]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - fs/ext4/super.c
projects / mirror_ubuntu-zesty-kernel.git / blob
? search:


c014edd12648b9c1573d10a27cce82434581251f
[mirror_ubuntu-zesty-kernel.git] / fs / ext4 / super.c
1 /*
2 * linux/fs/ext4/super.c
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
43
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
46
47 #include "ext4.h"
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
50 #include "xattr.h"
51 #include "acl.h"
52 #include "mballoc.h"
53
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
56
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
62
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 const char *ext4_decode_error(struct super_block *sb, int errno,
73 char nbuf[16]);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86
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 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96 #else
97 #define IS_EXT2_SB(sb) (0)
98 #endif
99
100
101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102 static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
104 .name = "ext3",
105 .mount = ext4_mount,
106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
108 };
109 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110 #else
111 #define IS_EXT3_SB(sb) (0)
112 #endif
113
114 static int ext4_verify_csum_type(struct super_block *sb,
115 struct ext4_super_block *es)
116 {
117 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
118 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
119 return 1;
120
121 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
122 }
123
124 static __le32 ext4_superblock_csum(struct super_block *sb,
125 struct ext4_super_block *es)
126 {
127 struct ext4_sb_info *sbi = EXT4_SB(sb);
128 int offset = offsetof(struct ext4_super_block, s_checksum);
129 __u32 csum;
130
131 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
132
133 return cpu_to_le32(csum);
134 }
135
136 int ext4_superblock_csum_verify(struct super_block *sb,
137 struct ext4_super_block *es)
138 {
139 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
140 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
141 return 1;
142
143 return es->s_checksum == ext4_superblock_csum(sb, es);
144 }
145
146 void ext4_superblock_csum_set(struct super_block *sb)
147 {
148 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
149
150 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
151 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
152 return;
153
154 es->s_checksum = ext4_superblock_csum(sb, es);
155 }
156
157 void *ext4_kvmalloc(size_t size, gfp_t flags)
158 {
159 void *ret;
160
161 ret = kmalloc(size, flags);
162 if (!ret)
163 ret = __vmalloc(size, flags, PAGE_KERNEL);
164 return ret;
165 }
166
167 void *ext4_kvzalloc(size_t size, gfp_t flags)
168 {
169 void *ret;
170
171 ret = kzalloc(size, flags);
172 if (!ret)
173 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
174 return ret;
175 }
176
177 void ext4_kvfree(void *ptr)
178 {
179 if (is_vmalloc_addr(ptr))
180 vfree(ptr);
181 else
182 kfree(ptr);
183
184 }
185
186 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
188 {
189 return le32_to_cpu(bg->bg_block_bitmap_lo) |
190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
192 }
193
194 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
195 struct ext4_group_desc *bg)
196 {
197 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
200 }
201
202 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
203 struct ext4_group_desc *bg)
204 {
205 return le32_to_cpu(bg->bg_inode_table_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
208 }
209
210 __u32 ext4_free_group_clusters(struct super_block *sb,
211 struct ext4_group_desc *bg)
212 {
213 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
216 }
217
218 __u32 ext4_free_inodes_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220 {
221 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
224 }
225
226 __u32 ext4_used_dirs_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
228 {
229 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
231 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
232 }
233
234 __u32 ext4_itable_unused_count(struct super_block *sb,
235 struct ext4_group_desc *bg)
236 {
237 return le16_to_cpu(bg->bg_itable_unused_lo) |
238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
239 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
240 }
241
242 void ext4_block_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
244 {
245 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
248 }
249
250 void ext4_inode_bitmap_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
252 {
253 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
256 }
257
258 void ext4_inode_table_set(struct super_block *sb,
259 struct ext4_group_desc *bg, ext4_fsblk_t blk)
260 {
261 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
264 }
265
266 void ext4_free_group_clusters_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
268 {
269 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
272 }
273
274 void ext4_free_inodes_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276 {
277 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
280 }
281
282 void ext4_used_dirs_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284 {
285 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
288 }
289
290 void ext4_itable_unused_set(struct super_block *sb,
291 struct ext4_group_desc *bg, __u32 count)
292 {
293 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
296 }
297
298
299 /* Just increment the non-pointer handle value */
300 static handle_t *ext4_get_nojournal(void)
301 {
302 handle_t *handle = current->journal_info;
303 unsigned long ref_cnt = (unsigned long)handle;
304
305 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
306
307 ref_cnt++;
308 handle = (handle_t *)ref_cnt;
309
310 current->journal_info = handle;
311 return handle;
312 }
313
314
315 /* Decrement the non-pointer handle value */
316 static void ext4_put_nojournal(handle_t *handle)
317 {
318 unsigned long ref_cnt = (unsigned long)handle;
319
320 BUG_ON(ref_cnt == 0);
321
322 ref_cnt--;
323 handle = (handle_t *)ref_cnt;
324
325 current->journal_info = handle;
326 }
327
328 /*
329 * Wrappers for jbd2_journal_start/end.
330 */
331 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
332 {
333 journal_t *journal;
334
335 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
336 if (sb->s_flags & MS_RDONLY)
337 return ERR_PTR(-EROFS);
338
339 WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE);
340 journal = EXT4_SB(sb)->s_journal;
341 if (!journal)
342 return ext4_get_nojournal();
343 /*
344 * Special case here: if the journal has aborted behind our
345 * backs (eg. EIO in the commit thread), then we still need to
346 * take the FS itself readonly cleanly.
347 */
348 if (is_journal_aborted(journal)) {
349 ext4_abort(sb, "Detected aborted journal");
350 return ERR_PTR(-EROFS);
351 }
352 return jbd2_journal_start(journal, nblocks);
353 }
354
355 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
356 {
357 struct super_block *sb;
358 int err;
359 int rc;
360
361 if (!ext4_handle_valid(handle)) {
362 ext4_put_nojournal(handle);
363 return 0;
364 }
365 sb = handle->h_transaction->t_journal->j_private;
366 err = handle->h_err;
367 rc = jbd2_journal_stop(handle);
368
369 if (!err)
370 err = rc;
371 if (err)
372 __ext4_std_error(sb, where, line, err);
373 return err;
374 }
375
376 void ext4_journal_abort_handle(const char *caller, unsigned int line,
377 const char *err_fn, struct buffer_head *bh,
378 handle_t *handle, int err)
379 {
380 char nbuf[16];
381 const char *errstr = ext4_decode_error(NULL, err, nbuf);
382
383 BUG_ON(!ext4_handle_valid(handle));
384
385 if (bh)
386 BUFFER_TRACE(bh, "abort");
387
388 if (!handle->h_err)
389 handle->h_err = err;
390
391 if (is_handle_aborted(handle))
392 return;
393
394 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
395 caller, line, errstr, err_fn);
396
397 jbd2_journal_abort_handle(handle);
398 }
399
400 static void __save_error_info(struct super_block *sb, const char *func,
401 unsigned int line)
402 {
403 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
404
405 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
406 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
407 es->s_last_error_time = cpu_to_le32(get_seconds());
408 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
409 es->s_last_error_line = cpu_to_le32(line);
410 if (!es->s_first_error_time) {
411 es->s_first_error_time = es->s_last_error_time;
412 strncpy(es->s_first_error_func, func,
413 sizeof(es->s_first_error_func));
414 es->s_first_error_line = cpu_to_le32(line);
415 es->s_first_error_ino = es->s_last_error_ino;
416 es->s_first_error_block = es->s_last_error_block;
417 }
418 /*
419 * Start the daily error reporting function if it hasn't been
420 * started already
421 */
422 if (!es->s_error_count)
423 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
424 le32_add_cpu(&es->s_error_count, 1);
425 }
426
427 static void save_error_info(struct super_block *sb, const char *func,
428 unsigned int line)
429 {
430 __save_error_info(sb, func, line);
431 ext4_commit_super(sb, 1);
432 }
433
434 /*
435 * The del_gendisk() function uninitializes the disk-specific data
436 * structures, including the bdi structure, without telling anyone
437 * else. Once this happens, any attempt to call mark_buffer_dirty()
438 * (for example, by ext4_commit_super), will cause a kernel OOPS.
439 * This is a kludge to prevent these oops until we can put in a proper
440 * hook in del_gendisk() to inform the VFS and file system layers.
441 */
442 static int block_device_ejected(struct super_block *sb)
443 {
444 struct inode *bd_inode = sb->s_bdev->bd_inode;
445 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
446
447 return bdi->dev == NULL;
448 }
449
450 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
451 {
452 struct super_block *sb = journal->j_private;
453 struct ext4_sb_info *sbi = EXT4_SB(sb);
454 int error = is_journal_aborted(journal);
455 struct ext4_journal_cb_entry *jce, *tmp;
456
457 spin_lock(&sbi->s_md_lock);
458 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
459 list_del_init(&jce->jce_list);
460 spin_unlock(&sbi->s_md_lock);
461 jce->jce_func(sb, jce, error);
462 spin_lock(&sbi->s_md_lock);
463 }
464 spin_unlock(&sbi->s_md_lock);
465 }
466
467 /* Deal with the reporting of failure conditions on a filesystem such as
468 * inconsistencies detected or read IO failures.
469 *
470 * On ext2, we can store the error state of the filesystem in the
471 * superblock. That is not possible on ext4, because we may have other
472 * write ordering constraints on the superblock which prevent us from
473 * writing it out straight away; and given that the journal is about to
474 * be aborted, we can't rely on the current, or future, transactions to
475 * write out the superblock safely.
476 *
477 * We'll just use the jbd2_journal_abort() error code to record an error in
478 * the journal instead. On recovery, the journal will complain about
479 * that error until we've noted it down and cleared it.
480 */
481
482 static void ext4_handle_error(struct super_block *sb)
483 {
484 if (sb->s_flags & MS_RDONLY)
485 return;
486
487 if (!test_opt(sb, ERRORS_CONT)) {
488 journal_t *journal = EXT4_SB(sb)->s_journal;
489
490 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
491 if (journal)
492 jbd2_journal_abort(journal, -EIO);
493 }
494 if (test_opt(sb, ERRORS_RO)) {
495 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
496 sb->s_flags |= MS_RDONLY;
497 }
498 if (test_opt(sb, ERRORS_PANIC))
499 panic("EXT4-fs (device %s): panic forced after error\n",
500 sb->s_id);
501 }
502
503 void __ext4_error(struct super_block *sb, const char *function,
504 unsigned int line, const char *fmt, ...)
505 {
506 struct va_format vaf;
507 va_list args;
508
509 va_start(args, fmt);
510 vaf.fmt = fmt;
511 vaf.va = &args;
512 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
513 sb->s_id, function, line, current->comm, &vaf);
514 va_end(args);
515 save_error_info(sb, function, line);
516
517 ext4_handle_error(sb);
518 }
519
520 void ext4_error_inode(struct inode *inode, const char *function,
521 unsigned int line, ext4_fsblk_t block,
522 const char *fmt, ...)
523 {
524 va_list args;
525 struct va_format vaf;
526 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
527
528 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
529 es->s_last_error_block = cpu_to_le64(block);
530 save_error_info(inode->i_sb, function, line);
531 va_start(args, fmt);
532 vaf.fmt = fmt;
533 vaf.va = &args;
534 if (block)
535 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
536 "inode #%lu: block %llu: comm %s: %pV\n",
537 inode->i_sb->s_id, function, line, inode->i_ino,
538 block, current->comm, &vaf);
539 else
540 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
541 "inode #%lu: comm %s: %pV\n",
542 inode->i_sb->s_id, function, line, inode->i_ino,
543 current->comm, &vaf);
544 va_end(args);
545
546 ext4_handle_error(inode->i_sb);
547 }
548
549 void ext4_error_file(struct file *file, const char *function,
550 unsigned int line, ext4_fsblk_t block,
551 const char *fmt, ...)
552 {
553 va_list args;
554 struct va_format vaf;
555 struct ext4_super_block *es;
556 struct inode *inode = file->f_dentry->d_inode;
557 char pathname[80], *path;
558
559 es = EXT4_SB(inode->i_sb)->s_es;
560 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
561 save_error_info(inode->i_sb, function, line);
562 path = d_path(&(file->f_path), pathname, sizeof(pathname));
563 if (IS_ERR(path))
564 path = "(unknown)";
565 va_start(args, fmt);
566 vaf.fmt = fmt;
567 vaf.va = &args;
568 if (block)
569 printk(KERN_CRIT
570 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
571 "block %llu: comm %s: path %s: %pV\n",
572 inode->i_sb->s_id, function, line, inode->i_ino,
573 block, current->comm, path, &vaf);
574 else
575 printk(KERN_CRIT
576 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
577 "comm %s: path %s: %pV\n",
578 inode->i_sb->s_id, function, line, inode->i_ino,
579 current->comm, path, &vaf);
580 va_end(args);
581
582 ext4_handle_error(inode->i_sb);
583 }
584
585 static const char *ext4_decode_error(struct super_block *sb, int errno,
586 char nbuf[16])
587 {
588 char *errstr = NULL;
589
590 switch (errno) {
591 case -EIO:
592 errstr = "IO failure";
593 break;
594 case -ENOMEM:
595 errstr = "Out of memory";
596 break;
597 case -EROFS:
598 if (!sb || (EXT4_SB(sb)->s_journal &&
599 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
600 errstr = "Journal has aborted";
601 else
602 errstr = "Readonly filesystem";
603 break;
604 default:
605 /* If the caller passed in an extra buffer for unknown
606 * errors, textualise them now. Else we just return
607 * NULL. */
608 if (nbuf) {
609 /* Check for truncated error codes... */
610 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
611 errstr = nbuf;
612 }
613 break;
614 }
615
616 return errstr;
617 }
618
619 /* __ext4_std_error decodes expected errors from journaling functions
620 * automatically and invokes the appropriate error response. */
621
622 void __ext4_std_error(struct super_block *sb, const char *function,
623 unsigned int line, int errno)
624 {
625 char nbuf[16];
626 const char *errstr;
627
628 /* Special case: if the error is EROFS, and we're not already
629 * inside a transaction, then there's really no point in logging
630 * an error. */
631 if (errno == -EROFS && journal_current_handle() == NULL &&
632 (sb->s_flags & MS_RDONLY))
633 return;
634
635 errstr = ext4_decode_error(sb, errno, nbuf);
636 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
637 sb->s_id, function, line, errstr);
638 save_error_info(sb, function, line);
639
640 ext4_handle_error(sb);
641 }
642
643 /*
644 * ext4_abort is a much stronger failure handler than ext4_error. The
645 * abort function may be used to deal with unrecoverable failures such
646 * as journal IO errors or ENOMEM at a critical moment in log management.
647 *
648 * We unconditionally force the filesystem into an ABORT|READONLY state,
649 * unless the error response on the fs has been set to panic in which
650 * case we take the easy way out and panic immediately.
651 */
652
653 void __ext4_abort(struct super_block *sb, const char *function,
654 unsigned int line, const char *fmt, ...)
655 {
656 va_list args;
657
658 save_error_info(sb, function, line);
659 va_start(args, fmt);
660 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
661 function, line);
662 vprintk(fmt, args);
663 printk("\n");
664 va_end(args);
665
666 if ((sb->s_flags & MS_RDONLY) == 0) {
667 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
668 sb->s_flags |= MS_RDONLY;
669 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
670 if (EXT4_SB(sb)->s_journal)
671 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
672 save_error_info(sb, function, line);
673 }
674 if (test_opt(sb, ERRORS_PANIC))
675 panic("EXT4-fs panic from previous error\n");
676 }
677
678 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
679 {
680 struct va_format vaf;
681 va_list args;
682
683 va_start(args, fmt);
684 vaf.fmt = fmt;
685 vaf.va = &args;
686 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
687 va_end(args);
688 }
689
690 void __ext4_warning(struct super_block *sb, const char *function,
691 unsigned int line, const char *fmt, ...)
692 {
693 struct va_format vaf;
694 va_list args;
695
696 va_start(args, fmt);
697 vaf.fmt = fmt;
698 vaf.va = &args;
699 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
700 sb->s_id, function, line, &vaf);
701 va_end(args);
702 }
703
704 void __ext4_grp_locked_error(const char *function, unsigned int line,
705 struct super_block *sb, ext4_group_t grp,
706 unsigned long ino, ext4_fsblk_t block,
707 const char *fmt, ...)
708 __releases(bitlock)
709 __acquires(bitlock)
710 {
711 struct va_format vaf;
712 va_list args;
713 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
714
715 es->s_last_error_ino = cpu_to_le32(ino);
716 es->s_last_error_block = cpu_to_le64(block);
717 __save_error_info(sb, function, line);
718
719 va_start(args, fmt);
720
721 vaf.fmt = fmt;
722 vaf.va = &args;
723 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
724 sb->s_id, function, line, grp);
725 if (ino)
726 printk(KERN_CONT "inode %lu: ", ino);
727 if (block)
728 printk(KERN_CONT "block %llu:", (unsigned long long) block);
729 printk(KERN_CONT "%pV\n", &vaf);
730 va_end(args);
731
732 if (test_opt(sb, ERRORS_CONT)) {
733 ext4_commit_super(sb, 0);
734 return;
735 }
736
737 ext4_unlock_group(sb, grp);
738 ext4_handle_error(sb);
739 /*
740 * We only get here in the ERRORS_RO case; relocking the group
741 * may be dangerous, but nothing bad will happen since the
742 * filesystem will have already been marked read/only and the
743 * journal has been aborted. We return 1 as a hint to callers
744 * who might what to use the return value from
745 * ext4_grp_locked_error() to distinguish between the
746 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
747 * aggressively from the ext4 function in question, with a
748 * more appropriate error code.
749 */
750 ext4_lock_group(sb, grp);
751 return;
752 }
753
754 void ext4_update_dynamic_rev(struct super_block *sb)
755 {
756 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
757
758 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
759 return;
760
761 ext4_warning(sb,
762 "updating to rev %d because of new feature flag, "
763 "running e2fsck is recommended",
764 EXT4_DYNAMIC_REV);
765
766 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
767 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
768 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
769 /* leave es->s_feature_*compat flags alone */
770 /* es->s_uuid will be set by e2fsck if empty */
771
772 /*
773 * The rest of the superblock fields should be zero, and if not it
774 * means they are likely already in use, so leave them alone. We
775 * can leave it up to e2fsck to clean up any inconsistencies there.
