]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blob - fs/ext4/super.c
projects / mirror_ubuntu-bionic-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
spi-imx: Implements handling of the SPI_READY mode flag.
[mirror_ubuntu-bionic-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
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
45
46 #include "ext4.h"
47 #include "ext4_extents.h" /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
49 #include "xattr.h"
50 #include "acl.h"
51 #include "mballoc.h"
52
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
55
56 static struct ext4_lazy_init *ext4_li_info;
57 static struct mutex ext4_li_mtx;
58 static struct ratelimit_state ext4_mount_msg_ratelimit;
59
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61 unsigned long journal_devnum);
62 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
63 static int ext4_commit_super(struct super_block *sb, int sync);
64 static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66 static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
68 static int ext4_sync_fs(struct super_block *sb, int wait);
69 static int ext4_remount(struct super_block *sb, int *flags, char *data);
70 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
71 static int ext4_unfreeze(struct super_block *sb);
72 static int ext4_freeze(struct super_block *sb);
73 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74 const char *dev_name, void *data);
75 static inline int ext2_feature_set_ok(struct super_block *sb);
76 static inline int ext3_feature_set_ok(struct super_block *sb);
77 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block *sb);
80 static void ext4_clear_request_list(void);
81 static struct inode *ext4_get_journal_inode(struct super_block *sb,
82 unsigned int journal_inum);
83
84 /*
85 * Lock ordering
86 *
87 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
88 * i_mmap_rwsem (inode->i_mmap_rwsem)!
89 *
90 * page fault path:
91 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
92 * page lock -> i_data_sem (rw)
93 *
94 * buffered write path:
95 * sb_start_write -> i_mutex -> mmap_sem
96 * sb_start_write -> i_mutex -> transaction start -> page lock ->
97 * i_data_sem (rw)
98 *
99 * truncate:
100 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101 * i_mmap_rwsem (w) -> page lock
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * transaction start -> i_data_sem (rw)
104 *
105 * direct IO:
106 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
107 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
108 * transaction start -> i_data_sem (rw)
109 *
110 * writepages:
111 * transaction start -> page lock(s) -> i_data_sem (rw)
112 */
113
114 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
115 static struct file_system_type ext2_fs_type = {
116 .owner = THIS_MODULE,
117 .name = "ext2",
118 .mount = ext4_mount,
119 .kill_sb = kill_block_super,
120 .fs_flags = FS_REQUIRES_DEV,
121 };
122 MODULE_ALIAS_FS("ext2");
123 MODULE_ALIAS("ext2");
124 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
125 #else
126 #define IS_EXT2_SB(sb) (0)
127 #endif
128
129
130 static struct file_system_type ext3_fs_type = {
131 .owner = THIS_MODULE,
132 .name = "ext3",
133 .mount = ext4_mount,
134 .kill_sb = kill_block_super,
135 .fs_flags = FS_REQUIRES_DEV,
136 };
137 MODULE_ALIAS_FS("ext3");
138 MODULE_ALIAS("ext3");
139 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
140
141 static int ext4_verify_csum_type(struct super_block *sb,
142 struct ext4_super_block *es)
143 {
144 if (!ext4_has_feature_metadata_csum(sb))
145 return 1;
146
147 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
148 }
149
150 static __le32 ext4_superblock_csum(struct super_block *sb,
151 struct ext4_super_block *es)
152 {
153 struct ext4_sb_info *sbi = EXT4_SB(sb);
154 int offset = offsetof(struct ext4_super_block, s_checksum);
155 __u32 csum;
156
157 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
158
159 return cpu_to_le32(csum);
160 }
161
162 static int ext4_superblock_csum_verify(struct super_block *sb,
163 struct ext4_super_block *es)
164 {
165 if (!ext4_has_metadata_csum(sb))
166 return 1;
167
168 return es->s_checksum == ext4_superblock_csum(sb, es);
169 }
170
171 void ext4_superblock_csum_set(struct super_block *sb)
172 {
173 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
174
175 if (!ext4_has_metadata_csum(sb))
176 return;
177
178 es->s_checksum = ext4_superblock_csum(sb, es);
179 }
180
181 void *ext4_kvmalloc(size_t size, gfp_t flags)
182 {
183 void *ret;
184
185 ret = kmalloc(size, flags | __GFP_NOWARN);
186 if (!ret)
187 ret = __vmalloc(size, flags, PAGE_KERNEL);
188 return ret;
189 }
190
191 void *ext4_kvzalloc(size_t size, gfp_t flags)
192 {
193 void *ret;
194
195 ret = kzalloc(size, flags | __GFP_NOWARN);
196 if (!ret)
197 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
198 return ret;
199 }
200
201 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
202 struct ext4_group_desc *bg)
203 {
204 return le32_to_cpu(bg->bg_block_bitmap_lo) |
205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
206 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
207 }
208
209 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
210 struct ext4_group_desc *bg)
211 {
212 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
214 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
215 }
216
217 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
218 struct ext4_group_desc *bg)
219 {
220 return le32_to_cpu(bg->bg_inode_table_lo) |
221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
222 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
223 }
224
225 __u32 ext4_free_group_clusters(struct super_block *sb,
226 struct ext4_group_desc *bg)
227 {
228 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
230 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
231 }
232
233 __u32 ext4_free_inodes_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
235 {
236 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
238 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
239 }
240
241 __u32 ext4_used_dirs_count(struct super_block *sb,
242 struct ext4_group_desc *bg)
243 {
244 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
245 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
246 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
247 }
248
249 __u32 ext4_itable_unused_count(struct super_block *sb,
250 struct ext4_group_desc *bg)
251 {
252 return le16_to_cpu(bg->bg_itable_unused_lo) |
253 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
254 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
255 }
256
257 void ext4_block_bitmap_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
259 {
260 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
263 }
264
265 void ext4_inode_bitmap_set(struct super_block *sb,
266 struct ext4_group_desc *bg, ext4_fsblk_t blk)
267 {
268 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
271 }
272
273 void ext4_inode_table_set(struct super_block *sb,
274 struct ext4_group_desc *bg, ext4_fsblk_t blk)
275 {
276 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
279 }
280
281 void ext4_free_group_clusters_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
283 {
284 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
287 }
288
289 void ext4_free_inodes_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
291 {
292 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
295 }
296
297 void ext4_used_dirs_set(struct super_block *sb,
298 struct ext4_group_desc *bg, __u32 count)
299 {
300 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
301 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
302 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
303 }
304
305 void ext4_itable_unused_set(struct super_block *sb,
306 struct ext4_group_desc *bg, __u32 count)
307 {
308 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
309 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
310 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
311 }
312
313
314 static void __save_error_info(struct super_block *sb, const char *func,
315 unsigned int line)
316 {
317 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
318
319 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
320 if (bdev_read_only(sb->s_bdev))
321 return;
322 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
323 es->s_last_error_time = cpu_to_le32(get_seconds());
324 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
325 es->s_last_error_line = cpu_to_le32(line);
326 if (!es->s_first_error_time) {
327 es->s_first_error_time = es->s_last_error_time;
328 strncpy(es->s_first_error_func, func,
329 sizeof(es->s_first_error_func));
330 es->s_first_error_line = cpu_to_le32(line);
331 es->s_first_error_ino = es->s_last_error_ino;
332 es->s_first_error_block = es->s_last_error_block;
333 }
334 /*
335 * Start the daily error reporting function if it hasn't been
336 * started already
337 */
338 if (!es->s_error_count)
339 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
340 le32_add_cpu(&es->s_error_count, 1);
341 }
342
343 static void save_error_info(struct super_block *sb, const char *func,
344 unsigned int line)
345 {
346 __save_error_info(sb, func, line);
347 ext4_commit_super(sb, 1);
348 }
349
350 /*
351 * The del_gendisk() function uninitializes the disk-specific data
352 * structures, including the bdi structure, without telling anyone
353 * else. Once this happens, any attempt to call mark_buffer_dirty()
354 * (for example, by ext4_commit_super), will cause a kernel OOPS.
355 * This is a kludge to prevent these oops until we can put in a proper
356 * hook in del_gendisk() to inform the VFS and file system layers.
357 */
358 static int block_device_ejected(struct super_block *sb)
359 {
360 struct inode *bd_inode = sb->s_bdev->bd_inode;
361 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
362
363 return bdi->dev == NULL;
364 }
365
366 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
367 {
368 struct super_block *sb = journal->j_private;
369 struct ext4_sb_info *sbi = EXT4_SB(sb);
370 int error = is_journal_aborted(journal);
371 struct ext4_journal_cb_entry *jce;
372
373 BUG_ON(txn->t_state == T_FINISHED);
374 spin_lock(&sbi->s_md_lock);
375 while (!list_empty(&txn->t_private_list)) {
376 jce = list_entry(txn->t_private_list.next,
377 struct ext4_journal_cb_entry, jce_list);
378 list_del_init(&jce->jce_list);
379 spin_unlock(&sbi->s_md_lock);
380 jce->jce_func(sb, jce, error);
381 spin_lock(&sbi->s_md_lock);
382 }
383 spin_unlock(&sbi->s_md_lock);
384 }
385
386 /* Deal with the reporting of failure conditions on a filesystem such as
387 * inconsistencies detected or read IO failures.
388 *
389 * On ext2, we can store the error state of the filesystem in the
390 * superblock. That is not possible on ext4, because we may have other
391 * write ordering constraints on the superblock which prevent us from
392 * writing it out straight away; and given that the journal is about to
393 * be aborted, we can't rely on the current, or future, transactions to
394 * write out the superblock safely.
395 *
396 * We'll just use the jbd2_journal_abort() error code to record an error in
397 * the journal instead. On recovery, the journal will complain about
398 * that error until we've noted it down and cleared it.
399 */
400
401 static void ext4_handle_error(struct super_block *sb)
402 {
403 if (sb->s_flags & MS_RDONLY)
404 return;
405
406 if (!test_opt(sb, ERRORS_CONT)) {
407 journal_t *journal = EXT4_SB(sb)->s_journal;
408
409 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
410 if (journal)
411 jbd2_journal_abort(journal, -EIO);
412 }
413 if (test_opt(sb, ERRORS_RO)) {
414 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
415 /*
416 * Make sure updated value of ->s_mount_flags will be visible
417 * before ->s_flags update
418 */
419 smp_wmb();
420 sb->s_flags |= MS_RDONLY;
421 }
422 if (test_opt(sb, ERRORS_PANIC)) {
423 if (EXT4_SB(sb)->s_journal &&
424 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
425 return;
426 panic("EXT4-fs (device %s): panic forced after error\n",
427 sb->s_id);
428 }
429 }
430
431 #define ext4_error_ratelimit(sb) \
432 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
433 "EXT4-fs error")
434
435 void __ext4_error(struct super_block *sb, const char *function,
436 unsigned int line, const char *fmt, ...)
437 {
438 struct va_format vaf;
439 va_list args;
440
441 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
442 return;
443
444 if (ext4_error_ratelimit(sb)) {
445 va_start(args, fmt);
446 vaf.fmt = fmt;
447 vaf.va = &args;
448 printk(KERN_CRIT
449 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
450 sb->s_id, function, line, current->comm, &vaf);
451 va_end(args);
452 }
453 save_error_info(sb, function, line);
454 ext4_handle_error(sb);
455 }
456
457 void __ext4_error_inode(struct inode *inode, const char *function,
458 unsigned int line, ext4_fsblk_t block,
459 const char *fmt, ...)
460 {
461 va_list args;
462 struct va_format vaf;
463 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
464
465 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
466 return;
467
468 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
469 es->s_last_error_block = cpu_to_le64(block);
470 if (ext4_error_ratelimit(inode->i_sb)) {
471 va_start(args, fmt);
472 vaf.fmt = fmt;
473 vaf.va = &args;
474 if (block)
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
476 "inode #%lu: block %llu: comm %s: %pV\n",
477 inode->i_sb->s_id, function, line, inode->i_ino,
478 block, current->comm, &vaf);
479 else
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
481 "inode #%lu: comm %s: %pV\n",
482 inode->i_sb->s_id, function, line, inode->i_ino,
483 current->comm, &vaf);
484 va_end(args);
485 }
486 save_error_info(inode->i_sb, function, line);
487 ext4_handle_error(inode->i_sb);
488 }
489
490 void __ext4_error_file(struct file *file, const char *function,
491 unsigned int line, ext4_fsblk_t block,
492 const char *fmt, ...)
493 {
494 va_list args;
495 struct va_format vaf;
496 struct ext4_super_block *es;
497 struct inode *inode = file_inode(file);
498 char pathname[80], *path;
499
500 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
501 return;
502
503 es = EXT4_SB(inode->i_sb)->s_es;
504 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
505 if (ext4_error_ratelimit(inode->i_sb)) {
506 path = file_path(file, pathname, sizeof(pathname));
507 if (IS_ERR(path))
508 path = "(unknown)";
509 va_start(args, fmt);
510 vaf.fmt = fmt;
511 vaf.va = &args;
512 if (block)
513 printk(KERN_CRIT
514 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
515 "block %llu: comm %s: path %s: %pV\n",
516 inode->i_sb->s_id, function, line, inode->i_ino,
517 block, current->comm, path, &vaf);
518 else
519 printk(KERN_CRIT
520 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
521 "comm %s: path %s: %pV\n",
522 inode->i_sb->s_id, function, line, inode->i_ino,
523 current->comm, path, &vaf);
524 va_end(args);
525 }
526 save_error_info(inode->i_sb, function, line);
527 ext4_handle_error(inode->i_sb);
528 }
529
530 const char *ext4_decode_error(struct super_block *sb, int errno,
531 char nbuf[16])
532 {
533 char *errstr = NULL;
534
535 switch (errno) {
536 case -EFSCORRUPTED:
537 errstr = "Corrupt filesystem";
538 break;
539 case -EFSBADCRC:
540 errstr = "Filesystem failed CRC";
541 break;
542 case -EIO:
543 errstr = "IO failure";
544 break;
545 case -ENOMEM:
546 errstr = "Out of memory";
547 break;
548 case -EROFS:
549 if (!sb || (EXT4_SB(sb)->s_journal &&
550 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
551 errstr = "Journal has aborted";
552 else
553 errstr = "Readonly filesystem";
554 break;
555 default:
556 /* If the caller passed in an extra buffer for unknown
557 * errors, textualise them now. Else we just return
558 * NULL. */
559 if (nbuf) {
560 /* Check for truncated error codes... */
561 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
562 errstr = nbuf;
563 }
564 break;
565 }
566
567 return errstr;
568 }
569
570 /* __ext4_std_error decodes expected errors from journaling functions
571 * automatically and invokes the appropriate error response. */
572
573 void __ext4_std_error(struct super_block *sb, const char *function,
574 unsigned int line, int errno)
575 {
576 char nbuf[16];
577 const char *errstr;
578
579 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
580 return;
581
582 /* Special case: if the error is EROFS, and we're not already
583 * inside a transaction, then there's really no point in logging
584 * an error. */
585 if (errno == -EROFS && journal_current_handle() == NULL &&
586 (sb->s_flags & MS_RDONLY))
587 return;
588
589 if (ext4_error_ratelimit(sb)) {
590 errstr = ext4_decode_error(sb, errno, nbuf);
591 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
592 sb->s_id, function, line, errstr);
593 }
594
595 save_error_info(sb, function, line);
596 ext4_handle_error(sb);
597 }
598
599 /*
600 * ext4_abort is a much stronger failure handler than ext4_error. The
601 * abort function may be used to deal with unrecoverable failures such
602 * as journal IO errors or ENOMEM at a critical moment in log management.
603 *
604 * We unconditionally force the filesystem into an ABORT|READONLY state,
605 * unless the error response on the fs has been set to panic in which
606 * case we take the easy way out and panic immediately.
607 */
608
609 void __ext4_abort(struct super_block *sb, const char *function,
610 unsigned int line, const char *fmt, ...)
611 {
612 struct va_format vaf;
613 va_list args;
614
615 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
616 return;
617
618 save_error_info(sb, function, line);
619 va_start(args, fmt);
620 vaf.fmt = fmt;
621 vaf.va = &args;
622 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
623 sb->s_id, function, line, &vaf);
624 va_end(args);
625
626 if ((sb->s_flags & MS_RDONLY) == 0) {
627 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
628 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
629 /*
630 * Make sure updated value of ->s_mount_flags will be visible
631 * before ->s_flags update
632 */
633 smp_wmb();
634 sb->s_flags |= MS_RDONLY;
635 if (EXT4_SB(sb)->s_journal)
636 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
637 save_error_info(sb, function, line);
638 }
639 if (test_opt(sb, ERRORS_PANIC)) {
640 if (EXT4_SB(sb)->s_journal &&
641 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
642 return;
643 panic("EXT4-fs panic from previous error\n");
644 }
645 }
646
647 void __ext4_msg(struct super_block *sb,
648 const char *prefix, const char *fmt, ...)
649 {
650 struct va_format vaf;
651 va_list args;
652
653 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
654 return;
655
656 va_start(args, fmt);
657 vaf.fmt = fmt;
658 vaf.va = &args;
659 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
660 va_end(args);
661 }
662
663 #define ext4_warning_ratelimit(sb) \
664 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
665 "EXT4-fs warning")
666
667 void __ext4_warning(struct super_block *sb, const char *function,
668 unsigned int line, const char *fmt, ...)
669 {
670 struct va_format vaf;
671 va_list args;
672
673 if (!ext4_warning_ratelimit(sb))
674 return;
675
676 va_start(args, fmt);
677 vaf.fmt = fmt;
678 vaf.va = &args;
679 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
680 sb->s_id, function, line, &vaf);
681 va_end(args);
682 }
683
684 void __ext4_warning_inode(const struct inode *inode, const char *function,
685 unsigned int line, const char *fmt, ...)
686 {
687 struct va_format vaf;
688 va_list args;
689
690 if (!ext4_warning_ratelimit(inode->i_sb))
691 return;
692
693 va_start(args, fmt);
694 vaf.fmt = fmt;
695 vaf.va = &args;
696 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
697 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
698 function, line, inode->i_ino, current->comm, &vaf);
699 va_end(args);
700 }
701
702 void __ext4_grp_locked_error(const char *function, unsigned int line,
703 struct super_block *sb, ext4_group_t grp,
704 unsigned long ino, ext4_fsblk_t block,
705 const char *fmt, ...)
706 __releases(bitlock)
707 __acquires(bitlock)
708 {
709 struct va_format vaf;
710 va_list args;
711 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
712
713 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
714 return;
715
716 es->s_last_error_ino = cpu_to_le32(ino);
717 es->s_last_error_block = cpu_to_le64(block);
718 __save_error_info(sb, function, line);
719
720 if (ext4_error_ratelimit(sb)) {
721 va_start(args, fmt);
722 vaf.fmt = fmt;
723 vaf.va = &args;
724 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
725 sb->s_id, function, line, grp);
726 if (ino)
727 printk(KERN_CONT "inode %lu: ", ino);
728 if (block)
729 printk(KERN_CONT "block %llu:",
730 (unsigned long long) block);
731 printk(KERN_CONT "%pV\n", &vaf);
732 va_end(args);
733 }
734
735 if (test_opt(sb, ERRORS_CONT)) {
736 ext4_commit_super(sb, 0);
737 return;
738 }
739
740 ext4_unlock_group(sb, grp);
741 ext4_handle_error(sb);
742 /*
743 * We only get here in the ERRORS_RO case; relocking the group
744 * may be dangerous, but nothing bad will happen since the
745 * filesystem will have already been marked read/only and the
746 * journal has been aborted. We return 1 as a hint to callers
747 * who might what to use the return value from
748 * ext4_grp_locked_error() to distinguish between the
749 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
750 * aggressively from the ext4 function in question, with a
751 * more appropriate error code.
