?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
vfs: allow O_PATH file descriptors for fstatfs()
[linux-2.6.git] / fs / ext4 / super.c
blob2c2e6cbc6bedc549938262e50be8e118d1cb6de7
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
43
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
46
47 #include "ext4.h"
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
50 #include "xattr.h"
51 #include "acl.h"
52 #include "mballoc.h"
53
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
56
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
62
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
86
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94 };
95 MODULE_ALIAS_FS("ext2");
96 MODULE_ALIAS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #else
99 #define IS_EXT2_SB(sb) (0)
100 #endif
101
102
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
106 .name = "ext3",
107 .mount = ext4_mount,
108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
110 };
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114 #else
115 #define IS_EXT3_SB(sb) (0)
116 #endif
117
118 static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
120 {
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
123 return 1;
124
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
126 }
127
128 static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
130 {
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
133 __u32 csum;
134
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136
137 return cpu_to_le32(csum);
138 }
139
140 int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
142 {
143 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
144 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
145 return 1;
146
147 return es->s_checksum == ext4_superblock_csum(sb, es);
148 }
149
150 void ext4_superblock_csum_set(struct super_block *sb)
151 {
152 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
153
154 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
155 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
156 return;
157
158 es->s_checksum = ext4_superblock_csum(sb, es);
159 }
160
161 void *ext4_kvmalloc(size_t size, gfp_t flags)
162 {
163 void *ret;
164
165 ret = kmalloc(size, flags | __GFP_NOWARN);
166 if (!ret)
167 ret = __vmalloc(size, flags, PAGE_KERNEL);
168 return ret;
169 }
170
171 void *ext4_kvzalloc(size_t size, gfp_t flags)
172 {
173 void *ret;
174
175 ret = kzalloc(size, flags | __GFP_NOWARN);
176 if (!ret)
177 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
178 return ret;
179 }
180
181 void ext4_kvfree(void *ptr)
182 {
183 if (is_vmalloc_addr(ptr))
184 vfree(ptr);
185 else
186 kfree(ptr);
187
188 }
189
190 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
191 struct ext4_group_desc *bg)
192 {
193 return le32_to_cpu(bg->bg_block_bitmap_lo) |
194 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
195 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
196 }
197
198 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
199 struct ext4_group_desc *bg)
200 {
201 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
202 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
203 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
204 }
205
206 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
207 struct ext4_group_desc *bg)
208 {
209 return le32_to_cpu(bg->bg_inode_table_lo) |
210 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
211 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
212 }
213
214 __u32 ext4_free_group_clusters(struct super_block *sb,
215 struct ext4_group_desc *bg)
216 {
217 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
218 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
219 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
220 }
221
222 __u32 ext4_free_inodes_count(struct super_block *sb,
223 struct ext4_group_desc *bg)
224 {
225 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
226 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
227 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
228 }
229
230 __u32 ext4_used_dirs_count(struct super_block *sb,
231 struct ext4_group_desc *bg)
232 {
233 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
234 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
235 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
236 }
237
238 __u32 ext4_itable_unused_count(struct super_block *sb,
239 struct ext4_group_desc *bg)
240 {
241 return le16_to_cpu(bg->bg_itable_unused_lo) |
242 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
243 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
244 }
245
246 void ext4_block_bitmap_set(struct super_block *sb,
247 struct ext4_group_desc *bg, ext4_fsblk_t blk)
248 {
249 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
250 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
251 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
252 }
253
254 void ext4_inode_bitmap_set(struct super_block *sb,
255 struct ext4_group_desc *bg, ext4_fsblk_t blk)
256 {
257 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
258 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
259 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
260 }
261
262 void ext4_inode_table_set(struct super_block *sb,
263 struct ext4_group_desc *bg, ext4_fsblk_t blk)
264 {
265 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
266 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
267 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
268 }
269
270 void ext4_free_group_clusters_set(struct super_block *sb,
271 struct ext4_group_desc *bg, __u32 count)
272 {
273 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
274 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
275 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
276 }
277
278 void ext4_free_inodes_set(struct super_block *sb,
279 struct ext4_group_desc *bg, __u32 count)
280 {
281 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
282 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
283 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
284 }
285
286 void ext4_used_dirs_set(struct super_block *sb,
287 struct ext4_group_desc *bg, __u32 count)
288 {
289 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
290 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
291 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
292 }
293
294 void ext4_itable_unused_set(struct super_block *sb,
295 struct ext4_group_desc *bg, __u32 count)
296 {
297 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
298 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
299 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
300 }
301
302
303 static void __save_error_info(struct super_block *sb, const char *func,
304 unsigned int line)
305 {
306 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
307
308 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
309 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
310 es->s_last_error_time = cpu_to_le32(get_seconds());
311 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
312 es->s_last_error_line = cpu_to_le32(line);
313 if (!es->s_first_error_time) {
314 es->s_first_error_time = es->s_last_error_time;
315 strncpy(es->s_first_error_func, func,
316 sizeof(es->s_first_error_func));
317 es->s_first_error_line = cpu_to_le32(line);
318 es->s_first_error_ino = es->s_last_error_ino;
319 es->s_first_error_block = es->s_last_error_block;
320 }
321 /*
322 * Start the daily error reporting function if it hasn't been
323 * started already
324 */
325 if (!es->s_error_count)
326 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
327 le32_add_cpu(&es->s_error_count, 1);
328 }
329
330 static void save_error_info(struct super_block *sb, const char *func,
331 unsigned int line)
332 {
333 __save_error_info(sb, func, line);
334 ext4_commit_super(sb, 1);
335 }
336
337 /*
338 * The del_gendisk() function uninitializes the disk-specific data
339 * structures, including the bdi structure, without telling anyone
340 * else. Once this happens, any attempt to call mark_buffer_dirty()
341 * (for example, by ext4_commit_super), will cause a kernel OOPS.
342 * This is a kludge to prevent these oops until we can put in a proper
343 * hook in del_gendisk() to inform the VFS and file system layers.
344 */
345 static int block_device_ejected(struct super_block *sb)
346 {
347 struct inode *bd_inode = sb->s_bdev->bd_inode;
348 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
349
350 return bdi->dev == NULL;
351 }
352
353 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
354 {
355 struct super_block *sb = journal->j_private;
356 struct ext4_sb_info *sbi = EXT4_SB(sb);
357 int error = is_journal_aborted(journal);
358 struct ext4_journal_cb_entry *jce;
359
360 BUG_ON(txn->t_state == T_FINISHED);
361 spin_lock(&sbi->s_md_lock);
362 while (!list_empty(&txn->t_private_list)) {
363 jce = list_entry(txn->t_private_list.next,
364 struct ext4_journal_cb_entry, jce_list);
365 list_del_init(&jce->jce_list);
366 spin_unlock(&sbi->s_md_lock);
367 jce->jce_func(sb, jce, error);
368 spin_lock(&sbi->s_md_lock);
369 }
370 spin_unlock(&sbi->s_md_lock);
371 }
372
373 /* Deal with the reporting of failure conditions on a filesystem such as
374 * inconsistencies detected or read IO failures.
375 *
376 * On ext2, we can store the error state of the filesystem in the
377 * superblock. That is not possible on ext4, because we may have other
378 * write ordering constraints on the superblock which prevent us from
379 * writing it out straight away; and given that the journal is about to
380 * be aborted, we can't rely on the current, or future, transactions to
381 * write out the superblock safely.
382 *
383 * We'll just use the jbd2_journal_abort() error code to record an error in
384 * the journal instead. On recovery, the journal will complain about
385 * that error until we've noted it down and cleared it.
386 */
387
388 static void ext4_handle_error(struct super_block *sb)
389 {
390 if (sb->s_flags & MS_RDONLY)
391 return;
392
393 if (!test_opt(sb, ERRORS_CONT)) {
394 journal_t *journal = EXT4_SB(sb)->s_journal;
395
396 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
397 if (journal)
398 jbd2_journal_abort(journal, -EIO);
399 }
400 if (test_opt(sb, ERRORS_RO)) {
401 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
402 /*
403 * Make sure updated value of ->s_mount_flags will be visible
404 * before ->s_flags update
405 */
406 smp_wmb();
407 sb->s_flags |= MS_RDONLY;
408 }
409 if (test_opt(sb, ERRORS_PANIC))
410 panic("EXT4-fs (device %s): panic forced after error\n",
411 sb->s_id);
412 }
413
414 void __ext4_error(struct super_block *sb, const char *function,
415 unsigned int line, const char *fmt, ...)
416 {
417 struct va_format vaf;
418 va_list args;
419
420 va_start(args, fmt);
421 vaf.fmt = fmt;
422 vaf.va = &args;
423 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
424 sb->s_id, function, line, current->comm, &vaf);
425 va_end(args);
426 save_error_info(sb, function, line);
427
428 ext4_handle_error(sb);
429 }
430
431 void __ext4_error_inode(struct inode *inode, const char *function,
432 unsigned int line, ext4_fsblk_t block,
433 const char *fmt, ...)
434 {
435 va_list args;
436 struct va_format vaf;
437 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
438
439 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
440 es->s_last_error_block = cpu_to_le64(block);
441 save_error_info(inode->i_sb, function, line);
442 va_start(args, fmt);
443 vaf.fmt = fmt;
444 vaf.va = &args;
445 if (block)
446 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
447 "inode #%lu: block %llu: comm %s: %pV\n",
448 inode->i_sb->s_id, function, line, inode->i_ino,
449 block, current->comm, &vaf);
450 else
451 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
452 "inode #%lu: comm %s: %pV\n",
453 inode->i_sb->s_id, function, line, inode->i_ino,
454 current->comm, &vaf);
455 va_end(args);
456
457 ext4_handle_error(inode->i_sb);
458 }
459
460 void __ext4_error_file(struct file *file, const char *function,
461 unsigned int line, ext4_fsblk_t block,
462 const char *fmt, ...)
463 {
464 va_list args;
465 struct va_format vaf;
466 struct ext4_super_block *es;
467 struct inode *inode = file_inode(file);
468 char pathname[80], *path;
469
470 es = EXT4_SB(inode->i_sb)->s_es;
471 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
472 save_error_info(inode->i_sb, function, line);
473 path = d_path(&(file->f_path), pathname, sizeof(pathname));
474 if (IS_ERR(path))
475 path = "(unknown)";
476 va_start(args, fmt);
477 vaf.fmt = fmt;
478 vaf.va = &args;
479 if (block)
480 printk(KERN_CRIT
481 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
482 "block %llu: comm %s: path %s: %pV\n",
483 inode->i_sb->s_id, function, line, inode->i_ino,
484 block, current->comm, path, &vaf);
485 else
486 printk(KERN_CRIT
487 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
488 "comm %s: path %s: %pV\n",
489 inode->i_sb->s_id, function, line, inode->i_ino,
490 current->comm, path, &vaf);
491 va_end(args);
492
493 ext4_handle_error(inode->i_sb);
494 }
495
496 const char *ext4_decode_error(struct super_block *sb, int errno,
497 char nbuf[16])
498 {
499 char *errstr = NULL;
500
501 switch (errno) {
502 case -EIO:
503 errstr = "IO failure";
504 break;
505 case -ENOMEM:
506 errstr = "Out of memory";
507 break;
508 case -EROFS:
509 if (!sb || (EXT4_SB(sb)->s_journal &&
510 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
511 errstr = "Journal has aborted";
512 else
513 errstr = "Readonly filesystem";
514 break;
515 default:
516 /* If the caller passed in an extra buffer for unknown
517 * errors, textualise them now. Else we just return
518 * NULL. */
519 if (nbuf) {
520 /* Check for truncated error codes... */
521 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
522 errstr = nbuf;
523 }
524 break;
525 }
526
527 return errstr;
528 }
529
530 /* __ext4_std_error decodes expected errors from journaling functions
531 * automatically and invokes the appropriate error response. */
532
533 void __ext4_std_error(struct super_block *sb, const char *function,
534 unsigned int line, int errno)
535 {
536 char nbuf[16];
537 const char *errstr;
538
539 /* Special case: if the error is EROFS, and we're not already
540 * inside a transaction, then there's really no point in logging
541 * an error. */
542 if (errno == -EROFS && journal_current_handle() == NULL &&
543 (sb->s_flags & MS_RDONLY))
544 return;
545
546 errstr = ext4_decode_error(sb, errno, nbuf);
547 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
548 sb->s_id, function, line, errstr);
549 save_error_info(sb, function, line);
550
551 ext4_handle_error(sb);
552 }
553
554 /*
555 * ext4_abort is a much stronger failure handler than ext4_error. The
556 * abort function may be used to deal with unrecoverable failures such
557 * as journal IO errors or ENOMEM at a critical moment in log management.
558 *
559 * We unconditionally force the filesystem into an ABORT|READONLY state,
560 * unless the error response on the fs has been set to panic in which
561 * case we take the easy way out and panic immediately.
562 */
563
564 void __ext4_abort(struct super_block *sb, const char *function,
565 unsigned int line, const char *fmt, ...)
566 {
567 va_list args;
568
569 save_error_info(sb, function, line);
570 va_start(args, fmt);
571 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
572 function, line);
573 vprintk(fmt, args);
574 printk("\n");
575 va_end(args);
576
577 if ((sb->s_flags & MS_RDONLY) == 0) {
578 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
579 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
580 /*
581 * Make sure updated value of ->s_mount_flags will be visible
582 * before ->s_flags update
583 */
584 smp_wmb();
585 sb->s_flags |= MS_RDONLY;
586 if (EXT4_SB(sb)->s_journal)
587 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
588 save_error_info(sb, function, line);
589 }
590 if (test_opt(sb, ERRORS_PANIC))
591 panic("EXT4-fs panic from previous error\n");
592 }
593
594 void __ext4_msg(struct super_block *sb,
595 const char *prefix, const char *fmt, ...)
596 {
597 struct va_format vaf;
598 va_list args;
599
600 va_start(args, fmt);
601 vaf.fmt = fmt;
602 vaf.va = &args;
603 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
604 va_end(args);
605 }
606
607 void __ext4_warning(struct super_block *sb, const char *function,
608 unsigned int line, const char *fmt, ...)
609 {
610 struct va_format vaf;
611 va_list args;
612
613 va_start(args, fmt);
614 vaf.fmt = fmt;
615 vaf.va = &args;
616 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
617 sb->s_id, function, line, &vaf);
618 va_end(args);
619 }
620
621 void __ext4_grp_locked_error(const char *function, unsigned int line,
622 struct super_block *sb, ext4_group_t grp,
623 unsigned long ino, ext4_fsblk_t block,
624 const char *fmt, ...)
625 __releases(bitlock)
626 __acquires(bitlock)
627 {
628 struct va_format vaf;
629 va_list args;
630 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
631
632 es->s_last_error_ino = cpu_to_le32(ino);
633 es->s_last_error_block = cpu_to_le64(block);
634 __save_error_info(sb, function, line);
635
636 va_start(args, fmt);
637
638 vaf.fmt = fmt;
639 vaf.va = &args;
640 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
641 sb->s_id, function, line, grp);
642 if (ino)
643 printk(KERN_CONT "inode %lu: ", ino);
644 if (block)
645 printk(KERN_CONT "block %llu:", (unsigned long long) block);
646 printk(KERN_CONT "%pV\n", &vaf);
647 va_end(args);
648
649 if (test_opt(sb, ERRORS_CONT)) {
650 ext4_commit_super(sb, 0);
651 return;
652 }
653
654 ext4_unlock_group(sb, grp);
655 ext4_handle_error(sb);
656 /*
657 * We only get here in the ERRORS_RO case; relocking the group
658 * may be dangerous, but nothing bad will happen since the
659 * filesystem will have already been marked read/only and the
660 * journal has been aborted. We return 1 as a hint to callers
661 * who might what to use the return value from
662 * ext4_grp_locked_error() to distinguish between the
663 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
664 * aggressively from the ext4 function in question, with a
665 * more appropriate error code.
666 */
667 ext4_lock_group(sb, grp);
668 return;
669 }
670
671 void ext4_update_dynamic_rev(struct super_block *sb)
672 {
673 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
674
675 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
676 return;
677
678 ext4_warning(sb,
679 "updating to rev %d because of new feature flag, "
680 "running e2fsck is recommended",
681 EXT4_DYNAMIC_REV);
682
683 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
684 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
685 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
686 /* leave es->s_feature_*compat flags alone */
687 /* es->s_uuid will be set by e2fsck if empty */
688
689 /*
690 * The rest of the superblock fields should be zero, and if not it
691 * means they are likely already in use, so leave them alone. We
692 * can leave it up to e2fsck to clean up any inconsistencies there.
693 */
694 }
695
696 /*
697 * Open the external journal device
698 */
699 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
700 {
701 struct block_device *bdev;
702 char b[BDEVNAME_SIZE];
703
704 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
705 if (IS_ERR(bdev))
706 goto fail;
707 return bdev;
708
709 fail:
710 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
711 __bdevname(dev, b), PTR_ERR(bdev));
712 return NULL;
713 }
714
715 /*
716 * Release the journal device
717 */
718 static void ext4_blkdev_put(struct block_device *bdev)
719 {
720 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
721 }
722
723 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
724 {
725 struct block_device *bdev;
726 bdev = sbi->journal_bdev;
727 if (bdev) {
728 ext4_blkdev_put(bdev);
729 sbi->journal_bdev = NULL;
730 }
731 }
732
733 static inline struct inode *orphan_list_entry(struct list_head *l)
734 {
735 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
736 }
737
738 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
739 {
740 struct list_head *l;
741
742 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
743 le32_to_cpu(sbi->s_es->s_last_orphan));
744
745 printk(KERN_ERR "sb_info orphan list:\n");
746 list_for_each(l, &sbi->s_orphan) {
747 struct inode *inode = orphan_list_entry(l);
748 printk(KERN_ERR " "
749 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
750 inode->i_sb->s_id, inode->i_ino, inode,
751 inode->i_mode, inode->i_nlink,
752 NEXT_ORPHAN(inode));
753 }
754 }
755
756 static void ext4_put_super(struct super_block *sb)
757 {
758 struct ext4_sb_info *sbi = EXT4_SB(sb);
759 struct ext4_super_block *es = sbi->s_es;
760 int i, err;
761
762 ext4_unregister_li_request(sb);
763 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
764
765 flush_workqueue(sbi->rsv_conversion_wq);
766 destroy_workqueue(sbi->rsv_conversion_wq);
767
768 if (sbi->s_journal) {
769 err = jbd2_journal_destroy(sbi->s_journal);
770 sbi->s_journal = NULL;
771 if (err < 0)
772 ext4_abort(sb, "Couldn't clean up the journal");
773 }
774
775 ext4_es_unregister_shrinker(sbi);
776 del_timer(&sbi->s_err_report);
777 ext4_release_system_zone(sb);
778 ext4_mb_release(sb);
779 ext4_ext_release(sb);
780 ext4_xattr_put_super(sb);
781
782 if (!(sb->s_flags & MS_RDONLY)) {
783 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
784 es->s_state = cpu_to_le16(sbi->s_mount_state);
785 }
786 if (!(sb->s_flags & MS_RDONLY))
787 ext4_commit_super(sb, 1);
788
789 if (sbi->s_proc) {
790 remove_proc_entry("options", sbi->s_proc);
791 remove_proc_entry(sb->s_id, ext4_proc_root);
792 }
793 kobject_del(&sbi->s_kobj);
794
795 for (i = 0; i < sbi->s_gdb_count; i++)
796 brelse(sbi->s_group_desc[i]);
797 ext4_kvfree(sbi->s_group_desc);
798 ext4_kvfree(sbi->s_flex_groups);
799 percpu_counter_destroy(&sbi->s_freeclusters_counter);
800 percpu_counter_destroy(&sbi->s_freeinodes_counter);
801 percpu_counter_destroy(&sbi->s_dirs_counter);
802 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
803 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
804 brelse(sbi->s_sbh);
805 #ifdef CONFIG_QUOTA
806 for (i = 0; i < MAXQUOTAS; i++)
807 kfree(sbi->s_qf_names[i]);
808 #endif
809
810 /* Debugging code just in case the in-memory inode orphan list
811 * isn't empty. The on-disk one can be non-empty if we've
812 * detected an error and taken the fs readonly, but the
813 * in-memory list had better be clean by this point. */
814 if (!list_empty(&sbi->s_orphan))
815 dump_orphan_list(sb, sbi);
816 J_ASSERT(list_empty(&sbi->s_orphan));
817
818 invalidate_bdev(sb->s_bdev);
819 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
820 /*
821 * Invalidate the journal device's buffers. We don't want them
822 * floating about in memory - the physical journal device may
823 * hotswapped, and it breaks the `ro-after' testing code.
