]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - fs/ext4/super.c
projects / mirror_ubuntu-zesty-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
UBUNTU: SAUCE: (namespace) ext4: Add support for unprivileged mounts from user namespaces
[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/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.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/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <linux/uaccess.h>
42 #include <linux/user_namespace.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 ext4_lazy_init *ext4_li_info;
58 static struct mutex ext4_li_mtx;
59 static struct ratelimit_state ext4_mount_msg_ratelimit;
60
61 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
62 unsigned long journal_devnum);
63 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
64 static int ext4_commit_super(struct super_block *sb, int sync);
65 static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67 static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
69 static int ext4_sync_fs(struct super_block *sb, int wait);
70 static int ext4_remount(struct super_block *sb, int *flags, char *data);
71 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
72 static int ext4_unfreeze(struct super_block *sb);
73 static int ext4_freeze(struct super_block *sb);
74 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
75 const char *dev_name, void *data);
76 static inline int ext2_feature_set_ok(struct super_block *sb);
77 static inline int ext3_feature_set_ok(struct super_block *sb);
78 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
79 static void ext4_destroy_lazyinit_thread(void);
80 static void ext4_unregister_li_request(struct super_block *sb);
81 static void ext4_clear_request_list(void);
82 static struct inode *ext4_get_journal_inode(struct super_block *sb,
83 unsigned int journal_inum);
84
85 /*
86 * Lock ordering
87 *
88 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
89 * i_mmap_rwsem (inode->i_mmap_rwsem)!
90 *
91 * page fault path:
92 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
93 * page lock -> i_data_sem (rw)
94 *
95 * buffered write path:
96 * sb_start_write -> i_mutex -> mmap_sem
97 * sb_start_write -> i_mutex -> transaction start -> page lock ->
98 * i_data_sem (rw)
99 *
100 * truncate:
101 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
102 * i_mmap_rwsem (w) -> page lock
103 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
104 * transaction start -> i_data_sem (rw)
105 *
106 * direct IO:
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
108 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
109 * transaction start -> i_data_sem (rw)
110 *
111 * writepages:
112 * transaction start -> page lock(s) -> i_data_sem (rw)
113 */
114
115 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
116 static struct file_system_type ext2_fs_type = {
117 .owner = THIS_MODULE,
118 .name = "ext2",
119 .mount = ext4_mount,
120 .kill_sb = kill_block_super,
121 .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT,
122 };
123 MODULE_ALIAS_FS("ext2");
124 MODULE_ALIAS("ext2");
125 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
126 #else
127 #define IS_EXT2_SB(sb) (0)
128 #endif
129
130
131 static struct file_system_type ext3_fs_type = {
132 .owner = THIS_MODULE,
133 .name = "ext3",
134 .mount = ext4_mount,
135 .kill_sb = kill_block_super,
136 .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT,
137 };
138 MODULE_ALIAS_FS("ext3");
139 MODULE_ALIAS("ext3");
140 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
141
142 static int ext4_verify_csum_type(struct super_block *sb,
143 struct ext4_super_block *es)
144 {
145 if (!ext4_has_feature_metadata_csum(sb))
146 return 1;
147
148 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
149 }
150
151 static __le32 ext4_superblock_csum(struct super_block *sb,
152 struct ext4_super_block *es)
153 {
154 struct ext4_sb_info *sbi = EXT4_SB(sb);
155 int offset = offsetof(struct ext4_super_block, s_checksum);
156 __u32 csum;
157
158 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
159
160 return cpu_to_le32(csum);
161 }
162
163 static int ext4_superblock_csum_verify(struct super_block *sb,
164 struct ext4_super_block *es)
165 {
166 if (!ext4_has_metadata_csum(sb))
167 return 1;
168
169 return es->s_checksum == ext4_superblock_csum(sb, es);
170 }
171
172 void ext4_superblock_csum_set(struct super_block *sb)
173 {
174 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
175
176 if (!ext4_has_metadata_csum(sb))
177 return;
178
179 es->s_checksum = ext4_superblock_csum(sb, es);
180 }
181
182 void *ext4_kvmalloc(size_t size, gfp_t flags)
183 {
184 void *ret;
185
186 ret = kmalloc(size, flags | __GFP_NOWARN);
187 if (!ret)
188 ret = __vmalloc(size, flags, PAGE_KERNEL);
189 return ret;
190 }
191
192 void *ext4_kvzalloc(size_t size, gfp_t flags)
193 {
194 void *ret;
195
196 ret = kzalloc(size, flags | __GFP_NOWARN);
197 if (!ret)
198 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
199 return ret;
200 }
201
202 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
203 struct ext4_group_desc *bg)
204 {
205 return le32_to_cpu(bg->bg_block_bitmap_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
208 }
209
210 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
211 struct ext4_group_desc *bg)
212 {
213 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
216 }
217
218 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
219 struct ext4_group_desc *bg)
220 {
221 return le32_to_cpu(bg->bg_inode_table_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
224 }
225
226 __u32 ext4_free_group_clusters(struct super_block *sb,
227 struct ext4_group_desc *bg)
228 {
229 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
231 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
232 }
233
234 __u32 ext4_free_inodes_count(struct super_block *sb,
235 struct ext4_group_desc *bg)
236 {
237 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
239 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
240 }
241
242 __u32 ext4_used_dirs_count(struct super_block *sb,
243 struct ext4_group_desc *bg)
244 {
245 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
246 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
247 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
248 }
249
250 __u32 ext4_itable_unused_count(struct super_block *sb,
251 struct ext4_group_desc *bg)
252 {
253 return le16_to_cpu(bg->bg_itable_unused_lo) |
254 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
255 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
256 }
257
258 void ext4_block_bitmap_set(struct super_block *sb,
259 struct ext4_group_desc *bg, ext4_fsblk_t blk)
260 {
261 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
264 }
265
266 void ext4_inode_bitmap_set(struct super_block *sb,
267 struct ext4_group_desc *bg, ext4_fsblk_t blk)
268 {
269 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
272 }
273
274 void ext4_inode_table_set(struct super_block *sb,
275 struct ext4_group_desc *bg, ext4_fsblk_t blk)
276 {
277 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
280 }
281
282 void ext4_free_group_clusters_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284 {
285 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
288 }
289
290 void ext4_free_inodes_set(struct super_block *sb,
291 struct ext4_group_desc *bg, __u32 count)
292 {
293 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
296 }
297
298 void ext4_used_dirs_set(struct super_block *sb,
299 struct ext4_group_desc *bg, __u32 count)
300 {
301 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
302 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
303 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
304 }
305
306 void ext4_itable_unused_set(struct super_block *sb,
307 struct ext4_group_desc *bg, __u32 count)
308 {
309 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
310 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
311 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
312 }
313
314
315 static void __save_error_info(struct super_block *sb, const char *func,
316 unsigned int line)
317 {
318 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
319
320 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
321 if (bdev_read_only(sb->s_bdev))
322 return;
323 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
324 es->s_last_error_time = cpu_to_le32(get_seconds());
325 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
326 es->s_last_error_line = cpu_to_le32(line);
327 if (!es->s_first_error_time) {
328 es->s_first_error_time = es->s_last_error_time;
329 strncpy(es->s_first_error_func, func,
330 sizeof(es->s_first_error_func));
331 es->s_first_error_line = cpu_to_le32(line);
332 es->s_first_error_ino = es->s_last_error_ino;
333 es->s_first_error_block = es->s_last_error_block;
334 }
335 /*
336 * Start the daily error reporting function if it hasn't been
337 * started already
338 */
339 if (!es->s_error_count)
340 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
341 le32_add_cpu(&es->s_error_count, 1);
342 }
343
344 static void save_error_info(struct super_block *sb, const char *func,
345 unsigned int line)
346 {
347 __save_error_info(sb, func, line);
348 ext4_commit_super(sb, 1);
349 }
350
351 /*
352 * The del_gendisk() function uninitializes the disk-specific data
353 * structures, including the bdi structure, without telling anyone
354 * else. Once this happens, any attempt to call mark_buffer_dirty()
355 * (for example, by ext4_commit_super), will cause a kernel OOPS.
356 * This is a kludge to prevent these oops until we can put in a proper
357 * hook in del_gendisk() to inform the VFS and file system layers.
358 */
359 static int block_device_ejected(struct super_block *sb)
360 {
361 struct inode *bd_inode = sb->s_bdev->bd_inode;
362 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
363
364 return bdi->dev == NULL;
365 }
366
367 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
368 {
369 struct super_block *sb = journal->j_private;
370 struct ext4_sb_info *sbi = EXT4_SB(sb);
371 int error = is_journal_aborted(journal);
372 struct ext4_journal_cb_entry *jce;
373
374 BUG_ON(txn->t_state == T_FINISHED);
375 spin_lock(&sbi->s_md_lock);
376 while (!list_empty(&txn->t_private_list)) {
377 jce = list_entry(txn->t_private_list.next,
378 struct ext4_journal_cb_entry, jce_list);
379 list_del_init(&jce->jce_list);
380 spin_unlock(&sbi->s_md_lock);
381 jce->jce_func(sb, jce, error);
382 spin_lock(&sbi->s_md_lock);
383 }
384 spin_unlock(&sbi->s_md_lock);
385 }
386
387 /* Deal with the reporting of failure conditions on a filesystem such as
388 * inconsistencies detected or read IO failures.
389 *
390 * On ext2, we can store the error state of the filesystem in the
391 * superblock. That is not possible on ext4, because we may have other
392 * write ordering constraints on the superblock which prevent us from
393 * writing it out straight away; and given that the journal is about to
394 * be aborted, we can't rely on the current, or future, transactions to
395 * write out the superblock safely.
396 *
397 * We'll just use the jbd2_journal_abort() error code to record an error in
398 * the journal instead. On recovery, the journal will complain about
399 * that error until we've noted it down and cleared it.
400 */
401
402 static void ext4_handle_error(struct super_block *sb)
403 {
404 if (sb->s_flags & MS_RDONLY)
405 return;
406
407 if (!test_opt(sb, ERRORS_CONT)) {
408 journal_t *journal = EXT4_SB(sb)->s_journal;
409
410 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
411 if (journal)
412 jbd2_journal_abort(journal, -EIO);
413 }
414 if (test_opt(sb, ERRORS_RO)) {
415 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
416 /*
417 * Make sure updated value of ->s_mount_flags will be visible
418 * before ->s_flags update
419 */
420 smp_wmb();
421 sb->s_flags |= MS_RDONLY;
422 }
423 if (test_opt(sb, ERRORS_PANIC)) {
424 if (EXT4_SB(sb)->s_journal &&
425 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
426 return;
427 panic("EXT4-fs (device %s): panic forced after error\n",
428 sb->s_id);
429 }
430 }
431
432 #define ext4_error_ratelimit(sb) \
433 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
434 "EXT4-fs error")
435
436 void __ext4_error(struct super_block *sb, const char *function,
437 unsigned int line, const char *fmt, ...)
438 {
439 struct va_format vaf;
440 va_list args;
441
442 if (ext4_error_ratelimit(sb)) {
443 va_start(args, fmt);
444 vaf.fmt = fmt;
445 vaf.va = &args;
446 printk(KERN_CRIT
447 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
448 sb->s_id, function, line, current->comm, &vaf);
449 va_end(args);
450 }
451 save_error_info(sb, function, line);
452 ext4_handle_error(sb);
453 }
454
455 void __ext4_error_inode(struct inode *inode, const char *function,
456 unsigned int line, ext4_fsblk_t block,
457 const char *fmt, ...)
458 {
459 va_list args;
460 struct va_format vaf;
461 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
462
463 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
464 es->s_last_error_block = cpu_to_le64(block);
465 if (ext4_error_ratelimit(inode->i_sb)) {
466 va_start(args, fmt);
467 vaf.fmt = fmt;
468 vaf.va = &args;
469 if (block)
470 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
471 "inode #%lu: block %llu: comm %s: %pV\n",
472 inode->i_sb->s_id, function, line, inode->i_ino,
473 block, current->comm, &vaf);
474 else
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
476 "inode #%lu: comm %s: %pV\n",
477 inode->i_sb->s_id, function, line, inode->i_ino,
478 current->comm, &vaf);
479 va_end(args);
480 }
481 save_error_info(inode->i_sb, function, line);
482 ext4_handle_error(inode->i_sb);
483 }
484
485 void __ext4_error_file(struct file *file, const char *function,
486 unsigned int line, ext4_fsblk_t block,
487 const char *fmt, ...)
488 {
489 va_list args;
490 struct va_format vaf;
491 struct ext4_super_block *es;
492 struct inode *inode = file_inode(file);
493 char pathname[80], *path;
494
495 es = EXT4_SB(inode->i_sb)->s_es;
496 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
497 if (ext4_error_ratelimit(inode->i_sb)) {
498 path = file_path(file, pathname, sizeof(pathname));
499 if (IS_ERR(path))
500 path = "(unknown)";
501 va_start(args, fmt);
502 vaf.fmt = fmt;
503 vaf.va = &args;
504 if (block)
505 printk(KERN_CRIT
506 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
507 "block %llu: comm %s: path %s: %pV\n",
508 inode->i_sb->s_id, function, line, inode->i_ino,
509 block, current->comm, path, &vaf);
510 else
511 printk(KERN_CRIT
512 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
513 "comm %s: path %s: %pV\n",
514 inode->i_sb->s_id, function, line, inode->i_ino,
515 current->comm, path, &vaf);
516 va_end(args);
517 }
518 save_error_info(inode->i_sb, function, line);
519 ext4_handle_error(inode->i_sb);
520 }
521
522 const char *ext4_decode_error(struct super_block *sb, int errno,
523 char nbuf[16])
524 {
525 char *errstr = NULL;
526
527 switch (errno) {
528 case -EFSCORRUPTED:
529 errstr = "Corrupt filesystem";
530 break;
531 case -EFSBADCRC:
532 errstr = "Filesystem failed CRC";
533 break;
534 case -EIO:
535 errstr = "IO failure";
536 break;
537 case -ENOMEM:
538 errstr = "Out of memory";
539 break;
540 case -EROFS:
541 if (!sb || (EXT4_SB(sb)->s_journal &&
542 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
543 errstr = "Journal has aborted";
544 else
545 errstr = "Readonly filesystem";
546 break;
547 default:
548 /* If the caller passed in an extra buffer for unknown
549 * errors, textualise them now. Else we just return
550 * NULL. */
551 if (nbuf) {
552 /* Check for truncated error codes... */
553 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
554 errstr = nbuf;
555 }
556 break;
557 }
558
559 return errstr;
560 }
561
562 /* __ext4_std_error decodes expected errors from journaling functions
563 * automatically and invokes the appropriate error response. */
564
565 void __ext4_std_error(struct super_block *sb, const char *function,
566 unsigned int line, int errno)
567 {
568 char nbuf[16];
569 const char *errstr;
570
571 /* Special case: if the error is EROFS, and we're not already
572 * inside a transaction, then there's really no point in logging
573 * an error. */
574 if (errno == -EROFS && journal_current_handle() == NULL &&
575 (sb->s_flags & MS_RDONLY))
576 return;
577
578 if (ext4_error_ratelimit(sb)) {
579 errstr = ext4_decode_error(sb, errno, nbuf);
580 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
581 sb->s_id, function, line, errstr);
582 }
583
584 save_error_info(sb, function, line);
585 ext4_handle_error(sb);
586 }
587
588 /*
589 * ext4_abort is a much stronger failure handler than ext4_error. The
590 * abort function may be used to deal with unrecoverable failures such
591 * as journal IO errors or ENOMEM at a critical moment in log management.
592 *
593 * We unconditionally force the filesystem into an ABORT|READONLY state,
594 * unless the error response on the fs has been set to panic in which
595 * case we take the easy way out and panic immediately.
596 */
597
598 void __ext4_abort(struct super_block *sb, const char *function,
599 unsigned int line, const char *fmt, ...)
600 {
601 struct va_format vaf;
602 va_list args;
603
604 save_error_info(sb, function, line);
605 va_start(args, fmt);
606 vaf.fmt = fmt;
607 vaf.va = &args;
608 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
609 sb->s_id, function, line, &vaf);
610 va_end(args);
611
612 if ((sb->s_flags & MS_RDONLY) == 0) {
613 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
614 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
615 /*
616 * Make sure updated value of ->s_mount_flags will be visible
617 * before ->s_flags update
618 */
619 smp_wmb();
620 sb->s_flags |= MS_RDONLY;
621 if (EXT4_SB(sb)->s_journal)
622 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
623 save_error_info(sb, function, line);
624 }
625 if (test_opt(sb, ERRORS_PANIC)) {
626 if (EXT4_SB(sb)->s_journal &&
627 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
628 return;
629 panic("EXT4-fs panic from previous error\n");
630 }
631 }
632
633 void __ext4_msg(struct super_block *sb,
634 const char *prefix, const char *fmt, ...)
635 {
636 struct va_format vaf;
637 va_list args;
638
639 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
640 return;
641
642 va_start(args, fmt);
643 vaf.fmt = fmt;
644 vaf.va = &args;
645 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
646 va_end(args);
647 }
648
649 #define ext4_warning_ratelimit(sb) \
650 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
651 "EXT4-fs warning")
652
653 void __ext4_warning(struct super_block *sb, const char *function,
654 unsigned int line, const char *fmt, ...)
655 {
656 struct va_format vaf;
657 va_list args;
658
659 if (!ext4_warning_ratelimit(sb))
660 return;
661
662 va_start(args, fmt);
663 vaf.fmt = fmt;
664 vaf.va = &args;
665 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
666 sb->s_id, function, line, &vaf);
667 va_end(args);
668 }
669
670 void __ext4_warning_inode(const struct inode *inode, const char *function,
671 unsigned int line, const char *fmt, ...)
672 {
673 struct va_format vaf;
674 va_list args;
675
676 if (!ext4_warning_ratelimit(inode->i_sb))
677 return;
678
679 va_start(args, fmt);
680 vaf.fmt = fmt;
681 vaf.va = &args;
682 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
683 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
684 function, line, inode->i_ino, current->comm, &vaf);
685 va_end(args);
686 }
687
688 void __ext4_grp_locked_error(const char *function, unsigned int line,
689 struct super_block *sb, ext4_group_t grp,
690 unsigned long ino, ext4_fsblk_t block,
691 const char *fmt, ...)
692 __releases(bitlock)
693 __acquires(bitlock)
694 {
695 struct va_format vaf;
696 va_list args;
697 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
698
699 es->s_last_error_ino = cpu_to_le32(ino);
700 es->s_last_error_block = cpu_to_le64(block);
701 __save_error_info(sb, function, line);
702
703 if (ext4_error_ratelimit(sb)) {
704 va_start(args, fmt);
705 vaf.fmt = fmt;
706 vaf.va = &args;
707 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
708 sb->s_id, function, line, grp);
709 if (ino)
710 printk(KERN_CONT "inode %lu: ", ino);
711 if (block)
712 printk(KERN_CONT "block %llu:",
713 (unsigned long long) block);
714 printk(KERN_CONT "%pV\n", &vaf);
715 va_end(args);
716 }
717
718 if (test_opt(sb, ERRORS_CONT)) {
719 ext4_commit_super(sb, 0);
720 return;
721 }
722
723 ext4_unlock_group(sb, grp);
724 ext4_handle_error(sb);
725 /*
726 * We only get here in the ERRORS_RO case; relocking the group
727 * may be dangerous, but nothing bad will happen since the
728 * filesystem will have already been marked read/only and the
729 * journal has been aborted. We return 1 as a hint to callers
730 * who might what to use the return value from
731 * ext4_grp_locked_error() to distinguish between the
732 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
733 * aggressively from the ext4 function in question, with a
734 * more appropriate error code.
735 */
736 ext4_lock_group(sb, grp);
737 return;
738 }
739
740 void ext4_update_dynamic_rev(struct super_block *sb)
741 {
742 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
743
744 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
745 return;
746
747 ext4_warning(sb,
748 "updating to rev %d because of new feature flag, "
749 "running e2fsck is recommended",
750 EXT4_DYNAMIC_REV);
751
752 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
753 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
754 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
755 /* leave es->s_feature_*compat flags alone */
756 /* es->s_uuid will be set by e2fsck if empty */
757
758 /*
759 * The rest of the superblock fields should be zero, and if not it
760 * means they are likely already in use, so leave them alone. We
761 * can leave it up to e2fsck to clean up any inconsistencies there.