752 */
753 ext4_lock_group(sb, grp);
754 return;
755 }
756
757 void ext4_update_dynamic_rev(struct super_block *sb)
758 {
759 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
760
761 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
762 return;
763
764 ext4_warning(sb,
765 "updating to rev %d because of new feature flag, "
766 "running e2fsck is recommended",
767 EXT4_DYNAMIC_REV);
768
769 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
770 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
771 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
772 /* leave es->s_feature_*compat flags alone */
773 /* es->s_uuid will be set by e2fsck if empty */
774
775 /*
776 * The rest of the superblock fields should be zero, and if not it
777 * means they are likely already in use, so leave them alone. We
778 * can leave it up to e2fsck to clean up any inconsistencies there.
779 */
780 }
781
782 /*
783 * Open the external journal device
784 */
785 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
786 {
787 struct block_device *bdev;
788 char b[BDEVNAME_SIZE];
789
790 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
791 if (IS_ERR(bdev))
792 goto fail;
793 return bdev;
794
795 fail:
796 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
797 __bdevname(dev, b), PTR_ERR(bdev));
798 return NULL;
799 }
800
801 /*
802 * Release the journal device
803 */
804 static void ext4_blkdev_put(struct block_device *bdev)
805 {
806 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
807 }
808
809 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
810 {
811 struct block_device *bdev;
812 bdev = sbi->journal_bdev;
813 if (bdev) {
814 ext4_blkdev_put(bdev);
815 sbi->journal_bdev = NULL;
816 }
817 }
818
819 static inline struct inode *orphan_list_entry(struct list_head *l)
820 {
821 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
822 }
823
824 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
825 {
826 struct list_head *l;
827
828 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
829 le32_to_cpu(sbi->s_es->s_last_orphan));
830
831 printk(KERN_ERR "sb_info orphan list:\n");
832 list_for_each(l, &sbi->s_orphan) {
833 struct inode *inode = orphan_list_entry(l);
834 printk(KERN_ERR " "
835 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
836 inode->i_sb->s_id, inode->i_ino, inode,
837 inode->i_mode, inode->i_nlink,
838 NEXT_ORPHAN(inode));
839 }
840 }
841
842 static void ext4_put_super(struct super_block *sb)
843 {
844 struct ext4_sb_info *sbi = EXT4_SB(sb);
845 struct ext4_super_block *es = sbi->s_es;
846 int aborted = 0;
847 int i, err;
848
849 ext4_unregister_li_request(sb);
850 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
851
852 flush_workqueue(sbi->rsv_conversion_wq);
853 destroy_workqueue(sbi->rsv_conversion_wq);
854
855 if (sbi->s_journal) {
856 aborted = is_journal_aborted(sbi->s_journal);
857 err = jbd2_journal_destroy(sbi->s_journal);
858 sbi->s_journal = NULL;
859 if ((err < 0) && !aborted)
860 ext4_abort(sb, "Couldn't clean up the journal");
861 }
862
863 ext4_unregister_sysfs(sb);
864 ext4_es_unregister_shrinker(sbi);
865 del_timer_sync(&sbi->s_err_report);
866 ext4_release_system_zone(sb);
867 ext4_mb_release(sb);
868 ext4_ext_release(sb);
869
870 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
871 ext4_clear_feature_journal_needs_recovery(sb);
872 es->s_state = cpu_to_le16(sbi->s_mount_state);
873 }
874 if (!(sb->s_flags & MS_RDONLY))
875 ext4_commit_super(sb, 1);
876
877 for (i = 0; i < sbi->s_gdb_count; i++)
878 brelse(sbi->s_group_desc[i]);
879 kvfree(sbi->s_group_desc);
880 kvfree(sbi->s_flex_groups);
881 percpu_counter_destroy(&sbi->s_freeclusters_counter);
882 percpu_counter_destroy(&sbi->s_freeinodes_counter);
883 percpu_counter_destroy(&sbi->s_dirs_counter);
884 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
885 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
886 #ifdef CONFIG_QUOTA
887 for (i = 0; i < EXT4_MAXQUOTAS; i++)
888 kfree(sbi->s_qf_names[i]);
889 #endif
890
891 /* Debugging code just in case the in-memory inode orphan list
892 * isn't empty. The on-disk one can be non-empty if we've
893 * detected an error and taken the fs readonly, but the
894 * in-memory list had better be clean by this point. */
895 if (!list_empty(&sbi->s_orphan))
896 dump_orphan_list(sb, sbi);
897 J_ASSERT(list_empty(&sbi->s_orphan));
898
899 sync_blockdev(sb->s_bdev);
900 invalidate_bdev(sb->s_bdev);
901 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
902 /*
903 * Invalidate the journal device's buffers. We don't want them
904 * floating about in memory - the physical journal device may
905 * hotswapped, and it breaks the `ro-after' testing code.
906 */
907 sync_blockdev(sbi->journal_bdev);
908 invalidate_bdev(sbi->journal_bdev);
909 ext4_blkdev_remove(sbi);
910 }
911 if (sbi->s_mb_cache) {
912 ext4_xattr_destroy_cache(sbi->s_mb_cache);
913 sbi->s_mb_cache = NULL;
914 }
915 if (sbi->s_mmp_tsk)
916 kthread_stop(sbi->s_mmp_tsk);
917 brelse(sbi->s_sbh);
918 sb->s_fs_info = NULL;
919 /*
920 * Now that we are completely done shutting down the
921 * superblock, we need to actually destroy the kobject.
922 */
923 kobject_put(&sbi->s_kobj);
924 wait_for_completion(&sbi->s_kobj_unregister);
925 if (sbi->s_chksum_driver)
926 crypto_free_shash(sbi->s_chksum_driver);
927 kfree(sbi->s_blockgroup_lock);
928 kfree(sbi);
929 }
930
931 static struct kmem_cache *ext4_inode_cachep;
932
933 /*
934 * Called inside transaction, so use GFP_NOFS
935 */
936 static struct inode *ext4_alloc_inode(struct super_block *sb)
937 {
938 struct ext4_inode_info *ei;
939
940 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
941 if (!ei)
942 return NULL;
943
944 ei->vfs_inode.i_version = 1;
945 spin_lock_init(&ei->i_raw_lock);
946 INIT_LIST_HEAD(&ei->i_prealloc_list);
947 spin_lock_init(&ei->i_prealloc_lock);
948 ext4_es_init_tree(&ei->i_es_tree);
949 rwlock_init(&ei->i_es_lock);
950 INIT_LIST_HEAD(&ei->i_es_list);
951 ei->i_es_all_nr = 0;
952 ei->i_es_shk_nr = 0;
953 ei->i_es_shrink_lblk = 0;
954 ei->i_reserved_data_blocks = 0;
955 ei->i_reserved_meta_blocks = 0;
956 ei->i_allocated_meta_blocks = 0;
957 ei->i_da_metadata_calc_len = 0;
958 ei->i_da_metadata_calc_last_lblock = 0;
959 spin_lock_init(&(ei->i_block_reservation_lock));
960 #ifdef CONFIG_QUOTA
961 ei->i_reserved_quota = 0;
962 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
963 #endif
964 ei->jinode = NULL;
965 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
966 spin_lock_init(&ei->i_completed_io_lock);
967 ei->i_sync_tid = 0;
968 ei->i_datasync_tid = 0;
969 atomic_set(&ei->i_unwritten, 0);
970 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
971 return &ei->vfs_inode;
972 }
973
974 static int ext4_drop_inode(struct inode *inode)
975 {
976 int drop = generic_drop_inode(inode);
977
978 trace_ext4_drop_inode(inode, drop);
979 return drop;
980 }
981
982 static void ext4_i_callback(struct rcu_head *head)
983 {
984 struct inode *inode = container_of(head, struct inode, i_rcu);
985 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
986 }
987
988 static void ext4_destroy_inode(struct inode *inode)
989 {
990 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
991 ext4_msg(inode->i_sb, KERN_ERR,
992 "Inode %lu (%p): orphan list check failed!",
993 inode->i_ino, EXT4_I(inode));
994 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
995 EXT4_I(inode), sizeof(struct ext4_inode_info),
996 true);
997 dump_stack();
998 }
999 call_rcu(&inode->i_rcu, ext4_i_callback);
1000 }
1001
1002 static void init_once(void *foo)
1003 {
1004 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1005
1006 INIT_LIST_HEAD(&ei->i_orphan);
1007 init_rwsem(&ei->xattr_sem);
1008 init_rwsem(&ei->i_data_sem);
1009 init_rwsem(&ei->i_mmap_sem);
1010 inode_init_once(&ei->vfs_inode);
1011 }
1012
1013 static int __init init_inodecache(void)
1014 {
1015 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1016 sizeof(struct ext4_inode_info),
1017 0, (SLAB_RECLAIM_ACCOUNT|
1018 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1019 init_once);
1020 if (ext4_inode_cachep == NULL)
1021 return -ENOMEM;
1022 return 0;
1023 }
1024
1025 static void destroy_inodecache(void)
1026 {
1027 /*
1028 * Make sure all delayed rcu free inodes are flushed before we
1029 * destroy cache.
1030 */
1031 rcu_barrier();
1032 kmem_cache_destroy(ext4_inode_cachep);
1033 }
1034
1035 void ext4_clear_inode(struct inode *inode)
1036 {
1037 invalidate_inode_buffers(inode);
1038 clear_inode(inode);
1039 dquot_drop(inode);
1040 ext4_discard_preallocations(inode);
1041 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1042 if (EXT4_I(inode)->jinode) {
1043 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1044 EXT4_I(inode)->jinode);
1045 jbd2_free_inode(EXT4_I(inode)->jinode);
1046 EXT4_I(inode)->jinode = NULL;
1047 }
1048 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1049 fscrypt_put_encryption_info(inode, NULL);
1050 #endif
1051 }
1052
1053 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1054 u64 ino, u32 generation)
1055 {
1056 struct inode *inode;
1057
1058 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1059 return ERR_PTR(-ESTALE);
1060 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1061 return ERR_PTR(-ESTALE);
1062
1063 /* iget isn't really right if the inode is currently unallocated!!
1064 *
1065 * ext4_read_inode will return a bad_inode if the inode had been
1066 * deleted, so we should be safe.
1067 *
1068 * Currently we don't know the generation for parent directory, so
1069 * a generation of 0 means "accept any"
1070 */
1071 inode = ext4_iget_normal(sb, ino);
1072 if (IS_ERR(inode))
1073 return ERR_CAST(inode);
1074 if (generation && inode->i_generation != generation) {
1075 iput(inode);
1076 return ERR_PTR(-ESTALE);
1077 }
1078
1079 return inode;
1080 }
1081
1082 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1083 int fh_len, int fh_type)
1084 {
1085 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1086 ext4_nfs_get_inode);
1087 }
1088
1089 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1090 int fh_len, int fh_type)
1091 {
1092 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1093 ext4_nfs_get_inode);
1094 }
1095
1096 /*
1097 * Try to release metadata pages (indirect blocks, directories) which are
1098 * mapped via the block device. Since these pages could have journal heads
1099 * which would prevent try_to_free_buffers() from freeing them, we must use
1100 * jbd2 layer's try_to_free_buffers() function to release them.
1101 */
1102 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1103 gfp_t wait)
1104 {
1105 journal_t *journal = EXT4_SB(sb)->s_journal;
1106
1107 WARN_ON(PageChecked(page));
1108 if (!page_has_buffers(page))
1109 return 0;
1110 if (journal)
1111 return jbd2_journal_try_to_free_buffers(journal, page,
1112 wait & ~__GFP_DIRECT_RECLAIM);
1113 return try_to_free_buffers(page);
1114 }
1115
1116 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1117 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1118 {
1119 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1120 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1121 }
1122
1123 static int ext4_prepare_context(struct inode *inode)
1124 {
1125 return ext4_convert_inline_data(inode);
1126 }
1127
1128 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1129 void *fs_data)
1130 {
1131 handle_t *handle = fs_data;
1132 int res, res2, retries = 0;
1133
1134 /*
1135 * If a journal handle was specified, then the encryption context is
1136 * being set on a new inode via inheritance and is part of a larger
1137 * transaction to create the inode. Otherwise the encryption context is
1138 * being set on an existing inode in its own transaction. Only in the
1139 * latter case should the "retry on ENOSPC" logic be used.
1140 */
1141
1142 if (handle) {
1143 res = ext4_xattr_set_handle(handle, inode,
1144 EXT4_XATTR_INDEX_ENCRYPTION,
1145 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1146 ctx, len, 0);
1147 if (!res) {
1148 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1149 ext4_clear_inode_state(inode,
1150 EXT4_STATE_MAY_INLINE_DATA);
1151 /*
1152 * Update inode->i_flags - e.g. S_DAX may get disabled
1153 */
1154 ext4_set_inode_flags(inode);
1155 }
1156 return res;
1157 }
1158
1159 retry:
1160 handle = ext4_journal_start(inode, EXT4_HT_MISC,
1161 ext4_jbd2_credits_xattr(inode));
1162 if (IS_ERR(handle))
1163 return PTR_ERR(handle);
1164
1165 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1166 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1167 ctx, len, 0);
1168 if (!res) {
1169 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1170 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1171 ext4_set_inode_flags(inode);
1172 res = ext4_mark_inode_dirty(handle, inode);
1173 if (res)
1174 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1175 }
1176 res2 = ext4_journal_stop(handle);
1177
1178 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1179 goto retry;
1180 if (!res)
1181 res = res2;
1182 return res;
1183 }
1184
1185 static int ext4_dummy_context(struct inode *inode)
1186 {
1187 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1188 }
1189
1190 static unsigned ext4_max_namelen(struct inode *inode)
1191 {
1192 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1193 EXT4_NAME_LEN;
1194 }
1195
1196 static const struct fscrypt_operations ext4_cryptops = {
1197 .key_prefix = "ext4:",
1198 .get_context = ext4_get_context,
1199 .prepare_context = ext4_prepare_context,
1200 .set_context = ext4_set_context,
1201 .dummy_context = ext4_dummy_context,
1202 .is_encrypted = ext4_encrypted_inode,
1203 .empty_dir = ext4_empty_dir,
1204 .max_namelen = ext4_max_namelen,
1205 };
1206 #else
1207 static const struct fscrypt_operations ext4_cryptops = {
1208 .is_encrypted = ext4_encrypted_inode,
1209 };
1210 #endif
1211
1212 #ifdef CONFIG_QUOTA
1213 static char *quotatypes[] = INITQFNAMES;
1214 #define QTYPE2NAME(t) (quotatypes[t])
1215
1216 static int ext4_write_dquot(struct dquot *dquot);
1217 static int ext4_acquire_dquot(struct dquot *dquot);
1218 static int ext4_release_dquot(struct dquot *dquot);
1219 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1220 static int ext4_write_info(struct super_block *sb, int type);
1221 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1222 const struct path *path);
1223 static int ext4_quota_off(struct super_block *sb, int type);
1224 static int ext4_quota_on_mount(struct super_block *sb, int type);
1225 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1226 size_t len, loff_t off);
1227 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1228 const char *data, size_t len, loff_t off);
1229 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1230 unsigned int flags);
1231 static int ext4_enable_quotas(struct super_block *sb);
1232 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1233
1234 static struct dquot **ext4_get_dquots(struct inode *inode)
1235 {
1236 return EXT4_I(inode)->i_dquot;
1237 }
1238
1239 static const struct dquot_operations ext4_quota_operations = {
1240 .get_reserved_space = ext4_get_reserved_space,
1241 .write_dquot = ext4_write_dquot,
1242 .acquire_dquot = ext4_acquire_dquot,
1243 .release_dquot = ext4_release_dquot,
1244 .mark_dirty = ext4_mark_dquot_dirty,
1245 .write_info = ext4_write_info,
1246 .alloc_dquot = dquot_alloc,
1247 .destroy_dquot = dquot_destroy,
1248 .get_projid = ext4_get_projid,
1249 .get_next_id = ext4_get_next_id,
1250 };
1251
1252 static const struct quotactl_ops ext4_qctl_operations = {
1253 .quota_on = ext4_quota_on,
1254 .quota_off = ext4_quota_off,
1255 .quota_sync = dquot_quota_sync,
1256 .get_state = dquot_get_state,
1257 .set_info = dquot_set_dqinfo,
1258 .get_dqblk = dquot_get_dqblk,
1259 .set_dqblk = dquot_set_dqblk,
1260 .get_nextdqblk = dquot_get_next_dqblk,
1261 };
1262 #endif
1263
1264 static const struct super_operations ext4_sops = {
1265 .alloc_inode = ext4_alloc_inode,
1266 .destroy_inode = ext4_destroy_inode,
1267 .write_inode = ext4_write_inode,
1268 .dirty_inode = ext4_dirty_inode,
1269 .drop_inode = ext4_drop_inode,
1270 .evict_inode = ext4_evict_inode,
1271 .put_super = ext4_put_super,
1272 .sync_fs = ext4_sync_fs,
1273 .freeze_fs = ext4_freeze,
1274 .unfreeze_fs = ext4_unfreeze,
1275 .statfs = ext4_statfs,
1276 .remount_fs = ext4_remount,
1277 .show_options = ext4_show_options,
1278 #ifdef CONFIG_QUOTA
1279 .quota_read = ext4_quota_read,
1280 .quota_write = ext4_quota_write,
1281 .get_dquots = ext4_get_dquots,
1282 #endif
1283 .bdev_try_to_free_page = bdev_try_to_free_page,
1284 };
1285
1286 static const struct export_operations ext4_export_ops = {
1287 .fh_to_dentry = ext4_fh_to_dentry,
1288 .fh_to_parent = ext4_fh_to_parent,
1289 .get_parent = ext4_get_parent,
1290 };
1291
1292 enum {
1293 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1294 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1295 Opt_nouid32, Opt_debug, Opt_removed,
1296 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1297 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1298 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1299 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1300 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1301 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1302 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1303 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1304 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1305 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1306 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1307 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1308 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1309 Opt_inode_readahead_blks, Opt_journal_ioprio,
1310 Opt_dioread_nolock, Opt_dioread_lock,
1311 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1312 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1313 };
1314
1315 static const match_table_t tokens = {
1316 {Opt_bsd_df, "bsddf"},
1317 {Opt_minix_df, "minixdf"},
1318 {Opt_grpid, "grpid"},
1319 {Opt_grpid, "bsdgroups"},
1320 {Opt_nogrpid, "nogrpid"},
1321 {Opt_nogrpid, "sysvgroups"},
1322 {Opt_resgid, "resgid=%u"},
1323 {Opt_resuid, "resuid=%u"},
1324 {Opt_sb, "sb=%u"},
1325 {Opt_err_cont, "errors=continue"},
1326 {Opt_err_panic, "errors=panic"},
1327 {Opt_err_ro, "errors=remount-ro"},
1328 {Opt_nouid32, "nouid32"},
1329 {Opt_debug, "debug"},
1330 {Opt_removed, "oldalloc"},
1331 {Opt_removed, "orlov"},
1332 {Opt_user_xattr, "user_xattr"},
1333 {Opt_nouser_xattr, "nouser_xattr"},
1334 {Opt_acl, "acl"},
1335 {Opt_noacl, "noacl"},