824 */
825 sync_blockdev(sbi->journal_bdev);
826 invalidate_bdev(sbi->journal_bdev);
827 ext4_blkdev_remove(sbi);
828 }
829 if (sbi->s_mmp_tsk)
830 kthread_stop(sbi->s_mmp_tsk);
831 sb->s_fs_info = NULL;
832 /*
833 * Now that we are completely done shutting down the
834 * superblock, we need to actually destroy the kobject.
835 */
836 kobject_put(&sbi->s_kobj);
837 wait_for_completion(&sbi->s_kobj_unregister);
838 if (sbi->s_chksum_driver)
839 crypto_free_shash(sbi->s_chksum_driver);
840 kfree(sbi->s_blockgroup_lock);
841 kfree(sbi);
842 }
843
844 static struct kmem_cache *ext4_inode_cachep;
845
846 /*
847 * Called inside transaction, so use GFP_NOFS
848 */
849 static struct inode *ext4_alloc_inode(struct super_block *sb)
850 {
851 struct ext4_inode_info *ei;
852
853 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
854 if (!ei)
855 return NULL;
856
857 ei->vfs_inode.i_version = 1;
858 INIT_LIST_HEAD(&ei->i_prealloc_list);
859 spin_lock_init(&ei->i_prealloc_lock);
860 ext4_es_init_tree(&ei->i_es_tree);
861 rwlock_init(&ei->i_es_lock);
862 INIT_LIST_HEAD(&ei->i_es_lru);
863 ei->i_es_lru_nr = 0;
864 ei->i_touch_when = 0;
865 ei->i_reserved_data_blocks = 0;
866 ei->i_reserved_meta_blocks = 0;
867 ei->i_allocated_meta_blocks = 0;
868 ei->i_da_metadata_calc_len = 0;
869 ei->i_da_metadata_calc_last_lblock = 0;
870 spin_lock_init(&(ei->i_block_reservation_lock));
871 #ifdef CONFIG_QUOTA
872 ei->i_reserved_quota = 0;
873 #endif
874 ei->jinode = NULL;
875 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
876 spin_lock_init(&ei->i_completed_io_lock);
877 ei->i_sync_tid = 0;
878 ei->i_datasync_tid = 0;
879 atomic_set(&ei->i_ioend_count, 0);
880 atomic_set(&ei->i_unwritten, 0);
881 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
882
883 return &ei->vfs_inode;
884 }
885
886 static int ext4_drop_inode(struct inode *inode)
887 {
888 int drop = generic_drop_inode(inode);
889
890 trace_ext4_drop_inode(inode, drop);
891 return drop;
892 }
893
894 static void ext4_i_callback(struct rcu_head *head)
895 {
896 struct inode *inode = container_of(head, struct inode, i_rcu);
897 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
898 }
899
900 static void ext4_destroy_inode(struct inode *inode)
901 {
902 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
903 ext4_msg(inode->i_sb, KERN_ERR,
904 "Inode %lu (%p): orphan list check failed!",
905 inode->i_ino, EXT4_I(inode));
906 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
907 EXT4_I(inode), sizeof(struct ext4_inode_info),
908 true);
909 dump_stack();
910 }
911 call_rcu(&inode->i_rcu, ext4_i_callback);
912 }
913
914 static void init_once(void *foo)
915 {
916 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
917
918 INIT_LIST_HEAD(&ei->i_orphan);
919 init_rwsem(&ei->xattr_sem);
920 init_rwsem(&ei->i_data_sem);
921 inode_init_once(&ei->vfs_inode);
922 }
923
924 static int init_inodecache(void)
925 {
926 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
927 sizeof(struct ext4_inode_info),
928 0, (SLAB_RECLAIM_ACCOUNT|
929 SLAB_MEM_SPREAD),
930 init_once);
931 if (ext4_inode_cachep == NULL)
932 return -ENOMEM;
933 return 0;
934 }
935
936 static void destroy_inodecache(void)
937 {
938 /*
939 * Make sure all delayed rcu free inodes are flushed before we
940 * destroy cache.
941 */
942 rcu_barrier();
943 kmem_cache_destroy(ext4_inode_cachep);
944 }
945
946 void ext4_clear_inode(struct inode *inode)
947 {
948 invalidate_inode_buffers(inode);
949 clear_inode(inode);
950 dquot_drop(inode);
951 ext4_discard_preallocations(inode);
952 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
953 ext4_es_lru_del(inode);
954 if (EXT4_I(inode)->jinode) {
955 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
956 EXT4_I(inode)->jinode);
957 jbd2_free_inode(EXT4_I(inode)->jinode);
958 EXT4_I(inode)->jinode = NULL;
959 }
960 }
961
962 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
963 u64 ino, u32 generation)
964 {
965 struct inode *inode;
966
967 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
968 return ERR_PTR(-ESTALE);
969 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
970 return ERR_PTR(-ESTALE);
971
972 /* iget isn't really right if the inode is currently unallocated!!
973 *
974 * ext4_read_inode will return a bad_inode if the inode had been
975 * deleted, so we should be safe.
976 *
977 * Currently we don't know the generation for parent directory, so
978 * a generation of 0 means "accept any"
979 */
980 inode = ext4_iget(sb, ino);
981 if (IS_ERR(inode))
982 return ERR_CAST(inode);
983 if (generation && inode->i_generation != generation) {
984 iput(inode);
985 return ERR_PTR(-ESTALE);
986 }
987
988 return inode;
989 }
990
991 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
992 int fh_len, int fh_type)
993 {
994 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
995 ext4_nfs_get_inode);
996 }
997
998 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
999 int fh_len, int fh_type)
1000 {
1001 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1002 ext4_nfs_get_inode);
1003 }
1004
1005 /*
1006 * Try to release metadata pages (indirect blocks, directories) which are
1007 * mapped via the block device. Since these pages could have journal heads
1008 * which would prevent try_to_free_buffers() from freeing them, we must use
1009 * jbd2 layer's try_to_free_buffers() function to release them.
1010 */
1011 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1012 gfp_t wait)
1013 {
1014 journal_t *journal = EXT4_SB(sb)->s_journal;
1015
1016 WARN_ON(PageChecked(page));
1017 if (!page_has_buffers(page))
1018 return 0;
1019 if (journal)
1020 return jbd2_journal_try_to_free_buffers(journal, page,
1021 wait & ~__GFP_WAIT);
1022 return try_to_free_buffers(page);
1023 }
1024
1025 #ifdef CONFIG_QUOTA
1026 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1027 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1028
1029 static int ext4_write_dquot(struct dquot *dquot);
1030 static int ext4_acquire_dquot(struct dquot *dquot);
1031 static int ext4_release_dquot(struct dquot *dquot);
1032 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1033 static int ext4_write_info(struct super_block *sb, int type);
1034 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1035 struct path *path);
1036 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1037 int format_id);
1038 static int ext4_quota_off(struct super_block *sb, int type);
1039 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1040 static int ext4_quota_on_mount(struct super_block *sb, int type);
1041 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1042 size_t len, loff_t off);
1043 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1044 const char *data, size_t len, loff_t off);
1045 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1046 unsigned int flags);
1047 static int ext4_enable_quotas(struct super_block *sb);
1048
1049 static const struct dquot_operations ext4_quota_operations = {
1050 .get_reserved_space = ext4_get_reserved_space,
1051 .write_dquot = ext4_write_dquot,
1052 .acquire_dquot = ext4_acquire_dquot,
1053 .release_dquot = ext4_release_dquot,
1054 .mark_dirty = ext4_mark_dquot_dirty,
1055 .write_info = ext4_write_info,
1056 .alloc_dquot = dquot_alloc,
1057 .destroy_dquot = dquot_destroy,
1058 };
1059
1060 static const struct quotactl_ops ext4_qctl_operations = {
1061 .quota_on = ext4_quota_on,
1062 .quota_off = ext4_quota_off,
1063 .quota_sync = dquot_quota_sync,
1064 .get_info = dquot_get_dqinfo,
1065 .set_info = dquot_set_dqinfo,
1066 .get_dqblk = dquot_get_dqblk,
1067 .set_dqblk = dquot_set_dqblk
1068 };
1069
1070 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1071 .quota_on_meta = ext4_quota_on_sysfile,
1072 .quota_off = ext4_quota_off_sysfile,
1073 .quota_sync = dquot_quota_sync,
1074 .get_info = dquot_get_dqinfo,
1075 .set_info = dquot_set_dqinfo,
1076 .get_dqblk = dquot_get_dqblk,
1077 .set_dqblk = dquot_set_dqblk
1078 };
1079 #endif
1080
1081 static const struct super_operations ext4_sops = {
1082 .alloc_inode = ext4_alloc_inode,
1083 .destroy_inode = ext4_destroy_inode,
1084 .write_inode = ext4_write_inode,
1085 .dirty_inode = ext4_dirty_inode,
1086 .drop_inode = ext4_drop_inode,
1087 .evict_inode = ext4_evict_inode,
1088 .put_super = ext4_put_super,
1089 .sync_fs = ext4_sync_fs,
1090 .freeze_fs = ext4_freeze,
1091 .unfreeze_fs = ext4_unfreeze,
1092 .statfs = ext4_statfs,
1093 .remount_fs = ext4_remount,
1094 .show_options = ext4_show_options,
1095 #ifdef CONFIG_QUOTA
1096 .quota_read = ext4_quota_read,
1097 .quota_write = ext4_quota_write,
1098 #endif
1099 .bdev_try_to_free_page = bdev_try_to_free_page,
1100 };
1101
1102 static const struct super_operations ext4_nojournal_sops = {
1103 .alloc_inode = ext4_alloc_inode,
1104 .destroy_inode = ext4_destroy_inode,
1105 .write_inode = ext4_write_inode,
1106 .dirty_inode = ext4_dirty_inode,
1107 .drop_inode = ext4_drop_inode,
1108 .evict_inode = ext4_evict_inode,
1109 .sync_fs = ext4_sync_fs_nojournal,
1110 .put_super = ext4_put_super,
1111 .statfs = ext4_statfs,
1112 .remount_fs = ext4_remount,
1113 .show_options = ext4_show_options,
1114 #ifdef CONFIG_QUOTA
1115 .quota_read = ext4_quota_read,
1116 .quota_write = ext4_quota_write,
1117 #endif
1118 .bdev_try_to_free_page = bdev_try_to_free_page,
1119 };
1120
1121 static const struct export_operations ext4_export_ops = {
1122 .fh_to_dentry = ext4_fh_to_dentry,
1123 .fh_to_parent = ext4_fh_to_parent,
1124 .get_parent = ext4_get_parent,
1125 };
1126
1127 enum {
1128 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1129 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1130 Opt_nouid32, Opt_debug, Opt_removed,
1131 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1132 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1133 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1134 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1135 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1136 Opt_data_err_abort, Opt_data_err_ignore,
1137 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1138 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1139 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1140 Opt_usrquota, Opt_grpquota, Opt_i_version,
1141 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1142 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1143 Opt_inode_readahead_blks, Opt_journal_ioprio,
1144 Opt_dioread_nolock, Opt_dioread_lock,
1145 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1146 Opt_max_dir_size_kb,
1147 };
1148
1149 static const match_table_t tokens = {
1150 {Opt_bsd_df, "bsddf"},
1151 {Opt_minix_df, "minixdf"},
1152 {Opt_grpid, "grpid"},
1153 {Opt_grpid, "bsdgroups"},
1154 {Opt_nogrpid, "nogrpid"},
1155 {Opt_nogrpid, "sysvgroups"},
1156 {Opt_resgid, "resgid=%u"},
1157 {Opt_resuid, "resuid=%u"},
1158 {Opt_sb, "sb=%u"},
1159 {Opt_err_cont, "errors=continue"},
1160 {Opt_err_panic, "errors=panic"},
1161 {Opt_err_ro, "errors=remount-ro"},
1162 {Opt_nouid32, "nouid32"},
1163 {Opt_debug, "debug"},
1164 {Opt_removed, "oldalloc"},
1165 {Opt_removed, "orlov"},
1166 {Opt_user_xattr, "user_xattr"},
1167 {Opt_nouser_xattr, "nouser_xattr"},
1168 {Opt_acl, "acl"},
1169 {Opt_noacl, "noacl"},
1170 {Opt_noload, "norecovery"},
1171 {Opt_noload, "noload"},
1172 {Opt_removed, "nobh"},
1173 {Opt_removed, "bh"},
1174 {Opt_commit, "commit=%u"},
1175 {Opt_min_batch_time, "min_batch_time=%u"},
1176 {Opt_max_batch_time, "max_batch_time=%u"},
1177 {Opt_journal_dev, "journal_dev=%u"},
1178 {Opt_journal_path, "journal_path=%s"},
1179 {Opt_journal_checksum, "journal_checksum"},
1180 {Opt_journal_async_commit, "journal_async_commit"},
1181 {Opt_abort, "abort"},
1182 {Opt_data_journal, "data=journal"},
1183 {Opt_data_ordered, "data=ordered"},
1184 {Opt_data_writeback, "data=writeback"},
1185 {Opt_data_err_abort, "data_err=abort"},
1186 {Opt_data_err_ignore, "data_err=ignore"},
1187 {Opt_offusrjquota, "usrjquota="},
1188 {Opt_usrjquota, "usrjquota=%s"},
1189 {Opt_offgrpjquota, "grpjquota="},
1190 {Opt_grpjquota, "grpjquota=%s"},
1191 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1192 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1193 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1194 {Opt_grpquota, "grpquota"},
1195 {Opt_noquota, "noquota"},
1196 {Opt_quota, "quota"},
1197 {Opt_usrquota, "usrquota"},
1198 {Opt_barrier, "barrier=%u"},
1199 {Opt_barrier, "barrier"},
1200 {Opt_nobarrier, "nobarrier"},
1201 {Opt_i_version, "i_version"},
1202 {Opt_stripe, "stripe=%u"},
1203 {Opt_delalloc, "delalloc"},
1204 {Opt_nodelalloc, "nodelalloc"},
1205 {Opt_removed, "mblk_io_submit"},
1206 {Opt_removed, "nomblk_io_submit"},
1207 {Opt_block_validity, "block_validity"},
1208 {Opt_noblock_validity, "noblock_validity"},
1209 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1210 {Opt_journal_ioprio, "journal_ioprio=%u"},
1211 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1212 {Opt_auto_da_alloc, "auto_da_alloc"},
1213 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1214 {Opt_dioread_nolock, "dioread_nolock"},
1215 {Opt_dioread_lock, "dioread_lock"},
1216 {Opt_discard, "discard"},
1217 {Opt_nodiscard, "nodiscard"},
1218 {Opt_init_itable, "init_itable=%u"},
1219 {Opt_init_itable, "init_itable"},
1220 {Opt_noinit_itable, "noinit_itable"},
1221 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1222 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1223 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1224 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1225 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1226 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1227 {Opt_err, NULL},
1228 };
1229
1230 static ext4_fsblk_t get_sb_block(void **data)
1231 {
1232 ext4_fsblk_t sb_block;
1233 char *options = (char *) *data;
1234
1235 if (!options || strncmp(options, "sb=", 3) != 0)
1236 return 1; /* Default location */
1237
1238 options += 3;
1239 /* TODO: use simple_strtoll with >32bit ext4 */
1240 sb_block = simple_strtoul(options, &options, 0);
1241 if (*options && *options != ',') {
1242 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1243 (char *) *data);
1244 return 1;
1245 }
1246 if (*options == ',')
1247 options++;
1248 *data = (void *) options;
1249
1250 return sb_block;
1251 }
1252
1253 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1254 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1255 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1256
1257 #ifdef CONFIG_QUOTA
1258 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1259 {
1260 struct ext4_sb_info *sbi = EXT4_SB(sb);
1261 char *qname;
1262 int ret = -1;
1263
1264 if (sb_any_quota_loaded(sb) &&
1265 !sbi->s_qf_names[qtype]) {
1266 ext4_msg(sb, KERN_ERR,
1267 "Cannot change journaled "
1268 "quota options when quota turned on");
1269 return -1;
1270 }
1271 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1272 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1273 "when QUOTA feature is enabled");
1274 return -1;
1275 }
1276 qname = match_strdup(args);
1277 if (!qname) {
1278 ext4_msg(sb, KERN_ERR,
1279 "Not enough memory for storing quotafile name");
1280 return -1;
1281 }
1282 if (sbi->s_qf_names[qtype]) {
1283 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1284 ret = 1;
1285 else
1286 ext4_msg(sb, KERN_ERR,
1287 "%s quota file already specified",
1288 QTYPE2NAME(qtype));
1289 goto errout;
1290 }
1291 if (strchr(qname, '/')) {
1292 ext4_msg(sb, KERN_ERR,
1293 "quotafile must be on filesystem root");
1294 goto errout;
1295 }
1296 sbi->s_qf_names[qtype] = qname;
1297 set_opt(sb, QUOTA);
1298 return 1;
1299 errout:
1300 kfree(qname);
1301 return ret;
1302 }
1303
1304 static int clear_qf_name(struct super_block *sb, int qtype)
1305 {
1306
1307 struct ext4_sb_info *sbi = EXT4_SB(sb);
1308
1309 if (sb_any_quota_loaded(sb) &&
1310 sbi->s_qf_names[qtype]) {
1311 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1312 " when quota turned on");
1313 return -1;
1314 }
1315 kfree(sbi->s_qf_names[qtype]);
1316 sbi->s_qf_names[qtype] = NULL;
1317 return 1;
1318 }
1319 #endif
1320
1321 #define MOPT_SET 0x0001
1322 #define MOPT_CLEAR 0x0002
1323 #define MOPT_NOSUPPORT 0x0004
1324 #define MOPT_EXPLICIT 0x0008
1325 #define MOPT_CLEAR_ERR 0x0010
1326 #define MOPT_GTE0 0x0020
1327 #ifdef CONFIG_QUOTA
1328 #define MOPT_Q 0
1329 #define MOPT_QFMT 0x0040
1330 #else
1331 #define MOPT_Q MOPT_NOSUPPORT
1332 #define MOPT_QFMT MOPT_NOSUPPORT
1333 #endif
1334 #define MOPT_DATAJ 0x0080
1335 #define MOPT_NO_EXT2 0x0100
1336 #define MOPT_NO_EXT3 0x0200
1337 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1338 #define MOPT_STRING 0x0400
1339
1340 static const struct mount_opts {
1341 int token;
1342 int mount_opt;
1343 int flags;
1344 } ext4_mount_opts[] = {
1345 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1346 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1347 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1348 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1349 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1350 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1351 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1352 MOPT_EXT4_ONLY | MOPT_SET},
1353 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1354 MOPT_EXT4_ONLY | MOPT_CLEAR},
1355 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1356 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1357 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1358 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1359 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1360 MOPT_EXT4_ONLY | MOPT_CLEAR},
1361 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1362 MOPT_EXT4_ONLY | MOPT_SET},
1363 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1364 EXT4_MOUNT_JOURNAL_CHECKSUM),
1365 MOPT_EXT4_ONLY | MOPT_SET},
1366 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1367 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1368 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1369 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1370 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1371 MOPT_NO_EXT2 | MOPT_SET},
1372 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1373 MOPT_NO_EXT2 | MOPT_CLEAR},
1374 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1375 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1376 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1377 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1378 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1379 {Opt_commit, 0, MOPT_GTE0},
1380 {Opt_max_batch_time, 0, MOPT_GTE0},
1381 {Opt_min_batch_time, 0, MOPT_GTE0},
1382 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1383 {Opt_init_itable, 0, MOPT_GTE0},
1384 {Opt_stripe, 0, MOPT_GTE0},
1385 {Opt_resuid, 0, MOPT_GTE0},
1386 {Opt_resgid, 0, MOPT_GTE0},
1387 {Opt_journal_dev, 0, MOPT_GTE0},
1388 {Opt_journal_path, 0, MOPT_STRING},
1389 {Opt_journal_ioprio, 0, MOPT_GTE0},
1390 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1391 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1392 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1393 MOPT_NO_EXT2 | MOPT_DATAJ},
1394 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1395 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1396 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1397 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1398 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1399 #else
1400 {Opt_acl, 0, MOPT_NOSUPPORT},
1401 {Opt_noacl, 0, MOPT_NOSUPPORT},
1402 #endif
1403 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1404 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1405 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1406 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1407 MOPT_SET | MOPT_Q},
1408 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1409 MOPT_SET | MOPT_Q},
1410 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1411 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1412 {Opt_usrjquota, 0, MOPT_Q},
1413 {Opt_grpjquota, 0, MOPT_Q},
1414 {Opt_offusrjquota, 0, MOPT_Q},
1415 {Opt_offgrpjquota, 0, MOPT_Q},
1416 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1417 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1418 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1419 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1420 {Opt_err, 0, 0}
1421 };
1422
1423 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1424 substring_t *args, unsigned long *journal_devnum,
1425 unsigned int *journal_ioprio, int is_remount)
1426 {
1427 struct ext4_sb_info *sbi = EXT4_SB(sb);
1428 const struct mount_opts *m;
1429 kuid_t uid;
1430 kgid_t gid;
1431 int arg = 0;
1432
1433 #ifdef CONFIG_QUOTA
1434 if (token == Opt_usrjquota)
1435 return set_qf_name(sb, USRQUOTA, &args[0]);
1436 else if (token == Opt_grpjquota)
1437 return set_qf_name(sb, GRPQUOTA, &args[0]);
1438 else if (token == Opt_offusrjquota)
1439 return clear_qf_name(sb, USRQUOTA);
1440 else if (token == Opt_offgrpjquota)
1441 return clear_qf_name(sb, GRPQUOTA);
1442 #endif
1443 switch (token) {
1444 case Opt_noacl:
1445 case Opt_nouser_xattr:
1446 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1447 break;
1448 case Opt_sb:
1449 return 1; /* handled by get_sb_block() */
1450 case Opt_removed:
1451 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1452 return 1;
1453 case Opt_abort:
1454 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1455 return 1;
1456 case Opt_i_version:
1457 sb->s_flags |= MS_I_VERSION;
1458 return 1;
1459 }
1460
1461 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1462 if (token == m->token)
1463 break;
1464
1465 if (m->token == Opt_err) {
1466 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1467 "or missing value", opt);
1468 return -1;
1469 }
1470
1471 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1472 ext4_msg(sb, KERN_ERR,
1473 "Mount option \"%s\" incompatible with ext2", opt);
1474 return -1;
1475 }
1476 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1477 ext4_msg(sb, KERN_ERR,
1478 "Mount option \"%s\" incompatible with ext3", opt);
1479 return -1;
1480 }
1481
1482 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1483 return -1;
1484 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1485 return -1;
1486 if (m->flags & MOPT_EXPLICIT)
1487 set_opt2(sb, EXPLICIT_DELALLOC);
1488 if (m->flags & MOPT_CLEAR_ERR)
1489 clear_opt(sb, ERRORS_MASK);
1490 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1491 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1492 "options when quota turned on");
1493 return -1;
1494 }
1495
1496 if (m->flags & MOPT_NOSUPPORT) {
1497 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1498 } else if (token == Opt_commit) {
1499 if (arg == 0)
1500 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1501 sbi->s_commit_interval = HZ * arg;
1502 } else if (token == Opt_max_batch_time) {
1503 if (arg == 0)
1504 arg = EXT4_DEF_MAX_BATCH_TIME;
1505 sbi->s_max_batch_time = arg;
1506 } else if (token == Opt_min_batch_time) {
1507 sbi->s_min_batch_time = arg;
1508 } else if (token == Opt_inode_readahead_blks) {
1509 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1510 ext4_msg(sb, KERN_ERR,
1511 "EXT4-fs: inode_readahead_blks must be "
1512 "0 or a power of 2 smaller than 2^31");
1513 return -1;
1514 }
1515 sbi->s_inode_readahead_blks = arg;
1516 } else if (token == Opt_init_itable) {
1517 set_opt(sb, INIT_INODE_TABLE);
1518 if (!args->from)
1519 arg = EXT4_DEF_LI_WAIT_MULT;
1520 sbi->s_li_wait_mult = arg;
1521 } else if (token == Opt_max_dir_size_kb) {
1522 sbi->s_max_dir_size_kb = arg;
1523 } else if (token == Opt_stripe) {
1524 sbi->s_stripe = arg;
1525 } else if (token == Opt_resuid) {
1526 uid = make_kuid(current_user_ns(), arg);
1527 if (!uid_valid(uid)) {
1528 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1529 return -1;
1530 }
1531 sbi->s_resuid = uid;
1532 } else if (token == Opt_resgid) {
1533 gid = make_kgid(current_user_ns(), arg);
1534 if (!gid_valid(gid)) {
1535 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1536 return -1;
1537 }
1538 sbi->s_resgid = gid;
1539 } else if (token == Opt_journal_dev) {
1540 if (is_remount) {
1541 ext4_msg(sb, KERN_ERR,
1542 "Cannot specify journal on remount");
1543 return -1;
1544 }
1545 *journal_devnum = arg;
1546 } else if (token == Opt_journal_path) {
1547 char *journal_path;
1548 struct inode *journal_inode;
1549 struct path path;
1550 int error;
1551
1552 if (is_remount) {
1553 ext4_msg(sb, KERN_ERR,
1554 "Cannot specify journal on remount");
1555 return -1;
1556 }
1557 journal_path = match_strdup(&args[0]);
1558 if (!journal_path) {
1559 ext4_msg(sb, KERN_ERR, "error: could not dup "
1560 "journal device string");
1561 return -1;
1562 }
1563
1564 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1565 if (error) {
1566 ext4_msg(sb, KERN_ERR, "error: could not find "
1567 "journal device path: error %d", error);
1568 kfree(journal_path);
1569 return -1;
1570 }
1571
1572 journal_inode = path.dentry->d_inode;
1573 if (!S_ISBLK(journal_inode->i_mode)) {
1574 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1575 "is not a block device", journal_path);
1576 path_put(&path);
1577 kfree(journal_path);
1578 return -1;
1579 }
1580
1581 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1582 path_put(&path);
1583 kfree(journal_path);
1584 } else if (token == Opt_journal_ioprio) {
1585 if (arg > 7) {
1586 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1587 " (must be 0-7)");
1588 return -1;
1589 }
1590 *journal_ioprio =
1591 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1592 } else if (m->flags & MOPT_DATAJ) {
1593 if (is_remount) {
1594 if (!sbi->s_journal)
1595 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1596 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1597 ext4_msg(sb, KERN_ERR,
1598 "Cannot change data mode on remount");
1599 return -1;
1600 }
1601 } else {
1602 clear_opt(sb, DATA_FLAGS);
1603 sbi->s_mount_opt |= m->mount_opt;
1604 }
1605 #ifdef CONFIG_QUOTA
1606 } else if (m->flags & MOPT_QFMT) {
1607 if (sb_any_quota_loaded(sb) &&
1608 sbi->s_jquota_fmt != m->mount_opt) {
1609 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1610 "quota options when quota turned on");
1611 return -1;
1612 }
1613 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1614 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1615 ext4_msg(sb, KERN_ERR,
1616 "Cannot set journaled quota options "
1617 "when QUOTA feature is enabled");
1618 return -1;
1619 }
1620 sbi->s_jquota_fmt = m->mount_opt;
1621 #endif
1622 } else {
1623 if (!args->from)
1624 arg = 1;
1625 if (m->flags & MOPT_CLEAR)
1626 arg = !arg;
1627 else if (unlikely(!(m->flags & MOPT_SET))) {
1628 ext4_msg(sb, KERN_WARNING,
1629 "buggy handling of option %s", opt);
1630 WARN_ON(1);
1631 return -1;
1632 }
1633 if (arg != 0)
1634 sbi->s_mount_opt |= m->mount_opt;
1635 else
1636 sbi->s_mount_opt &= ~m->mount_opt;
1637 }
1638 return 1;
1639 }
1640
1641 static int parse_options(char *options, struct super_block *sb,
1642 unsigned long *journal_devnum,
1643 unsigned int *journal_ioprio,
1644 int is_remount)
1645 {
1646 struct ext4_sb_info *sbi = EXT4_SB(sb);
1647 char *p;
1648 substring_t args[MAX_OPT_ARGS];
1649 int token;
1650
1651 if (!options)
1652 return 1;
1653
1654 while ((p = strsep(&options, ",")) != NULL) {
1655 if (!*p)
1656 continue;
1657 /*
1658 * Initialize args struct so we know whether arg was
1659 * found; some options take optional arguments.
1660 */
1661 args[0].to = args[0].from = NULL;
1662 token = match_token(p, tokens, args);
1663 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1664 journal_ioprio, is_remount) < 0)
1665 return 0;
1666 }
1667 #ifdef CONFIG_QUOTA
1668 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1669 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1670 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1671 "feature is enabled");
1672 return 0;
1673 }
1674 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1675 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1676 clear_opt(sb, USRQUOTA);
1677
1678 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1679 clear_opt(sb, GRPQUOTA);
1680
1681 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1682 ext4_msg(sb, KERN_ERR, "old and new quota "
1683 "format mixing");
1684 return 0;
1685 }
1686
1687 if (!sbi->s_jquota_fmt) {
1688 ext4_msg(sb, KERN_ERR, "journaled quota format "
1689 "not specified");
1690 return 0;
1691 }
1692 } else {
1693 if (sbi->s_jquota_fmt) {
1694 ext4_msg(sb, KERN_ERR, "journaled quota format "
1695 "specified with no journaling "
1696 "enabled");
1697 return 0;
1698 }
1699 }
1700 #endif
1701 if (test_opt(sb, DIOREAD_NOLOCK)) {
1702 int blocksize =
1703 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1704
1705 if (blocksize < PAGE_CACHE_SIZE) {
1706 ext4_msg(sb, KERN_ERR, "can't mount with "
1707 "dioread_nolock if block size != PAGE_SIZE");
1708 return 0;
1709 }
1710 }
1711 return 1;
1712 }
1713
1714 static inline void ext4_show_quota_options(struct seq_file *seq,
1715 struct super_block *sb)
1716 {
1717 #if defined(CONFIG_QUOTA)
1718 struct ext4_sb_info *sbi = EXT4_SB(sb);
1719
1720 if (sbi->s_jquota_fmt) {
1721 char *fmtname = "";
1722
1723 switch (sbi->s_jquota_fmt) {
1724 case QFMT_VFS_OLD:
1725 fmtname = "vfsold";
1726 break;
1727 case QFMT_VFS_V0:
1728 fmtname = "vfsv0";
1729 break;
1730 case QFMT_VFS_V1:
1731 fmtname = "vfsv1";
1732 break;
1733 }
1734 seq_printf(seq, ",jqfmt=%s", fmtname);
1735 }
1736
1737 if (sbi->s_qf_names[USRQUOTA])
1738 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1739
1740 if (sbi->s_qf_names[GRPQUOTA])
1741 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1742 #endif
1743 }
1744
1745 static const char *token2str(int token)
1746 {
1747 const struct match_token *t;
1748
1749 for (t = tokens; t->token != Opt_err; t++)
1750 if (t->token == token && !strchr(t->pattern, '='))
1751 break;
1752 return t->pattern;
1753 }
1754
1755 /*
1756 * Show an option if
1757 * - it's set to a non-default value OR
1758 * - if the per-sb default is different from the global default
1759 */
1760 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1761 int nodefs)
1762 {
1763 struct ext4_sb_info *sbi = EXT4_SB(sb);
1764 struct ext4_super_block *es = sbi->s_es;
1765 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1766 const struct mount_opts *m;
1767 char sep = nodefs ? '\n' : ',';
1768
1769 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1770 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1771
1772 if (sbi->s_sb_block != 1)
1773 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1774
1775 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1776 int want_set = m->flags & MOPT_SET;
1777 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1778 (m->flags & MOPT_CLEAR_ERR))
1779 continue;
1780 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1781 continue; /* skip if same as the default */
1782 if ((want_set &&
1783 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1784 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1785 continue; /* select Opt_noFoo vs Opt_Foo */
1786 SEQ_OPTS_PRINT("%s", token2str(m->token));
1787 }
1788
1789 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1790 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1791 SEQ_OPTS_PRINT("resuid=%u",
1792 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1793 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1794 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1795 SEQ_OPTS_PRINT("resgid=%u",
1796 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1797 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1798 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1799 SEQ_OPTS_PUTS("errors=remount-ro");
1800 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1801 SEQ_OPTS_PUTS("errors=continue");
1802 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1803 SEQ_OPTS_PUTS("errors=panic");
1804 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1805 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1806 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1807 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1808 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1809 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1810 if (sb->s_flags & MS_I_VERSION)
1811 SEQ_OPTS_PUTS("i_version");
1812 if (nodefs || sbi->s_stripe)
1813 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1814 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1815 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1816 SEQ_OPTS_PUTS("data=journal");
1817 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1818 SEQ_OPTS_PUTS("data=ordered");
1819 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1820 SEQ_OPTS_PUTS("data=writeback");
1821 }
1822 if (nodefs ||
1823 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1824 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1825 sbi->s_inode_readahead_blks);
1826
1827 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1828 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1829 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1830 if (nodefs || sbi->s_max_dir_size_kb)
1831 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1832
1833 ext4_show_quota_options(seq, sb);
1834 return 0;
1835 }
1836
1837 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1838 {
1839 return _ext4_show_options(seq, root->d_sb, 0);
1840 }
1841
1842 static int options_seq_show(struct seq_file *seq, void *offset)
1843 {
1844 struct super_block *sb = seq->private;
1845 int rc;
1846
1847 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1848 rc = _ext4_show_options(seq, sb, 1);
1849 seq_puts(seq, "\n");
1850 return rc;
1851 }
1852
1853 static int options_open_fs(struct inode *inode, struct file *file)
1854 {
1855 return single_open(file, options_seq_show, PDE_DATA(inode));
1856 }
1857
1858 static const struct file_operations ext4_seq_options_fops = {
1859 .owner = THIS_MODULE,
1860 .open = options_open_fs,
1861 .read = seq_read,
1862 .llseek = seq_lseek,
1863 .release = single_release,
1864 };
1865
1866 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1867 int read_only)
1868 {
1869 struct ext4_sb_info *sbi = EXT4_SB(sb);
1870 int res = 0;
1871
1872 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1873 ext4_msg(sb, KERN_ERR, "revision level too high, "
1874 "forcing read-only mode");
1875 res = MS_RDONLY;
1876 }
1877 if (read_only)
1878 goto done;
1879 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1880 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1881 "running e2fsck is recommended");
1882 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1883 ext4_msg(sb, KERN_WARNING,
1884 "warning: mounting fs with errors, "
1885 "running e2fsck is recommended");
1886 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1887 le16_to_cpu(es->s_mnt_count) >=
1888 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1889 ext4_msg(sb, KERN_WARNING,
1890 "warning: maximal mount count reached, "
1891 "running e2fsck is recommended");
1892 else if (le32_to_cpu(es->s_checkinterval) &&
1893 (le32_to_cpu(es->s_lastcheck) +
1894 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1895 ext4_msg(sb, KERN_WARNING,
1896 "warning: checktime reached, "
1897 "running e2fsck is recommended");
1898 if (!sbi->s_journal)
1899 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1900 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1901 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1902 le16_add_cpu(&es->s_mnt_count, 1);
1903 es->s_mtime = cpu_to_le32(get_seconds());
1904 ext4_update_dynamic_rev(sb);
1905 if (sbi->s_journal)
1906 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1907
1908 ext4_commit_super(sb, 1);
1909 done:
1910 if (test_opt(sb, DEBUG))
1911 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1912 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1913 sb->s_blocksize,
1914 sbi->s_groups_count,
1915 EXT4_BLOCKS_PER_GROUP(sb),
1916 EXT4_INODES_PER_GROUP(sb),
1917 sbi->s_mount_opt, sbi->s_mount_opt2);
1918
1919 cleancache_init_fs(sb);
1920 return res;
1921 }
1922
1923 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1924 {
1925 struct ext4_sb_info *sbi = EXT4_SB(sb);
1926 struct flex_groups *new_groups;
1927 int size;
1928
1929 if (!sbi->s_log_groups_per_flex)
1930 return 0;
1931
1932 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1933 if (size <= sbi->s_flex_groups_allocated)
1934 return 0;
1935
1936 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1937 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1938 if (!new_groups) {
1939 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1940 size / (int) sizeof(struct flex_groups));
1941 return -ENOMEM;
1942 }
1943
1944 if (sbi->s_flex_groups) {
1945 memcpy(new_groups, sbi->s_flex_groups,
1946 (sbi->s_flex_groups_allocated *
1947 sizeof(struct flex_groups)));
1948 ext4_kvfree(sbi->s_flex_groups);
1949 }
1950 sbi->s_flex_groups = new_groups;
1951 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1952 return 0;
1953 }
1954
1955 static int ext4_fill_flex_info(struct super_block *sb)
1956 {
1957 struct ext4_sb_info *sbi = EXT4_SB(sb);
1958 struct ext4_group_desc *gdp = NULL;
1959 ext4_group_t flex_group;
1960 int i, err;
1961
1962 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1963 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1964 sbi->s_log_groups_per_flex = 0;
1965 return 1;
1966 }
1967
1968 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1969 if (err)
1970 goto failed;
1971
1972 for (i = 0; i < sbi->s_groups_count; i++) {
1973 gdp = ext4_get_group_desc(sb, i, NULL);
1974
1975 flex_group = ext4_flex_group(sbi, i);
1976 atomic_add(ext4_free_inodes_count(sb, gdp),
1977 &sbi->s_flex_groups[flex_group].free_inodes);
1978 atomic64_add(ext4_free_group_clusters(sb, gdp),
1979 &sbi->s_flex_groups[flex_group].free_clusters);
1980 atomic_add(ext4_used_dirs_count(sb, gdp),
1981 &sbi->s_flex_groups[flex_group].used_dirs);
1982 }
1983
1984 return 1;
1985 failed:
1986 return 0;
1987 }
1988
1989 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1990 struct ext4_group_desc *gdp)
1991 {
1992 int offset;
1993 __u16 crc = 0;
1994 __le32 le_group = cpu_to_le32(block_group);
1995
1996 if ((sbi->s_es->s_feature_ro_compat &
1997 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1998 /* Use new metadata_csum algorithm */
1999 __le16 save_csum;
2000 __u32 csum32;
2001
2002 save_csum = gdp->bg_checksum;
2003 gdp->bg_checksum = 0;
2004 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2005 sizeof(le_group));
2006 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2007 sbi->s_desc_size);
2008 gdp->bg_checksum = save_csum;
2009
2010 crc = csum32 & 0xFFFF;
2011 goto out;
2012 }
2013
2014 /* old crc16 code */
2015 offset = offsetof(struct ext4_group_desc, bg_checksum);
2016
2017 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2018 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2019 crc = crc16(crc, (__u8 *)gdp, offset);
2020 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2021 /* for checksum of struct ext4_group_desc do the rest...*/
2022 if ((sbi->s_es->s_feature_incompat &
2023 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2024 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2025 crc = crc16(crc, (__u8 *)gdp + offset,
2026 le16_to_cpu(sbi->s_es->s_desc_size) -
2027 offset);
2028
2029 out:
2030 return cpu_to_le16(crc);
2031 }
2032
2033 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2034 struct ext4_group_desc *gdp)
2035 {
2036 if (ext4_has_group_desc_csum(sb) &&
2037 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2038 block_group, gdp)))
2039 return 0;
2040
2041 return 1;
2042 }
2043
2044 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2045 struct ext4_group_desc *gdp)
2046 {
2047 if (!ext4_has_group_desc_csum(sb))
2048 return;
2049 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2050 }
2051
2052 /* Called at mount-time, super-block is locked */
2053 static int ext4_check_descriptors(struct super_block *sb,
2054 ext4_group_t *first_not_zeroed)
2055 {
2056 struct ext4_sb_info *sbi = EXT4_SB(sb);
2057 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2058 ext4_fsblk_t last_block;
2059 ext4_fsblk_t block_bitmap;
2060 ext4_fsblk_t inode_bitmap;
2061 ext4_fsblk_t inode_table;
2062 int flexbg_flag = 0;
2063 ext4_group_t i, grp = sbi->s_groups_count;
2064
2065 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2066 flexbg_flag = 1;
2067
2068 ext4_debug("Checking group descriptors");
2069
2070 for (i = 0; i < sbi->s_groups_count; i++) {
2071 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2072
2073 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2074 last_block = ext4_blocks_count(sbi->s_es) - 1;
2075 else
2076 last_block = first_block +
2077 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2078
2079 if ((grp == sbi->s_groups_count) &&
2080 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2081 grp = i;
2082
2083 block_bitmap = ext4_block_bitmap(sb, gdp);
2084 if (block_bitmap < first_block || block_bitmap > last_block) {
2085 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2086 "Block bitmap for group %u not in group "
2087 "(block %llu)!", i, block_bitmap);
2088 return 0;
2089 }
2090 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2091 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2092 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2093 "Inode bitmap for group %u not in group "
2094 "(block %llu)!", i, inode_bitmap);
2095 return 0;
2096 }
2097 inode_table = ext4_inode_table(sb, gdp);
2098 if (inode_table < first_block ||
2099 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2100 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2101 "Inode table for group %u not in group "
2102 "(block %llu)!", i, inode_table);
2103 return 0;
2104 }
2105 ext4_lock_group(sb, i);
2106 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2107 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2108 "Checksum for group %u failed (%u!=%u)",
2109 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2110 gdp)), le16_to_cpu(gdp->bg_checksum));
2111 if (!(sb->s_flags & MS_RDONLY)) {
2112 ext4_unlock_group(sb, i);
2113 return 0;
2114 }
2115 }
2116 ext4_unlock_group(sb, i);
2117 if (!flexbg_flag)
2118 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2119 }
2120 if (NULL != first_not_zeroed)
2121 *first_not_zeroed = grp;
2122
2123 ext4_free_blocks_count_set(sbi->s_es,
2124 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2125 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2126 return 1;
2127 }
2128
2129 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2130 * the superblock) which were deleted from all directories, but held open by
2131 * a process at the time of a crash. We walk the list and try to delete these
2132 * inodes at recovery time (only with a read-write filesystem).