762 */
763 }
764
765 /*
766 * Open the external journal device
767 */
768 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
769 {
770 struct block_device *bdev;
771 char b[BDEVNAME_SIZE];
772
773 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
774 if (IS_ERR(bdev))
775 goto fail;
776 return bdev;
777
778 fail:
779 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
780 __bdevname(dev, b), PTR_ERR(bdev));
781 return NULL;
782 }
783
784 /*
785 * Release the journal device
786 */
787 static void ext4_blkdev_put(struct block_device *bdev)
788 {
789 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
790 }
791
792 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
793 {
794 struct block_device *bdev;
795 bdev = sbi->journal_bdev;
796 if (bdev) {
797 ext4_blkdev_put(bdev);
798 sbi->journal_bdev = NULL;
799 }
800 }
801
802 static inline struct inode *orphan_list_entry(struct list_head *l)
803 {
804 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
805 }
806
807 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
808 {
809 struct list_head *l;
810
811 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
812 le32_to_cpu(sbi->s_es->s_last_orphan));
813
814 printk(KERN_ERR "sb_info orphan list:\n");
815 list_for_each(l, &sbi->s_orphan) {
816 struct inode *inode = orphan_list_entry(l);
817 printk(KERN_ERR " "
818 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
819 inode->i_sb->s_id, inode->i_ino, inode,
820 inode->i_mode, inode->i_nlink,
821 NEXT_ORPHAN(inode));
822 }
823 }
824
825 static void ext4_put_super(struct super_block *sb)
826 {
827 struct ext4_sb_info *sbi = EXT4_SB(sb);
828 struct ext4_super_block *es = sbi->s_es;
829 int i, err;
830
831 ext4_unregister_li_request(sb);
832 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
833
834 flush_workqueue(sbi->rsv_conversion_wq);
835 destroy_workqueue(sbi->rsv_conversion_wq);
836
837 if (sbi->s_journal) {
838 err = jbd2_journal_destroy(sbi->s_journal);
839 sbi->s_journal = NULL;
840 if (err < 0)
841 ext4_abort(sb, "Couldn't clean up the journal");
842 }
843
844 ext4_unregister_sysfs(sb);
845 ext4_es_unregister_shrinker(sbi);
846 del_timer_sync(&sbi->s_err_report);
847 ext4_release_system_zone(sb);
848 ext4_mb_release(sb);
849 ext4_ext_release(sb);
850
851 if (!(sb->s_flags & MS_RDONLY)) {
852 ext4_clear_feature_journal_needs_recovery(sb);
853 es->s_state = cpu_to_le16(sbi->s_mount_state);
854 }
855 if (!(sb->s_flags & MS_RDONLY))
856 ext4_commit_super(sb, 1);
857
858 for (i = 0; i < sbi->s_gdb_count; i++)
859 brelse(sbi->s_group_desc[i]);
860 kvfree(sbi->s_group_desc);
861 kvfree(sbi->s_flex_groups);
862 percpu_counter_destroy(&sbi->s_freeclusters_counter);
863 percpu_counter_destroy(&sbi->s_freeinodes_counter);
864 percpu_counter_destroy(&sbi->s_dirs_counter);
865 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
866 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
867 #ifdef CONFIG_QUOTA
868 for (i = 0; i < EXT4_MAXQUOTAS; i++)
869 kfree(sbi->s_qf_names[i]);
870 #endif
871
872 /* Debugging code just in case the in-memory inode orphan list
873 * isn't empty. The on-disk one can be non-empty if we've
874 * detected an error and taken the fs readonly, but the
875 * in-memory list had better be clean by this point. */
876 if (!list_empty(&sbi->s_orphan))
877 dump_orphan_list(sb, sbi);
878 J_ASSERT(list_empty(&sbi->s_orphan));
879
880 sync_blockdev(sb->s_bdev);
881 invalidate_bdev(sb->s_bdev);
882 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
883 /*
884 * Invalidate the journal device's buffers. We don't want them
885 * floating about in memory - the physical journal device may
886 * hotswapped, and it breaks the `ro-after' testing code.
887 */
888 sync_blockdev(sbi->journal_bdev);
889 invalidate_bdev(sbi->journal_bdev);
890 ext4_blkdev_remove(sbi);
891 }
892 if (sbi->s_mb_cache) {
893 ext4_xattr_destroy_cache(sbi->s_mb_cache);
894 sbi->s_mb_cache = NULL;
895 }
896 if (sbi->s_mmp_tsk)
897 kthread_stop(sbi->s_mmp_tsk);
898 brelse(sbi->s_sbh);
899 sb->s_fs_info = NULL;
900 /*
901 * Now that we are completely done shutting down the
902 * superblock, we need to actually destroy the kobject.
903 */
904 kobject_put(&sbi->s_kobj);
905 wait_for_completion(&sbi->s_kobj_unregister);
906 if (sbi->s_chksum_driver)
907 crypto_free_shash(sbi->s_chksum_driver);
908 kfree(sbi->s_blockgroup_lock);
909 kfree(sbi);
910 }
911
912 static struct kmem_cache *ext4_inode_cachep;
913
914 /*
915 * Called inside transaction, so use GFP_NOFS
916 */
917 static struct inode *ext4_alloc_inode(struct super_block *sb)
918 {
919 struct ext4_inode_info *ei;
920
921 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
922 if (!ei)
923 return NULL;
924
925 ei->vfs_inode.i_version = 1;
926 spin_lock_init(&ei->i_raw_lock);
927 INIT_LIST_HEAD(&ei->i_prealloc_list);
928 spin_lock_init(&ei->i_prealloc_lock);
929 ext4_es_init_tree(&ei->i_es_tree);
930 rwlock_init(&ei->i_es_lock);
931 INIT_LIST_HEAD(&ei->i_es_list);
932 ei->i_es_all_nr = 0;
933 ei->i_es_shk_nr = 0;
934 ei->i_es_shrink_lblk = 0;
935 ei->i_reserved_data_blocks = 0;
936 ei->i_reserved_meta_blocks = 0;
937 ei->i_allocated_meta_blocks = 0;
938 ei->i_da_metadata_calc_len = 0;
939 ei->i_da_metadata_calc_last_lblock = 0;
940 spin_lock_init(&(ei->i_block_reservation_lock));
941 #ifdef CONFIG_QUOTA
942 ei->i_reserved_quota = 0;
943 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
944 #endif
945 ei->jinode = NULL;
946 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
947 spin_lock_init(&ei->i_completed_io_lock);
948 ei->i_sync_tid = 0;
949 ei->i_datasync_tid = 0;
950 atomic_set(&ei->i_unwritten, 0);
951 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
952 return &ei->vfs_inode;
953 }
954
955 static int ext4_drop_inode(struct inode *inode)
956 {
957 int drop = generic_drop_inode(inode);
958
959 trace_ext4_drop_inode(inode, drop);
960 return drop;
961 }
962
963 static void ext4_i_callback(struct rcu_head *head)
964 {
965 struct inode *inode = container_of(head, struct inode, i_rcu);
966 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
967 }
968
969 static void ext4_destroy_inode(struct inode *inode)
970 {
971 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
972 ext4_msg(inode->i_sb, KERN_ERR,
973 "Inode %lu (%p): orphan list check failed!",
974 inode->i_ino, EXT4_I(inode));
975 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
976 EXT4_I(inode), sizeof(struct ext4_inode_info),
977 true);
978 dump_stack();
979 }
980 call_rcu(&inode->i_rcu, ext4_i_callback);
981 }
982
983 static void init_once(void *foo)
984 {
985 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
986
987 INIT_LIST_HEAD(&ei->i_orphan);
988 init_rwsem(&ei->xattr_sem);
989 init_rwsem(&ei->i_data_sem);
990 init_rwsem(&ei->i_mmap_sem);
991 inode_init_once(&ei->vfs_inode);
992 }
993
994 static int __init init_inodecache(void)
995 {
996 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
997 sizeof(struct ext4_inode_info),
998 0, (SLAB_RECLAIM_ACCOUNT|
999 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1000 init_once);
1001 if (ext4_inode_cachep == NULL)
1002 return -ENOMEM;
1003 return 0;
1004 }
1005
1006 static void destroy_inodecache(void)
1007 {
1008 /*
1009 * Make sure all delayed rcu free inodes are flushed before we
1010 * destroy cache.
1011 */
1012 rcu_barrier();
1013 kmem_cache_destroy(ext4_inode_cachep);
1014 }
1015
1016 void ext4_clear_inode(struct inode *inode)
1017 {
1018 invalidate_inode_buffers(inode);
1019 clear_inode(inode);
1020 dquot_drop(inode);
1021 ext4_discard_preallocations(inode);
1022 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1023 if (EXT4_I(inode)->jinode) {
1024 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1025 EXT4_I(inode)->jinode);
1026 jbd2_free_inode(EXT4_I(inode)->jinode);
1027 EXT4_I(inode)->jinode = NULL;
1028 }
1029 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1030 fscrypt_put_encryption_info(inode, NULL);
1031 #endif
1032 }
1033
1034 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1035 u64 ino, u32 generation)
1036 {
1037 struct inode *inode;
1038
1039 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1040 return ERR_PTR(-ESTALE);
1041 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1042 return ERR_PTR(-ESTALE);
1043
1044 /* iget isn't really right if the inode is currently unallocated!!
1045 *
1046 * ext4_read_inode will return a bad_inode if the inode had been
1047 * deleted, so we should be safe.
1048 *
1049 * Currently we don't know the generation for parent directory, so
1050 * a generation of 0 means "accept any"
1051 */
1052 inode = ext4_iget_normal(sb, ino);
1053 if (IS_ERR(inode))
1054 return ERR_CAST(inode);
1055 if (generation && inode->i_generation != generation) {
1056 iput(inode);
1057 return ERR_PTR(-ESTALE);
1058 }
1059
1060 return inode;
1061 }
1062
1063 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1064 int fh_len, int fh_type)
1065 {
1066 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1067 ext4_nfs_get_inode);
1068 }
1069
1070 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1071 int fh_len, int fh_type)
1072 {
1073 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1074 ext4_nfs_get_inode);
1075 }
1076
1077 /*
1078 * Try to release metadata pages (indirect blocks, directories) which are
1079 * mapped via the block device. Since these pages could have journal heads
1080 * which would prevent try_to_free_buffers() from freeing them, we must use
1081 * jbd2 layer's try_to_free_buffers() function to release them.
1082 */
1083 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1084 gfp_t wait)
1085 {
1086 journal_t *journal = EXT4_SB(sb)->s_journal;
1087
1088 WARN_ON(PageChecked(page));
1089 if (!page_has_buffers(page))
1090 return 0;
1091 if (journal)
1092 return jbd2_journal_try_to_free_buffers(journal, page,
1093 wait & ~__GFP_DIRECT_RECLAIM);
1094 return try_to_free_buffers(page);
1095 }
1096
1097 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1098 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1099 {
1100 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1101 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1102 }
1103
1104 static int ext4_key_prefix(struct inode *inode, u8 **key)
1105 {
1106 *key = EXT4_SB(inode->i_sb)->key_prefix;
1107 return EXT4_SB(inode->i_sb)->key_prefix_size;
1108 }
1109
1110 static int ext4_prepare_context(struct inode *inode)
1111 {
1112 return ext4_convert_inline_data(inode);
1113 }
1114
1115 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1116 void *fs_data)
1117 {
1118 handle_t *handle = fs_data;
1119 int res, res2, retries = 0;
1120
1121 /*
1122 * If a journal handle was specified, then the encryption context is
1123 * being set on a new inode via inheritance and is part of a larger
1124 * transaction to create the inode. Otherwise the encryption context is
1125 * being set on an existing inode in its own transaction. Only in the
1126 * latter case should the "retry on ENOSPC" logic be used.
1127 */
1128
1129 if (handle) {
1130 res = ext4_xattr_set_handle(handle, inode,
1131 EXT4_XATTR_INDEX_ENCRYPTION,
1132 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1133 ctx, len, 0);
1134 if (!res) {
1135 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1136 ext4_clear_inode_state(inode,
1137 EXT4_STATE_MAY_INLINE_DATA);
1138 /*
1139 * Update inode->i_flags - e.g. S_DAX may get disabled
1140 */
1141 ext4_set_inode_flags(inode);
1142 }
1143 return res;
1144 }
1145
1146 retry:
1147 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1148 ext4_jbd2_credits_xattr(inode));
1149 if (IS_ERR(handle))
1150 return PTR_ERR(handle);
1151
1152 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1153 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1154 ctx, len, 0);
1155 if (!res) {
1156 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1157 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1158 ext4_set_inode_flags(inode);
1159 res = ext4_mark_inode_dirty(handle, inode);
1160 if (res)
1161 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1162 }
1163 res2 = ext4_journal_stop(handle);
1164
1165 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1166 goto retry;
1167 if (!res)
1168 res = res2;
1169 return res;
1170 }
1171
1172 static int ext4_dummy_context(struct inode *inode)
1173 {
1174 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1175 }
1176
1177 static unsigned ext4_max_namelen(struct inode *inode)
1178 {
1179 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1180 EXT4_NAME_LEN;
1181 }
1182
1183 static struct fscrypt_operations ext4_cryptops = {
1184 .get_context = ext4_get_context,
1185 .key_prefix = ext4_key_prefix,
1186 .prepare_context = ext4_prepare_context,
1187 .set_context = ext4_set_context,
1188 .dummy_context = ext4_dummy_context,
1189 .is_encrypted = ext4_encrypted_inode,
1190 .empty_dir = ext4_empty_dir,
1191 .max_namelen = ext4_max_namelen,
1192 };
1193 #else
1194 static struct fscrypt_operations ext4_cryptops = {
1195 .is_encrypted = ext4_encrypted_inode,
1196 };
1197 #endif
1198
1199 #ifdef CONFIG_QUOTA
1200 static char *quotatypes[] = INITQFNAMES;
1201 #define QTYPE2NAME(t) (quotatypes[t])
1202
1203 static int ext4_write_dquot(struct dquot *dquot);
1204 static int ext4_acquire_dquot(struct dquot *dquot);
1205 static int ext4_release_dquot(struct dquot *dquot);
1206 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1207 static int ext4_write_info(struct super_block *sb, int type);
1208 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1209 const struct path *path);
1210 static int ext4_quota_off(struct super_block *sb, int type);
1211 static int ext4_quota_on_mount(struct super_block *sb, int type);
1212 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1213 size_t len, loff_t off);
1214 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1215 const char *data, size_t len, loff_t off);
1216 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1217 unsigned int flags);
1218 static int ext4_enable_quotas(struct super_block *sb);
1219 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1220
1221 static struct dquot **ext4_get_dquots(struct inode *inode)
1222 {
1223 return EXT4_I(inode)->i_dquot;
1224 }
1225
1226 static const struct dquot_operations ext4_quota_operations = {
1227 .get_reserved_space = ext4_get_reserved_space,
1228 .write_dquot = ext4_write_dquot,
1229 .acquire_dquot = ext4_acquire_dquot,
1230 .release_dquot = ext4_release_dquot,
1231 .mark_dirty = ext4_mark_dquot_dirty,
1232 .write_info = ext4_write_info,
1233 .alloc_dquot = dquot_alloc,
1234 .destroy_dquot = dquot_destroy,
1235 .get_projid = ext4_get_projid,
1236 .get_next_id = ext4_get_next_id,
1237 };
1238
1239 static const struct quotactl_ops ext4_qctl_operations = {
1240 .quota_on = ext4_quota_on,
1241 .quota_off = ext4_quota_off,
1242 .quota_sync = dquot_quota_sync,
1243 .get_state = dquot_get_state,
1244 .set_info = dquot_set_dqinfo,
1245 .get_dqblk = dquot_get_dqblk,
1246 .set_dqblk = dquot_set_dqblk,
1247 .get_nextdqblk = dquot_get_next_dqblk,
1248 };
1249 #endif
1250
1251 static const struct super_operations ext4_sops = {
1252 .alloc_inode = ext4_alloc_inode,
1253 .destroy_inode = ext4_destroy_inode,
1254 .write_inode = ext4_write_inode,
1255 .dirty_inode = ext4_dirty_inode,
1256 .drop_inode = ext4_drop_inode,
1257 .evict_inode = ext4_evict_inode,
1258 .put_super = ext4_put_super,
1259 .sync_fs = ext4_sync_fs,
1260 .freeze_fs = ext4_freeze,
1261 .unfreeze_fs = ext4_unfreeze,
1262 .statfs = ext4_statfs,
1263 .remount_fs = ext4_remount,
1264 .show_options = ext4_show_options,
1265 #ifdef CONFIG_QUOTA
1266 .quota_read = ext4_quota_read,
1267 .quota_write = ext4_quota_write,
1268 .get_dquots = ext4_get_dquots,
1269 #endif
1270 .bdev_try_to_free_page = bdev_try_to_free_page,
1271 };
1272
1273 static const struct export_operations ext4_export_ops = {
1274 .fh_to_dentry = ext4_fh_to_dentry,
1275 .fh_to_parent = ext4_fh_to_parent,
1276 .get_parent = ext4_get_parent,
1277 };
1278
1279 enum {
1280 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1281 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1282 Opt_nouid32, Opt_debug, Opt_removed,
1283 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1284 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1285 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1286 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1287 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1288 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1289 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1290 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1291 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1292 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1293 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1294 Opt_lazytime, Opt_nolazytime,
1295 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1296 Opt_inode_readahead_blks, Opt_journal_ioprio,
1297 Opt_dioread_nolock, Opt_dioread_lock,
1298 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1299 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1300 };
1301
1302 static const match_table_t tokens = {
1303 {Opt_bsd_df, "bsddf"},
1304 {Opt_minix_df, "minixdf"},
1305 {Opt_grpid, "grpid"},
1306 {Opt_grpid, "bsdgroups"},
1307 {Opt_nogrpid, "nogrpid"},
1308 {Opt_nogrpid, "sysvgroups"},
1309 {Opt_resgid, "resgid=%u"},
1310 {Opt_resuid, "resuid=%u"},
1311 {Opt_sb, "sb=%u"},
1312 {Opt_err_cont, "errors=continue"},
1313 {Opt_err_panic, "errors=panic"},
1314 {Opt_err_ro, "errors=remount-ro"},
1315 {Opt_nouid32, "nouid32"},
1316 {Opt_debug, "debug"},
1317 {Opt_removed, "oldalloc"},
1318 {Opt_removed, "orlov"},
1319 {Opt_user_xattr, "user_xattr"},
1320 {Opt_nouser_xattr, "nouser_xattr"},
1321 {Opt_acl, "acl"},
1322 {Opt_noacl, "noacl"},
1323 {Opt_noload, "norecovery"},
1324 {Opt_noload, "noload"},
1325 {Opt_removed, "nobh"},
1326 {Opt_removed, "bh"},
1327 {Opt_commit, "commit=%u"},
1328 {Opt_min_batch_time, "min_batch_time=%u"},
1329 {Opt_max_batch_time, "max_batch_time=%u"},
1330 {Opt_journal_dev, "journal_dev=%u"},
1331 {Opt_journal_path, "journal_path=%s"},
1332 {Opt_journal_checksum, "journal_checksum"},
1333 {Opt_nojournal_checksum, "nojournal_checksum"},
1334 {Opt_journal_async_commit, "journal_async_commit"},
1335 {Opt_abort, "abort"},
1336 {Opt_data_journal, "data=journal"},
1337 {Opt_data_ordered, "data=ordered"},
1338 {Opt_data_writeback, "data=writeback"},
1339 {Opt_data_err_abort, "data_err=abort"},
1340 {Opt_data_err_ignore, "data_err=ignore"},
1341 {Opt_offusrjquota, "usrjquota="},
1342 {Opt_usrjquota, "usrjquota=%s"},
1343 {Opt_offgrpjquota, "grpjquota="},
1344 {Opt_grpjquota, "grpjquota=%s"},
1345 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1346 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1347 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1348 {Opt_grpquota, "grpquota"},
1349 {Opt_noquota, "noquota"},
1350 {Opt_quota, "quota"},
1351 {Opt_usrquota, "usrquota"},
1352 {Opt_prjquota, "prjquota"},
1353 {Opt_barrier, "barrier=%u"},
1354 {Opt_barrier, "barrier"},
1355 {Opt_nobarrier, "nobarrier"},
1356 {Opt_i_version, "i_version"},
1357 {Opt_dax, "dax"},
1358 {Opt_stripe, "stripe=%u"},
1359 {Opt_delalloc, "delalloc"},
1360 {Opt_lazytime, "lazytime"},
1361 {Opt_nolazytime, "nolazytime"},
1362 {Opt_nodelalloc, "nodelalloc"},
1363 {Opt_removed, "mblk_io_submit"},
1364 {Opt_removed, "nomblk_io_submit"},
1365 {Opt_block_validity, "block_validity"},
1366 {Opt_noblock_validity, "noblock_validity"},
1367 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1368 {Opt_journal_ioprio, "journal_ioprio=%u"},
1369 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1370 {Opt_auto_da_alloc, "auto_da_alloc"},
1371 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1372 {Opt_dioread_nolock, "dioread_nolock"},
1373 {Opt_dioread_lock, "dioread_lock"},
1374 {Opt_discard, "discard"},
1375 {Opt_nodiscard, "nodiscard"},
1376 {Opt_init_itable, "init_itable=%u"},
1377 {Opt_init_itable, "init_itable"},
1378 {Opt_noinit_itable, "noinit_itable"},
1379 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1380 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1381 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1382 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1383 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1384 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1385 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1386 {Opt_err, NULL},
1387 };
1388
1389 static ext4_fsblk_t get_sb_block(void **data)
1390 {
1391 ext4_fsblk_t sb_block;
1392 char *options = (char *) *data;
1393
1394 if (!options || strncmp(options, "sb=", 3) != 0)
1395 return 1; /* Default location */
1396
1397 options += 3;
1398 /* TODO: use simple_strtoll with >32bit ext4 */
1399 sb_block = simple_strtoul(options, &options, 0);
1400 if (*options && *options != ',') {
1401 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1402 (char *) *data);
1403 return 1;
1404 }
1405 if (*options == ',')
1406 options++;
1407 *data = (void *) options;
1408
1409 return sb_block;
1410 }
1411
1412 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1413 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1414 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1415
1416 #ifdef CONFIG_QUOTA
1417 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1418 {
1419 struct ext4_sb_info *sbi = EXT4_SB(sb);
1420 char *qname;
1421 int ret = -1;
1422
1423 if (sb_any_quota_loaded(sb) &&
1424 !sbi->s_qf_names[qtype]) {
1425 ext4_msg(sb, KERN_ERR,
1426 "Cannot change journaled "
1427 "quota options when quota turned on");
1428 return -1;
1429 }
1430 if (ext4_has_feature_quota(sb)) {
1431 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1432 "ignored when QUOTA feature is enabled");
1433 return 1;
1434 }
1435 qname = match_strdup(args);
1436 if (!qname) {
1437 ext4_msg(sb, KERN_ERR,
1438 "Not enough memory for storing quotafile name");
1439 return -1;
1440 }
1441 if (sbi->s_qf_names[qtype]) {
1442 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1443 ret = 1;
1444 else
1445 ext4_msg(sb, KERN_ERR,
1446 "%s quota file already specified",
1447 QTYPE2NAME(qtype));
1448 goto errout;
1449 }
1450 if (strchr(qname, '/')) {
1451 ext4_msg(sb, KERN_ERR,
1452 "quotafile must be on filesystem root");
1453 goto errout;
1454 }
1455 sbi->s_qf_names[qtype] = qname;
1456 set_opt(sb, QUOTA);
1457 return 1;
1458 errout:
1459 kfree(qname);
1460 return ret;
1461 }
1462
1463 static int clear_qf_name(struct super_block *sb, int qtype)
1464 {
1465
1466 struct ext4_sb_info *sbi = EXT4_SB(sb);
1467
1468 if (sb_any_quota_loaded(sb) &&
1469 sbi->s_qf_names[qtype]) {
1470 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1471 " when quota turned on");
1472 return -1;
1473 }
1474 kfree(sbi->s_qf_names[qtype]);
1475 sbi->s_qf_names[qtype] = NULL;
1476 return 1;
1477 }
1478 #endif
1479
1480 #define MOPT_SET 0x0001
1481 #define MOPT_CLEAR 0x0002
1482 #define MOPT_NOSUPPORT 0x0004
1483 #define MOPT_EXPLICIT 0x0008
1484 #define MOPT_CLEAR_ERR 0x0010
1485 #define MOPT_GTE0 0x0020
1486 #ifdef CONFIG_QUOTA
1487 #define MOPT_Q 0
1488 #define MOPT_QFMT 0x0040
1489 #else
1490 #define MOPT_Q MOPT_NOSUPPORT
1491 #define MOPT_QFMT MOPT_NOSUPPORT
1492 #endif
1493 #define MOPT_DATAJ 0x0080
1494 #define MOPT_NO_EXT2 0x0100
1495 #define MOPT_NO_EXT3 0x0200
1496 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1497 #define MOPT_STRING 0x0400
1498
1499 static const struct mount_opts {
1500 int token;
1501 int mount_opt;
1502 int flags;
1503 } ext4_mount_opts[] = {
1504 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1505 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1506 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1507 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1508 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1509 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1510 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1511 MOPT_EXT4_ONLY | MOPT_SET},
1512 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1513 MOPT_EXT4_ONLY | MOPT_CLEAR},
1514 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1515 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1516 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1517 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1518 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1519 MOPT_EXT4_ONLY | MOPT_CLEAR},
1520 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1521 MOPT_EXT4_ONLY | MOPT_CLEAR},
1522 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1523 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1524 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1525 EXT4_MOUNT_JOURNAL_CHECKSUM),
1526 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1527 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1528 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1529 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1530 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1531 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1532 MOPT_NO_EXT2},
1533 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1534 MOPT_NO_EXT2},
1535 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1536 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1537 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1538 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1539 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1540 {Opt_commit, 0, MOPT_GTE0},
1541 {Opt_max_batch_time, 0, MOPT_GTE0},
1542 {Opt_min_batch_time, 0, MOPT_GTE0},
1543 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1544 {Opt_init_itable, 0, MOPT_GTE0},
1545 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1546 {Opt_stripe, 0, MOPT_GTE0},
1547 {Opt_resuid, 0, MOPT_GTE0},
1548 {Opt_resgid, 0, MOPT_GTE0},