776 */
777 }
778
779 /*
780 * Open the external journal device
781 */
782 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
783 {
784 struct block_device *bdev;
785 char b[BDEVNAME_SIZE];
786
787 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
788 if (IS_ERR(bdev))
789 goto fail;
790 return bdev;
791
792 fail:
793 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
794 __bdevname(dev, b), PTR_ERR(bdev));
795 return NULL;
796 }
797
798 /*
799 * Release the journal device
800 */
801 static int ext4_blkdev_put(struct block_device *bdev)
802 {
803 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
804 }
805
806 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
807 {
808 struct block_device *bdev;
809 int ret = -ENODEV;
810
811 bdev = sbi->journal_bdev;
812 if (bdev) {
813 ret = ext4_blkdev_put(bdev);
814 sbi->journal_bdev = NULL;
815 }
816 return ret;
817 }
818
819 static inline struct inode *orphan_list_entry(struct list_head *l)
820 {
821 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
822 }
823
824 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
825 {
826 struct list_head *l;
827
828 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
829 le32_to_cpu(sbi->s_es->s_last_orphan));
830
831 printk(KERN_ERR "sb_info orphan list:\n");
832 list_for_each(l, &sbi->s_orphan) {
833 struct inode *inode = orphan_list_entry(l);
834 printk(KERN_ERR " "
835 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
836 inode->i_sb->s_id, inode->i_ino, inode,
837 inode->i_mode, inode->i_nlink,
838 NEXT_ORPHAN(inode));
839 }
840 }
841
842 static void ext4_put_super(struct super_block *sb)
843 {
844 struct ext4_sb_info *sbi = EXT4_SB(sb);
845 struct ext4_super_block *es = sbi->s_es;
846 int i, err;
847
848 ext4_unregister_li_request(sb);
849 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
850
851 flush_workqueue(sbi->dio_unwritten_wq);
852 destroy_workqueue(sbi->dio_unwritten_wq);
853
854 if (sbi->s_journal) {
855 err = jbd2_journal_destroy(sbi->s_journal);
856 sbi->s_journal = NULL;
857 if (err < 0)
858 ext4_abort(sb, "Couldn't clean up the journal");
859 }
860
861 del_timer(&sbi->s_err_report);
862 ext4_release_system_zone(sb);
863 ext4_mb_release(sb);
864 ext4_ext_release(sb);
865 ext4_xattr_put_super(sb);
866
867 if (!(sb->s_flags & MS_RDONLY)) {
868 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
869 es->s_state = cpu_to_le16(sbi->s_mount_state);
870 }
871 if (!(sb->s_flags & MS_RDONLY))
872 ext4_commit_super(sb, 1);
873
874 if (sbi->s_proc) {
875 remove_proc_entry("options", sbi->s_proc);
876 remove_proc_entry(sb->s_id, ext4_proc_root);
877 }
878 kobject_del(&sbi->s_kobj);
879
880 for (i = 0; i < sbi->s_gdb_count; i++)
881 brelse(sbi->s_group_desc[i]);
882 ext4_kvfree(sbi->s_group_desc);
883 ext4_kvfree(sbi->s_flex_groups);
884 percpu_counter_destroy(&sbi->s_freeclusters_counter);
885 percpu_counter_destroy(&sbi->s_freeinodes_counter);
886 percpu_counter_destroy(&sbi->s_dirs_counter);
887 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
888 brelse(sbi->s_sbh);
889 #ifdef CONFIG_QUOTA
890 for (i = 0; i < MAXQUOTAS; i++)
891 kfree(sbi->s_qf_names[i]);
892 #endif
893
894 /* Debugging code just in case the in-memory inode orphan list
895 * isn't empty. The on-disk one can be non-empty if we've
896 * detected an error and taken the fs readonly, but the
897 * in-memory list had better be clean by this point. */
898 if (!list_empty(&sbi->s_orphan))
899 dump_orphan_list(sb, sbi);
900 J_ASSERT(list_empty(&sbi->s_orphan));
901
902 invalidate_bdev(sb->s_bdev);
903 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
904 /*
905 * Invalidate the journal device's buffers. We don't want them
906 * floating about in memory - the physical journal device may
907 * hotswapped, and it breaks the `ro-after' testing code.
908 */
909 sync_blockdev(sbi->journal_bdev);
910 invalidate_bdev(sbi->journal_bdev);
911 ext4_blkdev_remove(sbi);
912 }
913 if (sbi->s_mmp_tsk)
914 kthread_stop(sbi->s_mmp_tsk);
915 sb->s_fs_info = NULL;
916 /*
917 * Now that we are completely done shutting down the
918 * superblock, we need to actually destroy the kobject.
919 */
920 kobject_put(&sbi->s_kobj);
921 wait_for_completion(&sbi->s_kobj_unregister);
922 if (sbi->s_chksum_driver)
923 crypto_free_shash(sbi->s_chksum_driver);
924 kfree(sbi->s_blockgroup_lock);
925 kfree(sbi);
926 }
927
928 static struct kmem_cache *ext4_inode_cachep;
929
930 /*
931 * Called inside transaction, so use GFP_NOFS
932 */
933 static struct inode *ext4_alloc_inode(struct super_block *sb)
934 {
935 struct ext4_inode_info *ei;
936
937 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
938 if (!ei)
939 return NULL;
940
941 ei->vfs_inode.i_version = 1;
942 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
943 INIT_LIST_HEAD(&ei->i_prealloc_list);
944 spin_lock_init(&ei->i_prealloc_lock);
945 ext4_es_init_tree(&ei->i_es_tree);
946 rwlock_init(&ei->i_es_lock);
947 ei->i_reserved_data_blocks = 0;
948 ei->i_reserved_meta_blocks = 0;
949 ei->i_allocated_meta_blocks = 0;
950 ei->i_da_metadata_calc_len = 0;
951 ei->i_da_metadata_calc_last_lblock = 0;
952 spin_lock_init(&(ei->i_block_reservation_lock));
953 #ifdef CONFIG_QUOTA
954 ei->i_reserved_quota = 0;
955 #endif
956 ei->jinode = NULL;
957 INIT_LIST_HEAD(&ei->i_completed_io_list);
958 spin_lock_init(&ei->i_completed_io_lock);
959 ei->i_sync_tid = 0;
960 ei->i_datasync_tid = 0;
961 atomic_set(&ei->i_ioend_count, 0);
962 atomic_set(&ei->i_unwritten, 0);
963
964 return &ei->vfs_inode;
965 }
966
967 static int ext4_drop_inode(struct inode *inode)
968 {
969 int drop = generic_drop_inode(inode);
970
971 trace_ext4_drop_inode(inode, drop);
972 return drop;
973 }
974
975 static void ext4_i_callback(struct rcu_head *head)
976 {
977 struct inode *inode = container_of(head, struct inode, i_rcu);
978 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
979 }
980
981 static void ext4_destroy_inode(struct inode *inode)
982 {
983 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
984 ext4_msg(inode->i_sb, KERN_ERR,
985 "Inode %lu (%p): orphan list check failed!",
986 inode->i_ino, EXT4_I(inode));
987 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
988 EXT4_I(inode), sizeof(struct ext4_inode_info),
989 true);
990 dump_stack();
991 }
992 call_rcu(&inode->i_rcu, ext4_i_callback);
993 }
994
995 static void init_once(void *foo)
996 {
997 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
998
999 INIT_LIST_HEAD(&ei->i_orphan);
1000 init_rwsem(&ei->xattr_sem);
1001 init_rwsem(&ei->i_data_sem);
1002 inode_init_once(&ei->vfs_inode);
1003 }
1004
1005 static int init_inodecache(void)
1006 {
1007 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1008 sizeof(struct ext4_inode_info),
1009 0, (SLAB_RECLAIM_ACCOUNT|
1010 SLAB_MEM_SPREAD),
1011 init_once);
1012 if (ext4_inode_cachep == NULL)
1013 return -ENOMEM;
1014 return 0;
1015 }
1016
1017 static void destroy_inodecache(void)
1018 {
1019 /*
1020 * Make sure all delayed rcu free inodes are flushed before we
1021 * destroy cache.
1022 */
1023 rcu_barrier();
1024 kmem_cache_destroy(ext4_inode_cachep);
1025 }
1026
1027 void ext4_clear_inode(struct inode *inode)
1028 {
1029 invalidate_inode_buffers(inode);
1030 clear_inode(inode);
1031 dquot_drop(inode);
1032 ext4_discard_preallocations(inode);
1033 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1034 if (EXT4_I(inode)->jinode) {
1035 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1036 EXT4_I(inode)->jinode);
1037 jbd2_free_inode(EXT4_I(inode)->jinode);
1038 EXT4_I(inode)->jinode = NULL;
1039 }
1040 }
1041
1042 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1043 u64 ino, u32 generation)
1044 {
1045 struct inode *inode;
1046
1047 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1048 return ERR_PTR(-ESTALE);
1049 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1050 return ERR_PTR(-ESTALE);
1051
1052 /* iget isn't really right if the inode is currently unallocated!!
1053 *
1054 * ext4_read_inode will return a bad_inode if the inode had been
1055 * deleted, so we should be safe.
1056 *
1057 * Currently we don't know the generation for parent directory, so
1058 * a generation of 0 means "accept any"
1059 */
1060 inode = ext4_iget(sb, ino);
1061 if (IS_ERR(inode))
1062 return ERR_CAST(inode);
1063 if (generation && inode->i_generation != generation) {
1064 iput(inode);
1065 return ERR_PTR(-ESTALE);
1066 }
1067
1068 return inode;
1069 }
1070
1071 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1072 int fh_len, int fh_type)
1073 {
1074 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1075 ext4_nfs_get_inode);
1076 }
1077
1078 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1079 int fh_len, int fh_type)
1080 {
1081 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1082 ext4_nfs_get_inode);
1083 }
1084
1085 /*
1086 * Try to release metadata pages (indirect blocks, directories) which are
1087 * mapped via the block device. Since these pages could have journal heads
1088 * which would prevent try_to_free_buffers() from freeing them, we must use
1089 * jbd2 layer's try_to_free_buffers() function to release them.
1090 */
1091 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1092 gfp_t wait)
1093 {
1094 journal_t *journal = EXT4_SB(sb)->s_journal;
1095
1096 WARN_ON(PageChecked(page));
1097 if (!page_has_buffers(page))
1098 return 0;
1099 if (journal)
1100 return jbd2_journal_try_to_free_buffers(journal, page,
1101 wait & ~__GFP_WAIT);
1102 return try_to_free_buffers(page);
1103 }
1104
1105 #ifdef CONFIG_QUOTA
1106 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1107 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1108
1109 static int ext4_write_dquot(struct dquot *dquot);
1110 static int ext4_acquire_dquot(struct dquot *dquot);
1111 static int ext4_release_dquot(struct dquot *dquot);
1112 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1113 static int ext4_write_info(struct super_block *sb, int type);
1114 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1115 struct path *path);
1116 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1117 int format_id);
1118 static int ext4_quota_off(struct super_block *sb, int type);
1119 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1120 static int ext4_quota_on_mount(struct super_block *sb, int type);
1121 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1122 size_t len, loff_t off);
1123 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1124 const char *data, size_t len, loff_t off);
1125 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1126 unsigned int flags);
1127 static int ext4_enable_quotas(struct super_block *sb);
1128
1129 static const struct dquot_operations ext4_quota_operations = {
1130 .get_reserved_space = ext4_get_reserved_space,
1131 .write_dquot = ext4_write_dquot,
1132 .acquire_dquot = ext4_acquire_dquot,
1133 .release_dquot = ext4_release_dquot,
1134 .mark_dirty = ext4_mark_dquot_dirty,
1135 .write_info = ext4_write_info,
1136 .alloc_dquot = dquot_alloc,
1137 .destroy_dquot = dquot_destroy,
1138 };
1139
1140 static const struct quotactl_ops ext4_qctl_operations = {
1141 .quota_on = ext4_quota_on,
1142 .quota_off = ext4_quota_off,
1143 .quota_sync = dquot_quota_sync,
1144 .get_info = dquot_get_dqinfo,
1145 .set_info = dquot_set_dqinfo,
1146 .get_dqblk = dquot_get_dqblk,
1147 .set_dqblk = dquot_set_dqblk
1148 };
1149
1150 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1151 .quota_on_meta = ext4_quota_on_sysfile,
1152 .quota_off = ext4_quota_off_sysfile,
1153 .quota_sync = dquot_quota_sync,
1154 .get_info = dquot_get_dqinfo,
1155 .set_info = dquot_set_dqinfo,
1156 .get_dqblk = dquot_get_dqblk,
1157 .set_dqblk = dquot_set_dqblk
1158 };
1159 #endif
1160
1161 static const struct super_operations ext4_sops = {
1162 .alloc_inode = ext4_alloc_inode,
1163 .destroy_inode = ext4_destroy_inode,
1164 .write_inode = ext4_write_inode,
1165 .dirty_inode = ext4_dirty_inode,
1166 .drop_inode = ext4_drop_inode,
1167 .evict_inode = ext4_evict_inode,
1168 .put_super = ext4_put_super,
1169 .sync_fs = ext4_sync_fs,
1170 .freeze_fs = ext4_freeze,
1171 .unfreeze_fs = ext4_unfreeze,
1172 .statfs = ext4_statfs,
1173 .remount_fs = ext4_remount,
1174 .show_options = ext4_show_options,
1175 #ifdef CONFIG_QUOTA
1176 .quota_read = ext4_quota_read,
1177 .quota_write = ext4_quota_write,
1178 #endif
1179 .bdev_try_to_free_page = bdev_try_to_free_page,
1180 };
1181
1182 static const struct super_operations ext4_nojournal_sops = {
1183 .alloc_inode = ext4_alloc_inode,
1184 .destroy_inode = ext4_destroy_inode,
1185 .write_inode = ext4_write_inode,
1186 .dirty_inode = ext4_dirty_inode,
1187 .drop_inode = ext4_drop_inode,
1188 .evict_inode = ext4_evict_inode,
1189 .put_super = ext4_put_super,
1190 .statfs = ext4_statfs,
1191 .remount_fs = ext4_remount,
1192 .show_options = ext4_show_options,
1193 #ifdef CONFIG_QUOTA
1194 .quota_read = ext4_quota_read,
1195 .quota_write = ext4_quota_write,
1196 #endif
1197 .bdev_try_to_free_page = bdev_try_to_free_page,
1198 };
1199
1200 static const struct export_operations ext4_export_ops = {
1201 .fh_to_dentry = ext4_fh_to_dentry,
1202 .fh_to_parent = ext4_fh_to_parent,
1203 .get_parent = ext4_get_parent,
1204 };
1205
1206 enum {
1207 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1208 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1209 Opt_nouid32, Opt_debug, Opt_removed,
1210 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1211 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1212 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1213 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1214 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1215 Opt_data_err_abort, Opt_data_err_ignore,
1216 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1217 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1218 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1219 Opt_usrquota, Opt_grpquota, Opt_i_version,
1220 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1221 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1222 Opt_inode_readahead_blks, Opt_journal_ioprio,
1223 Opt_dioread_nolock, Opt_dioread_lock,
1224 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1225 Opt_max_dir_size_kb,
1226 };
1227
1228 static const match_table_t tokens = {
1229 {Opt_bsd_df, "bsddf"},
1230 {Opt_minix_df, "minixdf"},
1231 {Opt_grpid, "grpid"},
1232 {Opt_grpid, "bsdgroups"},
1233 {Opt_nogrpid, "nogrpid"},
1234 {Opt_nogrpid, "sysvgroups"},
1235 {Opt_resgid, "resgid=%u"},
1236 {Opt_resuid, "resuid=%u"},
1237 {Opt_sb, "sb=%u"},
1238 {Opt_err_cont, "errors=continue"},
1239 {Opt_err_panic, "errors=panic"},
1240 {Opt_err_ro, "errors=remount-ro"},
1241 {Opt_nouid32, "nouid32"},
1242 {Opt_debug, "debug"},
1243 {Opt_removed, "oldalloc"},
1244 {Opt_removed, "orlov"},
1245 {Opt_user_xattr, "user_xattr"},
1246 {Opt_nouser_xattr, "nouser_xattr"},
1247 {Opt_acl, "acl"},
1248 {Opt_noacl, "noacl"},
1249 {Opt_noload, "norecovery"},
1250 {Opt_noload, "noload"},
1251 {Opt_removed, "nobh"},
1252 {Opt_removed, "bh"},
1253 {Opt_commit, "commit=%u"},
1254 {Opt_min_batch_time, "min_batch_time=%u"},
1255 {Opt_max_batch_time, "max_batch_time=%u"},
1256 {Opt_journal_dev, "journal_dev=%u"},
1257 {Opt_journal_checksum, "journal_checksum"},
1258 {Opt_journal_async_commit, "journal_async_commit"},
1259 {Opt_abort, "abort"},
1260 {Opt_data_journal, "data=journal"},
1261 {Opt_data_ordered, "data=ordered"},
1262 {Opt_data_writeback, "data=writeback"},
1263 {Opt_data_err_abort, "data_err=abort"},
1264 {Opt_data_err_ignore, "data_err=ignore"},
1265 {Opt_offusrjquota, "usrjquota="},
1266 {Opt_usrjquota, "usrjquota=%s"},
1267 {Opt_offgrpjquota, "grpjquota="},
1268 {Opt_grpjquota, "grpjquota=%s"},
1269 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1270 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1271 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1272 {Opt_grpquota, "grpquota"},
1273 {Opt_noquota, "noquota"},
1274 {Opt_quota, "quota"},
1275 {Opt_usrquota, "usrquota"},
1276 {Opt_barrier, "barrier=%u"},
1277 {Opt_barrier, "barrier"},
1278 {Opt_nobarrier, "nobarrier"},
1279 {Opt_i_version, "i_version"},
1280 {Opt_stripe, "stripe=%u"},
1281 {Opt_delalloc, "delalloc"},
1282 {Opt_nodelalloc, "nodelalloc"},
1283 {Opt_mblk_io_submit, "mblk_io_submit"},
1284 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1285 {Opt_block_validity, "block_validity"},
1286 {Opt_noblock_validity, "noblock_validity"},
1287 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1288 {Opt_journal_ioprio, "journal_ioprio=%u"},
1289 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1290 {Opt_auto_da_alloc, "auto_da_alloc"},
1291 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1292 {Opt_dioread_nolock, "dioread_nolock"},
1293 {Opt_dioread_lock, "dioread_lock"},
1294 {Opt_discard, "discard"},
1295 {Opt_nodiscard, "nodiscard"},
1296 {Opt_init_itable, "init_itable=%u"},
1297 {Opt_init_itable, "init_itable"},
1298 {Opt_noinit_itable, "noinit_itable"},
1299 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1300 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1301 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1302 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1303 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1304 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1305 {Opt_err, NULL},
1306 };
1307
1308 static ext4_fsblk_t get_sb_block(void **data)
1309 {
1310 ext4_fsblk_t sb_block;
1311 char *options = (char *) *data;
1312
1313 if (!options || strncmp(options, "sb=", 3) != 0)
1314 return 1; /* Default location */
1315
1316 options += 3;
1317 /* TODO: use simple_strtoll with >32bit ext4 */
1318 sb_block = simple_strtoul(options, &options, 0);
1319 if (*options && *options != ',') {
1320 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1321 (char *) *data);
1322 return 1;
1323 }
1324 if (*options == ',')
1325 options++;
1326 *data = (void *) options;
1327
1328 return sb_block;
1329 }
1330
1331 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1332 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1333 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1334
1335 #ifdef CONFIG_QUOTA
1336 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1337 {
1338 struct ext4_sb_info *sbi = EXT4_SB(sb);
1339 char *qname;
1340
1341 if (sb_any_quota_loaded(sb) &&
1342 !sbi->s_qf_names[qtype]) {
1343 ext4_msg(sb, KERN_ERR,
1344 "Cannot change journaled "
1345 "quota options when quota turned on");
1346 return -1;
1347 }
1348 qname = match_strdup(args);
1349 if (!qname) {
1350 ext4_msg(sb, KERN_ERR,
1351 "Not enough memory for storing quotafile name");
1352 return -1;
1353 }
1354 if (sbi->s_qf_names[qtype] &&
1355 strcmp(sbi->s_qf_names[qtype], qname)) {
1356 ext4_msg(sb, KERN_ERR,
1357 "%s quota file already specified", QTYPE2NAME(qtype));
1358 kfree(qname);
1359 return -1;
1360 }
1361 sbi->s_qf_names[qtype] = qname;
1362 if (strchr(sbi->s_qf_names[qtype], '/')) {
1363 ext4_msg(sb, KERN_ERR,
1364 "quotafile must be on filesystem root");
1365 kfree(sbi->s_qf_names[qtype]);
1366 sbi->s_qf_names[qtype] = NULL;
1367 return -1;
1368 }
1369 set_opt(sb, QUOTA);
1370 return 1;
1371 }
1372
1373 static int clear_qf_name(struct super_block *sb, int qtype)
1374 {
1375
1376 struct ext4_sb_info *sbi = EXT4_SB(sb);
1377
1378 if (sb_any_quota_loaded(sb) &&
1379 sbi->s_qf_names[qtype]) {
1380 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1381 " when quota turned on");
1382 return -1;
1383 }
1384 /*
1385 * The space will be released later when all options are confirmed
1386 * to be correct
1387 */
1388 sbi->s_qf_names[qtype] = NULL;
1389 return 1;
1390 }
1391 #endif
1392
1393 #define MOPT_SET 0x0001
1394 #define MOPT_CLEAR 0x0002
1395 #define MOPT_NOSUPPORT 0x0004
1396 #define MOPT_EXPLICIT 0x0008
1397 #define MOPT_CLEAR_ERR 0x0010
1398 #define MOPT_GTE0 0x0020
1399 #ifdef CONFIG_QUOTA
1400 #define MOPT_Q 0
1401 #define MOPT_QFMT 0x0040
1402 #else
1403 #define MOPT_Q MOPT_NOSUPPORT
1404 #define MOPT_QFMT MOPT_NOSUPPORT
1405 #endif
1406 #define MOPT_DATAJ 0x0080
1407
1408 static const struct mount_opts {
1409 int token;
1410 int mount_opt;
1411 int flags;
1412 } ext4_mount_opts[] = {
1413 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1414 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1415 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1416 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1417 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1418 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1419 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1420 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1421 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1422 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1423 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1424 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1425 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1426 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1427 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1428 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1429 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1430 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1431 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1432 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1433 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1434 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1435 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1436 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1437 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1438 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1439 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1440 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1441 {Opt_commit, 0, MOPT_GTE0},
1442 {Opt_max_batch_time, 0, MOPT_GTE0},
1443 {Opt_min_batch_time, 0, MOPT_GTE0},
1444 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1445 {Opt_init_itable, 0, MOPT_GTE0},
1446 {Opt_stripe, 0, MOPT_GTE0},