1336 {Opt_noload, "norecovery"},
1337 {Opt_noload, "noload"},
1338 {Opt_removed, "nobh"},
1339 {Opt_removed, "bh"},
1340 {Opt_commit, "commit=%u"},
1341 {Opt_min_batch_time, "min_batch_time=%u"},
1342 {Opt_max_batch_time, "max_batch_time=%u"},
1343 {Opt_journal_dev, "journal_dev=%u"},
1344 {Opt_journal_path, "journal_path=%s"},
1345 {Opt_journal_checksum, "journal_checksum"},
1346 {Opt_nojournal_checksum, "nojournal_checksum"},
1347 {Opt_journal_async_commit, "journal_async_commit"},
1348 {Opt_abort, "abort"},
1349 {Opt_data_journal, "data=journal"},
1350 {Opt_data_ordered, "data=ordered"},
1351 {Opt_data_writeback, "data=writeback"},
1352 {Opt_data_err_abort, "data_err=abort"},
1353 {Opt_data_err_ignore, "data_err=ignore"},
1354 {Opt_offusrjquota, "usrjquota="},
1355 {Opt_usrjquota, "usrjquota=%s"},
1356 {Opt_offgrpjquota, "grpjquota="},
1357 {Opt_grpjquota, "grpjquota=%s"},
1358 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1359 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1360 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1361 {Opt_grpquota, "grpquota"},
1362 {Opt_noquota, "noquota"},
1363 {Opt_quota, "quota"},
1364 {Opt_usrquota, "usrquota"},
1365 {Opt_prjquota, "prjquota"},
1366 {Opt_barrier, "barrier=%u"},
1367 {Opt_barrier, "barrier"},
1368 {Opt_nobarrier, "nobarrier"},
1369 {Opt_i_version, "i_version"},
1370 {Opt_dax, "dax"},
1371 {Opt_stripe, "stripe=%u"},
1372 {Opt_delalloc, "delalloc"},
1373 {Opt_lazytime, "lazytime"},
1374 {Opt_nolazytime, "nolazytime"},
1375 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1376 {Opt_nodelalloc, "nodelalloc"},
1377 {Opt_removed, "mblk_io_submit"},
1378 {Opt_removed, "nomblk_io_submit"},
1379 {Opt_block_validity, "block_validity"},
1380 {Opt_noblock_validity, "noblock_validity"},
1381 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1382 {Opt_journal_ioprio, "journal_ioprio=%u"},
1383 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1384 {Opt_auto_da_alloc, "auto_da_alloc"},
1385 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1386 {Opt_dioread_nolock, "dioread_nolock"},
1387 {Opt_dioread_lock, "dioread_lock"},
1388 {Opt_discard, "discard"},
1389 {Opt_nodiscard, "nodiscard"},
1390 {Opt_init_itable, "init_itable=%u"},
1391 {Opt_init_itable, "init_itable"},
1392 {Opt_noinit_itable, "noinit_itable"},
1393 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1394 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1395 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1396 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1397 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1398 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1399 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1400 {Opt_err, NULL},
1401 };
1402
1403 static ext4_fsblk_t get_sb_block(void **data)
1404 {
1405 ext4_fsblk_t sb_block;
1406 char *options = (char *) *data;
1407
1408 if (!options || strncmp(options, "sb=", 3) != 0)
1409 return 1; /* Default location */
1410
1411 options += 3;
1412 /* TODO: use simple_strtoll with >32bit ext4 */
1413 sb_block = simple_strtoul(options, &options, 0);
1414 if (*options && *options != ',') {
1415 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1416 (char *) *data);
1417 return 1;
1418 }
1419 if (*options == ',')
1420 options++;
1421 *data = (void *) options;
1422
1423 return sb_block;
1424 }
1425
1426 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1427 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1428 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1429
1430 #ifdef CONFIG_QUOTA
1431 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1432 {
1433 struct ext4_sb_info *sbi = EXT4_SB(sb);
1434 char *qname;
1435 int ret = -1;
1436
1437 if (sb_any_quota_loaded(sb) &&
1438 !sbi->s_qf_names[qtype]) {
1439 ext4_msg(sb, KERN_ERR,
1440 "Cannot change journaled "
1441 "quota options when quota turned on");
1442 return -1;
1443 }
1444 if (ext4_has_feature_quota(sb)) {
1445 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1446 "ignored when QUOTA feature is enabled");
1447 return 1;
1448 }
1449 qname = match_strdup(args);
1450 if (!qname) {
1451 ext4_msg(sb, KERN_ERR,
1452 "Not enough memory for storing quotafile name");
1453 return -1;
1454 }
1455 if (sbi->s_qf_names[qtype]) {
1456 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1457 ret = 1;
1458 else
1459 ext4_msg(sb, KERN_ERR,
1460 "%s quota file already specified",
1461 QTYPE2NAME(qtype));
1462 goto errout;
1463 }
1464 if (strchr(qname, '/')) {
1465 ext4_msg(sb, KERN_ERR,
1466 "quotafile must be on filesystem root");
1467 goto errout;
1468 }
1469 sbi->s_qf_names[qtype] = qname;
1470 set_opt(sb, QUOTA);
1471 return 1;
1472 errout:
1473 kfree(qname);
1474 return ret;
1475 }
1476
1477 static int clear_qf_name(struct super_block *sb, int qtype)
1478 {
1479
1480 struct ext4_sb_info *sbi = EXT4_SB(sb);
1481
1482 if (sb_any_quota_loaded(sb) &&
1483 sbi->s_qf_names[qtype]) {
1484 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1485 " when quota turned on");
1486 return -1;
1487 }
1488 kfree(sbi->s_qf_names[qtype]);
1489 sbi->s_qf_names[qtype] = NULL;
1490 return 1;
1491 }
1492 #endif
1493
1494 #define MOPT_SET 0x0001
1495 #define MOPT_CLEAR 0x0002
1496 #define MOPT_NOSUPPORT 0x0004
1497 #define MOPT_EXPLICIT 0x0008
1498 #define MOPT_CLEAR_ERR 0x0010
1499 #define MOPT_GTE0 0x0020
1500 #ifdef CONFIG_QUOTA
1501 #define MOPT_Q 0
1502 #define MOPT_QFMT 0x0040
1503 #else
1504 #define MOPT_Q MOPT_NOSUPPORT
1505 #define MOPT_QFMT MOPT_NOSUPPORT
1506 #endif
1507 #define MOPT_DATAJ 0x0080
1508 #define MOPT_NO_EXT2 0x0100
1509 #define MOPT_NO_EXT3 0x0200
1510 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1511 #define MOPT_STRING 0x0400
1512
1513 static const struct mount_opts {
1514 int token;
1515 int mount_opt;
1516 int flags;
1517 } ext4_mount_opts[] = {
1518 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1519 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1520 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1521 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1522 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1523 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1524 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1525 MOPT_EXT4_ONLY | MOPT_SET},
1526 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1527 MOPT_EXT4_ONLY | MOPT_CLEAR},
1528 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1529 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1530 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1531 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1532 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1533 MOPT_EXT4_ONLY | MOPT_CLEAR},
1534 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1535 MOPT_EXT4_ONLY | MOPT_CLEAR},
1536 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1537 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1538 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1539 EXT4_MOUNT_JOURNAL_CHECKSUM),
1540 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1541 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1542 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1543 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1544 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1545 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1546 MOPT_NO_EXT2},
1547 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1548 MOPT_NO_EXT2},
1549 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1550 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1551 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1552 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1553 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1554 {Opt_commit, 0, MOPT_GTE0},
1555 {Opt_max_batch_time, 0, MOPT_GTE0},
1556 {Opt_min_batch_time, 0, MOPT_GTE0},
1557 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1558 {Opt_init_itable, 0, MOPT_GTE0},
1559 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1560 {Opt_stripe, 0, MOPT_GTE0},
1561 {Opt_resuid, 0, MOPT_GTE0},
1562 {Opt_resgid, 0, MOPT_GTE0},
1563 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1564 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1565 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1566 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1567 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1568 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1569 MOPT_NO_EXT2 | MOPT_DATAJ},
1570 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1571 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1572 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1573 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1574 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1575 #else
1576 {Opt_acl, 0, MOPT_NOSUPPORT},
1577 {Opt_noacl, 0, MOPT_NOSUPPORT},
1578 #endif
1579 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1580 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1581 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1582 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1583 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1584 MOPT_SET | MOPT_Q},
1585 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1586 MOPT_SET | MOPT_Q},
1587 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1588 MOPT_SET | MOPT_Q},
1589 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1590 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1591 MOPT_CLEAR | MOPT_Q},
1592 {Opt_usrjquota, 0, MOPT_Q},
1593 {Opt_grpjquota, 0, MOPT_Q},
1594 {Opt_offusrjquota, 0, MOPT_Q},
1595 {Opt_offgrpjquota, 0, MOPT_Q},
1596 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1597 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1598 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1599 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1600 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1601 {Opt_err, 0, 0}
1602 };
1603
1604 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1605 substring_t *args, unsigned long *journal_devnum,
1606 unsigned int *journal_ioprio, int is_remount)
1607 {
1608 struct ext4_sb_info *sbi = EXT4_SB(sb);
1609 const struct mount_opts *m;
1610 kuid_t uid;
1611 kgid_t gid;
1612 int arg = 0;
1613
1614 #ifdef CONFIG_QUOTA
1615 if (token == Opt_usrjquota)
1616 return set_qf_name(sb, USRQUOTA, &args[0]);
1617 else if (token == Opt_grpjquota)
1618 return set_qf_name(sb, GRPQUOTA, &args[0]);
1619 else if (token == Opt_offusrjquota)
1620 return clear_qf_name(sb, USRQUOTA);
1621 else if (token == Opt_offgrpjquota)
1622 return clear_qf_name(sb, GRPQUOTA);
1623 #endif
1624 switch (token) {
1625 case Opt_noacl:
1626 case Opt_nouser_xattr:
1627 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1628 break;
1629 case Opt_sb:
1630 return 1; /* handled by get_sb_block() */
1631 case Opt_removed:
1632 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1633 return 1;
1634 case Opt_abort:
1635 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1636 return 1;
1637 case Opt_i_version:
1638 sb->s_flags |= MS_I_VERSION;
1639 return 1;
1640 case Opt_lazytime:
1641 sb->s_flags |= MS_LAZYTIME;
1642 return 1;
1643 case Opt_nolazytime:
1644 sb->s_flags &= ~MS_LAZYTIME;
1645 return 1;
1646 }
1647
1648 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1649 if (token == m->token)
1650 break;
1651
1652 if (m->token == Opt_err) {
1653 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1654 "or missing value", opt);
1655 return -1;
1656 }
1657
1658 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1659 ext4_msg(sb, KERN_ERR,
1660 "Mount option \"%s\" incompatible with ext2", opt);
1661 return -1;
1662 }
1663 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1664 ext4_msg(sb, KERN_ERR,
1665 "Mount option \"%s\" incompatible with ext3", opt);
1666 return -1;
1667 }
1668
1669 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1670 return -1;
1671 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1672 return -1;
1673 if (m->flags & MOPT_EXPLICIT) {
1674 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1675 set_opt2(sb, EXPLICIT_DELALLOC);
1676 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1677 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1678 } else
1679 return -1;
1680 }
1681 if (m->flags & MOPT_CLEAR_ERR)
1682 clear_opt(sb, ERRORS_MASK);
1683 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1684 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1685 "options when quota turned on");
1686 return -1;
1687 }
1688
1689 if (m->flags & MOPT_NOSUPPORT) {
1690 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1691 } else if (token == Opt_commit) {
1692 if (arg == 0)
1693 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1694 sbi->s_commit_interval = HZ * arg;
1695 } else if (token == Opt_debug_want_extra_isize) {
1696 sbi->s_want_extra_isize = arg;
1697 } else if (token == Opt_max_batch_time) {
1698 sbi->s_max_batch_time = arg;
1699 } else if (token == Opt_min_batch_time) {
1700 sbi->s_min_batch_time = arg;
1701 } else if (token == Opt_inode_readahead_blks) {
1702 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1703 ext4_msg(sb, KERN_ERR,
1704 "EXT4-fs: inode_readahead_blks must be "
1705 "0 or a power of 2 smaller than 2^31");
1706 return -1;
1707 }
1708 sbi->s_inode_readahead_blks = arg;
1709 } else if (token == Opt_init_itable) {
1710 set_opt(sb, INIT_INODE_TABLE);
1711 if (!args->from)
1712 arg = EXT4_DEF_LI_WAIT_MULT;
1713 sbi->s_li_wait_mult = arg;
1714 } else if (token == Opt_max_dir_size_kb) {
1715 sbi->s_max_dir_size_kb = arg;
1716 } else if (token == Opt_stripe) {
1717 sbi->s_stripe = arg;
1718 } else if (token == Opt_resuid) {
1719 uid = make_kuid(current_user_ns(), arg);
1720 if (!uid_valid(uid)) {
1721 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1722 return -1;
1723 }
1724 sbi->s_resuid = uid;
1725 } else if (token == Opt_resgid) {
1726 gid = make_kgid(current_user_ns(), arg);
1727 if (!gid_valid(gid)) {
1728 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1729 return -1;
1730 }
1731 sbi->s_resgid = gid;
1732 } else if (token == Opt_journal_dev) {
1733 if (is_remount) {
1734 ext4_msg(sb, KERN_ERR,
1735 "Cannot specify journal on remount");
1736 return -1;
1737 }
1738 *journal_devnum = arg;
1739 } else if (token == Opt_journal_path) {
1740 char *journal_path;
1741 struct inode *journal_inode;
1742 struct path path;
1743 int error;
1744
1745 if (is_remount) {
1746 ext4_msg(sb, KERN_ERR,
1747 "Cannot specify journal on remount");
1748 return -1;
1749 }
1750 journal_path = match_strdup(&args[0]);
1751 if (!journal_path) {
1752 ext4_msg(sb, KERN_ERR, "error: could not dup "
1753 "journal device string");
1754 return -1;
1755 }
1756
1757 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1758 if (error) {
1759 ext4_msg(sb, KERN_ERR, "error: could not find "
1760 "journal device path: error %d", error);
1761 kfree(journal_path);
1762 return -1;
1763 }
1764
1765 journal_inode = d_inode(path.dentry);
1766 if (!S_ISBLK(journal_inode->i_mode)) {
1767 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1768 "is not a block device", journal_path);
1769 path_put(&path);
1770 kfree(journal_path);
1771 return -1;
1772 }
1773
1774 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1775 path_put(&path);
1776 kfree(journal_path);
1777 } else if (token == Opt_journal_ioprio) {
1778 if (arg > 7) {
1779 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1780 " (must be 0-7)");
1781 return -1;
1782 }
1783 *journal_ioprio =
1784 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1785 } else if (token == Opt_test_dummy_encryption) {
1786 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1787 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1788 ext4_msg(sb, KERN_WARNING,
1789 "Test dummy encryption mode enabled");
1790 #else
1791 ext4_msg(sb, KERN_WARNING,
1792 "Test dummy encryption mount option ignored");
1793 #endif
1794 } else if (m->flags & MOPT_DATAJ) {
1795 if (is_remount) {
1796 if (!sbi->s_journal)
1797 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1798 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1799 ext4_msg(sb, KERN_ERR,
1800 "Cannot change data mode on remount");
1801 return -1;
1802 }
1803 } else {
1804 clear_opt(sb, DATA_FLAGS);
1805 sbi->s_mount_opt |= m->mount_opt;
1806 }
1807 #ifdef CONFIG_QUOTA
1808 } else if (m->flags & MOPT_QFMT) {
1809 if (sb_any_quota_loaded(sb) &&
1810 sbi->s_jquota_fmt != m->mount_opt) {
1811 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1812 "quota options when quota turned on");
1813 return -1;
1814 }
1815 if (ext4_has_feature_quota(sb)) {
1816 ext4_msg(sb, KERN_INFO,
1817 "Quota format mount options ignored "
1818 "when QUOTA feature is enabled");
1819 return 1;
1820 }
1821 sbi->s_jquota_fmt = m->mount_opt;
1822 #endif
1823 } else if (token == Opt_dax) {
1824 #ifdef CONFIG_FS_DAX
1825 ext4_msg(sb, KERN_WARNING,
1826 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1827 sbi->s_mount_opt |= m->mount_opt;
1828 #else
1829 ext4_msg(sb, KERN_INFO, "dax option not supported");
1830 return -1;
1831 #endif
1832 } else if (token == Opt_data_err_abort) {
1833 sbi->s_mount_opt |= m->mount_opt;
1834 } else if (token == Opt_data_err_ignore) {
1835 sbi->s_mount_opt &= ~m->mount_opt;
1836 } else {
1837 if (!args->from)
1838 arg = 1;
1839 if (m->flags & MOPT_CLEAR)
1840 arg = !arg;
1841 else if (unlikely(!(m->flags & MOPT_SET))) {
1842 ext4_msg(sb, KERN_WARNING,
1843 "buggy handling of option %s", opt);
1844 WARN_ON(1);
1845 return -1;
1846 }
1847 if (arg != 0)
1848 sbi->s_mount_opt |= m->mount_opt;
1849 else
1850 sbi->s_mount_opt &= ~m->mount_opt;
1851 }
1852 return 1;
1853 }
1854
1855 static int parse_options(char *options, struct super_block *sb,
1856 unsigned long *journal_devnum,
1857 unsigned int *journal_ioprio,
1858 int is_remount)
1859 {
1860 struct ext4_sb_info *sbi = EXT4_SB(sb);
1861 char *p;
1862 substring_t args[MAX_OPT_ARGS];
1863 int token;
1864
1865 if (!options)
1866 return 1;
1867
1868 while ((p = strsep(&options, ",")) != NULL) {
1869 if (!*p)
1870 continue;
1871 /*
1872 * Initialize args struct so we know whether arg was
1873 * found; some options take optional arguments.
1874 */
1875 args[0].to = args[0].from = NULL;
1876 token = match_token(p, tokens, args);
1877 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1878 journal_ioprio, is_remount) < 0)
1879 return 0;
1880 }
1881 #ifdef CONFIG_QUOTA
1882 /*
1883 * We do the test below only for project quotas. 'usrquota' and
1884 * 'grpquota' mount options are allowed even without quota feature
1885 * to support legacy quotas in quota files.