1549 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1550 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1551 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1552 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1553 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1554 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1555 MOPT_NO_EXT2 | MOPT_DATAJ},
1556 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1557 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1558 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1559 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1560 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1561 #else
1562 {Opt_acl, 0, MOPT_NOSUPPORT},
1563 {Opt_noacl, 0, MOPT_NOSUPPORT},
1564 #endif
1565 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1566 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1567 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1568 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1569 MOPT_SET | MOPT_Q},
1570 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1571 MOPT_SET | MOPT_Q},
1572 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1573 MOPT_SET | MOPT_Q},
1574 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1575 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1576 MOPT_CLEAR | MOPT_Q},
1577 {Opt_usrjquota, 0, MOPT_Q},
1578 {Opt_grpjquota, 0, MOPT_Q},
1579 {Opt_offusrjquota, 0, MOPT_Q},
1580 {Opt_offgrpjquota, 0, MOPT_Q},
1581 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1582 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1583 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1584 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1585 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1586 {Opt_err, 0, 0}
1587 };
1588
1589 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1590 substring_t *args, unsigned long *journal_devnum,
1591 unsigned int *journal_ioprio, int is_remount)
1592 {
1593 struct ext4_sb_info *sbi = EXT4_SB(sb);
1594 const struct mount_opts *m;
1595 kuid_t uid;
1596 kgid_t gid;
1597 int arg = 0;
1598
1599 #ifdef CONFIG_QUOTA
1600 if (token == Opt_usrjquota)
1601 return set_qf_name(sb, USRQUOTA, &args[0]);
1602 else if (token == Opt_grpjquota)
1603 return set_qf_name(sb, GRPQUOTA, &args[0]);
1604 else if (token == Opt_offusrjquota)
1605 return clear_qf_name(sb, USRQUOTA);
1606 else if (token == Opt_offgrpjquota)
1607 return clear_qf_name(sb, GRPQUOTA);
1608 #endif
1609 switch (token) {
1610 case Opt_noacl:
1611 case Opt_nouser_xattr:
1612 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1613 break;
1614 case Opt_sb:
1615 return 1; /* handled by get_sb_block() */
1616 case Opt_removed:
1617 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1618 return 1;
1619 case Opt_abort:
1620 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1621 return 1;
1622 case Opt_i_version:
1623 sb->s_flags |= MS_I_VERSION;
1624 return 1;
1625 case Opt_lazytime:
1626 sb->s_flags |= MS_LAZYTIME;
1627 return 1;
1628 case Opt_nolazytime:
1629 sb->s_flags &= ~MS_LAZYTIME;
1630 return 1;
1631 }
1632
1633 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1634 if (token == m->token)
1635 break;
1636
1637 if (m->token == Opt_err) {
1638 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1639 "or missing value", opt);
1640 return -1;
1641 }
1642
1643 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1644 ext4_msg(sb, KERN_ERR,
1645 "Mount option \"%s\" incompatible with ext2", opt);
1646 return -1;
1647 }
1648 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1649 ext4_msg(sb, KERN_ERR,
1650 "Mount option \"%s\" incompatible with ext3", opt);
1651 return -1;
1652 }
1653
1654 if (token == Opt_err_panic && !capable(CAP_SYS_ADMIN)) {
1655 ext4_msg(sb, KERN_ERR,
1656 "Mount option \"%s\" not allowed for unprivileged mounts",
1657 opt);
1658 return -1;
1659 }
1660
1661 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1662 return -1;
1663 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1664 return -1;
1665 if (m->flags & MOPT_EXPLICIT) {
1666 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1667 set_opt2(sb, EXPLICIT_DELALLOC);
1668 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1669 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1670 } else
1671 return -1;
1672 }
1673 if (m->flags & MOPT_CLEAR_ERR)
1674 clear_opt(sb, ERRORS_MASK);
1675 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1676 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1677 "options when quota turned on");
1678 return -1;
1679 }
1680
1681 if (m->flags & MOPT_NOSUPPORT) {
1682 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1683 } else if (token == Opt_commit) {
1684 if (arg == 0)
1685 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1686 sbi->s_commit_interval = HZ * arg;
1687 } else if (token == Opt_max_batch_time) {
1688 sbi->s_max_batch_time = arg;
1689 } else if (token == Opt_min_batch_time) {
1690 sbi->s_min_batch_time = arg;
1691 } else if (token == Opt_inode_readahead_blks) {
1692 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1693 ext4_msg(sb, KERN_ERR,
1694 "EXT4-fs: inode_readahead_blks must be "
1695 "0 or a power of 2 smaller than 2^31");
1696 return -1;
1697 }
1698 sbi->s_inode_readahead_blks = arg;
1699 } else if (token == Opt_init_itable) {
1700 set_opt(sb, INIT_INODE_TABLE);
1701 if (!args->from)
1702 arg = EXT4_DEF_LI_WAIT_MULT;
1703 sbi->s_li_wait_mult = arg;
1704 } else if (token == Opt_max_dir_size_kb) {
1705 sbi->s_max_dir_size_kb = arg;
1706 } else if (token == Opt_stripe) {
1707 sbi->s_stripe = arg;
1708 } else if (token == Opt_resuid) {
1709 uid = make_kuid(sb->s_user_ns, arg);
1710 if (!uid_valid(uid)) {
1711 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1712 return -1;
1713 }
1714 sbi->s_resuid = uid;
1715 } else if (token == Opt_resgid) {
1716 gid = make_kgid(sb->s_user_ns, arg);
1717 if (!gid_valid(gid)) {
1718 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1719 return -1;
1720 }
1721 sbi->s_resgid = gid;
1722 } else if (token == Opt_journal_dev) {
1723 if (is_remount) {
1724 ext4_msg(sb, KERN_ERR,
1725 "Cannot specify journal on remount");
1726 return -1;
1727 }
1728 *journal_devnum = arg;
1729 } else if (token == Opt_journal_path) {
1730 char *journal_path;
1731 struct inode *journal_inode;
1732 struct path path;
1733 int error;
1734
1735 if (is_remount) {
1736 ext4_msg(sb, KERN_ERR,
1737 "Cannot specify journal on remount");
1738 return -1;
1739 }
1740 journal_path = match_strdup(&args[0]);
1741 if (!journal_path) {
1742 ext4_msg(sb, KERN_ERR, "error: could not dup "
1743 "journal device string");
1744 return -1;
1745 }
1746
1747 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1748 if (error) {
1749 ext4_msg(sb, KERN_ERR, "error: could not find "
1750 "journal device path: error %d", error);
1751 kfree(journal_path);
1752 return -1;
1753 }
1754
1755 /*
1756 * Refuse access for unprivileged mounts if the user does
1757 * not have rw access to the journal device via the supplied
1758 * path.
1759 */
1760 if (!capable(CAP_SYS_ADMIN) &&
1761 inode_permission(d_inode(path.dentry), MAY_READ|MAY_WRITE)) {
1762 ext4_msg(sb, KERN_ERR,
1763 "error: Insufficient access to journal path %s",
1764 journal_path);
1765 return -1;
1766 }
1767
1768 journal_inode = d_inode(path.dentry);
1769 if (!S_ISBLK(journal_inode->i_mode)) {
1770 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1771 "is not a block device", journal_path);
1772 path_put(&path);
1773 kfree(journal_path);
1774 return -1;
1775 }
1776
1777 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1778 path_put(&path);
1779 kfree(journal_path);
1780 } else if (token == Opt_journal_ioprio) {
1781 if (arg > 7) {
1782 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1783 " (must be 0-7)");
1784 return -1;
1785 }
1786 *journal_ioprio =
1787 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1788 } else if (token == Opt_test_dummy_encryption) {
1789 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1790 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1791 ext4_msg(sb, KERN_WARNING,
1792 "Test dummy encryption mode enabled");
1793 #else
1794 ext4_msg(sb, KERN_WARNING,
1795 "Test dummy encryption mount option ignored");
1796 #endif
1797 } else if (m->flags & MOPT_DATAJ) {
1798 if (is_remount) {
1799 if (!sbi->s_journal)
1800 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1801 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1802 ext4_msg(sb, KERN_ERR,
1803 "Cannot change data mode on remount");
1804 return -1;
1805 }
1806 } else {
1807 clear_opt(sb, DATA_FLAGS);
1808 sbi->s_mount_opt |= m->mount_opt;
1809 }
1810 #ifdef CONFIG_QUOTA
1811 } else if (m->flags & MOPT_QFMT) {
1812 if (sb_any_quota_loaded(sb) &&
1813 sbi->s_jquota_fmt != m->mount_opt) {
1814 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1815 "quota options when quota turned on");
1816 return -1;
1817 }
1818 if (ext4_has_feature_quota(sb)) {
1819 ext4_msg(sb, KERN_INFO,
1820 "Quota format mount options ignored "
1821 "when QUOTA feature is enabled");
1822 return 1;
1823 }
1824 sbi->s_jquota_fmt = m->mount_opt;
1825 #endif
1826 } else if (token == Opt_dax) {
1827 #ifdef CONFIG_FS_DAX
1828 ext4_msg(sb, KERN_WARNING,
1829 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1830 sbi->s_mount_opt |= m->mount_opt;
1831 #else
1832 ext4_msg(sb, KERN_INFO, "dax option not supported");
1833 return -1;
1834 #endif
1835 } else if (token == Opt_data_err_abort) {
1836 sbi->s_mount_opt |= m->mount_opt;
1837 } else if (token == Opt_data_err_ignore) {
1838 sbi->s_mount_opt &= ~m->mount_opt;
1839 } else {
1840 if (!args->from)
1841 arg = 1;
1842 if (m->flags & MOPT_CLEAR)
1843 arg = !arg;
1844 else if (unlikely(!(m->flags & MOPT_SET))) {
1845 ext4_msg(sb, KERN_WARNING,
1846 "buggy handling of option %s", opt);
1847 WARN_ON(1);
1848 return -1;
1849 }
1850 if (arg != 0)
1851 sbi->s_mount_opt |= m->mount_opt;
1852 else
1853 sbi->s_mount_opt &= ~m->mount_opt;
1854 }
1855 return 1;
1856 }
1857
1858 static int parse_options(char *options, struct super_block *sb,
1859 unsigned long *journal_devnum,
1860 unsigned int *journal_ioprio,
1861 int is_remount)
1862 {
1863 struct ext4_sb_info *sbi = EXT4_SB(sb);
1864 char *p;
1865 substring_t args[MAX_OPT_ARGS];
1866 int token;
1867
1868 if (!options)
1869 return 1;
1870
1871 while ((p = strsep(&options, ",")) != NULL) {
1872 if (!*p)
1873 continue;
1874 /*
1875 * Initialize args struct so we know whether arg was
1876 * found; some options take optional arguments.
1877 */
1878 args[0].to = args[0].from = NULL;
1879 token = match_token(p, tokens, args);
1880 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1881 journal_ioprio, is_remount) < 0)
1882 return 0;
1883 }
1884 #ifdef CONFIG_QUOTA
1885 /*
1886 * We do the test below only for project quotas. 'usrquota' and
1887 * 'grpquota' mount options are allowed even without quota feature
1888 * to support legacy quotas in quota files.
1889 */
1890 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1891 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1892 "Cannot enable project quota enforcement.");
1893 return 0;
1894 }
1895 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1896 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1897 clear_opt(sb, USRQUOTA);
1898
1899 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1900 clear_opt(sb, GRPQUOTA);
1901
1902 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1903 ext4_msg(sb, KERN_ERR, "old and new quota "
1904 "format mixing");
1905 return 0;
1906 }
1907
1908 if (!sbi->s_jquota_fmt) {
1909 ext4_msg(sb, KERN_ERR, "journaled quota format "
1910 "not specified");
1911 return 0;
1912 }
1913 }
1914 #endif
1915 if (test_opt(sb, DIOREAD_NOLOCK)) {
1916 int blocksize =
1917 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1918
1919 if (blocksize < PAGE_SIZE) {
1920 ext4_msg(sb, KERN_ERR, "can't mount with "
1921 "dioread_nolock if block size != PAGE_SIZE");
1922 return 0;
1923 }
1924 }
1925 return 1;
1926 }
1927
1928 static inline void ext4_show_quota_options(struct seq_file *seq,
1929 struct super_block *sb)
1930 {
1931 #if defined(CONFIG_QUOTA)
1932 struct ext4_sb_info *sbi = EXT4_SB(sb);
1933
1934 if (sbi->s_jquota_fmt) {
1935 char *fmtname = "";
1936
1937 switch (sbi->s_jquota_fmt) {
1938 case QFMT_VFS_OLD:
1939 fmtname = "vfsold";
1940 break;
1941 case QFMT_VFS_V0:
1942 fmtname = "vfsv0";
1943 break;
1944 case QFMT_VFS_V1:
1945 fmtname = "vfsv1";
1946 break;
1947 }
1948 seq_printf(seq, ",jqfmt=%s", fmtname);
1949 }
1950
1951 if (sbi->s_qf_names[USRQUOTA])
1952 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1953
1954 if (sbi->s_qf_names[GRPQUOTA])
1955 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1956 #endif
1957 }
1958
1959 static const char *token2str(int token)
1960 {
1961 const struct match_token *t;
1962
1963 for (t = tokens; t->token != Opt_err; t++)
1964 if (t->token == token && !strchr(t->pattern, '='))
1965 break;
1966 return t->pattern;
1967 }
1968
1969 /*
1970 * Show an option if
1971 * - it's set to a non-default value OR
1972 * - if the per-sb default is different from the global default
1973 */
1974 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1975 int nodefs)
1976 {
1977 struct ext4_sb_info *sbi = EXT4_SB(sb);
1978 struct ext4_super_block *es = sbi->s_es;
1979 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1980 const struct mount_opts *m;
1981 char sep = nodefs ? '\n' : ',';
1982
1983 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1984 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1985
1986 if (sbi->s_sb_block != 1)
1987 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1988
1989 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1990 int want_set = m->flags & MOPT_SET;
1991 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1992 (m->flags & MOPT_CLEAR_ERR))
1993 continue;
1994 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1995 continue; /* skip if same as the default */
1996 if ((want_set &&
1997 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1998 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1999 continue; /* select Opt_noFoo vs Opt_Foo */
2000 SEQ_OPTS_PRINT("%s", token2str(m->token));
2001 }
2002
2003 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(sb->s_user_ns, EXT4_DEF_RESUID)) ||
2004 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2005 SEQ_OPTS_PRINT("resuid=%u",
2006 from_kuid_munged(sb->s_user_ns, sbi->s_resuid));
2007 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(sb->s_user_ns, EXT4_DEF_RESGID)) ||
2008 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2009 SEQ_OPTS_PRINT("resgid=%u",
2010 from_kgid_munged(sb->s_user_ns, sbi->s_resgid));
2011 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2012 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2013 SEQ_OPTS_PUTS("errors=remount-ro");
2014 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2015 SEQ_OPTS_PUTS("errors=continue");
2016 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2017 SEQ_OPTS_PUTS("errors=panic");
2018 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2019 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2020 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2021 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2022 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2023 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2024 if (sb->s_flags & MS_I_VERSION)
2025 SEQ_OPTS_PUTS("i_version");
2026 if (nodefs || sbi->s_stripe)
2027 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2028 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2029 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2030 SEQ_OPTS_PUTS("data=journal");
2031 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2032 SEQ_OPTS_PUTS("data=ordered");
2033 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2034 SEQ_OPTS_PUTS("data=writeback");
2035 }
2036 if (nodefs ||
2037 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2038 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2039 sbi->s_inode_readahead_blks);
2040
2041 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2042 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2043 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2044 if (nodefs || sbi->s_max_dir_size_kb)
2045 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2046 if (test_opt(sb, DATA_ERR_ABORT))
2047 SEQ_OPTS_PUTS("data_err=abort");
2048
2049 ext4_show_quota_options(seq, sb);
2050 return 0;
2051 }
2052
2053 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2054 {
2055 return _ext4_show_options(seq, root->d_sb, 0);
2056 }
2057
2058 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2059 {
2060 struct super_block *sb = seq->private;
2061 int rc;
2062
2063 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2064 rc = _ext4_show_options(seq, sb, 1);
2065 seq_puts(seq, "\n");
2066 return rc;
2067 }
2068
2069 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2070 int read_only)
2071 {
2072 struct ext4_sb_info *sbi = EXT4_SB(sb);
2073 int res = 0;
2074
2075 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2076 ext4_msg(sb, KERN_ERR, "revision level too high, "
2077 "forcing read-only mode");
2078 res = MS_RDONLY;
2079 }
2080 if (read_only)
2081 goto done;
2082 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2083 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2084 "running e2fsck is recommended");
2085 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2086 ext4_msg(sb, KERN_WARNING,
2087 "warning: mounting fs with errors, "
2088 "running e2fsck is recommended");
2089 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2090 le16_to_cpu(es->s_mnt_count) >=
2091 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2092 ext4_msg(sb, KERN_WARNING,
2093 "warning: maximal mount count reached, "
2094 "running e2fsck is recommended");
2095 else if (le32_to_cpu(es->s_checkinterval) &&
2096 (le32_to_cpu(es->s_lastcheck) +
2097 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2098 ext4_msg(sb, KERN_WARNING,
2099 "warning: checktime reached, "
2100 "running e2fsck is recommended");
2101 if (!sbi->s_journal)
2102 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2103 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2104 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2105 le16_add_cpu(&es->s_mnt_count, 1);
2106 es->s_mtime = cpu_to_le32(get_seconds());
2107 ext4_update_dynamic_rev(sb);
2108 if (sbi->s_journal)
2109 ext4_set_feature_journal_needs_recovery(sb);
2110
2111 ext4_commit_super(sb, 1);
2112 done:
2113 if (test_opt(sb, DEBUG))
2114 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2115 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2116 sb->s_blocksize,
2117 sbi->s_groups_count,
2118 EXT4_BLOCKS_PER_GROUP(sb),
2119 EXT4_INODES_PER_GROUP(sb),
2120 sbi->s_mount_opt, sbi->s_mount_opt2);
2121
2122 cleancache_init_fs(sb);
2123 return res;
2124 }
2125
2126 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2127 {
2128 struct ext4_sb_info *sbi = EXT4_SB(sb);
2129 struct flex_groups *new_groups;
2130 int size;
2131
2132 if (!sbi->s_log_groups_per_flex)
2133 return 0;
2134
2135 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2136 if (size <= sbi->s_flex_groups_allocated)
2137 return 0;
2138
2139 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2140 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2141 if (!new_groups) {
2142 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2143 size / (int) sizeof(struct flex_groups));
2144 return -ENOMEM;
2145 }
2146
2147 if (sbi->s_flex_groups) {
2148 memcpy(new_groups, sbi->s_flex_groups,
2149 (sbi->s_flex_groups_allocated *
2150 sizeof(struct flex_groups)));
2151 kvfree(sbi->s_flex_groups);
2152 }
2153 sbi->s_flex_groups = new_groups;
2154 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2155 return 0;
2156 }
2157
2158 static int ext4_fill_flex_info(struct super_block *sb)
2159 {
2160 struct ext4_sb_info *sbi = EXT4_SB(sb);
2161 struct ext4_group_desc *gdp = NULL;
2162 ext4_group_t flex_group;
2163 int i, err;
2164
2165 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2166 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2167 sbi->s_log_groups_per_flex = 0;
2168 return 1;
2169 }
2170
2171 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2172 if (err)
2173 goto failed;
2174
2175 for (i = 0; i < sbi->s_groups_count; i++) {
2176 gdp = ext4_get_group_desc(sb, i, NULL);
2177
2178 flex_group = ext4_flex_group(sbi, i);
2179 atomic_add(ext4_free_inodes_count(sb, gdp),
2180 &sbi->s_flex_groups[flex_group].free_inodes);
2181 atomic64_add(ext4_free_group_clusters(sb, gdp),
2182 &sbi->s_flex_groups[flex_group].free_clusters);
2183 atomic_add(ext4_used_dirs_count(sb, gdp),
2184 &sbi->s_flex_groups[flex_group].used_dirs);
2185 }
2186
2187 return 1;
2188 failed:
2189 return 0;
2190 }
2191
2192 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2193 struct ext4_group_desc *gdp)
2194 {
2195 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2196 __u16 crc = 0;
2197 __le32 le_group = cpu_to_le32(block_group);
2198 struct ext4_sb_info *sbi = EXT4_SB(sb);
2199
2200 if (ext4_has_metadata_csum(sbi->s_sb)) {
2201 /* Use new metadata_csum algorithm */
2202 __u32 csum32;
2203 __u16 dummy_csum = 0;
2204
2205 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2206 sizeof(le_group));
2207 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2208 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2209 sizeof(dummy_csum));
2210 offset += sizeof(dummy_csum);
2211 if (offset < sbi->s_desc_size)
2212 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2213 sbi->s_desc_size - offset);
2214
2215 crc = csum32 & 0xFFFF;
2216 goto out;
2217 }
2218
2219 /* old crc16 code */
2220 if (!ext4_has_feature_gdt_csum(sb))
2221 return 0;
2222
2223 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2224 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2225 crc = crc16(crc, (__u8 *)gdp, offset);
2226 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2227 /* for checksum of struct ext4_group_desc do the rest...*/
2228 if (ext4_has_feature_64bit(sb) &&
2229 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2230 crc = crc16(crc, (__u8 *)gdp + offset,
2231 le16_to_cpu(sbi->s_es->s_desc_size) -
2232 offset);
2233
2234 out:
2235 return cpu_to_le16(crc);
2236 }
2237
2238 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2239 struct ext4_group_desc *gdp)
2240 {
2241 if (ext4_has_group_desc_csum(sb) &&
2242 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2243 return 0;
2244
2245 return 1;
2246 }
2247
2248 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2249 struct ext4_group_desc *gdp)
2250 {
2251 if (!ext4_has_group_desc_csum(sb))
2252 return;
2253 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2254 }
2255
2256 /* Called at mount-time, super-block is locked */
2257 static int ext4_check_descriptors(struct super_block *sb,
2258 ext4_fsblk_t sb_block,
2259 ext4_group_t *first_not_zeroed)
2260 {
2261 struct ext4_sb_info *sbi = EXT4_SB(sb);
2262 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2263 ext4_fsblk_t last_block;
2264 ext4_fsblk_t block_bitmap;
2265 ext4_fsblk_t inode_bitmap;
2266 ext4_fsblk_t inode_table;
2267 int flexbg_flag = 0;
2268 ext4_group_t i, grp = sbi->s_groups_count;
2269
2270 if (ext4_has_feature_flex_bg(sb))
2271 flexbg_flag = 1;
2272
2273 ext4_debug("Checking group descriptors");
2274
2275 for (i = 0; i < sbi->s_groups_count; i++) {
2276 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2277
2278 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2279 last_block = ext4_blocks_count(sbi->s_es) - 1;
2280 else
2281 last_block = first_block +
2282 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2283
2284 if ((grp == sbi->s_groups_count) &&
2285 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2286 grp = i;
2287
2288 block_bitmap = ext4_block_bitmap(sb, gdp);
2289 if (block_bitmap == sb_block) {
2290 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2291 "Block bitmap for group %u overlaps "
2292 "superblock", i);
2293 }
2294 if (block_bitmap < first_block || block_bitmap > last_block) {
2295 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2296 "Block bitmap for group %u not in group "
2297 "(block %llu)!", i, block_bitmap);
2298 return 0;
2299 }
2300 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2301 if (inode_bitmap == sb_block) {
2302 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2303 "Inode bitmap for group %u overlaps "
2304 "superblock", i);
2305 }
2306 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2307 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2308 "Inode bitmap for group %u not in group "
2309 "(block %llu)!", i, inode_bitmap);
2310 return 0;
2311 }
2312 inode_table = ext4_inode_table(sb, gdp);
2313 if (inode_table == sb_block) {
2314 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2315 "Inode table for group %u overlaps "
2316 "superblock", i);
2317 }
2318 if (inode_table < first_block ||
2319 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2320 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2321 "Inode table for group %u not in group "
2322 "(block %llu)!", i, inode_table);
2323 return 0;
2324 }
2325 ext4_lock_group(sb, i);
2326 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2327 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2328 "Checksum for group %u failed (%u!=%u)",
2329 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2330 gdp)), le16_to_cpu(gdp->bg_checksum));
2331 if (!(sb->s_flags & MS_RDONLY)) {
2332 ext4_unlock_group(sb, i);
2333 return 0;
2334 }
2335 }
2336 ext4_unlock_group(sb, i);
2337 if (!flexbg_flag)
2338 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2339 }
2340 if (NULL != first_not_zeroed)
2341 *first_not_zeroed = grp;
2342 return 1;
2343 }
2344
2345 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2346 * the superblock) which were deleted from all directories, but held open by
2347 * a process at the time of a crash. We walk the list and try to delete these
2348 * inodes at recovery time (only with a read-write filesystem).