2133 *
2134 * In order to keep the orphan inode chain consistent during traversal (in
2135 * case of crash during recovery), we link each inode into the superblock
2136 * orphan list_head and handle it the same way as an inode deletion during
2137 * normal operation (which journals the operations for us).
2138 *
2139 * We only do an iget() and an iput() on each inode, which is very safe if we
2140 * accidentally point at an in-use or already deleted inode. The worst that
2141 * can happen in this case is that we get a "bit already cleared" message from
2142 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2143 * e2fsck was run on this filesystem, and it must have already done the orphan
2144 * inode cleanup for us, so we can safely abort without any further action.
2145 */
2146 static void ext4_orphan_cleanup(struct super_block *sb,
2147 struct ext4_super_block *es)
2148 {
2149 unsigned int s_flags = sb->s_flags;
2150 int nr_orphans = 0, nr_truncates = 0;
2151 #ifdef CONFIG_QUOTA
2152 int i;
2153 #endif
2154 if (!es->s_last_orphan) {
2155 jbd_debug(4, "no orphan inodes to clean up\n");
2156 return;
2157 }
2158
2159 if (bdev_read_only(sb->s_bdev)) {
2160 ext4_msg(sb, KERN_ERR, "write access "
2161 "unavailable, skipping orphan cleanup");
2162 return;
2163 }
2164
2165 /* Check if feature set would not allow a r/w mount */
2166 if (!ext4_feature_set_ok(sb, 0)) {
2167 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2168 "unknown ROCOMPAT features");
2169 return;
2170 }
2171
2172 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2173 /* don't clear list on RO mount w/ errors */
2174 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2175 jbd_debug(1, "Errors on filesystem, "
2176 "clearing orphan list.\n");
2177 es->s_last_orphan = 0;
2178 }
2179 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2180 return;
2181 }
2182
2183 if (s_flags & MS_RDONLY) {
2184 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2185 sb->s_flags &= ~MS_RDONLY;
2186 }
2187 #ifdef CONFIG_QUOTA
2188 /* Needed for iput() to work correctly and not trash data */
2189 sb->s_flags |= MS_ACTIVE;
2190 /* Turn on quotas so that they are updated correctly */
2191 for (i = 0; i < MAXQUOTAS; i++) {
2192 if (EXT4_SB(sb)->s_qf_names[i]) {
2193 int ret = ext4_quota_on_mount(sb, i);
2194 if (ret < 0)
2195 ext4_msg(sb, KERN_ERR,
2196 "Cannot turn on journaled "
2197 "quota: error %d", ret);
2198 }
2199 }
2200 #endif
2201
2202 while (es->s_last_orphan) {
2203 struct inode *inode;
2204
2205 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2206 if (IS_ERR(inode)) {
2207 es->s_last_orphan = 0;
2208 break;
2209 }
2210
2211 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2212 dquot_initialize(inode);
2213 if (inode->i_nlink) {
2214 if (test_opt(sb, DEBUG))
2215 ext4_msg(sb, KERN_DEBUG,
2216 "%s: truncating inode %lu to %lld bytes",
2217 __func__, inode->i_ino, inode->i_size);
2218 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2219 inode->i_ino, inode->i_size);
2220 mutex_lock(&inode->i_mutex);
2221 truncate_inode_pages(inode->i_mapping, inode->i_size);
2222 ext4_truncate(inode);
2223 mutex_unlock(&inode->i_mutex);
2224 nr_truncates++;
2225 } else {
2226 if (test_opt(sb, DEBUG))
2227 ext4_msg(sb, KERN_DEBUG,
2228 "%s: deleting unreferenced inode %lu",
2229 __func__, inode->i_ino);
2230 jbd_debug(2, "deleting unreferenced inode %lu\n",
2231 inode->i_ino);
2232 nr_orphans++;
2233 }
2234 iput(inode); /* The delete magic happens here! */
2235 }
2236
2237 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2238
2239 if (nr_orphans)
2240 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2241 PLURAL(nr_orphans));
2242 if (nr_truncates)
2243 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2244 PLURAL(nr_truncates));
2245 #ifdef CONFIG_QUOTA
2246 /* Turn quotas off */
2247 for (i = 0; i < MAXQUOTAS; i++) {
2248 if (sb_dqopt(sb)->files[i])
2249 dquot_quota_off(sb, i);
2250 }
2251 #endif
2252 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2253 }
2254
2255 /*
2256 * Maximal extent format file size.
2257 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2258 * extent format containers, within a sector_t, and within i_blocks
2259 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2260 * so that won't be a limiting factor.
2261 *
2262 * However there is other limiting factor. We do store extents in the form
2263 * of starting block and length, hence the resulting length of the extent
2264 * covering maximum file size must fit into on-disk format containers as
2265 * well. Given that length is always by 1 unit bigger than max unit (because
2266 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2267 *
2268 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2269 */
2270 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2271 {
2272 loff_t res;
2273 loff_t upper_limit = MAX_LFS_FILESIZE;
2274
2275 /* small i_blocks in vfs inode? */
2276 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2277 /*
2278 * CONFIG_LBDAF is not enabled implies the inode
2279 * i_block represent total blocks in 512 bytes
2280 * 32 == size of vfs inode i_blocks * 8
2281 */
2282 upper_limit = (1LL << 32) - 1;
2283
2284 /* total blocks in file system block size */
2285 upper_limit >>= (blkbits - 9);
2286 upper_limit <<= blkbits;
2287 }
2288
2289 /*
2290 * 32-bit extent-start container, ee_block. We lower the maxbytes
2291 * by one fs block, so ee_len can cover the extent of maximum file
2292 * size
2293 */
2294 res = (1LL << 32) - 1;
2295 res <<= blkbits;
2296
2297 /* Sanity check against vm- & vfs- imposed limits */
2298 if (res > upper_limit)
2299 res = upper_limit;
2300
2301 return res;
2302 }
2303
2304 /*
2305 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2306 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2307 * We need to be 1 filesystem block less than the 2^48 sector limit.
2308 */
2309 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2310 {
2311 loff_t res = EXT4_NDIR_BLOCKS;
2312 int meta_blocks;
2313 loff_t upper_limit;
2314 /* This is calculated to be the largest file size for a dense, block
2315 * mapped file such that the file's total number of 512-byte sectors,
2316 * including data and all indirect blocks, does not exceed (2^48 - 1).
2317 *
2318 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2319 * number of 512-byte sectors of the file.
2320 */
2321
2322 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2323 /*
2324 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2325 * the inode i_block field represents total file blocks in
2326 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2327 */
2328 upper_limit = (1LL << 32) - 1;
2329
2330 /* total blocks in file system block size */
2331 upper_limit >>= (bits - 9);
2332
2333 } else {
2334 /*
2335 * We use 48 bit ext4_inode i_blocks
2336 * With EXT4_HUGE_FILE_FL set the i_blocks
2337 * represent total number of blocks in
2338 * file system block size
2339 */
2340 upper_limit = (1LL << 48) - 1;
2341
2342 }
2343
2344 /* indirect blocks */
2345 meta_blocks = 1;
2346 /* double indirect blocks */
2347 meta_blocks += 1 + (1LL << (bits-2));
2348 /* tripple indirect blocks */
2349 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2350
2351 upper_limit -= meta_blocks;
2352 upper_limit <<= bits;
2353
2354 res += 1LL << (bits-2);
2355 res += 1LL << (2*(bits-2));
2356 res += 1LL << (3*(bits-2));
2357 res <<= bits;
2358 if (res > upper_limit)
2359 res = upper_limit;
2360
2361 if (res > MAX_LFS_FILESIZE)
2362 res = MAX_LFS_FILESIZE;
2363
2364 return res;
2365 }
2366
2367 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2368 ext4_fsblk_t logical_sb_block, int nr)
2369 {
2370 struct ext4_sb_info *sbi = EXT4_SB(sb);
2371 ext4_group_t bg, first_meta_bg;
2372 int has_super = 0;
2373
2374 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2375
2376 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2377 nr < first_meta_bg)
2378 return logical_sb_block + nr + 1;
2379 bg = sbi->s_desc_per_block * nr;
2380 if (ext4_bg_has_super(sb, bg))
2381 has_super = 1;
2382
2383 return (has_super + ext4_group_first_block_no(sb, bg));
2384 }
2385
2386 /**
2387 * ext4_get_stripe_size: Get the stripe size.
2388 * @sbi: In memory super block info
2389 *
2390 * If we have specified it via mount option, then
2391 * use the mount option value. If the value specified at mount time is
2392 * greater than the blocks per group use the super block value.
2393 * If the super block value is greater than blocks per group return 0.
2394 * Allocator needs it be less than blocks per group.
2395 *
2396 */
2397 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2398 {
2399 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2400 unsigned long stripe_width =
2401 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2402 int ret;
2403
2404 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2405 ret = sbi->s_stripe;
2406 else if (stripe_width <= sbi->s_blocks_per_group)
2407 ret = stripe_width;
2408 else if (stride <= sbi->s_blocks_per_group)
2409 ret = stride;
2410 else
2411 ret = 0;
2412
2413 /*
2414 * If the stripe width is 1, this makes no sense and
2415 * we set it to 0 to turn off stripe handling code.
2416 */
2417 if (ret <= 1)
2418 ret = 0;
2419
2420 return ret;
2421 }
2422
2423 /* sysfs supprt */
2424
2425 struct ext4_attr {
2426 struct attribute attr;
2427 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2428 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2429 const char *, size_t);
2430 union {
2431 int offset;
2432 int deprecated_val;
2433 } u;
2434 };
2435
2436 static int parse_strtoull(const char *buf,
2437 unsigned long long max, unsigned long long *value)
2438 {
2439 int ret;
2440
2441 ret = kstrtoull(skip_spaces(buf), 0, value);
2442 if (!ret && *value > max)
2443 ret = -EINVAL;
2444 return ret;
2445 }
2446
2447 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2448 struct ext4_sb_info *sbi,
2449 char *buf)
2450 {
2451 return snprintf(buf, PAGE_SIZE, "%llu\n",
2452 (s64) EXT4_C2B(sbi,
2453 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2454 }
2455
2456 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2457 struct ext4_sb_info *sbi, char *buf)
2458 {
2459 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2460
2461 if (!sb->s_bdev->bd_part)
2462 return snprintf(buf, PAGE_SIZE, "0\n");
2463 return snprintf(buf, PAGE_SIZE, "%lu\n",
2464 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2465 sbi->s_sectors_written_start) >> 1);
2466 }
2467
2468 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2469 struct ext4_sb_info *sbi, char *buf)
2470 {
2471 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2472
2473 if (!sb->s_bdev->bd_part)
2474 return snprintf(buf, PAGE_SIZE, "0\n");
2475 return snprintf(buf, PAGE_SIZE, "%llu\n",
2476 (unsigned long long)(sbi->s_kbytes_written +
2477 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2478 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2479 }
2480
2481 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2482 struct ext4_sb_info *sbi,
2483 const char *buf, size_t count)
2484 {
2485 unsigned long t;
2486 int ret;
2487
2488 ret = kstrtoul(skip_spaces(buf), 0, &t);
2489 if (ret)
2490 return ret;
2491
2492 if (t && (!is_power_of_2(t) || t > 0x40000000))
2493 return -EINVAL;
2494
2495 sbi->s_inode_readahead_blks = t;
2496 return count;
2497 }
2498
2499 static ssize_t sbi_ui_show(struct ext4_attr *a,
2500 struct ext4_sb_info *sbi, char *buf)
2501 {
2502 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2503
2504 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2505 }
2506
2507 static ssize_t sbi_ui_store(struct ext4_attr *a,
2508 struct ext4_sb_info *sbi,
2509 const char *buf, size_t count)
2510 {
2511 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2512 unsigned long t;
2513 int ret;
2514
2515 ret = kstrtoul(skip_spaces(buf), 0, &t);
2516 if (ret)
2517 return ret;
2518 *ui = t;
2519 return count;
2520 }
2521
2522 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2523 struct ext4_sb_info *sbi, char *buf)
2524 {
2525 return snprintf(buf, PAGE_SIZE, "%llu\n",
2526 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2527 }
2528
2529 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2530 struct ext4_sb_info *sbi,
2531 const char *buf, size_t count)
2532 {
2533 unsigned long long val;
2534 int ret;
2535
2536 if (parse_strtoull(buf, -1ULL, &val))
2537 return -EINVAL;
2538 ret = ext4_reserve_clusters(sbi, val);
2539
2540 return ret ? ret : count;
2541 }
2542
2543 static ssize_t trigger_test_error(struct ext4_attr *a,
2544 struct ext4_sb_info *sbi,
2545 const char *buf, size_t count)
2546 {
2547 int len = count;
2548
2549 if (!capable(CAP_SYS_ADMIN))
2550 return -EPERM;
2551
2552 if (len && buf[len-1] == '\n')
2553 len--;
2554
2555 if (len)
2556 ext4_error(sbi->s_sb, "%.*s", len, buf);
2557 return count;
2558 }
2559
2560 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2561 struct ext4_sb_info *sbi, char *buf)
2562 {
2563 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2564 }
2565
2566 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2567 static struct ext4_attr ext4_attr_##_name = { \
2568 .attr = {.name = __stringify(_name), .mode = _mode }, \
2569 .show = _show, \
2570 .store = _store, \
2571 .u = { \
2572 .offset = offsetof(struct ext4_sb_info, _elname),\
2573 }, \
2574 }
2575 #define EXT4_ATTR(name, mode, show, store) \
2576 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2577
2578 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2579 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2580 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2581 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2582 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2583 #define ATTR_LIST(name) &ext4_attr_##name.attr
2584 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2585 static struct ext4_attr ext4_attr_##_name = { \
2586 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2587 .show = sbi_deprecated_show, \
2588 .u = { \
2589 .deprecated_val = _val, \
2590 }, \
2591 }
2592
2593 EXT4_RO_ATTR(delayed_allocation_blocks);
2594 EXT4_RO_ATTR(session_write_kbytes);
2595 EXT4_RO_ATTR(lifetime_write_kbytes);
2596 EXT4_RW_ATTR(reserved_clusters);
2597 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2598 inode_readahead_blks_store, s_inode_readahead_blks);
2599 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2600 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2601 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2602 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2603 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2604 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2605 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2606 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2607 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2608 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2609
2610 static struct attribute *ext4_attrs[] = {
2611 ATTR_LIST(delayed_allocation_blocks),
2612 ATTR_LIST(session_write_kbytes),
2613 ATTR_LIST(lifetime_write_kbytes),
2614 ATTR_LIST(reserved_clusters),
2615 ATTR_LIST(inode_readahead_blks),
2616 ATTR_LIST(inode_goal),
2617 ATTR_LIST(mb_stats),
2618 ATTR_LIST(mb_max_to_scan),
2619 ATTR_LIST(mb_min_to_scan),
2620 ATTR_LIST(mb_order2_req),
2621 ATTR_LIST(mb_stream_req),
2622 ATTR_LIST(mb_group_prealloc),
2623 ATTR_LIST(max_writeback_mb_bump),
2624 ATTR_LIST(extent_max_zeroout_kb),
2625 ATTR_LIST(trigger_fs_error),
2626 NULL,
2627 };
2628
2629 /* Features this copy of ext4 supports */
2630 EXT4_INFO_ATTR(lazy_itable_init);
2631 EXT4_INFO_ATTR(batched_discard);
2632 EXT4_INFO_ATTR(meta_bg_resize);
2633
2634 static struct attribute *ext4_feat_attrs[] = {
2635 ATTR_LIST(lazy_itable_init),
2636 ATTR_LIST(batched_discard),
2637 ATTR_LIST(meta_bg_resize),
2638 NULL,
2639 };
2640
2641 static ssize_t ext4_attr_show(struct kobject *kobj,
2642 struct attribute *attr, char *buf)
2643 {
2644 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2645 s_kobj);
2646 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2647
2648 return a->show ? a->show(a, sbi, buf) : 0;
2649 }
2650
2651 static ssize_t ext4_attr_store(struct kobject *kobj,
2652 struct attribute *attr,
2653 const char *buf, size_t len)
2654 {
2655 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2656 s_kobj);
2657 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2658
2659 return a->store ? a->store(a, sbi, buf, len) : 0;
2660 }
2661
2662 static void ext4_sb_release(struct kobject *kobj)
2663 {
2664 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2665 s_kobj);
2666 complete(&sbi->s_kobj_unregister);
2667 }
2668
2669 static const struct sysfs_ops ext4_attr_ops = {
2670 .show = ext4_attr_show,
2671 .store = ext4_attr_store,
2672 };
2673
2674 static struct kobj_type ext4_ktype = {
2675 .default_attrs = ext4_attrs,
2676 .sysfs_ops = &ext4_attr_ops,
2677 .release = ext4_sb_release,
2678 };
2679
2680 static void ext4_feat_release(struct kobject *kobj)
2681 {
2682 complete(&ext4_feat->f_kobj_unregister);
2683 }
2684
2685 static struct kobj_type ext4_feat_ktype = {
2686 .default_attrs = ext4_feat_attrs,
2687 .sysfs_ops = &ext4_attr_ops,
2688 .release = ext4_feat_release,
2689 };
2690
2691 /*
2692 * Check whether this filesystem can be mounted based on
2693 * the features present and the RDONLY/RDWR mount requested.
2694 * Returns 1 if this filesystem can be mounted as requested,
2695 * 0 if it cannot be.
