }
2207 }
2208 #endif
2209
2210 while (es->s_last_orphan) {
2211 struct inode *inode;
2212
2213 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2214 if (IS_ERR(inode)) {
2215 es->s_last_orphan = 0;
2216 break;
2217 }
2218
2219 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2220 dquot_initialize(inode);
2221 if (inode->i_nlink) {
2222 if (test_opt(sb, DEBUG))
2223 ext4_msg(sb, KERN_DEBUG,
2224 "%s: truncating inode %lu to %lld bytes",
2225 __func__, inode->i_ino, inode->i_size);
2226 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2227 inode->i_ino, inode->i_size);
2228 mutex_lock(&inode->i_mutex);
2229 truncate_inode_pages(inode->i_mapping, inode->i_size);
2230 ext4_truncate(inode);
2231 mutex_unlock(&inode->i_mutex);
2232 nr_truncates++;
2233 } else {
2234 if (test_opt(sb, DEBUG))
2235 ext4_msg(sb, KERN_DEBUG,
2236 "%s: deleting unreferenced inode %lu",
2237 __func__, inode->i_ino);
2238 jbd_debug(2, "deleting unreferenced inode %lu\n",
2239 inode->i_ino);
2240 nr_orphans++;
2241 }
2242 iput(inode); /* The delete magic happens here! */
2243 }
2244
2245 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2246
2247 if (nr_orphans)
2248 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2249 PLURAL(nr_orphans));
2250 if (nr_truncates)
2251 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2252 PLURAL(nr_truncates));
2253 #ifdef CONFIG_QUOTA
2254 /* Turn quotas off */
2255 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2256 if (sb_dqopt(sb)->files[i])
2257 dquot_quota_off(sb, i);
2258 }
2259 #endif
2260 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2261 }
2262
2263 /*
2264 * Maximal extent format file size.
2265 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2266 * extent format containers, within a sector_t, and within i_blocks
2267 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2268 * so that won't be a limiting factor.
2269 *
2270 * However there is other limiting factor. We do store extents in the form
2271 * of starting block and length, hence the resulting length of the extent
2272 * covering maximum file size must fit into on-disk format containers as
2273 * well. Given that length is always by 1 unit bigger than max unit (because
2274 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2275 *
2276 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2277 */
2278 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2279 {
2280 loff_t res;
2281 loff_t upper_limit = MAX_LFS_FILESIZE;
2282
2283 /* small i_blocks in vfs inode? */
2284 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2285 /*
2286 * CONFIG_LBDAF is not enabled implies the inode
2287 * i_block represent total blocks in 512 bytes
2288 * 32 == size of vfs inode i_blocks * 8
2289 */
2290 upper_limit = (1LL << 32) - 1;
2291
2292 /* total blocks in file system block size */
2293 upper_limit >>= (blkbits - 9);
2294 upper_limit <<= blkbits;
2295 }
2296
2297 /*
2298 * 32-bit extent-start container, ee_block. We lower the maxbytes
2299 * by one fs block, so ee_len can cover the extent of maximum file
2300 * size
2301 */
2302 res = (1LL << 32) - 1;
2303 res <<= blkbits;
2304
2305 /* Sanity check against vm- & vfs- imposed limits */
2306 if (res > upper_limit)
2307 res = upper_limit;
2308
2309 return res;
2310 }
2311
2312 /*
2313 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2314 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2315 * We need to be 1 filesystem block less than the 2^48 sector limit.
2316 */
2317 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2318 {
2319 loff_t res = EXT4_NDIR_BLOCKS;
2320 int meta_blocks;
2321 loff_t upper_limit;
2322 /* This is calculated to be the largest file size for a dense, block
2323 * mapped file such that the file's total number of 512-byte sectors,
2324 * including data and all indirect blocks, does not exceed (2^48 - 1).
2325 *
2326 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2327 * number of 512-byte sectors of the file.
2328 */
2329
2330 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2331 /*
2332 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2333 * the inode i_block field represents total file blocks in
2334 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2335 */
2336 upper_limit = (1LL << 32) - 1;
2337
2338 /* total blocks in file system block size */
2339 upper_limit >>= (bits - 9);
2340
2341 } else {
2342 /*
2343 * We use 48 bit ext4_inode i_blocks
2344 * With EXT4_HUGE_FILE_FL set the i_blocks
2345 * represent total number of blocks in
2346 * file system block size
2347 */
2348 upper_limit = (1LL << 48) - 1;
2349
2350 }
2351
2352 /* indirect blocks */
2353 meta_blocks = 1;
2354 /* double indirect blocks */
2355 meta_blocks += 1 + (1LL << (bits-2));
2356 /* tripple indirect blocks */
2357 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2358
2359 upper_limit -= meta_blocks;
2360 upper_limit <<= bits;
2361
2362 res += 1LL << (bits-2);
2363 res += 1LL << (2*(bits-2));
2364 res += 1LL << (3*(bits-2));
2365 res <<= bits;
2366 if (res > upper_limit)
2367 res = upper_limit;
2368
2369 if (res > MAX_LFS_FILESIZE)
2370 res = MAX_LFS_FILESIZE;
2371
2372 return res;
2373 }
2374
2375 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2376 ext4_fsblk_t logical_sb_block, int nr)
2377 {
2378 struct ext4_sb_info *sbi = EXT4_SB(sb);
2379 ext4_group_t bg, first_meta_bg;
2380 int has_super = 0;
2381
2382 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2383
2384 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2385 nr < first_meta_bg)
2386 return logical_sb_block + nr + 1;
2387 bg = sbi->s_desc_per_block * nr;
2388 if (ext4_bg_has_super(sb, bg))
2389 has_super = 1;
2390
2391 /*
2392 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2393 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2394 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2395 * compensate.
2396 */
2397 if (sb->s_blocksize == 1024 && nr == 0 &&
2398 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2399 has_super++;
2400
2401 return (has_super + ext4_group_first_block_no(sb, bg));
2402 }
2403
2404 /**
2405 * ext4_get_stripe_size: Get the stripe size.
2406 * @sbi: In memory super block info
2407 *
2408 * If we have specified it via mount option, then
2409 * use the mount option value. If the value specified at mount time is
2410 * greater than the blocks per group use the super block value.
2411 * If the super block value is greater than blocks per group return 0.
2412 * Allocator needs it be less than blocks per group.
2413 *
2414 */
2415 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2416 {
2417 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2418 unsigned long stripe_width =
2419 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2420 int ret;
2421
2422 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2423 ret = sbi->s_stripe;
2424 else if (stripe_width <= sbi->s_blocks_per_group)
2425 ret = stripe_width;
2426 else if (stride <= sbi->s_blocks_per_group)
2427 ret = stride;
2428 else
2429 ret = 0;
2430
2431 /*
2432 * If the stripe width is 1, this makes no sense and
2433 * we set it to 0 to turn off stripe handling code.
2434 */
2435 if (ret <= 1)
2436 ret = 0;
2437
2438 return ret;
2439 }
2440
2441 /* sysfs supprt */
2442
2443 struct ext4_attr {
2444 struct attribute attr;
2445 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2446 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2447 const char *, size_t);
2448 union {
2449 int offset;
2450 int deprecated_val;
2451 } u;
2452 };
2453
2454 static int parse_strtoull(const char *buf,
2455 unsigned long long max, unsigned long long *value)
2456 {
2457 int ret;
2458
2459 ret = kstrtoull(skip_spaces(buf), 0, value);
2460 if (!ret && *value > max)
2461 ret = -EINVAL;
2462 return ret;
2463 }
2464
2465 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2466 struct ext4_sb_info *sbi,
2467 char *buf)
2468 {
2469 return snprintf(buf, PAGE_SIZE, "%llu\n",
2470 (s64) EXT4_C2B(sbi,
2471 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2472 }
2473
2474 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2475 struct ext4_sb_info *sbi, char *buf)
2476 {
2477 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2478
2479 if (!sb->s_bdev->bd_part)
2480 return snprintf(buf, PAGE_SIZE, "0\n");
2481 return snprintf(buf, PAGE_SIZE, "%lu\n",
2482 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2483 sbi->s_sectors_written_start) >> 1);
2484 }
2485
2486 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2487 struct ext4_sb_info *sbi, char *buf)
2488 {
2489 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2490
2491 if (!sb->s_bdev->bd_part)
2492 return snprintf(buf, PAGE_SIZE, "0\n");
2493 return snprintf(buf, PAGE_SIZE, "%llu\n",
2494 (unsigned long long)(sbi->s_kbytes_written +
2495 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2496 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2497 }
2498
2499 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2500 struct ext4_sb_info *sbi,
2501 const char *buf, size_t count)
2502 {
2503 unsigned long t;
2504 int ret;
2505
2506 ret = kstrtoul(skip_spaces(buf), 0, &t);
2507 if (ret)
2508 return ret;
2509
2510 if (t && (!is_power_of_2(t) || t > 0x40000000))
2511 return -EINVAL;
2512
2513 sbi->s_inode_readahead_blks = t;
2514 return count;
2515 }
2516
2517 static ssize_t sbi_ui_show(struct ext4_attr *a,
2518 struct ext4_sb_info *sbi, char *buf)
2519 {
2520 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2521
2522 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2523 }
2524
2525 static ssize_t sbi_ui_store(struct ext4_attr *a,
2526 struct ext4_sb_info *sbi,
2527 const char *buf, size_t count)
2528 {
2529 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2530 unsigned long t;
2531 int ret;
2532
2533 ret = kstrtoul(skip_spaces(buf), 0, &t);
2534 if (ret)
2535 return ret;
2536 *ui = t;
2537 return count;
2538 }
2539
2540 static ssize_t es_ui_show(struct ext4_attr *a,
2541 struct ext4_sb_info *sbi, char *buf)
2542 {
2543
2544 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2545 a->u.offset);
2546
2547 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2548 }
2549
2550 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2551 struct ext4_sb_info *sbi, char *buf)
2552 {
2553 return snprintf(buf, PAGE_SIZE, "%llu\n",
2554 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2555 }
2556
2557 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2558 struct ext4_sb_info *sbi,
2559 const char *buf, size_t count)
2560 {
2561 unsigned long long val;
2562 int ret;
2563
2564 if (parse_strtoull(buf, -1ULL, &val))
2565 return -EINVAL;
2566 ret = ext4_reserve_clusters(sbi, val);
2567
2568 return ret ? ret : count;
2569 }
2570
2571 static ssize_t trigger_test_error(struct ext4_attr *a,
2572 struct ext4_sb_info *sbi,
2573 const char *buf, size_t count)
2574 {
2575 int len = count;
2576
2577 if (!capable(CAP_SYS_ADMIN))
2578 return -EPERM;
2579
2580 if (len && buf[len-1] == '\n')
2581 len--;
2582
2583 if (len)
2584 ext4_error(sbi->s_sb, "%.*s", len, buf);
2585 return count;
2586 }
2587
2588 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2589 struct ext4_sb_info *sbi, char *buf)
2590 {
2591 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2592 }
2593
2594 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2595 static struct ext4_attr ext4_attr_##_name = { \
2596 .attr = {.name = __stringify(_name), .mode = _mode }, \
2597 .show = _show, \
2598 .store = _store, \
2599 .u = { \
2600 .offset = offsetof(struct ext4_sb_info, _elname),\
2601 }, \
2602 }
2603
2604 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2605 static struct ext4_attr ext4_attr_##_name = { \
2606 .attr = {.name = __stringify(_name), .mode = _mode }, \
2607 .show = _show, \
2608 .store = _store, \
2609 .u = { \
2610 .offset = offsetof(struct ext4_super_block, _elname), \
2611 }, \
2612 }
2613
2614 #define EXT4_ATTR(name, mode, show, store) \
2615 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2616
2617 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2618 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2619 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2620
2621 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2622 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2623 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2624 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2625
2626 #define ATTR_LIST(name) &ext4_attr_##name.attr
2627 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2628 static struct ext4_attr ext4_attr_##_name = { \
2629 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2630 .show = sbi_deprecated_show, \
2631 .u = { \
2632 .deprecated_val = _val, \
2633 }, \
2634 }
2635
2636 EXT4_RO_ATTR(delayed_allocation_blocks);
2637 EXT4_RO_ATTR(session_write_kbytes);
2638 EXT4_RO_ATTR(lifetime_write_kbytes);
2639 EXT4_RW_ATTR(reserved_clusters);
2640 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2641 inode_readahead_blks_store, s_inode_readahead_blks);
2642 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2643 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2644 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2645 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2646 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2647 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2648 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2649 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2650 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2651 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2652 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2653 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2654 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2655 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2656 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2657 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2658 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2659 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2660 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2661
2662 static struct attribute *ext4_attrs[] = {
2663 ATTR_LIST(delayed_allocation_blocks),
2664 ATTR_LIST(session_write_kbytes),
2665 ATTR_LIST(lifetime_write_kbytes),
2666 ATTR_LIST(reserved_clusters),
2667 ATTR_LIST(inode_readahead_blks),
2668 ATTR_LIST(inode_goal),
2669 ATTR_LIST(mb_stats),
2670 ATTR_LIST(mb_max_to_scan),
2671 ATTR_LIST(mb_min_to_scan),
2672 ATTR_LIST(mb_order2_req),
2673 ATTR_LIST(mb_stream_req),
2674 ATTR_LIST(mb_group_prealloc),
2675 ATTR_LIST(max_writeback_mb_bump),
2676 ATTR_LIST(extent_max_zeroout_kb),
2677 ATTR_LIST(trigger_fs_error),
2678 ATTR_LIST(err_ratelimit_interval_ms),
2679 ATTR_LIST(err_ratelimit_burst),
2680 ATTR_LIST(warning_ratelimit_interval_ms),
2681 ATTR_LIST(warning_ratelimit_burst),
2682 ATTR_LIST(msg_ratelimit_interval_ms),
2683 ATTR_LIST(msg_ratelimit_burst),
2684 ATTR_LIST(errors_count),
2685 ATTR_LIST(first_error_time),
2686 ATTR_LIST(last_error_time),
2687 NULL,
2688 };
2689
2690 /* Features this copy of ext4 supports */
2691 EXT4_INFO_ATTR(lazy_itable_init);
2692 EXT4_INFO_ATTR(batched_discard);
2693 EXT4_INFO_ATTR(meta_bg_resize);
2694
2695 static struct attribute *ext4_feat_attrs[] = {
2696 ATTR_LIST(lazy_itable_init),
2697 ATTR_LIST(batched_discard),
2698 ATTR_LIST(meta_bg_resize),
2699 NULL,
2700 };
2701
2702 static ssize_t ext4_attr_show(struct kobject *kobj,
2703 struct attribute *attr, char *buf)
2704 {
2705 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2706 s_kobj);
2707 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2708
2709 return a->show ? a->show(a, sbi, buf) : 0;
2710 }
2711
2712 static ssize_t ext4_attr_store(struct kobject *kobj,
2713 struct attribute *attr,
2714 const char *buf, size_t len)
2715 {
2716 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2717 s_kobj);
2718 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2719
2720 return a->store ? a->store(a, sbi, buf, len) : 0;
2721 }
2722
2723 static void ext4_sb_release(struct kobject *kobj)
2724 {
2725 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2726 s_kobj);
2727 complete(&sbi->s_kobj_unregister);
2728 }
2729
2730 static const struct sysfs_ops ext4_attr_ops = {
2731 .show = ext4_attr_show,
2732 .store = ext4_attr_store,
2733 };
2734
2735 static struct kobj_type ext4_ktype = {
2736 .default_attrs = ext4_attrs,
2737 .sysfs_ops = &ext4_attr_ops,
2738 .release = ext4_sb_release,
2739 };
2740
2741 static void ext4_feat_release(struct kobject *kobj)
2742 {
2743 complete(&ext4_feat->f_kobj_unregister);
2744 }
2745
2746 static ssize_t ext4_feat_show(struct kobject *kobj,
2747 struct attribute *attr, char *buf)
2748 {
2749 return snprintf(buf, PAGE_SIZE, "supported\n");
2750 }
2751
2752 /*
2753 * We can not use ext4_attr_show/store because it relies on the kobject
2754 * being embedded in the ext4_sb_info structure which is definitely not
2755 * true in this case.