2349 *
2350 * In order to keep the orphan inode chain consistent during traversal (in
2351 * case of crash during recovery), we link each inode into the superblock
2352 * orphan list_head and handle it the same way as an inode deletion during
2353 * normal operation (which journals the operations for us).
2354 *
2355 * We only do an iget() and an iput() on each inode, which is very safe if we
2356 * accidentally point at an in-use or already deleted inode. The worst that
2357 * can happen in this case is that we get a "bit already cleared" message from
2358 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2359 * e2fsck was run on this filesystem, and it must have already done the orphan
2360 * inode cleanup for us, so we can safely abort without any further action.
2361 */
2362 static void ext4_orphan_cleanup(struct super_block *sb,
2363 struct ext4_super_block *es)
2364 {
2365 unsigned int s_flags = sb->s_flags;
2366 int ret, nr_orphans = 0, nr_truncates = 0;
2367 #ifdef CONFIG_QUOTA
2368 int i;
2369 #endif
2370 if (!es->s_last_orphan) {
2371 jbd_debug(4, "no orphan inodes to clean up\n");
2372 return;
2373 }
2374
2375 if (bdev_read_only(sb->s_bdev)) {
2376 ext4_msg(sb, KERN_ERR, "write access "
2377 "unavailable, skipping orphan cleanup");
2378 return;
2379 }
2380
2381 /* Check if feature set would not allow a r/w mount */
2382 if (!ext4_feature_set_ok(sb, 0)) {
2383 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2384 "unknown ROCOMPAT features");
2385 return;
2386 }
2387
2388 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2389 /* don't clear list on RO mount w/ errors */
2390 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2391 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2392 "clearing orphan list.\n");
2393 es->s_last_orphan = 0;
2394 }
2395 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2396 return;
2397 }
2398
2399 if (s_flags & MS_RDONLY) {
2400 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2401 sb->s_flags &= ~MS_RDONLY;
2402 }
2403 #ifdef CONFIG_QUOTA
2404 /* Needed for iput() to work correctly and not trash data */
2405 sb->s_flags |= MS_ACTIVE;
2406 /* Turn on quotas so that they are updated correctly */
2407 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2408 if (EXT4_SB(sb)->s_qf_names[i]) {
2409 int ret = ext4_quota_on_mount(sb, i);
2410 if (ret < 0)
2411 ext4_msg(sb, KERN_ERR,
2412 "Cannot turn on journaled "
2413 "quota: error %d", ret);
2414 }
2415 }
2416 #endif
2417
2418 while (es->s_last_orphan) {
2419 struct inode *inode;
2420
2421 /*
2422 * We may have encountered an error during cleanup; if
2423 * so, skip the rest.
2424 */
2425 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2426 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2427 es->s_last_orphan = 0;
2428 break;
2429 }
2430
2431 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2432 if (IS_ERR(inode)) {
2433 es->s_last_orphan = 0;
2434 break;
2435 }
2436
2437 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2438 dquot_initialize(inode);
2439 if (inode->i_nlink) {
2440 if (test_opt(sb, DEBUG))
2441 ext4_msg(sb, KERN_DEBUG,
2442 "%s: truncating inode %lu to %lld bytes",
2443 __func__, inode->i_ino, inode->i_size);
2444 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2445 inode->i_ino, inode->i_size);
2446 inode_lock(inode);
2447 truncate_inode_pages(inode->i_mapping, inode->i_size);
2448 ret = ext4_truncate(inode);
2449 if (ret)
2450 ext4_std_error(inode->i_sb, ret);
2451 inode_unlock(inode);
2452 nr_truncates++;
2453 } else {
2454 if (test_opt(sb, DEBUG))
2455 ext4_msg(sb, KERN_DEBUG,
2456 "%s: deleting unreferenced inode %lu",
2457 __func__, inode->i_ino);
2458 jbd_debug(2, "deleting unreferenced inode %lu\n",
2459 inode->i_ino);
2460 nr_orphans++;
2461 }
2462 iput(inode); /* The delete magic happens here! */
2463 }
2464
2465 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2466
2467 if (nr_orphans)
2468 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2469 PLURAL(nr_orphans));
2470 if (nr_truncates)
2471 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2472 PLURAL(nr_truncates));
2473 #ifdef CONFIG_QUOTA
2474 /* Turn quotas off */
2475 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2476 if (sb_dqopt(sb)->files[i])
2477 dquot_quota_off(sb, i);
2478 }
2479 #endif
2480 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2481 }
2482
2483 /*
2484 * Maximal extent format file size.
2485 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2486 * extent format containers, within a sector_t, and within i_blocks
2487 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2488 * so that won't be a limiting factor.
2489 *
2490 * However there is other limiting factor. We do store extents in the form
2491 * of starting block and length, hence the resulting length of the extent
2492 * covering maximum file size must fit into on-disk format containers as
2493 * well. Given that length is always by 1 unit bigger than max unit (because
2494 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2495 *
2496 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2497 */
2498 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2499 {
2500 loff_t res;
2501 loff_t upper_limit = MAX_LFS_FILESIZE;
2502
2503 /* small i_blocks in vfs inode? */
2504 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2505 /*
2506 * CONFIG_LBDAF is not enabled implies the inode
2507 * i_block represent total blocks in 512 bytes
2508 * 32 == size of vfs inode i_blocks * 8
2509 */
2510 upper_limit = (1LL << 32) - 1;
2511
2512 /* total blocks in file system block size */
2513 upper_limit >>= (blkbits - 9);
2514 upper_limit <<= blkbits;
2515 }
2516
2517 /*
2518 * 32-bit extent-start container, ee_block. We lower the maxbytes
2519 * by one fs block, so ee_len can cover the extent of maximum file
2520 * size
2521 */
2522 res = (1LL << 32) - 1;
2523 res <<= blkbits;
2524
2525 /* Sanity check against vm- & vfs- imposed limits */
2526 if (res > upper_limit)
2527 res = upper_limit;
2528
2529 return res;
2530 }
2531
2532 /*
2533 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2534 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2535 * We need to be 1 filesystem block less than the 2^48 sector limit.
2536 */
2537 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2538 {
2539 loff_t res = EXT4_NDIR_BLOCKS;
2540 int meta_blocks;
2541 loff_t upper_limit;
2542 /* This is calculated to be the largest file size for a dense, block
2543 * mapped file such that the file's total number of 512-byte sectors,
2544 * including data and all indirect blocks, does not exceed (2^48 - 1).
2545 *
2546 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2547 * number of 512-byte sectors of the file.
2548 */
2549
2550 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2551 /*
2552 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2553 * the inode i_block field represents total file blocks in
2554 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2555 */
2556 upper_limit = (1LL << 32) - 1;
2557
2558 /* total blocks in file system block size */
2559 upper_limit >>= (bits - 9);
2560
2561 } else {
2562 /*
2563 * We use 48 bit ext4_inode i_blocks
2564 * With EXT4_HUGE_FILE_FL set the i_blocks
2565 * represent total number of blocks in
2566 * file system block size
2567 */
2568 upper_limit = (1LL << 48) - 1;
2569
2570 }
2571
2572 /* indirect blocks */
2573 meta_blocks = 1;
2574 /* double indirect blocks */
2575 meta_blocks += 1 + (1LL << (bits-2));
2576 /* tripple indirect blocks */
2577 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2578
2579 upper_limit -= meta_blocks;
2580 upper_limit <<= bits;
2581
2582 res += 1LL << (bits-2);
2583 res += 1LL << (2*(bits-2));
2584 res += 1LL << (3*(bits-2));
2585 res <<= bits;
2586 if (res > upper_limit)
2587 res = upper_limit;
2588
2589 if (res > MAX_LFS_FILESIZE)
2590 res = MAX_LFS_FILESIZE;
2591
2592 return res;
2593 }
2594
2595 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2596 ext4_fsblk_t logical_sb_block, int nr)
2597 {
2598 struct ext4_sb_info *sbi = EXT4_SB(sb);
2599 ext4_group_t bg, first_meta_bg;
2600 int has_super = 0;
2601
2602 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2603
2604 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2605 return logical_sb_block + nr + 1;
2606 bg = sbi->s_desc_per_block * nr;
2607 if (ext4_bg_has_super(sb, bg))
2608 has_super = 1;
2609
2610 /*
2611 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2612 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2613 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2614 * compensate.
2615 */
2616 if (sb->s_blocksize == 1024 && nr == 0 &&
2617 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2618 has_super++;
2619
2620 return (has_super + ext4_group_first_block_no(sb, bg));
2621 }
2622
2623 /**
2624 * ext4_get_stripe_size: Get the stripe size.
2625 * @sbi: In memory super block info
2626 *
2627 * If we have specified it via mount option, then
2628 * use the mount option value. If the value specified at mount time is
2629 * greater than the blocks per group use the super block value.
2630 * If the super block value is greater than blocks per group return 0.
2631 * Allocator needs it be less than blocks per group.
2632 *
2633 */
2634 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2635 {
2636 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2637 unsigned long stripe_width =
2638 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2639 int ret;
2640
2641 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2642 ret = sbi->s_stripe;
2643 else if (stripe_width <= sbi->s_blocks_per_group)
2644 ret = stripe_width;
2645 else if (stride <= sbi->s_blocks_per_group)
2646 ret = stride;
2647 else
2648 ret = 0;
2649
2650 /*
2651 * If the stripe width is 1, this makes no sense and
2652 * we set it to 0 to turn off stripe handling code.
2653 */
2654 if (ret <= 1)
2655 ret = 0;
2656
2657 return ret;
2658 }
2659
2660 /*
2661 * Check whether this filesystem can be mounted based on
2662 * the features present and the RDONLY/RDWR mount requested.
2663 * Returns 1 if this filesystem can be mounted as requested,
2664 * 0 if it cannot be.
2665 */
2666 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2667 {
2668 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2669 ext4_msg(sb, KERN_ERR,
2670 "Couldn't mount because of "
2671 "unsupported optional features (%x)",
2672 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2673 ~EXT4_FEATURE_INCOMPAT_SUPP));
2674 return 0;
2675 }
2676
2677 if (readonly)
2678 return 1;
2679
2680 if (ext4_has_feature_readonly(sb)) {
2681 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2682 sb->s_flags |= MS_RDONLY;
2683 return 1;
2684 }
2685
2686 /* Check that feature set is OK for a read-write mount */
2687 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2688 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2689 "unsupported optional features (%x)",
2690 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2691 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2692 return 0;
2693 }
2694 /*
2695 * Large file size enabled file system can only be mounted
2696 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2697 */
2698 if (ext4_has_feature_huge_file(sb)) {
2699 if (sizeof(blkcnt_t) < sizeof(u64)) {
2700 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2701 "cannot be mounted RDWR without "
2702 "CONFIG_LBDAF");
2703 return 0;
2704 }
2705 }
2706 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2707 ext4_msg(sb, KERN_ERR,
2708 "Can't support bigalloc feature without "
2709 "extents feature\n");
2710 return 0;
2711 }
2712
2713 #ifndef CONFIG_QUOTA
2714 if (ext4_has_feature_quota(sb) && !readonly) {
2715 ext4_msg(sb, KERN_ERR,
2716 "Filesystem with quota feature cannot be mounted RDWR "
2717 "without CONFIG_QUOTA");
2718 return 0;
2719 }
2720 if (ext4_has_feature_project(sb) && !readonly) {
2721 ext4_msg(sb, KERN_ERR,
2722 "Filesystem with project quota feature cannot be mounted RDWR "
2723 "without CONFIG_QUOTA");
2724 return 0;
2725 }
2726 #endif /* CONFIG_QUOTA */
2727 return 1;
2728 }
2729
2730 /*
2731 * This function is called once a day if we have errors logged
2732 * on the file system
2733 */
2734 static void print_daily_error_info(unsigned long arg)
2735 {
2736 struct super_block *sb = (struct super_block *) arg;
2737 struct ext4_sb_info *sbi;
2738 struct ext4_super_block *es;
2739
2740 sbi = EXT4_SB(sb);
2741 es = sbi->s_es;
2742
2743 if (es->s_error_count)
2744 /* fsck newer than v1.41.13 is needed to clean this condition. */
2745 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2746 le32_to_cpu(es->s_error_count));
2747 if (es->s_first_error_time) {
2748 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2749 sb->s_id, le32_to_cpu(es->s_first_error_time),
2750 (int) sizeof(es->s_first_error_func),
2751 es->s_first_error_func,
2752 le32_to_cpu(es->s_first_error_line));
2753 if (es->s_first_error_ino)
2754 printk(KERN_CONT ": inode %u",
2755 le32_to_cpu(es->s_first_error_ino));
2756 if (es->s_first_error_block)
2757 printk(KERN_CONT ": block %llu", (unsigned long long)
2758 le64_to_cpu(es->s_first_error_block));
2759 printk(KERN_CONT "\n");
2760 }
2761 if (es->s_last_error_time) {
2762 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2763 sb->s_id, le32_to_cpu(es->s_last_error_time),
2764 (int) sizeof(es->s_last_error_func),
2765 es->s_last_error_func,
2766 le32_to_cpu(es->s_last_error_line));
2767 if (es->s_last_error_ino)
2768 printk(KERN_CONT ": inode %u",
2769 le32_to_cpu(es->s_last_error_ino));
2770 if (es->s_last_error_block)
2771 printk(KERN_CONT ": block %llu", (unsigned long long)
2772 le64_to_cpu(es->s_last_error_block));
2773 printk(KERN_CONT "\n");
2774 }
2775 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2776 }
2777
2778 /* Find next suitable group and run ext4_init_inode_table */
2779 static int ext4_run_li_request(struct ext4_li_request *elr)
2780 {
2781 struct ext4_group_desc *gdp = NULL;
2782 ext4_group_t group, ngroups;
2783 struct super_block *sb;
2784 unsigned long timeout = 0;
2785 int ret = 0;
2786
2787 sb = elr->lr_super;
2788 ngroups = EXT4_SB(sb)->s_groups_count;
2789
2790 for (group = elr->lr_next_group; group < ngroups; group++) {
2791 gdp = ext4_get_group_desc(sb, group, NULL);
2792 if (!gdp) {
2793 ret = 1;
2794 break;
2795 }
2796
2797 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2798 break;
2799 }
2800
2801 if (group >= ngroups)
2802 ret = 1;
2803
2804 if (!ret) {
2805 timeout = jiffies;
2806 ret = ext4_init_inode_table(sb, group,
2807 elr->lr_timeout ? 0 : 1);
2808 if (elr->lr_timeout == 0) {
2809 timeout = (jiffies - timeout) *
2810 elr->lr_sbi->s_li_wait_mult;
2811 elr->lr_timeout = timeout;
2812 }
2813 elr->lr_next_sched = jiffies + elr->lr_timeout;
2814 elr->lr_next_group = group + 1;
2815 }
2816 return ret;
2817 }
2818
2819 /*
2820 * Remove lr_request from the list_request and free the
2821 * request structure. Should be called with li_list_mtx held
2822 */
2823 static void ext4_remove_li_request(struct ext4_li_request *elr)
2824 {
2825 struct ext4_sb_info *sbi;
2826
2827 if (!elr)
2828 return;
2829
2830 sbi = elr->lr_sbi;
2831
2832 list_del(&elr->lr_request);
2833 sbi->s_li_request = NULL;
2834 kfree(elr);
2835 }
2836
2837 static void ext4_unregister_li_request(struct super_block *sb)
2838 {
2839 mutex_lock(&ext4_li_mtx);
2840 if (!ext4_li_info) {
2841 mutex_unlock(&ext4_li_mtx);
2842 return;
2843 }
2844
2845 mutex_lock(&ext4_li_info->li_list_mtx);
2846 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2847 mutex_unlock(&ext4_li_info->li_list_mtx);
2848 mutex_unlock(&ext4_li_mtx);
2849 }
2850
2851 static struct task_struct *ext4_lazyinit_task;
2852
2853 /*
2854 * This is the function where ext4lazyinit thread lives. It walks
2855 * through the request list searching for next scheduled filesystem.
2856 * When such a fs is found, run the lazy initialization request
2857 * (ext4_rn_li_request) and keep track of the time spend in this
2858 * function. Based on that time we compute next schedule time of
2859 * the request. When walking through the list is complete, compute
2860 * next waking time and put itself into sleep.