1447 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1448 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1449 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1450 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1451 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1452 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1453 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1454 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1455 #else
1456 {Opt_acl, 0, MOPT_NOSUPPORT},
1457 {Opt_noacl, 0, MOPT_NOSUPPORT},
1458 #endif
1459 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1460 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1461 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1462 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1463 MOPT_SET | MOPT_Q},
1464 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1465 MOPT_SET | MOPT_Q},
1466 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1467 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1468 {Opt_usrjquota, 0, MOPT_Q},
1469 {Opt_grpjquota, 0, MOPT_Q},
1470 {Opt_offusrjquota, 0, MOPT_Q},
1471 {Opt_offgrpjquota, 0, MOPT_Q},
1472 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1473 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1474 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1475 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1476 {Opt_err, 0, 0}
1477 };
1478
1479 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1480 substring_t *args, unsigned long *journal_devnum,
1481 unsigned int *journal_ioprio, int is_remount)
1482 {
1483 struct ext4_sb_info *sbi = EXT4_SB(sb);
1484 const struct mount_opts *m;
1485 kuid_t uid;
1486 kgid_t gid;
1487 int arg = 0;
1488
1489 #ifdef CONFIG_QUOTA
1490 if (token == Opt_usrjquota)
1491 return set_qf_name(sb, USRQUOTA, &args[0]);
1492 else if (token == Opt_grpjquota)
1493 return set_qf_name(sb, GRPQUOTA, &args[0]);
1494 else if (token == Opt_offusrjquota)
1495 return clear_qf_name(sb, USRQUOTA);
1496 else if (token == Opt_offgrpjquota)
1497 return clear_qf_name(sb, GRPQUOTA);
1498 #endif
1499 if (args->from && match_int(args, &arg))
1500 return -1;
1501 switch (token) {
1502 case Opt_noacl:
1503 case Opt_nouser_xattr:
1504 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1505 break;
1506 case Opt_sb:
1507 return 1; /* handled by get_sb_block() */
1508 case Opt_removed:
1509 ext4_msg(sb, KERN_WARNING,
1510 "Ignoring removed %s option", opt);
1511 return 1;
1512 case Opt_resuid:
1513 uid = make_kuid(current_user_ns(), arg);
1514 if (!uid_valid(uid)) {
1515 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1516 return -1;
1517 }
1518 sbi->s_resuid = uid;
1519 return 1;
1520 case Opt_resgid:
1521 gid = make_kgid(current_user_ns(), arg);
1522 if (!gid_valid(gid)) {
1523 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1524 return -1;
1525 }
1526 sbi->s_resgid = gid;
1527 return 1;
1528 case Opt_abort:
1529 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1530 return 1;
1531 case Opt_i_version:
1532 sb->s_flags |= MS_I_VERSION;
1533 return 1;
1534 case Opt_journal_dev:
1535 if (is_remount) {
1536 ext4_msg(sb, KERN_ERR,
1537 "Cannot specify journal on remount");
1538 return -1;
1539 }
1540 *journal_devnum = arg;
1541 return 1;
1542 case Opt_journal_ioprio:
1543 if (arg < 0 || arg > 7)
1544 return -1;
1545 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1546 return 1;
1547 }
1548
1549 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1550 if (token != m->token)
1551 continue;
1552 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1553 return -1;
1554 if (m->flags & MOPT_EXPLICIT)
1555 set_opt2(sb, EXPLICIT_DELALLOC);
1556 if (m->flags & MOPT_CLEAR_ERR)
1557 clear_opt(sb, ERRORS_MASK);
1558 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1559 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1560 "options when quota turned on");
1561 return -1;
1562 }
1563
1564 if (m->flags & MOPT_NOSUPPORT) {
1565 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1566 } else if (token == Opt_commit) {
1567 if (arg == 0)
1568 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1569 sbi->s_commit_interval = HZ * arg;
1570 } else if (token == Opt_max_batch_time) {
1571 if (arg == 0)
1572 arg = EXT4_DEF_MAX_BATCH_TIME;
1573 sbi->s_max_batch_time = arg;
1574 } else if (token == Opt_min_batch_time) {
1575 sbi->s_min_batch_time = arg;
1576 } else if (token == Opt_inode_readahead_blks) {
1577 if (arg > (1 << 30))
1578 return -1;
1579 if (arg && !is_power_of_2(arg)) {
1580 ext4_msg(sb, KERN_ERR,
1581 "EXT4-fs: inode_readahead_blks"
1582 " must be a power of 2");
1583 return -1;
1584 }
1585 sbi->s_inode_readahead_blks = arg;
1586 } else if (token == Opt_init_itable) {
1587 set_opt(sb, INIT_INODE_TABLE);
1588 if (!args->from)
1589 arg = EXT4_DEF_LI_WAIT_MULT;
1590 sbi->s_li_wait_mult = arg;
1591 } else if (token == Opt_max_dir_size_kb) {
1592 sbi->s_max_dir_size_kb = arg;
1593 } else if (token == Opt_stripe) {
1594 sbi->s_stripe = arg;
1595 } else if (m->flags & MOPT_DATAJ) {
1596 if (is_remount) {
1597 if (!sbi->s_journal)
1598 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1599 else if (test_opt(sb, DATA_FLAGS) !=
1600 m->mount_opt) {
1601 ext4_msg(sb, KERN_ERR,
1602 "Cannot change data mode on remount");
1603 return -1;
1604 }
1605 } else {
1606 clear_opt(sb, DATA_FLAGS);
1607 sbi->s_mount_opt |= m->mount_opt;
1608 }
1609 #ifdef CONFIG_QUOTA
1610 } else if (m->flags & MOPT_QFMT) {
1611 if (sb_any_quota_loaded(sb) &&
1612 sbi->s_jquota_fmt != m->mount_opt) {
1613 ext4_msg(sb, KERN_ERR, "Cannot "
1614 "change journaled quota options "
1615 "when quota turned on");
1616 return -1;
1617 }
1618 sbi->s_jquota_fmt = m->mount_opt;
1619 #endif
1620 } else {
1621 if (!args->from)
1622 arg = 1;
1623 if (m->flags & MOPT_CLEAR)
1624 arg = !arg;
1625 else if (unlikely(!(m->flags & MOPT_SET))) {
1626 ext4_msg(sb, KERN_WARNING,
1627 "buggy handling of option %s", opt);
1628 WARN_ON(1);
1629 return -1;
1630 }
1631 if (arg != 0)
1632 sbi->s_mount_opt |= m->mount_opt;
1633 else
1634 sbi->s_mount_opt &= ~m->mount_opt;
1635 }
1636 return 1;
1637 }
1638 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1639 "or missing value", opt);
1640 return -1;
1641 }
1642
1643 static int parse_options(char *options, struct super_block *sb,
1644 unsigned long *journal_devnum,
1645 unsigned int *journal_ioprio,
1646 int is_remount)
1647 {
1648 struct ext4_sb_info *sbi = EXT4_SB(sb);
1649 char *p;
1650 substring_t args[MAX_OPT_ARGS];
1651 int token;
1652
1653 if (!options)
1654 return 1;
1655
1656 while ((p = strsep(&options, ",")) != NULL) {
1657 if (!*p)
1658 continue;
1659 /*
1660 * Initialize args struct so we know whether arg was
1661 * found; some options take optional arguments.
1662 */
1663 args[0].to = args[0].from = NULL;
1664 token = match_token(p, tokens, args);
1665 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1666 journal_ioprio, is_remount) < 0)
1667 return 0;
1668 }
1669 #ifdef CONFIG_QUOTA
1670 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1671 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1672 clear_opt(sb, USRQUOTA);
1673
1674 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1675 clear_opt(sb, GRPQUOTA);
1676
1677 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1678 ext4_msg(sb, KERN_ERR, "old and new quota "
1679 "format mixing");
1680 return 0;
1681 }
1682
1683 if (!sbi->s_jquota_fmt) {
1684 ext4_msg(sb, KERN_ERR, "journaled quota format "
1685 "not specified");
1686 return 0;
1687 }
1688 } else {
1689 if (sbi->s_jquota_fmt) {
1690 ext4_msg(sb, KERN_ERR, "journaled quota format "
1691 "specified with no journaling "
1692 "enabled");
1693 return 0;
1694 }
1695 }
1696 #endif
1697 if (test_opt(sb, DIOREAD_NOLOCK)) {
1698 int blocksize =
1699 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1700
1701 if (blocksize < PAGE_CACHE_SIZE) {
1702 ext4_msg(sb, KERN_ERR, "can't mount with "
1703 "dioread_nolock if block size != PAGE_SIZE");
1704 return 0;
1705 }
1706 }
1707 return 1;
1708 }
1709
1710 static inline void ext4_show_quota_options(struct seq_file *seq,
1711 struct super_block *sb)
1712 {
1713 #if defined(CONFIG_QUOTA)
1714 struct ext4_sb_info *sbi = EXT4_SB(sb);
1715
1716 if (sbi->s_jquota_fmt) {
1717 char *fmtname = "";
1718
1719 switch (sbi->s_jquota_fmt) {
1720 case QFMT_VFS_OLD:
1721 fmtname = "vfsold";
1722 break;
1723 case QFMT_VFS_V0:
1724 fmtname = "vfsv0";
1725 break;
1726 case QFMT_VFS_V1:
1727 fmtname = "vfsv1";
1728 break;
1729 }
1730 seq_printf(seq, ",jqfmt=%s", fmtname);
1731 }
1732
1733 if (sbi->s_qf_names[USRQUOTA])
1734 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1735
1736 if (sbi->s_qf_names[GRPQUOTA])
1737 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1738
1739 if (test_opt(sb, USRQUOTA))
1740 seq_puts(seq, ",usrquota");
1741
1742 if (test_opt(sb, GRPQUOTA))
1743 seq_puts(seq, ",grpquota");
1744 #endif
1745 }
1746
1747 static const char *token2str(int token)
1748 {
1749 const struct match_token *t;
1750
1751 for (t = tokens; t->token != Opt_err; t++)
1752 if (t->token == token && !strchr(t->pattern, '='))
1753 break;
1754 return t->pattern;
1755 }
1756
1757 /*
1758 * Show an option if
1759 * - it's set to a non-default value OR
1760 * - if the per-sb default is different from the global default
1761 */
1762 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1763 int nodefs)
1764 {
1765 struct ext4_sb_info *sbi = EXT4_SB(sb);
1766 struct ext4_super_block *es = sbi->s_es;
1767 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1768 const struct mount_opts *m;
1769 char sep = nodefs ? '\n' : ',';
1770
1771 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1772 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1773
1774 if (sbi->s_sb_block != 1)
1775 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1776
1777 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1778 int want_set = m->flags & MOPT_SET;
1779 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1780 (m->flags & MOPT_CLEAR_ERR))
1781 continue;
1782 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1783 continue; /* skip if same as the default */
1784 if ((want_set &&
1785 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1786 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1787 continue; /* select Opt_noFoo vs Opt_Foo */
1788 SEQ_OPTS_PRINT("%s", token2str(m->token));
1789 }
1790
1791 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1792 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1793 SEQ_OPTS_PRINT("resuid=%u",
1794 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1795 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1796 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1797 SEQ_OPTS_PRINT("resgid=%u",
1798 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1799 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1800 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1801 SEQ_OPTS_PUTS("errors=remount-ro");
1802 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1803 SEQ_OPTS_PUTS("errors=continue");
1804 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1805 SEQ_OPTS_PUTS("errors=panic");
1806 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1807 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1808 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1809 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1810 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1811 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1812 if (sb->s_flags & MS_I_VERSION)
1813 SEQ_OPTS_PUTS("i_version");
1814 if (nodefs || sbi->s_stripe)
1815 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1816 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1817 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1818 SEQ_OPTS_PUTS("data=journal");
1819 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1820 SEQ_OPTS_PUTS("data=ordered");
1821 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1822 SEQ_OPTS_PUTS("data=writeback");
1823 }
1824 if (nodefs ||
1825 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1826 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1827 sbi->s_inode_readahead_blks);
1828
1829 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1830 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1831 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1832 if (nodefs || sbi->s_max_dir_size_kb)
1833 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1834
1835 ext4_show_quota_options(seq, sb);
1836 return 0;
1837 }
1838
1839 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1840 {
1841 return _ext4_show_options(seq, root->d_sb, 0);
1842 }
1843
1844 static int options_seq_show(struct seq_file *seq, void *offset)
1845 {
1846 struct super_block *sb = seq->private;
1847 int rc;
1848
1849 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1850 rc = _ext4_show_options(seq, sb, 1);
1851 seq_puts(seq, "\n");
1852 return rc;
1853 }
1854
1855 static int options_open_fs(struct inode *inode, struct file *file)
1856 {
1857 return single_open(file, options_seq_show, PDE(inode)->data);
1858 }
1859
1860 static const struct file_operations ext4_seq_options_fops = {
1861 .owner = THIS_MODULE,
1862 .open = options_open_fs,
1863 .read = seq_read,
1864 .llseek = seq_lseek,
1865 .release = single_release,
1866 };
1867
1868 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1869 int read_only)
1870 {
1871 struct ext4_sb_info *sbi = EXT4_SB(sb);
1872 int res = 0;
1873
1874 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1875 ext4_msg(sb, KERN_ERR, "revision level too high, "
1876 "forcing read-only mode");
1877 res = MS_RDONLY;
1878 }
1879 if (read_only)
1880 goto done;
1881 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1882 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1883 "running e2fsck is recommended");
1884 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1885 ext4_msg(sb, KERN_WARNING,
1886 "warning: mounting fs with errors, "
1887 "running e2fsck is recommended");
1888 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1889 le16_to_cpu(es->s_mnt_count) >=
1890 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1891 ext4_msg(sb, KERN_WARNING,
1892 "warning: maximal mount count reached, "
1893 "running e2fsck is recommended");
1894 else if (le32_to_cpu(es->s_checkinterval) &&
1895 (le32_to_cpu(es->s_lastcheck) +
1896 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1897 ext4_msg(sb, KERN_WARNING,
1898 "warning: checktime reached, "
1899 "running e2fsck is recommended");
1900 if (!sbi->s_journal)
1901 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1902 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1903 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1904 le16_add_cpu(&es->s_mnt_count, 1);
1905 es->s_mtime = cpu_to_le32(get_seconds());
1906 ext4_update_dynamic_rev(sb);
1907 if (sbi->s_journal)
1908 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1909
1910 ext4_commit_super(sb, 1);
1911 done:
1912 if (test_opt(sb, DEBUG))
1913 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1914 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1915 sb->s_blocksize,
1916 sbi->s_groups_count,
1917 EXT4_BLOCKS_PER_GROUP(sb),
1918 EXT4_INODES_PER_GROUP(sb),
1919 sbi->s_mount_opt, sbi->s_mount_opt2);
1920
1921 cleancache_init_fs(sb);
1922 return res;
1923 }
1924
1925 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1926 {
1927 struct ext4_sb_info *sbi = EXT4_SB(sb);
1928 struct flex_groups *new_groups;
1929 int size;
1930
1931 if (!sbi->s_log_groups_per_flex)
1932 return 0;
1933
1934 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1935 if (size <= sbi->s_flex_groups_allocated)
1936 return 0;
1937
1938 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1939 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1940 if (!new_groups) {
1941 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1942 size / (int) sizeof(struct flex_groups));
1943 return -ENOMEM;
1944 }
1945
1946 if (sbi->s_flex_groups) {
1947 memcpy(new_groups, sbi->s_flex_groups,
1948 (sbi->s_flex_groups_allocated *
1949 sizeof(struct flex_groups)));
1950 ext4_kvfree(sbi->s_flex_groups);
1951 }
1952 sbi->s_flex_groups = new_groups;
1953 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1954 return 0;
1955 }
1956
1957 static int ext4_fill_flex_info(struct super_block *sb)
1958 {
1959 struct ext4_sb_info *sbi = EXT4_SB(sb);
1960 struct ext4_group_desc *gdp = NULL;
1961 ext4_group_t flex_group;
1962 unsigned int groups_per_flex = 0;
1963 int i, err;
1964
1965 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1966 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1967 sbi->s_log_groups_per_flex = 0;
1968 return 1;
1969 }
1970 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
1971
1972 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1973 if (err)
1974 goto failed;
1975
1976 for (i = 0; i < sbi->s_groups_count; i++) {
1977 gdp = ext4_get_group_desc(sb, i, NULL);
1978
1979 flex_group = ext4_flex_group(sbi, i);
1980 atomic_add(ext4_free_inodes_count(sb, gdp),
1981 &sbi->s_flex_groups[flex_group].free_inodes);
1982 atomic_add(ext4_free_group_clusters(sb, gdp),
1983 &sbi->s_flex_groups[flex_group].free_clusters);
1984 atomic_add(ext4_used_dirs_count(sb, gdp),
1985 &sbi->s_flex_groups[flex_group].used_dirs);
1986 }
1987
1988 return 1;
1989 failed:
1990 return 0;
1991 }
1992
1993 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1994 struct ext4_group_desc *gdp)
1995 {
1996 int offset;
1997 __u16 crc = 0;
1998 __le32 le_group = cpu_to_le32(block_group);
1999
2000 if ((sbi->s_es->s_feature_ro_compat &
2001 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2002 /* Use new metadata_csum algorithm */
2003 __u16 old_csum;
2004 __u32 csum32;
2005
2006 old_csum = gdp->bg_checksum;
2007 gdp->bg_checksum = 0;
2008 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2009 sizeof(le_group));
2010 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2011 sbi->s_desc_size);
2012 gdp->bg_checksum = old_csum;
2013
2014 crc = csum32 & 0xFFFF;
2015 goto out;
2016 }
2017
2018 /* old crc16 code */
2019 offset = offsetof(struct ext4_group_desc, bg_checksum);
2020
2021 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2022 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2023 crc = crc16(crc, (__u8 *)gdp, offset);
2024 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2025 /* for checksum of struct ext4_group_desc do the rest...*/
2026 if ((sbi->s_es->s_feature_incompat &
2027 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2028 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2029 crc = crc16(crc, (__u8 *)gdp + offset,
2030 le16_to_cpu(sbi->s_es->s_desc_size) -
2031 offset);
2032
2033 out:
2034 return cpu_to_le16(crc);
2035 }
2036
2037 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2038 struct ext4_group_desc *gdp)
2039 {
2040 if (ext4_has_group_desc_csum(sb) &&
2041 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2042 block_group, gdp)))
2043 return 0;
2044
2045 return 1;
2046 }
2047
2048 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2049 struct ext4_group_desc *gdp)
2050 {
2051 if (!ext4_has_group_desc_csum(sb))
2052 return;
2053 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2054 }
2055
2056 /* Called at mount-time, super-block is locked */
2057 static int ext4_check_descriptors(struct super_block *sb,
2058 ext4_group_t *first_not_zeroed)
2059 {
2060 struct ext4_sb_info *sbi = EXT4_SB(sb);
2061 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2062 ext4_fsblk_t last_block;
2063 ext4_fsblk_t block_bitmap;
2064 ext4_fsblk_t inode_bitmap;
2065 ext4_fsblk_t inode_table;
2066 int flexbg_flag = 0;
2067 ext4_group_t i, grp = sbi->s_groups_count;
2068
2069 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2070 flexbg_flag = 1;
2071
2072 ext4_debug("Checking group descriptors");
2073
2074 for (i = 0; i < sbi->s_groups_count; i++) {
2075 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2076
2077 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2078 last_block = ext4_blocks_count(sbi->s_es) - 1;
2079 else
2080 last_block = first_block +
2081 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2082
2083 if ((grp == sbi->s_groups_count) &&
2084 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2085 grp = i;
2086
2087 block_bitmap = ext4_block_bitmap(sb, gdp);
2088 if (block_bitmap < first_block || block_bitmap > last_block) {
2089 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2090 "Block bitmap for group %u not in group "
2091 "(block %llu)!", i, block_bitmap);
2092 return 0;
2093 }
2094 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2095 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2096 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2097 "Inode bitmap for group %u not in group "
2098 "(block %llu)!", i, inode_bitmap);
2099 return 0;
2100 }
2101 inode_table = ext4_inode_table(sb, gdp);
2102 if (inode_table < first_block ||
2103 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2104 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2105 "Inode table for group %u not in group "
2106 "(block %llu)!", i, inode_table);
2107 return 0;
2108 }
2109 ext4_lock_group(sb, i);
2110 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2111 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2112 "Checksum for group %u failed (%u!=%u)",
2113 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2114 gdp)), le16_to_cpu(gdp->bg_checksum));
2115 if (!(sb->s_flags & MS_RDONLY)) {
2116 ext4_unlock_group(sb, i);
2117 return 0;
2118 }
2119 }
2120 ext4_unlock_group(sb, i);
2121 if (!flexbg_flag)
2122 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2123 }
2124 if (NULL != first_not_zeroed)
2125 *first_not_zeroed = grp;
2126
2127 ext4_free_blocks_count_set(sbi->s_es,
2128 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2129 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2130 return 1;
2131 }
2132
2133 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2134 * the superblock) which were deleted from all directories, but held open by
2135 * a process at the time of a crash. We walk the list and try to delete these
2136 * inodes at recovery time (only with a read-write filesystem).