2756 */
2757 static const struct sysfs_ops ext4_feat_ops = {
2758 .show = ext4_feat_show,
2759 .store = NULL,
2760 };
2761
2762 static struct kobj_type ext4_feat_ktype = {
2763 .default_attrs = ext4_feat_attrs,
2764 .sysfs_ops = &ext4_feat_ops,
2765 .release = ext4_feat_release,
2766 };
2767
2768 /*
2769 * Check whether this filesystem can be mounted based on
2770 * the features present and the RDONLY/RDWR mount requested.
2771 * Returns 1 if this filesystem can be mounted as requested,
2772 * 0 if it cannot be.
2773 */
2774 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2775 {
2776 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2777 ext4_msg(sb, KERN_ERR,
2778 "Couldn't mount because of "
2779 "unsupported optional features (%x)",
2780 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2781 ~EXT4_FEATURE_INCOMPAT_SUPP));
2782 return 0;
2783 }
2784
2785 if (readonly)
2786 return 1;
2787
2788 /* Check that feature set is OK for a read-write mount */
2789 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2790 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2791 "unsupported optional features (%x)",
2792 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2793 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2794 return 0;
2795 }
2796 /*
2797 * Large file size enabled file system can only be mounted
2798 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2799 */
2800 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2801 if (sizeof(blkcnt_t) < sizeof(u64)) {
2802 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2803 "cannot be mounted RDWR without "
2804 "CONFIG_LBDAF");
2805 return 0;
2806 }
2807 }
2808 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2809 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2810 ext4_msg(sb, KERN_ERR,
2811 "Can't support bigalloc feature without "
2812 "extents feature\n");
2813 return 0;
2814 }
2815
2816 #ifndef CONFIG_QUOTA
2817 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2818 !readonly) {
2819 ext4_msg(sb, KERN_ERR,
2820 "Filesystem with quota feature cannot be mounted RDWR "
2821 "without CONFIG_QUOTA");
2822 return 0;
2823 }
2824 #endif /* CONFIG_QUOTA */
2825 return 1;
2826 }
2827
2828 /*
2829 * This function is called once a day if we have errors logged
2830 * on the file system
2831 */
2832 static void print_daily_error_info(unsigned long arg)
2833 {
2834 struct super_block *sb = (struct super_block *) arg;
2835 struct ext4_sb_info *sbi;
2836 struct ext4_super_block *es;
2837
2838 sbi = EXT4_SB(sb);
2839 es = sbi->s_es;
2840
2841 if (es->s_error_count)
2842 /* fsck newer than v1.41.13 is needed to clean this condition. */
2843 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2844 le32_to_cpu(es->s_error_count));
2845 if (es->s_first_error_time) {
2846 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2847 sb->s_id, le32_to_cpu(es->s_first_error_time),
2848 (int) sizeof(es->s_first_error_func),
2849 es->s_first_error_func,
2850 le32_to_cpu(es->s_first_error_line));
2851 if (es->s_first_error_ino)
2852 printk(": inode %u",
2853 le32_to_cpu(es->s_first_error_ino));
2854 if (es->s_first_error_block)
2855 printk(": block %llu", (unsigned long long)
2856 le64_to_cpu(es->s_first_error_block));
2857 printk("\n");
2858 }
2859 if (es->s_last_error_time) {
2860 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2861 sb->s_id, le32_to_cpu(es->s_last_error_time),
2862 (int) sizeof(es->s_last_error_func),
2863 es->s_last_error_func,
2864 le32_to_cpu(es->s_last_error_line));
2865 if (es->s_last_error_ino)
2866 printk(": inode %u",
2867 le32_to_cpu(es->s_last_error_ino));
2868 if (es->s_last_error_block)
2869 printk(": block %llu", (unsigned long long)
2870 le64_to_cpu(es->s_last_error_block));
2871 printk("\n");
2872 }
2873 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2874 }
2875
2876 /* Find next suitable group and run ext4_init_inode_table */
2877 static int ext4_run_li_request(struct ext4_li_request *elr)
2878 {
2879 struct ext4_group_desc *gdp = NULL;
2880 ext4_group_t group, ngroups;
2881 struct super_block *sb;
2882 unsigned long timeout = 0;
2883 int ret = 0;
2884
2885 sb = elr->lr_super;
2886 ngroups = EXT4_SB(sb)->s_groups_count;
2887
2888 sb_start_write(sb);
2889 for (group = elr->lr_next_group; group < ngroups; group++) {
2890 gdp = ext4_get_group_desc(sb, group, NULL);
2891 if (!gdp) {
2892 ret = 1;
2893 break;
2894 }
2895
2896 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2897 break;
2898 }
2899
2900 if (group >= ngroups)
2901 ret = 1;
2902
2903 if (!ret) {
2904 timeout = jiffies;
2905 ret = ext4_init_inode_table(sb, group,
2906 elr->lr_timeout ? 0 : 1);
2907 if (elr->lr_timeout == 0) {
2908 timeout = (jiffies - timeout) *
2909 elr->lr_sbi->s_li_wait_mult;
2910 elr->lr_timeout = timeout;
2911 }
2912 elr->lr_next_sched = jiffies + elr->lr_timeout;
2913 elr->lr_next_group = group + 1;
2914 }
2915 sb_end_write(sb);
2916
2917 return ret;
2918 }
2919
2920 /*
2921 * Remove lr_request from the list_request and free the
2922 * request structure. Should be called with li_list_mtx held
2923 */
2924 static void ext4_remove_li_request(struct ext4_li_request *elr)
2925 {
2926 struct ext4_sb_info *sbi;
2927
2928 if (!elr)
2929 return;
2930
2931 sbi = elr->lr_sbi;
2932
2933 list_del(&elr->lr_request);
2934 sbi->s_li_request = NULL;
2935 kfree(elr);
2936 }
2937
2938 static void ext4_unregister_li_request(struct super_block *sb)
2939 {
2940 mutex_lock(&ext4_li_mtx);
2941 if (!ext4_li_info) {
2942 mutex_unlock(&ext4_li_mtx);
2943 return;
2944 }
2945
2946 mutex_lock(&ext4_li_info->li_list_mtx);
2947 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2948 mutex_unlock(&ext4_li_info->li_list_mtx);
2949 mutex_unlock(&ext4_li_mtx);
2950 }
2951
2952 static struct task_struct *ext4_lazyinit_task;
2953
2954 /*
2955 * This is the function where ext4lazyinit thread lives. It walks
2956 * through the request list searching for next scheduled filesystem.
2957 * When such a fs is found, run the lazy initialization request
2958 * (ext4_rn_li_request) and keep track of the time spend in this
2959 * function. Based on that time we compute next schedule time of
2960 * the request. When walking through the list is complete, compute
2961 * next waking time and put itself into sleep.
2962 */
2963 static int ext4_lazyinit_thread(void *arg)
2964 {
2965 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2966 struct list_head *pos, *n;
2967 struct ext4_li_request *elr;
2968 unsigned long next_wakeup, cur;
2969
2970 BUG_ON(NULL == eli);
2971
2972 cont_thread:
2973 while (true) {
2974 next_wakeup = MAX_JIFFY_OFFSET;
2975
2976 mutex_lock(&eli->li_list_mtx);
2977 if (list_empty(&eli->li_request_list)) {
2978 mutex_unlock(&eli->li_list_mtx);
2979 goto exit_thread;
2980 }
2981
2982 list_for_each_safe(pos, n, &eli->li_request_list) {
2983 elr = list_entry(pos, struct ext4_li_request,
2984 lr_request);
2985
2986 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2987 if (ext4_run_li_request(elr) != 0) {
2988 /* error, remove the lazy_init job */
2989 ext4_remove_li_request(elr);
2990 continue;
2991 }
2992 }
2993
2994 if (time_before(elr->lr_next_sched, next_wakeup))
2995 next_wakeup = elr->lr_next_sched;
2996 }
2997 mutex_unlock(&eli->li_list_mtx);
2998
2999 try_to_freeze();
3000
3001 cur = jiffies;
3002 if ((time_after_eq(cur, next_wakeup)) ||
3003 (MAX_JIFFY_OFFSET == next_wakeup)) {
3004 cond_resched();
3005 continue;
3006 }
3007
3008 schedule_timeout_interruptible(next_wakeup - cur);
3009
3010 if (kthread_should_stop()) {
3011 ext4_clear_request_list();
3012 goto exit_thread;
3013 }
3014 }
3015
3016 exit_thread:
3017 /*
3018 * It looks like the request list is empty, but we need
3019 * to check it under the li_list_mtx lock, to prevent any
3020 * additions into it, and of course we should lock ext4_li_mtx
3021 * to atomically free the list and ext4_li_info, because at
3022 * this point another ext4 filesystem could be registering
3023 * new one.
3024 */
3025 mutex_lock(&ext4_li_mtx);
3026 mutex_lock(&eli->li_list_mtx);
3027 if (!list_empty(&eli->li_request_list)) {
3028 mutex_unlock(&eli->li_list_mtx);
3029 mutex_unlock(&ext4_li_mtx);
3030 goto cont_thread;
3031 }
3032 mutex_unlock(&eli->li_list_mtx);
3033 kfree(ext4_li_info);
3034 ext4_li_info = NULL;
3035 mutex_unlock(&ext4_li_mtx);
3036
3037 return 0;
3038 }
3039
3040 static void ext4_clear_request_list(void)
3041 {
3042 struct list_head *pos, *n;
3043 struct ext4_li_request *elr;
3044
3045 mutex_lock(&ext4_li_info->li_list_mtx);
3046 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3047 elr = list_entry(pos, struct ext4_li_request,
3048 lr_request);
3049 ext4_remove_li_request(elr);
3050 }
3051 mutex_unlock(&ext4_li_info->li_list_mtx);
3052 }
3053
3054 static int ext4_run_lazyinit_thread(void)
3055 {
3056 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3057 ext4_li_info, "ext4lazyinit");
3058 if (IS_ERR(ext4_lazyinit_task)) {
3059 int err = PTR_ERR(ext4_lazyinit_task);
3060 ext4_clear_request_list();
3061 kfree(ext4_li_info);
3062 ext4_li_info = NULL;
3063 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3064 "initialization thread\n",
3065 err);
3066 return err;
3067 }
3068 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3069 return 0;
3070 }
3071
3072 /*
3073 * Check whether it make sense to run itable init. thread or not.