1886 */
1887 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1888 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1889 "Cannot enable project quota enforcement.");
1890 return 0;
1891 }
1892 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1893 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1894 clear_opt(sb, USRQUOTA);
1895
1896 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1897 clear_opt(sb, GRPQUOTA);
1898
1899 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1900 ext4_msg(sb, KERN_ERR, "old and new quota "
1901 "format mixing");
1902 return 0;
1903 }
1904
1905 if (!sbi->s_jquota_fmt) {
1906 ext4_msg(sb, KERN_ERR, "journaled quota format "
1907 "not specified");
1908 return 0;
1909 }
1910 }
1911 #endif
1912 if (test_opt(sb, DIOREAD_NOLOCK)) {
1913 int blocksize =
1914 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1915
1916 if (blocksize < PAGE_SIZE) {
1917 ext4_msg(sb, KERN_ERR, "can't mount with "
1918 "dioread_nolock if block size != PAGE_SIZE");
1919 return 0;
1920 }
1921 }
1922 return 1;
1923 }
1924
1925 static inline void ext4_show_quota_options(struct seq_file *seq,
1926 struct super_block *sb)
1927 {
1928 #if defined(CONFIG_QUOTA)
1929 struct ext4_sb_info *sbi = EXT4_SB(sb);
1930
1931 if (sbi->s_jquota_fmt) {
1932 char *fmtname = "";
1933
1934 switch (sbi->s_jquota_fmt) {
1935 case QFMT_VFS_OLD:
1936 fmtname = "vfsold";
1937 break;
1938 case QFMT_VFS_V0:
1939 fmtname = "vfsv0";
1940 break;
1941 case QFMT_VFS_V1:
1942 fmtname = "vfsv1";
1943 break;
1944 }
1945 seq_printf(seq, ",jqfmt=%s", fmtname);
1946 }
1947
1948 if (sbi->s_qf_names[USRQUOTA])
1949 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1950
1951 if (sbi->s_qf_names[GRPQUOTA])
1952 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1953 #endif
1954 }
1955
1956 static const char *token2str(int token)
1957 {
1958 const struct match_token *t;
1959
1960 for (t = tokens; t->token != Opt_err; t++)
1961 if (t->token == token && !strchr(t->pattern, '='))
1962 break;
1963 return t->pattern;
1964 }
1965
1966 /*
1967 * Show an option if
1968 * - it's set to a non-default value OR
1969 * - if the per-sb default is different from the global default
1970 */
1971 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1972 int nodefs)
1973 {
1974 struct ext4_sb_info *sbi = EXT4_SB(sb);
1975 struct ext4_super_block *es = sbi->s_es;
1976 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1977 const struct mount_opts *m;
1978 char sep = nodefs ? '\n' : ',';
1979
1980 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1981 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1982
1983 if (sbi->s_sb_block != 1)
1984 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1985
1986 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1987 int want_set = m->flags & MOPT_SET;
1988 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1989 (m->flags & MOPT_CLEAR_ERR))
1990 continue;
1991 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1992 continue; /* skip if same as the default */
1993 if ((want_set &&
1994 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1995 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1996 continue; /* select Opt_noFoo vs Opt_Foo */
1997 SEQ_OPTS_PRINT("%s", token2str(m->token));
1998 }
1999
2000 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2001 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2002 SEQ_OPTS_PRINT("resuid=%u",
2003 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2004 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2005 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2006 SEQ_OPTS_PRINT("resgid=%u",
2007 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2008 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2009 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2010 SEQ_OPTS_PUTS("errors=remount-ro");
2011 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2012 SEQ_OPTS_PUTS("errors=continue");
2013 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2014 SEQ_OPTS_PUTS("errors=panic");
2015 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2016 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2017 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2018 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2019 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2020 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2021 if (sb->s_flags & MS_I_VERSION)
2022 SEQ_OPTS_PUTS("i_version");
2023 if (nodefs || sbi->s_stripe)
2024 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2025 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2026 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2027 SEQ_OPTS_PUTS("data=journal");
2028 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2029 SEQ_OPTS_PUTS("data=ordered");
2030 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2031 SEQ_OPTS_PUTS("data=writeback");
2032 }
2033 if (nodefs ||
2034 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2035 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2036 sbi->s_inode_readahead_blks);
2037
2038 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2039 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2040 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2041 if (nodefs || sbi->s_max_dir_size_kb)
2042 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2043 if (test_opt(sb, DATA_ERR_ABORT))
2044 SEQ_OPTS_PUTS("data_err=abort");
2045
2046 ext4_show_quota_options(seq, sb);
2047 return 0;
2048 }
2049
2050 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2051 {
2052 return _ext4_show_options(seq, root->d_sb, 0);
2053 }
2054
2055 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2056 {
2057 struct super_block *sb = seq->private;
2058 int rc;
2059
2060 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2061 rc = _ext4_show_options(seq, sb, 1);
2062 seq_puts(seq, "\n");
2063 return rc;
2064 }
2065
2066 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2067 int read_only)
2068 {
2069 struct ext4_sb_info *sbi = EXT4_SB(sb);
2070 int res = 0;
2071
2072 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2073 ext4_msg(sb, KERN_ERR, "revision level too high, "
2074 "forcing read-only mode");
2075 res = MS_RDONLY;
2076 }
2077 if (read_only)
2078 goto done;
2079 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2080 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2081 "running e2fsck is recommended");
2082 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2083 ext4_msg(sb, KERN_WARNING,
2084 "warning: mounting fs with errors, "
2085 "running e2fsck is recommended");
2086 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2087 le16_to_cpu(es->s_mnt_count) >=
2088 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2089 ext4_msg(sb, KERN_WARNING,
2090 "warning: maximal mount count reached, "
2091 "running e2fsck is recommended");
2092 else if (le32_to_cpu(es->s_checkinterval) &&
2093 (le32_to_cpu(es->s_lastcheck) +
2094 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2095 ext4_msg(sb, KERN_WARNING,
2096 "warning: checktime reached, "
2097 "running e2fsck is recommended");
2098 if (!sbi->s_journal)
2099 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2100 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2101 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2102 le16_add_cpu(&es->s_mnt_count, 1);
2103 es->s_mtime = cpu_to_le32(get_seconds());
2104 ext4_update_dynamic_rev(sb);
2105 if (sbi->s_journal)
2106 ext4_set_feature_journal_needs_recovery(sb);
2107
2108 ext4_commit_super(sb, 1);
2109 done:
2110 if (test_opt(sb, DEBUG))
2111 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2112 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2113 sb->s_blocksize,
2114 sbi->s_groups_count,
2115 EXT4_BLOCKS_PER_GROUP(sb),
2116 EXT4_INODES_PER_GROUP(sb),
2117 sbi->s_mount_opt, sbi->s_mount_opt2);
2118
2119 cleancache_init_fs(sb);
2120 return res;
2121 }
2122
2123 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2124 {
2125 struct ext4_sb_info *sbi = EXT4_SB(sb);
2126 struct flex_groups *new_groups;
2127 int size;
2128
2129 if (!sbi->s_log_groups_per_flex)
2130 return 0;
2131
2132 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2133 if (size <= sbi->s_flex_groups_allocated)
2134 return 0;
2135
2136 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2137 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2138 if (!new_groups) {
2139 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2140 size / (int) sizeof(struct flex_groups));
2141 return -ENOMEM;
2142 }
2143
2144 if (sbi->s_flex_groups) {
2145 memcpy(new_groups, sbi->s_flex_groups,
2146 (sbi->s_flex_groups_allocated *
2147 sizeof(struct flex_groups)));
2148 kvfree(sbi->s_flex_groups);
2149 }
2150 sbi->s_flex_groups = new_groups;
2151 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2152 return 0;
2153 }
2154
2155 static int ext4_fill_flex_info(struct super_block *sb)
2156 {
2157 struct ext4_sb_info *sbi = EXT4_SB(sb);
2158 struct ext4_group_desc *gdp = NULL;
2159 ext4_group_t flex_group;
2160 int i, err;
2161
2162 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2163 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2164 sbi->s_log_groups_per_flex = 0;
2165 return 1;
2166 }
2167
2168 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2169 if (err)
2170 goto failed;
2171
2172 for (i = 0; i < sbi->s_groups_count; i++) {
2173 gdp = ext4_get_group_desc(sb, i, NULL);
2174
2175 flex_group = ext4_flex_group(sbi, i);
2176 atomic_add(ext4_free_inodes_count(sb, gdp),
2177 &sbi->s_flex_groups[flex_group].free_inodes);
2178 atomic64_add(ext4_free_group_clusters(sb, gdp),
2179 &sbi->s_flex_groups[flex_group].free_clusters);
2180 atomic_add(ext4_used_dirs_count(sb, gdp),
2181 &sbi->s_flex_groups[flex_group].used_dirs);
2182 }
2183
2184 return 1;
2185 failed:
2186 return 0;
2187 }
2188
2189 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2190 struct ext4_group_desc *gdp)
2191 {
2192 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2193 __u16 crc = 0;
2194 __le32 le_group = cpu_to_le32(block_group);
2195 struct ext4_sb_info *sbi = EXT4_SB(sb);
2196
2197 if (ext4_has_metadata_csum(sbi->s_sb)) {
2198 /* Use new metadata_csum algorithm */
2199 __u32 csum32;
2200 __u16 dummy_csum = 0;
2201
2202 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2203 sizeof(le_group));
2204 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2205 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2206 sizeof(dummy_csum));
2207 offset += sizeof(dummy_csum);
2208 if (offset < sbi->s_desc_size)
2209 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2210 sbi->s_desc_size - offset);
2211
2212 crc = csum32 & 0xFFFF;
2213 goto out;
2214 }
2215
2216 /* old crc16 code */
2217 if (!ext4_has_feature_gdt_csum(sb))
2218 return 0;
2219
2220 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2221 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2222 crc = crc16(crc, (__u8 *)gdp, offset);
2223 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2224 /* for checksum of struct ext4_group_desc do the rest...*/
2225 if (ext4_has_feature_64bit(sb) &&
2226 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2227 crc = crc16(crc, (__u8 *)gdp + offset,
2228 le16_to_cpu(sbi->s_es->s_desc_size) -
2229 offset);
2230
2231 out:
2232 return cpu_to_le16(crc);
2233 }
2234
2235 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2236 struct ext4_group_desc *gdp)
2237 {
2238 if (ext4_has_group_desc_csum(sb) &&
2239 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2240 return 0;
2241
2242 return 1;
2243 }
2244
2245 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2246 struct ext4_group_desc *gdp)
2247 {
2248 if (!ext4_has_group_desc_csum(sb))
2249 return;
2250 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2251 }
2252
2253 /* Called at mount-time, super-block is locked */
2254 static int ext4_check_descriptors(struct super_block *sb,
2255 ext4_fsblk_t sb_block,
2256 ext4_group_t *first_not_zeroed)
2257 {
2258 struct ext4_sb_info *sbi = EXT4_SB(sb);
2259 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2260 ext4_fsblk_t last_block;
2261 ext4_fsblk_t block_bitmap;
2262 ext4_fsblk_t inode_bitmap;
2263 ext4_fsblk_t inode_table;
2264 int flexbg_flag = 0;
2265 ext4_group_t i, grp = sbi->s_groups_count;
2266
2267 if (ext4_has_feature_flex_bg(sb))
2268 flexbg_flag = 1;
2269
2270 ext4_debug("Checking group descriptors");
2271
2272 for (i = 0; i < sbi->s_groups_count; i++) {
2273 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2274
2275 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2276 last_block = ext4_blocks_count(sbi->s_es) - 1;
2277 else
2278 last_block = first_block +
2279 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2280
2281 if ((grp == sbi->s_groups_count) &&
2282 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2283 grp = i;
2284
2285 block_bitmap = ext4_block_bitmap(sb, gdp);
2286 if (block_bitmap == sb_block) {
2287 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2288 "Block bitmap for group %u overlaps "
2289 "superblock", i);
2290 }
2291 if (block_bitmap < first_block || block_bitmap > last_block) {
2292 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2293 "Block bitmap for group %u not in group "
2294 "(block %llu)!", i, block_bitmap);
2295 return 0;
2296 }
2297 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2298 if (inode_bitmap == sb_block) {
2299 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2300 "Inode bitmap for group %u overlaps "
2301 "superblock", i);
2302 }
2303 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2304 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2305 "Inode bitmap for group %u not in group "
2306 "(block %llu)!", i, inode_bitmap);
2307 return 0;
2308 }
2309 inode_table = ext4_inode_table(sb, gdp);
2310 if (inode_table == sb_block) {
2311 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2312 "Inode table for group %u overlaps "
2313 "superblock", i);
2314 }
2315 if (inode_table < first_block ||
2316 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2317 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2318 "Inode table for group %u not in group "
2319 "(block %llu)!", i, inode_table);
2320 return 0;
2321 }
2322 ext4_lock_group(sb, i);
2323 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2324 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2325 "Checksum for group %u failed (%u!=%u)",
2326 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2327 gdp)), le16_to_cpu(gdp->bg_checksum));
2328 if (!(sb->s_flags & MS_RDONLY)) {
2329 ext4_unlock_group(sb, i);
2330 return 0;
2331 }
2332 }
2333 ext4_unlock_group(sb, i);
2334 if (!flexbg_flag)
2335 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2336 }
2337 if (NULL != first_not_zeroed)
2338 *first_not_zeroed = grp;
2339 return 1;
2340 }
2341
2342 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2343 * the superblock) which were deleted from all directories, but held open by
2344 * a process at the time of a crash. We walk the list and try to delete these
2345 * inodes at recovery time (only with a read-write filesystem).
2346 *
2347 * In order to keep the orphan inode chain consistent during traversal (in
2348 * case of crash during recovery), we link each inode into the superblock
2349 * orphan list_head and handle it the same way as an inode deletion during
2350 * normal operation (which journals the operations for us).
2351 *
2352 * We only do an iget() and an iput() on each inode, which is very safe if we
2353 * accidentally point at an in-use or already deleted inode. The worst that
2354 * can happen in this case is that we get a "bit already cleared" message from
2355 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2356 * e2fsck was run on this filesystem, and it must have already done the orphan
2357 * inode cleanup for us, so we can safely abort without any further action.
2358 */
2359 static void ext4_orphan_cleanup(struct super_block *sb,
2360 struct ext4_super_block *es)
2361 {
2362 unsigned int s_flags = sb->s_flags;
2363 int ret, nr_orphans = 0, nr_truncates = 0;
2364 #ifdef CONFIG_QUOTA
2365 int i;
2366 #endif
2367 if (!es->s_last_orphan) {
2368 jbd_debug(4, "no orphan inodes to clean up\n");
2369 return;
2370 }
2371
2372 if (bdev_read_only(sb->s_bdev)) {
2373 ext4_msg(sb, KERN_ERR, "write access "
2374 "unavailable, skipping orphan cleanup");
2375 return;
2376 }
2377
2378 /* Check if feature set would not allow a r/w mount */
2379 if (!ext4_feature_set_ok(sb, 0)) {
2380 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2381 "unknown ROCOMPAT features");
2382 return;
2383 }
2384
2385 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2386 /* don't clear list on RO mount w/ errors */
2387 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2388 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2389 "clearing orphan list.\n");
2390 es->s_last_orphan = 0;
2391 }
2392 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2393 return;
2394 }
2395
2396 if (s_flags & MS_RDONLY) {
2397 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2398 sb->s_flags &= ~MS_RDONLY;
2399 }
2400 #ifdef CONFIG_QUOTA
2401 /* Needed for iput() to work correctly and not trash data */
2402 sb->s_flags |= MS_ACTIVE;
2403 /* Turn on quotas so that they are updated correctly */
2404 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2405 if (EXT4_SB(sb)->s_qf_names[i]) {
2406 int ret = ext4_quota_on_mount(sb, i);
2407 if (ret < 0)
2408 ext4_msg(sb, KERN_ERR,
2409 "Cannot turn on journaled "
2410 "quota: error %d", ret);
2411 }
2412 }
2413 #endif
2414
2415 while (es->s_last_orphan) {
2416 struct inode *inode;
2417
2418 /*
2419 * We may have encountered an error during cleanup; if
2420 * so, skip the rest.
2421 */
2422 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2423 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2424 es->s_last_orphan = 0;
2425 break;
2426 }
2427
2428 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2429 if (IS_ERR(inode)) {
2430 es->s_last_orphan = 0;
2431 break;
2432 }
2433
2434 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2435 dquot_initialize(inode);
2436 if (inode->i_nlink) {
2437 if (test_opt(sb, DEBUG))
2438 ext4_msg(sb, KERN_DEBUG,
2439 "%s: truncating inode %lu to %lld bytes",
2440 __func__, inode->i_ino, inode->i_size);
2441 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2442 inode->i_ino, inode->i_size);
2443 inode_lock(inode);
2444 truncate_inode_pages(inode->i_mapping, inode->i_size);
2445 ret = ext4_truncate(inode);
2446 if (ret)
2447 ext4_std_error(inode->i_sb, ret);
2448 inode_unlock(inode);
2449 nr_truncates++;
2450 } else {
2451 if (test_opt(sb, DEBUG))
2452 ext4_msg(sb, KERN_DEBUG,
2453 "%s: deleting unreferenced inode %lu",
2454 __func__, inode->i_ino);
2455 jbd_debug(2, "deleting unreferenced inode %lu\n",
2456 inode->i_ino);
2457 nr_orphans++;
2458 }
2459 iput(inode); /* The delete magic happens here! */
2460 }
2461
2462 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2463
2464 if (nr_orphans)
2465 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2466 PLURAL(nr_orphans));
2467 if (nr_truncates)
2468 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2469 PLURAL(nr_truncates));
2470 #ifdef CONFIG_QUOTA
2471 /* Turn quotas off */
2472 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2473 if (sb_dqopt(sb)->files[i])
2474 dquot_quota_off(sb, i);
2475 }
2476 #endif
2477 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2478 }
2479
2480 /*
2481 * Maximal extent format file size.
2482 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2483 * extent format containers, within a sector_t, and within i_blocks
2484 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2485 * so that won't be a limiting factor.
2486 *
2487 * However there is other limiting factor. We do store extents in the form
2488 * of starting block and length, hence the resulting length of the extent
2489 * covering maximum file size must fit into on-disk format containers as
2490 * well. Given that length is always by 1 unit bigger than max unit (because
2491 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2492 *
2493 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2494 */
2495 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2496 {
2497 loff_t res;
2498 loff_t upper_limit = MAX_LFS_FILESIZE;
2499
2500 /* small i_blocks in vfs inode? */
2501 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2502 /*
2503 * CONFIG_LBDAF is not enabled implies the inode
2504 * i_block represent total blocks in 512 bytes
2505 * 32 == size of vfs inode i_blocks * 8
2506 */
2507 upper_limit = (1LL << 32) - 1;
2508
2509 /* total blocks in file system block size */
2510 upper_limit >>= (blkbits - 9);
2511 upper_limit <<= blkbits;
2512 }
2513
2514 /*
2515 * 32-bit extent-start container, ee_block. We lower the maxbytes
2516 * by one fs block, so ee_len can cover the extent of maximum file
2517 * size
2518 */
2519 res = (1LL << 32) - 1;
2520 res <<= blkbits;
2521
2522 /* Sanity check against vm- & vfs- imposed limits */
2523 if (res > upper_limit)
2524 res = upper_limit;
2525
2526 return res;
2527 }
2528
2529 /*
2530 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2531 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2532 * We need to be 1 filesystem block less than the 2^48 sector limit.