2137 *
2138 * In order to keep the orphan inode chain consistent during traversal (in
2139 * case of crash during recovery), we link each inode into the superblock
2140 * orphan list_head and handle it the same way as an inode deletion during
2141 * normal operation (which journals the operations for us).
2142 *
2143 * We only do an iget() and an iput() on each inode, which is very safe if we
2144 * accidentally point at an in-use or already deleted inode. The worst that
2145 * can happen in this case is that we get a "bit already cleared" message from
2146 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2147 * e2fsck was run on this filesystem, and it must have already done the orphan
2148 * inode cleanup for us, so we can safely abort without any further action.
2149 */
2150 static void ext4_orphan_cleanup(struct super_block *sb,
2151 struct ext4_super_block *es)
2152 {
2153 unsigned int s_flags = sb->s_flags;
2154 int nr_orphans = 0, nr_truncates = 0;
2155 #ifdef CONFIG_QUOTA
2156 int i;
2157 #endif
2158 if (!es->s_last_orphan) {
2159 jbd_debug(4, "no orphan inodes to clean up\n");
2160 return;
2161 }
2162
2163 if (bdev_read_only(sb->s_bdev)) {
2164 ext4_msg(sb, KERN_ERR, "write access "
2165 "unavailable, skipping orphan cleanup");
2166 return;
2167 }
2168
2169 /* Check if feature set would not allow a r/w mount */
2170 if (!ext4_feature_set_ok(sb, 0)) {
2171 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2172 "unknown ROCOMPAT features");
2173 return;
2174 }
2175
2176 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2177 /* don't clear list on RO mount w/ errors */
2178 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2179 jbd_debug(1, "Errors on filesystem, "
2180 "clearing orphan list.\n");
2181 es->s_last_orphan = 0;
2182 }
2183 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2184 return;
2185 }
2186
2187 if (s_flags & MS_RDONLY) {
2188 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2189 sb->s_flags &= ~MS_RDONLY;
2190 }
2191 #ifdef CONFIG_QUOTA
2192 /* Needed for iput() to work correctly and not trash data */
2193 sb->s_flags |= MS_ACTIVE;
2194 /* Turn on quotas so that they are updated correctly */
2195 for (i = 0; i < MAXQUOTAS; i++) {
2196 if (EXT4_SB(sb)->s_qf_names[i]) {
2197 int ret = ext4_quota_on_mount(sb, i);
2198 if (ret < 0)
2199 ext4_msg(sb, KERN_ERR,
2200 "Cannot turn on journaled "
2201 "quota: error %d", ret);
2202 }
2203 }
2204 #endif
2205
2206 while (es->s_last_orphan) {
2207 struct inode *inode;
2208
2209 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2210 if (IS_ERR(inode)) {
2211 es->s_last_orphan = 0;
2212 break;
2213 }
2214
2215 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2216 dquot_initialize(inode);
2217 if (inode->i_nlink) {
2218 ext4_msg(sb, KERN_DEBUG,
2219 "%s: truncating inode %lu to %lld bytes",
2220 __func__, inode->i_ino, inode->i_size);
2221 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2222 inode->i_ino, inode->i_size);
2223 mutex_lock(&inode->i_mutex);
2224 ext4_truncate(inode);
2225 mutex_unlock(&inode->i_mutex);
2226 nr_truncates++;
2227 } else {
2228 ext4_msg(sb, KERN_DEBUG,
2229 "%s: deleting unreferenced inode %lu",
2230 __func__, inode->i_ino);
2231 jbd_debug(2, "deleting unreferenced inode %lu\n",
2232 inode->i_ino);
2233 nr_orphans++;
2234 }
2235 iput(inode); /* The delete magic happens here! */
2236 }
2237
2238 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2239
2240 if (nr_orphans)
2241 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2242 PLURAL(nr_orphans));
2243 if (nr_truncates)
2244 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2245 PLURAL(nr_truncates));
2246 #ifdef CONFIG_QUOTA
2247 /* Turn quotas off */
2248 for (i = 0; i < MAXQUOTAS; i++) {
2249 if (sb_dqopt(sb)->files[i])
2250 dquot_quota_off(sb, i);
2251 }
2252 #endif
2253 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2254 }
2255
2256 /*
2257 * Maximal extent format file size.
2258 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2259 * extent format containers, within a sector_t, and within i_blocks
2260 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2261 * so that won't be a limiting factor.
2262 *
2263 * However there is other limiting factor. We do store extents in the form
2264 * of starting block and length, hence the resulting length of the extent
2265 * covering maximum file size must fit into on-disk format containers as
2266 * well. Given that length is always by 1 unit bigger than max unit (because
2267 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2268 *
2269 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2270 */
2271 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2272 {
2273 loff_t res;
2274 loff_t upper_limit = MAX_LFS_FILESIZE;
2275
2276 /* small i_blocks in vfs inode? */
2277 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2278 /*
2279 * CONFIG_LBDAF is not enabled implies the inode
2280 * i_block represent total blocks in 512 bytes
2281 * 32 == size of vfs inode i_blocks * 8
2282 */
2283 upper_limit = (1LL << 32) - 1;
2284
2285 /* total blocks in file system block size */
2286 upper_limit >>= (blkbits - 9);
2287 upper_limit <<= blkbits;
2288 }
2289
2290 /*
2291 * 32-bit extent-start container, ee_block. We lower the maxbytes
2292 * by one fs block, so ee_len can cover the extent of maximum file
2293 * size
2294 */
2295 res = (1LL << 32) - 1;
2296 res <<= blkbits;
2297
2298 /* Sanity check against vm- & vfs- imposed limits */
2299 if (res > upper_limit)
2300 res = upper_limit;
2301
2302 return res;
2303 }
2304
2305 /*
2306 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2307 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2308 * We need to be 1 filesystem block less than the 2^48 sector limit.
2309 */
2310 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2311 {
2312 loff_t res = EXT4_NDIR_BLOCKS;
2313 int meta_blocks;
2314 loff_t upper_limit;
2315 /* This is calculated to be the largest file size for a dense, block
2316 * mapped file such that the file's total number of 512-byte sectors,
2317 * including data and all indirect blocks, does not exceed (2^48 - 1).
2318 *
2319 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2320 * number of 512-byte sectors of the file.
2321 */
2322
2323 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2324 /*
2325 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2326 * the inode i_block field represents total file blocks in
2327 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2328 */
2329 upper_limit = (1LL << 32) - 1;
2330
2331 /* total blocks in file system block size */
2332 upper_limit >>= (bits - 9);
2333
2334 } else {
2335 /*
2336 * We use 48 bit ext4_inode i_blocks
2337 * With EXT4_HUGE_FILE_FL set the i_blocks
2338 * represent total number of blocks in
2339 * file system block size
2340 */
2341 upper_limit = (1LL << 48) - 1;
2342
2343 }
2344
2345 /* indirect blocks */
2346 meta_blocks = 1;
2347 /* double indirect blocks */
2348 meta_blocks += 1 + (1LL << (bits-2));
2349 /* tripple indirect blocks */
2350 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2351
2352 upper_limit -= meta_blocks;
2353 upper_limit <<= bits;
2354
2355 res += 1LL << (bits-2);
2356 res += 1LL << (2*(bits-2));
2357 res += 1LL << (3*(bits-2));
2358 res <<= bits;
2359 if (res > upper_limit)
2360 res = upper_limit;
2361
2362 if (res > MAX_LFS_FILESIZE)
2363 res = MAX_LFS_FILESIZE;
2364
2365 return res;
2366 }
2367
2368 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2369 ext4_fsblk_t logical_sb_block, int nr)
2370 {
2371 struct ext4_sb_info *sbi = EXT4_SB(sb);
2372 ext4_group_t bg, first_meta_bg;
2373 int has_super = 0;
2374
2375 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2376
2377 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2378 nr < first_meta_bg)
2379 return logical_sb_block + nr + 1;
2380 bg = sbi->s_desc_per_block * nr;
2381 if (ext4_bg_has_super(sb, bg))
2382 has_super = 1;
2383
2384 return (has_super + ext4_group_first_block_no(sb, bg));
2385 }
2386
2387 /**
2388 * ext4_get_stripe_size: Get the stripe size.
2389 * @sbi: In memory super block info
2390 *
2391 * If we have specified it via mount option, then
2392 * use the mount option value. If the value specified at mount time is
2393 * greater than the blocks per group use the super block value.
2394 * If the super block value is greater than blocks per group return 0.
2395 * Allocator needs it be less than blocks per group.
2396 *
2397 */
2398 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2399 {
2400 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2401 unsigned long stripe_width =
2402 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2403 int ret;
2404
2405 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2406 ret = sbi->s_stripe;
2407 else if (stripe_width <= sbi->s_blocks_per_group)
2408 ret = stripe_width;
2409 else if (stride <= sbi->s_blocks_per_group)
2410 ret = stride;
2411 else
2412 ret = 0;
2413
2414 /*
2415 * If the stripe width is 1, this makes no sense and
2416 * we set it to 0 to turn off stripe handling code.
2417 */
2418 if (ret <= 1)
2419 ret = 0;
2420
2421 return ret;
2422 }
2423
2424 /* sysfs supprt */
2425
2426 struct ext4_attr {
2427 struct attribute attr;
2428 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2429 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2430 const char *, size_t);
2431 int offset;
2432 };
2433
2434 static int parse_strtoul(const char *buf,
2435 unsigned long max, unsigned long *value)
2436 {
2437 char *endp;
2438
2439 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2440 endp = skip_spaces(endp);
2441 if (*endp || *value > max)
2442 return -EINVAL;
2443
2444 return 0;
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
2487 if (parse_strtoul(buf, 0x40000000, &t))
2488 return -EINVAL;
2489
2490 if (t && !is_power_of_2(t))
2491 return -EINVAL;
2492
2493 sbi->s_inode_readahead_blks = t;
2494 return count;
2495 }
2496
2497 static ssize_t sbi_ui_show(struct ext4_attr *a,
2498 struct ext4_sb_info *sbi, char *buf)
2499 {
2500 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2501
2502 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2503 }
2504
2505 static ssize_t sbi_ui_store(struct ext4_attr *a,
2506 struct ext4_sb_info *sbi,
2507 const char *buf, size_t count)
2508 {
2509 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2510 unsigned long t;
2511
2512 if (parse_strtoul(buf, 0xffffffff, &t))
2513 return -EINVAL;
2514 *ui = t;
2515 return count;
2516 }
2517
2518 static ssize_t trigger_test_error(struct ext4_attr *a,
2519 struct ext4_sb_info *sbi,
2520 const char *buf, size_t count)
2521 {
2522 int len = count;
2523
2524 if (!capable(CAP_SYS_ADMIN))
2525 return -EPERM;
2526
2527 if (len && buf[len-1] == '\n')
2528 len--;
2529
2530 if (len)
2531 ext4_error(sbi->s_sb, "%.*s", len, buf);
2532 return count;
2533 }
2534
2535 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2536 static struct ext4_attr ext4_attr_##_name = { \
2537 .attr = {.name = __stringify(_name), .mode = _mode }, \
2538 .show = _show, \
2539 .store = _store, \
2540 .offset = offsetof(struct ext4_sb_info, _elname), \
2541 }
2542 #define EXT4_ATTR(name, mode, show, store) \
2543 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2544
2545 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2546 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2547 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2548 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2549 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2550 #define ATTR_LIST(name) &ext4_attr_##name.attr
2551
2552 EXT4_RO_ATTR(delayed_allocation_blocks);
2553 EXT4_RO_ATTR(session_write_kbytes);
2554 EXT4_RO_ATTR(lifetime_write_kbytes);
2555 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2556 inode_readahead_blks_store, s_inode_readahead_blks);
2557 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2558 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2559 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2560 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2561 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2562 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2563 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2564 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2565 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2566 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2567
2568 static struct attribute *ext4_attrs[] = {
2569 ATTR_LIST(delayed_allocation_blocks),
2570 ATTR_LIST(session_write_kbytes),
2571 ATTR_LIST(lifetime_write_kbytes),
2572 ATTR_LIST(inode_readahead_blks),
2573 ATTR_LIST(inode_goal),
2574 ATTR_LIST(mb_stats),
2575 ATTR_LIST(mb_max_to_scan),
2576 ATTR_LIST(mb_min_to_scan),
2577 ATTR_LIST(mb_order2_req),
2578 ATTR_LIST(mb_stream_req),
2579 ATTR_LIST(mb_group_prealloc),
2580 ATTR_LIST(max_writeback_mb_bump),
2581 ATTR_LIST(extent_max_zeroout_kb),
2582 ATTR_LIST(trigger_fs_error),
2583 NULL,
2584 };
2585
2586 /* Features this copy of ext4 supports */
2587 EXT4_INFO_ATTR(lazy_itable_init);
2588 EXT4_INFO_ATTR(batched_discard);
2589 EXT4_INFO_ATTR(meta_bg_resize);
2590
2591 static struct attribute *ext4_feat_attrs[] = {
2592 ATTR_LIST(lazy_itable_init),
2593 ATTR_LIST(batched_discard),
2594 ATTR_LIST(meta_bg_resize),
2595 NULL,
2596 };
2597
2598 static ssize_t ext4_attr_show(struct kobject *kobj,
2599 struct attribute *attr, char *buf)
2600 {
2601 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2602 s_kobj);
2603 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2604
2605 return a->show ? a->show(a, sbi, buf) : 0;
2606 }
2607
2608 static ssize_t ext4_attr_store(struct kobject *kobj,
2609 struct attribute *attr,
2610 const char *buf, size_t len)
2611 {
2612 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2613 s_kobj);
2614 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2615
2616 return a->store ? a->store(a, sbi, buf, len) : 0;
2617 }
2618
2619 static void ext4_sb_release(struct kobject *kobj)
2620 {
2621 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2622 s_kobj);
2623 complete(&sbi->s_kobj_unregister);
2624 }
2625
2626 static const struct sysfs_ops ext4_attr_ops = {
2627 .show = ext4_attr_show,
2628 .store = ext4_attr_store,
2629 };
2630
2631 static struct kobj_type ext4_ktype = {
2632 .default_attrs = ext4_attrs,
2633 .sysfs_ops = &ext4_attr_ops,
2634 .release = ext4_sb_release,
2635 };
2636
2637 static void ext4_feat_release(struct kobject *kobj)
2638 {
2639 complete(&ext4_feat->f_kobj_unregister);
2640 }
2641
2642 static struct kobj_type ext4_feat_ktype = {
2643 .default_attrs = ext4_feat_attrs,
2644 .sysfs_ops = &ext4_attr_ops,
2645 .release = ext4_feat_release,
2646 };
2647
2648 /*
2649 * Check whether this filesystem can be mounted based on
2650 * the features present and the RDONLY/RDWR mount requested.
2651 * Returns 1 if this filesystem can be mounted as requested,
2652 * 0 if it cannot be.