3074 * If there is at least one uninitialized inode table, return
3075 * corresponding group number, else the loop goes through all
3076 * groups and return total number of groups.
3077 */
3078 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3079 {
3080 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3081 struct ext4_group_desc *gdp = NULL;
3082
3083 for (group = 0; group < ngroups; group++) {
3084 gdp = ext4_get_group_desc(sb, group, NULL);
3085 if (!gdp)
3086 continue;
3087
3088 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3089 break;
3090 }
3091
3092 return group;
3093 }
3094
3095 static int ext4_li_info_new(void)
3096 {
3097 struct ext4_lazy_init *eli = NULL;
3098
3099 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3100 if (!eli)
3101 return -ENOMEM;
3102
3103 INIT_LIST_HEAD(&eli->li_request_list);
3104 mutex_init(&eli->li_list_mtx);
3105
3106 eli->li_state |= EXT4_LAZYINIT_QUIT;
3107
3108 ext4_li_info = eli;
3109
3110 return 0;
3111 }
3112
3113 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3114 ext4_group_t start)
3115 {
3116 struct ext4_sb_info *sbi = EXT4_SB(sb);
3117 struct ext4_li_request *elr;
3118
3119 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3120 if (!elr)
3121 return NULL;
3122
3123 elr->lr_super = sb;
3124 elr->lr_sbi = sbi;
3125 elr->lr_next_group = start;
3126
3127 /*
3128 * Randomize first schedule time of the request to
3129 * spread the inode table initialization requests
3130 * better.
3131 */
3132 elr->lr_next_sched = jiffies + (prandom_u32() %
3133 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3134 return elr;
3135 }
3136
3137 int ext4_register_li_request(struct super_block *sb,
3138 ext4_group_t first_not_zeroed)
3139 {
3140 struct ext4_sb_info *sbi = EXT4_SB(sb);
3141 struct ext4_li_request *elr = NULL;
3142 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3143 int ret = 0;
3144
3145 mutex_lock(&ext4_li_mtx);
3146 if (sbi->s_li_request != NULL) {
3147 /*
3148 * Reset timeout so it can be computed again, because
3149 * s_li_wait_mult might have changed.
3150 */
3151 sbi->s_li_request->lr_timeout = 0;
3152 goto out;
3153 }
3154
3155 if (first_not_zeroed == ngroups ||
3156 (sb->s_flags & MS_RDONLY) ||
3157 !test_opt(sb, INIT_INODE_TABLE))
3158 goto out;
3159
3160 elr = ext4_li_request_new(sb, first_not_zeroed);
3161 if (!elr) {
3162 ret = -ENOMEM;
3163 goto out;
3164 }
3165
3166 if (NULL == ext4_li_info) {
3167 ret = ext4_li_info_new();
3168 if (ret)
3169 goto out;
3170 }
3171
3172 mutex_lock(&ext4_li_info->li_list_mtx);
3173 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3174 mutex_unlock(&ext4_li_info->li_list_mtx);
3175
3176 sbi->s_li_request = elr;
3177 /*
3178 * set elr to NULL here since it has been inserted to
3179 * the request_list and the removal and free of it is
3180 * handled by ext4_clear_request_list from now on.
3181 */
3182 elr = NULL;
3183
3184 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3185 ret = ext4_run_lazyinit_thread();
3186 if (ret)
3187 goto out;
3188 }
3189 out:
3190 mutex_unlock(&ext4_li_mtx);
3191 if (ret)
3192 kfree(elr);
3193 return ret;
3194 }
3195
3196 /*
3197 * We do not need to lock anything since this is called on
3198 * module unload.
3199 */
3200 static void ext4_destroy_lazyinit_thread(void)
3201 {
3202 /*
3203 * If thread exited earlier
3204 * there's nothing to be done.
3205 */
3206 if (!ext4_li_info || !ext4_lazyinit_task)
3207 return;
3208
3209 kthread_stop(ext4_lazyinit_task);
3210 }
3211
3212 static int set_journal_csum_feature_set(struct super_block *sb)
3213 {
3214 int ret = 1;
3215 int compat, incompat;
3216 struct ext4_sb_info *sbi = EXT4_SB(sb);
3217
3218 if (ext4_has_metadata_csum(sb)) {
3219 /* journal checksum v3 */
3220 compat = 0;
3221 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3222 } else {
3223 /* journal checksum v1 */
3224 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3225 incompat = 0;
3226 }
3227
3228 jbd2_journal_clear_features(sbi->s_journal,
3229 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3230 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3231 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3232 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3233 ret = jbd2_journal_set_features(sbi->s_journal,
3234 compat, 0,
3235 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3236 incompat);
3237 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3238 ret = jbd2_journal_set_features(sbi->s_journal,
3239 compat, 0,
3240 incompat);
3241 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3242 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3243 } else {
3244 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3245 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3246 }
3247
3248 return ret;
3249 }
3250
3251 /*
3252 * Note: calculating the overhead so we can be compatible with
3253 * historical BSD practice is quite difficult in the face of
3254 * clusters/bigalloc. This is because multiple metadata blocks from
3255 * different block group can end up in the same allocation cluster.
3256 * Calculating the exact overhead in the face of clustered allocation
3257 * requires either O(all block bitmaps) in memory or O(number of block
3258 * groups**2) in time. We will still calculate the superblock for
3259 * older file systems --- and if we come across with a bigalloc file
3260 * system with zero in s_overhead_clusters the estimate will be close to
3261 * correct especially for very large cluster sizes --- but for newer
3262 * file systems, it's better to calculate this figure once at mkfs
3263 * time, and store it in the superblock. If the superblock value is
3264 * present (even for non-bigalloc file systems), we will use it.
3265 */
3266 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3267 char *buf)
3268 {
3269 struct ext4_sb_info *sbi = EXT4_SB(sb);
3270 struct ext4_group_desc *gdp;
3271 ext4_fsblk_t first_block, last_block, b;
3272 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3273 int s, j, count = 0;
3274
3275 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3276 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3277 sbi->s_itb_per_group + 2);
3278
3279 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3280 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3281 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3282 for (i = 0; i < ngroups; i++) {
3283 gdp = ext4_get_group_desc(sb, i, NULL);
3284 b = ext4_block_bitmap(sb, gdp);
3285 if (b >= first_block && b <= last_block) {
3286 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3287 count++;
3288 }
3289 b = ext4_inode_bitmap(sb, gdp);
3290 if (b >= first_block && b <= last_block) {
3291 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3292 count++;
3293 }
3294 b = ext4_inode_table(sb, gdp);
3295 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3296 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3297 int c = EXT4_B2C(sbi, b - first_block);
3298 ext4_set_bit(c, buf);
3299 count++;
3300 }
3301 if (i != grp)
3302 continue;
3303 s = 0;
3304 if (ext4_bg_has_super(sb, grp)) {
3305 ext4_set_bit(s++, buf);
3306 count++;
3307 }
3308 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3309 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3310 count++;
3311 }
3312 }
3313 if (!count)
3314 return 0;
3315 return EXT4_CLUSTERS_PER_GROUP(sb) -
3316 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3317 }
3318
3319 /*
3320 * Compute the overhead and stash it in sbi->s_overhead
3321 */
3322 int ext4_calculate_overhead(struct super_block *sb)
3323 {
3324 struct ext4_sb_info *sbi = EXT4_SB(sb);
3325 struct ext4_super_block *es = sbi->s_es;
3326 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3327 ext4_fsblk_t overhead = 0;
3328 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3329
3330 if (!buf)
3331 return -ENOMEM;
3332
3333 /*
3334 * Compute the overhead (FS structures). This is constant
3335 * for a given filesystem unless the number of block groups
3336 * changes so we cache the previous value until it does.
3337 */
3338
3339 /*
3340 * All of the blocks before first_data_block are overhead
3341 */
3342 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3343
3344 /*
3345 * Add the overhead found in each block group
3346 */
3347 for (i = 0; i < ngroups; i++) {
3348 int blks;
3349
3350 blks = count_overhead(sb, i, buf);
3351 overhead += blks;
3352 if (blks)
3353 memset(buf, 0, PAGE_SIZE);
3354 cond_resched();
3355 }
3356 /* Add the internal journal blocks as well */
3357 if (sbi->s_journal && !sbi->journal_bdev)
3358 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3359
3360 sbi->s_overhead = overhead;
3361 smp_wmb();
3362 free_page((unsigned long) buf);
3363 return 0;
3364 }
3365
3366
3367 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3368 {
3369 ext4_fsblk_t resv_clusters;
3370
3371 /*
3372 * There's no need to reserve anything when we aren't using extents.
3373 * The space estimates are exact, there are no unwritten extents,
3374 * hole punching doesn't need new metadata... This is needed especially
3375 * to keep ext2/3 backward compatibility.
3376 */
3377 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3378 return 0;
3379 /*
3380 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3381 * This should cover the situations where we can not afford to run
3382 * out of space like for example punch hole, or converting
3383 * unwritten extents in delalloc path. In most cases such
3384 * allocation would require 1, or 2 blocks, higher numbers are
3385 * very rare.
3386 */
3387 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3388 EXT4_SB(sb)->s_cluster_bits;
3389
3390 do_div(resv_clusters, 50);
3391 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3392
3393 return resv_clusters;
3394 }
3395
3396
3397 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3398 {
3399 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3400 sbi->s_cluster_bits;
3401
3402 if (count >= clusters)
3403 return -EINVAL;
3404
3405 atomic64_set(&sbi->s_resv_clusters, count);
3406 return 0;
3407 }
3408
3409 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3410 {
3411 char *orig_data = kstrdup(data, GFP_KERNEL);
3412 struct buffer_head *bh;
3413 struct ext4_super_block *es = NULL;
3414 struct ext4_sb_info *sbi;
3415 ext4_fsblk_t block;
3416 ext4_fsblk_t sb_block = get_sb_block(&data);
3417 ext4_fsblk_t logical_sb_block;
3418 unsigned long offset = 0;
3419 unsigned long journal_devnum = 0;
3420 unsigned long def_mount_opts;
3421 struct inode *root;
3422 char *cp;
3423 const char *descr;
3424 int ret = -ENOMEM;
3425 int blocksize, clustersize;
3426 unsigned int db_count;
3427 unsigned int i;
3428 int needs_recovery, has_huge_files, has_bigalloc;
3429 __u64 blocks_count;
3430 int err = 0;
3431 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3432 ext4_group_t first_not_zeroed;
3433
3434 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3435 if (!sbi)
3436 goto out_free_orig;
3437
3438 sbi->s_blockgroup_lock =
3439 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3440 if (!sbi->s_blockgroup_lock) {
3441 kfree(sbi);
3442 goto out_free_orig;
3443 }
3444 sb->s_fs_info = sbi;
3445 sbi->s_sb = sb;
3446 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3447 sbi->s_sb_block = sb_block;
3448 if (sb->s_bdev->bd_part)
3449 sbi->s_sectors_written_start =
3450 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3451
3452 /* Cleanup superblock name */
3453 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3454 *cp = '!';
3455
3456 /* -EINVAL is default */
3457 ret = -EINVAL;
3458 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3459 if (!blocksize) {
3460 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3461 goto out_fail;
3462 }
3463
3464 /*
3465 * The ext4 superblock will not be buffer aligned for other than 1kB
3466 * block sizes. We need to calculate the offset from buffer start.