2696 */
2697 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2698 {
2699 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2700 ext4_msg(sb, KERN_ERR,
2701 "Couldn't mount because of "
2702 "unsupported optional features (%x)",
2703 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2704 ~EXT4_FEATURE_INCOMPAT_SUPP));
2705 return 0;
2706 }
2707
2708 if (readonly)
2709 return 1;
2710
2711 /* Check that feature set is OK for a read-write mount */
2712 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2713 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2714 "unsupported optional features (%x)",
2715 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2716 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2717 return 0;
2718 }
2719 /*
2720 * Large file size enabled file system can only be mounted
2721 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2722 */
2723 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2724 if (sizeof(blkcnt_t) < sizeof(u64)) {
2725 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2726 "cannot be mounted RDWR without "
2727 "CONFIG_LBDAF");
2728 return 0;
2729 }
2730 }
2731 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2732 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2733 ext4_msg(sb, KERN_ERR,
2734 "Can't support bigalloc feature without "
2735 "extents feature\n");
2736 return 0;
2737 }
2738
2739 #ifndef CONFIG_QUOTA
2740 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2741 !readonly) {
2742 ext4_msg(sb, KERN_ERR,
2743 "Filesystem with quota feature cannot be mounted RDWR "
2744 "without CONFIG_QUOTA");
2745 return 0;
2746 }
2747 #endif /* CONFIG_QUOTA */
2748 return 1;
2749 }
2750
2751 /*
2752 * This function is called once a day if we have errors logged
2753 * on the file system
2754 */
2755 static void print_daily_error_info(unsigned long arg)
2756 {
2757 struct super_block *sb = (struct super_block *) arg;
2758 struct ext4_sb_info *sbi;
2759 struct ext4_super_block *es;
2760
2761 sbi = EXT4_SB(sb);
2762 es = sbi->s_es;
2763
2764 if (es->s_error_count)
2765 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2766 le32_to_cpu(es->s_error_count));
2767 if (es->s_first_error_time) {
2768 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2769 sb->s_id, le32_to_cpu(es->s_first_error_time),
2770 (int) sizeof(es->s_first_error_func),
2771 es->s_first_error_func,
2772 le32_to_cpu(es->s_first_error_line));
2773 if (es->s_first_error_ino)
2774 printk(": inode %u",
2775 le32_to_cpu(es->s_first_error_ino));
2776 if (es->s_first_error_block)
2777 printk(": block %llu", (unsigned long long)
2778 le64_to_cpu(es->s_first_error_block));
2779 printk("\n");
2780 }
2781 if (es->s_last_error_time) {
2782 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2783 sb->s_id, le32_to_cpu(es->s_last_error_time),
2784 (int) sizeof(es->s_last_error_func),
2785 es->s_last_error_func,
2786 le32_to_cpu(es->s_last_error_line));
2787 if (es->s_last_error_ino)
2788 printk(": inode %u",
2789 le32_to_cpu(es->s_last_error_ino));
2790 if (es->s_last_error_block)
2791 printk(": block %llu", (unsigned long long)
2792 le64_to_cpu(es->s_last_error_block));
2793 printk("\n");
2794 }
2795 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2796 }
2797
2798 /* Find next suitable group and run ext4_init_inode_table */
2799 static int ext4_run_li_request(struct ext4_li_request *elr)
2800 {
2801 struct ext4_group_desc *gdp = NULL;
2802 ext4_group_t group, ngroups;
2803 struct super_block *sb;
2804 unsigned long timeout = 0;
2805 int ret = 0;
2806
2807 sb = elr->lr_super;
2808 ngroups = EXT4_SB(sb)->s_groups_count;
2809
2810 sb_start_write(sb);
2811 for (group = elr->lr_next_group; group < ngroups; group++) {
2812 gdp = ext4_get_group_desc(sb, group, NULL);
2813 if (!gdp) {
2814 ret = 1;
2815 break;
2816 }
2817
2818 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2819 break;
2820 }
2821
2822 if (group >= ngroups)
2823 ret = 1;
2824
2825 if (!ret) {
2826 timeout = jiffies;
2827 ret = ext4_init_inode_table(sb, group,
2828 elr->lr_timeout ? 0 : 1);
2829 if (elr->lr_timeout == 0) {
2830 timeout = (jiffies - timeout) *
2831 elr->lr_sbi->s_li_wait_mult;
2832 elr->lr_timeout = timeout;
2833 }
2834 elr->lr_next_sched = jiffies + elr->lr_timeout;
2835 elr->lr_next_group = group + 1;
2836 }
2837 sb_end_write(sb);
2838
2839 return ret;
2840 }
2841
2842 /*
2843 * Remove lr_request from the list_request and free the
2844 * request structure. Should be called with li_list_mtx held
2845 */
2846 static void ext4_remove_li_request(struct ext4_li_request *elr)
2847 {
2848 struct ext4_sb_info *sbi;
2849
2850 if (!elr)
2851 return;
2852
2853 sbi = elr->lr_sbi;
2854
2855 list_del(&elr->lr_request);
2856 sbi->s_li_request = NULL;
2857 kfree(elr);
2858 }
2859
2860 static void ext4_unregister_li_request(struct super_block *sb)
2861 {
2862 mutex_lock(&ext4_li_mtx);
2863 if (!ext4_li_info) {
2864 mutex_unlock(&ext4_li_mtx);
2865 return;
2866 }
2867
2868 mutex_lock(&ext4_li_info->li_list_mtx);
2869 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2870 mutex_unlock(&ext4_li_info->li_list_mtx);
2871 mutex_unlock(&ext4_li_mtx);
2872 }
2873
2874 static struct task_struct *ext4_lazyinit_task;
2875
2876 /*
2877 * This is the function where ext4lazyinit thread lives. It walks
2878 * through the request list searching for next scheduled filesystem.
2879 * When such a fs is found, run the lazy initialization request
2880 * (ext4_rn_li_request) and keep track of the time spend in this
2881 * function. Based on that time we compute next schedule time of
2882 * the request. When walking through the list is complete, compute
2883 * next waking time and put itself into sleep.
2884 */
2885 static int ext4_lazyinit_thread(void *arg)
2886 {
2887 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2888 struct list_head *pos, *n;
2889 struct ext4_li_request *elr;
2890 unsigned long next_wakeup, cur;
2891
2892 BUG_ON(NULL == eli);
2893
2894 cont_thread:
2895 while (true) {
2896 next_wakeup = MAX_JIFFY_OFFSET;
2897
2898 mutex_lock(&eli->li_list_mtx);
2899 if (list_empty(&eli->li_request_list)) {
2900 mutex_unlock(&eli->li_list_mtx);
2901 goto exit_thread;
2902 }
2903
2904 list_for_each_safe(pos, n, &eli->li_request_list) {
2905 elr = list_entry(pos, struct ext4_li_request,
2906 lr_request);
2907
2908 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2909 if (ext4_run_li_request(elr) != 0) {
2910 /* error, remove the lazy_init job */
2911 ext4_remove_li_request(elr);
2912 continue;
2913 }
2914 }
2915
2916 if (time_before(elr->lr_next_sched, next_wakeup))
2917 next_wakeup = elr->lr_next_sched;
2918 }
2919 mutex_unlock(&eli->li_list_mtx);
2920
2921 try_to_freeze();
2922
2923 cur = jiffies;
2924 if ((time_after_eq(cur, next_wakeup)) ||
2925 (MAX_JIFFY_OFFSET == next_wakeup)) {
2926 cond_resched();
2927 continue;
2928 }
2929
2930 schedule_timeout_interruptible(next_wakeup - cur);
2931
2932 if (kthread_should_stop()) {
2933 ext4_clear_request_list();
2934 goto exit_thread;
2935 }
2936 }
2937
2938 exit_thread:
2939 /*
2940 * It looks like the request list is empty, but we need
2941 * to check it under the li_list_mtx lock, to prevent any
2942 * additions into it, and of course we should lock ext4_li_mtx
2943 * to atomically free the list and ext4_li_info, because at
2944 * this point another ext4 filesystem could be registering
2945 * new one.
2946 */
2947 mutex_lock(&ext4_li_mtx);
2948 mutex_lock(&eli->li_list_mtx);
2949 if (!list_empty(&eli->li_request_list)) {
2950 mutex_unlock(&eli->li_list_mtx);
2951 mutex_unlock(&ext4_li_mtx);
2952 goto cont_thread;
2953 }
2954 mutex_unlock(&eli->li_list_mtx);
2955 kfree(ext4_li_info);
2956 ext4_li_info = NULL;
2957 mutex_unlock(&ext4_li_mtx);
2958
2959 return 0;
2960 }
2961
2962 static void ext4_clear_request_list(void)
2963 {
2964 struct list_head *pos, *n;
2965 struct ext4_li_request *elr;
2966
2967 mutex_lock(&ext4_li_info->li_list_mtx);
2968 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2969 elr = list_entry(pos, struct ext4_li_request,
2970 lr_request);
2971 ext4_remove_li_request(elr);
2972 }
2973 mutex_unlock(&ext4_li_info->li_list_mtx);
2974 }
2975
2976 static int ext4_run_lazyinit_thread(void)
2977 {
2978 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2979 ext4_li_info, "ext4lazyinit");
2980 if (IS_ERR(ext4_lazyinit_task)) {
2981 int err = PTR_ERR(ext4_lazyinit_task);
2982 ext4_clear_request_list();
2983 kfree(ext4_li_info);
2984 ext4_li_info = NULL;
2985 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2986 "initialization thread\n",
2987 err);
2988 return err;
2989 }
2990 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2991 return 0;
2992 }
2993
2994 /*
2995 * Check whether it make sense to run itable init. thread or not.
2996 * If there is at least one uninitialized inode table, return
2997 * corresponding group number, else the loop goes through all
2998 * groups and return total number of groups.
2999 */
3000 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3001 {
3002 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3003 struct ext4_group_desc *gdp = NULL;
3004
3005 for (group = 0; group < ngroups; group++) {
3006 gdp = ext4_get_group_desc(sb, group, NULL);
3007 if (!gdp)
3008 continue;
3009
3010 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3011 break;
3012 }
3013
3014 return group;
3015 }
3016
3017 static int ext4_li_info_new(void)
3018 {
3019 struct ext4_lazy_init *eli = NULL;
3020
3021 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3022 if (!eli)
3023 return -ENOMEM;
3024
3025 INIT_LIST_HEAD(&eli->li_request_list);
3026 mutex_init(&eli->li_list_mtx);
3027
3028 eli->li_state |= EXT4_LAZYINIT_QUIT;
3029
3030 ext4_li_info = eli;
3031
3032 return 0;
3033 }
3034
3035 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3036 ext4_group_t start)
3037 {
3038 struct ext4_sb_info *sbi = EXT4_SB(sb);
3039 struct ext4_li_request *elr;
3040 unsigned long rnd;
3041
3042 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3043 if (!elr)
3044 return NULL;
3045
3046 elr->lr_super = sb;
3047 elr->lr_sbi = sbi;
3048 elr->lr_next_group = start;
3049
3050 /*
3051 * Randomize first schedule time of the request to
3052 * spread the inode table initialization requests
3053 * better.
3054 */
3055 get_random_bytes(&rnd, sizeof(rnd));
3056 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3057 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3058
3059 return elr;
3060 }
3061
3062 int ext4_register_li_request(struct super_block *sb,
3063 ext4_group_t first_not_zeroed)
3064 {
3065 struct ext4_sb_info *sbi = EXT4_SB(sb);
3066 struct ext4_li_request *elr = NULL;
3067 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3068 int ret = 0;
3069
3070 mutex_lock(&ext4_li_mtx);
3071 if (sbi->s_li_request != NULL) {
3072 /*
3073 * Reset timeout so it can be computed again, because
3074 * s_li_wait_mult might have changed.
3075 */
3076 sbi->s_li_request->lr_timeout = 0;
3077 goto out;
3078 }
3079
3080 if (first_not_zeroed == ngroups ||
3081 (sb->s_flags & MS_RDONLY) ||
3082 !test_opt(sb, INIT_INODE_TABLE))
3083 goto out;
3084
3085 elr = ext4_li_request_new(sb, first_not_zeroed);
3086 if (!elr) {
3087 ret = -ENOMEM;
3088 goto out;
3089 }
3090
3091 if (NULL == ext4_li_info) {
3092 ret = ext4_li_info_new();
3093 if (ret)
3094 goto out;
3095 }
3096
3097 mutex_lock(&ext4_li_info->li_list_mtx);
3098 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3099 mutex_unlock(&ext4_li_info->li_list_mtx);
3100
3101 sbi->s_li_request = elr;
3102 /*
3103 * set elr to NULL here since it has been inserted to
3104 * the request_list and the removal and free of it is
3105 * handled by ext4_clear_request_list from now on.
3106 */
3107 elr = NULL;
3108
3109 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3110 ret = ext4_run_lazyinit_thread();
3111 if (ret)
3112 goto out;
3113 }
3114 out:
3115 mutex_unlock(&ext4_li_mtx);
3116 if (ret)
3117 kfree(elr);
3118 return ret;
3119 }
3120
3121 /*
3122 * We do not need to lock anything since this is called on
3123 * module unload.
3124 */
3125 static void ext4_destroy_lazyinit_thread(void)
3126 {
3127 /*
3128 * If thread exited earlier
3129 * there's nothing to be done.
3130 */
3131 if (!ext4_li_info || !ext4_lazyinit_task)
3132 return;
3133
3134 kthread_stop(ext4_lazyinit_task);
3135 }
3136
3137 static int set_journal_csum_feature_set(struct super_block *sb)
3138 {
3139 int ret = 1;
3140 int compat, incompat;
3141 struct ext4_sb_info *sbi = EXT4_SB(sb);
3142
3143 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3144 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3145 /* journal checksum v2 */
3146 compat = 0;
3147 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3148 } else {
3149 /* journal checksum v1 */
3150 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3151 incompat = 0;
3152 }
3153
3154 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3155 ret = jbd2_journal_set_features(sbi->s_journal,
3156 compat, 0,
3157 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3158 incompat);
3159 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3160 ret = jbd2_journal_set_features(sbi->s_journal,
3161 compat, 0,
3162 incompat);
3163 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3164 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3165 } else {
3166 jbd2_journal_clear_features(sbi->s_journal,
3167 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3168 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3169 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3170 }
3171
3172 return ret;
3173 }
3174
3175 /*
3176 * Note: calculating the overhead so we can be compatible with
3177 * historical BSD practice is quite difficult in the face of
3178 * clusters/bigalloc. This is because multiple metadata blocks from
3179 * different block group can end up in the same allocation cluster.
3180 * Calculating the exact overhead in the face of clustered allocation
3181 * requires either O(all block bitmaps) in memory or O(number of block
3182 * groups**2) in time. We will still calculate the superblock for
3183 * older file systems --- and if we come across with a bigalloc file
3184 * system with zero in s_overhead_clusters the estimate will be close to
3185 * correct especially for very large cluster sizes --- but for newer
3186 * file systems, it's better to calculate this figure once at mkfs
3187 * time, and store it in the superblock. If the superblock value is
3188 * present (even for non-bigalloc file systems), we will use it.
3189 */
3190 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3191 char *buf)
3192 {
3193 struct ext4_sb_info *sbi = EXT4_SB(sb);
3194 struct ext4_group_desc *gdp;
3195 ext4_fsblk_t first_block, last_block, b;
3196 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3197 int s, j, count = 0;
3198
3199 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3200 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3201 sbi->s_itb_per_group + 2);
3202
3203 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3204 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3205 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3206 for (i = 0; i < ngroups; i++) {
3207 gdp = ext4_get_group_desc(sb, i, NULL);
3208 b = ext4_block_bitmap(sb, gdp);
3209 if (b >= first_block && b <= last_block) {
3210 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3211 count++;
3212 }
3213 b = ext4_inode_bitmap(sb, gdp);
3214 if (b >= first_block && b <= last_block) {
3215 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3216 count++;
3217 }
3218 b = ext4_inode_table(sb, gdp);
3219 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3220 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3221 int c = EXT4_B2C(sbi, b - first_block);
3222 ext4_set_bit(c, buf);
3223 count++;
3224 }
3225 if (i != grp)
3226 continue;
3227 s = 0;
3228 if (ext4_bg_has_super(sb, grp)) {
3229 ext4_set_bit(s++, buf);
3230 count++;
3231 }
3232 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3233 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3234 count++;
3235 }
3236 }
3237 if (!count)
3238 return 0;
3239 return EXT4_CLUSTERS_PER_GROUP(sb) -
3240 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3241 }
3242
3243 /*
3244 * Compute the overhead and stash it in sbi->s_overhead
3245 */
3246 int ext4_calculate_overhead(struct super_block *sb)
3247 {
3248 struct ext4_sb_info *sbi = EXT4_SB(sb);
3249 struct ext4_super_block *es = sbi->s_es;
3250 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3251 ext4_fsblk_t overhead = 0;
3252 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3253
3254 if (!buf)
3255 return -ENOMEM;
3256
3257 /*
3258 * Compute the overhead (FS structures). This is constant
3259 * for a given filesystem unless the number of block groups
3260 * changes so we cache the previous value until it does.
3261 */
3262
3263 /*
3264 * All of the blocks before first_data_block are overhead
3265 */
3266 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3267
3268 /*
3269 * Add the overhead found in each block group
3270 */
3271 for (i = 0; i < ngroups; i++) {
3272 int blks;
3273
3274 blks = count_overhead(sb, i, buf);
3275 overhead += blks;
3276 if (blks)
3277 memset(buf, 0, PAGE_SIZE);
3278 cond_resched();
3279 }
3280 /* Add the journal blocks as well */
3281 if (sbi->s_journal)
3282 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3283
3284 sbi->s_overhead = overhead;
3285 smp_wmb();
3286 free_page((unsigned long) buf);
3287 return 0;
3288 }
3289
3290
3291 static ext4_fsblk_t ext4_calculate_resv_clusters(struct ext4_sb_info *sbi)
3292 {
3293 ext4_fsblk_t resv_clusters;
3294
3295 /*
3296 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3297 * This should cover the situations where we can not afford to run
3298 * out of space like for example punch hole, or converting
3299 * uninitialized extents in delalloc path. In most cases such
3300 * allocation would require 1, or 2 blocks, higher numbers are
3301 * very rare.
3302 */
3303 resv_clusters = ext4_blocks_count(sbi->s_es) >> sbi->s_cluster_bits;
3304
3305 do_div(resv_clusters, 50);
3306 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3307
3308 return resv_clusters;
3309 }
3310
3311
3312 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3313 {
3314 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3315 sbi->s_cluster_bits;
3316
3317 if (count >= clusters)
3318 return -EINVAL;
3319
3320 atomic64_set(&sbi->s_resv_clusters, count);
3321 return 0;
3322 }
3323
3324 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3325 {
3326 char *orig_data = kstrdup(data, GFP_KERNEL);
3327 struct buffer_head *bh;
3328 struct ext4_super_block *es = NULL;
3329 struct ext4_sb_info *sbi;
3330 ext4_fsblk_t block;
3331 ext4_fsblk_t sb_block = get_sb_block(&data);
3332 ext4_fsblk_t logical_sb_block;
3333 unsigned long offset = 0;
3334 unsigned long journal_devnum = 0;
3335 unsigned long def_mount_opts;
3336 struct inode *root;
3337 char *cp;
3338 const char *descr;
3339 int ret = -ENOMEM;
3340 int blocksize, clustersize;
3341 unsigned int db_count;
3342 unsigned int i;
3343 int needs_recovery, has_huge_files, has_bigalloc;
3344 __u64 blocks_count;
3345 int err = 0;
3346 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3347 ext4_group_t first_not_zeroed;
3348
3349 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3350 if (!sbi)
3351 goto out_free_orig;
3352
3353 sbi->s_blockgroup_lock =
3354 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3355 if (!sbi->s_blockgroup_lock) {
3356 kfree(sbi);
3357 goto out_free_orig;
3358 }
3359 sb->s_fs_info = sbi;
3360 sbi->s_sb = sb;
3361 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3362 sbi->s_sb_block = sb_block;
3363 if (sb->s_bdev->bd_part)
3364 sbi->s_sectors_written_start =
3365 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3366
3367 /* Cleanup superblock name */
3368 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3369 *cp = '!';
3370
3371 /* -EINVAL is default */
3372 ret = -EINVAL;
3373 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3374 if (!blocksize) {
3375 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3376 goto out_fail;
3377 }
3378
3379 /*
3380 * The ext4 superblock will not be buffer aligned for other than 1kB
3381 * block sizes. We need to calculate the offset from buffer start.
3382 */
3383 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3384 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3385 offset = do_div(logical_sb_block, blocksize);
3386 } else {
3387 logical_sb_block = sb_block;
3388 }
3389
3390 if (!(bh = sb_bread(sb, logical_sb_block))) {
3391 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3392 goto out_fail;
3393 }
3394 /*
3395 * Note: s_es must be initialized as soon as possible because
3396 * some ext4 macro-instructions depend on its value
3397 */
3398 es = (struct ext4_super_block *) (bh->b_data + offset);
3399 sbi->s_es = es;
3400 sb->s_magic = le16_to_cpu(es->s_magic);
3401 if (sb->s_magic != EXT4_SUPER_MAGIC)
3402 goto cantfind_ext4;
3403 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3404
3405 /* Warn if metadata_csum and gdt_csum are both set. */
3406 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3407 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3408 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3409 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3410 "redundant flags; please run fsck.");
3411
3412 /* Check for a known checksum algorithm */
3413 if (!ext4_verify_csum_type(sb, es)) {
3414 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3415 "unknown checksum algorithm.");
3416 silent = 1;
3417 goto cantfind_ext4;
3418 }
3419
3420 /* Load the checksum driver */
3421 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3422 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3423 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3424 if (IS_ERR(sbi->s_chksum_driver)) {
3425 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3426 ret = PTR_ERR(sbi->s_chksum_driver);
3427 sbi->s_chksum_driver = NULL;
3428 goto failed_mount;
3429 }
3430 }
3431
3432 /* Check superblock checksum */
3433 if (!ext4_superblock_csum_verify(sb, es)) {
3434 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3435 "invalid superblock checksum. Run e2fsck?");
3436 silent = 1;
3437 goto cantfind_ext4;
3438 }
3439
3440 /* Precompute checksum seed for all metadata */
3441 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3442 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3443 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3444 sizeof(es->s_uuid));
3445
3446 /* Set defaults before we parse the mount options */
3447 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3448 set_opt(sb, INIT_INODE_TABLE);
3449 if (def_mount_opts & EXT4_DEFM_DEBUG)
3450 set_opt(sb, DEBUG);
3451 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3452 set_opt(sb, GRPID);
3453 if (def_mount_opts & EXT4_DEFM_UID16)
3454 set_opt(sb, NO_UID32);
3455 /* xattr user namespace & acls are now defaulted on */
3456 set_opt(sb, XATTR_USER);
3457 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3458 set_opt(sb, POSIX_ACL);
3459 #endif
3460 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3461 set_opt(sb, JOURNAL_DATA);
3462 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3463 set_opt(sb, ORDERED_DATA);
3464 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3465 set_opt(sb, WRITEBACK_DATA);
3466
3467 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3468 set_opt(sb, ERRORS_PANIC);
3469 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3470 set_opt(sb, ERRORS_CONT);
3471 else
3472 set_opt(sb, ERRORS_RO);
3473 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3474 set_opt(sb, BLOCK_VALIDITY);
3475 if (def_mount_opts & EXT4_DEFM_DISCARD)
3476 set_opt(sb, DISCARD);
3477
3478 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3479 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3480 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3481 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3482 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3483
3484 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3485 set_opt(sb, BARRIER);
3486
3487 /*
3488 * enable delayed allocation by default
3489 * Use -o nodelalloc to turn it off
3490 */
3491 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3492 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3493 set_opt(sb, DELALLOC);
3494
3495 /*
3496 * set default s_li_wait_mult for lazyinit, for the case there is
3497 * no mount option specified.