2653 */
2654 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2655 {
2656 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2657 ext4_msg(sb, KERN_ERR,
2658 "Couldn't mount because of "
2659 "unsupported optional features (%x)",
2660 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2661 ~EXT4_FEATURE_INCOMPAT_SUPP));
2662 return 0;
2663 }
2664
2665 if (readonly)
2666 return 1;
2667
2668 /* Check that feature set is OK for a read-write mount */
2669 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2670 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2671 "unsupported optional features (%x)",
2672 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2673 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2674 return 0;
2675 }
2676 /*
2677 * Large file size enabled file system can only be mounted
2678 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2679 */
2680 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2681 if (sizeof(blkcnt_t) < sizeof(u64)) {
2682 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2683 "cannot be mounted RDWR without "
2684 "CONFIG_LBDAF");
2685 return 0;
2686 }
2687 }
2688 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2689 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2690 ext4_msg(sb, KERN_ERR,
2691 "Can't support bigalloc feature without "
2692 "extents feature\n");
2693 return 0;
2694 }
2695
2696 #ifndef CONFIG_QUOTA
2697 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2698 !readonly) {
2699 ext4_msg(sb, KERN_ERR,
2700 "Filesystem with quota feature cannot be mounted RDWR "
2701 "without CONFIG_QUOTA");
2702 return 0;
2703 }
2704 #endif /* CONFIG_QUOTA */
2705 return 1;
2706 }
2707
2708 /*
2709 * This function is called once a day if we have errors logged
2710 * on the file system
2711 */
2712 static void print_daily_error_info(unsigned long arg)
2713 {
2714 struct super_block *sb = (struct super_block *) arg;
2715 struct ext4_sb_info *sbi;
2716 struct ext4_super_block *es;
2717
2718 sbi = EXT4_SB(sb);
2719 es = sbi->s_es;
2720
2721 if (es->s_error_count)
2722 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2723 le32_to_cpu(es->s_error_count));
2724 if (es->s_first_error_time) {
2725 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2726 sb->s_id, le32_to_cpu(es->s_first_error_time),
2727 (int) sizeof(es->s_first_error_func),
2728 es->s_first_error_func,
2729 le32_to_cpu(es->s_first_error_line));
2730 if (es->s_first_error_ino)
2731 printk(": inode %u",
2732 le32_to_cpu(es->s_first_error_ino));
2733 if (es->s_first_error_block)
2734 printk(": block %llu", (unsigned long long)
2735 le64_to_cpu(es->s_first_error_block));
2736 printk("\n");
2737 }
2738 if (es->s_last_error_time) {
2739 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2740 sb->s_id, le32_to_cpu(es->s_last_error_time),
2741 (int) sizeof(es->s_last_error_func),
2742 es->s_last_error_func,
2743 le32_to_cpu(es->s_last_error_line));
2744 if (es->s_last_error_ino)
2745 printk(": inode %u",
2746 le32_to_cpu(es->s_last_error_ino));
2747 if (es->s_last_error_block)
2748 printk(": block %llu", (unsigned long long)
2749 le64_to_cpu(es->s_last_error_block));
2750 printk("\n");
2751 }
2752 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2753 }
2754
2755 /* Find next suitable group and run ext4_init_inode_table */
2756 static int ext4_run_li_request(struct ext4_li_request *elr)
2757 {
2758 struct ext4_group_desc *gdp = NULL;
2759 ext4_group_t group, ngroups;
2760 struct super_block *sb;
2761 unsigned long timeout = 0;
2762 int ret = 0;
2763
2764 sb = elr->lr_super;
2765 ngroups = EXT4_SB(sb)->s_groups_count;
2766
2767 sb_start_write(sb);
2768 for (group = elr->lr_next_group; group < ngroups; group++) {
2769 gdp = ext4_get_group_desc(sb, group, NULL);
2770 if (!gdp) {
2771 ret = 1;
2772 break;
2773 }
2774
2775 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2776 break;
2777 }
2778
2779 if (group >= ngroups)
2780 ret = 1;
2781
2782 if (!ret) {
2783 timeout = jiffies;
2784 ret = ext4_init_inode_table(sb, group,
2785 elr->lr_timeout ? 0 : 1);
2786 if (elr->lr_timeout == 0) {
2787 timeout = (jiffies - timeout) *
2788 elr->lr_sbi->s_li_wait_mult;
2789 elr->lr_timeout = timeout;
2790 }
2791 elr->lr_next_sched = jiffies + elr->lr_timeout;
2792 elr->lr_next_group = group + 1;
2793 }
2794 sb_end_write(sb);
2795
2796 return ret;
2797 }
2798
2799 /*
2800 * Remove lr_request from the list_request and free the
2801 * request structure. Should be called with li_list_mtx held
2802 */
2803 static void ext4_remove_li_request(struct ext4_li_request *elr)
2804 {
2805 struct ext4_sb_info *sbi;
2806
2807 if (!elr)
2808 return;
2809
2810 sbi = elr->lr_sbi;
2811
2812 list_del(&elr->lr_request);
2813 sbi->s_li_request = NULL;
2814 kfree(elr);
2815 }
2816
2817 static void ext4_unregister_li_request(struct super_block *sb)
2818 {
2819 mutex_lock(&ext4_li_mtx);
2820 if (!ext4_li_info) {
2821 mutex_unlock(&ext4_li_mtx);
2822 return;
2823 }
2824
2825 mutex_lock(&ext4_li_info->li_list_mtx);
2826 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2827 mutex_unlock(&ext4_li_info->li_list_mtx);
2828 mutex_unlock(&ext4_li_mtx);
2829 }
2830
2831 static struct task_struct *ext4_lazyinit_task;
2832
2833 /*
2834 * This is the function where ext4lazyinit thread lives. It walks
2835 * through the request list searching for next scheduled filesystem.
2836 * When such a fs is found, run the lazy initialization request
2837 * (ext4_rn_li_request) and keep track of the time spend in this
2838 * function. Based on that time we compute next schedule time of
2839 * the request. When walking through the list is complete, compute
2840 * next waking time and put itself into sleep.
2841 */
2842 static int ext4_lazyinit_thread(void *arg)
2843 {
2844 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2845 struct list_head *pos, *n;
2846 struct ext4_li_request *elr;
2847 unsigned long next_wakeup, cur;
2848
2849 BUG_ON(NULL == eli);
2850
2851 cont_thread:
2852 while (true) {
2853 next_wakeup = MAX_JIFFY_OFFSET;
2854
2855 mutex_lock(&eli->li_list_mtx);
2856 if (list_empty(&eli->li_request_list)) {
2857 mutex_unlock(&eli->li_list_mtx);
2858 goto exit_thread;
2859 }
2860
2861 list_for_each_safe(pos, n, &eli->li_request_list) {
2862 elr = list_entry(pos, struct ext4_li_request,
2863 lr_request);
2864
2865 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2866 if (ext4_run_li_request(elr) != 0) {
2867 /* error, remove the lazy_init job */
2868 ext4_remove_li_request(elr);
2869 continue;
2870 }
2871 }
2872
2873 if (time_before(elr->lr_next_sched, next_wakeup))
2874 next_wakeup = elr->lr_next_sched;
2875 }
2876 mutex_unlock(&eli->li_list_mtx);
2877
2878 try_to_freeze();
2879
2880 cur = jiffies;
2881 if ((time_after_eq(cur, next_wakeup)) ||
2882 (MAX_JIFFY_OFFSET == next_wakeup)) {
2883 cond_resched();
2884 continue;
2885 }
2886
2887 schedule_timeout_interruptible(next_wakeup - cur);
2888
2889 if (kthread_should_stop()) {
2890 ext4_clear_request_list();
2891 goto exit_thread;
2892 }
2893 }
2894
2895 exit_thread:
2896 /*
2897 * It looks like the request list is empty, but we need
2898 * to check it under the li_list_mtx lock, to prevent any
2899 * additions into it, and of course we should lock ext4_li_mtx
2900 * to atomically free the list and ext4_li_info, because at
2901 * this point another ext4 filesystem could be registering
2902 * new one.
2903 */
2904 mutex_lock(&ext4_li_mtx);
2905 mutex_lock(&eli->li_list_mtx);
2906 if (!list_empty(&eli->li_request_list)) {
2907 mutex_unlock(&eli->li_list_mtx);
2908 mutex_unlock(&ext4_li_mtx);
2909 goto cont_thread;
2910 }
2911 mutex_unlock(&eli->li_list_mtx);
2912 kfree(ext4_li_info);
2913 ext4_li_info = NULL;
2914 mutex_unlock(&ext4_li_mtx);
2915
2916 return 0;
2917 }
2918
2919 static void ext4_clear_request_list(void)
2920 {
2921 struct list_head *pos, *n;
2922 struct ext4_li_request *elr;
2923
2924 mutex_lock(&ext4_li_info->li_list_mtx);
2925 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2926 elr = list_entry(pos, struct ext4_li_request,
2927 lr_request);
2928 ext4_remove_li_request(elr);
2929 }
2930 mutex_unlock(&ext4_li_info->li_list_mtx);
2931 }
2932
2933 static int ext4_run_lazyinit_thread(void)
2934 {
2935 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2936 ext4_li_info, "ext4lazyinit");
2937 if (IS_ERR(ext4_lazyinit_task)) {
2938 int err = PTR_ERR(ext4_lazyinit_task);
2939 ext4_clear_request_list();
2940 kfree(ext4_li_info);
2941 ext4_li_info = NULL;
2942 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2943 "initialization thread\n",
2944 err);
2945 return err;
2946 }
2947 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2948 return 0;
2949 }
2950
2951 /*
2952 * Check whether it make sense to run itable init. thread or not.
2953 * If there is at least one uninitialized inode table, return
2954 * corresponding group number, else the loop goes through all
2955 * groups and return total number of groups.
2956 */
2957 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2958 {
2959 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2960 struct ext4_group_desc *gdp = NULL;
2961
2962 for (group = 0; group < ngroups; group++) {
2963 gdp = ext4_get_group_desc(sb, group, NULL);
2964 if (!gdp)
2965 continue;
2966
2967 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2968 break;
2969 }
2970
2971 return group;
2972 }
2973
2974 static int ext4_li_info_new(void)
2975 {
2976 struct ext4_lazy_init *eli = NULL;
2977
2978 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2979 if (!eli)
2980 return -ENOMEM;
2981
2982 INIT_LIST_HEAD(&eli->li_request_list);
2983 mutex_init(&eli->li_list_mtx);
2984
2985 eli->li_state |= EXT4_LAZYINIT_QUIT;
2986
2987 ext4_li_info = eli;
2988
2989 return 0;
2990 }
2991
2992 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2993 ext4_group_t start)
2994 {
2995 struct ext4_sb_info *sbi = EXT4_SB(sb);
2996 struct ext4_li_request *elr;
2997 unsigned long rnd;
2998
2999 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3000 if (!elr)
3001 return NULL;
3002
3003 elr->lr_super = sb;
3004 elr->lr_sbi = sbi;
3005 elr->lr_next_group = start;
3006
3007 /*
3008 * Randomize first schedule time of the request to
3009 * spread the inode table initialization requests
3010 * better.
3011 */
3012 get_random_bytes(&rnd, sizeof(rnd));
3013 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3014 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3015
3016 return elr;
3017 }
3018
3019 int ext4_register_li_request(struct super_block *sb,
3020 ext4_group_t first_not_zeroed)
3021 {
3022 struct ext4_sb_info *sbi = EXT4_SB(sb);
3023 struct ext4_li_request *elr = NULL;
3024 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3025 int ret = 0;
3026
3027 mutex_lock(&ext4_li_mtx);
3028 if (sbi->s_li_request != NULL) {
3029 /*
3030 * Reset timeout so it can be computed again, because
3031 * s_li_wait_mult might have changed.
3032 */
3033 sbi->s_li_request->lr_timeout = 0;
3034 goto out;
3035 }
3036
3037 if (first_not_zeroed == ngroups ||
3038 (sb->s_flags & MS_RDONLY) ||
3039 !test_opt(sb, INIT_INODE_TABLE))
3040 goto out;
3041
3042 elr = ext4_li_request_new(sb, first_not_zeroed);
3043 if (!elr) {
3044 ret = -ENOMEM;
3045 goto out;
3046 }
3047
3048 if (NULL == ext4_li_info) {
3049 ret = ext4_li_info_new();
3050 if (ret)
3051 goto out;
3052 }
3053
3054 mutex_lock(&ext4_li_info->li_list_mtx);
3055 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3056 mutex_unlock(&ext4_li_info->li_list_mtx);
3057
3058 sbi->s_li_request = elr;
3059 /*
3060 * set elr to NULL here since it has been inserted to
3061 * the request_list and the removal and free of it is
3062 * handled by ext4_clear_request_list from now on.
3063 */
3064 elr = NULL;
3065
3066 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3067 ret = ext4_run_lazyinit_thread();
3068 if (ret)
3069 goto out;
3070 }
3071 out:
3072 mutex_unlock(&ext4_li_mtx);
3073 if (ret)
3074 kfree(elr);
3075 return ret;
3076 }
3077
3078 /*
3079 * We do not need to lock anything since this is called on
3080 * module unload.
3081 */
3082 static void ext4_destroy_lazyinit_thread(void)
3083 {
3084 /*
3085 * If thread exited earlier
3086 * there's nothing to be done.
3087 */
3088 if (!ext4_li_info || !ext4_lazyinit_task)
3089 return;
3090
3091 kthread_stop(ext4_lazyinit_task);
3092 }
3093
3094 static int set_journal_csum_feature_set(struct super_block *sb)
3095 {
3096 int ret = 1;
3097 int compat, incompat;
3098 struct ext4_sb_info *sbi = EXT4_SB(sb);
3099
3100 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3101 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3102 /* journal checksum v2 */
3103 compat = 0;
3104 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3105 } else {
3106 /* journal checksum v1 */
3107 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3108 incompat = 0;
3109 }
3110
3111 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3112 ret = jbd2_journal_set_features(sbi->s_journal,
3113 compat, 0,
3114 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3115 incompat);
3116 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3117 ret = jbd2_journal_set_features(sbi->s_journal,
3118 compat, 0,
3119 incompat);
3120 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3121 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3122 } else {
3123 jbd2_journal_clear_features(sbi->s_journal,
3124 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3125 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3126 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3127 }
3128
3129 return ret;
3130 }
3131
3132 /*
3133 * Note: calculating the overhead so we can be compatible with
3134 * historical BSD practice is quite difficult in the face of
3135 * clusters/bigalloc. This is because multiple metadata blocks from
3136 * different block group can end up in the same allocation cluster.
3137 * Calculating the exact overhead in the face of clustered allocation
3138 * requires either O(all block bitmaps) in memory or O(number of block
3139 * groups**2) in time. We will still calculate the superblock for
3140 * older file systems --- and if we come across with a bigalloc file
3141 * system with zero in s_overhead_clusters the estimate will be close to
3142 * correct especially for very large cluster sizes --- but for newer
3143 * file systems, it's better to calculate this figure once at mkfs
3144 * time, and store it in the superblock. If the superblock value is
3145 * present (even for non-bigalloc file systems), we will use it.
3146 */
3147 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3148 char *buf)
3149 {
3150 struct ext4_sb_info *sbi = EXT4_SB(sb);
3151 struct ext4_group_desc *gdp;
3152 ext4_fsblk_t first_block, last_block, b;
3153 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3154 int s, j, count = 0;
3155
3156 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3157 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3158 sbi->s_itb_per_group + 2);
3159
3160 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3161 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3162 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3163 for (i = 0; i < ngroups; i++) {
3164 gdp = ext4_get_group_desc(sb, i, NULL);
3165 b = ext4_block_bitmap(sb, gdp);
3166 if (b >= first_block && b <= last_block) {
3167 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3168 count++;
3169 }
3170 b = ext4_inode_bitmap(sb, gdp);
3171 if (b >= first_block && b <= last_block) {
3172 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3173 count++;
3174 }
3175 b = ext4_inode_table(sb, gdp);
3176 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3177 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3178 int c = EXT4_B2C(sbi, b - first_block);
3179 ext4_set_bit(c, buf);
3180 count++;
3181 }
3182 if (i != grp)
3183 continue;
3184 s = 0;
3185 if (ext4_bg_has_super(sb, grp)) {
3186 ext4_set_bit(s++, buf);
3187 count++;
3188 }
3189 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3190 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3191 count++;
3192 }
3193 }
3194 if (!count)
3195 return 0;
3196 return EXT4_CLUSTERS_PER_GROUP(sb) -
3197 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3198 }
3199
3200 /*
3201 * Compute the overhead and stash it in sbi->s_overhead
3202 */
3203 int ext4_calculate_overhead(struct super_block *sb)
3204 {
3205 struct ext4_sb_info *sbi = EXT4_SB(sb);
3206 struct ext4_super_block *es = sbi->s_es;
3207 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3208 ext4_fsblk_t overhead = 0;
3209 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3210
3211 if (!buf)
3212 return -ENOMEM;
3213
3214 /*
3215 * Compute the overhead (FS structures). This is constant
3216 * for a given filesystem unless the number of block groups
3217 * changes so we cache the previous value until it does.
3218 */
3219
3220 /*
3221 * All of the blocks before first_data_block are overhead
3222 */
3223 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3224
3225 /*
3226 * Add the overhead found in each block group
3227 */
3228 for (i = 0; i < ngroups; i++) {
3229 int blks;
3230
3231 blks = count_overhead(sb, i, buf);
3232 overhead += blks;
3233 if (blks)
3234 memset(buf, 0, PAGE_SIZE);
3235 cond_resched();
3236 }
3237 /* Add the journal blocks as well */
3238 if (sbi->s_journal)
3239 overhead += EXT4_B2C(sbi, sbi->s_journal->j_maxlen);
3240
3241 sbi->s_overhead = overhead;
3242 smp_wmb();
3243 free_page((unsigned long) buf);
3244 return 0;
3245 }
3246
3247 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3248 {
3249 char *orig_data = kstrdup(data, GFP_KERNEL);
3250 struct buffer_head *bh;
3251 struct ext4_super_block *es = NULL;
3252 struct ext4_sb_info *sbi;
3253 ext4_fsblk_t block;
3254 ext4_fsblk_t sb_block = get_sb_block(&data);
3255 ext4_fsblk_t logical_sb_block;
3256 unsigned long offset = 0;
3257 unsigned long journal_devnum = 0;
3258 unsigned long def_mount_opts;
3259 struct inode *root;
3260 char *cp;
3261 const char *descr;
3262 int ret = -ENOMEM;
3263 int blocksize, clustersize;
3264 unsigned int db_count;
3265 unsigned int i;
3266 int needs_recovery, has_huge_files, has_bigalloc;
3267 __u64 blocks_count;
3268 int err = 0;
3269 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3270 ext4_group_t first_not_zeroed;
3271
3272 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3273 if (!sbi)
3274 goto out_free_orig;
3275
3276 sbi->s_blockgroup_lock =
3277 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3278 if (!sbi->s_blockgroup_lock) {
3279 kfree(sbi);
3280 goto out_free_orig;
3281 }
3282 sb->s_fs_info = sbi;
3283 sbi->s_sb = sb;
3284 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3285 sbi->s_sb_block = sb_block;
3286 if (sb->s_bdev->bd_part)
3287 sbi->s_sectors_written_start =
3288 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3289
3290 /* Cleanup superblock name */
3291 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3292 *cp = '!';
3293
3294 /* -EINVAL is default */
3295 ret = -EINVAL;
3296 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3297 if (!blocksize) {
3298 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3299 goto out_fail;
3300 }
3301
3302 /*
3303 * The ext4 superblock will not be buffer aligned for other than 1kB
3304 * block sizes. We need to calculate the offset from buffer start.
3305 */
3306 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3307 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3308 offset = do_div(logical_sb_block, blocksize);
3309 } else {
3310 logical_sb_block = sb_block;
3311 }
3312
3313 if (!(bh = sb_bread(sb, logical_sb_block))) {
3314 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3315 goto out_fail;
3316 }
3317 /*
3318 * Note: s_es must be initialized as soon as possible because
3319 * some ext4 macro-instructions depend on its value
3320 */
3321 es = (struct ext4_super_block *) (bh->b_data + offset);
3322 sbi->s_es = es;
3323 sb->s_magic = le16_to_cpu(es->s_magic);
3324 if (sb->s_magic != EXT4_SUPER_MAGIC)
3325 goto cantfind_ext4;
3326 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3327
3328 /* Warn if metadata_csum and gdt_csum are both set. */
3329 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3330 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3331 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3332 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3333 "redundant flags; please run fsck.");
3334
3335 /* Check for a known checksum algorithm */
3336 if (!ext4_verify_csum_type(sb, es)) {
3337 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3338 "unknown checksum algorithm.");
3339 silent = 1;
3340 goto cantfind_ext4;
3341 }
3342
3343 /* Load the checksum driver */
3344 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3345 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3346 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3347 if (IS_ERR(sbi->s_chksum_driver)) {
3348 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3349 ret = PTR_ERR(sbi->s_chksum_driver);
3350 sbi->s_chksum_driver = NULL;
3351 goto failed_mount;
3352 }
3353 }
3354
3355 /* Check superblock checksum */
3356 if (!ext4_superblock_csum_verify(sb, es)) {
3357 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3358 "invalid superblock checksum. Run e2fsck?");
3359 silent = 1;
3360 goto cantfind_ext4;
3361 }
3362
3363 /* Precompute checksum seed for all metadata */
3364 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3365 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3366 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3367 sizeof(es->s_uuid));
3368
3369 /* Set defaults before we parse the mount options */
3370 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3371 set_opt(sb, INIT_INODE_TABLE);
3372 if (def_mount_opts & EXT4_DEFM_DEBUG)
3373 set_opt(sb, DEBUG);
3374 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3375 set_opt(sb, GRPID);
3376 if (def_mount_opts & EXT4_DEFM_UID16)
3377 set_opt(sb, NO_UID32);
3378 /* xattr user namespace & acls are now defaulted on */
3379 set_opt(sb, XATTR_USER);
3380 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3381 set_opt(sb, POSIX_ACL);
3382 #endif
3383 set_opt(sb, MBLK_IO_SUBMIT);
3384 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3385 set_opt(sb, JOURNAL_DATA);
3386 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3387 set_opt(sb, ORDERED_DATA);
3388 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3389 set_opt(sb, WRITEBACK_DATA);
3390
3391 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3392 set_opt(sb, ERRORS_PANIC);
3393 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3394 set_opt(sb, ERRORS_CONT);
3395 else
3396 set_opt(sb, ERRORS_RO);
3397 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3398 set_opt(sb, BLOCK_VALIDITY);
3399 if (def_mount_opts & EXT4_DEFM_DISCARD)
3400 set_opt(sb, DISCARD);
3401
3402 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3403 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3404 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3405 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3406 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3407
3408 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3409 set_opt(sb, BARRIER);
3410
3411 /*
3412 * enable delayed allocation by default
3413 * Use -o nodelalloc to turn it off
3414 */
3415 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3416 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3417 set_opt(sb, DELALLOC);
3418
3419 /*
3420 * set default s_li_wait_mult for lazyinit, for the case there is
3421 * no mount option specified.