3467 */
3468 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3469 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3470 offset = do_div(logical_sb_block, blocksize);
3471 } else {
3472 logical_sb_block = sb_block;
3473 }
3474
3475 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3476 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3477 goto out_fail;
3478 }
3479 /*
3480 * Note: s_es must be initialized as soon as possible because
3481 * some ext4 macro-instructions depend on its value
3482 */
3483 es = (struct ext4_super_block *) (bh->b_data + offset);
3484 sbi->s_es = es;
3485 sb->s_magic = le16_to_cpu(es->s_magic);
3486 if (sb->s_magic != EXT4_SUPER_MAGIC)
3487 goto cantfind_ext4;
3488 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3489
3490 /* Warn if metadata_csum and gdt_csum are both set. */
3491 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3492 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3493 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3494 ext4_warning(sb, "metadata_csum and uninit_bg are "
3495 "redundant flags; please run fsck.");
3496
3497 /* Check for a known checksum algorithm */
3498 if (!ext4_verify_csum_type(sb, es)) {
3499 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3500 "unknown checksum algorithm.");
3501 silent = 1;
3502 goto cantfind_ext4;
3503 }
3504
3505 /* Load the checksum driver */
3506 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3507 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3508 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3509 if (IS_ERR(sbi->s_chksum_driver)) {
3510 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3511 ret = PTR_ERR(sbi->s_chksum_driver);
3512 sbi->s_chksum_driver = NULL;
3513 goto failed_mount;
3514 }
3515 }
3516
3517 /* Check superblock checksum */
3518 if (!ext4_superblock_csum_verify(sb, es)) {
3519 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3520 "invalid superblock checksum. Run e2fsck?");
3521 silent = 1;
3522 goto cantfind_ext4;
3523 }
3524
3525 /* Precompute checksum seed for all metadata */
3526 if (ext4_has_metadata_csum(sb))
3527 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3528 sizeof(es->s_uuid));
3529
3530 /* Set defaults before we parse the mount options */
3531 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3532 set_opt(sb, INIT_INODE_TABLE);
3533 if (def_mount_opts & EXT4_DEFM_DEBUG)
3534 set_opt(sb, DEBUG);
3535 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3536 set_opt(sb, GRPID);
3537 if (def_mount_opts & EXT4_DEFM_UID16)
3538 set_opt(sb, NO_UID32);
3539 /* xattr user namespace & acls are now defaulted on */
3540 set_opt(sb, XATTR_USER);
3541 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3542 set_opt(sb, POSIX_ACL);
3543 #endif
3544 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3545 if (ext4_has_metadata_csum(sb))
3546 set_opt(sb, JOURNAL_CHECKSUM);
3547
3548 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3549 set_opt(sb, JOURNAL_DATA);
3550 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3551 set_opt(sb, ORDERED_DATA);
3552 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3553 set_opt(sb, WRITEBACK_DATA);
3554
3555 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3556 set_opt(sb, ERRORS_PANIC);
3557 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3558 set_opt(sb, ERRORS_CONT);
3559 else
3560 set_opt(sb, ERRORS_RO);
3561 /* block_validity enabled by default; disable with noblock_validity */
3562 set_opt(sb, BLOCK_VALIDITY);
3563 if (def_mount_opts & EXT4_DEFM_DISCARD)
3564 set_opt(sb, DISCARD);
3565
3566 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3567 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3568 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3569 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3570 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3571
3572 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3573 set_opt(sb, BARRIER);
3574
3575 /*
3576 * enable delayed allocation by default
3577 * Use -o nodelalloc to turn it off
3578 */
3579 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3580 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3581 set_opt(sb, DELALLOC);
3582
3583 /*
3584 * set default s_li_wait_mult for lazyinit, for the case there is
3585 * no mount option specified.
3586 */
3587 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3588
3589 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3590 &journal_devnum, &journal_ioprio, 0)) {
3591 ext4_msg(sb, KERN_WARNING,
3592 "failed to parse options in superblock: %s",
3593 sbi->s_es->s_mount_opts);
3594 }
3595 sbi->s_def_mount_opt = sbi->s_mount_opt;
3596 if (!parse_options((char *) data, sb, &journal_devnum,
3597 &journal_ioprio, 0))
3598 goto failed_mount;
3599
3600 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3601 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3602 "with data=journal disables delayed "
3603 "allocation and O_DIRECT support!\n");
3604 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3605 ext4_msg(sb, KERN_ERR, "can't mount with "
3606 "both data=journal and delalloc");
3607 goto failed_mount;
3608 }
3609 if (test_opt(sb, DIOREAD_NOLOCK)) {
3610 ext4_msg(sb, KERN_ERR, "can't mount with "
3611 "both data=journal and dioread_nolock");
3612 goto failed_mount;
3613 }
3614 if (test_opt(sb, DELALLOC))
3615 clear_opt(sb, DELALLOC);
3616 }
3617
3618 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3619 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3620
3621 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3622 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3623 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3624 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3625 ext4_msg(sb, KERN_WARNING,
3626 "feature flags set on rev 0 fs, "
3627 "running e2fsck is recommended");
3628
3629 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3630 set_opt2(sb, HURD_COMPAT);
3631 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3632 EXT4_FEATURE_INCOMPAT_64BIT)) {
3633 ext4_msg(sb, KERN_ERR,
3634 "The Hurd can't support 64-bit file systems");
3635 goto failed_mount;
3636 }
3637 }
3638
3639 if (IS_EXT2_SB(sb)) {
3640 if (ext2_feature_set_ok(sb))
3641 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3642 "using the ext4 subsystem");
3643 else {
3644 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3645 "to feature incompatibilities");
3646 goto failed_mount;
3647 }
3648 }
3649
3650 if (IS_EXT3_SB(sb)) {
3651 if (ext3_feature_set_ok(sb))
3652 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3653 "using the ext4 subsystem");
3654 else {
3655 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3656 "to feature incompatibilities");
3657 goto failed_mount;
3658 }
3659 }
3660
3661 /*
3662 * Check feature flags regardless of the revision level, since we
3663 * previously didn't change the revision level when setting the flags,
3664 * so there is a chance incompat flags are set on a rev 0 filesystem.
3665 */
3666 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3667 goto failed_mount;
3668
3669 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3670 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3671 blocksize > EXT4_MAX_BLOCK_SIZE) {
3672 ext4_msg(sb, KERN_ERR,
3673 "Unsupported filesystem blocksize %d", blocksize);
3674 goto failed_mount;
3675 }
3676
3677 if (sb->s_blocksize != blocksize) {
3678 /* Validate the filesystem blocksize */
3679 if (!sb_set_blocksize(sb, blocksize)) {
3680 ext4_msg(sb, KERN_ERR, "bad block size %d",
3681 blocksize);
3682 goto failed_mount;
3683 }
3684
3685 brelse(bh);
3686 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3687 offset = do_div(logical_sb_block, blocksize);
3688 bh = sb_bread_unmovable(sb, logical_sb_block);
3689 if (!bh) {
3690 ext4_msg(sb, KERN_ERR,
3691 "Can't read superblock on 2nd try");
3692 goto failed_mount;
3693 }
3694 es = (struct ext4_super_block *)(bh->b_data + offset);
3695 sbi->s_es = es;
3696 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3697 ext4_msg(sb, KERN_ERR,
3698 "Magic mismatch, very weird!");
3699 goto failed_mount;
3700 }
3701 }
3702
3703 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3704 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3705 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3706 has_huge_files);
3707 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3708
3709 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3710 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3711 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3712 } else {
3713 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3714 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3715 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3716 (!is_power_of_2(sbi->s_inode_size)) ||
3717 (sbi->s_inode_size > blocksize)) {
3718 ext4_msg(sb, KERN_ERR,
3719 "unsupported inode size: %d",
3720 sbi->s_inode_size);
3721 goto failed_mount;
3722 }
3723 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3724 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3725 }
3726
3727 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3728 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3729 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3730 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3731 !is_power_of_2(sbi->s_desc_size)) {
3732 ext4_msg(sb, KERN_ERR,
3733 "unsupported descriptor size %lu",
3734 sbi->s_desc_size);
3735 goto failed_mount;
3736 }
3737 } else
3738 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3739
3740 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3741 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3742 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3743 goto cantfind_ext4;
3744
3745 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3746 if (sbi->s_inodes_per_block == 0)
3747 goto cantfind_ext4;
3748 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3749 sbi->s_inodes_per_block;
3750 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3751 sbi->s_sbh = bh;
3752 sbi->s_mount_state = le16_to_cpu(es->s_state);
3753 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3754 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3755
3756 for (i = 0; i < 4; i++)
3757 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3758 sbi->s_def_hash_version = es->s_def_hash_version;
3759 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3760 i = le32_to_cpu(es->s_flags);
3761 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3762 sbi->s_hash_unsigned = 3;
3763 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3764 #ifdef __CHAR_UNSIGNED__
3765 if (!(sb->s_flags & MS_RDONLY))
3766 es->s_flags |=
3767 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3768 sbi->s_hash_unsigned = 3;
3769 #else
3770 if (!(sb->s_flags & MS_RDONLY))
3771 es->s_flags |=
3772 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3773 #endif
3774 }
3775 }
3776
3777 /* Handle clustersize */
3778 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3779 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3780 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3781 if (has_bigalloc) {
3782 if (clustersize < blocksize) {
3783 ext4_msg(sb, KERN_ERR,
3784 "cluster size (%d) smaller than "
3785 "block size (%d)", clustersize, blocksize);
3786 goto failed_mount;
3787 }
3788 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3789 le32_to_cpu(es->s_log_block_size);
3790 sbi->s_clusters_per_group =
3791 le32_to_cpu(es->s_clusters_per_group);
3792 if (sbi->s_clusters_per_group > blocksize * 8) {
3793 ext4_msg(sb, KERN_ERR,
3794 "#clusters per group too big: %lu",
3795 sbi->s_clusters_per_group);
3796 goto failed_mount;
3797 }
3798 if (sbi->s_blocks_per_group !=
3799 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3800 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3801 "clusters per group (%lu) inconsistent",
3802 sbi->s_blocks_per_group,
3803 sbi->s_clusters_per_group);
3804 goto failed_mount;
3805 }
3806 } else {
3807 if (clustersize != blocksize) {
3808 ext4_warning(sb, "fragment/cluster size (%d) != "
3809 "block size (%d)", clustersize,
3810 blocksize);
3811 clustersize = blocksize;
3812 }
3813 if (sbi->s_blocks_per_group > blocksize * 8) {
3814 ext4_msg(sb, KERN_ERR,
3815 "#blocks per group too big: %lu",
3816 sbi->s_blocks_per_group);
3817 goto failed_mount;
3818 }
3819 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3820 sbi->s_cluster_bits = 0;
3821 }
3822 sbi->s_cluster_ratio = clustersize / blocksize;
3823
3824 if (sbi->s_inodes_per_group > blocksize * 8) {
3825 ext4_msg(sb, KERN_ERR,
3826 "#inodes per group too big: %lu",
3827 sbi->s_inodes_per_group);
3828 goto failed_mount;
3829 }
3830
3831 /* Do we have standard group size of clustersize * 8 blocks ? */
3832 if (sbi->s_blocks_per_group == clustersize << 3)
3833 set_opt2(sb, STD_GROUP_SIZE);
3834
3835 /*
3836 * Test whether we have more sectors than will fit in sector_t,
3837 * and whether the max offset is addressable by the page cache.
3838 */
3839 err = generic_check_addressable(sb->s_blocksize_bits,
3840 ext4_blocks_count(es));
3841 if (err) {
3842 ext4_msg(sb, KERN_ERR, "filesystem"
3843 " too large to mount safely on this system");
3844 if (sizeof(sector_t) < 8)
3845 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3846 goto failed_mount;
3847 }
3848
3849 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3850 goto cantfind_ext4;
3851
3852 /* check blocks count against device size */
3853 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3854 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3855 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3856 "exceeds size of device (%llu blocks)",
3857 ext4_blocks_count(es), blocks_count);
3858 goto failed_mount;
3859 }
3860
3861 /*
3862 * It makes no sense for the first data block to be beyond the end
3863 * of the filesystem.
3864 */
3865 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3866 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3867 "block %u is beyond end of filesystem (%llu)",
3868 le32_to_cpu(es->s_first_data_block),
3869 ext4_blocks_count(es));
3870 goto failed_mount;
3871 }
3872 blocks_count = (ext4_blocks_count(es) -
3873 le32_to_cpu(es->s_first_data_block) +
3874 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3875 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3876 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3877 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3878 "(block count %llu, first data block %u, "
3879 "blocks per group %lu)", sbi->s_groups_count,
3880 ext4_blocks_count(es),
3881 le32_to_cpu(es->s_first_data_block),
3882 EXT4_BLOCKS_PER_GROUP(sb));
3883 goto failed_mount;
3884 }
3885 sbi->s_groups_count = blocks_count;
3886 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3887 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3888 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3889 EXT4_DESC_PER_BLOCK(sb);
3890 sbi->s_group_desc = ext4_kvmalloc(db_count *
3891 sizeof(struct buffer_head *),
3892 GFP_KERNEL);
3893 if (sbi->s_group_desc == NULL) {
3894 ext4_msg(sb, KERN_ERR, "not enough memory");
3895 ret = -ENOMEM;
3896 goto failed_mount;
3897 }
3898
3899 if (ext4_proc_root)
3900 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3901
3902 if (sbi->s_proc)
3903 proc_create_data("options", S_IRUGO, sbi->s_proc,
3904 &ext4_seq_options_fops, sb);
3905
3906 bgl_lock_init(sbi->s_blockgroup_lock);
3907
3908 for (i = 0; i < db_count; i++) {
3909 block = descriptor_loc(sb, logical_sb_block, i);
3910 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3911 if (!sbi->s_group_desc[i]) {
3912 ext4_msg(sb, KERN_ERR,
3913 "can't read group descriptor %d", i);
3914 db_count = i;
3915 goto failed_mount2;
3916 }
3917 }
3918 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3919 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3920 goto failed_mount2;
3921 }
3922
3923 sbi->s_gdb_count = db_count;
3924 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3925 spin_lock_init(&sbi->s_next_gen_lock);
3926
3927 init_timer(&sbi->s_err_report);
3928 sbi->s_err_report.function = print_daily_error_info;
3929 sbi->s_err_report.data = (unsigned long) sb;
3930
3931 /* Register extent status tree shrinker */
3932 if (ext4_es_register_shrinker(sbi))
3933 goto failed_mount3;
3934
3935 sbi->s_stripe = ext4_get_stripe_size(sbi);
3936 sbi->s_extent_max_zeroout_kb = 32;
3937
3938 /*
3939 * set up enough so that it can read an inode
3940 */
3941 sb->s_op = &ext4_sops;
3942 sb->s_export_op = &ext4_export_ops;
3943 sb->s_xattr = ext4_xattr_handlers;
3944 #ifdef CONFIG_QUOTA
3945 sb->dq_op = &ext4_quota_operations;
3946 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3947 sb->s_qcop = &ext4_qctl_sysfile_operations;
3948 else
3949 sb->s_qcop = &ext4_qctl_operations;
3950 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
3951 #endif
3952 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3953
3954 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3955 mutex_init(&sbi->s_orphan_lock);
3956
3957 sb->s_root = NULL;
3958
3959 needs_recovery = (es->s_last_orphan != 0 ||
3960 EXT4_HAS_INCOMPAT_FEATURE(sb,
3961 EXT4_FEATURE_INCOMPAT_RECOVER));
3962
3963 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3964 !(sb->s_flags & MS_RDONLY))
3965 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3966 goto failed_mount3a;
3967
3968 /*
3969 * The first inode we look at is the journal inode. Don't try
3970 * root first: it may be modified in the journal!