2861 */
2862 static int ext4_lazyinit_thread(void *arg)
2863 {
2864 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2865 struct list_head *pos, *n;
2866 struct ext4_li_request *elr;
2867 unsigned long next_wakeup, cur;
2868
2869 BUG_ON(NULL == eli);
2870
2871 cont_thread:
2872 while (true) {
2873 next_wakeup = MAX_JIFFY_OFFSET;
2874
2875 mutex_lock(&eli->li_list_mtx);
2876 if (list_empty(&eli->li_request_list)) {
2877 mutex_unlock(&eli->li_list_mtx);
2878 goto exit_thread;
2879 }
2880 list_for_each_safe(pos, n, &eli->li_request_list) {
2881 int err = 0;
2882 int progress = 0;
2883 elr = list_entry(pos, struct ext4_li_request,
2884 lr_request);
2885
2886 if (time_before(jiffies, elr->lr_next_sched)) {
2887 if (time_before(elr->lr_next_sched, next_wakeup))
2888 next_wakeup = elr->lr_next_sched;
2889 continue;
2890 }
2891 if (down_read_trylock(&elr->lr_super->s_umount)) {
2892 if (sb_start_write_trylock(elr->lr_super)) {
2893 progress = 1;
2894 /*
2895 * We hold sb->s_umount, sb can not
2896 * be removed from the list, it is
2897 * now safe to drop li_list_mtx
2898 */
2899 mutex_unlock(&eli->li_list_mtx);
2900 err = ext4_run_li_request(elr);
2901 sb_end_write(elr->lr_super);
2902 mutex_lock(&eli->li_list_mtx);
2903 n = pos->next;
2904 }
2905 up_read((&elr->lr_super->s_umount));
2906 }
2907 /* error, remove the lazy_init job */
2908 if (err) {
2909 ext4_remove_li_request(elr);
2910 continue;
2911 }
2912 if (!progress) {
2913 elr->lr_next_sched = jiffies +
2914 (prandom_u32()
2915 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2916 }
2917 if (time_before(elr->lr_next_sched, next_wakeup))
2918 next_wakeup = elr->lr_next_sched;
2919 }
2920 mutex_unlock(&eli->li_list_mtx);
2921
2922 try_to_freeze();
2923
2924 cur = jiffies;
2925 if ((time_after_eq(cur, next_wakeup)) ||
2926 (MAX_JIFFY_OFFSET == next_wakeup)) {
2927 cond_resched();
2928 continue;
2929 }
2930
2931 schedule_timeout_interruptible(next_wakeup - cur);
2932
2933 if (kthread_should_stop()) {
2934 ext4_clear_request_list();
2935 goto exit_thread;
2936 }
2937 }
2938
2939 exit_thread:
2940 /*
2941 * It looks like the request list is empty, but we need
2942 * to check it under the li_list_mtx lock, to prevent any
2943 * additions into it, and of course we should lock ext4_li_mtx
2944 * to atomically free the list and ext4_li_info, because at
2945 * this point another ext4 filesystem could be registering
2946 * new one.
2947 */
2948 mutex_lock(&ext4_li_mtx);
2949 mutex_lock(&eli->li_list_mtx);
2950 if (!list_empty(&eli->li_request_list)) {
2951 mutex_unlock(&eli->li_list_mtx);
2952 mutex_unlock(&ext4_li_mtx);
2953 goto cont_thread;
2954 }
2955 mutex_unlock(&eli->li_list_mtx);
2956 kfree(ext4_li_info);
2957 ext4_li_info = NULL;
2958 mutex_unlock(&ext4_li_mtx);
2959
2960 return 0;
2961 }
2962
2963 static void ext4_clear_request_list(void)
2964 {
2965 struct list_head *pos, *n;
2966 struct ext4_li_request *elr;
2967
2968 mutex_lock(&ext4_li_info->li_list_mtx);
2969 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2970 elr = list_entry(pos, struct ext4_li_request,
2971 lr_request);
2972 ext4_remove_li_request(elr);
2973 }
2974 mutex_unlock(&ext4_li_info->li_list_mtx);
2975 }
2976
2977 static int ext4_run_lazyinit_thread(void)
2978 {
2979 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2980 ext4_li_info, "ext4lazyinit");
2981 if (IS_ERR(ext4_lazyinit_task)) {
2982 int err = PTR_ERR(ext4_lazyinit_task);
2983 ext4_clear_request_list();
2984 kfree(ext4_li_info);
2985 ext4_li_info = NULL;
2986 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2987 "initialization thread\n",
2988 err);
2989 return err;
2990 }
2991 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2992 return 0;
2993 }
2994
2995 /*
2996 * Check whether it make sense to run itable init. thread or not.
2997 * If there is at least one uninitialized inode table, return
2998 * corresponding group number, else the loop goes through all
2999 * groups and return total number of groups.
3000 */
3001 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3002 {
3003 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3004 struct ext4_group_desc *gdp = NULL;
3005
3006 for (group = 0; group < ngroups; group++) {
3007 gdp = ext4_get_group_desc(sb, group, NULL);
3008 if (!gdp)
3009 continue;
3010
3011 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3012 break;
3013 }
3014
3015 return group;
3016 }
3017
3018 static int ext4_li_info_new(void)
3019 {
3020 struct ext4_lazy_init *eli = NULL;
3021
3022 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3023 if (!eli)
3024 return -ENOMEM;
3025
3026 INIT_LIST_HEAD(&eli->li_request_list);
3027 mutex_init(&eli->li_list_mtx);
3028
3029 eli->li_state |= EXT4_LAZYINIT_QUIT;
3030
3031 ext4_li_info = eli;
3032
3033 return 0;
3034 }
3035
3036 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3037 ext4_group_t start)
3038 {
3039 struct ext4_sb_info *sbi = EXT4_SB(sb);
3040 struct ext4_li_request *elr;
3041
3042 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3043 if (!elr)
3044 return NULL;
3045
3046 elr->lr_super = sb;
3047 elr->lr_sbi = sbi;
3048 elr->lr_next_group = start;
3049
3050 /*
3051 * Randomize first schedule time of the request to
3052 * spread the inode table initialization requests
3053 * better.
3054 */
3055 elr->lr_next_sched = jiffies + (prandom_u32() %
3056 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3057 return elr;
3058 }
3059
3060 int ext4_register_li_request(struct super_block *sb,
3061 ext4_group_t first_not_zeroed)
3062 {
3063 struct ext4_sb_info *sbi = EXT4_SB(sb);
3064 struct ext4_li_request *elr = NULL;
3065 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3066 int ret = 0;
3067
3068 mutex_lock(&ext4_li_mtx);
3069 if (sbi->s_li_request != NULL) {
3070 /*
3071 * Reset timeout so it can be computed again, because
3072 * s_li_wait_mult might have changed.
3073 */
3074 sbi->s_li_request->lr_timeout = 0;
3075 goto out;
3076 }
3077
3078 if (first_not_zeroed == ngroups ||
3079 (sb->s_flags & MS_RDONLY) ||
3080 !test_opt(sb, INIT_INODE_TABLE))
3081 goto out;
3082
3083 elr = ext4_li_request_new(sb, first_not_zeroed);
3084 if (!elr) {
3085 ret = -ENOMEM;
3086 goto out;
3087 }
3088
3089 if (NULL == ext4_li_info) {
3090 ret = ext4_li_info_new();
3091 if (ret)
3092 goto out;
3093 }
3094
3095 mutex_lock(&ext4_li_info->li_list_mtx);
3096 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3097 mutex_unlock(&ext4_li_info->li_list_mtx);
3098
3099 sbi->s_li_request = elr;
3100 /*
3101 * set elr to NULL here since it has been inserted to
3102 * the request_list and the removal and free of it is
3103 * handled by ext4_clear_request_list from now on.
3104 */
3105 elr = NULL;
3106
3107 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3108 ret = ext4_run_lazyinit_thread();
3109 if (ret)
3110 goto out;
3111 }
3112 out:
3113 mutex_unlock(&ext4_li_mtx);
3114 if (ret)
3115 kfree(elr);
3116 return ret;
3117 }
3118
3119 /*
3120 * We do not need to lock anything since this is called on
3121 * module unload.
3122 */
3123 static void ext4_destroy_lazyinit_thread(void)
3124 {
3125 /*
3126 * If thread exited earlier
3127 * there's nothing to be done.
3128 */
3129 if (!ext4_li_info || !ext4_lazyinit_task)
3130 return;
3131
3132 kthread_stop(ext4_lazyinit_task);
3133 }
3134
3135 static int set_journal_csum_feature_set(struct super_block *sb)
3136 {
3137 int ret = 1;
3138 int compat, incompat;
3139 struct ext4_sb_info *sbi = EXT4_SB(sb);
3140
3141 if (ext4_has_metadata_csum(sb)) {
3142 /* journal checksum v3 */
3143 compat = 0;
3144 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3145 } else {
3146 /* journal checksum v1 */
3147 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3148 incompat = 0;
3149 }
3150
3151 jbd2_journal_clear_features(sbi->s_journal,
3152 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3153 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3154 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3155 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3156 ret = jbd2_journal_set_features(sbi->s_journal,
3157 compat, 0,
3158 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3159 incompat);
3160 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3161 ret = jbd2_journal_set_features(sbi->s_journal,
3162 compat, 0,
3163 incompat);
3164 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3165 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3166 } else {
3167 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3168 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3169 }
3170
3171 return ret;
3172 }
3173
3174 /*
3175 * Note: calculating the overhead so we can be compatible with
3176 * historical BSD practice is quite difficult in the face of
3177 * clusters/bigalloc. This is because multiple metadata blocks from
3178 * different block group can end up in the same allocation cluster.
3179 * Calculating the exact overhead in the face of clustered allocation
3180 * requires either O(all block bitmaps) in memory or O(number of block
3181 * groups**2) in time. We will still calculate the superblock for
3182 * older file systems --- and if we come across with a bigalloc file
3183 * system with zero in s_overhead_clusters the estimate will be close to
3184 * correct especially for very large cluster sizes --- but for newer
3185 * file systems, it's better to calculate this figure once at mkfs
3186 * time, and store it in the superblock. If the superblock value is
3187 * present (even for non-bigalloc file systems), we will use it.
3188 */
3189 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3190 char *buf)
3191 {
3192 struct ext4_sb_info *sbi = EXT4_SB(sb);
3193 struct ext4_group_desc *gdp;
3194 ext4_fsblk_t first_block, last_block, b;
3195 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3196 int s, j, count = 0;
3197
3198 if (!ext4_has_feature_bigalloc(sb))
3199 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3200 sbi->s_itb_per_group + 2);
3201
3202 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3203 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3204 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3205 for (i = 0; i < ngroups; i++) {
3206 gdp = ext4_get_group_desc(sb, i, NULL);
3207 b = ext4_block_bitmap(sb, gdp);
3208 if (b >= first_block && b <= last_block) {
3209 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3210 count++;
3211 }
3212 b = ext4_inode_bitmap(sb, gdp);
3213 if (b >= first_block && b <= last_block) {
3214 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3215 count++;
3216 }
3217 b = ext4_inode_table(sb, gdp);
3218 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3219 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3220 int c = EXT4_B2C(sbi, b - first_block);
3221 ext4_set_bit(c, buf);
3222 count++;
3223 }
3224 if (i != grp)
3225 continue;
3226 s = 0;
3227 if (ext4_bg_has_super(sb, grp)) {
3228 ext4_set_bit(s++, buf);
3229 count++;
3230 }
3231 j = ext4_bg_num_gdb(sb, grp);
3232 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3233 ext4_error(sb, "Invalid number of block group "
3234 "descriptor blocks: %d", j);
3235 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3236 }
3237 count += j;
3238 for (; j > 0; j--)
3239 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3240 }
3241 if (!count)
3242 return 0;
3243 return EXT4_CLUSTERS_PER_GROUP(sb) -
3244 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3245 }
3246
3247 /*
3248 * Compute the overhead and stash it in sbi->s_overhead
3249 */
3250 int ext4_calculate_overhead(struct super_block *sb)
3251 {
3252 struct ext4_sb_info *sbi = EXT4_SB(sb);
3253 struct ext4_super_block *es = sbi->s_es;
3254 struct inode *j_inode;
3255 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3256 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3257 ext4_fsblk_t overhead = 0;
3258 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3259
3260 if (!buf)
3261 return -ENOMEM;
3262
3263 /*
3264 * Compute the overhead (FS structures). This is constant
3265 * for a given filesystem unless the number of block groups
3266 * changes so we cache the previous value until it does.
3267 */
3268
3269 /*
3270 * All of the blocks before first_data_block are overhead
3271 */
3272 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3273
3274 /*
3275 * Add the overhead found in each block group
3276 */
3277 for (i = 0; i < ngroups; i++) {
3278 int blks;
3279
3280 blks = count_overhead(sb, i, buf);
3281 overhead += blks;
3282 if (blks)
3283 memset(buf, 0, PAGE_SIZE);
3284 cond_resched();
3285 }
3286
3287 /*
3288 * Add the internal journal blocks whether the journal has been
3289 * loaded or not
3290 */
3291 if (sbi->s_journal && !sbi->journal_bdev)
3292 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3293 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3294 j_inode = ext4_get_journal_inode(sb, j_inum);
3295 if (j_inode) {
3296 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3297 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3298 iput(j_inode);
3299 } else {
3300 ext4_msg(sb, KERN_ERR, "can't get journal size");
3301 }
3302 }
3303 sbi->s_overhead = overhead;
3304 smp_wmb();
3305 free_page((unsigned long) buf);
3306 return 0;
3307 }
3308
3309 static void ext4_set_resv_clusters(struct super_block *sb)
3310 {
3311 ext4_fsblk_t resv_clusters;
3312 struct ext4_sb_info *sbi = EXT4_SB(sb);
3313
3314 /*
3315 * There's no need to reserve anything when we aren't using extents.
3316 * The space estimates are exact, there are no unwritten extents,
3317 * hole punching doesn't need new metadata... This is needed especially
3318 * to keep ext2/3 backward compatibility.
3319 */
3320 if (!ext4_has_feature_extents(sb))
3321 return;
3322 /*
3323 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3324 * This should cover the situations where we can not afford to run
3325 * out of space like for example punch hole, or converting
3326 * unwritten extents in delalloc path. In most cases such
3327 * allocation would require 1, or 2 blocks, higher numbers are
3328 * very rare.
3329 */
3330 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3331 sbi->s_cluster_bits);
3332
3333 do_div(resv_clusters, 50);
3334 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3335
3336 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3337 }
3338
3339 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3340 {
3341 char *orig_data = kstrdup(data, GFP_KERNEL);
3342 struct buffer_head *bh;
3343 struct ext4_super_block *es = NULL;
3344 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3345 ext4_fsblk_t block;
3346 ext4_fsblk_t sb_block = get_sb_block(&data);
3347 ext4_fsblk_t logical_sb_block;
3348 unsigned long offset = 0;
3349 unsigned long journal_devnum = 0;
3350 unsigned long def_mount_opts;
3351 struct inode *root;
3352 const char *descr;
3353 int ret = -ENOMEM;
3354 int blocksize, clustersize;
3355 unsigned int db_count;
3356 unsigned int i;
3357 int needs_recovery, has_huge_files, has_bigalloc;
3358 __u64 blocks_count;
3359 int err = 0;
3360 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3361 ext4_group_t first_not_zeroed;
3362
3363 if ((data && !orig_data) || !sbi)
3364 goto out_free_base;
3365
3366 sbi->s_blockgroup_lock =
3367 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3368 if (!sbi->s_blockgroup_lock)
3369 goto out_free_base;
3370
3371 sb->s_fs_info = sbi;
3372 sbi->s_sb = sb;
3373 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3374 sbi->s_sb_block = sb_block;
3375 if (sb->s_bdev->bd_part)
3376 sbi->s_sectors_written_start =
3377 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3378
3379 /* Cleanup superblock name */
3380 strreplace(sb->s_id, '/', '!');
3381
3382 /* -EINVAL is default */
3383 ret = -EINVAL;
3384 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3385 if (!blocksize) {
3386 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3387 goto out_fail;
3388 }
3389
3390 /*
3391 * The ext4 superblock will not be buffer aligned for other than 1kB
3392 * block sizes. We need to calculate the offset from buffer start.
3393 */
3394 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3395 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3396 offset = do_div(logical_sb_block, blocksize);
3397 } else {
3398 logical_sb_block = sb_block;
3399 }
3400
3401 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3402 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3403 goto out_fail;
3404 }
3405 /*
3406 * Note: s_es must be initialized as soon as possible because
3407 * some ext4 macro-instructions depend on its value
3408 */
3409 es = (struct ext4_super_block *) (bh->b_data + offset);
3410 sbi->s_es = es;
3411 sb->s_magic = le16_to_cpu(es->s_magic);
3412 if (sb->s_magic != EXT4_SUPER_MAGIC)
3413 goto cantfind_ext4;
3414 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3415
3416 /* Warn if metadata_csum and gdt_csum are both set. */
3417 if (ext4_has_feature_metadata_csum(sb) &&
3418 ext4_has_feature_gdt_csum(sb))
3419 ext4_warning(sb, "metadata_csum and uninit_bg are "
3420 "redundant flags; please run fsck.");
3421
3422 /* Check for a known checksum algorithm */
3423 if (!ext4_verify_csum_type(sb, es)) {
3424 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3425 "unknown checksum algorithm.");
3426 silent = 1;
3427 goto cantfind_ext4;
3428 }
3429
3430 /* Load the checksum driver */
3431 if (ext4_has_feature_metadata_csum(sb)) {
3432 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3433 if (IS_ERR(sbi->s_chksum_driver)) {
3434 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3435 ret = PTR_ERR(sbi->s_chksum_driver);
3436 sbi->s_chksum_driver = NULL;
3437 goto failed_mount;
3438 }
3439 }
3440
3441 /* Check superblock checksum */
3442 if (!ext4_superblock_csum_verify(sb, es)) {
3443 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3444 "invalid superblock checksum. Run e2fsck?");
3445 silent = 1;
3446 ret = -EFSBADCRC;
3447 goto cantfind_ext4;
3448 }
3449
3450 /* Precompute checksum seed for all metadata */
3451 if (ext4_has_feature_csum_seed(sb))
3452 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3453 else if (ext4_has_metadata_csum(sb))
3454 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3455 sizeof(es->s_uuid));
3456
3457 /* Set defaults before we parse the mount options */
3458 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3459 set_opt(sb, INIT_INODE_TABLE);
3460 if (def_mount_opts & EXT4_DEFM_DEBUG)
3461 set_opt(sb, DEBUG);
3462 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3463 set_opt(sb, GRPID);
3464 if (def_mount_opts & EXT4_DEFM_UID16)
3465 set_opt(sb, NO_UID32);
3466 /* xattr user namespace & acls are now defaulted on */
3467 set_opt(sb, XATTR_USER);
3468 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3469 set_opt(sb, POSIX_ACL);
3470 #endif
3471 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3472 if (ext4_has_metadata_csum(sb))
3473 set_opt(sb, JOURNAL_CHECKSUM);
3474
3475 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3476 set_opt(sb, JOURNAL_DATA);
3477 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3478 set_opt(sb, ORDERED_DATA);
3479 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3480 set_opt(sb, WRITEBACK_DATA);
3481
3482 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC) {
3483 if (!capable(CAP_SYS_ADMIN))
3484 goto failed_mount;
3485 set_opt(sb, ERRORS_PANIC);
3486 } else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE) {
3487 set_opt(sb, ERRORS_CONT);
3488 } else {
3489 set_opt(sb, ERRORS_RO);
3490 }
3491 /* block_validity enabled by default; disable with noblock_validity */
3492 set_opt(sb, BLOCK_VALIDITY);
3493 if (def_mount_opts & EXT4_DEFM_DISCARD)
3494 set_opt(sb, DISCARD);
3495
3496 sbi->s_resuid = make_kuid(sb->s_user_ns, le16_to_cpu(es->s_def_resuid));
3497 if (!uid_valid(sbi->s_resuid))
3498 sbi->s_resuid = make_kuid(sb->s_user_ns, EXT4_DEF_RESUID);
3499 sbi->s_resgid = make_kgid(sb->s_user_ns, le16_to_cpu(es->s_def_resgid));
3500 if (!gid_valid(sbi->s_resgid))
3501 sbi->s_resgid = make_kgid(sb->s_user_ns, EXT4_DEF_RESGID);
3502 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3503 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3504 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3505
3506 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3507 set_opt(sb, BARRIER);
3508
3509 /*
3510 * enable delayed allocation by default
3511 * Use -o nodelalloc to turn it off
3512 */
3513 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3514 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3515 set_opt(sb, DELALLOC);
3516
3517 /*
3518 * set default s_li_wait_mult for lazyinit, for the case there is
3519 * no mount option specified.
3520 */
3521 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3522
3523 if (sbi->s_es->s_mount_opts[0]) {
3524 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3525 sizeof(sbi->s_es->s_mount_opts),
3526 GFP_KERNEL);
3527 if (!s_mount_opts)
3528 goto failed_mount;
3529 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3530 &journal_ioprio, 0)) {
3531 ext4_msg(sb, KERN_WARNING,
3532 "failed to parse options in superblock: %s",
3533 s_mount_opts);
3534 }
3535 kfree(s_mount_opts);
3536 }
3537 sbi->s_def_mount_opt = sbi->s_mount_opt;
3538 if (!parse_options((char *) data, sb, &journal_devnum,
3539 &journal_ioprio, 0))
3540 goto failed_mount;
3541
3542 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3543 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3544 "with data=journal disables delayed "
3545 "allocation and O_DIRECT support!\n");
3546 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3547 ext4_msg(sb, KERN_ERR, "can't mount with "
3548 "both data=journal and delalloc");
3549 goto failed_mount;
3550 }
3551 if (test_opt(sb, DIOREAD_NOLOCK)) {
3552 ext4_msg(sb, KERN_ERR, "can't mount with "
3553 "both data=journal and dioread_nolock");
3554 goto failed_mount;
3555 }
3556 if (test_opt(sb, DAX)) {
3557 ext4_msg(sb, KERN_ERR, "can't mount with "
3558 "both data=journal and dax");
3559 goto failed_mount;
3560 }
3561 if (ext4_has_feature_encrypt(sb)) {
3562 ext4_msg(sb, KERN_WARNING,
3563 "encrypted files will use data=ordered "
3564 "instead of data journaling mode");
3565 }
3566 if (test_opt(sb, DELALLOC))
3567 clear_opt(sb, DELALLOC);
3568 } else {
3569 sb->s_iflags |= SB_I_CGROUPWB;
3570 }
3571
3572 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3573 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3574
3575 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3576 (ext4_has_compat_features(sb) ||
3577 ext4_has_ro_compat_features(sb) ||
3578 ext4_has_incompat_features(sb)))
3579 ext4_msg(sb, KERN_WARNING,
3580 "feature flags set on rev 0 fs, "
3581 "running e2fsck is recommended");
3582
3583 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3584 set_opt2(sb, HURD_COMPAT);
3585 if (ext4_has_feature_64bit(sb)) {
3586 ext4_msg(sb, KERN_ERR,
3587 "The Hurd can't support 64-bit file systems");
3588 goto failed_mount;
3589 }
3590 }
3591
3592 if (IS_EXT2_SB(sb)) {
3593 if (ext2_feature_set_ok(sb))
3594 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3595 "using the ext4 subsystem");
3596 else {
3597 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3598 "to feature incompatibilities");
3599 goto failed_mount;
3600 }
3601 }
3602
3603 if (IS_EXT3_SB(sb)) {
3604 if (ext3_feature_set_ok(sb))
3605 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3606 "using the ext4 subsystem");
3607 else {
3608 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3609 "to feature incompatibilities");
3610 goto failed_mount;
3611 }
3612 }
3613
3614 /*
3615 * Check feature flags regardless of the revision level, since we
3616 * previously didn't change the revision level when setting the flags,
3617 * so there is a chance incompat flags are set on a rev 0 filesystem.