3422 */
3423 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3424
3425 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3426 &journal_devnum, &journal_ioprio, 0)) {
3427 ext4_msg(sb, KERN_WARNING,
3428 "failed to parse options in superblock: %s",
3429 sbi->s_es->s_mount_opts);
3430 }
3431 sbi->s_def_mount_opt = sbi->s_mount_opt;
3432 if (!parse_options((char *) data, sb, &journal_devnum,
3433 &journal_ioprio, 0))
3434 goto failed_mount;
3435
3436 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3437 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3438 "with data=journal disables delayed "
3439 "allocation and O_DIRECT support!\n");
3440 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3441 ext4_msg(sb, KERN_ERR, "can't mount with "
3442 "both data=journal and delalloc");
3443 goto failed_mount;
3444 }
3445 if (test_opt(sb, DIOREAD_NOLOCK)) {
3446 ext4_msg(sb, KERN_ERR, "can't mount with "
3447 "both data=journal and delalloc");
3448 goto failed_mount;
3449 }
3450 if (test_opt(sb, DELALLOC))
3451 clear_opt(sb, DELALLOC);
3452 }
3453
3454 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3455 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3456
3457 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3458 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3459 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3460 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3461 ext4_msg(sb, KERN_WARNING,
3462 "feature flags set on rev 0 fs, "
3463 "running e2fsck is recommended");
3464
3465 if (IS_EXT2_SB(sb)) {
3466 if (ext2_feature_set_ok(sb))
3467 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3468 "using the ext4 subsystem");
3469 else {
3470 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3471 "to feature incompatibilities");
3472 goto failed_mount;
3473 }
3474 }
3475
3476 if (IS_EXT3_SB(sb)) {
3477 if (ext3_feature_set_ok(sb))
3478 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3479 "using the ext4 subsystem");
3480 else {
3481 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3482 "to feature incompatibilities");
3483 goto failed_mount;
3484 }
3485 }
3486
3487 /*
3488 * Check feature flags regardless of the revision level, since we
3489 * previously didn't change the revision level when setting the flags,
3490 * so there is a chance incompat flags are set on a rev 0 filesystem.
3491 */
3492 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3493 goto failed_mount;
3494
3495 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3496 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3497 blocksize > EXT4_MAX_BLOCK_SIZE) {
3498 ext4_msg(sb, KERN_ERR,
3499 "Unsupported filesystem blocksize %d", blocksize);
3500 goto failed_mount;
3501 }
3502
3503 if (sb->s_blocksize != blocksize) {
3504 /* Validate the filesystem blocksize */
3505 if (!sb_set_blocksize(sb, blocksize)) {
3506 ext4_msg(sb, KERN_ERR, "bad block size %d",
3507 blocksize);
3508 goto failed_mount;
3509 }
3510
3511 brelse(bh);
3512 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3513 offset = do_div(logical_sb_block, blocksize);
3514 bh = sb_bread(sb, logical_sb_block);
3515 if (!bh) {
3516 ext4_msg(sb, KERN_ERR,
3517 "Can't read superblock on 2nd try");
3518 goto failed_mount;
3519 }
3520 es = (struct ext4_super_block *)(bh->b_data + offset);
3521 sbi->s_es = es;
3522 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3523 ext4_msg(sb, KERN_ERR,
3524 "Magic mismatch, very weird!");
3525 goto failed_mount;
3526 }
3527 }
3528
3529 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3530 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3531 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3532 has_huge_files);
3533 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3534
3535 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3536 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3537 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3538 } else {
3539 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3540 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3541 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3542 (!is_power_of_2(sbi->s_inode_size)) ||
3543 (sbi->s_inode_size > blocksize)) {
3544 ext4_msg(sb, KERN_ERR,
3545 "unsupported inode size: %d",
3546 sbi->s_inode_size);
3547 goto failed_mount;
3548 }
3549 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3550 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3551 }
3552
3553 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3554 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3555 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3556 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3557 !is_power_of_2(sbi->s_desc_size)) {
3558 ext4_msg(sb, KERN_ERR,
3559 "unsupported descriptor size %lu",
3560 sbi->s_desc_size);
3561 goto failed_mount;
3562 }
3563 } else
3564 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3565
3566 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3567 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3568 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3569 goto cantfind_ext4;
3570
3571 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3572 if (sbi->s_inodes_per_block == 0)
3573 goto cantfind_ext4;
3574 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3575 sbi->s_inodes_per_block;
3576 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3577 sbi->s_sbh = bh;
3578 sbi->s_mount_state = le16_to_cpu(es->s_state);
3579 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3580 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3581
3582 for (i = 0; i < 4; i++)
3583 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3584 sbi->s_def_hash_version = es->s_def_hash_version;
3585 i = le32_to_cpu(es->s_flags);
3586 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3587 sbi->s_hash_unsigned = 3;
3588 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3589 #ifdef __CHAR_UNSIGNED__
3590 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3591 sbi->s_hash_unsigned = 3;
3592 #else
3593 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3594 #endif
3595 }
3596
3597 /* Handle clustersize */
3598 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3599 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3600 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3601 if (has_bigalloc) {
3602 if (clustersize < blocksize) {
3603 ext4_msg(sb, KERN_ERR,
3604 "cluster size (%d) smaller than "
3605 "block size (%d)", clustersize, blocksize);
3606 goto failed_mount;
3607 }
3608 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3609 le32_to_cpu(es->s_log_block_size);
3610 sbi->s_clusters_per_group =
3611 le32_to_cpu(es->s_clusters_per_group);
3612 if (sbi->s_clusters_per_group > blocksize * 8) {
3613 ext4_msg(sb, KERN_ERR,
3614 "#clusters per group too big: %lu",
3615 sbi->s_clusters_per_group);
3616 goto failed_mount;
3617 }
3618 if (sbi->s_blocks_per_group !=
3619 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3620 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3621 "clusters per group (%lu) inconsistent",
3622 sbi->s_blocks_per_group,
3623 sbi->s_clusters_per_group);
3624 goto failed_mount;
3625 }
3626 } else {
3627 if (clustersize != blocksize) {
3628 ext4_warning(sb, "fragment/cluster size (%d) != "
3629 "block size (%d)", clustersize,
3630 blocksize);
3631 clustersize = blocksize;
3632 }
3633 if (sbi->s_blocks_per_group > blocksize * 8) {
3634 ext4_msg(sb, KERN_ERR,
3635 "#blocks per group too big: %lu",
3636 sbi->s_blocks_per_group);
3637 goto failed_mount;
3638 }
3639 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3640 sbi->s_cluster_bits = 0;
3641 }
3642 sbi->s_cluster_ratio = clustersize / blocksize;
3643
3644 if (sbi->s_inodes_per_group > blocksize * 8) {
3645 ext4_msg(sb, KERN_ERR,
3646 "#inodes per group too big: %lu",
3647 sbi->s_inodes_per_group);
3648 goto failed_mount;
3649 }
3650
3651 /*
3652 * Test whether we have more sectors than will fit in sector_t,
3653 * and whether the max offset is addressable by the page cache.
3654 */
3655 err = generic_check_addressable(sb->s_blocksize_bits,
3656 ext4_blocks_count(es));
3657 if (err) {
3658 ext4_msg(sb, KERN_ERR, "filesystem"
3659 " too large to mount safely on this system");
3660 if (sizeof(sector_t) < 8)
3661 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3662 goto failed_mount;
3663 }
3664
3665 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3666 goto cantfind_ext4;
3667
3668 /* check blocks count against device size */
3669 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3670 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3671 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3672 "exceeds size of device (%llu blocks)",
3673 ext4_blocks_count(es), blocks_count);
3674 goto failed_mount;
3675 }
3676
3677 /*
3678 * It makes no sense for the first data block to be beyond the end
3679 * of the filesystem.
3680 */
3681 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3682 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3683 "block %u is beyond end of filesystem (%llu)",
3684 le32_to_cpu(es->s_first_data_block),
3685 ext4_blocks_count(es));
3686 goto failed_mount;
3687 }
3688 blocks_count = (ext4_blocks_count(es) -
3689 le32_to_cpu(es->s_first_data_block) +
3690 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3691 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3692 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3693 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3694 "(block count %llu, first data block %u, "
3695 "blocks per group %lu)", sbi->s_groups_count,
3696 ext4_blocks_count(es),
3697 le32_to_cpu(es->s_first_data_block),
3698 EXT4_BLOCKS_PER_GROUP(sb));
3699 goto failed_mount;
3700 }
3701 sbi->s_groups_count = blocks_count;
3702 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3703 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3704 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3705 EXT4_DESC_PER_BLOCK(sb);
3706 sbi->s_group_desc = ext4_kvmalloc(db_count *
3707 sizeof(struct buffer_head *),
3708 GFP_KERNEL);
3709 if (sbi->s_group_desc == NULL) {
3710 ext4_msg(sb, KERN_ERR, "not enough memory");
3711 ret = -ENOMEM;
3712 goto failed_mount;
3713 }
3714
3715 if (ext4_proc_root)
3716 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3717
3718 if (sbi->s_proc)
3719 proc_create_data("options", S_IRUGO, sbi->s_proc,
3720 &ext4_seq_options_fops, sb);
3721
3722 bgl_lock_init(sbi->s_blockgroup_lock);
3723
3724 for (i = 0; i < db_count; i++) {
3725 block = descriptor_loc(sb, logical_sb_block, i);
3726 sbi->s_group_desc[i] = sb_bread(sb, block);
3727 if (!sbi->s_group_desc[i]) {
3728 ext4_msg(sb, KERN_ERR,
3729 "can't read group descriptor %d", i);
3730 db_count = i;
3731 goto failed_mount2;
3732 }
3733 }
3734 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3735 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3736 goto failed_mount2;
3737 }
3738 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3739 if (!ext4_fill_flex_info(sb)) {
3740 ext4_msg(sb, KERN_ERR,
3741 "unable to initialize "
3742 "flex_bg meta info!");
3743 goto failed_mount2;
3744 }
3745
3746 sbi->s_gdb_count = db_count;
3747 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3748 spin_lock_init(&sbi->s_next_gen_lock);
3749
3750 init_timer(&sbi->s_err_report);
3751 sbi->s_err_report.function = print_daily_error_info;
3752 sbi->s_err_report.data = (unsigned long) sb;
3753
3754 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3755 ext4_count_free_clusters(sb));
3756 if (!err) {
3757 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3758 ext4_count_free_inodes(sb));
3759 }
3760 if (!err) {
3761 err = percpu_counter_init(&sbi->s_dirs_counter,
3762 ext4_count_dirs(sb));
3763 }
3764 if (!err) {
3765 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3766 }
3767 if (err) {
3768 ext4_msg(sb, KERN_ERR, "insufficient memory");
3769 goto failed_mount3;
3770 }
3771
3772 sbi->s_stripe = ext4_get_stripe_size(sbi);
3773 sbi->s_max_writeback_mb_bump = 128;
3774 sbi->s_extent_max_zeroout_kb = 32;
3775
3776 /*
3777 * set up enough so that it can read an inode
3778 */
3779 if (!test_opt(sb, NOLOAD) &&
3780 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3781 sb->s_op = &ext4_sops;
3782 else
3783 sb->s_op = &ext4_nojournal_sops;
3784 sb->s_export_op = &ext4_export_ops;
3785 sb->s_xattr = ext4_xattr_handlers;
3786 #ifdef CONFIG_QUOTA
3787 sb->s_qcop = &ext4_qctl_operations;
3788 sb->dq_op = &ext4_quota_operations;
3789
3790 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3791 /* Use qctl operations for hidden quota files. */
3792 sb->s_qcop = &ext4_qctl_sysfile_operations;
3793 }
3794 #endif
3795 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3796
3797 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3798 mutex_init(&sbi->s_orphan_lock);
3799
3800 sb->s_root = NULL;
3801
3802 needs_recovery = (es->s_last_orphan != 0 ||
3803 EXT4_HAS_INCOMPAT_FEATURE(sb,
3804 EXT4_FEATURE_INCOMPAT_RECOVER));
3805
3806 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3807 !(sb->s_flags & MS_RDONLY))
3808 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3809 goto failed_mount3;
3810
3811 /*
3812 * The first inode we look at is the journal inode. Don't try
3813 * root first: it may be modified in the journal!
3814 */
3815 if (!test_opt(sb, NOLOAD) &&
3816 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3817 if (ext4_load_journal(sb, es, journal_devnum))
3818 goto failed_mount3;
3819 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3820 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3821 ext4_msg(sb, KERN_ERR, "required journal recovery "
3822 "suppressed and not mounted read-only");
3823 goto failed_mount_wq;
3824 } else {
3825 clear_opt(sb, DATA_FLAGS);
3826 sbi->s_journal = NULL;
3827 needs_recovery = 0;
3828 goto no_journal;
3829 }
3830
3831 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3832 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3833 JBD2_FEATURE_INCOMPAT_64BIT)) {
3834 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3835 goto failed_mount_wq;
3836 }
3837
3838 if (!set_journal_csum_feature_set(sb)) {
3839 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3840 "feature set");
3841 goto failed_mount_wq;
3842 }
3843
3844 /* We have now updated the journal if required, so we can
3845 * validate the data journaling mode. */
3846 switch (test_opt(sb, DATA_FLAGS)) {
3847 case 0:
3848 /* No mode set, assume a default based on the journal
3849 * capabilities: ORDERED_DATA if the journal can
3850 * cope, else JOURNAL_DATA
3851 */
3852 if (jbd2_journal_check_available_features
3853 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3854 set_opt(sb, ORDERED_DATA);
3855 else
3856 set_opt(sb, JOURNAL_DATA);
3857 break;
3858
3859 case EXT4_MOUNT_ORDERED_DATA:
3860 case EXT4_MOUNT_WRITEBACK_DATA:
3861 if (!jbd2_journal_check_available_features
3862 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3863 ext4_msg(sb, KERN_ERR, "Journal does not support "
3864 "requested data journaling mode");
3865 goto failed_mount_wq;
3866 }
3867 default:
3868 break;
3869 }
3870 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3871
3872 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3873
3874 /*
3875 * The journal may have updated the bg summary counts, so we
3876 * need to update the global counters.
3877 */
3878 percpu_counter_set(&sbi->s_freeclusters_counter,
3879 ext4_count_free_clusters(sb));
3880 percpu_counter_set(&sbi->s_freeinodes_counter,
3881 ext4_count_free_inodes(sb));
3882 percpu_counter_set(&sbi->s_dirs_counter,
3883 ext4_count_dirs(sb));
3884 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3885
3886 no_journal:
3887 /*
3888 * Get the # of file system overhead blocks from the
3889 * superblock if present.
3890 */
3891 if (es->s_overhead_clusters)
3892 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3893 else {
3894 err = ext4_calculate_overhead(sb);
3895 if (err)
3896 goto failed_mount_wq;
3897 }
3898
3899 /*
3900 * The maximum number of concurrent works can be high and
3901 * concurrency isn't really necessary. Limit it to 1.
3902 */
3903 EXT4_SB(sb)->dio_unwritten_wq =
3904 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3905 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3906 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3907 ret = -ENOMEM;
3908 goto failed_mount_wq;
3909 }
3910
3911 /*
3912 * The jbd2_journal_load will have done any necessary log recovery,
3913 * so we can safely mount the rest of the filesystem now.
3914 */
3915
3916 root = ext4_iget(sb, EXT4_ROOT_INO);
3917 if (IS_ERR(root)) {
3918 ext4_msg(sb, KERN_ERR, "get root inode failed");
3919 ret = PTR_ERR(root);
3920 root = NULL;
3921 goto failed_mount4;
3922 }
3923 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3924 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3925 iput(root);
3926 goto failed_mount4;
3927 }
3928 sb->s_root = d_make_root(root);
3929 if (!sb->s_root) {
3930 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3931 ret = -ENOMEM;
3932 goto failed_mount4;
3933 }
3934
3935 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3936 sb->s_flags |= MS_RDONLY;
3937
3938 /* determine the minimum size of new large inodes, if present */
3939 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3940 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3941 EXT4_GOOD_OLD_INODE_SIZE;
3942 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3943 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3944 if (sbi->s_want_extra_isize <
3945 le16_to_cpu(es->s_want_extra_isize))
3946 sbi->s_want_extra_isize =
3947 le16_to_cpu(es->s_want_extra_isize);
3948 if (sbi->s_want_extra_isize <
3949 le16_to_cpu(es->s_min_extra_isize))
3950 sbi->s_want_extra_isize =
3951 le16_to_cpu(es->s_min_extra_isize);
3952 }
3953 }
3954 /* Check if enough inode space is available */
3955 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3956 sbi->s_inode_size) {
3957 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3958 EXT4_GOOD_OLD_INODE_SIZE;
3959 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3960 "available");
3961 }
3962
3963 err = ext4_setup_system_zone(sb);
3964 if (err) {
3965 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3966 "zone (%d)", err);
3967 goto failed_mount4a;
3968 }
3969
3970 ext4_ext_init(sb);
3971 err = ext4_mb_init(sb);
3972 if (err) {
3973 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3974 err);
3975 goto failed_mount5;
3976 }
3977
3978 err = ext4_register_li_request(sb, first_not_zeroed);
3979 if (err)
3980 goto failed_mount6;
3981
3982 sbi->s_kobj.kset = ext4_kset;
3983 init_completion(&sbi->s_kobj_unregister);
3984 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3985 "%s", sb->s_id);
3986 if (err)
3987 goto failed_mount7;
3988
3989 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3990 ext4_orphan_cleanup(sb, es);
3991 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3992 if (needs_recovery) {
3993 ext4_msg(sb, KERN_INFO, "recovery complete");
3994 ext4_mark_recovery_complete(sb, es);
3995 }
3996 if (EXT4_SB(sb)->s_journal) {
3997 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3998 descr = " journalled data mode";
3999 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4000 descr = " ordered data mode";
4001 else
4002 descr = " writeback data mode";
4003 } else
4004 descr = "out journal";
4005
4006 #ifdef CONFIG_QUOTA
4007 /* Enable quota usage during mount. */
4008 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4009 !(sb->s_flags & MS_RDONLY)) {
4010 err = ext4_enable_quotas(sb);
4011 if (err)
4012 goto failed_mount7;
4013 }
4014 #endif /* CONFIG_QUOTA */
4015
4016 if (test_opt(sb, DISCARD)) {
4017 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4018 if (!blk_queue_discard(q))
4019 ext4_msg(sb, KERN_WARNING,
4020 "mounting with \"discard\" option, but "
4021 "the device does not support discard");
4022 }
4023
4024 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4025 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4026 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4027
4028 if (es->s_error_count)
4029 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4030
4031 kfree(orig_data);
4032 return 0;
4033
4034 cantfind_ext4:
4035 if (!silent)
4036 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4037 goto failed_mount;
4038
4039 failed_mount7:
4040 ext4_unregister_li_request(sb);
4041 failed_mount6:
4042 ext4_mb_release(sb);
4043 failed_mount5:
4044 ext4_ext_release(sb);
4045 ext4_release_system_zone(sb);
4046 failed_mount4a:
4047 dput(sb->s_root);
4048 sb->s_root = NULL;
4049 failed_mount4:
4050 ext4_msg(sb, KERN_ERR, "mount failed");
4051 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4052 failed_mount_wq:
4053 if (sbi->s_journal) {
4054 jbd2_journal_destroy(sbi->s_journal);
4055 sbi->s_journal = NULL;
4056 }
4057 failed_mount3:
4058 del_timer(&sbi->s_err_report);
4059 if (sbi->s_flex_groups)
4060 ext4_kvfree(sbi->s_flex_groups);
4061 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4062 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4063 percpu_counter_destroy(&sbi->s_dirs_counter);
4064 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4065 if (sbi->s_mmp_tsk)
4066 kthread_stop(sbi->s_mmp_tsk);
4067 failed_mount2:
4068 for (i = 0; i < db_count; i++)
4069 brelse(sbi->s_group_desc[i]);
4070 ext4_kvfree(sbi->s_group_desc);
4071 failed_mount:
4072 if (sbi->s_chksum_driver)
4073 crypto_free_shash(sbi->s_chksum_driver);
4074 if (sbi->s_proc) {
4075 remove_proc_entry("options", sbi->s_proc);
4076 remove_proc_entry(sb->s_id, ext4_proc_root);
4077 }
4078 #ifdef CONFIG_QUOTA
4079 for (i = 0; i < MAXQUOTAS; i++)
4080 kfree(sbi->s_qf_names[i]);
4081 #endif
4082 ext4_blkdev_remove(sbi);
4083 brelse(bh);
4084 out_fail:
4085 sb->s_fs_info = NULL;
4086 kfree(sbi->s_blockgroup_lock);
4087 kfree(sbi);
4088 out_free_orig:
4089 kfree(orig_data);
4090 return err ? err : ret;
4091 }
4092
4093 /*
4094 * Setup any per-fs journal parameters now. We'll do this both on
4095 * initial mount, once the journal has been initialised but before we've
4096 * done any recovery; and again on any subsequent remount.