3971 */
3972 if (!test_opt(sb, NOLOAD) &&
3973 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3974 if (ext4_load_journal(sb, es, journal_devnum))
3975 goto failed_mount3a;
3976 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3977 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3978 ext4_msg(sb, KERN_ERR, "required journal recovery "
3979 "suppressed and not mounted read-only");
3980 goto failed_mount_wq;
3981 } else {
3982 clear_opt(sb, DATA_FLAGS);
3983 sbi->s_journal = NULL;
3984 needs_recovery = 0;
3985 goto no_journal;
3986 }
3987
3988 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3989 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3990 JBD2_FEATURE_INCOMPAT_64BIT)) {
3991 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3992 goto failed_mount_wq;
3993 }
3994
3995 if (!set_journal_csum_feature_set(sb)) {
3996 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3997 "feature set");
3998 goto failed_mount_wq;
3999 }
4000
4001 /* We have now updated the journal if required, so we can
4002 * validate the data journaling mode. */
4003 switch (test_opt(sb, DATA_FLAGS)) {
4004 case 0:
4005 /* No mode set, assume a default based on the journal
4006 * capabilities: ORDERED_DATA if the journal can
4007 * cope, else JOURNAL_DATA
4008 */
4009 if (jbd2_journal_check_available_features
4010 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4011 set_opt(sb, ORDERED_DATA);
4012 else
4013 set_opt(sb, JOURNAL_DATA);
4014 break;
4015
4016 case EXT4_MOUNT_ORDERED_DATA:
4017 case EXT4_MOUNT_WRITEBACK_DATA:
4018 if (!jbd2_journal_check_available_features
4019 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4020 ext4_msg(sb, KERN_ERR, "Journal does not support "
4021 "requested data journaling mode");
4022 goto failed_mount_wq;
4023 }
4024 default:
4025 break;
4026 }
4027 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4028
4029 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4030
4031 no_journal:
4032 if (ext4_mballoc_ready) {
4033 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4034 if (!sbi->s_mb_cache) {
4035 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4036 goto failed_mount_wq;
4037 }
4038 }
4039
4040 /*
4041 * Get the # of file system overhead blocks from the
4042 * superblock if present.
4043 */
4044 if (es->s_overhead_clusters)
4045 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4046 else {
4047 err = ext4_calculate_overhead(sb);
4048 if (err)
4049 goto failed_mount_wq;
4050 }
4051
4052 /*
4053 * The maximum number of concurrent works can be high and
4054 * concurrency isn't really necessary. Limit it to 1.
4055 */
4056 EXT4_SB(sb)->rsv_conversion_wq =
4057 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4058 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4059 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4060 ret = -ENOMEM;
4061 goto failed_mount4;
4062 }
4063
4064 /*
4065 * The jbd2_journal_load will have done any necessary log recovery,
4066 * so we can safely mount the rest of the filesystem now.
4067 */
4068
4069 root = ext4_iget(sb, EXT4_ROOT_INO);
4070 if (IS_ERR(root)) {
4071 ext4_msg(sb, KERN_ERR, "get root inode failed");
4072 ret = PTR_ERR(root);
4073 root = NULL;
4074 goto failed_mount4;
4075 }
4076 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4077 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4078 iput(root);
4079 goto failed_mount4;
4080 }
4081 sb->s_root = d_make_root(root);
4082 if (!sb->s_root) {
4083 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4084 ret = -ENOMEM;
4085 goto failed_mount4;
4086 }
4087
4088 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4089 sb->s_flags |= MS_RDONLY;
4090
4091 /* determine the minimum size of new large inodes, if present */
4092 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4093 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4094 EXT4_GOOD_OLD_INODE_SIZE;
4095 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4096 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4097 if (sbi->s_want_extra_isize <
4098 le16_to_cpu(es->s_want_extra_isize))
4099 sbi->s_want_extra_isize =
4100 le16_to_cpu(es->s_want_extra_isize);
4101 if (sbi->s_want_extra_isize <
4102 le16_to_cpu(es->s_min_extra_isize))
4103 sbi->s_want_extra_isize =
4104 le16_to_cpu(es->s_min_extra_isize);
4105 }
4106 }
4107 /* Check if enough inode space is available */
4108 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4109 sbi->s_inode_size) {
4110 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4111 EXT4_GOOD_OLD_INODE_SIZE;
4112 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4113 "available");
4114 }
4115
4116 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4117 if (err) {
4118 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4119 "reserved pool", ext4_calculate_resv_clusters(sb));
4120 goto failed_mount4a;
4121 }
4122
4123 err = ext4_setup_system_zone(sb);
4124 if (err) {
4125 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4126 "zone (%d)", err);
4127 goto failed_mount4a;
4128 }
4129
4130 ext4_ext_init(sb);
4131 err = ext4_mb_init(sb);
4132 if (err) {
4133 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4134 err);
4135 goto failed_mount5;
4136 }
4137
4138 block = ext4_count_free_clusters(sb);
4139 ext4_free_blocks_count_set(sbi->s_es,
4140 EXT4_C2B(sbi, block));
4141 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4142 GFP_KERNEL);
4143 if (!err) {
4144 unsigned long freei = ext4_count_free_inodes(sb);
4145 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4146 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4147 GFP_KERNEL);
4148 }
4149 if (!err)
4150 err = percpu_counter_init(&sbi->s_dirs_counter,
4151 ext4_count_dirs(sb), GFP_KERNEL);
4152 if (!err)
4153 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4154 GFP_KERNEL);
4155 if (err) {
4156 ext4_msg(sb, KERN_ERR, "insufficient memory");
4157 goto failed_mount6;
4158 }
4159
4160 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4161 if (!ext4_fill_flex_info(sb)) {
4162 ext4_msg(sb, KERN_ERR,
4163 "unable to initialize "
4164 "flex_bg meta info!");
4165 goto failed_mount6;
4166 }
4167
4168 err = ext4_register_li_request(sb, first_not_zeroed);
4169 if (err)
4170 goto failed_mount6;
4171
4172 sbi->s_kobj.kset = ext4_kset;
4173 init_completion(&sbi->s_kobj_unregister);
4174 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4175 "%s", sb->s_id);
4176 if (err)
4177 goto failed_mount7;
4178
4179 #ifdef CONFIG_QUOTA
4180 /* Enable quota usage during mount. */
4181 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4182 !(sb->s_flags & MS_RDONLY)) {
4183 err = ext4_enable_quotas(sb);
4184 if (err)
4185 goto failed_mount8;
4186 }
4187 #endif /* CONFIG_QUOTA */
4188
4189 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4190 ext4_orphan_cleanup(sb, es);
4191 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4192 if (needs_recovery) {
4193 ext4_msg(sb, KERN_INFO, "recovery complete");
4194 ext4_mark_recovery_complete(sb, es);
4195 }
4196 if (EXT4_SB(sb)->s_journal) {
4197 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4198 descr = " journalled data mode";
4199 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4200 descr = " ordered data mode";
4201 else
4202 descr = " writeback data mode";
4203 } else
4204 descr = "out journal";
4205
4206 if (test_opt(sb, DISCARD)) {
4207 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4208 if (!blk_queue_discard(q))
4209 ext4_msg(sb, KERN_WARNING,
4210 "mounting with \"discard\" option, but "
4211 "the device does not support discard");
4212 }
4213
4214 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4215 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4216 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4217
4218 if (es->s_error_count)
4219 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4220
4221 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4222 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4223 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4224 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4225
4226 kfree(orig_data);
4227 return 0;
4228
4229 cantfind_ext4:
4230 if (!silent)
4231 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4232 goto failed_mount;
4233
4234 #ifdef CONFIG_QUOTA
4235 failed_mount8:
4236 kobject_del(&sbi->s_kobj);
4237 #endif
4238 failed_mount7:
4239 ext4_unregister_li_request(sb);
4240 failed_mount6:
4241 ext4_mb_release(sb);
4242 if (sbi->s_flex_groups)
4243 kvfree(sbi->s_flex_groups);
4244 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4245 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4246 percpu_counter_destroy(&sbi->s_dirs_counter);
4247 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4248 failed_mount5:
4249 ext4_ext_release(sb);
4250 ext4_release_system_zone(sb);
4251 failed_mount4a:
4252 dput(sb->s_root);
4253 sb->s_root = NULL;
4254 failed_mount4:
4255 ext4_msg(sb, KERN_ERR, "mount failed");
4256 if (EXT4_SB(sb)->rsv_conversion_wq)
4257 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4258 failed_mount_wq:
4259 if (sbi->s_journal) {
4260 jbd2_journal_destroy(sbi->s_journal);
4261 sbi->s_journal = NULL;
4262 }
4263 failed_mount3a:
4264 ext4_es_unregister_shrinker(sbi);
4265 failed_mount3:
4266 del_timer_sync(&sbi->s_err_report);
4267 if (sbi->s_mmp_tsk)
4268 kthread_stop(sbi->s_mmp_tsk);
4269 failed_mount2:
4270 for (i = 0; i < db_count; i++)
4271 brelse(sbi->s_group_desc[i]);
4272 kvfree(sbi->s_group_desc);
4273 failed_mount:
4274 if (sbi->s_chksum_driver)
4275 crypto_free_shash(sbi->s_chksum_driver);
4276 if (sbi->s_proc) {
4277 remove_proc_entry("options", sbi->s_proc);
4278 remove_proc_entry(sb->s_id, ext4_proc_root);
4279 }
4280 #ifdef CONFIG_QUOTA
4281 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4282 kfree(sbi->s_qf_names[i]);
4283 #endif
4284 ext4_blkdev_remove(sbi);
4285 brelse(bh);
4286 out_fail:
4287 sb->s_fs_info = NULL;
4288 kfree(sbi->s_blockgroup_lock);
4289 kfree(sbi);
4290 out_free_orig:
4291 kfree(orig_data);
4292 return err ? err : ret;
4293 }
4294
4295 /*
4296 * Setup any per-fs journal parameters now. We'll do this both on
4297 * initial mount, once the journal has been initialised but before we've
4298 * done any recovery; and again on any subsequent remount.