3618 */
3619 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3620 goto failed_mount;
3621
3622 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3623 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3624 blocksize > EXT4_MAX_BLOCK_SIZE) {
3625 ext4_msg(sb, KERN_ERR,
3626 "Unsupported filesystem blocksize %d (%d log_block_size)",
3627 blocksize, le32_to_cpu(es->s_log_block_size));
3628 goto failed_mount;
3629 }
3630 if (le32_to_cpu(es->s_log_block_size) >
3631 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3632 ext4_msg(sb, KERN_ERR,
3633 "Invalid log block size: %u",
3634 le32_to_cpu(es->s_log_block_size));
3635 goto failed_mount;
3636 }
3637
3638 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3639 ext4_msg(sb, KERN_ERR,
3640 "Number of reserved GDT blocks insanely large: %d",
3641 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3642 goto failed_mount;
3643 }
3644
3645 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3646 err = bdev_dax_supported(sb, blocksize);
3647 if (err)
3648 goto failed_mount;
3649 }
3650
3651 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3652 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3653 es->s_encryption_level);
3654 goto failed_mount;
3655 }
3656
3657 if (sb->s_blocksize != blocksize) {
3658 /* Validate the filesystem blocksize */
3659 if (!sb_set_blocksize(sb, blocksize)) {
3660 ext4_msg(sb, KERN_ERR, "bad block size %d",
3661 blocksize);
3662 goto failed_mount;
3663 }
3664
3665 brelse(bh);
3666 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3667 offset = do_div(logical_sb_block, blocksize);
3668 bh = sb_bread_unmovable(sb, logical_sb_block);
3669 if (!bh) {
3670 ext4_msg(sb, KERN_ERR,
3671 "Can't read superblock on 2nd try");
3672 goto failed_mount;
3673 }
3674 es = (struct ext4_super_block *)(bh->b_data + offset);
3675 sbi->s_es = es;
3676 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3677 ext4_msg(sb, KERN_ERR,
3678 "Magic mismatch, very weird!");
3679 goto failed_mount;
3680 }
3681 }
3682
3683 has_huge_files = ext4_has_feature_huge_file(sb);
3684 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3685 has_huge_files);
3686 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3687
3688 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3689 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3690 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3691 } else {
3692 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3693 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3694 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3695 (!is_power_of_2(sbi->s_inode_size)) ||
3696 (sbi->s_inode_size > blocksize)) {
3697 ext4_msg(sb, KERN_ERR,
3698 "unsupported inode size: %d",
3699 sbi->s_inode_size);
3700 goto failed_mount;
3701 }
3702 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3703 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3704 }
3705
3706 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3707 if (ext4_has_feature_64bit(sb)) {
3708 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3709 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3710 !is_power_of_2(sbi->s_desc_size)) {
3711 ext4_msg(sb, KERN_ERR,
3712 "unsupported descriptor size %lu",
3713 sbi->s_desc_size);
3714 goto failed_mount;
3715 }
3716 } else
3717 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3718
3719 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3720 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3721
3722 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3723 if (sbi->s_inodes_per_block == 0)
3724 goto cantfind_ext4;
3725 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3726 sbi->s_inodes_per_group > blocksize * 8) {
3727 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3728 sbi->s_blocks_per_group);
3729 goto failed_mount;
3730 }
3731 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3732 sbi->s_inodes_per_block;
3733 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3734 sbi->s_sbh = bh;
3735 sbi->s_mount_state = le16_to_cpu(es->s_state);
3736 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3737 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3738
3739 for (i = 0; i < 4; i++)
3740 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3741 sbi->s_def_hash_version = es->s_def_hash_version;
3742 if (ext4_has_feature_dir_index(sb)) {
3743 i = le32_to_cpu(es->s_flags);
3744 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3745 sbi->s_hash_unsigned = 3;
3746 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3747 #ifdef __CHAR_UNSIGNED__
3748 if (!(sb->s_flags & MS_RDONLY))
3749 es->s_flags |=
3750 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3751 sbi->s_hash_unsigned = 3;
3752 #else
3753 if (!(sb->s_flags & MS_RDONLY))
3754 es->s_flags |=
3755 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3756 #endif
3757 }
3758 }
3759
3760 /* Handle clustersize */
3761 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3762 has_bigalloc = ext4_has_feature_bigalloc(sb);
3763 if (has_bigalloc) {
3764 if (clustersize < blocksize) {
3765 ext4_msg(sb, KERN_ERR,
3766 "cluster size (%d) smaller than "
3767 "block size (%d)", clustersize, blocksize);
3768 goto failed_mount;
3769 }
3770 if (le32_to_cpu(es->s_log_cluster_size) >
3771 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3772 ext4_msg(sb, KERN_ERR,
3773 "Invalid log cluster size: %u",
3774 le32_to_cpu(es->s_log_cluster_size));
3775 goto failed_mount;
3776 }
3777 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3778 le32_to_cpu(es->s_log_block_size);
3779 sbi->s_clusters_per_group =
3780 le32_to_cpu(es->s_clusters_per_group);
3781 if (sbi->s_clusters_per_group > blocksize * 8) {
3782 ext4_msg(sb, KERN_ERR,
3783 "#clusters per group too big: %lu",
3784 sbi->s_clusters_per_group);
3785 goto failed_mount;
3786 }
3787 if (sbi->s_blocks_per_group !=
3788 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3789 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3790 "clusters per group (%lu) inconsistent",
3791 sbi->s_blocks_per_group,
3792 sbi->s_clusters_per_group);
3793 goto failed_mount;
3794 }
3795 } else {
3796 if (clustersize != blocksize) {
3797 ext4_warning(sb, "fragment/cluster size (%d) != "
3798 "block size (%d)", clustersize,
3799 blocksize);
3800 clustersize = blocksize;
3801 }
3802 if (sbi->s_blocks_per_group > blocksize * 8) {
3803 ext4_msg(sb, KERN_ERR,
3804 "#blocks per group too big: %lu",
3805 sbi->s_blocks_per_group);
3806 goto failed_mount;
3807 }
3808 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3809 sbi->s_cluster_bits = 0;
3810 }
3811 sbi->s_cluster_ratio = clustersize / blocksize;
3812
3813 /* Do we have standard group size of clustersize * 8 blocks ? */
3814 if (sbi->s_blocks_per_group == clustersize << 3)
3815 set_opt2(sb, STD_GROUP_SIZE);
3816
3817 /*
3818 * Test whether we have more sectors than will fit in sector_t,
3819 * and whether the max offset is addressable by the page cache.
3820 */
3821 err = generic_check_addressable(sb->s_blocksize_bits,
3822 ext4_blocks_count(es));
3823 if (err) {
3824 ext4_msg(sb, KERN_ERR, "filesystem"
3825 " too large to mount safely on this system");
3826 if (sizeof(sector_t) < 8)
3827 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3828 goto failed_mount;
3829 }
3830
3831 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3832 goto cantfind_ext4;
3833
3834 /* check blocks count against device size */
3835 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3836 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3837 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3838 "exceeds size of device (%llu blocks)",
3839 ext4_blocks_count(es), blocks_count);
3840 goto failed_mount;
3841 }
3842
3843 /*
3844 * It makes no sense for the first data block to be beyond the end
3845 * of the filesystem.
3846 */
3847 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3848 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3849 "block %u is beyond end of filesystem (%llu)",
3850 le32_to_cpu(es->s_first_data_block),
3851 ext4_blocks_count(es));
3852 goto failed_mount;
3853 }
3854 blocks_count = (ext4_blocks_count(es) -
3855 le32_to_cpu(es->s_first_data_block) +
3856 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3857 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3858 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3859 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3860 "(block count %llu, first data block %u, "
3861 "blocks per group %lu)", sbi->s_groups_count,
3862 ext4_blocks_count(es),
3863 le32_to_cpu(es->s_first_data_block),
3864 EXT4_BLOCKS_PER_GROUP(sb));
3865 goto failed_mount;
3866 }
3867 sbi->s_groups_count = blocks_count;
3868 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3869 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3870 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3871 EXT4_DESC_PER_BLOCK(sb);
3872 if (ext4_has_feature_meta_bg(sb)) {
3873 if (le32_to_cpu(es->s_first_meta_bg) >= db_count) {
3874 ext4_msg(sb, KERN_WARNING,
3875 "first meta block group too large: %u "
3876 "(group descriptor block count %u)",
3877 le32_to_cpu(es->s_first_meta_bg), db_count);
3878 goto failed_mount;
3879 }
3880 }
3881 sbi->s_group_desc = ext4_kvmalloc(db_count *
3882 sizeof(struct buffer_head *),
3883 GFP_KERNEL);
3884 if (sbi->s_group_desc == NULL) {
3885 ext4_msg(sb, KERN_ERR, "not enough memory");
3886 ret = -ENOMEM;
3887 goto failed_mount;
3888 }
3889
3890 bgl_lock_init(sbi->s_blockgroup_lock);
3891
3892 for (i = 0; i < db_count; i++) {
3893 block = descriptor_loc(sb, logical_sb_block, i);
3894 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3895 if (!sbi->s_group_desc[i]) {
3896 ext4_msg(sb, KERN_ERR,
3897 "can't read group descriptor %d", i);
3898 db_count = i;
3899 goto failed_mount2;
3900 }
3901 }
3902 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3903 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3904 ret = -EFSCORRUPTED;
3905 goto failed_mount2;
3906 }
3907
3908 sbi->s_gdb_count = db_count;
3909 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3910 spin_lock_init(&sbi->s_next_gen_lock);
3911
3912 setup_timer(&sbi->s_err_report, print_daily_error_info,
3913 (unsigned long) sb);
3914
3915 /* Register extent status tree shrinker */
3916 if (ext4_es_register_shrinker(sbi))
3917 goto failed_mount3;
3918
3919 sbi->s_stripe = ext4_get_stripe_size(sbi);
3920 sbi->s_extent_max_zeroout_kb = 32;
3921
3922 /*
3923 * set up enough so that it can read an inode
3924 */
3925 sb->s_op = &ext4_sops;
3926 sb->s_export_op = &ext4_export_ops;
3927 sb->s_xattr = ext4_xattr_handlers;
3928 sb->s_cop = &ext4_cryptops;
3929 #ifdef CONFIG_QUOTA
3930 sb->dq_op = &ext4_quota_operations;
3931 if (ext4_has_feature_quota(sb))
3932 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3933 else
3934 sb->s_qcop = &ext4_qctl_operations;
3935 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3936 #endif
3937 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3938
3939 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3940 mutex_init(&sbi->s_orphan_lock);
3941
3942 sb->s_root = NULL;
3943
3944 needs_recovery = (es->s_last_orphan != 0 ||
3945 ext4_has_feature_journal_needs_recovery(sb));
3946
3947 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3948 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3949 goto failed_mount3a;
3950
3951 /*
3952 * The first inode we look at is the journal inode. Don't try
3953 * root first: it may be modified in the journal!
3954 */
3955 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3956 if (ext4_load_journal(sb, es, journal_devnum))
3957 goto failed_mount3a;
3958 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3959 ext4_has_feature_journal_needs_recovery(sb)) {
3960 ext4_msg(sb, KERN_ERR, "required journal recovery "
3961 "suppressed and not mounted read-only");
3962 goto failed_mount_wq;
3963 } else {
3964 /* Nojournal mode, all journal mount options are illegal */
3965 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3966 ext4_msg(sb, KERN_ERR, "can't mount with "
3967 "journal_checksum, fs mounted w/o journal");
3968 goto failed_mount_wq;
3969 }
3970 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3971 ext4_msg(sb, KERN_ERR, "can't mount with "
3972 "journal_async_commit, fs mounted w/o journal");
3973 goto failed_mount_wq;
3974 }
3975 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3976 ext4_msg(sb, KERN_ERR, "can't mount with "
3977 "commit=%lu, fs mounted w/o journal",
3978 sbi->s_commit_interval / HZ);
3979 goto failed_mount_wq;
3980 }
3981 if (EXT4_MOUNT_DATA_FLAGS &
3982 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3983 ext4_msg(sb, KERN_ERR, "can't mount with "
3984 "data=, fs mounted w/o journal");
3985 goto failed_mount_wq;
3986 }
3987 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3988 clear_opt(sb, JOURNAL_CHECKSUM);
3989 clear_opt(sb, DATA_FLAGS);
3990 sbi->s_journal = NULL;
3991 needs_recovery = 0;
3992 goto no_journal;
3993 }
3994
3995 if (ext4_has_feature_64bit(sb) &&
3996 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3997 JBD2_FEATURE_INCOMPAT_64BIT)) {
3998 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3999 goto failed_mount_wq;
4000 }
4001
4002 if (!set_journal_csum_feature_set(sb)) {
4003 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4004 "feature set");
4005 goto failed_mount_wq;
4006 }
4007
4008 /* We have now updated the journal if required, so we can
4009 * validate the data journaling mode. */
4010 switch (test_opt(sb, DATA_FLAGS)) {
4011 case 0:
4012 /* No mode set, assume a default based on the journal
4013 * capabilities: ORDERED_DATA if the journal can
4014 * cope, else JOURNAL_DATA
4015 */
4016 if (jbd2_journal_check_available_features
4017 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4018 set_opt(sb, ORDERED_DATA);
4019 else
4020 set_opt(sb, JOURNAL_DATA);
4021 break;
4022
4023 case EXT4_MOUNT_ORDERED_DATA:
4024 case EXT4_MOUNT_WRITEBACK_DATA:
4025 if (!jbd2_journal_check_available_features
4026 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4027 ext4_msg(sb, KERN_ERR, "Journal does not support "
4028 "requested data journaling mode");
4029 goto failed_mount_wq;
4030 }
4031 default:
4032 break;
4033 }
4034
4035 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4036 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4037 ext4_msg(sb, KERN_ERR, "can't mount with "
4038 "journal_async_commit in data=ordered mode");
4039 goto failed_mount_wq;
4040 }
4041
4042 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4043
4044 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4045
4046 no_journal:
4047 sbi->s_mb_cache = ext4_xattr_create_cache();
4048 if (!sbi->s_mb_cache) {
4049 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4050 goto failed_mount_wq;
4051 }
4052
4053 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4054 (blocksize != PAGE_SIZE)) {
4055 ext4_msg(sb, KERN_ERR,
4056 "Unsupported blocksize for fs encryption");
4057 goto failed_mount_wq;
4058 }
4059
4060 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4061 !ext4_has_feature_encrypt(sb)) {
4062 ext4_set_feature_encrypt(sb);
4063 ext4_commit_super(sb, 1);
4064 }
4065
4066 /*
4067 * Get the # of file system overhead blocks from the
4068 * superblock if present.
4069 */
4070 if (es->s_overhead_clusters)
4071 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4072 else {
4073 err = ext4_calculate_overhead(sb);
4074 if (err)
4075 goto failed_mount_wq;
4076 }
4077
4078 /*
4079 * The maximum number of concurrent works can be high and
4080 * concurrency isn't really necessary. Limit it to 1.
4081 */
4082 EXT4_SB(sb)->rsv_conversion_wq =
4083 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4084 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4085 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4086 ret = -ENOMEM;
4087 goto failed_mount4;
4088 }
4089
4090 /*
4091 * The jbd2_journal_load will have done any necessary log recovery,
4092 * so we can safely mount the rest of the filesystem now.
4093 */
4094
4095 root = ext4_iget(sb, EXT4_ROOT_INO);
4096 if (IS_ERR(root)) {
4097 ext4_msg(sb, KERN_ERR, "get root inode failed");
4098 ret = PTR_ERR(root);
4099 root = NULL;
4100 goto failed_mount4;
4101 }
4102 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4103 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4104 iput(root);
4105 goto failed_mount4;
4106 }
4107 sb->s_root = d_make_root(root);
4108 if (!sb->s_root) {
4109 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4110 ret = -ENOMEM;
4111 goto failed_mount4;
4112 }
4113
4114 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4115 sb->s_flags |= MS_RDONLY;
4116
4117 /* determine the minimum size of new large inodes, if present */
4118 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4119 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4120 EXT4_GOOD_OLD_INODE_SIZE;
4121 if (ext4_has_feature_extra_isize(sb)) {
4122 if (sbi->s_want_extra_isize <
4123 le16_to_cpu(es->s_want_extra_isize))
4124 sbi->s_want_extra_isize =
4125 le16_to_cpu(es->s_want_extra_isize);
4126 if (sbi->s_want_extra_isize <
4127 le16_to_cpu(es->s_min_extra_isize))
4128 sbi->s_want_extra_isize =
4129 le16_to_cpu(es->s_min_extra_isize);
4130 }
4131 }
4132 /* Check if enough inode space is available */
4133 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4134 sbi->s_inode_size) {
4135 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4136 EXT4_GOOD_OLD_INODE_SIZE;
4137 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4138 "available");
4139 }
4140
4141 ext4_set_resv_clusters(sb);
4142
4143 err = ext4_setup_system_zone(sb);
4144 if (err) {
4145 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4146 "zone (%d)", err);
4147 goto failed_mount4a;
4148 }
4149
4150 ext4_ext_init(sb);
4151 err = ext4_mb_init(sb);
4152 if (err) {
4153 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4154 err);
4155 goto failed_mount5;
4156 }
4157
4158 block = ext4_count_free_clusters(sb);
4159 ext4_free_blocks_count_set(sbi->s_es,
4160 EXT4_C2B(sbi, block));
4161 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4162 GFP_KERNEL);
4163 if (!err) {
4164 unsigned long freei = ext4_count_free_inodes(sb);
4165 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4166 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4167 GFP_KERNEL);
4168 }
4169 if (!err)
4170 err = percpu_counter_init(&sbi->s_dirs_counter,
4171 ext4_count_dirs(sb), GFP_KERNEL);
4172 if (!err)
4173 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4174 GFP_KERNEL);
4175 if (!err)
4176 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4177
4178 if (err) {
4179 ext4_msg(sb, KERN_ERR, "insufficient memory");
4180 goto failed_mount6;
4181 }
4182
4183 if (ext4_has_feature_flex_bg(sb))
4184 if (!ext4_fill_flex_info(sb)) {
4185 ext4_msg(sb, KERN_ERR,
4186 "unable to initialize "
4187 "flex_bg meta info!");
4188 goto failed_mount6;
4189 }
4190
4191 err = ext4_register_li_request(sb, first_not_zeroed);
4192 if (err)
4193 goto failed_mount6;
4194
4195 err = ext4_register_sysfs(sb);
4196 if (err)
4197 goto failed_mount7;
4198
4199 #ifdef CONFIG_QUOTA
4200 /* Enable quota usage during mount. */
4201 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4202 err = ext4_enable_quotas(sb);
4203 if (err)
4204 goto failed_mount8;
4205 }
4206 #endif /* CONFIG_QUOTA */
4207
4208 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4209 ext4_orphan_cleanup(sb, es);
4210 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4211 if (needs_recovery) {
4212 ext4_msg(sb, KERN_INFO, "recovery complete");
4213 ext4_mark_recovery_complete(sb, es);
4214 }
4215 if (EXT4_SB(sb)->s_journal) {
4216 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4217 descr = " journalled data mode";
4218 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4219 descr = " ordered data mode";
4220 else
4221 descr = " writeback data mode";
4222 } else
4223 descr = "out journal";
4224
4225 if (test_opt(sb, DISCARD)) {
4226 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4227 if (!blk_queue_discard(q))
4228 ext4_msg(sb, KERN_WARNING,
4229 "mounting with \"discard\" option, but "
4230 "the device does not support discard");
4231 }
4232
4233 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4234 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4235 "Opts: %.*s%s%s", descr,
4236 (int) sizeof(sbi->s_es->s_mount_opts),
4237 sbi->s_es->s_mount_opts,
4238 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4239
4240 if (es->s_error_count)
4241 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4242
4243 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4244 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4245 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4246 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4247
4248 kfree(orig_data);
4249 #ifdef CONFIG_EXT4_FS_ENCRYPTION
4250 memcpy(sbi->key_prefix, EXT4_KEY_DESC_PREFIX,
4251 EXT4_KEY_DESC_PREFIX_SIZE);
4252 sbi->key_prefix_size = EXT4_KEY_DESC_PREFIX_SIZE;
4253 #endif
4254 return 0;
4255
4256 cantfind_ext4:
4257 if (!silent)
4258 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4259 goto failed_mount;
4260
4261 #ifdef CONFIG_QUOTA
4262 failed_mount8:
4263 ext4_unregister_sysfs(sb);
4264 #endif
4265 failed_mount7:
4266 ext4_unregister_li_request(sb);
4267 failed_mount6:
4268 ext4_mb_release(sb);
4269 if (sbi->s_flex_groups)
4270 kvfree(sbi->s_flex_groups);
4271 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4272 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4273 percpu_counter_destroy(&sbi->s_dirs_counter);
4274 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4275 failed_mount5:
4276 ext4_ext_release(sb);
4277 ext4_release_system_zone(sb);
4278 failed_mount4a:
4279 dput(sb->s_root);
4280 sb->s_root = NULL;
4281 failed_mount4:
4282 ext4_msg(sb, KERN_ERR, "mount failed");
4283 if (EXT4_SB(sb)->rsv_conversion_wq)
4284 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4285 failed_mount_wq:
4286 if (sbi->s_mb_cache) {
4287 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4288 sbi->s_mb_cache = NULL;
4289 }
4290 if (sbi->s_journal) {
4291 jbd2_journal_destroy(sbi->s_journal);
4292 sbi->s_journal = NULL;
4293 }
4294 failed_mount3a:
4295 ext4_es_unregister_shrinker(sbi);
4296 failed_mount3:
4297 del_timer_sync(&sbi->s_err_report);
4298 if (sbi->s_mmp_tsk)
4299 kthread_stop(sbi->s_mmp_tsk);
4300 failed_mount2:
4301 for (i = 0; i < db_count; i++)
4302 brelse(sbi->s_group_desc[i]);
4303 kvfree(sbi->s_group_desc);
4304 failed_mount:
4305 if (sbi->s_chksum_driver)
4306 crypto_free_shash(sbi->s_chksum_driver);
4307 #ifdef CONFIG_QUOTA
4308 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4309 kfree(sbi->s_qf_names[i]);
4310 #endif
4311 ext4_blkdev_remove(sbi);
4312 brelse(bh);
4313 out_fail:
4314 /* sb->s_user_ns will be put when sb is destroyed */
4315 sb->s_fs_info = NULL;
4316 kfree(sbi->s_blockgroup_lock);
4317 out_free_base:
4318 kfree(sbi);
4319 kfree(orig_data);
4320 return err ? err : ret;
4321 }
4322
4323 /*
4324 * Setup any per-fs journal parameters now. We'll do this both on
4325 * initial mount, once the journal has been initialised but before we've
4326 * done any recovery; and again on any subsequent remount.