4097 */
4098 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4099 {
4100 struct ext4_sb_info *sbi = EXT4_SB(sb);
4101
4102 journal->j_commit_interval = sbi->s_commit_interval;
4103 journal->j_min_batch_time = sbi->s_min_batch_time;
4104 journal->j_max_batch_time = sbi->s_max_batch_time;
4105
4106 write_lock(&journal->j_state_lock);
4107 if (test_opt(sb, BARRIER))
4108 journal->j_flags |= JBD2_BARRIER;
4109 else
4110 journal->j_flags &= ~JBD2_BARRIER;
4111 if (test_opt(sb, DATA_ERR_ABORT))
4112 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4113 else
4114 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4115 write_unlock(&journal->j_state_lock);
4116 }
4117
4118 static journal_t *ext4_get_journal(struct super_block *sb,
4119 unsigned int journal_inum)
4120 {
4121 struct inode *journal_inode;
4122 journal_t *journal;
4123
4124 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4125
4126 /* First, test for the existence of a valid inode on disk. Bad
4127 * things happen if we iget() an unused inode, as the subsequent
4128 * iput() will try to delete it. */
4129
4130 journal_inode = ext4_iget(sb, journal_inum);
4131 if (IS_ERR(journal_inode)) {
4132 ext4_msg(sb, KERN_ERR, "no journal found");
4133 return NULL;
4134 }
4135 if (!journal_inode->i_nlink) {
4136 make_bad_inode(journal_inode);
4137 iput(journal_inode);
4138 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4139 return NULL;
4140 }
4141
4142 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4143 journal_inode, journal_inode->i_size);
4144 if (!S_ISREG(journal_inode->i_mode)) {
4145 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4146 iput(journal_inode);
4147 return NULL;
4148 }
4149
4150 journal = jbd2_journal_init_inode(journal_inode);
4151 if (!journal) {
4152 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4153 iput(journal_inode);
4154 return NULL;
4155 }
4156 journal->j_private = sb;
4157 ext4_init_journal_params(sb, journal);
4158 return journal;
4159 }
4160
4161 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4162 dev_t j_dev)
4163 {
4164 struct buffer_head *bh;
4165 journal_t *journal;
4166 ext4_fsblk_t start;
4167 ext4_fsblk_t len;
4168 int hblock, blocksize;
4169 ext4_fsblk_t sb_block;
4170 unsigned long offset;
4171 struct ext4_super_block *es;
4172 struct block_device *bdev;
4173
4174 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4175
4176 bdev = ext4_blkdev_get(j_dev, sb);
4177 if (bdev == NULL)
4178 return NULL;
4179
4180 blocksize = sb->s_blocksize;
4181 hblock = bdev_logical_block_size(bdev);
4182 if (blocksize < hblock) {
4183 ext4_msg(sb, KERN_ERR,
4184 "blocksize too small for journal device");
4185 goto out_bdev;
4186 }
4187
4188 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4189 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4190 set_blocksize(bdev, blocksize);
4191 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4192 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4193 "external journal");
4194 goto out_bdev;
4195 }
4196
4197 es = (struct ext4_super_block *) (bh->b_data + offset);
4198 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4199 !(le32_to_cpu(es->s_feature_incompat) &
4200 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4201 ext4_msg(sb, KERN_ERR, "external journal has "
4202 "bad superblock");
4203 brelse(bh);
4204 goto out_bdev;
4205 }
4206
4207 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4208 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4209 brelse(bh);
4210 goto out_bdev;
4211 }
4212
4213 len = ext4_blocks_count(es);
4214 start = sb_block + 1;
4215 brelse(bh); /* we're done with the superblock */
4216
4217 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4218 start, len, blocksize);
4219 if (!journal) {
4220 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4221 goto out_bdev;
4222 }
4223 journal->j_private = sb;
4224 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4225 wait_on_buffer(journal->j_sb_buffer);
4226 if (!buffer_uptodate(journal->j_sb_buffer)) {
4227 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4228 goto out_journal;
4229 }
4230 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4231 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4232 "user (unsupported) - %d",
4233 be32_to_cpu(journal->j_superblock->s_nr_users));
4234 goto out_journal;
4235 }
4236 EXT4_SB(sb)->journal_bdev = bdev;
4237 ext4_init_journal_params(sb, journal);
4238 return journal;
4239
4240 out_journal:
4241 jbd2_journal_destroy(journal);
4242 out_bdev:
4243 ext4_blkdev_put(bdev);
4244 return NULL;
4245 }
4246
4247 static int ext4_load_journal(struct super_block *sb,
4248 struct ext4_super_block *es,
4249 unsigned long journal_devnum)
4250 {
4251 journal_t *journal;
4252 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4253 dev_t journal_dev;
4254 int err = 0;
4255 int really_read_only;
4256
4257 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4258
4259 if (journal_devnum &&
4260 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4261 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4262 "numbers have changed");
4263 journal_dev = new_decode_dev(journal_devnum);
4264 } else
4265 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4266
4267 really_read_only = bdev_read_only(sb->s_bdev);
4268
4269 /*
4270 * Are we loading a blank journal or performing recovery after a
4271 * crash? For recovery, we need to check in advance whether we
4272 * can get read-write access to the device.
4273 */
4274 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4275 if (sb->s_flags & MS_RDONLY) {
4276 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4277 "required on readonly filesystem");
4278 if (really_read_only) {
4279 ext4_msg(sb, KERN_ERR, "write access "
4280 "unavailable, cannot proceed");
4281 return -EROFS;
4282 }
4283 ext4_msg(sb, KERN_INFO, "write access will "
4284 "be enabled during recovery");
4285 }
4286 }
4287
4288 if (journal_inum && journal_dev) {
4289 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4290 "and inode journals!");
4291 return -EINVAL;
4292 }
4293
4294 if (journal_inum) {
4295 if (!(journal = ext4_get_journal(sb, journal_inum)))
4296 return -EINVAL;
4297 } else {
4298 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4299 return -EINVAL;
4300 }
4301
4302 if (!(journal->j_flags & JBD2_BARRIER))
4303 ext4_msg(sb, KERN_INFO, "barriers disabled");
4304
4305 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4306 err = jbd2_journal_wipe(journal, !really_read_only);
4307 if (!err) {
4308 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4309 if (save)
4310 memcpy(save, ((char *) es) +
4311 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4312 err = jbd2_journal_load(journal);
4313 if (save)
4314 memcpy(((char *) es) + EXT4_S_ERR_START,
4315 save, EXT4_S_ERR_LEN);
4316 kfree(save);
4317 }
4318
4319 if (err) {
4320 ext4_msg(sb, KERN_ERR, "error loading journal");
4321 jbd2_journal_destroy(journal);
4322 return err;
4323 }
4324
4325 EXT4_SB(sb)->s_journal = journal;
4326 ext4_clear_journal_err(sb, es);
4327
4328 if (!really_read_only && journal_devnum &&
4329 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4330 es->s_journal_dev = cpu_to_le32(journal_devnum);
4331
4332 /* Make sure we flush the recovery flag to disk. */
4333 ext4_commit_super(sb, 1);
4334 }
4335
4336 return 0;
4337 }
4338
4339 static int ext4_commit_super(struct super_block *sb, int sync)
4340 {
4341 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4342 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4343 int error = 0;
4344
4345 if (!sbh || block_device_ejected(sb))
4346 return error;
4347 if (buffer_write_io_error(sbh)) {
4348 /*
4349 * Oh, dear. A previous attempt to write the
4350 * superblock failed. This could happen because the
4351 * USB device was yanked out. Or it could happen to
4352 * be a transient write error and maybe the block will
4353 * be remapped. Nothing we can do but to retry the
4354 * write and hope for the best.
4355 */
4356 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4357 "superblock detected");
4358 clear_buffer_write_io_error(sbh);
4359 set_buffer_uptodate(sbh);
4360 }
4361 /*
4362 * If the file system is mounted read-only, don't update the
4363 * superblock write time. This avoids updating the superblock
4364 * write time when we are mounting the root file system
4365 * read/only but we need to replay the journal; at that point,
4366 * for people who are east of GMT and who make their clock
4367 * tick in localtime for Windows bug-for-bug compatibility,
4368 * the clock is set in the future, and this will cause e2fsck
4369 * to complain and force a full file system check.
4370 */
4371 if (!(sb->s_flags & MS_RDONLY))
4372 es->s_wtime = cpu_to_le32(get_seconds());
4373 if (sb->s_bdev->bd_part)
4374 es->s_kbytes_written =
4375 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4376 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4377 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4378 else
4379 es->s_kbytes_written =
4380 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4381 ext4_free_blocks_count_set(es,
4382 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4383 &EXT4_SB(sb)->s_freeclusters_counter)));
4384 es->s_free_inodes_count =
4385 cpu_to_le32(percpu_counter_sum_positive(
4386 &EXT4_SB(sb)->s_freeinodes_counter));
4387 BUFFER_TRACE(sbh, "marking dirty");
4388 ext4_superblock_csum_set(sb);
4389 mark_buffer_dirty(sbh);
4390 if (sync) {
4391 error = sync_dirty_buffer(sbh);
4392 if (error)
4393 return error;
4394
4395 error = buffer_write_io_error(sbh);
4396 if (error) {
4397 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4398 "superblock");
4399 clear_buffer_write_io_error(sbh);
4400 set_buffer_uptodate(sbh);
4401 }
4402 }
4403 return error;
4404 }
4405
4406 /*
4407 * Have we just finished recovery? If so, and if we are mounting (or
4408 * remounting) the filesystem readonly, then we will end up with a
4409 * consistent fs on disk. Record that fact.
4410 */
4411 static void ext4_mark_recovery_complete(struct super_block *sb,
4412 struct ext4_super_block *es)
4413 {
4414 journal_t *journal = EXT4_SB(sb)->s_journal;
4415
4416 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4417 BUG_ON(journal != NULL);
4418 return;
4419 }
4420 jbd2_journal_lock_updates(journal);
4421 if (jbd2_journal_flush(journal) < 0)
4422 goto out;
4423
4424 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4425 sb->s_flags & MS_RDONLY) {
4426 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4427 ext4_commit_super(sb, 1);
4428 }
4429
4430 out:
4431 jbd2_journal_unlock_updates(journal);
4432 }
4433
4434 /*
4435 * If we are mounting (or read-write remounting) a filesystem whose journal
4436 * has recorded an error from a previous lifetime, move that error to the
4437 * main filesystem now.
4438 */
4439 static void ext4_clear_journal_err(struct super_block *sb,
4440 struct ext4_super_block *es)
4441 {
4442 journal_t *journal;
4443 int j_errno;
4444 const char *errstr;
4445
4446 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4447
4448 journal = EXT4_SB(sb)->s_journal;
4449
4450 /*
4451 * Now check for any error status which may have been recorded in the
4452 * journal by a prior ext4_error() or ext4_abort()
4453 */
4454
4455 j_errno = jbd2_journal_errno(journal);
4456 if (j_errno) {
4457 char nbuf[16];
4458
4459 errstr = ext4_decode_error(sb, j_errno, nbuf);
4460 ext4_warning(sb, "Filesystem error recorded "
4461 "from previous mount: %s", errstr);
4462 ext4_warning(sb, "Marking fs in need of filesystem check.");
4463
4464 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4465 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4466 ext4_commit_super(sb, 1);
4467
4468 jbd2_journal_clear_err(journal);
4469 jbd2_journal_update_sb_errno(journal);
4470 }
4471 }
4472
4473 /*
4474 * Force the running and committing transactions to commit,
4475 * and wait on the commit.
4476 */
4477 int ext4_force_commit(struct super_block *sb)
4478 {
4479 journal_t *journal;
4480 int ret = 0;
4481
4482 if (sb->s_flags & MS_RDONLY)
4483 return 0;
4484
4485 journal = EXT4_SB(sb)->s_journal;
4486 if (journal)
4487 ret = ext4_journal_force_commit(journal);
4488
4489 return ret;
4490 }
4491
4492 static int ext4_sync_fs(struct super_block *sb, int wait)
4493 {
4494 int ret = 0;
4495 tid_t target;
4496 struct ext4_sb_info *sbi = EXT4_SB(sb);
4497
4498 trace_ext4_sync_fs(sb, wait);
4499 flush_workqueue(sbi->dio_unwritten_wq);
4500 /*
4501 * Writeback quota in non-journalled quota case - journalled quota has
4502 * no dirty dquots
4503 */
4504 dquot_writeback_dquots(sb, -1);
4505 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4506 if (wait)
4507 jbd2_log_wait_commit(sbi->s_journal, target);
4508 }
4509 return ret;
4510 }
4511
4512 /*
4513 * LVM calls this function before a (read-only) snapshot is created. This
4514 * gives us a chance to flush the journal completely and mark the fs clean.
4515 *
4516 * Note that only this function cannot bring a filesystem to be in a clean
4517 * state independently. It relies on upper layer to stop all data & metadata
4518 * modifications.
4519 */
4520 static int ext4_freeze(struct super_block *sb)
4521 {
4522 int error = 0;
4523 journal_t *journal;
4524
4525 if (sb->s_flags & MS_RDONLY)
4526 return 0;
4527
4528 journal = EXT4_SB(sb)->s_journal;
4529
4530 /* Now we set up the journal barrier. */
4531 jbd2_journal_lock_updates(journal);
4532
4533 /*
4534 * Don't clear the needs_recovery flag if we failed to flush
4535 * the journal.
4536 */
4537 error = jbd2_journal_flush(journal);
4538 if (error < 0)
4539 goto out;
4540
4541 /* Journal blocked and flushed, clear needs_recovery flag. */
4542 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4543 error = ext4_commit_super(sb, 1);
4544 out:
4545 /* we rely on upper layer to stop further updates */
4546 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4547 return error;
4548 }
4549
4550 /*
4551 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4552 * flag here, even though the filesystem is not technically dirty yet.
4553 */
4554 static int ext4_unfreeze(struct super_block *sb)
4555 {
4556 if (sb->s_flags & MS_RDONLY)
4557 return 0;
4558
4559 /* Reset the needs_recovery flag before the fs is unlocked. */
4560 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4561 ext4_commit_super(sb, 1);
4562 return 0;
4563 }
4564
4565 /*
4566 * Structure to save mount options for ext4_remount's benefit
4567 */
4568 struct ext4_mount_options {
4569 unsigned long s_mount_opt;
4570 unsigned long s_mount_opt2;
4571 kuid_t s_resuid;
4572 kgid_t s_resgid;
4573 unsigned long s_commit_interval;
4574 u32 s_min_batch_time, s_max_batch_time;
4575 #ifdef CONFIG_QUOTA
4576 int s_jquota_fmt;
4577 char *s_qf_names[MAXQUOTAS];
4578 #endif
4579 };
4580
4581 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4582 {
4583 struct ext4_super_block *es;
4584 struct ext4_sb_info *sbi = EXT4_SB(sb);
4585 unsigned long old_sb_flags;
4586 struct ext4_mount_options old_opts;
4587 int enable_quota = 0;
4588 ext4_group_t g;
4589 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4590 int err = 0;
4591 #ifdef CONFIG_QUOTA
4592 int i;
4593 #endif
4594 char *orig_data = kstrdup(data, GFP_KERNEL);
4595
4596 /* Store the original options */
4597 old_sb_flags = sb->s_flags;
4598 old_opts.s_mount_opt = sbi->s_mount_opt;
4599 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4600 old_opts.s_resuid = sbi->s_resuid;
4601 old_opts.s_resgid = sbi->s_resgid;
4602 old_opts.s_commit_interval = sbi->s_commit_interval;
4603 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4604 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4605 #ifdef CONFIG_QUOTA
4606 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4607 for (i = 0; i < MAXQUOTAS; i++)
4608 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4609 #endif
4610 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4611 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4612
4613 /*
4614 * Allow the "check" option to be passed as a remount option.
4615 */
4616 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4617 err = -EINVAL;
4618 goto restore_opts;
4619 }
4620
4621 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4622 ext4_abort(sb, "Abort forced by user");
4623
4624 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4625 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4626
4627 es = sbi->s_es;
4628
4629 if (sbi->s_journal) {
4630 ext4_init_journal_params(sb, sbi->s_journal);
4631 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4632 }
4633
4634 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4635 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4636 err = -EROFS;
4637 goto restore_opts;
4638 }
4639
4640 if (*flags & MS_RDONLY) {
4641 err = dquot_suspend(sb, -1);
4642 if (err < 0)
4643 goto restore_opts;
4644
4645 /*
4646 * First of all, the unconditional stuff we have to do
4647 * to disable replay of the journal when we next remount
4648 */
4649 sb->s_flags |= MS_RDONLY;
4650
4651 /*
4652 * OK, test if we are remounting a valid rw partition
4653 * readonly, and if so set the rdonly flag and then
4654 * mark the partition as valid again.
4655 */
4656 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4657 (sbi->s_mount_state & EXT4_VALID_FS))
4658 es->s_state = cpu_to_le16(sbi->s_mount_state);
4659
4660 if (sbi->s_journal)
4661 ext4_mark_recovery_complete(sb, es);
4662 } else {
4663 /* Make sure we can mount this feature set readwrite */
4664 if (!ext4_feature_set_ok(sb, 0)) {
4665 err = -EROFS;
4666 goto restore_opts;
4667 }
4668 /*
4669 * Make sure the group descriptor checksums
4670 * are sane. If they aren't, refuse to remount r/w.
4671 */
4672 for (g = 0; g < sbi->s_groups_count; g++) {
4673 struct ext4_group_desc *gdp =
4674 ext4_get_group_desc(sb, g, NULL);
4675
4676 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4677 ext4_msg(sb, KERN_ERR,
4678 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4679 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4680 le16_to_cpu(gdp->bg_checksum));
4681 err = -EINVAL;
4682 goto restore_opts;
4683 }
4684 }
4685
4686 /*
4687 * If we have an unprocessed orphan list hanging
4688 * around from a previously readonly bdev mount,
4689 * require a full umount/remount for now.