4299 */
4300 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4301 {
4302 struct ext4_sb_info *sbi = EXT4_SB(sb);
4303
4304 journal->j_commit_interval = sbi->s_commit_interval;
4305 journal->j_min_batch_time = sbi->s_min_batch_time;
4306 journal->j_max_batch_time = sbi->s_max_batch_time;
4307
4308 write_lock(&journal->j_state_lock);
4309 if (test_opt(sb, BARRIER))
4310 journal->j_flags |= JBD2_BARRIER;
4311 else
4312 journal->j_flags &= ~JBD2_BARRIER;
4313 if (test_opt(sb, DATA_ERR_ABORT))
4314 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4315 else
4316 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4317 write_unlock(&journal->j_state_lock);
4318 }
4319
4320 static journal_t *ext4_get_journal(struct super_block *sb,
4321 unsigned int journal_inum)
4322 {
4323 struct inode *journal_inode;
4324 journal_t *journal;
4325
4326 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4327
4328 /* First, test for the existence of a valid inode on disk. Bad
4329 * things happen if we iget() an unused inode, as the subsequent
4330 * iput() will try to delete it. */
4331
4332 journal_inode = ext4_iget(sb, journal_inum);
4333 if (IS_ERR(journal_inode)) {
4334 ext4_msg(sb, KERN_ERR, "no journal found");
4335 return NULL;
4336 }
4337 if (!journal_inode->i_nlink) {
4338 make_bad_inode(journal_inode);
4339 iput(journal_inode);
4340 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4341 return NULL;
4342 }
4343
4344 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4345 journal_inode, journal_inode->i_size);
4346 if (!S_ISREG(journal_inode->i_mode)) {
4347 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4348 iput(journal_inode);
4349 return NULL;
4350 }
4351
4352 journal = jbd2_journal_init_inode(journal_inode);
4353 if (!journal) {
4354 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4355 iput(journal_inode);
4356 return NULL;
4357 }
4358 journal->j_private = sb;
4359 ext4_init_journal_params(sb, journal);
4360 return journal;
4361 }
4362
4363 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4364 dev_t j_dev)
4365 {
4366 struct buffer_head *bh;
4367 journal_t *journal;
4368 ext4_fsblk_t start;
4369 ext4_fsblk_t len;
4370 int hblock, blocksize;
4371 ext4_fsblk_t sb_block;
4372 unsigned long offset;
4373 struct ext4_super_block *es;
4374 struct block_device *bdev;
4375
4376 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4377
4378 bdev = ext4_blkdev_get(j_dev, sb);
4379 if (bdev == NULL)
4380 return NULL;
4381
4382 blocksize = sb->s_blocksize;
4383 hblock = bdev_logical_block_size(bdev);
4384 if (blocksize < hblock) {
4385 ext4_msg(sb, KERN_ERR,
4386 "blocksize too small for journal device");
4387 goto out_bdev;
4388 }
4389
4390 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4391 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4392 set_blocksize(bdev, blocksize);
4393 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4394 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4395 "external journal");
4396 goto out_bdev;
4397 }
4398
4399 es = (struct ext4_super_block *) (bh->b_data + offset);
4400 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4401 !(le32_to_cpu(es->s_feature_incompat) &
4402 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4403 ext4_msg(sb, KERN_ERR, "external journal has "
4404 "bad superblock");
4405 brelse(bh);
4406 goto out_bdev;
4407 }
4408
4409 if ((le32_to_cpu(es->s_feature_ro_compat) &
4410 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4411 es->s_checksum != ext4_superblock_csum(sb, es)) {
4412 ext4_msg(sb, KERN_ERR, "external journal has "
4413 "corrupt superblock");
4414 brelse(bh);
4415 goto out_bdev;
4416 }
4417
4418 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4419 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4420 brelse(bh);
4421 goto out_bdev;
4422 }
4423
4424 len = ext4_blocks_count(es);
4425 start = sb_block + 1;
4426 brelse(bh); /* we're done with the superblock */
4427
4428 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4429 start, len, blocksize);
4430 if (!journal) {
4431 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4432 goto out_bdev;
4433 }
4434 journal->j_private = sb;
4435 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4436 wait_on_buffer(journal->j_sb_buffer);
4437 if (!buffer_uptodate(journal->j_sb_buffer)) {
4438 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4439 goto out_journal;
4440 }
4441 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4442 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4443 "user (unsupported) - %d",
4444 be32_to_cpu(journal->j_superblock->s_nr_users));
4445 goto out_journal;
4446 }
4447 EXT4_SB(sb)->journal_bdev = bdev;
4448 ext4_init_journal_params(sb, journal);
4449 return journal;
4450
4451 out_journal:
4452 jbd2_journal_destroy(journal);
4453 out_bdev:
4454 ext4_blkdev_put(bdev);
4455 return NULL;
4456 }
4457
4458 static int ext4_load_journal(struct super_block *sb,
4459 struct ext4_super_block *es,
4460 unsigned long journal_devnum)
4461 {
4462 journal_t *journal;
4463 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4464 dev_t journal_dev;
4465 int err = 0;
4466 int really_read_only;
4467
4468 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4469
4470 if (journal_devnum &&
4471 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4472 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4473 "numbers have changed");
4474 journal_dev = new_decode_dev(journal_devnum);
4475 } else
4476 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4477
4478 really_read_only = bdev_read_only(sb->s_bdev);
4479
4480 /*
4481 * Are we loading a blank journal or performing recovery after a
4482 * crash? For recovery, we need to check in advance whether we
4483 * can get read-write access to the device.
4484 */
4485 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4486 if (sb->s_flags & MS_RDONLY) {
4487 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4488 "required on readonly filesystem");
4489 if (really_read_only) {
4490 ext4_msg(sb, KERN_ERR, "write access "
4491 "unavailable, cannot proceed");
4492 return -EROFS;
4493 }
4494 ext4_msg(sb, KERN_INFO, "write access will "
4495 "be enabled during recovery");
4496 }
4497 }
4498
4499 if (journal_inum && journal_dev) {
4500 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4501 "and inode journals!");
4502 return -EINVAL;
4503 }
4504
4505 if (journal_inum) {
4506 if (!(journal = ext4_get_journal(sb, journal_inum)))
4507 return -EINVAL;
4508 } else {
4509 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4510 return -EINVAL;
4511 }
4512
4513 if (!(journal->j_flags & JBD2_BARRIER))
4514 ext4_msg(sb, KERN_INFO, "barriers disabled");
4515
4516 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4517 err = jbd2_journal_wipe(journal, !really_read_only);
4518 if (!err) {
4519 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4520 if (save)
4521 memcpy(save, ((char *) es) +
4522 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4523 err = jbd2_journal_load(journal);
4524 if (save)
4525 memcpy(((char *) es) + EXT4_S_ERR_START,
4526 save, EXT4_S_ERR_LEN);
4527 kfree(save);
4528 }
4529
4530 if (err) {
4531 ext4_msg(sb, KERN_ERR, "error loading journal");
4532 jbd2_journal_destroy(journal);
4533 return err;
4534 }
4535
4536 EXT4_SB(sb)->s_journal = journal;
4537 ext4_clear_journal_err(sb, es);
4538
4539 if (!really_read_only && journal_devnum &&
4540 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4541 es->s_journal_dev = cpu_to_le32(journal_devnum);
4542
4543 /* Make sure we flush the recovery flag to disk. */
4544 ext4_commit_super(sb, 1);
4545 }
4546
4547 return 0;
4548 }
4549
4550 static int ext4_commit_super(struct super_block *sb, int sync)
4551 {
4552 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4553 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4554 int error = 0;
4555
4556 if (!sbh || block_device_ejected(sb))
4557 return error;
4558 if (buffer_write_io_error(sbh)) {
4559 /*
4560 * Oh, dear. A previous attempt to write the
4561 * superblock failed. This could happen because the
4562 * USB device was yanked out. Or it could happen to
4563 * be a transient write error and maybe the block will
4564 * be remapped. Nothing we can do but to retry the
4565 * write and hope for the best.
4566 */
4567 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4568 "superblock detected");
4569 clear_buffer_write_io_error(sbh);
4570 set_buffer_uptodate(sbh);
4571 }
4572 /*
4573 * If the file system is mounted read-only, don't update the
4574 * superblock write time. This avoids updating the superblock
4575 * write time when we are mounting the root file system
4576 * read/only but we need to replay the journal; at that point,
4577 * for people who are east of GMT and who make their clock
4578 * tick in localtime for Windows bug-for-bug compatibility,
4579 * the clock is set in the future, and this will cause e2fsck
4580 * to complain and force a full file system check.
4581 */
4582 if (!(sb->s_flags & MS_RDONLY))
4583 es->s_wtime = cpu_to_le32(get_seconds());
4584 if (sb->s_bdev->bd_part)
4585 es->s_kbytes_written =
4586 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4587 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4588 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4589 else
4590 es->s_kbytes_written =
4591 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4592 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4593 ext4_free_blocks_count_set(es,
4594 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4595 &EXT4_SB(sb)->s_freeclusters_counter)));
4596 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4597 es->s_free_inodes_count =
4598 cpu_to_le32(percpu_counter_sum_positive(
4599 &EXT4_SB(sb)->s_freeinodes_counter));
4600 BUFFER_TRACE(sbh, "marking dirty");
4601 ext4_superblock_csum_set(sb);
4602 mark_buffer_dirty(sbh);
4603 if (sync) {
4604 error = sync_dirty_buffer(sbh);
4605 if (error)
4606 return error;
4607
4608 error = buffer_write_io_error(sbh);
4609 if (error) {
4610 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4611 "superblock");
4612 clear_buffer_write_io_error(sbh);
4613 set_buffer_uptodate(sbh);
4614 }
4615 }
4616 return error;
4617 }
4618
4619 /*
4620 * Have we just finished recovery? If so, and if we are mounting (or
4621 * remounting) the filesystem readonly, then we will end up with a
4622 * consistent fs on disk. Record that fact.
4623 */
4624 static void ext4_mark_recovery_complete(struct super_block *sb,
4625 struct ext4_super_block *es)
4626 {
4627 journal_t *journal = EXT4_SB(sb)->s_journal;
4628
4629 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4630 BUG_ON(journal != NULL);
4631 return;
4632 }
4633 jbd2_journal_lock_updates(journal);
4634 if (jbd2_journal_flush(journal) < 0)
4635 goto out;
4636
4637 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4638 sb->s_flags & MS_RDONLY) {
4639 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4640 ext4_commit_super(sb, 1);
4641 }
4642
4643 out:
4644 jbd2_journal_unlock_updates(journal);
4645 }
4646
4647 /*
4648 * If we are mounting (or read-write remounting) a filesystem whose journal
4649 * has recorded an error from a previous lifetime, move that error to the
4650 * main filesystem now.
4651 */
4652 static void ext4_clear_journal_err(struct super_block *sb,
4653 struct ext4_super_block *es)
4654 {
4655 journal_t *journal;
4656 int j_errno;
4657 const char *errstr;
4658
4659 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4660
4661 journal = EXT4_SB(sb)->s_journal;
4662
4663 /*
4664 * Now check for any error status which may have been recorded in the
4665 * journal by a prior ext4_error() or ext4_abort()
4666 */
4667
4668 j_errno = jbd2_journal_errno(journal);
4669 if (j_errno) {
4670 char nbuf[16];
4671
4672 errstr = ext4_decode_error(sb, j_errno, nbuf);
4673 ext4_warning(sb, "Filesystem error recorded "
4674 "from previous mount: %s", errstr);
4675 ext4_warning(sb, "Marking fs in need of filesystem check.");
4676
4677 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4678 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4679 ext4_commit_super(sb, 1);
4680
4681 jbd2_journal_clear_err(journal);
4682 jbd2_journal_update_sb_errno(journal);
4683 }
4684 }
4685
4686 /*
4687 * Force the running and committing transactions to commit,
4688 * and wait on the commit.
4689 */
4690 int ext4_force_commit(struct super_block *sb)
4691 {
4692 journal_t *journal;
4693
4694 if (sb->s_flags & MS_RDONLY)
4695 return 0;
4696
4697 journal = EXT4_SB(sb)->s_journal;
4698 return ext4_journal_force_commit(journal);
4699 }
4700
4701 static int ext4_sync_fs(struct super_block *sb, int wait)
4702 {
4703 int ret = 0;
4704 tid_t target;
4705 bool needs_barrier = false;
4706 struct ext4_sb_info *sbi = EXT4_SB(sb);
4707
4708 trace_ext4_sync_fs(sb, wait);
4709 flush_workqueue(sbi->rsv_conversion_wq);
4710 /*
4711 * Writeback quota in non-journalled quota case - journalled quota has
4712 * no dirty dquots
4713 */
4714 dquot_writeback_dquots(sb, -1);
4715 /*
4716 * Data writeback is possible w/o journal transaction, so barrier must
4717 * being sent at the end of the function. But we can skip it if
4718 * transaction_commit will do it for us.
4719 */
4720 if (sbi->s_journal) {
4721 target = jbd2_get_latest_transaction(sbi->s_journal);
4722 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4723 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4724 needs_barrier = true;
4725
4726 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4727 if (wait)
4728 ret = jbd2_log_wait_commit(sbi->s_journal,
4729 target);
4730 }
4731 } else if (wait && test_opt(sb, BARRIER))
4732 needs_barrier = true;
4733 if (needs_barrier) {
4734 int err;
4735 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4736 if (!ret)
4737 ret = err;
4738 }
4739
4740 return ret;
4741 }
4742
4743 /*
4744 * LVM calls this function before a (read-only) snapshot is created. This
4745 * gives us a chance to flush the journal completely and mark the fs clean.
4746 *
4747 * Note that only this function cannot bring a filesystem to be in a clean
4748 * state independently. It relies on upper layer to stop all data & metadata
4749 * modifications.