3498 */
3499 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3500
3501 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3502 &journal_devnum, &journal_ioprio, 0)) {
3503 ext4_msg(sb, KERN_WARNING,
3504 "failed to parse options in superblock: %s",
3505 sbi->s_es->s_mount_opts);
3506 }
3507 sbi->s_def_mount_opt = sbi->s_mount_opt;
3508 if (!parse_options((char *) data, sb, &journal_devnum,
3509 &journal_ioprio, 0))
3510 goto failed_mount;
3511
3512 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3513 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3514 "with data=journal disables delayed "
3515 "allocation and O_DIRECT support!\n");
3516 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3517 ext4_msg(sb, KERN_ERR, "can't mount with "
3518 "both data=journal and delalloc");
3519 goto failed_mount;
3520 }
3521 if (test_opt(sb, DIOREAD_NOLOCK)) {
3522 ext4_msg(sb, KERN_ERR, "can't mount with "
3523 "both data=journal and dioread_nolock");
3524 goto failed_mount;
3525 }
3526 if (test_opt(sb, DELALLOC))
3527 clear_opt(sb, DELALLOC);
3528 }
3529
3530 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3531 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3532
3533 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3534 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3535 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3536 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3537 ext4_msg(sb, KERN_WARNING,
3538 "feature flags set on rev 0 fs, "
3539 "running e2fsck is recommended");
3540
3541 if (IS_EXT2_SB(sb)) {
3542 if (ext2_feature_set_ok(sb))
3543 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3544 "using the ext4 subsystem");
3545 else {
3546 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3547 "to feature incompatibilities");
3548 goto failed_mount;
3549 }
3550 }
3551
3552 if (IS_EXT3_SB(sb)) {
3553 if (ext3_feature_set_ok(sb))
3554 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3555 "using the ext4 subsystem");
3556 else {
3557 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3558 "to feature incompatibilities");
3559 goto failed_mount;
3560 }
3561 }
3562
3563 /*
3564 * Check feature flags regardless of the revision level, since we
3565 * previously didn't change the revision level when setting the flags,
3566 * so there is a chance incompat flags are set on a rev 0 filesystem.
3567 */
3568 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3569 goto failed_mount;
3570
3571 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3572 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3573 blocksize > EXT4_MAX_BLOCK_SIZE) {
3574 ext4_msg(sb, KERN_ERR,
3575 "Unsupported filesystem blocksize %d", blocksize);
3576 goto failed_mount;
3577 }
3578
3579 if (sb->s_blocksize != blocksize) {
3580 /* Validate the filesystem blocksize */
3581 if (!sb_set_blocksize(sb, blocksize)) {
3582 ext4_msg(sb, KERN_ERR, "bad block size %d",
3583 blocksize);
3584 goto failed_mount;
3585 }
3586
3587 brelse(bh);
3588 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3589 offset = do_div(logical_sb_block, blocksize);
3590 bh = sb_bread(sb, logical_sb_block);
3591 if (!bh) {
3592 ext4_msg(sb, KERN_ERR,
3593 "Can't read superblock on 2nd try");
3594 goto failed_mount;
3595 }
3596 es = (struct ext4_super_block *)(bh->b_data + offset);
3597 sbi->s_es = es;
3598 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3599 ext4_msg(sb, KERN_ERR,
3600 "Magic mismatch, very weird!");
3601 goto failed_mount;
3602 }
3603 }
3604
3605 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3606 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3607 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3608 has_huge_files);
3609 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3610
3611 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3612 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3613 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3614 } else {
3615 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3616 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3617 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3618 (!is_power_of_2(sbi->s_inode_size)) ||
3619 (sbi->s_inode_size > blocksize)) {
3620 ext4_msg(sb, KERN_ERR,
3621 "unsupported inode size: %d",
3622 sbi->s_inode_size);
3623 goto failed_mount;
3624 }
3625 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3626 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3627 }
3628
3629 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3630 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3631 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3632 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3633 !is_power_of_2(sbi->s_desc_size)) {
3634 ext4_msg(sb, KERN_ERR,
3635 "unsupported descriptor size %lu",
3636 sbi->s_desc_size);
3637 goto failed_mount;
3638 }
3639 } else
3640 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3641
3642 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3643 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3644 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3645 goto cantfind_ext4;
3646
3647 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3648 if (sbi->s_inodes_per_block == 0)
3649 goto cantfind_ext4;
3650 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3651 sbi->s_inodes_per_block;
3652 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3653 sbi->s_sbh = bh;
3654 sbi->s_mount_state = le16_to_cpu(es->s_state);
3655 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3656 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3657
3658 for (i = 0; i < 4; i++)
3659 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3660 sbi->s_def_hash_version = es->s_def_hash_version;
3661 i = le32_to_cpu(es->s_flags);
3662 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3663 sbi->s_hash_unsigned = 3;
3664 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3665 #ifdef __CHAR_UNSIGNED__
3666 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3667 sbi->s_hash_unsigned = 3;
3668 #else
3669 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3670 #endif
3671 }
3672
3673 /* Handle clustersize */
3674 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3675 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3676 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3677 if (has_bigalloc) {
3678 if (clustersize < blocksize) {
3679 ext4_msg(sb, KERN_ERR,
3680 "cluster size (%d) smaller than "
3681 "block size (%d)", clustersize, blocksize);
3682 goto failed_mount;
3683 }
3684 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3685 le32_to_cpu(es->s_log_block_size);
3686 sbi->s_clusters_per_group =
3687 le32_to_cpu(es->s_clusters_per_group);
3688 if (sbi->s_clusters_per_group > blocksize * 8) {
3689 ext4_msg(sb, KERN_ERR,
3690 "#clusters per group too big: %lu",
3691 sbi->s_clusters_per_group);
3692 goto failed_mount;
3693 }
3694 if (sbi->s_blocks_per_group !=
3695 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3696 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3697 "clusters per group (%lu) inconsistent",
3698 sbi->s_blocks_per_group,
3699 sbi->s_clusters_per_group);
3700 goto failed_mount;
3701 }
3702 } else {
3703 if (clustersize != blocksize) {
3704 ext4_warning(sb, "fragment/cluster size (%d) != "
3705 "block size (%d)", clustersize,
3706 blocksize);
3707 clustersize = blocksize;
3708 }
3709 if (sbi->s_blocks_per_group > blocksize * 8) {
3710 ext4_msg(sb, KERN_ERR,
3711 "#blocks per group too big: %lu",
3712 sbi->s_blocks_per_group);
3713 goto failed_mount;
3714 }
3715 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3716 sbi->s_cluster_bits = 0;
3717 }
3718 sbi->s_cluster_ratio = clustersize / blocksize;
3719
3720 if (sbi->s_inodes_per_group > blocksize * 8) {
3721 ext4_msg(sb, KERN_ERR,
3722 "#inodes per group too big: %lu",
3723 sbi->s_inodes_per_group);
3724 goto failed_mount;
3725 }
3726
3727 /* Do we have standard group size of clustersize * 8 blocks ? */
3728 if (sbi->s_blocks_per_group == clustersize << 3)
3729 set_opt2(sb, STD_GROUP_SIZE);
3730
3731 /*
3732 * Test whether we have more sectors than will fit in sector_t,
3733 * and whether the max offset is addressable by the page cache.
3734 */
3735 err = generic_check_addressable(sb->s_blocksize_bits,
3736 ext4_blocks_count(es));
3737 if (err) {
3738 ext4_msg(sb, KERN_ERR, "filesystem"
3739 " too large to mount safely on this system");
3740 if (sizeof(sector_t) < 8)
3741 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3742 goto failed_mount;
3743 }
3744
3745 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3746 goto cantfind_ext4;
3747
3748 /* check blocks count against device size */
3749 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3750 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3751 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3752 "exceeds size of device (%llu blocks)",
3753 ext4_blocks_count(es), blocks_count);
3754 goto failed_mount;
3755 }
3756
3757 /*
3758 * It makes no sense for the first data block to be beyond the end
3759 * of the filesystem.
3760 */
3761 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3762 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3763 "block %u is beyond end of filesystem (%llu)",
3764 le32_to_cpu(es->s_first_data_block),
3765 ext4_blocks_count(es));
3766 goto failed_mount;
3767 }
3768 blocks_count = (ext4_blocks_count(es) -
3769 le32_to_cpu(es->s_first_data_block) +
3770 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3771 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3772 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3773 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3774 "(block count %llu, first data block %u, "
3775 "blocks per group %lu)", sbi->s_groups_count,
3776 ext4_blocks_count(es),
3777 le32_to_cpu(es->s_first_data_block),
3778 EXT4_BLOCKS_PER_GROUP(sb));
3779 goto failed_mount;
3780 }
3781 sbi->s_groups_count = blocks_count;
3782 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3783 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3784 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3785 EXT4_DESC_PER_BLOCK(sb);
3786 sbi->s_group_desc = ext4_kvmalloc(db_count *
3787 sizeof(struct buffer_head *),
3788 GFP_KERNEL);
3789 if (sbi->s_group_desc == NULL) {
3790 ext4_msg(sb, KERN_ERR, "not enough memory");
3791 ret = -ENOMEM;
3792 goto failed_mount;
3793 }
3794
3795 if (ext4_proc_root)
3796 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3797
3798 if (sbi->s_proc)
3799 proc_create_data("options", S_IRUGO, sbi->s_proc,
3800 &ext4_seq_options_fops, sb);
3801
3802 bgl_lock_init(sbi->s_blockgroup_lock);
3803
3804 for (i = 0; i < db_count; i++) {
3805 block = descriptor_loc(sb, logical_sb_block, i);
3806 sbi->s_group_desc[i] = sb_bread(sb, block);
3807 if (!sbi->s_group_desc[i]) {
3808 ext4_msg(sb, KERN_ERR,
3809 "can't read group descriptor %d", i);
3810 db_count = i;
3811 goto failed_mount2;
3812 }
3813 }
3814 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3815 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3816 goto failed_mount2;
3817 }
3818 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3819 if (!ext4_fill_flex_info(sb)) {
3820 ext4_msg(sb, KERN_ERR,
3821 "unable to initialize "
3822 "flex_bg meta info!");
3823 goto failed_mount2;
3824 }
3825
3826 sbi->s_gdb_count = db_count;
3827 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3828 spin_lock_init(&sbi->s_next_gen_lock);
3829
3830 init_timer(&sbi->s_err_report);
3831 sbi->s_err_report.function = print_daily_error_info;
3832 sbi->s_err_report.data = (unsigned long) sb;
3833
3834 /* Register extent status tree shrinker */
3835 ext4_es_register_shrinker(sbi);
3836
3837 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3838 ext4_count_free_clusters(sb));
3839 if (!err) {
3840 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3841 ext4_count_free_inodes(sb));
3842 }
3843 if (!err) {
3844 err = percpu_counter_init(&sbi->s_dirs_counter,
3845 ext4_count_dirs(sb));
3846 }
3847 if (!err) {
3848 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3849 }
3850 if (!err) {
3851 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3852 }
3853 if (err) {
3854 ext4_msg(sb, KERN_ERR, "insufficient memory");
3855 goto failed_mount3;
3856 }
3857
3858 sbi->s_stripe = ext4_get_stripe_size(sbi);
3859 sbi->s_extent_max_zeroout_kb = 32;
3860
3861 /*
3862 * set up enough so that it can read an inode
3863 */
3864 if (!test_opt(sb, NOLOAD) &&
3865 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3866 sb->s_op = &ext4_sops;
3867 else
3868 sb->s_op = &ext4_nojournal_sops;
3869 sb->s_export_op = &ext4_export_ops;
3870 sb->s_xattr = ext4_xattr_handlers;
3871 #ifdef CONFIG_QUOTA
3872 sb->dq_op = &ext4_quota_operations;
3873 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
3874 sb->s_qcop = &ext4_qctl_sysfile_operations;
3875 else
3876 sb->s_qcop = &ext4_qctl_operations;
3877 #endif
3878 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3879
3880 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3881 mutex_init(&sbi->s_orphan_lock);
3882
3883 sb->s_root = NULL;
3884
3885 needs_recovery = (es->s_last_orphan != 0 ||
3886 EXT4_HAS_INCOMPAT_FEATURE(sb,
3887 EXT4_FEATURE_INCOMPAT_RECOVER));
3888
3889 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3890 !(sb->s_flags & MS_RDONLY))
3891 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3892 goto failed_mount3;
3893
3894 /*
3895 * The first inode we look at is the journal inode. Don't try
3896 * root first: it may be modified in the journal!
3897 */
3898 if (!test_opt(sb, NOLOAD) &&
3899 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3900 if (ext4_load_journal(sb, es, journal_devnum))
3901 goto failed_mount3;
3902 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3903 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3904 ext4_msg(sb, KERN_ERR, "required journal recovery "
3905 "suppressed and not mounted read-only");
3906 goto failed_mount_wq;
3907 } else {
3908 clear_opt(sb, DATA_FLAGS);
3909 sbi->s_journal = NULL;
3910 needs_recovery = 0;
3911 goto no_journal;
3912 }
3913
3914 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3915 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3916 JBD2_FEATURE_INCOMPAT_64BIT)) {
3917 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3918 goto failed_mount_wq;
3919 }
3920
3921 if (!set_journal_csum_feature_set(sb)) {
3922 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3923 "feature set");
3924 goto failed_mount_wq;
3925 }
3926
3927 /* We have now updated the journal if required, so we can
3928 * validate the data journaling mode. */
3929 switch (test_opt(sb, DATA_FLAGS)) {
3930 case 0:
3931 /* No mode set, assume a default based on the journal
3932 * capabilities: ORDERED_DATA if the journal can
3933 * cope, else JOURNAL_DATA
3934 */
3935 if (jbd2_journal_check_available_features
3936 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3937 set_opt(sb, ORDERED_DATA);
3938 else
3939 set_opt(sb, JOURNAL_DATA);
3940 break;
3941
3942 case EXT4_MOUNT_ORDERED_DATA:
3943 case EXT4_MOUNT_WRITEBACK_DATA:
3944 if (!jbd2_journal_check_available_features
3945 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3946 ext4_msg(sb, KERN_ERR, "Journal does not support "
3947 "requested data journaling mode");
3948 goto failed_mount_wq;
3949 }
3950 default:
3951 break;
3952 }
3953 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3954
3955 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3956
3957 /*
3958 * The journal may have updated the bg summary counts, so we
3959 * need to update the global counters.
3960 */
3961 percpu_counter_set(&sbi->s_freeclusters_counter,
3962 ext4_count_free_clusters(sb));
3963 percpu_counter_set(&sbi->s_freeinodes_counter,
3964 ext4_count_free_inodes(sb));
3965 percpu_counter_set(&sbi->s_dirs_counter,
3966 ext4_count_dirs(sb));
3967 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3968
3969 no_journal:
3970 /*
3971 * Get the # of file system overhead blocks from the
3972 * superblock if present.
3973 */
3974 if (es->s_overhead_clusters)
3975 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3976 else {
3977 err = ext4_calculate_overhead(sb);
3978 if (err)
3979 goto failed_mount_wq;
3980 }
3981
3982 /*
3983 * The maximum number of concurrent works can be high and
3984 * concurrency isn't really necessary. Limit it to 1.
3985 */
3986 EXT4_SB(sb)->rsv_conversion_wq =
3987 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3988 if (!EXT4_SB(sb)->rsv_conversion_wq) {
3989 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
3990 ret = -ENOMEM;
3991 goto failed_mount4;
3992 }
3993
3994 /*
3995 * The jbd2_journal_load will have done any necessary log recovery,
3996 * so we can safely mount the rest of the filesystem now.
3997 */
3998
3999 root = ext4_iget(sb, EXT4_ROOT_INO);
4000 if (IS_ERR(root)) {
4001 ext4_msg(sb, KERN_ERR, "get root inode failed");
4002 ret = PTR_ERR(root);
4003 root = NULL;
4004 goto failed_mount4;
4005 }
4006 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4007 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4008 iput(root);
4009 goto failed_mount4;
4010 }
4011 sb->s_root = d_make_root(root);
4012 if (!sb->s_root) {
4013 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4014 ret = -ENOMEM;
4015 goto failed_mount4;
4016 }
4017
4018 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4019 sb->s_flags |= MS_RDONLY;
4020
4021 /* determine the minimum size of new large inodes, if present */
4022 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4023 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4024 EXT4_GOOD_OLD_INODE_SIZE;
4025 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4026 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4027 if (sbi->s_want_extra_isize <
4028 le16_to_cpu(es->s_want_extra_isize))
4029 sbi->s_want_extra_isize =
4030 le16_to_cpu(es->s_want_extra_isize);
4031 if (sbi->s_want_extra_isize <
4032 le16_to_cpu(es->s_min_extra_isize))
4033 sbi->s_want_extra_isize =
4034 le16_to_cpu(es->s_min_extra_isize);
4035 }
4036 }
4037 /* Check if enough inode space is available */
4038 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4039 sbi->s_inode_size) {
4040 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4041 EXT4_GOOD_OLD_INODE_SIZE;
4042 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4043 "available");
4044 }
4045
4046 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sbi));
4047 if (err) {
4048 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4049 "reserved pool", ext4_calculate_resv_clusters(sbi));
4050 goto failed_mount4a;
4051 }
4052
4053 err = ext4_setup_system_zone(sb);
4054 if (err) {
4055 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4056 "zone (%d)", err);
4057 goto failed_mount4a;
4058 }
4059
4060 ext4_ext_init(sb);
4061 err = ext4_mb_init(sb);
4062 if (err) {
4063 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4064 err);
4065 goto failed_mount5;
4066 }
4067
4068 err = ext4_register_li_request(sb, first_not_zeroed);
4069 if (err)
4070 goto failed_mount6;
4071
4072 sbi->s_kobj.kset = ext4_kset;
4073 init_completion(&sbi->s_kobj_unregister);
4074 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4075 "%s", sb->s_id);
4076 if (err)
4077 goto failed_mount7;
4078
4079 #ifdef CONFIG_QUOTA
4080 /* Enable quota usage during mount. */
4081 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4082 !(sb->s_flags & MS_RDONLY)) {
4083 err = ext4_enable_quotas(sb);
4084 if (err)
4085 goto failed_mount8;
4086 }
4087 #endif /* CONFIG_QUOTA */
4088
4089 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4090 ext4_orphan_cleanup(sb, es);
4091 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4092 if (needs_recovery) {
4093 ext4_msg(sb, KERN_INFO, "recovery complete");
4094 ext4_mark_recovery_complete(sb, es);
4095 }
4096 if (EXT4_SB(sb)->s_journal) {
4097 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4098 descr = " journalled data mode";
4099 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4100 descr = " ordered data mode";
4101 else
4102 descr = " writeback data mode";
4103 } else
4104 descr = "out journal";
4105
4106 if (test_opt(sb, DISCARD)) {
4107 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4108 if (!blk_queue_discard(q))
4109 ext4_msg(sb, KERN_WARNING,
4110 "mounting with \"discard\" option, but "
4111 "the device does not support discard");
4112 }
4113
4114 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4115 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4116 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4117
4118 if (es->s_error_count)
4119 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4120
4121 kfree(orig_data);
4122 return 0;
4123
4124 cantfind_ext4:
4125 if (!silent)
4126 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4127 goto failed_mount;
4128
4129 #ifdef CONFIG_QUOTA
4130 failed_mount8:
4131 kobject_del(&sbi->s_kobj);
4132 #endif
4133 failed_mount7:
4134 ext4_unregister_li_request(sb);
4135 failed_mount6:
4136 ext4_mb_release(sb);
4137 failed_mount5:
4138 ext4_ext_release(sb);
4139 ext4_release_system_zone(sb);
4140 failed_mount4a:
4141 dput(sb->s_root);
4142 sb->s_root = NULL;
4143 failed_mount4:
4144 ext4_msg(sb, KERN_ERR, "mount failed");
4145 if (EXT4_SB(sb)->rsv_conversion_wq)
4146 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4147 failed_mount_wq:
4148 if (sbi->s_journal) {
4149 jbd2_journal_destroy(sbi->s_journal);
4150 sbi->s_journal = NULL;
4151 }
4152 failed_mount3:
4153 ext4_es_unregister_shrinker(sbi);
4154 del_timer(&sbi->s_err_report);
4155 if (sbi->s_flex_groups)
4156 ext4_kvfree(sbi->s_flex_groups);
4157 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4158 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4159 percpu_counter_destroy(&sbi->s_dirs_counter);
4160 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4161 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
4162 if (sbi->s_mmp_tsk)
4163 kthread_stop(sbi->s_mmp_tsk);
4164 failed_mount2:
4165 for (i = 0; i < db_count; i++)
4166 brelse(sbi->s_group_desc[i]);
4167 ext4_kvfree(sbi->s_group_desc);
4168 failed_mount:
4169 if (sbi->s_chksum_driver)
4170 crypto_free_shash(sbi->s_chksum_driver);
4171 if (sbi->s_proc) {
4172 remove_proc_entry("options", sbi->s_proc);
4173 remove_proc_entry(sb->s_id, ext4_proc_root);
4174 }
4175 #ifdef CONFIG_QUOTA
4176 for (i = 0; i < MAXQUOTAS; i++)
4177 kfree(sbi->s_qf_names[i]);
4178 #endif
4179 ext4_blkdev_remove(sbi);
4180 brelse(bh);
4181 out_fail:
4182 sb->s_fs_info = NULL;
4183 kfree(sbi->s_blockgroup_lock);
4184 kfree(sbi);
4185 out_free_orig:
4186 kfree(orig_data);
4187 return err ? err : ret;
4188 }
4189
4190 /*
4191 * Setup any per-fs journal parameters now. We'll do this both on
4192 * initial mount, once the journal has been initialised but before we've
4193 * done any recovery; and again on any subsequent remount.