4690 */
4691 if (es->s_last_orphan) {
4692 ext4_msg(sb, KERN_WARNING, "Couldn't "
4693 "remount RDWR because of unprocessed "
4694 "orphan inode list. Please "
4695 "umount/remount instead");
4696 err = -EINVAL;
4697 goto restore_opts;
4698 }
4699
4700 /*
4701 * Mounting a RDONLY partition read-write, so reread
4702 * and store the current valid flag. (It may have
4703 * been changed by e2fsck since we originally mounted
4704 * the partition.)
4705 */
4706 if (sbi->s_journal)
4707 ext4_clear_journal_err(sb, es);
4708 sbi->s_mount_state = le16_to_cpu(es->s_state);
4709 if (!ext4_setup_super(sb, es, 0))
4710 sb->s_flags &= ~MS_RDONLY;
4711 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4712 EXT4_FEATURE_INCOMPAT_MMP))
4713 if (ext4_multi_mount_protect(sb,
4714 le64_to_cpu(es->s_mmp_block))) {
4715 err = -EROFS;
4716 goto restore_opts;
4717 }
4718 enable_quota = 1;
4719 }
4720 }
4721
4722 /*
4723 * Reinitialize lazy itable initialization thread based on
4724 * current settings
4725 */
4726 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4727 ext4_unregister_li_request(sb);
4728 else {
4729 ext4_group_t first_not_zeroed;
4730 first_not_zeroed = ext4_has_uninit_itable(sb);
4731 ext4_register_li_request(sb, first_not_zeroed);
4732 }
4733
4734 ext4_setup_system_zone(sb);
4735 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
4736 ext4_commit_super(sb, 1);
4737
4738 #ifdef CONFIG_QUOTA
4739 /* Release old quota file names */
4740 for (i = 0; i < MAXQUOTAS; i++)
4741 if (old_opts.s_qf_names[i] &&
4742 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4743 kfree(old_opts.s_qf_names[i]);
4744 if (enable_quota) {
4745 if (sb_any_quota_suspended(sb))
4746 dquot_resume(sb, -1);
4747 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4748 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4749 err = ext4_enable_quotas(sb);
4750 if (err)
4751 goto restore_opts;
4752 }
4753 }
4754 #endif
4755
4756 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4757 kfree(orig_data);
4758 return 0;
4759
4760 restore_opts:
4761 sb->s_flags = old_sb_flags;
4762 sbi->s_mount_opt = old_opts.s_mount_opt;
4763 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4764 sbi->s_resuid = old_opts.s_resuid;
4765 sbi->s_resgid = old_opts.s_resgid;
4766 sbi->s_commit_interval = old_opts.s_commit_interval;
4767 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4768 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4769 #ifdef CONFIG_QUOTA
4770 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4771 for (i = 0; i < MAXQUOTAS; i++) {
4772 if (sbi->s_qf_names[i] &&
4773 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4774 kfree(sbi->s_qf_names[i]);
4775 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4776 }
4777 #endif
4778 kfree(orig_data);
4779 return err;
4780 }
4781
4782 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4783 {
4784 struct super_block *sb = dentry->d_sb;
4785 struct ext4_sb_info *sbi = EXT4_SB(sb);
4786 struct ext4_super_block *es = sbi->s_es;
4787 ext4_fsblk_t overhead = 0;
4788 u64 fsid;
4789 s64 bfree;
4790
4791 if (!test_opt(sb, MINIX_DF))
4792 overhead = sbi->s_overhead;
4793
4794 buf->f_type = EXT4_SUPER_MAGIC;
4795 buf->f_bsize = sb->s_blocksize;
4796 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
4797 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4798 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4799 /* prevent underflow in case that few free space is available */
4800 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4801 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4802 if (buf->f_bfree < ext4_r_blocks_count(es))
4803 buf->f_bavail = 0;
4804 buf->f_files = le32_to_cpu(es->s_inodes_count);
4805 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4806 buf->f_namelen = EXT4_NAME_LEN;
4807 fsid = le64_to_cpup((void *)es->s_uuid) ^
4808 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4809 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4810 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4811
4812 return 0;
4813 }
4814
4815 /* Helper function for writing quotas on sync - we need to start transaction
4816 * before quota file is locked for write. Otherwise the are possible deadlocks:
4817 * Process 1 Process 2
4818 * ext4_create() quota_sync()
4819 * jbd2_journal_start() write_dquot()
4820 * dquot_initialize() down(dqio_mutex)
4821 * down(dqio_mutex) jbd2_journal_start()
4822 *
4823 */
4824
4825 #ifdef CONFIG_QUOTA
4826
4827 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4828 {
4829 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
4830 }
4831
4832 static int ext4_write_dquot(struct dquot *dquot)
4833 {
4834 int ret, err;
4835 handle_t *handle;
4836 struct inode *inode;
4837
4838 inode = dquot_to_inode(dquot);
4839 handle = ext4_journal_start(inode,
4840 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4841 if (IS_ERR(handle))
4842 return PTR_ERR(handle);
4843 ret = dquot_commit(dquot);
4844 err = ext4_journal_stop(handle);
4845 if (!ret)
4846 ret = err;
4847 return ret;
4848 }
4849
4850 static int ext4_acquire_dquot(struct dquot *dquot)
4851 {
4852 int ret, err;
4853 handle_t *handle;
4854
4855 handle = ext4_journal_start(dquot_to_inode(dquot),
4856 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4857 if (IS_ERR(handle))
4858 return PTR_ERR(handle);
4859 ret = dquot_acquire(dquot);
4860 err = ext4_journal_stop(handle);
4861 if (!ret)
4862 ret = err;
4863 return ret;
4864 }
4865
4866 static int ext4_release_dquot(struct dquot *dquot)
4867 {
4868 int ret, err;
4869 handle_t *handle;
4870
4871 handle = ext4_journal_start(dquot_to_inode(dquot),
4872 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4873 if (IS_ERR(handle)) {
4874 /* Release dquot anyway to avoid endless cycle in dqput() */
4875 dquot_release(dquot);
4876 return PTR_ERR(handle);
4877 }
4878 ret = dquot_release(dquot);
4879 err = ext4_journal_stop(handle);
4880 if (!ret)
4881 ret = err;
4882 return ret;
4883 }
4884
4885 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4886 {
4887 /* Are we journaling quotas? */
4888 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4889 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4890 dquot_mark_dquot_dirty(dquot);
4891 return ext4_write_dquot(dquot);
4892 } else {
4893 return dquot_mark_dquot_dirty(dquot);
4894 }
4895 }
4896
4897 static int ext4_write_info(struct super_block *sb, int type)
4898 {
4899 int ret, err;
4900 handle_t *handle;
4901
4902 /* Data block + inode block */
4903 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4904 if (IS_ERR(handle))
4905 return PTR_ERR(handle);
4906 ret = dquot_commit_info(sb, type);
4907 err = ext4_journal_stop(handle);
4908 if (!ret)
4909 ret = err;
4910 return ret;
4911 }
4912
4913 /*
4914 * Turn on quotas during mount time - we need to find
4915 * the quota file and such...
4916 */
4917 static int ext4_quota_on_mount(struct super_block *sb, int type)
4918 {
4919 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4920 EXT4_SB(sb)->s_jquota_fmt, type);
4921 }
4922
4923 /*
4924 * Standard function to be called on quota_on
4925 */
4926 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4927 struct path *path)
4928 {
4929 int err;
4930
4931 if (!test_opt(sb, QUOTA))
4932 return -EINVAL;
4933
4934 /* Quotafile not on the same filesystem? */
4935 if (path->dentry->d_sb != sb)
4936 return -EXDEV;
4937 /* Journaling quota? */
4938 if (EXT4_SB(sb)->s_qf_names[type]) {
4939 /* Quotafile not in fs root? */
4940 if (path->dentry->d_parent != sb->s_root)
4941 ext4_msg(sb, KERN_WARNING,
4942 "Quota file not on filesystem root. "
4943 "Journaled quota will not work");
4944 }
4945
4946 /*
4947 * When we journal data on quota file, we have to flush journal to see
4948 * all updates to the file when we bypass pagecache...
4949 */
4950 if (EXT4_SB(sb)->s_journal &&
4951 ext4_should_journal_data(path->dentry->d_inode)) {
4952 /*
4953 * We don't need to lock updates but journal_flush() could
4954 * otherwise be livelocked...
4955 */
4956 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4957 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4958 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4959 if (err)
4960 return err;
4961 }
4962
4963 return dquot_quota_on(sb, type, format_id, path);
4964 }
4965
4966 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4967 unsigned int flags)
4968 {
4969 int err;
4970 struct inode *qf_inode;
4971 unsigned long qf_inums[MAXQUOTAS] = {
4972 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4973 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4974 };
4975
4976 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4977
4978 if (!qf_inums[type])
4979 return -EPERM;
4980
4981 qf_inode = ext4_iget(sb, qf_inums[type]);
4982 if (IS_ERR(qf_inode)) {
4983 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4984 return PTR_ERR(qf_inode);
4985 }
4986
4987 err = dquot_enable(qf_inode, type, format_id, flags);
4988 iput(qf_inode);
4989
4990 return err;
4991 }
4992
4993 /* Enable usage tracking for all quota types. */
4994 static int ext4_enable_quotas(struct super_block *sb)
4995 {
4996 int type, err = 0;
4997 unsigned long qf_inums[MAXQUOTAS] = {
4998 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4999 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5000 };
5001
5002 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5003 for (type = 0; type < MAXQUOTAS; type++) {
5004 if (qf_inums[type]) {
5005 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5006 DQUOT_USAGE_ENABLED);
5007 if (err) {
5008 ext4_warning(sb,
5009 "Failed to enable quota (type=%d) "
5010 "tracking. Please run e2fsck to fix.",
5011 type);
5012 return err;
5013 }
5014 }
5015 }
5016 return 0;
5017 }
5018
5019 /*
5020 * quota_on function that is used when QUOTA feature is set.
5021 */
5022 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5023 int format_id)
5024 {
5025 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5026 return -EINVAL;
5027
5028 /*
5029 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5030 */
5031 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5032 }
5033
5034 static int ext4_quota_off(struct super_block *sb, int type)
5035 {
5036 struct inode *inode = sb_dqopt(sb)->files[type];
5037 handle_t *handle;
5038
5039 /* Force all delayed allocation blocks to be allocated.
5040 * Caller already holds s_umount sem */
5041 if (test_opt(sb, DELALLOC))
5042 sync_filesystem(sb);
5043
5044 if (!inode)
5045 goto out;
5046
5047 /* Update modification times of quota files when userspace can
5048 * start looking at them */
5049 handle = ext4_journal_start(inode, 1);
5050 if (IS_ERR(handle))
5051 goto out;
5052 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5053 ext4_mark_inode_dirty(handle, inode);
5054 ext4_journal_stop(handle);
5055
5056 out:
5057 return dquot_quota_off(sb, type);
5058 }
5059
5060 /*
5061 * quota_off function that is used when QUOTA feature is set.
5062 */
5063 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5064 {
5065 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5066 return -EINVAL;
5067
5068 /* Disable only the limits. */
5069 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5070 }
5071
5072 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5073 * acquiring the locks... As quota files are never truncated and quota code
5074 * itself serializes the operations (and no one else should touch the files)
5075 * we don't have to be afraid of races */
5076 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5077 size_t len, loff_t off)
5078 {
5079 struct inode *inode = sb_dqopt(sb)->files[type];
5080 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5081 int err = 0;
5082 int offset = off & (sb->s_blocksize - 1);
5083 int tocopy;
5084 size_t toread;
5085 struct buffer_head *bh;
5086 loff_t i_size = i_size_read(inode);
5087
5088 if (off > i_size)
5089 return 0;
5090 if (off+len > i_size)
5091 len = i_size-off;
5092 toread = len;
5093 while (toread > 0) {
5094 tocopy = sb->s_blocksize - offset < toread ?
5095 sb->s_blocksize - offset : toread;
5096 bh = ext4_bread(NULL, inode, blk, 0, &err);
5097 if (err)
5098 return err;
5099 if (!bh) /* A hole? */
5100 memset(data, 0, tocopy);
5101 else
5102 memcpy(data, bh->b_data+offset, tocopy);
5103 brelse(bh);
5104 offset = 0;
5105 toread -= tocopy;
5106 data += tocopy;
5107 blk++;
5108 }
5109 return len;
5110 }
5111
5112 /* Write to quotafile (we know the transaction is already started and has
5113 * enough credits) */
5114 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5115 const char *data, size_t len, loff_t off)
5116 {
5117 struct inode *inode = sb_dqopt(sb)->files[type];
5118 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5119 int err = 0;
5120 int offset = off & (sb->s_blocksize - 1);
5121 struct buffer_head *bh;
5122 handle_t *handle = journal_current_handle();
5123
5124 if (EXT4_SB(sb)->s_journal && !handle) {
5125 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5126 " cancelled because transaction is not started",
5127 (unsigned long long)off, (unsigned long long)len);
5128 return -EIO;
5129 }
5130 /*
5131 * Since we account only one data block in transaction credits,
5132 * then it is impossible to cross a block boundary.
5133 */
5134 if (sb->s_blocksize - offset < len) {
5135 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5136 " cancelled because not block aligned",
5137 (unsigned long long)off, (unsigned long long)len);
5138 return -EIO;
5139 }
5140
5141 bh = ext4_bread(handle, inode, blk, 1, &err);
5142 if (!bh)
5143 goto out;
5144 err = ext4_journal_get_write_access(handle, bh);
5145 if (err) {
5146 brelse(bh);
5147 goto out;
5148 }
5149 lock_buffer(bh);
5150 memcpy(bh->b_data+offset, data, len);
5151 flush_dcache_page(bh->b_page);
5152 unlock_buffer(bh);
5153 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5154 brelse(bh);
5155 out:
5156 if (err)
5157 return err;
5158 if (inode->i_size < off + len) {
5159 i_size_write(inode, off + len);
5160 EXT4_I(inode)->i_disksize = inode->i_size;
5161 ext4_mark_inode_dirty(handle, inode);
5162 }
5163 return len;
5164 }
5165
5166 #endif
5167
5168 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5169 const char *dev_name, void *data)
5170 {
5171 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5172 }
5173
5174 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5175 static inline void register_as_ext2(void)
5176 {
5177 int err = register_filesystem(&ext2_fs_type);
5178 if (err)
5179 printk(KERN_WARNING
5180 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5181 }
5182
5183 static inline void unregister_as_ext2(void)
5184 {
5185 unregister_filesystem(&ext2_fs_type);
5186 }
5187
5188 static inline int ext2_feature_set_ok(struct super_block *sb)
5189 {
5190 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5191 return 0;
5192 if (sb->s_flags & MS_RDONLY)
5193 return 1;
5194 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5195 return 0;
5196 return 1;
5197 }
5198 MODULE_ALIAS("ext2");
5199 #else
5200 static inline void register_as_ext2(void) { }
5201 static inline void unregister_as_ext2(void) { }
5202 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5203 #endif
5204
5205 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5206 static inline void register_as_ext3(void)
5207 {
5208 int err = register_filesystem(&ext3_fs_type);
5209 if (err)
5210 printk(KERN_WARNING
5211 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5212 }
5213
5214 static inline void unregister_as_ext3(void)
5215 {
5216 unregister_filesystem(&ext3_fs_type);
5217 }
5218
5219 static inline int ext3_feature_set_ok(struct super_block *sb)
5220 {
5221 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5222 return 0;
5223 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5224 return 0;
5225 if (sb->s_flags & MS_RDONLY)
5226 return 1;
5227 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5228 return 0;
5229 return 1;
5230 }
5231 MODULE_ALIAS("ext3");
5232 #else
5233 static inline void register_as_ext3(void) { }
5234 static inline void unregister_as_ext3(void) { }
5235 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5236 #endif
5237
5238 static struct file_system_type ext4_fs_type = {
5239 .owner = THIS_MODULE,
5240 .name = "ext4",
5241 .mount = ext4_mount,
5242 .kill_sb = kill_block_super,
5243 .fs_flags = FS_REQUIRES_DEV,
5244 };
5245
5246 static int __init ext4_init_feat_adverts(void)
5247 {
5248 struct ext4_features *ef;
5249 int ret = -ENOMEM;
5250
5251 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5252 if (!ef)
5253 goto out;
5254
5255 ef->f_kobj.kset = ext4_kset;
5256 init_completion(&ef->f_kobj_unregister);
5257 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5258 "features");
5259 if (ret) {
5260 kfree(ef);
5261 goto out;
5262 }
5263
5264 ext4_feat = ef;
5265 ret = 0;
5266 out:
5267 return ret;
5268 }
5269
5270 static void ext4_exit_feat_adverts(void)
5271 {
5272 kobject_put(&ext4_feat->f_kobj);
5273 wait_for_completion(&ext4_feat->f_kobj_unregister);
5274 kfree(ext4_feat);
5275 }
5276
5277 /* Shared across all ext4 file systems */
5278 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5279 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5280
5281 static int __init ext4_init_fs(void)
5282 {
5283 int i, err;
5284
5285 ext4_li_info = NULL;
5286 mutex_init(&ext4_li_mtx);
5287
5288 /* Build-time check for flags consistency */
5289 ext4_check_flag_values();
5290
5291 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5292 mutex_init(&ext4__aio_mutex[i]);
5293 init_waitqueue_head(&ext4__ioend_wq[i]);
5294 }
5295
5296 err = ext4_init_es();
5297 if (err)
5298 return err;
5299
5300 err = ext4_init_pageio();
5301 if (err)
5302 goto out7;
5303
5304 err = ext4_init_system_zone();
5305 if (err)
5306 goto out6;
5307 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5308 if (!ext4_kset) {
5309 err = -ENOMEM;
5310 goto out5;
5311 }
5312 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5313
5314 err = ext4_init_feat_adverts();
5315 if (err)
5316 goto out4;
5317
5318 err = ext4_init_mballoc();
5319 if (err)
5320 goto out3;
5321
5322 err = ext4_init_xattr();
5323 if (err)
5324 goto out2;
5325 err = init_inodecache();
5326 if (err)
5327 goto out1;
5328 register_as_ext3();
5329 register_as_ext2();
5330 err = register_filesystem(&ext4_fs_type);
5331 if (err)
5332 goto out;
5333
5334 return 0;
5335 out:
5336 unregister_as_ext2();
5337 unregister_as_ext3();
5338 destroy_inodecache();
5339 out1:
5340 ext4_exit_xattr();
5341 out2:
5342 ext4_exit_mballoc();
5343 out3:
5344 ext4_exit_feat_adverts();
5345 out4:
5346 if (ext4_proc_root)
5347 remove_proc_entry("fs/ext4", NULL);
5348 kset_unregister(ext4_kset);
5349 out5:
5350 ext4_exit_system_zone();
5351 out6:
5352 ext4_exit_pageio();
5353 out7:
5354 ext4_exit_es();
5355
5356 return err;
5357 }
5358
5359 static void __exit ext4_exit_fs(void)
5360 {
5361 ext4_destroy_lazyinit_thread();
5362 unregister_as_ext2();
5363 unregister_as_ext3();
5364 unregister_filesystem(&ext4_fs_type);
5365 destroy_inodecache();
5366 ext4_exit_xattr();
5367 ext4_exit_mballoc();
5368 ext4_exit_feat_adverts();
5369 remove_proc_entry("fs/ext4", NULL);
5370 kset_unregister(ext4_kset);
5371 ext4_exit_system_zone();
5372 ext4_exit_pageio();
5373 }
5374
5375 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5376 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5377 MODULE_LICENSE("GPL");
5378 module_init(ext4_init_fs)
5379 module_exit(ext4_exit_fs)
ubuntu-zesty-kernel mirror