4750 */
4751 static int ext4_freeze(struct super_block *sb)
4752 {
4753 int error = 0;
4754 journal_t *journal;
4755
4756 if (sb->s_flags & MS_RDONLY)
4757 return 0;
4758
4759 journal = EXT4_SB(sb)->s_journal;
4760
4761 if (journal) {
4762 /* Now we set up the journal barrier. */
4763 jbd2_journal_lock_updates(journal);
4764
4765 /*
4766 * Don't clear the needs_recovery flag if we failed to
4767 * flush the journal.
4768 */
4769 error = jbd2_journal_flush(journal);
4770 if (error < 0)
4771 goto out;
4772 }
4773
4774 /* Journal blocked and flushed, clear needs_recovery flag. */
4775 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4776 error = ext4_commit_super(sb, 1);
4777 out:
4778 if (journal)
4779 /* we rely on upper layer to stop further updates */
4780 jbd2_journal_unlock_updates(journal);
4781 return error;
4782 }
4783
4784 /*
4785 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4786 * flag here, even though the filesystem is not technically dirty yet.
4787 */
4788 static int ext4_unfreeze(struct super_block *sb)
4789 {
4790 if (sb->s_flags & MS_RDONLY)
4791 return 0;
4792
4793 /* Reset the needs_recovery flag before the fs is unlocked. */
4794 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4795 ext4_commit_super(sb, 1);
4796 return 0;
4797 }
4798
4799 /*
4800 * Structure to save mount options for ext4_remount's benefit
4801 */
4802 struct ext4_mount_options {
4803 unsigned long s_mount_opt;
4804 unsigned long s_mount_opt2;
4805 kuid_t s_resuid;
4806 kgid_t s_resgid;
4807 unsigned long s_commit_interval;
4808 u32 s_min_batch_time, s_max_batch_time;
4809 #ifdef CONFIG_QUOTA
4810 int s_jquota_fmt;
4811 char *s_qf_names[EXT4_MAXQUOTAS];
4812 #endif
4813 };
4814
4815 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4816 {
4817 struct ext4_super_block *es;
4818 struct ext4_sb_info *sbi = EXT4_SB(sb);
4819 unsigned long old_sb_flags;
4820 struct ext4_mount_options old_opts;
4821 int enable_quota = 0;
4822 ext4_group_t g;
4823 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4824 int err = 0;
4825 #ifdef CONFIG_QUOTA
4826 int i, j;
4827 #endif
4828 char *orig_data = kstrdup(data, GFP_KERNEL);
4829
4830 /* Store the original options */
4831 old_sb_flags = sb->s_flags;
4832 old_opts.s_mount_opt = sbi->s_mount_opt;
4833 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4834 old_opts.s_resuid = sbi->s_resuid;
4835 old_opts.s_resgid = sbi->s_resgid;
4836 old_opts.s_commit_interval = sbi->s_commit_interval;
4837 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4838 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4839 #ifdef CONFIG_QUOTA
4840 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4841 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4842 if (sbi->s_qf_names[i]) {
4843 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4844 GFP_KERNEL);
4845 if (!old_opts.s_qf_names[i]) {
4846 for (j = 0; j < i; j++)
4847 kfree(old_opts.s_qf_names[j]);
4848 kfree(orig_data);
4849 return -ENOMEM;
4850 }
4851 } else
4852 old_opts.s_qf_names[i] = NULL;
4853 #endif
4854 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4855 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4856
4857 /*
4858 * Allow the "check" option to be passed as a remount option.
4859 */
4860 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4861 err = -EINVAL;
4862 goto restore_opts;
4863 }
4864
4865 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4866 test_opt(sb, JOURNAL_CHECKSUM)) {
4867 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4868 "during remount not supported");
4869 err = -EINVAL;
4870 goto restore_opts;
4871 }
4872
4873 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4874 test_opt(sb, JOURNAL_CHECKSUM)) {
4875 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4876 "during remount not supported");
4877 err = -EINVAL;
4878 goto restore_opts;
4879 }
4880
4881 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4882 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4883 ext4_msg(sb, KERN_ERR, "can't mount with "
4884 "both data=journal and delalloc");
4885 err = -EINVAL;
4886 goto restore_opts;
4887 }
4888 if (test_opt(sb, DIOREAD_NOLOCK)) {
4889 ext4_msg(sb, KERN_ERR, "can't mount with "
4890 "both data=journal and dioread_nolock");
4891 err = -EINVAL;
4892 goto restore_opts;
4893 }
4894 }
4895
4896 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4897 ext4_abort(sb, "Abort forced by user");
4898
4899 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4900 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4901
4902 es = sbi->s_es;
4903
4904 if (sbi->s_journal) {
4905 ext4_init_journal_params(sb, sbi->s_journal);
4906 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4907 }
4908
4909 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4910 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4911 err = -EROFS;
4912 goto restore_opts;
4913 }
4914
4915 if (*flags & MS_RDONLY) {
4916 err = sync_filesystem(sb);
4917 if (err < 0)
4918 goto restore_opts;
4919 err = dquot_suspend(sb, -1);
4920 if (err < 0)
4921 goto restore_opts;
4922
4923 /*
4924 * First of all, the unconditional stuff we have to do
4925 * to disable replay of the journal when we next remount
4926 */
4927 sb->s_flags |= MS_RDONLY;
4928
4929 /*
4930 * OK, test if we are remounting a valid rw partition
4931 * readonly, and if so set the rdonly flag and then
4932 * mark the partition as valid again.
4933 */
4934 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4935 (sbi->s_mount_state & EXT4_VALID_FS))
4936 es->s_state = cpu_to_le16(sbi->s_mount_state);
4937
4938 if (sbi->s_journal)
4939 ext4_mark_recovery_complete(sb, es);
4940 } else {
4941 /* Make sure we can mount this feature set readwrite */
4942 if (!ext4_feature_set_ok(sb, 0)) {
4943 err = -EROFS;
4944 goto restore_opts;
4945 }
4946 /*
4947 * Make sure the group descriptor checksums
4948 * are sane. If they aren't, refuse to remount r/w.
4949 */
4950 for (g = 0; g < sbi->s_groups_count; g++) {
4951 struct ext4_group_desc *gdp =
4952 ext4_get_group_desc(sb, g, NULL);
4953
4954 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4955 ext4_msg(sb, KERN_ERR,
4956 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4957 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4958 le16_to_cpu(gdp->bg_checksum));
4959 err = -EINVAL;
4960 goto restore_opts;
4961 }
4962 }
4963
4964 /*
4965 * If we have an unprocessed orphan list hanging
4966 * around from a previously readonly bdev mount,
4967 * require a full umount/remount for now.
4968 */
4969 if (es->s_last_orphan) {
4970 ext4_msg(sb, KERN_WARNING, "Couldn't "
4971 "remount RDWR because of unprocessed "
4972 "orphan inode list. Please "
4973 "umount/remount instead");
4974 err = -EINVAL;
4975 goto restore_opts;
4976 }
4977
4978 /*
4979 * Mounting a RDONLY partition read-write, so reread
4980 * and store the current valid flag. (It may have
4981 * been changed by e2fsck since we originally mounted
4982 * the partition.)
4983 */
4984 if (sbi->s_journal)
4985 ext4_clear_journal_err(sb, es);
4986 sbi->s_mount_state = le16_to_cpu(es->s_state);
4987 if (!ext4_setup_super(sb, es, 0))
4988 sb->s_flags &= ~MS_RDONLY;
4989 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4990 EXT4_FEATURE_INCOMPAT_MMP))
4991 if (ext4_multi_mount_protect(sb,
4992 le64_to_cpu(es->s_mmp_block))) {
4993 err = -EROFS;
4994 goto restore_opts;
4995 }
4996 enable_quota = 1;
4997 }
4998 }
4999
5000 /*
5001 * Reinitialize lazy itable initialization thread based on
5002 * current settings
5003 */
5004 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5005 ext4_unregister_li_request(sb);
5006 else {
5007 ext4_group_t first_not_zeroed;
5008 first_not_zeroed = ext4_has_uninit_itable(sb);
5009 ext4_register_li_request(sb, first_not_zeroed);
5010 }
5011
5012 ext4_setup_system_zone(sb);
5013 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5014 ext4_commit_super(sb, 1);
5015
5016 #ifdef CONFIG_QUOTA
5017 /* Release old quota file names */
5018 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5019 kfree(old_opts.s_qf_names[i]);
5020 if (enable_quota) {
5021 if (sb_any_quota_suspended(sb))
5022 dquot_resume(sb, -1);
5023 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5024 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5025 err = ext4_enable_quotas(sb);
5026 if (err)
5027 goto restore_opts;
5028 }
5029 }
5030 #endif
5031
5032 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5033 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5034 kfree(orig_data);
5035 return 0;
5036
5037 restore_opts:
5038 sb->s_flags = old_sb_flags;
5039 sbi->s_mount_opt = old_opts.s_mount_opt;
5040 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5041 sbi->s_resuid = old_opts.s_resuid;
5042 sbi->s_resgid = old_opts.s_resgid;
5043 sbi->s_commit_interval = old_opts.s_commit_interval;
5044 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5045 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5046 #ifdef CONFIG_QUOTA
5047 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5048 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5049 kfree(sbi->s_qf_names[i]);
5050 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5051 }
5052 #endif
5053 kfree(orig_data);
5054 return err;
5055 }
5056
5057 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5058 {
5059 struct super_block *sb = dentry->d_sb;
5060 struct ext4_sb_info *sbi = EXT4_SB(sb);
5061 struct ext4_super_block *es = sbi->s_es;
5062 ext4_fsblk_t overhead = 0, resv_blocks;
5063 u64 fsid;
5064 s64 bfree;
5065 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5066
5067 if (!test_opt(sb, MINIX_DF))
5068 overhead = sbi->s_overhead;
5069
5070 buf->f_type = EXT4_SUPER_MAGIC;
5071 buf->f_bsize = sb->s_blocksize;
5072 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5073 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5074 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5075 /* prevent underflow in case that few free space is available */
5076 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5077 buf->f_bavail = buf->f_bfree -
5078 (ext4_r_blocks_count(es) + resv_blocks);
5079 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5080 buf->f_bavail = 0;
5081 buf->f_files = le32_to_cpu(es->s_inodes_count);
5082 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5083 buf->f_namelen = EXT4_NAME_LEN;
5084 fsid = le64_to_cpup((void *)es->s_uuid) ^
5085 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5086 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5087 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5088
5089 return 0;
5090 }
5091
5092 /* Helper function for writing quotas on sync - we need to start transaction
5093 * before quota file is locked for write. Otherwise the are possible deadlocks:
5094 * Process 1 Process 2
5095 * ext4_create() quota_sync()
5096 * jbd2_journal_start() write_dquot()
5097 * dquot_initialize() down(dqio_mutex)
5098 * down(dqio_mutex) jbd2_journal_start()
5099 *
5100 */
5101
5102 #ifdef CONFIG_QUOTA
5103
5104 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5105 {
5106 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5107 }
5108
5109 static int ext4_write_dquot(struct dquot *dquot)
5110 {
5111 int ret, err;
5112 handle_t *handle;
5113 struct inode *inode;
5114
5115 inode = dquot_to_inode(dquot);
5116 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5117 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5118 if (IS_ERR(handle))
5119 return PTR_ERR(handle);
5120 ret = dquot_commit(dquot);
5121 err = ext4_journal_stop(handle);
5122 if (!ret)
5123 ret = err;
5124 return ret;
5125 }
5126
5127 static int ext4_acquire_dquot(struct dquot *dquot)
5128 {
5129 int ret, err;
5130 handle_t *handle;
5131
5132 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5133 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5134 if (IS_ERR(handle))
5135 return PTR_ERR(handle);
5136 ret = dquot_acquire(dquot);
5137 err = ext4_journal_stop(handle);
5138 if (!ret)
5139 ret = err;
5140 return ret;
5141 }
5142
5143 static int ext4_release_dquot(struct dquot *dquot)
5144 {
5145 int ret, err;
5146 handle_t *handle;
5147
5148 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5149 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5150 if (IS_ERR(handle)) {
5151 /* Release dquot anyway to avoid endless cycle in dqput() */
5152 dquot_release(dquot);
5153 return PTR_ERR(handle);
5154 }
5155 ret = dquot_release(dquot);
5156 err = ext4_journal_stop(handle);
5157 if (!ret)
5158 ret = err;
5159 return ret;
5160 }
5161
5162 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5163 {
5164 struct super_block *sb = dquot->dq_sb;
5165 struct ext4_sb_info *sbi = EXT4_SB(sb);
5166
5167 /* Are we journaling quotas? */
5168 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5169 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5170 dquot_mark_dquot_dirty(dquot);
5171 return ext4_write_dquot(dquot);
5172 } else {
5173 return dquot_mark_dquot_dirty(dquot);
5174 }
5175 }
5176
5177 static int ext4_write_info(struct super_block *sb, int type)
5178 {
5179 int ret, err;
5180 handle_t *handle;
5181
5182 /* Data block + inode block */
5183 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5184 if (IS_ERR(handle))
5185 return PTR_ERR(handle);
5186 ret = dquot_commit_info(sb, type);
5187 err = ext4_journal_stop(handle);
5188 if (!ret)
5189 ret = err;
5190 return ret;
5191 }
5192
5193 /*
5194 * Turn on quotas during mount time - we need to find
5195 * the quota file and such...