2533 */
2534 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2535 {
2536 loff_t res = EXT4_NDIR_BLOCKS;
2537 int meta_blocks;
2538 loff_t upper_limit;
2539 /* This is calculated to be the largest file size for a dense, block
2540 * mapped file such that the file's total number of 512-byte sectors,
2541 * including data and all indirect blocks, does not exceed (2^48 - 1).
2542 *
2543 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2544 * number of 512-byte sectors of the file.
2545 */
2546
2547 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2548 /*
2549 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2550 * the inode i_block field represents total file blocks in
2551 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2552 */
2553 upper_limit = (1LL << 32) - 1;
2554
2555 /* total blocks in file system block size */
2556 upper_limit >>= (bits - 9);
2557
2558 } else {
2559 /*
2560 * We use 48 bit ext4_inode i_blocks
2561 * With EXT4_HUGE_FILE_FL set the i_blocks
2562 * represent total number of blocks in
2563 * file system block size
2564 */
2565 upper_limit = (1LL << 48) - 1;
2566
2567 }
2568
2569 /* indirect blocks */
2570 meta_blocks = 1;
2571 /* double indirect blocks */
2572 meta_blocks += 1 + (1LL << (bits-2));
2573 /* tripple indirect blocks */
2574 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2575
2576 upper_limit -= meta_blocks;
2577 upper_limit <<= bits;
2578
2579 res += 1LL << (bits-2);
2580 res += 1LL << (2*(bits-2));
2581 res += 1LL << (3*(bits-2));
2582 res <<= bits;
2583 if (res > upper_limit)
2584 res = upper_limit;
2585
2586 if (res > MAX_LFS_FILESIZE)
2587 res = MAX_LFS_FILESIZE;
2588
2589 return res;
2590 }
2591
2592 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2593 ext4_fsblk_t logical_sb_block, int nr)
2594 {
2595 struct ext4_sb_info *sbi = EXT4_SB(sb);
2596 ext4_group_t bg, first_meta_bg;
2597 int has_super = 0;
2598
2599 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2600
2601 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2602 return logical_sb_block + nr + 1;
2603 bg = sbi->s_desc_per_block * nr;
2604 if (ext4_bg_has_super(sb, bg))
2605 has_super = 1;
2606
2607 /*
2608 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2609 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2610 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2611 * compensate.
2612 */
2613 if (sb->s_blocksize == 1024 && nr == 0 &&
2614 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2615 has_super++;
2616
2617 return (has_super + ext4_group_first_block_no(sb, bg));
2618 }
2619
2620 /**
2621 * ext4_get_stripe_size: Get the stripe size.
2622 * @sbi: In memory super block info
2623 *
2624 * If we have specified it via mount option, then
2625 * use the mount option value. If the value specified at mount time is
2626 * greater than the blocks per group use the super block value.
2627 * If the super block value is greater than blocks per group return 0.
2628 * Allocator needs it be less than blocks per group.
2629 *
2630 */
2631 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2632 {
2633 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2634 unsigned long stripe_width =
2635 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2636 int ret;
2637
2638 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2639 ret = sbi->s_stripe;
2640 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2641 ret = stripe_width;
2642 else if (stride && stride <= sbi->s_blocks_per_group)
2643 ret = stride;
2644 else
2645 ret = 0;
2646
2647 /*
2648 * If the stripe width is 1, this makes no sense and
2649 * we set it to 0 to turn off stripe handling code.
2650 */
2651 if (ret <= 1)
2652 ret = 0;
2653
2654 return ret;
2655 }
2656
2657 /*
2658 * Check whether this filesystem can be mounted based on
2659 * the features present and the RDONLY/RDWR mount requested.
2660 * Returns 1 if this filesystem can be mounted as requested,
2661 * 0 if it cannot be.
2662 */
2663 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2664 {
2665 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2666 ext4_msg(sb, KERN_ERR,
2667 "Couldn't mount because of "
2668 "unsupported optional features (%x)",
2669 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2670 ~EXT4_FEATURE_INCOMPAT_SUPP));
2671 return 0;
2672 }
2673
2674 if (readonly)
2675 return 1;
2676
2677 if (ext4_has_feature_readonly(sb)) {
2678 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2679 sb->s_flags |= MS_RDONLY;
2680 return 1;
2681 }
2682
2683 /* Check that feature set is OK for a read-write mount */
2684 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2685 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2686 "unsupported optional features (%x)",
2687 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2688 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2689 return 0;
2690 }
2691 /*
2692 * Large file size enabled file system can only be mounted
2693 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2694 */
2695 if (ext4_has_feature_huge_file(sb)) {
2696 if (sizeof(blkcnt_t) < sizeof(u64)) {
2697 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2698 "cannot be mounted RDWR without "
2699 "CONFIG_LBDAF");
2700 return 0;
2701 }
2702 }
2703 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2704 ext4_msg(sb, KERN_ERR,
2705 "Can't support bigalloc feature without "
2706 "extents feature\n");
2707 return 0;
2708 }
2709
2710 #ifndef CONFIG_QUOTA
2711 if (ext4_has_feature_quota(sb) && !readonly) {
2712 ext4_msg(sb, KERN_ERR,
2713 "Filesystem with quota feature cannot be mounted RDWR "
2714 "without CONFIG_QUOTA");
2715 return 0;
2716 }
2717 if (ext4_has_feature_project(sb) && !readonly) {
2718 ext4_msg(sb, KERN_ERR,
2719 "Filesystem with project quota feature cannot be mounted RDWR "
2720 "without CONFIG_QUOTA");
2721 return 0;
2722 }
2723 #endif /* CONFIG_QUOTA */
2724 return 1;
2725 }
2726
2727 /*
2728 * This function is called once a day if we have errors logged
2729 * on the file system
2730 */
2731 static void print_daily_error_info(unsigned long arg)
2732 {
2733 struct super_block *sb = (struct super_block *) arg;
2734 struct ext4_sb_info *sbi;
2735 struct ext4_super_block *es;
2736
2737 sbi = EXT4_SB(sb);
2738 es = sbi->s_es;
2739
2740 if (es->s_error_count)
2741 /* fsck newer than v1.41.13 is needed to clean this condition. */
2742 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2743 le32_to_cpu(es->s_error_count));
2744 if (es->s_first_error_time) {
2745 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2746 sb->s_id, le32_to_cpu(es->s_first_error_time),
2747 (int) sizeof(es->s_first_error_func),
2748 es->s_first_error_func,
2749 le32_to_cpu(es->s_first_error_line));
2750 if (es->s_first_error_ino)
2751 printk(KERN_CONT ": inode %u",
2752 le32_to_cpu(es->s_first_error_ino));
2753 if (es->s_first_error_block)
2754 printk(KERN_CONT ": block %llu", (unsigned long long)
2755 le64_to_cpu(es->s_first_error_block));
2756 printk(KERN_CONT "\n");
2757 }
2758 if (es->s_last_error_time) {
2759 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2760 sb->s_id, le32_to_cpu(es->s_last_error_time),
2761 (int) sizeof(es->s_last_error_func),
2762 es->s_last_error_func,
2763 le32_to_cpu(es->s_last_error_line));
2764 if (es->s_last_error_ino)
2765 printk(KERN_CONT ": inode %u",
2766 le32_to_cpu(es->s_last_error_ino));
2767 if (es->s_last_error_block)
2768 printk(KERN_CONT ": block %llu", (unsigned long long)
2769 le64_to_cpu(es->s_last_error_block));
2770 printk(KERN_CONT "\n");
2771 }
2772 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2773 }
2774
2775 /* Find next suitable group and run ext4_init_inode_table */
2776 static int ext4_run_li_request(struct ext4_li_request *elr)
2777 {
2778 struct ext4_group_desc *gdp = NULL;
2779 ext4_group_t group, ngroups;
2780 struct super_block *sb;
2781 unsigned long timeout = 0;
2782 int ret = 0;
2783
2784 sb = elr->lr_super;
2785 ngroups = EXT4_SB(sb)->s_groups_count;
2786
2787 for (group = elr->lr_next_group; group < ngroups; group++) {
2788 gdp = ext4_get_group_desc(sb, group, NULL);
2789 if (!gdp) {
2790 ret = 1;
2791 break;
2792 }
2793
2794 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2795 break;
2796 }
2797
2798 if (group >= ngroups)
2799 ret = 1;
2800
2801 if (!ret) {
2802 timeout = jiffies;
2803 ret = ext4_init_inode_table(sb, group,
2804 elr->lr_timeout ? 0 : 1);
2805 if (elr->lr_timeout == 0) {
2806 timeout = (jiffies - timeout) *
2807 elr->lr_sbi->s_li_wait_mult;
2808 elr->lr_timeout = timeout;
2809 }
2810 elr->lr_next_sched = jiffies + elr->lr_timeout;
2811 elr->lr_next_group = group + 1;
2812 }
2813 return ret;
2814 }
2815
2816 /*
2817 * Remove lr_request from the list_request and free the
2818 * request structure. Should be called with li_list_mtx held
2819 */
2820 static void ext4_remove_li_request(struct ext4_li_request *elr)
2821 {
2822 struct ext4_sb_info *sbi;
2823
2824 if (!elr)
2825 return;
2826
2827 sbi = elr->lr_sbi;
2828
2829 list_del(&elr->lr_request);
2830 sbi->s_li_request = NULL;
2831 kfree(elr);
2832 }
2833
2834 static void ext4_unregister_li_request(struct super_block *sb)
2835 {
2836 mutex_lock(&ext4_li_mtx);
2837 if (!ext4_li_info) {
2838 mutex_unlock(&ext4_li_mtx);
2839 return;
2840 }
2841
2842 mutex_lock(&ext4_li_info->li_list_mtx);
2843 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2844 mutex_unlock(&ext4_li_info->li_list_mtx);
2845 mutex_unlock(&ext4_li_mtx);
2846 }
2847
2848 static struct task_struct *ext4_lazyinit_task;
2849
2850 /*
2851 * This is the function where ext4lazyinit thread lives. It walks
2852 * through the request list searching for next scheduled filesystem.
2853 * When such a fs is found, run the lazy initialization request
2854 * (ext4_rn_li_request) and keep track of the time spend in this
2855 * function. Based on that time we compute next schedule time of
2856 * the request. When walking through the list is complete, compute
2857 * next waking time and put itself into sleep.
2858 */
2859 static int ext4_lazyinit_thread(void *arg)
2860 {
2861 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2862 struct list_head *pos, *n;
2863 struct ext4_li_request *elr;
2864 unsigned long next_wakeup, cur;
2865
2866 BUG_ON(NULL == eli);
2867
2868 cont_thread:
2869 while (true) {
2870 next_wakeup = MAX_JIFFY_OFFSET;
2871
2872 mutex_lock(&eli->li_list_mtx);
2873 if (list_empty(&eli->li_request_list)) {
2874 mutex_unlock(&eli->li_list_mtx);
2875 goto exit_thread;
2876 }
2877 list_for_each_safe(pos, n, &eli->li_request_list) {
2878 int err = 0;
2879 int progress = 0;
2880 elr = list_entry(pos, struct ext4_li_request,
2881 lr_request);
2882
2883 if (time_before(jiffies, elr->lr_next_sched)) {
2884 if (time_before(elr->lr_next_sched, next_wakeup))
2885 next_wakeup = elr->lr_next_sched;
2886 continue;
2887 }
2888 if (down_read_trylock(&elr->lr_super->s_umount)) {
2889 if (sb_start_write_trylock(elr->lr_super)) {
2890 progress = 1;
2891 /*
2892 * We hold sb->s_umount, sb can not
2893 * be removed from the list, it is
2894 * now safe to drop li_list_mtx
2895 */
2896 mutex_unlock(&eli->li_list_mtx);
2897 err = ext4_run_li_request(elr);
2898 sb_end_write(elr->lr_super);
2899 mutex_lock(&eli->li_list_mtx);
2900 n = pos->next;
2901 }
2902 up_read((&elr->lr_super->s_umount));
2903 }
2904 /* error, remove the lazy_init job */
2905 if (err) {
2906 ext4_remove_li_request(elr);
2907 continue;
2908 }
2909 if (!progress) {
2910 elr->lr_next_sched = jiffies +
2911 (prandom_u32()
2912 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2913 }
2914 if (time_before(elr->lr_next_sched, next_wakeup))
2915 next_wakeup = elr->lr_next_sched;
2916 }
2917 mutex_unlock(&eli->li_list_mtx);
2918
2919 try_to_freeze();
2920
2921 cur = jiffies;
2922 if ((time_after_eq(cur, next_wakeup)) ||
2923 (MAX_JIFFY_OFFSET == next_wakeup)) {
2924 cond_resched();
2925 continue;
2926 }
2927
2928 schedule_timeout_interruptible(next_wakeup - cur);
2929
2930 if (kthread_should_stop()) {
2931 ext4_clear_request_list();
2932 goto exit_thread;
2933 }
2934 }
2935
2936 exit_thread:
2937 /*
2938 * It looks like the request list is empty, but we need
2939 * to check it under the li_list_mtx lock, to prevent any
2940 * additions into it, and of course we should lock ext4_li_mtx
2941 * to atomically free the list and ext4_li_info, because at
2942 * this point another ext4 filesystem could be registering
2943 * new one.
2944 */
2945 mutex_lock(&ext4_li_mtx);
2946 mutex_lock(&eli->li_list_mtx);
2947 if (!list_empty(&eli->li_request_list)) {
2948 mutex_unlock(&eli->li_list_mtx);
2949 mutex_unlock(&ext4_li_mtx);
2950 goto cont_thread;
2951 }
2952 mutex_unlock(&eli->li_list_mtx);
2953 kfree(ext4_li_info);
2954 ext4_li_info = NULL;
2955 mutex_unlock(&ext4_li_mtx);
2956
2957 return 0;
2958 }
2959
2960 static void ext4_clear_request_list(void)
2961 {
2962 struct list_head *pos, *n;
2963 struct ext4_li_request *elr;
2964
2965 mutex_lock(&ext4_li_info->li_list_mtx);
2966 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2967 elr = list_entry(pos, struct ext4_li_request,
2968 lr_request);
2969 ext4_remove_li_request(elr);
2970 }
2971 mutex_unlock(&ext4_li_info->li_list_mtx);
2972 }
2973
2974 static int ext4_run_lazyinit_thread(void)
2975 {
2976 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2977 ext4_li_info, "ext4lazyinit");
2978 if (IS_ERR(ext4_lazyinit_task)) {
2979 int err = PTR_ERR(ext4_lazyinit_task);
2980 ext4_clear_request_list();
2981 kfree(ext4_li_info);
2982 ext4_li_info = NULL;
2983 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2984 "initialization thread\n",
2985 err);
2986 return err;
2987 }
2988 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2989 return 0;
2990 }
2991
2992 /*
2993 * Check whether it make sense to run itable init. thread or not.
2994 * If there is at least one uninitialized inode table, return
2995 * corresponding group number, else the loop goes through all
2996 * groups and return total number of groups.
2997 */
2998 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2999 {
3000 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3001 struct ext4_group_desc *gdp = NULL;
3002
3003 for (group = 0; group < ngroups; group++) {
3004 gdp = ext4_get_group_desc(sb, group, NULL);
3005 if (!gdp)
3006 continue;
3007
3008 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3009 break;
3010 }
3011
3012 return group;
3013 }
3014
3015 static int ext4_li_info_new(void)
3016 {
3017 struct ext4_lazy_init *eli = NULL;
3018
3019 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3020 if (!eli)
3021 return -ENOMEM;
3022
3023 INIT_LIST_HEAD(&eli->li_request_list);
3024 mutex_init(&eli->li_list_mtx);
3025
3026 eli->li_state |= EXT4_LAZYINIT_QUIT;
3027
3028 ext4_li_info = eli;
3029
3030 return 0;
3031 }
3032
3033 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3034 ext4_group_t start)
3035 {
3036 struct ext4_sb_info *sbi = EXT4_SB(sb);
3037 struct ext4_li_request *elr;
3038
3039 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3040 if (!elr)
3041 return NULL;
3042
3043 elr->lr_super = sb;
3044 elr->lr_sbi = sbi;
3045 elr->lr_next_group = start;
3046
3047 /*
3048 * Randomize first schedule time of the request to
3049 * spread the inode table initialization requests
3050 * better.
3051 */
3052 elr->lr_next_sched = jiffies + (prandom_u32() %
3053 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3054 return elr;
3055 }
3056
3057 int ext4_register_li_request(struct super_block *sb,
3058 ext4_group_t first_not_zeroed)
3059 {
3060 struct ext4_sb_info *sbi = EXT4_SB(sb);
3061 struct ext4_li_request *elr = NULL;
3062 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3063 int ret = 0;
3064
3065 mutex_lock(&ext4_li_mtx);
3066 if (sbi->s_li_request != NULL) {
3067 /*
3068 * Reset timeout so it can be computed again, because
3069 * s_li_wait_mult might have changed.
3070 */
3071 sbi->s_li_request->lr_timeout = 0;
3072 goto out;
3073 }
3074
3075 if (first_not_zeroed == ngroups ||
3076 (sb->s_flags & MS_RDONLY) ||
3077 !test_opt(sb, INIT_INODE_TABLE))
3078 goto out;
3079
3080 elr = ext4_li_request_new(sb, first_not_zeroed);
3081 if (!elr) {
3082 ret = -ENOMEM;
3083 goto out;
3084 }
3085
3086 if (NULL == ext4_li_info) {
3087 ret = ext4_li_info_new();
3088 if (ret)
3089 goto out;
3090 }
3091
3092 mutex_lock(&ext4_li_info->li_list_mtx);
3093 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3094 mutex_unlock(&ext4_li_info->li_list_mtx);
3095
3096 sbi->s_li_request = elr;
3097 /*
3098 * set elr to NULL here since it has been inserted to
3099 * the request_list and the removal and free of it is
3100 * handled by ext4_clear_request_list from now on.
3101 */
3102 elr = NULL;
3103
3104 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3105 ret = ext4_run_lazyinit_thread();
3106 if (ret)
3107 goto out;
3108 }
3109 out:
3110 mutex_unlock(&ext4_li_mtx);
3111 if (ret)
3112 kfree(elr);
3113 return ret;
3114 }
3115
3116 /*
3117 * We do not need to lock anything since this is called on
3118 * module unload.
3119 */
3120 static void ext4_destroy_lazyinit_thread(void)
3121 {
3122 /*
3123 * If thread exited earlier
3124 * there's nothing to be done.