4194 */
4195 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4196 {
4197 struct ext4_sb_info *sbi = EXT4_SB(sb);
4198
4199 journal->j_commit_interval = sbi->s_commit_interval;
4200 journal->j_min_batch_time = sbi->s_min_batch_time;
4201 journal->j_max_batch_time = sbi->s_max_batch_time;
4202
4203 write_lock(&journal->j_state_lock);
4204 if (test_opt(sb, BARRIER))
4205 journal->j_flags |= JBD2_BARRIER;
4206 else
4207 journal->j_flags &= ~JBD2_BARRIER;
4208 if (test_opt(sb, DATA_ERR_ABORT))
4209 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4210 else
4211 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4212 write_unlock(&journal->j_state_lock);
4213 }
4214
4215 static journal_t *ext4_get_journal(struct super_block *sb,
4216 unsigned int journal_inum)
4217 {
4218 struct inode *journal_inode;
4219 journal_t *journal;
4220
4221 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4222
4223 /* First, test for the existence of a valid inode on disk. Bad
4224 * things happen if we iget() an unused inode, as the subsequent
4225 * iput() will try to delete it. */
4226
4227 journal_inode = ext4_iget(sb, journal_inum);
4228 if (IS_ERR(journal_inode)) {
4229 ext4_msg(sb, KERN_ERR, "no journal found");
4230 return NULL;
4231 }
4232 if (!journal_inode->i_nlink) {
4233 make_bad_inode(journal_inode);
4234 iput(journal_inode);
4235 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4236 return NULL;
4237 }
4238
4239 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4240 journal_inode, journal_inode->i_size);
4241 if (!S_ISREG(journal_inode->i_mode)) {
4242 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4243 iput(journal_inode);
4244 return NULL;
4245 }
4246
4247 journal = jbd2_journal_init_inode(journal_inode);
4248 if (!journal) {
4249 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4250 iput(journal_inode);
4251 return NULL;
4252 }
4253 journal->j_private = sb;
4254 ext4_init_journal_params(sb, journal);
4255 return journal;
4256 }
4257
4258 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4259 dev_t j_dev)
4260 {
4261 struct buffer_head *bh;
4262 journal_t *journal;
4263 ext4_fsblk_t start;
4264 ext4_fsblk_t len;
4265 int hblock, blocksize;
4266 ext4_fsblk_t sb_block;
4267 unsigned long offset;
4268 struct ext4_super_block *es;
4269 struct block_device *bdev;
4270
4271 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4272
4273 bdev = ext4_blkdev_get(j_dev, sb);
4274 if (bdev == NULL)
4275 return NULL;
4276
4277 blocksize = sb->s_blocksize;
4278 hblock = bdev_logical_block_size(bdev);
4279 if (blocksize < hblock) {
4280 ext4_msg(sb, KERN_ERR,
4281 "blocksize too small for journal device");
4282 goto out_bdev;
4283 }
4284
4285 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4286 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4287 set_blocksize(bdev, blocksize);
4288 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4289 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4290 "external journal");
4291 goto out_bdev;
4292 }
4293
4294 es = (struct ext4_super_block *) (bh->b_data + offset);
4295 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4296 !(le32_to_cpu(es->s_feature_incompat) &
4297 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4298 ext4_msg(sb, KERN_ERR, "external journal has "
4299 "bad superblock");
4300 brelse(bh);
4301 goto out_bdev;
4302 }
4303
4304 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4305 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4306 brelse(bh);
4307 goto out_bdev;
4308 }
4309
4310 len = ext4_blocks_count(es);
4311 start = sb_block + 1;
4312 brelse(bh); /* we're done with the superblock */
4313
4314 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4315 start, len, blocksize);
4316 if (!journal) {
4317 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4318 goto out_bdev;
4319 }
4320 journal->j_private = sb;
4321 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4322 wait_on_buffer(journal->j_sb_buffer);
4323 if (!buffer_uptodate(journal->j_sb_buffer)) {
4324 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4325 goto out_journal;
4326 }
4327 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4328 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4329 "user (unsupported) - %d",
4330 be32_to_cpu(journal->j_superblock->s_nr_users));
4331 goto out_journal;
4332 }
4333 EXT4_SB(sb)->journal_bdev = bdev;
4334 ext4_init_journal_params(sb, journal);
4335 return journal;
4336
4337 out_journal:
4338 jbd2_journal_destroy(journal);
4339 out_bdev:
4340 ext4_blkdev_put(bdev);
4341 return NULL;
4342 }
4343
4344 static int ext4_load_journal(struct super_block *sb,
4345 struct ext4_super_block *es,
4346 unsigned long journal_devnum)
4347 {
4348 journal_t *journal;
4349 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4350 dev_t journal_dev;
4351 int err = 0;
4352 int really_read_only;
4353
4354 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4355
4356 if (journal_devnum &&
4357 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4358 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4359 "numbers have changed");
4360 journal_dev = new_decode_dev(journal_devnum);
4361 } else
4362 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4363
4364 really_read_only = bdev_read_only(sb->s_bdev);
4365
4366 /*
4367 * Are we loading a blank journal or performing recovery after a
4368 * crash? For recovery, we need to check in advance whether we
4369 * can get read-write access to the device.
4370 */
4371 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4372 if (sb->s_flags & MS_RDONLY) {
4373 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4374 "required on readonly filesystem");
4375 if (really_read_only) {
4376 ext4_msg(sb, KERN_ERR, "write access "
4377 "unavailable, cannot proceed");
4378 return -EROFS;
4379 }
4380 ext4_msg(sb, KERN_INFO, "write access will "
4381 "be enabled during recovery");
4382 }
4383 }
4384
4385 if (journal_inum && journal_dev) {
4386 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4387 "and inode journals!");
4388 return -EINVAL;
4389 }
4390
4391 if (journal_inum) {
4392 if (!(journal = ext4_get_journal(sb, journal_inum)))
4393 return -EINVAL;
4394 } else {
4395 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4396 return -EINVAL;
4397 }
4398
4399 if (!(journal->j_flags & JBD2_BARRIER))
4400 ext4_msg(sb, KERN_INFO, "barriers disabled");
4401
4402 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4403 err = jbd2_journal_wipe(journal, !really_read_only);
4404 if (!err) {
4405 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4406 if (save)
4407 memcpy(save, ((char *) es) +
4408 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4409 err = jbd2_journal_load(journal);
4410 if (save)
4411 memcpy(((char *) es) + EXT4_S_ERR_START,
4412 save, EXT4_S_ERR_LEN);
4413 kfree(save);
4414 }
4415
4416 if (err) {
4417 ext4_msg(sb, KERN_ERR, "error loading journal");
4418 jbd2_journal_destroy(journal);
4419 return err;
4420 }
4421
4422 EXT4_SB(sb)->s_journal = journal;
4423 ext4_clear_journal_err(sb, es);
4424
4425 if (!really_read_only && journal_devnum &&
4426 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4427 es->s_journal_dev = cpu_to_le32(journal_devnum);
4428
4429 /* Make sure we flush the recovery flag to disk. */
4430 ext4_commit_super(sb, 1);
4431 }
4432
4433 return 0;
4434 }
4435
4436 static int ext4_commit_super(struct super_block *sb, int sync)
4437 {
4438 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4439 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4440 int error = 0;
4441
4442 if (!sbh || block_device_ejected(sb))
4443 return error;
4444 if (buffer_write_io_error(sbh)) {
4445 /*
4446 * Oh, dear. A previous attempt to write the
4447 * superblock failed. This could happen because the
4448 * USB device was yanked out. Or it could happen to
4449 * be a transient write error and maybe the block will
4450 * be remapped. Nothing we can do but to retry the
4451 * write and hope for the best.
4452 */
4453 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4454 "superblock detected");
4455 clear_buffer_write_io_error(sbh);
4456 set_buffer_uptodate(sbh);
4457 }
4458 /*
4459 * If the file system is mounted read-only, don't update the
4460 * superblock write time. This avoids updating the superblock
4461 * write time when we are mounting the root file system
4462 * read/only but we need to replay the journal; at that point,
4463 * for people who are east of GMT and who make their clock
4464 * tick in localtime for Windows bug-for-bug compatibility,
4465 * the clock is set in the future, and this will cause e2fsck
4466 * to complain and force a full file system check.
4467 */
4468 if (!(sb->s_flags & MS_RDONLY))
4469 es->s_wtime = cpu_to_le32(get_seconds());
4470 if (sb->s_bdev->bd_part)
4471 es->s_kbytes_written =
4472 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4473 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4474 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4475 else
4476 es->s_kbytes_written =
4477 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4478 ext4_free_blocks_count_set(es,
4479 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4480 &EXT4_SB(sb)->s_freeclusters_counter)));
4481 es->s_free_inodes_count =
4482 cpu_to_le32(percpu_counter_sum_positive(
4483 &EXT4_SB(sb)->s_freeinodes_counter));
4484 BUFFER_TRACE(sbh, "marking dirty");
4485 ext4_superblock_csum_set(sb);
4486 mark_buffer_dirty(sbh);
4487 if (sync) {
4488 error = sync_dirty_buffer(sbh);
4489 if (error)
4490 return error;
4491
4492 error = buffer_write_io_error(sbh);
4493 if (error) {
4494 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4495 "superblock");
4496 clear_buffer_write_io_error(sbh);
4497 set_buffer_uptodate(sbh);
4498 }
4499 }
4500 return error;
4501 }
4502
4503 /*
4504 * Have we just finished recovery? If so, and if we are mounting (or
4505 * remounting) the filesystem readonly, then we will end up with a
4506 * consistent fs on disk. Record that fact.
4507 */
4508 static void ext4_mark_recovery_complete(struct super_block *sb,
4509 struct ext4_super_block *es)
4510 {
4511 journal_t *journal = EXT4_SB(sb)->s_journal;
4512
4513 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4514 BUG_ON(journal != NULL);
4515 return;
4516 }
4517 jbd2_journal_lock_updates(journal);
4518 if (jbd2_journal_flush(journal) < 0)
4519 goto out;
4520
4521 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4522 sb->s_flags & MS_RDONLY) {
4523 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4524 ext4_commit_super(sb, 1);
4525 }
4526
4527 out:
4528 jbd2_journal_unlock_updates(journal);
4529 }
4530
4531 /*
4532 * If we are mounting (or read-write remounting) a filesystem whose journal
4533 * has recorded an error from a previous lifetime, move that error to the
4534 * main filesystem now.
4535 */
4536 static void ext4_clear_journal_err(struct super_block *sb,
4537 struct ext4_super_block *es)
4538 {
4539 journal_t *journal;
4540 int j_errno;
4541 const char *errstr;
4542
4543 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4544
4545 journal = EXT4_SB(sb)->s_journal;
4546
4547 /*
4548 * Now check for any error status which may have been recorded in the
4549 * journal by a prior ext4_error() or ext4_abort()
4550 */
4551
4552 j_errno = jbd2_journal_errno(journal);
4553 if (j_errno) {
4554 char nbuf[16];
4555
4556 errstr = ext4_decode_error(sb, j_errno, nbuf);
4557 ext4_warning(sb, "Filesystem error recorded "
4558 "from previous mount: %s", errstr);
4559 ext4_warning(sb, "Marking fs in need of filesystem check.");
4560
4561 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4562 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4563 ext4_commit_super(sb, 1);
4564
4565 jbd2_journal_clear_err(journal);
4566 jbd2_journal_update_sb_errno(journal);
4567 }
4568 }
4569
4570 /*
4571 * Force the running and committing transactions to commit,
4572 * and wait on the commit.
4573 */
4574 int ext4_force_commit(struct super_block *sb)
4575 {
4576 journal_t *journal;
4577
4578 if (sb->s_flags & MS_RDONLY)
4579 return 0;
4580
4581 journal = EXT4_SB(sb)->s_journal;
4582 return ext4_journal_force_commit(journal);
4583 }
4584
4585 static int ext4_sync_fs(struct super_block *sb, int wait)
4586 {
4587 int ret = 0;
4588 tid_t target;
4589 bool needs_barrier = false;
4590 struct ext4_sb_info *sbi = EXT4_SB(sb);
4591
4592 trace_ext4_sync_fs(sb, wait);
4593 flush_workqueue(sbi->rsv_conversion_wq);
4594 /*
4595 * Writeback quota in non-journalled quota case - journalled quota has
4596 * no dirty dquots
4597 */
4598 dquot_writeback_dquots(sb, -1);
4599 /*
4600 * Data writeback is possible w/o journal transaction, so barrier must
4601 * being sent at the end of the function. But we can skip it if
4602 * transaction_commit will do it for us.
4603 */
4604 target = jbd2_get_latest_transaction(sbi->s_journal);
4605 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4606 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4607 needs_barrier = true;
4608
4609 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4610 if (wait)
4611 ret = jbd2_log_wait_commit(sbi->s_journal, target);
4612 }
4613 if (needs_barrier) {
4614 int err;
4615 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4616 if (!ret)
4617 ret = err;
4618 }
4619
4620 return ret;
4621 }
4622
4623 static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4624 {
4625 int ret = 0;
4626
4627 trace_ext4_sync_fs(sb, wait);
4628 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4629 dquot_writeback_dquots(sb, -1);
4630 if (wait && test_opt(sb, BARRIER))
4631 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4632
4633 return ret;
4634 }
4635
4636 /*
4637 * LVM calls this function before a (read-only) snapshot is created. This
4638 * gives us a chance to flush the journal completely and mark the fs clean.
4639 *
4640 * Note that only this function cannot bring a filesystem to be in a clean
4641 * state independently. It relies on upper layer to stop all data & metadata
4642 * modifications.
4643 */
4644 static int ext4_freeze(struct super_block *sb)
4645 {
4646 int error = 0;
4647 journal_t *journal;
4648
4649 if (sb->s_flags & MS_RDONLY)
4650 return 0;
4651
4652 journal = EXT4_SB(sb)->s_journal;
4653
4654 /* Now we set up the journal barrier. */
4655 jbd2_journal_lock_updates(journal);
4656
4657 /*
4658 * Don't clear the needs_recovery flag if we failed to flush
4659 * the journal.
4660 */
4661 error = jbd2_journal_flush(journal);
4662 if (error < 0)
4663 goto out;
4664
4665 /* Journal blocked and flushed, clear needs_recovery flag. */
4666 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4667 error = ext4_commit_super(sb, 1);
4668 out:
4669 /* we rely on upper layer to stop further updates */
4670 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4671 return error;
4672 }
4673
4674 /*
4675 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4676 * flag here, even though the filesystem is not technically dirty yet.
4677 */
4678 static int ext4_unfreeze(struct super_block *sb)
4679 {
4680 if (sb->s_flags & MS_RDONLY)
4681 return 0;
4682
4683 /* Reset the needs_recovery flag before the fs is unlocked. */
4684 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4685 ext4_commit_super(sb, 1);
4686 return 0;
4687 }
4688
4689 /*
4690 * Structure to save mount options for ext4_remount's benefit
4691 */
4692 struct ext4_mount_options {
4693 unsigned long s_mount_opt;
4694 unsigned long s_mount_opt2;
4695 kuid_t s_resuid;
4696 kgid_t s_resgid;
4697 unsigned long s_commit_interval;
4698 u32 s_min_batch_time, s_max_batch_time;
4699 #ifdef CONFIG_QUOTA
4700 int s_jquota_fmt;
4701 char *s_qf_names[MAXQUOTAS];
4702 #endif
4703 };
4704
4705 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4706 {
4707 struct ext4_super_block *es;
4708 struct ext4_sb_info *sbi = EXT4_SB(sb);
4709 unsigned long old_sb_flags;
4710 struct ext4_mount_options old_opts;
4711 int enable_quota = 0;
4712 ext4_group_t g;
4713 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4714 int err = 0;
4715 #ifdef CONFIG_QUOTA
4716 int i, j;
4717 #endif
4718 char *orig_data = kstrdup(data, GFP_KERNEL);
4719
4720 /* Store the original options */
4721 old_sb_flags = sb->s_flags;
4722 old_opts.s_mount_opt = sbi->s_mount_opt;
4723 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4724 old_opts.s_resuid = sbi->s_resuid;
4725 old_opts.s_resgid = sbi->s_resgid;
4726 old_opts.s_commit_interval = sbi->s_commit_interval;
4727 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4728 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4729 #ifdef CONFIG_QUOTA
4730 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4731 for (i = 0; i < MAXQUOTAS; i++)
4732 if (sbi->s_qf_names[i]) {
4733 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4734 GFP_KERNEL);
4735 if (!old_opts.s_qf_names[i]) {
4736 for (j = 0; j < i; j++)
4737 kfree(old_opts.s_qf_names[j]);
4738 kfree(orig_data);
4739 return -ENOMEM;
4740 }
4741 } else
4742 old_opts.s_qf_names[i] = NULL;
4743 #endif
4744 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4745 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4746
4747 /*
4748 * Allow the "check" option to be passed as a remount option.
4749 */
4750 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4751 err = -EINVAL;
4752 goto restore_opts;
4753 }
4754
4755 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4756 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4757 ext4_msg(sb, KERN_ERR, "can't mount with "
4758 "both data=journal and delalloc");
4759 err = -EINVAL;
4760 goto restore_opts;
4761 }
4762 if (test_opt(sb, DIOREAD_NOLOCK)) {
4763 ext4_msg(sb, KERN_ERR, "can't mount with "
4764 "both data=journal and dioread_nolock");
4765 err = -EINVAL;
4766 goto restore_opts;
4767 }
4768 }
4769
4770 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4771 ext4_abort(sb, "Abort forced by user");
4772
4773 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4774 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4775
4776 es = sbi->s_es;
4777
4778 if (sbi->s_journal) {
4779 ext4_init_journal_params(sb, sbi->s_journal);
4780 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4781 }
4782
4783 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4784 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4785 err = -EROFS;
4786 goto restore_opts;
4787 }
4788
4789 if (*flags & MS_RDONLY) {
4790 err = dquot_suspend(sb, -1);
4791 if (err < 0)
4792 goto restore_opts;
4793
4794 /*
4795 * First of all, the unconditional stuff we have to do
4796 * to disable replay of the journal when we next remount
4797 */
4798 sb->s_flags |= MS_RDONLY;
4799
4800 /*
4801 * OK, test if we are remounting a valid rw partition
4802 * readonly, and if so set the rdonly flag and then
4803 * mark the partition as valid again.