5196 */
5197 static int ext4_quota_on_mount(struct super_block *sb, int type)
5198 {
5199 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5200 EXT4_SB(sb)->s_jquota_fmt, type);
5201 }
5202
5203 /*
5204 * Standard function to be called on quota_on
5205 */
5206 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5207 struct path *path)
5208 {
5209 int err;
5210
5211 if (!test_opt(sb, QUOTA))
5212 return -EINVAL;
5213
5214 /* Quotafile not on the same filesystem? */
5215 if (path->dentry->d_sb != sb)
5216 return -EXDEV;
5217 /* Journaling quota? */
5218 if (EXT4_SB(sb)->s_qf_names[type]) {
5219 /* Quotafile not in fs root? */
5220 if (path->dentry->d_parent != sb->s_root)
5221 ext4_msg(sb, KERN_WARNING,
5222 "Quota file not on filesystem root. "
5223 "Journaled quota will not work");
5224 }
5225
5226 /*
5227 * When we journal data on quota file, we have to flush journal to see
5228 * all updates to the file when we bypass pagecache...
5229 */
5230 if (EXT4_SB(sb)->s_journal &&
5231 ext4_should_journal_data(path->dentry->d_inode)) {
5232 /*
5233 * We don't need to lock updates but journal_flush() could
5234 * otherwise be livelocked...
5235 */
5236 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5237 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5238 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5239 if (err)
5240 return err;
5241 }
5242
5243 return dquot_quota_on(sb, type, format_id, path);
5244 }
5245
5246 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5247 unsigned int flags)
5248 {
5249 int err;
5250 struct inode *qf_inode;
5251 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5252 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5253 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5254 };
5255
5256 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5257
5258 if (!qf_inums[type])
5259 return -EPERM;
5260
5261 qf_inode = ext4_iget(sb, qf_inums[type]);
5262 if (IS_ERR(qf_inode)) {
5263 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5264 return PTR_ERR(qf_inode);
5265 }
5266
5267 /* Don't account quota for quota files to avoid recursion */
5268 qf_inode->i_flags |= S_NOQUOTA;
5269 err = dquot_enable(qf_inode, type, format_id, flags);
5270 iput(qf_inode);
5271
5272 return err;
5273 }
5274
5275 /* Enable usage tracking for all quota types. */
5276 static int ext4_enable_quotas(struct super_block *sb)
5277 {
5278 int type, err = 0;
5279 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5280 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5281 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5282 };
5283
5284 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5285 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5286 if (qf_inums[type]) {
5287 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5288 DQUOT_USAGE_ENABLED);
5289 if (err) {
5290 ext4_warning(sb,
5291 "Failed to enable quota tracking "
5292 "(type=%d, err=%d). Please run "
5293 "e2fsck to fix.", type, err);
5294 return err;
5295 }
5296 }
5297 }
5298 return 0;
5299 }
5300
5301 /*
5302 * quota_on function that is used when QUOTA feature is set.
5303 */
5304 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5305 int format_id)
5306 {
5307 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5308 return -EINVAL;
5309
5310 /*
5311 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5312 */
5313 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5314 }
5315
5316 static int ext4_quota_off(struct super_block *sb, int type)
5317 {
5318 struct inode *inode = sb_dqopt(sb)->files[type];
5319 handle_t *handle;
5320
5321 /* Force all delayed allocation blocks to be allocated.
5322 * Caller already holds s_umount sem */
5323 if (test_opt(sb, DELALLOC))
5324 sync_filesystem(sb);
5325
5326 if (!inode)
5327 goto out;
5328
5329 /* Update modification times of quota files when userspace can
5330 * start looking at them */
5331 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5332 if (IS_ERR(handle))
5333 goto out;
5334 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5335 ext4_mark_inode_dirty(handle, inode);
5336 ext4_journal_stop(handle);
5337
5338 out:
5339 return dquot_quota_off(sb, type);
5340 }
5341
5342 /*
5343 * quota_off function that is used when QUOTA feature is set.
5344 */
5345 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5346 {
5347 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5348 return -EINVAL;
5349
5350 /* Disable only the limits. */
5351 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5352 }
5353
5354 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5355 * acquiring the locks... As quota files are never truncated and quota code
5356 * itself serializes the operations (and no one else should touch the files)
5357 * we don't have to be afraid of races */
5358 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5359 size_t len, loff_t off)
5360 {
5361 struct inode *inode = sb_dqopt(sb)->files[type];
5362 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5363 int offset = off & (sb->s_blocksize - 1);
5364 int tocopy;
5365 size_t toread;
5366 struct buffer_head *bh;
5367 loff_t i_size = i_size_read(inode);
5368
5369 if (off > i_size)
5370 return 0;
5371 if (off+len > i_size)
5372 len = i_size-off;
5373 toread = len;
5374 while (toread > 0) {
5375 tocopy = sb->s_blocksize - offset < toread ?
5376 sb->s_blocksize - offset : toread;
5377 bh = ext4_bread(NULL, inode, blk, 0);
5378 if (IS_ERR(bh))
5379 return PTR_ERR(bh);
5380 if (!bh) /* A hole? */
5381 memset(data, 0, tocopy);
5382 else
5383 memcpy(data, bh->b_data+offset, tocopy);
5384 brelse(bh);
5385 offset = 0;
5386 toread -= tocopy;
5387 data += tocopy;
5388 blk++;
5389 }
5390 return len;
5391 }
5392
5393 /* Write to quotafile (we know the transaction is already started and has
5394 * enough credits) */
5395 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5396 const char *data, size_t len, loff_t off)
5397 {
5398 struct inode *inode = sb_dqopt(sb)->files[type];
5399 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5400 int err, offset = off & (sb->s_blocksize - 1);
5401 struct buffer_head *bh;
5402 handle_t *handle = journal_current_handle();
5403
5404 if (EXT4_SB(sb)->s_journal && !handle) {
5405 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5406 " cancelled because transaction is not started",
5407 (unsigned long long)off, (unsigned long long)len);
5408 return -EIO;
5409 }
5410 /*
5411 * Since we account only one data block in transaction credits,
5412 * then it is impossible to cross a block boundary.
5413 */
5414 if (sb->s_blocksize - offset < len) {
5415 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5416 " cancelled because not block aligned",
5417 (unsigned long long)off, (unsigned long long)len);
5418 return -EIO;
5419 }
5420
5421 bh = ext4_bread(handle, inode, blk, 1);
5422 if (IS_ERR(bh))
5423 return PTR_ERR(bh);
5424 if (!bh)
5425 goto out;
5426 BUFFER_TRACE(bh, "get write access");
5427 err = ext4_journal_get_write_access(handle, bh);
5428 if (err) {
5429 brelse(bh);
5430 return err;
5431 }
5432 lock_buffer(bh);
5433 memcpy(bh->b_data+offset, data, len);
5434 flush_dcache_page(bh->b_page);
5435 unlock_buffer(bh);
5436 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5437 brelse(bh);
5438 out:
5439 if (inode->i_size < off + len) {
5440 i_size_write(inode, off + len);
5441 EXT4_I(inode)->i_disksize = inode->i_size;
5442 ext4_mark_inode_dirty(handle, inode);
5443 }
5444 return len;
5445 }
5446
5447 #endif
5448
5449 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5450 const char *dev_name, void *data)
5451 {
5452 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5453 }
5454
5455 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5456 static inline void register_as_ext2(void)
5457 {
5458 int err = register_filesystem(&ext2_fs_type);
5459 if (err)
5460 printk(KERN_WARNING
5461 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5462 }
5463
5464 static inline void unregister_as_ext2(void)
5465 {
5466 unregister_filesystem(&ext2_fs_type);
5467 }
5468
5469 static inline int ext2_feature_set_ok(struct super_block *sb)
5470 {
5471 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5472 return 0;
5473 if (sb->s_flags & MS_RDONLY)
5474 return 1;
5475 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5476 return 0;
5477 return 1;
5478 }
5479 #else
5480 static inline void register_as_ext2(void) { }
5481 static inline void unregister_as_ext2(void) { }
5482 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5483 #endif
5484
5485 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5486 static inline void register_as_ext3(void)
5487 {
5488 int err = register_filesystem(&ext3_fs_type);
5489 if (err)
5490 printk(KERN_WARNING
5491 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5492 }
5493
5494 static inline void unregister_as_ext3(void)
5495 {
5496 unregister_filesystem(&ext3_fs_type);
5497 }
5498
5499 static inline int ext3_feature_set_ok(struct super_block *sb)
5500 {
5501 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5502 return 0;
5503 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5504 return 0;
5505 if (sb->s_flags & MS_RDONLY)
5506 return 1;
5507 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5508 return 0;
5509 return 1;
5510 }
5511 #else
5512 static inline void register_as_ext3(void) { }
5513 static inline void unregister_as_ext3(void) { }
5514 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5515 #endif
5516
5517 static struct file_system_type ext4_fs_type = {
5518 .owner = THIS_MODULE,
5519 .name = "ext4",
5520 .mount = ext4_mount,
5521 .kill_sb = kill_block_super,
5522 .fs_flags = FS_REQUIRES_DEV,
5523 };
5524 MODULE_ALIAS_FS("ext4");
5525
5526 static int __init ext4_init_feat_adverts(void)
5527 {
5528 struct ext4_features *ef;
5529 int ret = -ENOMEM;
5530
5531 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5532 if (!ef)
5533 goto out;
5534
5535 ef->f_kobj.kset = ext4_kset;
5536 init_completion(&ef->f_kobj_unregister);
5537 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5538 "features");
5539 if (ret) {
5540 kfree(ef);
5541 goto out;
5542 }
5543
5544 ext4_feat = ef;
5545 ret = 0;
5546 out:
5547 return ret;
5548 }
5549
5550 static void ext4_exit_feat_adverts(void)
5551 {
5552 kobject_put(&ext4_feat->f_kobj);
5553 wait_for_completion(&ext4_feat->f_kobj_unregister);
5554 kfree(ext4_feat);
5555 }
5556
5557 /* Shared across all ext4 file systems */
5558 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5559 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5560
5561 static int __init ext4_init_fs(void)
5562 {
5563 int i, err;
5564
5565 ext4_li_info = NULL;
5566 mutex_init(&ext4_li_mtx);
5567
5568 /* Build-time check for flags consistency */
5569 ext4_check_flag_values();
5570
5571 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5572 mutex_init(&ext4__aio_mutex[i]);
5573 init_waitqueue_head(&ext4__ioend_wq[i]);
5574 }
5575
5576 err = ext4_init_es();
5577 if (err)
5578 return err;
5579
5580 err = ext4_init_pageio();
5581 if (err)
5582 goto out7;
5583
5584 err = ext4_init_system_zone();
5585 if (err)
5586 goto out6;
5587 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5588 if (!ext4_kset) {
5589 err = -ENOMEM;
5590 goto out5;
5591 }
5592 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5593
5594 err = ext4_init_feat_adverts();
5595 if (err)
5596 goto out4;
5597
5598 err = ext4_init_mballoc();
5599 if (err)
5600 goto out2;
5601 else
5602 ext4_mballoc_ready = 1;
5603 err = init_inodecache();
5604 if (err)
5605 goto out1;
5606 register_as_ext3();
5607 register_as_ext2();
5608 err = register_filesystem(&ext4_fs_type);
5609 if (err)
5610 goto out;
5611
5612 return 0;
5613 out:
5614 unregister_as_ext2();
5615 unregister_as_ext3();
5616 destroy_inodecache();
5617 out1:
5618 ext4_mballoc_ready = 0;
5619 ext4_exit_mballoc();
5620 out2:
5621 ext4_exit_feat_adverts();
5622 out4:
5623 if (ext4_proc_root)
5624 remove_proc_entry("fs/ext4", NULL);
5625 kset_unregister(ext4_kset);
5626 out5:
5627 ext4_exit_system_zone();
5628 out6:
5629 ext4_exit_pageio();
5630 out7:
5631 ext4_exit_es();
5632
5633 return err;
5634 }
5635
5636 static void __exit ext4_exit_fs(void)
5637 {
5638 ext4_destroy_lazyinit_thread();
5639 unregister_as_ext2();
5640 unregister_as_ext3();
5641 unregister_filesystem(&ext4_fs_type);
5642 destroy_inodecache();
5643 ext4_exit_mballoc();
5644 ext4_exit_feat_adverts();
5645 remove_proc_entry("fs/ext4", NULL);
5646 kset_unregister(ext4_kset);
5647 ext4_exit_system_zone();
5648 ext4_exit_pageio();
5649 ext4_exit_es();
5650 }
5651
5652 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5653 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5654 MODULE_LICENSE("GPL");
5655 module_init(ext4_init_fs)
5656 module_exit(ext4_exit_fs)
Ubuntu Artful kernel mirror