3125 */
3126 if (!ext4_li_info || !ext4_lazyinit_task)
3127 return;
3128
3129 kthread_stop(ext4_lazyinit_task);
3130 }
3131
3132 static int set_journal_csum_feature_set(struct super_block *sb)
3133 {
3134 int ret = 1;
3135 int compat, incompat;
3136 struct ext4_sb_info *sbi = EXT4_SB(sb);
3137
3138 if (ext4_has_metadata_csum(sb)) {
3139 /* journal checksum v3 */
3140 compat = 0;
3141 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3142 } else {
3143 /* journal checksum v1 */
3144 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3145 incompat = 0;
3146 }
3147
3148 jbd2_journal_clear_features(sbi->s_journal,
3149 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3150 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3151 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3152 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3153 ret = jbd2_journal_set_features(sbi->s_journal,
3154 compat, 0,
3155 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3156 incompat);
3157 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3158 ret = jbd2_journal_set_features(sbi->s_journal,
3159 compat, 0,
3160 incompat);
3161 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3162 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3163 } else {
3164 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3165 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3166 }
3167
3168 return ret;
3169 }
3170
3171 /*
3172 * Note: calculating the overhead so we can be compatible with
3173 * historical BSD practice is quite difficult in the face of
3174 * clusters/bigalloc. This is because multiple metadata blocks from
3175 * different block group can end up in the same allocation cluster.
3176 * Calculating the exact overhead in the face of clustered allocation
3177 * requires either O(all block bitmaps) in memory or O(number of block
3178 * groups**2) in time. We will still calculate the superblock for
3179 * older file systems --- and if we come across with a bigalloc file
3180 * system with zero in s_overhead_clusters the estimate will be close to
3181 * correct especially for very large cluster sizes --- but for newer
3182 * file systems, it's better to calculate this figure once at mkfs
3183 * time, and store it in the superblock. If the superblock value is
3184 * present (even for non-bigalloc file systems), we will use it.
3185 */
3186 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3187 char *buf)
3188 {
3189 struct ext4_sb_info *sbi = EXT4_SB(sb);
3190 struct ext4_group_desc *gdp;
3191 ext4_fsblk_t first_block, last_block, b;
3192 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3193 int s, j, count = 0;
3194
3195 if (!ext4_has_feature_bigalloc(sb))
3196 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3197 sbi->s_itb_per_group + 2);
3198
3199 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3200 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3201 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3202 for (i = 0; i < ngroups; i++) {
3203 gdp = ext4_get_group_desc(sb, i, NULL);
3204 b = ext4_block_bitmap(sb, gdp);
3205 if (b >= first_block && b <= last_block) {
3206 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3207 count++;
3208 }
3209 b = ext4_inode_bitmap(sb, gdp);
3210 if (b >= first_block && b <= last_block) {
3211 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3212 count++;
3213 }
3214 b = ext4_inode_table(sb, gdp);
3215 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3216 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3217 int c = EXT4_B2C(sbi, b - first_block);
3218 ext4_set_bit(c, buf);
3219 count++;
3220 }
3221 if (i != grp)
3222 continue;
3223 s = 0;
3224 if (ext4_bg_has_super(sb, grp)) {
3225 ext4_set_bit(s++, buf);
3226 count++;
3227 }
3228 j = ext4_bg_num_gdb(sb, grp);
3229 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3230 ext4_error(sb, "Invalid number of block group "
3231 "descriptor blocks: %d", j);
3232 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3233 }
3234 count += j;
3235 for (; j > 0; j--)
3236 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3237 }
3238 if (!count)
3239 return 0;
3240 return EXT4_CLUSTERS_PER_GROUP(sb) -
3241 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3242 }
3243
3244 /*
3245 * Compute the overhead and stash it in sbi->s_overhead
3246 */
3247 int ext4_calculate_overhead(struct super_block *sb)
3248 {
3249 struct ext4_sb_info *sbi = EXT4_SB(sb);
3250 struct ext4_super_block *es = sbi->s_es;
3251 struct inode *j_inode;
3252 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3253 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3254 ext4_fsblk_t overhead = 0;
3255 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3256
3257 if (!buf)
3258 return -ENOMEM;
3259
3260 /*
3261 * Compute the overhead (FS structures). This is constant
3262 * for a given filesystem unless the number of block groups
3263 * changes so we cache the previous value until it does.
3264 */
3265
3266 /*
3267 * All of the blocks before first_data_block are overhead
3268 */
3269 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3270
3271 /*
3272 * Add the overhead found in each block group
3273 */
3274 for (i = 0; i < ngroups; i++) {
3275 int blks;
3276
3277 blks = count_overhead(sb, i, buf);
3278 overhead += blks;
3279 if (blks)
3280 memset(buf, 0, PAGE_SIZE);
3281 cond_resched();
3282 }
3283
3284 /*
3285 * Add the internal journal blocks whether the journal has been
3286 * loaded or not
3287 */
3288 if (sbi->s_journal && !sbi->journal_bdev)
3289 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3290 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3291 j_inode = ext4_get_journal_inode(sb, j_inum);
3292 if (j_inode) {
3293 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3294 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3295 iput(j_inode);
3296 } else {
3297 ext4_msg(sb, KERN_ERR, "can't get journal size");
3298 }
3299 }
3300 sbi->s_overhead = overhead;
3301 smp_wmb();
3302 free_page((unsigned long) buf);
3303 return 0;
3304 }
3305
3306 static void ext4_set_resv_clusters(struct super_block *sb)
3307 {
3308 ext4_fsblk_t resv_clusters;
3309 struct ext4_sb_info *sbi = EXT4_SB(sb);
3310
3311 /*
3312 * There's no need to reserve anything when we aren't using extents.
3313 * The space estimates are exact, there are no unwritten extents,
3314 * hole punching doesn't need new metadata... This is needed especially
3315 * to keep ext2/3 backward compatibility.
3316 */
3317 if (!ext4_has_feature_extents(sb))
3318 return;
3319 /*
3320 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3321 * This should cover the situations where we can not afford to run
3322 * out of space like for example punch hole, or converting
3323 * unwritten extents in delalloc path. In most cases such
3324 * allocation would require 1, or 2 blocks, higher numbers are
3325 * very rare.
3326 */
3327 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3328 sbi->s_cluster_bits);
3329
3330 do_div(resv_clusters, 50);
3331 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3332
3333 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3334 }
3335
3336 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3337 {
3338 char *orig_data = kstrdup(data, GFP_KERNEL);
3339 struct buffer_head *bh;
3340 struct ext4_super_block *es = NULL;
3341 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3342 ext4_fsblk_t block;
3343 ext4_fsblk_t sb_block = get_sb_block(&data);
3344 ext4_fsblk_t logical_sb_block;
3345 unsigned long offset = 0;
3346 unsigned long journal_devnum = 0;
3347 unsigned long def_mount_opts;
3348 struct inode *root;
3349 const char *descr;
3350 int ret = -ENOMEM;
3351 int blocksize, clustersize;
3352 unsigned int db_count;
3353 unsigned int i;
3354 int needs_recovery, has_huge_files, has_bigalloc;
3355 __u64 blocks_count;
3356 int err = 0;
3357 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3358 ext4_group_t first_not_zeroed;
3359
3360 if ((data && !orig_data) || !sbi)
3361 goto out_free_base;
3362
3363 sbi->s_blockgroup_lock =
3364 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3365 if (!sbi->s_blockgroup_lock)
3366 goto out_free_base;
3367
3368 sb->s_fs_info = sbi;
3369 sbi->s_sb = sb;
3370 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3371 sbi->s_sb_block = sb_block;
3372 if (sb->s_bdev->bd_part)
3373 sbi->s_sectors_written_start =
3374 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3375
3376 /* Cleanup superblock name */
3377 strreplace(sb->s_id, '/', '!');
3378
3379 /* -EINVAL is default */
3380 ret = -EINVAL;
3381 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3382 if (!blocksize) {
3383 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3384 goto out_fail;
3385 }
3386
3387 /*
3388 * The ext4 superblock will not be buffer aligned for other than 1kB
3389 * block sizes. We need to calculate the offset from buffer start.
3390 */
3391 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3392 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3393 offset = do_div(logical_sb_block, blocksize);
3394 } else {
3395 logical_sb_block = sb_block;
3396 }
3397
3398 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3399 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3400 goto out_fail;
3401 }
3402 /*
3403 * Note: s_es must be initialized as soon as possible because
3404 * some ext4 macro-instructions depend on its value
3405 */
3406 es = (struct ext4_super_block *) (bh->b_data + offset);
3407 sbi->s_es = es;
3408 sb->s_magic = le16_to_cpu(es->s_magic);
3409 if (sb->s_magic != EXT4_SUPER_MAGIC)
3410 goto cantfind_ext4;
3411 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3412
3413 /* Warn if metadata_csum and gdt_csum are both set. */
3414 if (ext4_has_feature_metadata_csum(sb) &&
3415 ext4_has_feature_gdt_csum(sb))
3416 ext4_warning(sb, "metadata_csum and uninit_bg are "
3417 "redundant flags; please run fsck.");
3418
3419 /* Check for a known checksum algorithm */
3420 if (!ext4_verify_csum_type(sb, es)) {
3421 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3422 "unknown checksum algorithm.");
3423 silent = 1;
3424 goto cantfind_ext4;
3425 }
3426
3427 /* Load the checksum driver */
3428 if (ext4_has_feature_metadata_csum(sb)) {
3429 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3430 if (IS_ERR(sbi->s_chksum_driver)) {
3431 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3432 ret = PTR_ERR(sbi->s_chksum_driver);
3433 sbi->s_chksum_driver = NULL;
3434 goto failed_mount;
3435 }
3436 }
3437
3438 /* Check superblock checksum */
3439 if (!ext4_superblock_csum_verify(sb, es)) {
3440 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3441 "invalid superblock checksum. Run e2fsck?");
3442 silent = 1;
3443 ret = -EFSBADCRC;
3444 goto cantfind_ext4;
3445 }
3446
3447 /* Precompute checksum seed for all metadata */
3448 if (ext4_has_feature_csum_seed(sb))
3449 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3450 else if (ext4_has_metadata_csum(sb))
3451 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3452 sizeof(es->s_uuid));
3453
3454 /* Set defaults before we parse the mount options */
3455 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3456 set_opt(sb, INIT_INODE_TABLE);
3457 if (def_mount_opts & EXT4_DEFM_DEBUG)
3458 set_opt(sb, DEBUG);
3459 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3460 set_opt(sb, GRPID);
3461 if (def_mount_opts & EXT4_DEFM_UID16)
3462 set_opt(sb, NO_UID32);
3463 /* xattr user namespace & acls are now defaulted on */
3464 set_opt(sb, XATTR_USER);
3465 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3466 set_opt(sb, POSIX_ACL);
3467 #endif
3468 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3469 if (ext4_has_metadata_csum(sb))
3470 set_opt(sb, JOURNAL_CHECKSUM);
3471
3472 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3473 set_opt(sb, JOURNAL_DATA);
3474 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3475 set_opt(sb, ORDERED_DATA);
3476 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3477 set_opt(sb, WRITEBACK_DATA);
3478
3479 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3480 set_opt(sb, ERRORS_PANIC);
3481 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3482 set_opt(sb, ERRORS_CONT);
3483 else
3484 set_opt(sb, ERRORS_RO);
3485 /* block_validity enabled by default; disable with noblock_validity */
3486 set_opt(sb, BLOCK_VALIDITY);
3487 if (def_mount_opts & EXT4_DEFM_DISCARD)
3488 set_opt(sb, DISCARD);
3489
3490 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3491 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3492 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3493 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3494 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3495
3496 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3497 set_opt(sb, BARRIER);
3498
3499 /*
3500 * enable delayed allocation by default
3501 * Use -o nodelalloc to turn it off
3502 */
3503 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3504 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3505 set_opt(sb, DELALLOC);
3506
3507 /*
3508 * set default s_li_wait_mult for lazyinit, for the case there is
3509 * no mount option specified.
3510 */
3511 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3512
3513 if (sbi->s_es->s_mount_opts[0]) {
3514 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3515 sizeof(sbi->s_es->s_mount_opts),
3516 GFP_KERNEL);
3517 if (!s_mount_opts)
3518 goto failed_mount;
3519 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3520 &journal_ioprio, 0)) {
3521 ext4_msg(sb, KERN_WARNING,
3522 "failed to parse options in superblock: %s",
3523 s_mount_opts);
3524 }
3525 kfree(s_mount_opts);
3526 }
3527 sbi->s_def_mount_opt = sbi->s_mount_opt;
3528 if (!parse_options((char *) data, sb, &journal_devnum,
3529 &journal_ioprio, 0))
3530 goto failed_mount;
3531
3532 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3533 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3534 "with data=journal disables delayed "
3535 "allocation and O_DIRECT support!\n");
3536 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3537 ext4_msg(sb, KERN_ERR, "can't mount with "
3538 "both data=journal and delalloc");
3539 goto failed_mount;
3540 }
3541 if (test_opt(sb, DIOREAD_NOLOCK)) {
3542 ext4_msg(sb, KERN_ERR, "can't mount with "
3543 "both data=journal and dioread_nolock");
3544 goto failed_mount;
3545 }
3546 if (test_opt(sb, DAX)) {
3547 ext4_msg(sb, KERN_ERR, "can't mount with "
3548 "both data=journal and dax");
3549 goto failed_mount;
3550 }
3551 if (ext4_has_feature_encrypt(sb)) {
3552 ext4_msg(sb, KERN_WARNING,
3553 "encrypted files will use data=ordered "
3554 "instead of data journaling mode");
3555 }
3556 if (test_opt(sb, DELALLOC))
3557 clear_opt(sb, DELALLOC);
3558 } else {
3559 sb->s_iflags |= SB_I_CGROUPWB;
3560 }
3561
3562 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3563 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3564
3565 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3566 (ext4_has_compat_features(sb) ||
3567 ext4_has_ro_compat_features(sb) ||
3568 ext4_has_incompat_features(sb)))
3569 ext4_msg(sb, KERN_WARNING,
3570 "feature flags set on rev 0 fs, "
3571 "running e2fsck is recommended");
3572
3573 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3574 set_opt2(sb, HURD_COMPAT);
3575 if (ext4_has_feature_64bit(sb)) {
3576 ext4_msg(sb, KERN_ERR,
3577 "The Hurd can't support 64-bit file systems");
3578 goto failed_mount;
3579 }
3580 }
3581
3582 if (IS_EXT2_SB(sb)) {
3583 if (ext2_feature_set_ok(sb))
3584 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3585 "using the ext4 subsystem");
3586 else {
3587 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3588 "to feature incompatibilities");
3589 goto failed_mount;
3590 }
3591 }
3592
3593 if (IS_EXT3_SB(sb)) {
3594 if (ext3_feature_set_ok(sb))
3595 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3596 "using the ext4 subsystem");
3597 else {
3598 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3599 "to feature incompatibilities");
3600 goto failed_mount;
3601 }
3602 }
3603
3604 /*
3605 * Check feature flags regardless of the revision level, since we
3606 * previously didn't change the revision level when setting the flags,
3607 * so there is a chance incompat flags are set on a rev 0 filesystem.
3608 */
3609 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3610 goto failed_mount;
3611
3612 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3613 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3614 blocksize > EXT4_MAX_BLOCK_SIZE) {
3615 ext4_msg(sb, KERN_ERR,
3616 "Unsupported filesystem blocksize %d (%d log_block_size)",
3617 blocksize, le32_to_cpu(es->s_log_block_size));
3618 goto failed_mount;
3619 }
3620 if (le32_to_cpu(es->s_log_block_size) >
3621 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3622 ext4_msg(sb, KERN_ERR,
3623 "Invalid log block size: %u",
3624 le32_to_cpu(es->s_log_block_size));
3625 goto failed_mount;
3626 }
3627
3628 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3629 ext4_msg(sb, KERN_ERR,
3630 "Number of reserved GDT blocks insanely large: %d",
3631 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3632 goto failed_mount;
3633 }
3634
3635 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3636 err = bdev_dax_supported(sb, blocksize);
3637 if (err)
3638 goto failed_mount;
3639 }
3640
3641 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3642 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3643 es->s_encryption_level);
3644 goto failed_mount;
3645 }
3646
3647 if (sb->s_blocksize != blocksize) {
3648 /* Validate the filesystem blocksize */
3649 if (!sb_set_blocksize(sb, blocksize)) {
3650 ext4_msg(sb, KERN_ERR, "bad block size %d",
3651 blocksize);
3652 goto failed_mount;
3653 }
3654
3655 brelse(bh);
3656 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3657 offset = do_div(logical_sb_block, blocksize);
3658 bh = sb_bread_unmovable(sb, logical_sb_block);
3659 if (!bh) {
3660 ext4_msg(sb, KERN_ERR,
3661 "Can't read superblock on 2nd try");
3662 goto failed_mount;
3663 }
3664 es = (struct ext4_super_block *)(bh->b_data + offset);
3665 sbi->s_es = es;
3666 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3667 ext4_msg(sb, KERN_ERR,
3668 "Magic mismatch, very weird!");
3669 goto failed_mount;
3670 }
3671 }
3672
3673 has_huge_files = ext4_has_feature_huge_file(sb);
3674 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3675 has_huge_files);
3676 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3677
3678 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3679 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3680 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3681 } else {
3682 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3683 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3684 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3685 (!is_power_of_2(sbi->s_inode_size)) ||
3686 (sbi->s_inode_size > blocksize)) {
3687 ext4_msg(sb, KERN_ERR,
3688 "unsupported inode size: %d",
3689 sbi->s_inode_size);
3690 goto failed_mount;
3691 }
3692 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3693 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3694 }
3695
3696 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3697 if (ext4_has_feature_64bit(sb)) {
3698 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3699 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3700 !is_power_of_2(sbi->s_desc_size)) {
3701 ext4_msg(sb, KERN_ERR,
3702 "unsupported descriptor size %lu",
3703 sbi->s_desc_size);
3704 goto failed_mount;
3705 }
3706 } else
3707 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3708
3709 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3710 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3711
3712 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3713 if (sbi->s_inodes_per_block == 0)
3714 goto cantfind_ext4;
3715 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3716 sbi->s_inodes_per_group > blocksize * 8) {
3717 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3718 sbi->s_blocks_per_group);
3719 goto failed_mount;
3720 }
3721 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3722 sbi->s_inodes_per_block;
3723 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3724 sbi->s_sbh = bh;
3725 sbi->s_mount_state = le16_to_cpu(es->s_state);
3726 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3727 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3728
3729 for (i = 0; i < 4; i++)
3730 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3731 sbi->s_def_hash_version = es->s_def_hash_version;
3732 if (ext4_has_feature_dir_index(sb)) {
3733 i = le32_to_cpu(es->s_flags);
3734 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3735 sbi->s_hash_unsigned = 3;
3736 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3737 #ifdef __CHAR_UNSIGNED__
3738 if (!(sb->s_flags & MS_RDONLY))
3739 es->s_flags |=
3740 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3741 sbi->s_hash_unsigned = 3;
3742 #else
3743 if (!(sb->s_flags & MS_RDONLY))
3744 es->s_flags |=
3745 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3746 #endif
3747 }
3748 }
3749
3750 /* Handle clustersize */
3751 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3752 has_bigalloc = ext4_has_feature_bigalloc(sb);
3753 if (has_bigalloc) {
3754 if (clustersize < blocksize) {
3755 ext4_msg(sb, KERN_ERR,
3756 "cluster size (%d) smaller than "
3757 "block size (%d)", clustersize, blocksize);
3758 goto failed_mount;
3759 }
3760 if (le32_to_cpu(es->s_log_cluster_size) >
3761 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3762 ext4_msg(sb, KERN_ERR,
3763 "Invalid log cluster size: %u",
3764 le32_to_cpu(es->s_log_cluster_size));
3765 goto failed_mount;
3766 }
3767 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3768 le32_to_cpu(es->s_log_block_size);
3769 sbi->s_clusters_per_group =
3770 le32_to_cpu(es->s_clusters_per_group);
3771 if (sbi->s_clusters_per_group > blocksize * 8) {
3772 ext4_msg(sb, KERN_ERR,
3773 "#clusters per group too big: %lu",
3774 sbi->s_clusters_per_group);
3775 goto failed_mount;
3776 }
3777 if (sbi->s_blocks_per_group !=
3778 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3779 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3780 "clusters per group (%lu) inconsistent",
3781 sbi->s_blocks_per_group,
3782 sbi->s_clusters_per_group);
3783 goto failed_mount;
3784 }
3785 } else {
3786 if (clustersize != blocksize) {
3787 ext4_warning(sb, "fragment/cluster size (%d) != "
3788 "block size (%d)", clustersize,
3789 blocksize);
3790 clustersize = blocksize;
3791 }
3792 if (sbi->s_blocks_per_group > blocksize * 8) {
3793 ext4_msg(sb, KERN_ERR,
3794 "#blocks per group too big: %lu",
3795 sbi->s_blocks_per_group);
3796 goto failed_mount;
3797 }
3798 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3799 sbi->s_cluster_bits = 0;
3800 }
3801 sbi->s_cluster_ratio = clustersize / blocksize;
3802
3803 /* Do we have standard group size of clustersize * 8 blocks ? */
3804 if (sbi->s_blocks_per_group == clustersize << 3)
3805 set_opt2(sb, STD_GROUP_SIZE);
3806
3807 /*
3808 * Test whether we have more sectors than will fit in sector_t,
3809 * and whether the max offset is addressable by the page cache.