4804 */
4805 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4806 (sbi->s_mount_state & EXT4_VALID_FS))
4807 es->s_state = cpu_to_le16(sbi->s_mount_state);
4808
4809 if (sbi->s_journal)
4810 ext4_mark_recovery_complete(sb, es);
4811 } else {
4812 /* Make sure we can mount this feature set readwrite */
4813 if (!ext4_feature_set_ok(sb, 0)) {
4814 err = -EROFS;
4815 goto restore_opts;
4816 }
4817 /*
4818 * Make sure the group descriptor checksums
4819 * are sane. If they aren't, refuse to remount r/w.
4820 */
4821 for (g = 0; g < sbi->s_groups_count; g++) {
4822 struct ext4_group_desc *gdp =
4823 ext4_get_group_desc(sb, g, NULL);
4824
4825 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4826 ext4_msg(sb, KERN_ERR,
4827 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4828 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4829 le16_to_cpu(gdp->bg_checksum));
4830 err = -EINVAL;
4831 goto restore_opts;
4832 }
4833 }
4834
4835 /*
4836 * If we have an unprocessed orphan list hanging
4837 * around from a previously readonly bdev mount,
4838 * require a full umount/remount for now.
4839 */
4840 if (es->s_last_orphan) {
4841 ext4_msg(sb, KERN_WARNING, "Couldn't "
4842 "remount RDWR because of unprocessed "
4843 "orphan inode list. Please "
4844 "umount/remount instead");
4845 err = -EINVAL;
4846 goto restore_opts;
4847 }
4848
4849 /*
4850 * Mounting a RDONLY partition read-write, so reread
4851 * and store the current valid flag. (It may have
4852 * been changed by e2fsck since we originally mounted
4853 * the partition.)
4854 */
4855 if (sbi->s_journal)
4856 ext4_clear_journal_err(sb, es);
4857 sbi->s_mount_state = le16_to_cpu(es->s_state);
4858 if (!ext4_setup_super(sb, es, 0))
4859 sb->s_flags &= ~MS_RDONLY;
4860 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4861 EXT4_FEATURE_INCOMPAT_MMP))
4862 if (ext4_multi_mount_protect(sb,
4863 le64_to_cpu(es->s_mmp_block))) {
4864 err = -EROFS;
4865 goto restore_opts;
4866 }
4867 enable_quota = 1;
4868 }
4869 }
4870
4871 /*
4872 * Reinitialize lazy itable initialization thread based on
4873 * current settings
4874 */
4875 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4876 ext4_unregister_li_request(sb);
4877 else {
4878 ext4_group_t first_not_zeroed;
4879 first_not_zeroed = ext4_has_uninit_itable(sb);
4880 ext4_register_li_request(sb, first_not_zeroed);
4881 }
4882
4883 ext4_setup_system_zone(sb);
4884 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4885 ext4_commit_super(sb, 1);
4886
4887 #ifdef CONFIG_QUOTA
4888 /* Release old quota file names */
4889 for (i = 0; i < MAXQUOTAS; i++)
4890 kfree(old_opts.s_qf_names[i]);
4891 if (enable_quota) {
4892 if (sb_any_quota_suspended(sb))
4893 dquot_resume(sb, -1);
4894 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4895 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4896 err = ext4_enable_quotas(sb);
4897 if (err)
4898 goto restore_opts;
4899 }
4900 }
4901 #endif
4902
4903 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4904 kfree(orig_data);
4905 return 0;
4906
4907 restore_opts:
4908 sb->s_flags = old_sb_flags;
4909 sbi->s_mount_opt = old_opts.s_mount_opt;
4910 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4911 sbi->s_resuid = old_opts.s_resuid;
4912 sbi->s_resgid = old_opts.s_resgid;
4913 sbi->s_commit_interval = old_opts.s_commit_interval;
4914 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4915 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4916 #ifdef CONFIG_QUOTA
4917 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4918 for (i = 0; i < MAXQUOTAS; i++) {
4919 kfree(sbi->s_qf_names[i]);
4920 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4921 }
4922 #endif
4923 kfree(orig_data);
4924 return err;
4925 }
4926
4927 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4928 {
4929 struct super_block *sb = dentry->d_sb;
4930 struct ext4_sb_info *sbi = EXT4_SB(sb);
4931 struct ext4_super_block *es = sbi->s_es;
4932 ext4_fsblk_t overhead = 0, resv_blocks;
4933 u64 fsid;
4934 s64 bfree;
4935 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
4936
4937 if (!test_opt(sb, MINIX_DF))
4938 overhead = sbi->s_overhead;
4939
4940 buf->f_type = EXT4_SUPER_MAGIC;
4941 buf->f_bsize = sb->s_blocksize;
4942 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4943 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4944 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4945 /* prevent underflow in case that few free space is available */
4946 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4947 buf->f_bavail = buf->f_bfree -
4948 (ext4_r_blocks_count(es) + resv_blocks);
4949 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
4950 buf->f_bavail = 0;
4951 buf->f_files = le32_to_cpu(es->s_inodes_count);
4952 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4953 buf->f_namelen = EXT4_NAME_LEN;
4954 fsid = le64_to_cpup((void *)es->s_uuid) ^
4955 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4956 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4957 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4958
4959 return 0;
4960 }
4961
4962 /* Helper function for writing quotas on sync - we need to start transaction
4963 * before quota file is locked for write. Otherwise the are possible deadlocks:
4964 * Process 1 Process 2
4965 * ext4_create() quota_sync()
4966 * jbd2_journal_start() write_dquot()
4967 * dquot_initialize() down(dqio_mutex)
4968 * down(dqio_mutex) jbd2_journal_start()
4969 *
4970 */
4971
4972 #ifdef CONFIG_QUOTA
4973
4974 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4975 {
4976 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4977 }
4978
4979 static int ext4_write_dquot(struct dquot *dquot)
4980 {
4981 int ret, err;
4982 handle_t *handle;
4983 struct inode *inode;
4984
4985 inode = dquot_to_inode(dquot);
4986 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
4987 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4988 if (IS_ERR(handle))
4989 return PTR_ERR(handle);
4990 ret = dquot_commit(dquot);
4991 err = ext4_journal_stop(handle);
4992 if (!ret)
4993 ret = err;
4994 return ret;
4995 }
4996
4997 static int ext4_acquire_dquot(struct dquot *dquot)
4998 {
4999 int ret, err;
5000 handle_t *handle;
5001
5002 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5003 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5004 if (IS_ERR(handle))
5005 return PTR_ERR(handle);
5006 ret = dquot_acquire(dquot);
5007 err = ext4_journal_stop(handle);
5008 if (!ret)
5009 ret = err;
5010 return ret;
5011 }
5012
5013 static int ext4_release_dquot(struct dquot *dquot)
5014 {
5015 int ret, err;
5016 handle_t *handle;
5017
5018 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5019 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5020 if (IS_ERR(handle)) {
5021 /* Release dquot anyway to avoid endless cycle in dqput() */
5022 dquot_release(dquot);
5023 return PTR_ERR(handle);
5024 }
5025 ret = dquot_release(dquot);
5026 err = ext4_journal_stop(handle);
5027 if (!ret)
5028 ret = err;
5029 return ret;
5030 }
5031
5032 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5033 {
5034 struct super_block *sb = dquot->dq_sb;
5035 struct ext4_sb_info *sbi = EXT4_SB(sb);
5036
5037 /* Are we journaling quotas? */
5038 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5039 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5040 dquot_mark_dquot_dirty(dquot);
5041 return ext4_write_dquot(dquot);
5042 } else {
5043 return dquot_mark_dquot_dirty(dquot);
5044 }
5045 }
5046
5047 static int ext4_write_info(struct super_block *sb, int type)
5048 {
5049 int ret, err;
5050 handle_t *handle;
5051
5052 /* Data block + inode block */
5053 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
5054 if (IS_ERR(handle))
5055 return PTR_ERR(handle);
5056 ret = dquot_commit_info(sb, type);
5057 err = ext4_journal_stop(handle);
5058 if (!ret)
5059 ret = err;
5060 return ret;
5061 }
5062
5063 /*
5064 * Turn on quotas during mount time - we need to find
5065 * the quota file and such...
5066 */
5067 static int ext4_quota_on_mount(struct super_block *sb, int type)
5068 {
5069 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5070 EXT4_SB(sb)->s_jquota_fmt, type);
5071 }
5072
5073 /*
5074 * Standard function to be called on quota_on
5075 */
5076 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5077 struct path *path)
5078 {
5079 int err;
5080
5081 if (!test_opt(sb, QUOTA))
5082 return -EINVAL;
5083
5084 /* Quotafile not on the same filesystem? */
5085 if (path->dentry->d_sb != sb)
5086 return -EXDEV;
5087 /* Journaling quota? */
5088 if (EXT4_SB(sb)->s_qf_names[type]) {
5089 /* Quotafile not in fs root? */
5090 if (path->dentry->d_parent != sb->s_root)
5091 ext4_msg(sb, KERN_WARNING,
5092 "Quota file not on filesystem root. "
5093 "Journaled quota will not work");
5094 }
5095
5096 /*
5097 * When we journal data on quota file, we have to flush journal to see
5098 * all updates to the file when we bypass pagecache...
5099 */
5100 if (EXT4_SB(sb)->s_journal &&
5101 ext4_should_journal_data(path->dentry->d_inode)) {
5102 /*
5103 * We don't need to lock updates but journal_flush() could
5104 * otherwise be livelocked...
5105 */
5106 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5107 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5108 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5109 if (err)
5110 return err;
5111 }
5112
5113 return dquot_quota_on(sb, type, format_id, path);
5114 }
5115
5116 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5117 unsigned int flags)
5118 {
5119 int err;
5120 struct inode *qf_inode;
5121 unsigned long qf_inums[MAXQUOTAS] = {
5122 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5123 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5124 };
5125
5126 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5127
5128 if (!qf_inums[type])
5129 return -EPERM;
5130
5131 qf_inode = ext4_iget(sb, qf_inums[type]);
5132 if (IS_ERR(qf_inode)) {
5133 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5134 return PTR_ERR(qf_inode);
5135 }
5136
5137 /* Don't account quota for quota files to avoid recursion */
5138 qf_inode->i_flags |= S_NOQUOTA;
5139 err = dquot_enable(qf_inode, type, format_id, flags);
5140 iput(qf_inode);
5141
5142 return err;
5143 }
5144
5145 /* Enable usage tracking for all quota types. */
5146 static int ext4_enable_quotas(struct super_block *sb)
5147 {
5148 int type, err = 0;
5149 unsigned long qf_inums[MAXQUOTAS] = {
5150 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5151 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5152 };
5153
5154 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5155 for (type = 0; type < MAXQUOTAS; type++) {
5156 if (qf_inums[type]) {
5157 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5158 DQUOT_USAGE_ENABLED);
5159 if (err) {
5160 ext4_warning(sb,
5161 "Failed to enable quota tracking "
5162 "(type=%d, err=%d). Please run "
5163 "e2fsck to fix.", type, err);
5164 return err;
5165 }
5166 }
5167 }
5168 return 0;
5169 }
5170
5171 /*
5172 * quota_on function that is used when QUOTA feature is set.
5173 */
5174 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5175 int format_id)
5176 {
5177 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5178 return -EINVAL;
5179
5180 /*
5181 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5182 */
5183 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5184 }
5185
5186 static int ext4_quota_off(struct super_block *sb, int type)
5187 {
5188 struct inode *inode = sb_dqopt(sb)->files[type];
5189 handle_t *handle;
5190
5191 /* Force all delayed allocation blocks to be allocated.
5192 * Caller already holds s_umount sem */
5193 if (test_opt(sb, DELALLOC))
5194 sync_filesystem(sb);
5195
5196 if (!inode)
5197 goto out;
5198
5199 /* Update modification times of quota files when userspace can
5200 * start looking at them */
5201 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5202 if (IS_ERR(handle))
5203 goto out;
5204 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5205 ext4_mark_inode_dirty(handle, inode);
5206 ext4_journal_stop(handle);
5207
5208 out:
5209 return dquot_quota_off(sb, type);
5210 }
5211
5212 /*
5213 * quota_off function that is used when QUOTA feature is set.
5214 */
5215 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5216 {
5217 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5218 return -EINVAL;
5219
5220 /* Disable only the limits. */
5221 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5222 }
5223
5224 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5225 * acquiring the locks... As quota files are never truncated and quota code
5226 * itself serializes the operations (and no one else should touch the files)
5227 * we don't have to be afraid of races */
5228 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5229 size_t len, loff_t off)
5230 {
5231 struct inode *inode = sb_dqopt(sb)->files[type];
5232 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5233 int err = 0;
5234 int offset = off & (sb->s_blocksize - 1);
5235 int tocopy;
5236 size_t toread;
5237 struct buffer_head *bh;
5238 loff_t i_size = i_size_read(inode);
5239
5240 if (off > i_size)
5241 return 0;
5242 if (off+len > i_size)
5243 len = i_size-off;
5244 toread = len;
5245 while (toread > 0) {
5246 tocopy = sb->s_blocksize - offset < toread ?
5247 sb->s_blocksize - offset : toread;
5248 bh = ext4_bread(NULL, inode, blk, 0, &err);
5249 if (err)
5250 return err;
5251 if (!bh) /* A hole? */
5252 memset(data, 0, tocopy);
5253 else
5254 memcpy(data, bh->b_data+offset, tocopy);
5255 brelse(bh);
5256 offset = 0;
5257 toread -= tocopy;
5258 data += tocopy;
5259 blk++;
5260 }
5261 return len;
5262 }
5263
5264 /* Write to quotafile (we know the transaction is already started and has
5265 * enough credits) */
5266 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5267 const char *data, size_t len, loff_t off)
5268 {
5269 struct inode *inode = sb_dqopt(sb)->files[type];
5270 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5271 int err = 0;
5272 int offset = off & (sb->s_blocksize - 1);
5273 struct buffer_head *bh;
5274 handle_t *handle = journal_current_handle();
5275
5276 if (EXT4_SB(sb)->s_journal && !handle) {
5277 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5278 " cancelled because transaction is not started",
5279 (unsigned long long)off, (unsigned long long)len);
5280 return -EIO;
5281 }
5282 /*
5283 * Since we account only one data block in transaction credits,
5284 * then it is impossible to cross a block boundary.
5285 */
5286 if (sb->s_blocksize - offset < len) {
5287 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5288 " cancelled because not block aligned",
5289 (unsigned long long)off, (unsigned long long)len);
5290 return -EIO;
5291 }
5292
5293 bh = ext4_bread(handle, inode, blk, 1, &err);
5294 if (!bh)
5295 goto out;
5296 err = ext4_journal_get_write_access(handle, bh);
5297 if (err) {
5298 brelse(bh);
5299 goto out;
5300 }
5301 lock_buffer(bh);
5302 memcpy(bh->b_data+offset, data, len);
5303 flush_dcache_page(bh->b_page);
5304 unlock_buffer(bh);
5305 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5306 brelse(bh);
5307 out:
5308 if (err)
5309 return err;
5310 if (inode->i_size < off + len) {
5311 i_size_write(inode, off + len);
5312 EXT4_I(inode)->i_disksize = inode->i_size;
5313 ext4_mark_inode_dirty(handle, inode);
5314 }
5315 return len;
5316 }
5317
5318 #endif
5319
5320 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5321 const char *dev_name, void *data)
5322 {
5323 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5324 }
5325
5326 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5327 static inline void register_as_ext2(void)
5328 {
5329 int err = register_filesystem(&ext2_fs_type);
5330 if (err)
5331 printk(KERN_WARNING
5332 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5333 }
5334
5335 static inline void unregister_as_ext2(void)
5336 {
5337 unregister_filesystem(&ext2_fs_type);
5338 }
5339
5340 static inline int ext2_feature_set_ok(struct super_block *sb)
5341 {
5342 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5343 return 0;
5344 if (sb->s_flags & MS_RDONLY)
5345 return 1;
5346 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5347 return 0;
5348 return 1;
5349 }
5350 #else
5351 static inline void register_as_ext2(void) { }
5352 static inline void unregister_as_ext2(void) { }
5353 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5354 #endif
5355
5356 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5357 static inline void register_as_ext3(void)
5358 {
5359 int err = register_filesystem(&ext3_fs_type);
5360 if (err)
5361 printk(KERN_WARNING
5362 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5363 }
5364
5365 static inline void unregister_as_ext3(void)
5366 {
5367 unregister_filesystem(&ext3_fs_type);
5368 }
5369
5370 static inline int ext3_feature_set_ok(struct super_block *sb)
5371 {
5372 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5373 return 0;
5374 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5375 return 0;
5376 if (sb->s_flags & MS_RDONLY)
5377 return 1;
5378 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5379 return 0;
5380 return 1;
5381 }
5382 #else
5383 static inline void register_as_ext3(void) { }
5384 static inline void unregister_as_ext3(void) { }
5385 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5386 #endif
5387
5388 static struct file_system_type ext4_fs_type = {
5389 .owner = THIS_MODULE,
5390 .name = "ext4",
5391 .mount = ext4_mount,
5392 .kill_sb = kill_block_super,
5393 .fs_flags = FS_REQUIRES_DEV,
5394 };
5395 MODULE_ALIAS_FS("ext4");
5396
5397 static int __init ext4_init_feat_adverts(void)
5398 {
5399 struct ext4_features *ef;
5400 int ret = -ENOMEM;
5401
5402 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5403 if (!ef)
5404 goto out;
5405
5406 ef->f_kobj.kset = ext4_kset;
5407 init_completion(&ef->f_kobj_unregister);
5408 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5409 "features");
5410 if (ret) {
5411 kfree(ef);
5412 goto out;
5413 }
5414
5415 ext4_feat = ef;
5416 ret = 0;
5417 out:
5418 return ret;
5419 }
5420
5421 static void ext4_exit_feat_adverts(void)
5422 {
5423 kobject_put(&ext4_feat->f_kobj);
5424 wait_for_completion(&ext4_feat->f_kobj_unregister);
5425 kfree(ext4_feat);
5426 }
5427
5428 /* Shared across all ext4 file systems */
5429 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5430 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5431
5432 static int __init ext4_init_fs(void)
5433 {
5434 int i, err;
5435
5436 ext4_li_info = NULL;
5437 mutex_init(&ext4_li_mtx);
5438
5439 /* Build-time check for flags consistency */
5440 ext4_check_flag_values();
5441
5442 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5443 mutex_init(&ext4__aio_mutex[i]);
5444 init_waitqueue_head(&ext4__ioend_wq[i]);
5445 }
5446
5447 err = ext4_init_es();
5448 if (err)
5449 return err;
5450
5451 err = ext4_init_pageio();
5452 if (err)
5453 goto out7;
5454
5455 err = ext4_init_system_zone();
5456 if (err)
5457 goto out6;
5458 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5459 if (!ext4_kset) {
5460 err = -ENOMEM;
5461 goto out5;
5462 }
5463 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5464
5465 err = ext4_init_feat_adverts();
5466 if (err)
5467 goto out4;
5468
5469 err = ext4_init_mballoc();
5470 if (err)
5471 goto out3;
5472
5473 err = ext4_init_xattr();
5474 if (err)
5475 goto out2;
5476 err = init_inodecache();
5477 if (err)
5478 goto out1;
5479 register_as_ext3();
5480 register_as_ext2();
5481 err = register_filesystem(&ext4_fs_type);
5482 if (err)
5483 goto out;
5484
5485 return 0;
5486 out:
5487 unregister_as_ext2();
5488 unregister_as_ext3();
5489 destroy_inodecache();
5490 out1:
5491 ext4_exit_xattr();
5492 out2:
5493 ext4_exit_mballoc();
5494 out3:
5495 ext4_exit_feat_adverts();
5496 out4:
5497 if (ext4_proc_root)
5498 remove_proc_entry("fs/ext4", NULL);
5499 kset_unregister(ext4_kset);
5500 out5:
5501 ext4_exit_system_zone();
5502 out6:
5503 ext4_exit_pageio();
5504 out7:
5505 ext4_exit_es();
5506
5507 return err;
5508 }
5509
5510 static void __exit ext4_exit_fs(void)
5511 {
5512 ext4_destroy_lazyinit_thread();
5513 unregister_as_ext2();
5514 unregister_as_ext3();
5515 unregister_filesystem(&ext4_fs_type);
5516 destroy_inodecache();
5517 ext4_exit_xattr();
5518 ext4_exit_mballoc();
5519 ext4_exit_feat_adverts();
5520 remove_proc_entry("fs/ext4", NULL);
5521 kset_unregister(ext4_kset);
5522 ext4_exit_system_zone();
5523 ext4_exit_pageio();
5524 ext4_exit_es();
5525 }
5526
5527 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5528 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5529 MODULE_LICENSE("GPL");
5530 module_init(ext4_init_fs)
5531 module_exit(ext4_exit_fs)
Linus' kernel tree