4327 */
4328 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4329 {
4330 struct ext4_sb_info *sbi = EXT4_SB(sb);
4331
4332 journal->j_commit_interval = sbi->s_commit_interval;
4333 journal->j_min_batch_time = sbi->s_min_batch_time;
4334 journal->j_max_batch_time = sbi->s_max_batch_time;
4335
4336 write_lock(&journal->j_state_lock);
4337 if (test_opt(sb, BARRIER))
4338 journal->j_flags |= JBD2_BARRIER;
4339 else
4340 journal->j_flags &= ~JBD2_BARRIER;
4341 if (test_opt(sb, DATA_ERR_ABORT))
4342 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4343 else
4344 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4345 write_unlock(&journal->j_state_lock);
4346 }
4347
4348 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4349 unsigned int journal_inum)
4350 {
4351 struct inode *journal_inode;
4352
4353 /*
4354 * Test for the existence of a valid inode on disk. Bad things
4355 * happen if we iget() an unused inode, as the subsequent iput()
4356 * will try to delete it.
4357 */
4358 journal_inode = ext4_iget(sb, journal_inum);
4359 if (IS_ERR(journal_inode)) {
4360 ext4_msg(sb, KERN_ERR, "no journal found");
4361 return NULL;
4362 }
4363 if (!journal_inode->i_nlink) {
4364 make_bad_inode(journal_inode);
4365 iput(journal_inode);
4366 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4367 return NULL;
4368 }
4369
4370 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4371 journal_inode, journal_inode->i_size);
4372 if (!S_ISREG(journal_inode->i_mode)) {
4373 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4374 iput(journal_inode);
4375 return NULL;
4376 }
4377 return journal_inode;
4378 }
4379
4380 static journal_t *ext4_get_journal(struct super_block *sb,
4381 unsigned int journal_inum)
4382 {
4383 struct inode *journal_inode;
4384 journal_t *journal;
4385
4386 BUG_ON(!ext4_has_feature_journal(sb));
4387
4388 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4389 if (!journal_inode)
4390 return NULL;
4391
4392 journal = jbd2_journal_init_inode(journal_inode);
4393 if (!journal) {
4394 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4395 iput(journal_inode);
4396 return NULL;
4397 }
4398 journal->j_private = sb;
4399 ext4_init_journal_params(sb, journal);
4400 return journal;
4401 }
4402
4403 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4404 dev_t j_dev)
4405 {
4406 struct buffer_head *bh;
4407 journal_t *journal;
4408 ext4_fsblk_t start;
4409 ext4_fsblk_t len;
4410 int hblock, blocksize;
4411 ext4_fsblk_t sb_block;
4412 unsigned long offset;
4413 struct ext4_super_block *es;
4414 struct block_device *bdev;
4415
4416 BUG_ON(!ext4_has_feature_journal(sb));
4417
4418 bdev = ext4_blkdev_get(j_dev, sb);
4419 if (bdev == NULL)
4420 return NULL;
4421
4422 blocksize = sb->s_blocksize;
4423 hblock = bdev_logical_block_size(bdev);
4424 if (blocksize < hblock) {
4425 ext4_msg(sb, KERN_ERR,
4426 "blocksize too small for journal device");
4427 goto out_bdev;
4428 }
4429
4430 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4431 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4432 set_blocksize(bdev, blocksize);
4433 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4434 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4435 "external journal");
4436 goto out_bdev;
4437 }
4438
4439 es = (struct ext4_super_block *) (bh->b_data + offset);
4440 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4441 !(le32_to_cpu(es->s_feature_incompat) &
4442 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4443 ext4_msg(sb, KERN_ERR, "external journal has "
4444 "bad superblock");
4445 brelse(bh);
4446 goto out_bdev;
4447 }
4448
4449 if ((le32_to_cpu(es->s_feature_ro_compat) &
4450 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4451 es->s_checksum != ext4_superblock_csum(sb, es)) {
4452 ext4_msg(sb, KERN_ERR, "external journal has "
4453 "corrupt superblock");
4454 brelse(bh);
4455 goto out_bdev;
4456 }
4457
4458 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4459 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4460 brelse(bh);
4461 goto out_bdev;
4462 }
4463
4464 len = ext4_blocks_count(es);
4465 start = sb_block + 1;
4466 brelse(bh); /* we're done with the superblock */
4467
4468 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4469 start, len, blocksize);
4470 if (!journal) {
4471 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4472 goto out_bdev;
4473 }
4474 journal->j_private = sb;
4475 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4476 wait_on_buffer(journal->j_sb_buffer);
4477 if (!buffer_uptodate(journal->j_sb_buffer)) {
4478 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4479 goto out_journal;
4480 }
4481 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4482 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4483 "user (unsupported) - %d",
4484 be32_to_cpu(journal->j_superblock->s_nr_users));
4485 goto out_journal;
4486 }
4487 EXT4_SB(sb)->journal_bdev = bdev;
4488 ext4_init_journal_params(sb, journal);
4489 return journal;
4490
4491 out_journal:
4492 jbd2_journal_destroy(journal);
4493 out_bdev:
4494 ext4_blkdev_put(bdev);
4495 return NULL;
4496 }
4497
4498 static int ext4_load_journal(struct super_block *sb,
4499 struct ext4_super_block *es,
4500 unsigned long journal_devnum)
4501 {
4502 journal_t *journal;
4503 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4504 dev_t journal_dev;
4505 int err = 0;
4506 int really_read_only;
4507
4508 BUG_ON(!ext4_has_feature_journal(sb));
4509
4510 if (journal_devnum &&
4511 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4512 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4513 "numbers have changed");
4514 journal_dev = new_decode_dev(journal_devnum);
4515 } else
4516 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4517
4518 really_read_only = bdev_read_only(sb->s_bdev);
4519
4520 /*
4521 * Are we loading a blank journal or performing recovery after a
4522 * crash? For recovery, we need to check in advance whether we
4523 * can get read-write access to the device.
4524 */
4525 if (ext4_has_feature_journal_needs_recovery(sb)) {
4526 if (sb->s_flags & MS_RDONLY) {
4527 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4528 "required on readonly filesystem");
4529 if (really_read_only) {
4530 ext4_msg(sb, KERN_ERR, "write access "
4531 "unavailable, cannot proceed");
4532 return -EROFS;
4533 }
4534 ext4_msg(sb, KERN_INFO, "write access will "
4535 "be enabled during recovery");
4536 }
4537 }
4538
4539 if (journal_inum && journal_dev) {
4540 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4541 "and inode journals!");
4542 return -EINVAL;
4543 }
4544
4545 if (journal_inum) {
4546 if (!(journal = ext4_get_journal(sb, journal_inum)))
4547 return -EINVAL;
4548 } else {
4549 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4550 return -EINVAL;
4551 }
4552
4553 if (!(journal->j_flags & JBD2_BARRIER))
4554 ext4_msg(sb, KERN_INFO, "barriers disabled");
4555
4556 if (!ext4_has_feature_journal_needs_recovery(sb))
4557 err = jbd2_journal_wipe(journal, !really_read_only);
4558 if (!err) {
4559 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4560 if (save)
4561 memcpy(save, ((char *) es) +
4562 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4563 err = jbd2_journal_load(journal);
4564 if (save)
4565 memcpy(((char *) es) + EXT4_S_ERR_START,
4566 save, EXT4_S_ERR_LEN);
4567 kfree(save);
4568 }
4569
4570 if (err) {
4571 ext4_msg(sb, KERN_ERR, "error loading journal");
4572 jbd2_journal_destroy(journal);
4573 return err;
4574 }
4575
4576 EXT4_SB(sb)->s_journal = journal;
4577 ext4_clear_journal_err(sb, es);
4578
4579 if (!really_read_only && journal_devnum &&
4580 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4581 es->s_journal_dev = cpu_to_le32(journal_devnum);
4582
4583 /* Make sure we flush the recovery flag to disk. */
4584 ext4_commit_super(sb, 1);
4585 }
4586
4587 return 0;
4588 }
4589
4590 static int ext4_commit_super(struct super_block *sb, int sync)
4591 {
4592 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4593 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4594 int error = 0;
4595
4596 if (!sbh || block_device_ejected(sb))
4597 return error;
4598 /*
4599 * If the file system is mounted read-only, don't update the
4600 * superblock write time. This avoids updating the superblock
4601 * write time when we are mounting the root file system
4602 * read/only but we need to replay the journal; at that point,
4603 * for people who are east of GMT and who make their clock
4604 * tick in localtime for Windows bug-for-bug compatibility,
4605 * the clock is set in the future, and this will cause e2fsck
4606 * to complain and force a full file system check.
4607 */
4608 if (!(sb->s_flags & MS_RDONLY))
4609 es->s_wtime = cpu_to_le32(get_seconds());
4610 if (sb->s_bdev->bd_part)
4611 es->s_kbytes_written =
4612 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4613 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4614 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4615 else
4616 es->s_kbytes_written =
4617 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4618 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4619 ext4_free_blocks_count_set(es,
4620 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4621 &EXT4_SB(sb)->s_freeclusters_counter)));
4622 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4623 es->s_free_inodes_count =
4624 cpu_to_le32(percpu_counter_sum_positive(
4625 &EXT4_SB(sb)->s_freeinodes_counter));
4626 BUFFER_TRACE(sbh, "marking dirty");
4627 ext4_superblock_csum_set(sb);
4628 if (sync)
4629 lock_buffer(sbh);
4630 if (buffer_write_io_error(sbh)) {
4631 /*
4632 * Oh, dear. A previous attempt to write the
4633 * superblock failed. This could happen because the
4634 * USB device was yanked out. Or it could happen to
4635 * be a transient write error and maybe the block will
4636 * be remapped. Nothing we can do but to retry the
4637 * write and hope for the best.
4638 */
4639 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4640 "superblock detected");
4641 clear_buffer_write_io_error(sbh);
4642 set_buffer_uptodate(sbh);
4643 }
4644 mark_buffer_dirty(sbh);
4645 if (sync) {
4646 unlock_buffer(sbh);
4647 error = __sync_dirty_buffer(sbh,
4648 test_opt(sb, BARRIER) ? REQ_FUA : REQ_SYNC);
4649 if (error)
4650 return error;
4651
4652 error = buffer_write_io_error(sbh);
4653 if (error) {
4654 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4655 "superblock");
4656 clear_buffer_write_io_error(sbh);
4657 set_buffer_uptodate(sbh);
4658 }
4659 }
4660 return error;
4661 }
4662
4663 /*
4664 * Have we just finished recovery? If so, and if we are mounting (or
4665 * remounting) the filesystem readonly, then we will end up with a
4666 * consistent fs on disk. Record that fact.
4667 */
4668 static void ext4_mark_recovery_complete(struct super_block *sb,
4669 struct ext4_super_block *es)
4670 {
4671 journal_t *journal = EXT4_SB(sb)->s_journal;
4672
4673 if (!ext4_has_feature_journal(sb)) {
4674 BUG_ON(journal != NULL);
4675 return;
4676 }
4677 jbd2_journal_lock_updates(journal);
4678 if (jbd2_journal_flush(journal) < 0)
4679 goto out;
4680
4681 if (ext4_has_feature_journal_needs_recovery(sb) &&
4682 sb->s_flags & MS_RDONLY) {
4683 ext4_clear_feature_journal_needs_recovery(sb);
4684 ext4_commit_super(sb, 1);
4685 }
4686
4687 out:
4688 jbd2_journal_unlock_updates(journal);
4689 }
4690
4691 /*
4692 * If we are mounting (or read-write remounting) a filesystem whose journal
4693 * has recorded an error from a previous lifetime, move that error to the
4694 * main filesystem now.
4695 */
4696 static void ext4_clear_journal_err(struct super_block *sb,
4697 struct ext4_super_block *es)
4698 {
4699 journal_t *journal;
4700 int j_errno;
4701 const char *errstr;
4702
4703 BUG_ON(!ext4_has_feature_journal(sb));
4704
4705 journal = EXT4_SB(sb)->s_journal;
4706
4707 /*
4708 * Now check for any error status which may have been recorded in the
4709 * journal by a prior ext4_error() or ext4_abort()
4710 */
4711
4712 j_errno = jbd2_journal_errno(journal);
4713 if (j_errno) {
4714 char nbuf[16];
4715
4716 errstr = ext4_decode_error(sb, j_errno, nbuf);
4717 ext4_warning(sb, "Filesystem error recorded "
4718 "from previous mount: %s", errstr);
4719 ext4_warning(sb, "Marking fs in need of filesystem check.");
4720
4721 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4722 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4723 ext4_commit_super(sb, 1);
4724
4725 jbd2_journal_clear_err(journal);
4726 jbd2_journal_update_sb_errno(journal);
4727 }
4728 }
4729
4730 /*
4731 * Force the running and committing transactions to commit,
4732 * and wait on the commit.
4733 */
4734 int ext4_force_commit(struct super_block *sb)
4735 {
4736 journal_t *journal;
4737
4738 if (sb->s_flags & MS_RDONLY)
4739 return 0;
4740
4741 journal = EXT4_SB(sb)->s_journal;
4742 return ext4_journal_force_commit(journal);
4743 }
4744
4745 static int ext4_sync_fs(struct super_block *sb, int wait)
4746 {
4747 int ret = 0;
4748 tid_t target;
4749 bool needs_barrier = false;
4750 struct ext4_sb_info *sbi = EXT4_SB(sb);
4751
4752 trace_ext4_sync_fs(sb, wait);
4753 flush_workqueue(sbi->rsv_conversion_wq);
4754 /*
4755 * Writeback quota in non-journalled quota case - journalled quota has
4756 * no dirty dquots
4757 */
4758 dquot_writeback_dquots(sb, -1);
4759 /*
4760 * Data writeback is possible w/o journal transaction, so barrier must
4761 * being sent at the end of the function. But we can skip it if
4762 * transaction_commit will do it for us.
4763 */
4764 if (sbi->s_journal) {
4765 target = jbd2_get_latest_transaction(sbi->s_journal);
4766 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4767 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4768 needs_barrier = true;
4769
4770 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4771 if (wait)
4772 ret = jbd2_log_wait_commit(sbi->s_journal,
4773 target);
4774 }
4775 } else if (wait && test_opt(sb, BARRIER))
4776 needs_barrier = true;
4777 if (needs_barrier) {
4778 int err;
4779 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4780 if (!ret)
4781 ret = err;
4782 }
4783
4784 return ret;
4785 }
4786
4787 /*
4788 * LVM calls this function before a (read-only) snapshot is created. This
4789 * gives us a chance to flush the journal completely and mark the fs clean.
4790 *
4791 * Note that only this function cannot bring a filesystem to be in a clean
4792 * state independently. It relies on upper layer to stop all data & metadata
4793 * modifications.
4794 */
4795 static int ext4_freeze(struct super_block *sb)
4796 {
4797 int error = 0;
4798 journal_t *journal;
4799
4800 if (sb->s_flags & MS_RDONLY)
4801 return 0;
4802
4803 journal = EXT4_SB(sb)->s_journal;
4804
4805 if (journal) {
4806 /* Now we set up the journal barrier. */
4807 jbd2_journal_lock_updates(journal);
4808
4809 /*
4810 * Don't clear the needs_recovery flag if we failed to
4811 * flush the journal.
4812 */
4813 error = jbd2_journal_flush(journal);
4814 if (error < 0)
4815 goto out;
4816
4817 /* Journal blocked and flushed, clear needs_recovery flag. */
4818 ext4_clear_feature_journal_needs_recovery(sb);
4819 }
4820
4821 error = ext4_commit_super(sb, 1);
4822 out:
4823 if (journal)
4824 /* we rely on upper layer to stop further updates */
4825 jbd2_journal_unlock_updates(journal);
4826 return error;
4827 }
4828
4829 /*
4830 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4831 * flag here, even though the filesystem is not technically dirty yet.
4832 */
4833 static int ext4_unfreeze(struct super_block *sb)
4834 {
4835 if (sb->s_flags & MS_RDONLY)
4836 return 0;
4837
4838 if (EXT4_SB(sb)->s_journal) {
4839 /* Reset the needs_recovery flag before the fs is unlocked. */
4840 ext4_set_feature_journal_needs_recovery(sb);
4841 }
4842
4843 ext4_commit_super(sb, 1);
4844 return 0;
4845 }
4846
4847 /*
4848 * Structure to save mount options for ext4_remount's benefit
4849 */
4850 struct ext4_mount_options {
4851 unsigned long s_mount_opt;
4852 unsigned long s_mount_opt2;
4853 kuid_t s_resuid;
4854 kgid_t s_resgid;
4855 unsigned long s_commit_interval;
4856 u32 s_min_batch_time, s_max_batch_time;
4857 #ifdef CONFIG_QUOTA
4858 int s_jquota_fmt;
4859 char *s_qf_names[EXT4_MAXQUOTAS];
4860 #endif
4861 };
4862
4863 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4864 {
4865 struct ext4_super_block *es;
4866 struct ext4_sb_info *sbi = EXT4_SB(sb);
4867 unsigned long old_sb_flags;
4868 struct ext4_mount_options old_opts;
4869 int enable_quota = 0;
4870 ext4_group_t g;
4871 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4872 int err = 0;
4873 #ifdef CONFIG_QUOTA
4874 int i, j;
4875 #endif
4876 char *orig_data = kstrdup(data, GFP_KERNEL);
4877
4878 /* Store the original options */
4879 old_sb_flags = sb->s_flags;
4880 old_opts.s_mount_opt = sbi->s_mount_opt;
4881 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4882 old_opts.s_resuid = sbi->s_resuid;
4883 old_opts.s_resgid = sbi->s_resgid;
4884 old_opts.s_commit_interval = sbi->s_commit_interval;
4885 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4886 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4887 #ifdef CONFIG_QUOTA
4888 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4889 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4890 if (sbi->s_qf_names[i]) {
4891 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4892 GFP_KERNEL);
4893 if (!old_opts.s_qf_names[i]) {
4894 for (j = 0; j < i; j++)
4895 kfree(old_opts.s_qf_names[j]);
4896 kfree(orig_data);
4897 return -ENOMEM;
4898 }
4899 } else
4900 old_opts.s_qf_names[i] = NULL;
4901 #endif
4902 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4903 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4904
4905 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4906 err = -EINVAL;
4907 goto restore_opts;
4908 }
4909
4910 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4911 test_opt(sb, JOURNAL_CHECKSUM)) {
4912 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4913 "during remount not supported; ignoring");
4914 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4915 }
4916
4917 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4918 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4919 ext4_msg(sb, KERN_ERR, "can't mount with "
4920 "both data=journal and delalloc");
4921 err = -EINVAL;
4922 goto restore_opts;
4923 }
4924 if (test_opt(sb, DIOREAD_NOLOCK)) {
4925 ext4_msg(sb, KERN_ERR, "can't mount with "
4926 "both data=journal and dioread_nolock");
4927 err = -EINVAL;
4928 goto restore_opts;
4929 }
4930 if (test_opt(sb, DAX)) {
4931 ext4_msg(sb, KERN_ERR, "can't mount with "
4932 "both data=journal and dax");
4933 err = -EINVAL;
4934 goto restore_opts;
4935 }
4936 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
4937 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4938 ext4_msg(sb, KERN_ERR, "can't mount with "
4939 "journal_async_commit in data=ordered mode");
4940 err = -EINVAL;
4941 goto restore_opts;
4942 }
4943 }
4944
4945 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4946 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4947 "dax flag with busy inodes while remounting");
4948 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4949 }
4950
4951 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4952 ext4_abort(sb, "Abort forced by user");
4953
4954 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4955 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4956
4957 es = sbi->s_es;
4958
4959 if (sbi->s_journal) {
4960 ext4_init_journal_params(sb, sbi->s_journal);
4961 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4962 }
4963
4964 if (*flags & MS_LAZYTIME)
4965 sb->s_flags |= MS_LAZYTIME;
4966
4967 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4968 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4969 err = -EROFS;
4970 goto restore_opts;
4971 }
4972
4973 if (*flags & MS_RDONLY) {
4974 err = sync_filesystem(sb);
4975 if (err < 0)
4976 goto restore_opts;
4977 err = dquot_suspend(sb, -1);
4978 if (err < 0)
4979 goto restore_opts;
4980
4981 /*
4982 * First of all, the unconditional stuff we have to do
4983 * to disable replay of the journal when we next remount
4984 */
4985 sb->s_flags |= MS_RDONLY;
4986
4987 /*
4988 * OK, test if we are remounting a valid rw partition
4989 * readonly, and if so set the rdonly flag and then
4990 * mark the partition as valid again.