3810 */
3811 err = generic_check_addressable(sb->s_blocksize_bits,
3812 ext4_blocks_count(es));
3813 if (err) {
3814 ext4_msg(sb, KERN_ERR, "filesystem"
3815 " too large to mount safely on this system");
3816 if (sizeof(sector_t) < 8)
3817 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3818 goto failed_mount;
3819 }
3820
3821 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3822 goto cantfind_ext4;
3823
3824 /* check blocks count against device size */
3825 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3826 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3827 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3828 "exceeds size of device (%llu blocks)",
3829 ext4_blocks_count(es), blocks_count);
3830 goto failed_mount;
3831 }
3832
3833 /*
3834 * It makes no sense for the first data block to be beyond the end
3835 * of the filesystem.
3836 */
3837 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3838 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3839 "block %u is beyond end of filesystem (%llu)",
3840 le32_to_cpu(es->s_first_data_block),
3841 ext4_blocks_count(es));
3842 goto failed_mount;
3843 }
3844 blocks_count = (ext4_blocks_count(es) -
3845 le32_to_cpu(es->s_first_data_block) +
3846 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3847 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3848 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3849 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3850 "(block count %llu, first data block %u, "
3851 "blocks per group %lu)", sbi->s_groups_count,
3852 ext4_blocks_count(es),
3853 le32_to_cpu(es->s_first_data_block),
3854 EXT4_BLOCKS_PER_GROUP(sb));
3855 goto failed_mount;
3856 }
3857 sbi->s_groups_count = blocks_count;
3858 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3859 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3860 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3861 EXT4_DESC_PER_BLOCK(sb);
3862 if (ext4_has_feature_meta_bg(sb)) {
3863 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3864 ext4_msg(sb, KERN_WARNING,
3865 "first meta block group too large: %u "
3866 "(group descriptor block count %u)",
3867 le32_to_cpu(es->s_first_meta_bg), db_count);
3868 goto failed_mount;
3869 }
3870 }
3871 sbi->s_group_desc = ext4_kvmalloc(db_count *
3872 sizeof(struct buffer_head *),
3873 GFP_KERNEL);
3874 if (sbi->s_group_desc == NULL) {
3875 ext4_msg(sb, KERN_ERR, "not enough memory");
3876 ret = -ENOMEM;
3877 goto failed_mount;
3878 }
3879
3880 bgl_lock_init(sbi->s_blockgroup_lock);
3881
3882 for (i = 0; i < db_count; i++) {
3883 block = descriptor_loc(sb, logical_sb_block, i);
3884 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3885 if (!sbi->s_group_desc[i]) {
3886 ext4_msg(sb, KERN_ERR,
3887 "can't read group descriptor %d", i);
3888 db_count = i;
3889 goto failed_mount2;
3890 }
3891 }
3892 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3893 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3894 ret = -EFSCORRUPTED;
3895 goto failed_mount2;
3896 }
3897
3898 sbi->s_gdb_count = db_count;
3899 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3900 spin_lock_init(&sbi->s_next_gen_lock);
3901
3902 setup_timer(&sbi->s_err_report, print_daily_error_info,
3903 (unsigned long) sb);
3904
3905 /* Register extent status tree shrinker */
3906 if (ext4_es_register_shrinker(sbi))
3907 goto failed_mount3;
3908
3909 sbi->s_stripe = ext4_get_stripe_size(sbi);
3910 sbi->s_extent_max_zeroout_kb = 32;
3911
3912 /*
3913 * set up enough so that it can read an inode
3914 */
3915 sb->s_op = &ext4_sops;
3916 sb->s_export_op = &ext4_export_ops;
3917 sb->s_xattr = ext4_xattr_handlers;
3918 sb->s_cop = &ext4_cryptops;
3919 #ifdef CONFIG_QUOTA
3920 sb->dq_op = &ext4_quota_operations;
3921 if (ext4_has_feature_quota(sb))
3922 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3923 else
3924 sb->s_qcop = &ext4_qctl_operations;
3925 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3926 #endif
3927 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3928
3929 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3930 mutex_init(&sbi->s_orphan_lock);
3931
3932 sb->s_root = NULL;
3933
3934 needs_recovery = (es->s_last_orphan != 0 ||
3935 ext4_has_feature_journal_needs_recovery(sb));
3936
3937 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
3938 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3939 goto failed_mount3a;
3940
3941 /*
3942 * The first inode we look at is the journal inode. Don't try
3943 * root first: it may be modified in the journal!
3944 */
3945 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
3946 err = ext4_load_journal(sb, es, journal_devnum);
3947 if (err)
3948 goto failed_mount3a;
3949 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3950 ext4_has_feature_journal_needs_recovery(sb)) {
3951 ext4_msg(sb, KERN_ERR, "required journal recovery "
3952 "suppressed and not mounted read-only");
3953 goto failed_mount_wq;
3954 } else {
3955 /* Nojournal mode, all journal mount options are illegal */
3956 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
3957 ext4_msg(sb, KERN_ERR, "can't mount with "
3958 "journal_checksum, fs mounted w/o journal");
3959 goto failed_mount_wq;
3960 }
3961 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3962 ext4_msg(sb, KERN_ERR, "can't mount with "
3963 "journal_async_commit, fs mounted w/o journal");
3964 goto failed_mount_wq;
3965 }
3966 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
3967 ext4_msg(sb, KERN_ERR, "can't mount with "
3968 "commit=%lu, fs mounted w/o journal",
3969 sbi->s_commit_interval / HZ);
3970 goto failed_mount_wq;
3971 }
3972 if (EXT4_MOUNT_DATA_FLAGS &
3973 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3974 ext4_msg(sb, KERN_ERR, "can't mount with "
3975 "data=, fs mounted w/o journal");
3976 goto failed_mount_wq;
3977 }
3978 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
3979 clear_opt(sb, JOURNAL_CHECKSUM);
3980 clear_opt(sb, DATA_FLAGS);
3981 sbi->s_journal = NULL;
3982 needs_recovery = 0;
3983 goto no_journal;
3984 }
3985
3986 if (ext4_has_feature_64bit(sb) &&
3987 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3988 JBD2_FEATURE_INCOMPAT_64BIT)) {
3989 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3990 goto failed_mount_wq;
3991 }
3992
3993 if (!set_journal_csum_feature_set(sb)) {
3994 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3995 "feature set");
3996 goto failed_mount_wq;
3997 }
3998
3999 /* We have now updated the journal if required, so we can
4000 * validate the data journaling mode. */
4001 switch (test_opt(sb, DATA_FLAGS)) {
4002 case 0:
4003 /* No mode set, assume a default based on the journal
4004 * capabilities: ORDERED_DATA if the journal can
4005 * cope, else JOURNAL_DATA
4006 */
4007 if (jbd2_journal_check_available_features
4008 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4009 set_opt(sb, ORDERED_DATA);
4010 else
4011 set_opt(sb, JOURNAL_DATA);
4012 break;
4013
4014 case EXT4_MOUNT_ORDERED_DATA:
4015 case EXT4_MOUNT_WRITEBACK_DATA:
4016 if (!jbd2_journal_check_available_features
4017 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4018 ext4_msg(sb, KERN_ERR, "Journal does not support "
4019 "requested data journaling mode");
4020 goto failed_mount_wq;
4021 }
4022 default:
4023 break;
4024 }
4025
4026 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4027 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4028 ext4_msg(sb, KERN_ERR, "can't mount with "
4029 "journal_async_commit in data=ordered mode");
4030 goto failed_mount_wq;
4031 }
4032
4033 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4034
4035 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4036
4037 no_journal:
4038 sbi->s_mb_cache = ext4_xattr_create_cache();
4039 if (!sbi->s_mb_cache) {
4040 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4041 goto failed_mount_wq;
4042 }
4043
4044 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4045 (blocksize != PAGE_SIZE)) {
4046 ext4_msg(sb, KERN_ERR,
4047 "Unsupported blocksize for fs encryption");
4048 goto failed_mount_wq;
4049 }
4050
4051 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4052 !ext4_has_feature_encrypt(sb)) {
4053 ext4_set_feature_encrypt(sb);
4054 ext4_commit_super(sb, 1);
4055 }
4056
4057 /*
4058 * Get the # of file system overhead blocks from the
4059 * superblock if present.
4060 */
4061 if (es->s_overhead_clusters)
4062 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4063 else {
4064 err = ext4_calculate_overhead(sb);
4065 if (err)
4066 goto failed_mount_wq;
4067 }
4068
4069 /*
4070 * The maximum number of concurrent works can be high and
4071 * concurrency isn't really necessary. Limit it to 1.
4072 */
4073 EXT4_SB(sb)->rsv_conversion_wq =
4074 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4075 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4076 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4077 ret = -ENOMEM;
4078 goto failed_mount4;
4079 }
4080
4081 /*
4082 * The jbd2_journal_load will have done any necessary log recovery,
4083 * so we can safely mount the rest of the filesystem now.
4084 */
4085
4086 root = ext4_iget(sb, EXT4_ROOT_INO);
4087 if (IS_ERR(root)) {
4088 ext4_msg(sb, KERN_ERR, "get root inode failed");
4089 ret = PTR_ERR(root);
4090 root = NULL;
4091 goto failed_mount4;
4092 }
4093 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4094 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4095 iput(root);
4096 goto failed_mount4;
4097 }
4098 sb->s_root = d_make_root(root);
4099 if (!sb->s_root) {
4100 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4101 ret = -ENOMEM;
4102 goto failed_mount4;
4103 }
4104
4105 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4106 sb->s_flags |= MS_RDONLY;
4107
4108 /* determine the minimum size of new large inodes, if present */
4109 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4110 sbi->s_want_extra_isize == 0) {
4111 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4112 EXT4_GOOD_OLD_INODE_SIZE;
4113 if (ext4_has_feature_extra_isize(sb)) {
4114 if (sbi->s_want_extra_isize <
4115 le16_to_cpu(es->s_want_extra_isize))
4116 sbi->s_want_extra_isize =
4117 le16_to_cpu(es->s_want_extra_isize);
4118 if (sbi->s_want_extra_isize <
4119 le16_to_cpu(es->s_min_extra_isize))
4120 sbi->s_want_extra_isize =
4121 le16_to_cpu(es->s_min_extra_isize);
4122 }
4123 }
4124 /* Check if enough inode space is available */
4125 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4126 sbi->s_inode_size) {
4127 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4128 EXT4_GOOD_OLD_INODE_SIZE;
4129 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4130 "available");
4131 }
4132
4133 ext4_set_resv_clusters(sb);
4134
4135 err = ext4_setup_system_zone(sb);
4136 if (err) {
4137 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4138 "zone (%d)", err);
4139 goto failed_mount4a;
4140 }
4141
4142 ext4_ext_init(sb);
4143 err = ext4_mb_init(sb);
4144 if (err) {
4145 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4146 err);
4147 goto failed_mount5;
4148 }
4149
4150 block = ext4_count_free_clusters(sb);
4151 ext4_free_blocks_count_set(sbi->s_es,
4152 EXT4_C2B(sbi, block));
4153 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4154 GFP_KERNEL);
4155 if (!err) {
4156 unsigned long freei = ext4_count_free_inodes(sb);
4157 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4158 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4159 GFP_KERNEL);
4160 }
4161 if (!err)
4162 err = percpu_counter_init(&sbi->s_dirs_counter,
4163 ext4_count_dirs(sb), GFP_KERNEL);
4164 if (!err)
4165 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4166 GFP_KERNEL);
4167 if (!err)
4168 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4169
4170 if (err) {
4171 ext4_msg(sb, KERN_ERR, "insufficient memory");
4172 goto failed_mount6;
4173 }
4174
4175 if (ext4_has_feature_flex_bg(sb))
4176 if (!ext4_fill_flex_info(sb)) {
4177 ext4_msg(sb, KERN_ERR,
4178 "unable to initialize "
4179 "flex_bg meta info!");
4180 goto failed_mount6;
4181 }
4182
4183 err = ext4_register_li_request(sb, first_not_zeroed);
4184 if (err)
4185 goto failed_mount6;
4186
4187 err = ext4_register_sysfs(sb);
4188 if (err)
4189 goto failed_mount7;
4190
4191 #ifdef CONFIG_QUOTA
4192 /* Enable quota usage during mount. */
4193 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4194 err = ext4_enable_quotas(sb);
4195 if (err)
4196 goto failed_mount8;
4197 }
4198 #endif /* CONFIG_QUOTA */
4199
4200 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4201 ext4_orphan_cleanup(sb, es);
4202 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4203 if (needs_recovery) {
4204 ext4_msg(sb, KERN_INFO, "recovery complete");
4205 ext4_mark_recovery_complete(sb, es);
4206 }
4207 if (EXT4_SB(sb)->s_journal) {
4208 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4209 descr = " journalled data mode";
4210 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4211 descr = " ordered data mode";
4212 else
4213 descr = " writeback data mode";
4214 } else
4215 descr = "out journal";
4216
4217 if (test_opt(sb, DISCARD)) {
4218 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4219 if (!blk_queue_discard(q))
4220 ext4_msg(sb, KERN_WARNING,
4221 "mounting with \"discard\" option, but "
4222 "the device does not support discard");
4223 }
4224
4225 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4226 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4227 "Opts: %.*s%s%s", descr,
4228 (int) sizeof(sbi->s_es->s_mount_opts),
4229 sbi->s_es->s_mount_opts,
4230 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4231
4232 if (es->s_error_count)
4233 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4234
4235 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4236 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4237 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4238 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4239
4240 kfree(orig_data);
4241 return 0;
4242
4243 cantfind_ext4:
4244 if (!silent)
4245 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4246 goto failed_mount;
4247
4248 #ifdef CONFIG_QUOTA
4249 failed_mount8:
4250 ext4_unregister_sysfs(sb);
4251 #endif
4252 failed_mount7:
4253 ext4_unregister_li_request(sb);
4254 failed_mount6:
4255 ext4_mb_release(sb);
4256 if (sbi->s_flex_groups)
4257 kvfree(sbi->s_flex_groups);
4258 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4259 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4260 percpu_counter_destroy(&sbi->s_dirs_counter);
4261 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4262 failed_mount5:
4263 ext4_ext_release(sb);
4264 ext4_release_system_zone(sb);
4265 failed_mount4a:
4266 dput(sb->s_root);
4267 sb->s_root = NULL;
4268 failed_mount4:
4269 ext4_msg(sb, KERN_ERR, "mount failed");
4270 if (EXT4_SB(sb)->rsv_conversion_wq)
4271 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4272 failed_mount_wq:
4273 if (sbi->s_mb_cache) {
4274 ext4_xattr_destroy_cache(sbi->s_mb_cache);
4275 sbi->s_mb_cache = NULL;
4276 }
4277 if (sbi->s_journal) {
4278 jbd2_journal_destroy(sbi->s_journal);
4279 sbi->s_journal = NULL;
4280 }
4281 failed_mount3a:
4282 ext4_es_unregister_shrinker(sbi);
4283 failed_mount3:
4284 del_timer_sync(&sbi->s_err_report);
4285 if (sbi->s_mmp_tsk)
4286 kthread_stop(sbi->s_mmp_tsk);
4287 failed_mount2:
4288 for (i = 0; i < db_count; i++)
4289 brelse(sbi->s_group_desc[i]);
4290 kvfree(sbi->s_group_desc);
4291 failed_mount:
4292 if (sbi->s_chksum_driver)
4293 crypto_free_shash(sbi->s_chksum_driver);
4294 #ifdef CONFIG_QUOTA
4295 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4296 kfree(sbi->s_qf_names[i]);
4297 #endif
4298 ext4_blkdev_remove(sbi);
4299 brelse(bh);
4300 out_fail:
4301 sb->s_fs_info = NULL;
4302 kfree(sbi->s_blockgroup_lock);
4303 out_free_base:
4304 kfree(sbi);
4305 kfree(orig_data);
4306 return err ? err : ret;
4307 }
4308
4309 /*
4310 * Setup any per-fs journal parameters now. We'll do this both on
4311 * initial mount, once the journal has been initialised but before we've
4312 * done any recovery; and again on any subsequent remount.
4313 */
4314 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4315 {
4316 struct ext4_sb_info *sbi = EXT4_SB(sb);
4317
4318 journal->j_commit_interval = sbi->s_commit_interval;
4319 journal->j_min_batch_time = sbi->s_min_batch_time;
4320 journal->j_max_batch_time = sbi->s_max_batch_time;
4321
4322 write_lock(&journal->j_state_lock);
4323 if (test_opt(sb, BARRIER))
4324 journal->j_flags |= JBD2_BARRIER;
4325 else
4326 journal->j_flags &= ~JBD2_BARRIER;
4327 if (test_opt(sb, DATA_ERR_ABORT))
4328 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4329 else
4330 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4331 write_unlock(&journal->j_state_lock);
4332 }
4333
4334 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4335 unsigned int journal_inum)
4336 {
4337 struct inode *journal_inode;
4338
4339 /*
4340 * Test for the existence of a valid inode on disk. Bad things
4341 * happen if we iget() an unused inode, as the subsequent iput()
4342 * will try to delete it.