4991 */
4992 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4993 (sbi->s_mount_state & EXT4_VALID_FS))
4994 es->s_state = cpu_to_le16(sbi->s_mount_state);
4995
4996 if (sbi->s_journal)
4997 ext4_mark_recovery_complete(sb, es);
4998 } else {
4999 /* Make sure we can mount this feature set readwrite */
5000 if (ext4_has_feature_readonly(sb) ||
5001 !ext4_feature_set_ok(sb, 0)) {
5002 err = -EROFS;
5003 goto restore_opts;
5004 }
5005 /*
5006 * Make sure the group descriptor checksums
5007 * are sane. If they aren't, refuse to remount r/w.
5008 */
5009 for (g = 0; g < sbi->s_groups_count; g++) {
5010 struct ext4_group_desc *gdp =
5011 ext4_get_group_desc(sb, g, NULL);
5012
5013 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5014 ext4_msg(sb, KERN_ERR,
5015 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5016 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5017 le16_to_cpu(gdp->bg_checksum));
5018 err = -EFSBADCRC;
5019 goto restore_opts;
5020 }
5021 }
5022
5023 /*
5024 * If we have an unprocessed orphan list hanging
5025 * around from a previously readonly bdev mount,
5026 * require a full umount/remount for now.
5027 */
5028 if (es->s_last_orphan) {
5029 ext4_msg(sb, KERN_WARNING, "Couldn't "
5030 "remount RDWR because of unprocessed "
5031 "orphan inode list. Please "
5032 "umount/remount instead");
5033 err = -EINVAL;
5034 goto restore_opts;
5035 }
5036
5037 /*
5038 * Mounting a RDONLY partition read-write, so reread
5039 * and store the current valid flag. (It may have
5040 * been changed by e2fsck since we originally mounted
5041 * the partition.)
5042 */
5043 if (sbi->s_journal)
5044 ext4_clear_journal_err(sb, es);
5045 sbi->s_mount_state = le16_to_cpu(es->s_state);
5046 if (!ext4_setup_super(sb, es, 0))
5047 sb->s_flags &= ~MS_RDONLY;
5048 if (ext4_has_feature_mmp(sb))
5049 if (ext4_multi_mount_protect(sb,
5050 le64_to_cpu(es->s_mmp_block))) {
5051 err = -EROFS;
5052 goto restore_opts;
5053 }
5054 enable_quota = 1;
5055 }
5056 }
5057
5058 /*
5059 * Reinitialize lazy itable initialization thread based on
5060 * current settings
5061 */
5062 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5063 ext4_unregister_li_request(sb);
5064 else {
5065 ext4_group_t first_not_zeroed;
5066 first_not_zeroed = ext4_has_uninit_itable(sb);
5067 ext4_register_li_request(sb, first_not_zeroed);
5068 }
5069
5070 ext4_setup_system_zone(sb);
5071 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5072 ext4_commit_super(sb, 1);
5073
5074 #ifdef CONFIG_QUOTA
5075 /* Release old quota file names */
5076 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5077 kfree(old_opts.s_qf_names[i]);
5078 if (enable_quota) {
5079 if (sb_any_quota_suspended(sb))
5080 dquot_resume(sb, -1);
5081 else if (ext4_has_feature_quota(sb)) {
5082 err = ext4_enable_quotas(sb);
5083 if (err)
5084 goto restore_opts;
5085 }
5086 }
5087 #endif
5088
5089 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5090 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5091 kfree(orig_data);
5092 return 0;
5093
5094 restore_opts:
5095 sb->s_flags = old_sb_flags;
5096 sbi->s_mount_opt = old_opts.s_mount_opt;
5097 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5098 sbi->s_resuid = old_opts.s_resuid;
5099 sbi->s_resgid = old_opts.s_resgid;
5100 sbi->s_commit_interval = old_opts.s_commit_interval;
5101 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5102 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5103 #ifdef CONFIG_QUOTA
5104 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5105 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5106 kfree(sbi->s_qf_names[i]);
5107 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5108 }
5109 #endif
5110 kfree(orig_data);
5111 return err;
5112 }
5113
5114 #ifdef CONFIG_QUOTA
5115 static int ext4_statfs_project(struct super_block *sb,
5116 kprojid_t projid, struct kstatfs *buf)
5117 {
5118 struct kqid qid;
5119 struct dquot *dquot;
5120 u64 limit;
5121 u64 curblock;
5122
5123 qid = make_kqid_projid(projid);
5124 dquot = dqget(sb, qid);
5125 if (IS_ERR(dquot))
5126 return PTR_ERR(dquot);
5127 spin_lock(&dq_data_lock);
5128
5129 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5130 dquot->dq_dqb.dqb_bsoftlimit :
5131 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5132 if (limit && buf->f_blocks > limit) {
5133 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5134 buf->f_blocks = limit;
5135 buf->f_bfree = buf->f_bavail =
5136 (buf->f_blocks > curblock) ?
5137 (buf->f_blocks - curblock) : 0;
5138 }
5139
5140 limit = dquot->dq_dqb.dqb_isoftlimit ?
5141 dquot->dq_dqb.dqb_isoftlimit :
5142 dquot->dq_dqb.dqb_ihardlimit;
5143 if (limit && buf->f_files > limit) {
5144 buf->f_files = limit;
5145 buf->f_ffree =
5146 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5147 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5148 }
5149
5150 spin_unlock(&dq_data_lock);
5151 dqput(dquot);
5152 return 0;
5153 }
5154 #endif
5155
5156 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5157 {
5158 struct super_block *sb = dentry->d_sb;
5159 struct ext4_sb_info *sbi = EXT4_SB(sb);
5160 struct ext4_super_block *es = sbi->s_es;
5161 ext4_fsblk_t overhead = 0, resv_blocks;
5162 u64 fsid;
5163 s64 bfree;
5164 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5165
5166 if (!test_opt(sb, MINIX_DF))
5167 overhead = sbi->s_overhead;
5168
5169 buf->f_type = EXT4_SUPER_MAGIC;
5170 buf->f_bsize = sb->s_blocksize;
5171 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5172 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5173 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5174 /* prevent underflow in case that few free space is available */
5175 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5176 buf->f_bavail = buf->f_bfree -
5177 (ext4_r_blocks_count(es) + resv_blocks);
5178 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5179 buf->f_bavail = 0;
5180 buf->f_files = le32_to_cpu(es->s_inodes_count);
5181 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5182 buf->f_namelen = EXT4_NAME_LEN;
5183 fsid = le64_to_cpup((void *)es->s_uuid) ^
5184 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5185 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5186 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5187
5188 #ifdef CONFIG_QUOTA
5189 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5190 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5191 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5192 #endif
5193 return 0;
5194 }
5195
5196 /* Helper function for writing quotas on sync - we need to start transaction
5197 * before quota file is locked for write. Otherwise the are possible deadlocks:
5198 * Process 1 Process 2
5199 * ext4_create() quota_sync()
5200 * jbd2_journal_start() write_dquot()
5201 * dquot_initialize() down(dqio_mutex)
5202 * down(dqio_mutex) jbd2_journal_start()
5203 *
5204 */
5205
5206 #ifdef CONFIG_QUOTA
5207
5208 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5209 {
5210 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5211 }
5212
5213 static int ext4_write_dquot(struct dquot *dquot)
5214 {
5215 int ret, err;
5216 handle_t *handle;
5217 struct inode *inode;
5218
5219 inode = dquot_to_inode(dquot);
5220 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5221 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5222 if (IS_ERR(handle))
5223 return PTR_ERR(handle);
5224 ret = dquot_commit(dquot);
5225 err = ext4_journal_stop(handle);
5226 if (!ret)
5227 ret = err;
5228 return ret;
5229 }
5230
5231 static int ext4_acquire_dquot(struct dquot *dquot)
5232 {
5233 int ret, err;
5234 handle_t *handle;
5235
5236 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5237 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5238 if (IS_ERR(handle))
5239 return PTR_ERR(handle);
5240 ret = dquot_acquire(dquot);
5241 err = ext4_journal_stop(handle);
5242 if (!ret)
5243 ret = err;
5244 return ret;
5245 }
5246
5247 static int ext4_release_dquot(struct dquot *dquot)
5248 {
5249 int ret, err;
5250 handle_t *handle;
5251
5252 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5253 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5254 if (IS_ERR(handle)) {
5255 /* Release dquot anyway to avoid endless cycle in dqput() */
5256 dquot_release(dquot);
5257 return PTR_ERR(handle);
5258 }
5259 ret = dquot_release(dquot);
5260 err = ext4_journal_stop(handle);
5261 if (!ret)
5262 ret = err;
5263 return ret;
5264 }
5265
5266 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5267 {
5268 struct super_block *sb = dquot->dq_sb;
5269 struct ext4_sb_info *sbi = EXT4_SB(sb);
5270
5271 /* Are we journaling quotas? */
5272 if (ext4_has_feature_quota(sb) ||
5273 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5274 dquot_mark_dquot_dirty(dquot);
5275 return ext4_write_dquot(dquot);
5276 } else {
5277 return dquot_mark_dquot_dirty(dquot);
5278 }
5279 }
5280
5281 static int ext4_write_info(struct super_block *sb, int type)
5282 {
5283 int ret, err;
5284 handle_t *handle;
5285
5286 /* Data block + inode block */
5287 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5288 if (IS_ERR(handle))
5289 return PTR_ERR(handle);
5290 ret = dquot_commit_info(sb, type);
5291 err = ext4_journal_stop(handle);
5292 if (!ret)
5293 ret = err;
5294 return ret;
5295 }
5296
5297 /*
5298 * Turn on quotas during mount time - we need to find
5299 * the quota file and such...
5300 */
5301 static int ext4_quota_on_mount(struct super_block *sb, int type)
5302 {
5303 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5304 EXT4_SB(sb)->s_jquota_fmt, type);
5305 }
5306
5307 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5308 {
5309 struct ext4_inode_info *ei = EXT4_I(inode);
5310
5311 /* The first argument of lockdep_set_subclass has to be
5312 * *exactly* the same as the argument to init_rwsem() --- in
5313 * this case, in init_once() --- or lockdep gets unhappy
5314 * because the name of the lock is set using the
5315 * stringification of the argument to init_rwsem().
5316 */
5317 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5318 lockdep_set_subclass(&ei->i_data_sem, subclass);
5319 }
5320
5321 /*
5322 * Standard function to be called on quota_on
5323 */
5324 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5325 const struct path *path)
5326 {
5327 int err;
5328
5329 if (!test_opt(sb, QUOTA))
5330 return -EINVAL;
5331
5332 /* Quotafile not on the same filesystem? */
5333 if (path->dentry->d_sb != sb)
5334 return -EXDEV;
5335 /* Journaling quota? */
5336 if (EXT4_SB(sb)->s_qf_names[type]) {
5337 /* Quotafile not in fs root? */
5338 if (path->dentry->d_parent != sb->s_root)
5339 ext4_msg(sb, KERN_WARNING,
5340 "Quota file not on filesystem root. "
5341 "Journaled quota will not work");
5342 }
5343
5344 /*
5345 * When we journal data on quota file, we have to flush journal to see
5346 * all updates to the file when we bypass pagecache...
5347 */
5348 if (EXT4_SB(sb)->s_journal &&
5349 ext4_should_journal_data(d_inode(path->dentry))) {
5350 /*
5351 * We don't need to lock updates but journal_flush() could
5352 * otherwise be livelocked...
5353 */
5354 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5355 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5356 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5357 if (err)
5358 return err;
5359 }
5360 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5361 err = dquot_quota_on(sb, type, format_id, path);
5362 if (err)
5363 lockdep_set_quota_inode(path->dentry->d_inode,
5364 I_DATA_SEM_NORMAL);
5365 return err;
5366 }
5367
5368 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5369 unsigned int flags)
5370 {
5371 int err;
5372 struct inode *qf_inode;
5373 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5374 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5375 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5376 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5377 };
5378
5379 BUG_ON(!ext4_has_feature_quota(sb));
5380
5381 if (!qf_inums[type])
5382 return -EPERM;
5383
5384 qf_inode = ext4_iget(sb, qf_inums[type]);
5385 if (IS_ERR(qf_inode)) {
5386 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5387 return PTR_ERR(qf_inode);
5388 }
5389
5390 /* Don't account quota for quota files to avoid recursion */
5391 qf_inode->i_flags |= S_NOQUOTA;
5392 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5393 err = dquot_enable(qf_inode, type, format_id, flags);
5394 iput(qf_inode);
5395 if (err)
5396 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5397
5398 return err;
5399 }
5400
5401 /* Enable usage tracking for all quota types. */
5402 static int ext4_enable_quotas(struct super_block *sb)
5403 {
5404 int type, err = 0;
5405 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5406 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5407 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5408 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5409 };
5410 bool quota_mopt[EXT4_MAXQUOTAS] = {
5411 test_opt(sb, USRQUOTA),
5412 test_opt(sb, GRPQUOTA),
5413 test_opt(sb, PRJQUOTA),
5414 };
5415
5416 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5417 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5418 if (qf_inums[type]) {
5419 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5420 DQUOT_USAGE_ENABLED |
5421 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5422 if (err) {
5423 ext4_warning(sb,
5424 "Failed to enable quota tracking "
5425 "(type=%d, err=%d). Please run "
5426 "e2fsck to fix.", type, err);
5427 return err;
5428 }
5429 }
5430 }
5431 return 0;
5432 }
5433
5434 static int ext4_quota_off(struct super_block *sb, int type)
5435 {
5436 struct inode *inode = sb_dqopt(sb)->files[type];
5437 handle_t *handle;
5438
5439 /* Force all delayed allocation blocks to be allocated.
5440 * Caller already holds s_umount sem */
5441 if (test_opt(sb, DELALLOC))
5442 sync_filesystem(sb);
5443
5444 if (!inode)
5445 goto out;
5446
5447 /* Update modification times of quota files when userspace can
5448 * start looking at them */
5449 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5450 if (IS_ERR(handle))
5451 goto out;
5452 inode->i_mtime = inode->i_ctime = current_time(inode);
5453 ext4_mark_inode_dirty(handle, inode);
5454 ext4_journal_stop(handle);
5455
5456 out:
5457 return dquot_quota_off(sb, type);
5458 }
5459
5460 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5461 * acquiring the locks... As quota files are never truncated and quota code
5462 * itself serializes the operations (and no one else should touch the files)
5463 * we don't have to be afraid of races */
5464 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5465 size_t len, loff_t off)
5466 {
5467 struct inode *inode = sb_dqopt(sb)->files[type];
5468 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5469 int offset = off & (sb->s_blocksize - 1);
5470 int tocopy;
5471 size_t toread;
5472 struct buffer_head *bh;
5473 loff_t i_size = i_size_read(inode);
5474
5475 if (off > i_size)
5476 return 0;
5477 if (off+len > i_size)
5478 len = i_size-off;
5479 toread = len;
5480 while (toread > 0) {
5481 tocopy = sb->s_blocksize - offset < toread ?
5482 sb->s_blocksize - offset : toread;
5483 bh = ext4_bread(NULL, inode, blk, 0);
5484 if (IS_ERR(bh))
5485 return PTR_ERR(bh);
5486 if (!bh) /* A hole? */
5487 memset(data, 0, tocopy);
5488 else
5489 memcpy(data, bh->b_data+offset, tocopy);
5490 brelse(bh);
5491 offset = 0;
5492 toread -= tocopy;
5493 data += tocopy;
5494 blk++;
5495 }
5496 return len;
5497 }
5498
5499 /* Write to quotafile (we know the transaction is already started and has
5500 * enough credits) */
5501 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5502 const char *data, size_t len, loff_t off)
5503 {
5504 struct inode *inode = sb_dqopt(sb)->files[type];
5505 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5506 int err, offset = off & (sb->s_blocksize - 1);
5507 int retries = 0;
5508 struct buffer_head *bh;
5509 handle_t *handle = journal_current_handle();
5510
5511 if (EXT4_SB(sb)->s_journal && !handle) {
5512 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5513 " cancelled because transaction is not started",
5514 (unsigned long long)off, (unsigned long long)len);
5515 return -EIO;
5516 }
5517 /*
5518 * Since we account only one data block in transaction credits,
5519 * then it is impossible to cross a block boundary.
5520 */
5521 if (sb->s_blocksize - offset < len) {
5522 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5523 " cancelled because not block aligned",
5524 (unsigned long long)off, (unsigned long long)len);
5525 return -EIO;
5526 }
5527
5528 do {
5529 bh = ext4_bread(handle, inode, blk,
5530 EXT4_GET_BLOCKS_CREATE |
5531 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5532 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5533 ext4_should_retry_alloc(inode->i_sb, &retries));
5534 if (IS_ERR(bh))
5535 return PTR_ERR(bh);
5536 if (!bh)
5537 goto out;
5538 BUFFER_TRACE(bh, "get write access");
5539 err = ext4_journal_get_write_access(handle, bh);
5540 if (err) {
5541 brelse(bh);
5542 return err;
5543 }
5544 lock_buffer(bh);
5545 memcpy(bh->b_data+offset, data, len);
5546 flush_dcache_page(bh->b_page);
5547 unlock_buffer(bh);
5548 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5549 brelse(bh);
5550 out:
5551 if (inode->i_size < off + len) {
5552 i_size_write(inode, off + len);
5553 EXT4_I(inode)->i_disksize = inode->i_size;
5554 ext4_mark_inode_dirty(handle, inode);
5555 }
5556 return len;
5557 }
5558
5559 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5560 {
5561 const struct quota_format_ops *ops;
5562
5563 if (!sb_has_quota_loaded(sb, qid->type))
5564 return -ESRCH;
5565 ops = sb_dqopt(sb)->ops[qid->type];
5566 if (!ops || !ops->get_next_id)
5567 return -ENOSYS;
5568 return dquot_get_next_id(sb, qid);
5569 }
5570 #endif
5571
5572 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5573 const char *dev_name, void *data)
5574 {
5575 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5576 }
5577
5578 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5579 static inline void register_as_ext2(void)
5580 {
5581 int err = register_filesystem(&ext2_fs_type);
5582 if (err)
5583 printk(KERN_WARNING
5584 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5585 }
5586
5587 static inline void unregister_as_ext2(void)
5588 {
5589 unregister_filesystem(&ext2_fs_type);
5590 }
5591
5592 static inline int ext2_feature_set_ok(struct super_block *sb)
5593 {
5594 if (ext4_has_unknown_ext2_incompat_features(sb))
5595 return 0;
5596 if (sb->s_flags & MS_RDONLY)
5597 return 1;
5598 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5599 return 0;
5600 return 1;
5601 }
5602 #else
5603 static inline void register_as_ext2(void) { }
5604 static inline void unregister_as_ext2(void) { }
5605 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5606 #endif
5607
5608 static inline void register_as_ext3(void)
5609 {
5610 int err = register_filesystem(&ext3_fs_type);
5611 if (err)
5612 printk(KERN_WARNING
5613 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5614 }
5615
5616 static inline void unregister_as_ext3(void)
5617 {
5618 unregister_filesystem(&ext3_fs_type);
5619 }
5620
5621 static inline int ext3_feature_set_ok(struct super_block *sb)
5622 {
5623 if (ext4_has_unknown_ext3_incompat_features(sb))
5624 return 0;
5625 if (!ext4_has_feature_journal(sb))
5626 return 0;
5627 if (sb->s_flags & MS_RDONLY)
5628 return 1;
5629 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5630 return 0;
5631 return 1;
5632 }
5633
5634 static struct file_system_type ext4_fs_type = {
5635 .owner = THIS_MODULE,
5636 .name = "ext4",
5637 .mount = ext4_mount,
5638 .kill_sb = kill_block_super,
5639 .fs_flags = FS_REQUIRES_DEV | FS_USERNS_MOUNT,
5640 };
5641 MODULE_ALIAS_FS("ext4");
5642
5643 /* Shared across all ext4 file systems */
5644 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5645
5646 static int __init ext4_init_fs(void)
5647 {
5648 int i, err;
5649
5650 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5651 ext4_li_info = NULL;
5652 mutex_init(&ext4_li_mtx);
5653
5654 /* Build-time check for flags consistency */
5655 ext4_check_flag_values();
5656
5657 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5658 init_waitqueue_head(&ext4__ioend_wq[i]);
5659
5660 err = ext4_init_es();
5661 if (err)
5662 return err;
5663
5664 err = ext4_init_pageio();
5665 if (err)
5666 goto out5;
5667
5668 err = ext4_init_system_zone();
5669 if (err)
5670 goto out4;
5671
5672 err = ext4_init_sysfs();
5673 if (err)
5674 goto out3;
5675
5676 err = ext4_init_mballoc();
5677 if (err)
5678 goto out2;
5679 err = init_inodecache();
5680 if (err)
5681 goto out1;
5682 register_as_ext3();
5683 register_as_ext2();
5684 err = register_filesystem(&ext4_fs_type);
5685 if (err)
5686 goto out;
5687
5688 return 0;
5689 out:
5690 unregister_as_ext2();
5691 unregister_as_ext3();
5692 destroy_inodecache();
5693 out1:
5694 ext4_exit_mballoc();
5695 out2:
5696 ext4_exit_sysfs();
5697 out3:
5698 ext4_exit_system_zone();
5699 out4:
5700 ext4_exit_pageio();
5701 out5:
5702 ext4_exit_es();
5703
5704 return err;
5705 }
5706
5707 static void __exit ext4_exit_fs(void)
5708 {
5709 ext4_destroy_lazyinit_thread();
5710 unregister_as_ext2();
5711 unregister_as_ext3();
5712 unregister_filesystem(&ext4_fs_type);
5713 destroy_inodecache();
5714 ext4_exit_mballoc();
5715 ext4_exit_sysfs();
5716 ext4_exit_system_zone();
5717 ext4_exit_pageio();
5718 ext4_exit_es();
5719 }
5720
5721 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5722 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5723 MODULE_LICENSE("GPL");
5724 module_init(ext4_init_fs)
5725 module_exit(ext4_exit_fs)
ubuntu-zesty-kernel mirror