4343 */
4344 journal_inode = ext4_iget(sb, journal_inum);
4345 if (IS_ERR(journal_inode)) {
4346 ext4_msg(sb, KERN_ERR, "no journal found");
4347 return NULL;
4348 }
4349 if (!journal_inode->i_nlink) {
4350 make_bad_inode(journal_inode);
4351 iput(journal_inode);
4352 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4353 return NULL;
4354 }
4355
4356 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4357 journal_inode, journal_inode->i_size);
4358 if (!S_ISREG(journal_inode->i_mode)) {
4359 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4360 iput(journal_inode);
4361 return NULL;
4362 }
4363 return journal_inode;
4364 }
4365
4366 static journal_t *ext4_get_journal(struct super_block *sb,
4367 unsigned int journal_inum)
4368 {
4369 struct inode *journal_inode;
4370 journal_t *journal;
4371
4372 BUG_ON(!ext4_has_feature_journal(sb));
4373
4374 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4375 if (!journal_inode)
4376 return NULL;
4377
4378 journal = jbd2_journal_init_inode(journal_inode);
4379 if (!journal) {
4380 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4381 iput(journal_inode);
4382 return NULL;
4383 }
4384 journal->j_private = sb;
4385 ext4_init_journal_params(sb, journal);
4386 return journal;
4387 }
4388
4389 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4390 dev_t j_dev)
4391 {
4392 struct buffer_head *bh;
4393 journal_t *journal;
4394 ext4_fsblk_t start;
4395 ext4_fsblk_t len;
4396 int hblock, blocksize;
4397 ext4_fsblk_t sb_block;
4398 unsigned long offset;
4399 struct ext4_super_block *es;
4400 struct block_device *bdev;
4401
4402 BUG_ON(!ext4_has_feature_journal(sb));
4403
4404 bdev = ext4_blkdev_get(j_dev, sb);
4405 if (bdev == NULL)
4406 return NULL;
4407
4408 blocksize = sb->s_blocksize;
4409 hblock = bdev_logical_block_size(bdev);
4410 if (blocksize < hblock) {
4411 ext4_msg(sb, KERN_ERR,
4412 "blocksize too small for journal device");
4413 goto out_bdev;
4414 }
4415
4416 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4417 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4418 set_blocksize(bdev, blocksize);
4419 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4420 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4421 "external journal");
4422 goto out_bdev;
4423 }
4424
4425 es = (struct ext4_super_block *) (bh->b_data + offset);
4426 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4427 !(le32_to_cpu(es->s_feature_incompat) &
4428 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4429 ext4_msg(sb, KERN_ERR, "external journal has "
4430 "bad superblock");
4431 brelse(bh);
4432 goto out_bdev;
4433 }
4434
4435 if ((le32_to_cpu(es->s_feature_ro_compat) &
4436 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4437 es->s_checksum != ext4_superblock_csum(sb, es)) {
4438 ext4_msg(sb, KERN_ERR, "external journal has "
4439 "corrupt superblock");
4440 brelse(bh);
4441 goto out_bdev;
4442 }
4443
4444 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4445 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4446 brelse(bh);
4447 goto out_bdev;
4448 }
4449
4450 len = ext4_blocks_count(es);
4451 start = sb_block + 1;
4452 brelse(bh); /* we're done with the superblock */
4453
4454 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4455 start, len, blocksize);
4456 if (!journal) {
4457 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4458 goto out_bdev;
4459 }
4460 journal->j_private = sb;
4461 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4462 wait_on_buffer(journal->j_sb_buffer);
4463 if (!buffer_uptodate(journal->j_sb_buffer)) {
4464 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4465 goto out_journal;
4466 }
4467 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4468 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4469 "user (unsupported) - %d",
4470 be32_to_cpu(journal->j_superblock->s_nr_users));
4471 goto out_journal;
4472 }
4473 EXT4_SB(sb)->journal_bdev = bdev;
4474 ext4_init_journal_params(sb, journal);
4475 return journal;
4476
4477 out_journal:
4478 jbd2_journal_destroy(journal);
4479 out_bdev:
4480 ext4_blkdev_put(bdev);
4481 return NULL;
4482 }
4483
4484 static int ext4_load_journal(struct super_block *sb,
4485 struct ext4_super_block *es,
4486 unsigned long journal_devnum)
4487 {
4488 journal_t *journal;
4489 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4490 dev_t journal_dev;
4491 int err = 0;
4492 int really_read_only;
4493
4494 BUG_ON(!ext4_has_feature_journal(sb));
4495
4496 if (journal_devnum &&
4497 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4498 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4499 "numbers have changed");
4500 journal_dev = new_decode_dev(journal_devnum);
4501 } else
4502 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4503
4504 really_read_only = bdev_read_only(sb->s_bdev);
4505
4506 /*
4507 * Are we loading a blank journal or performing recovery after a
4508 * crash? For recovery, we need to check in advance whether we
4509 * can get read-write access to the device.
4510 */
4511 if (ext4_has_feature_journal_needs_recovery(sb)) {
4512 if (sb->s_flags & MS_RDONLY) {
4513 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4514 "required on readonly filesystem");
4515 if (really_read_only) {
4516 ext4_msg(sb, KERN_ERR, "write access "
4517 "unavailable, cannot proceed");
4518 return -EROFS;
4519 }
4520 ext4_msg(sb, KERN_INFO, "write access will "
4521 "be enabled during recovery");
4522 }
4523 }
4524
4525 if (journal_inum && journal_dev) {
4526 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4527 "and inode journals!");
4528 return -EINVAL;
4529 }
4530
4531 if (journal_inum) {
4532 if (!(journal = ext4_get_journal(sb, journal_inum)))
4533 return -EINVAL;
4534 } else {
4535 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4536 return -EINVAL;
4537 }
4538
4539 if (!(journal->j_flags & JBD2_BARRIER))
4540 ext4_msg(sb, KERN_INFO, "barriers disabled");
4541
4542 if (!ext4_has_feature_journal_needs_recovery(sb))
4543 err = jbd2_journal_wipe(journal, !really_read_only);
4544 if (!err) {
4545 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4546 if (save)
4547 memcpy(save, ((char *) es) +
4548 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4549 err = jbd2_journal_load(journal);
4550 if (save)
4551 memcpy(((char *) es) + EXT4_S_ERR_START,
4552 save, EXT4_S_ERR_LEN);
4553 kfree(save);
4554 }
4555
4556 if (err) {
4557 ext4_msg(sb, KERN_ERR, "error loading journal");
4558 jbd2_journal_destroy(journal);
4559 return err;
4560 }
4561
4562 EXT4_SB(sb)->s_journal = journal;
4563 ext4_clear_journal_err(sb, es);
4564
4565 if (!really_read_only && journal_devnum &&
4566 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4567 es->s_journal_dev = cpu_to_le32(journal_devnum);
4568
4569 /* Make sure we flush the recovery flag to disk. */
4570 ext4_commit_super(sb, 1);
4571 }
4572
4573 return 0;
4574 }
4575
4576 static int ext4_commit_super(struct super_block *sb, int sync)
4577 {
4578 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4579 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4580 int error = 0;
4581
4582 if (!sbh || block_device_ejected(sb))
4583 return error;
4584 /*
4585 * If the file system is mounted read-only, don't update the
4586 * superblock write time. This avoids updating the superblock
4587 * write time when we are mounting the root file system
4588 * read/only but we need to replay the journal; at that point,
4589 * for people who are east of GMT and who make their clock
4590 * tick in localtime for Windows bug-for-bug compatibility,
4591 * the clock is set in the future, and this will cause e2fsck
4592 * to complain and force a full file system check.
4593 */
4594 if (!(sb->s_flags & MS_RDONLY))
4595 es->s_wtime = cpu_to_le32(get_seconds());
4596 if (sb->s_bdev->bd_part)
4597 es->s_kbytes_written =
4598 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4599 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4600 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4601 else
4602 es->s_kbytes_written =
4603 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4604 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4605 ext4_free_blocks_count_set(es,
4606 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4607 &EXT4_SB(sb)->s_freeclusters_counter)));
4608 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4609 es->s_free_inodes_count =
4610 cpu_to_le32(percpu_counter_sum_positive(
4611 &EXT4_SB(sb)->s_freeinodes_counter));
4612 BUFFER_TRACE(sbh, "marking dirty");
4613 ext4_superblock_csum_set(sb);
4614 if (sync)
4615 lock_buffer(sbh);
4616 if (buffer_write_io_error(sbh)) {
4617 /*
4618 * Oh, dear. A previous attempt to write the
4619 * superblock failed. This could happen because the
4620 * USB device was yanked out. Or it could happen to
4621 * be a transient write error and maybe the block will
4622 * be remapped. Nothing we can do but to retry the
4623 * write and hope for the best.
4624 */
4625 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4626 "superblock detected");
4627 clear_buffer_write_io_error(sbh);
4628 set_buffer_uptodate(sbh);
4629 }
4630 mark_buffer_dirty(sbh);
4631 if (sync) {
4632 unlock_buffer(sbh);
4633 error = __sync_dirty_buffer(sbh,
4634 test_opt(sb, BARRIER) ? REQ_FUA : REQ_SYNC);
4635 if (error)
4636 return error;
4637
4638 error = buffer_write_io_error(sbh);
4639 if (error) {
4640 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4641 "superblock");
4642 clear_buffer_write_io_error(sbh);
4643 set_buffer_uptodate(sbh);
4644 }
4645 }
4646 return error;
4647 }
4648
4649 /*
4650 * Have we just finished recovery? If so, and if we are mounting (or
4651 * remounting) the filesystem readonly, then we will end up with a
4652 * consistent fs on disk. Record that fact.
4653 */
4654 static void ext4_mark_recovery_complete(struct super_block *sb,
4655 struct ext4_super_block *es)
4656 {
4657 journal_t *journal = EXT4_SB(sb)->s_journal;
4658
4659 if (!ext4_has_feature_journal(sb)) {
4660 BUG_ON(journal != NULL);
4661 return;
4662 }
4663 jbd2_journal_lock_updates(journal);
4664 if (jbd2_journal_flush(journal) < 0)
4665 goto out;
4666
4667 if (ext4_has_feature_journal_needs_recovery(sb) &&
4668 sb->s_flags & MS_RDONLY) {
4669 ext4_clear_feature_journal_needs_recovery(sb);
4670 ext4_commit_super(sb, 1);
4671 }
4672
4673 out:
4674 jbd2_journal_unlock_updates(journal);
4675 }
4676
4677 /*
4678 * If we are mounting (or read-write remounting) a filesystem whose journal
4679 * has recorded an error from a previous lifetime, move that error to the
4680 * main filesystem now.
4681 */
4682 static void ext4_clear_journal_err(struct super_block *sb,
4683 struct ext4_super_block *es)
4684 {
4685 journal_t *journal;
4686 int j_errno;
4687 const char *errstr;
4688
4689 BUG_ON(!ext4_has_feature_journal(sb));
4690
4691 journal = EXT4_SB(sb)->s_journal;
4692
4693 /*
4694 * Now check for any error status which may have been recorded in the
4695 * journal by a prior ext4_error() or ext4_abort()
4696 */
4697
4698 j_errno = jbd2_journal_errno(journal);
4699 if (j_errno) {
4700 char nbuf[16];
4701
4702 errstr = ext4_decode_error(sb, j_errno, nbuf);
4703 ext4_warning(sb, "Filesystem error recorded "
4704 "from previous mount: %s", errstr);
4705 ext4_warning(sb, "Marking fs in need of filesystem check.");
4706
4707 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4708 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4709 ext4_commit_super(sb, 1);
4710
4711 jbd2_journal_clear_err(journal);
4712 jbd2_journal_update_sb_errno(journal);
4713 }
4714 }
4715
4716 /*
4717 * Force the running and committing transactions to commit,
4718 * and wait on the commit.
4719 */
4720 int ext4_force_commit(struct super_block *sb)
4721 {
4722 journal_t *journal;
4723
4724 if (sb->s_flags & MS_RDONLY)
4725 return 0;
4726
4727 journal = EXT4_SB(sb)->s_journal;
4728 return ext4_journal_force_commit(journal);
4729 }
4730
4731 static int ext4_sync_fs(struct super_block *sb, int wait)
4732 {
4733 int ret = 0;
4734 tid_t target;
4735 bool needs_barrier = false;
4736 struct ext4_sb_info *sbi = EXT4_SB(sb);
4737
4738 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4739 return 0;
4740
4741 trace_ext4_sync_fs(sb, wait);
4742 flush_workqueue(sbi->rsv_conversion_wq);
4743 /*
4744 * Writeback quota in non-journalled quota case - journalled quota has
4745 * no dirty dquots
4746 */
4747 dquot_writeback_dquots(sb, -1);
4748 /*
4749 * Data writeback is possible w/o journal transaction, so barrier must
4750 * being sent at the end of the function. But we can skip it if
4751 * transaction_commit will do it for us.
4752 */
4753 if (sbi->s_journal) {
4754 target = jbd2_get_latest_transaction(sbi->s_journal);
4755 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4756 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4757 needs_barrier = true;
4758
4759 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4760 if (wait)
4761 ret = jbd2_log_wait_commit(sbi->s_journal,
4762 target);
4763 }
4764 } else if (wait && test_opt(sb, BARRIER))
4765 needs_barrier = true;
4766 if (needs_barrier) {
4767 int err;
4768 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4769 if (!ret)
4770 ret = err;
4771 }
4772
4773 return ret;
4774 }
4775
4776 /*
4777 * LVM calls this function before a (read-only) snapshot is created. This
4778 * gives us a chance to flush the journal completely and mark the fs clean.
4779 *
4780 * Note that only this function cannot bring a filesystem to be in a clean
4781 * state independently. It relies on upper layer to stop all data & metadata
4782 * modifications.
4783 */
4784 static int ext4_freeze(struct super_block *sb)
4785 {
4786 int error = 0;
4787 journal_t *journal;
4788
4789 if (sb->s_flags & MS_RDONLY)
4790 return 0;
4791
4792 journal = EXT4_SB(sb)->s_journal;
4793
4794 if (journal) {
4795 /* Now we set up the journal barrier. */
4796 jbd2_journal_lock_updates(journal);
4797
4798 /*
4799 * Don't clear the needs_recovery flag if we failed to
4800 * flush the journal.
4801 */
4802 error = jbd2_journal_flush(journal);
4803 if (error < 0)
4804 goto out;
4805
4806 /* Journal blocked and flushed, clear needs_recovery flag. */
4807 ext4_clear_feature_journal_needs_recovery(sb);
4808 }
4809
4810 error = ext4_commit_super(sb, 1);
4811 out:
4812 if (journal)
4813 /* we rely on upper layer to stop further updates */
4814 jbd2_journal_unlock_updates(journal);
4815 return error;
4816 }
4817
4818 /*
4819 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4820 * flag here, even though the filesystem is not technically dirty yet.
4821 */
4822 static int ext4_unfreeze(struct super_block *sb)
4823 {
4824 if ((sb->s_flags & MS_RDONLY) || ext4_forced_shutdown(EXT4_SB(sb)))
4825 return 0;
4826
4827 if (EXT4_SB(sb)->s_journal) {
4828 /* Reset the needs_recovery flag before the fs is unlocked. */
4829 ext4_set_feature_journal_needs_recovery(sb);
4830 }
4831
4832 ext4_commit_super(sb, 1);
4833 return 0;
4834 }
4835
4836 /*
4837 * Structure to save mount options for ext4_remount's benefit
4838 */
4839 struct ext4_mount_options {
4840 unsigned long s_mount_opt;
4841 unsigned long s_mount_opt2;
4842 kuid_t s_resuid;
4843 kgid_t s_resgid;
4844 unsigned long s_commit_interval;
4845 u32 s_min_batch_time, s_max_batch_time;
4846 #ifdef CONFIG_QUOTA
4847 int s_jquota_fmt;
4848 char *s_qf_names[EXT4_MAXQUOTAS];
4849 #endif
4850 };
4851
4852 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4853 {
4854 struct ext4_super_block *es;
4855 struct ext4_sb_info *sbi = EXT4_SB(sb);
4856 unsigned long old_sb_flags;
4857 struct ext4_mount_options old_opts;
4858 int enable_quota = 0;
4859 ext4_group_t g;
4860 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4861 int err = 0;
4862 #ifdef CONFIG_QUOTA
4863 int i, j;
4864 #endif
4865 char *orig_data = kstrdup(data, GFP_KERNEL);
4866
4867 /* Store the original options */
4868 old_sb_flags = sb->s_flags;
4869 old_opts.s_mount_opt = sbi->s_mount_opt;
4870 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4871 old_opts.s_resuid = sbi->s_resuid;
4872 old_opts.s_resgid = sbi->s_resgid;
4873 old_opts.s_commit_interval = sbi->s_commit_interval;
4874 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4875 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4876 #ifdef CONFIG_QUOTA
4877 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4878 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4879 if (sbi->s_qf_names[i]) {
4880 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4881 GFP_KERNEL);
4882 if (!old_opts.s_qf_names[i]) {
4883 for (j = 0; j < i; j++)
4884 kfree(old_opts.s_qf_names[j]);
4885 kfree(orig_data);
4886 return -ENOMEM;
4887 }
4888 } else
4889 old_opts.s_qf_names[i] = NULL;
4890 #endif
4891 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4892 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4893
4894 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4895 err = -EINVAL;
4896 goto restore_opts;
4897 }
4898
4899 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4900 test_opt(sb, JOURNAL_CHECKSUM)) {
4901 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4902 "during remount not supported; ignoring");
4903 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4904 }
4905
4906 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4907 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4908 ext4_msg(sb, KERN_ERR, "can't mount with "
4909 "both data=journal and delalloc");
4910 err = -EINVAL;
4911 goto restore_opts;
4912 }
4913 if (test_opt(sb, DIOREAD_NOLOCK)) {
4914 ext4_msg(sb, KERN_ERR, "can't mount with "
4915 "both data=journal and dioread_nolock");
4916 err = -EINVAL;
4917 goto restore_opts;
4918 }
4919 if (test_opt(sb, DAX)) {
4920 ext4_msg(sb, KERN_ERR, "can't mount with "
4921 "both data=journal and dax");
4922 err = -EINVAL;
4923 goto restore_opts;
4924 }
4925 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
4926 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4927 ext4_msg(sb, KERN_ERR, "can't mount with "
4928 "journal_async_commit in data=ordered mode");
4929 err = -EINVAL;
4930 goto restore_opts;
4931 }
4932 }
4933
4934 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4935 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4936 "dax flag with busy inodes while remounting");
4937 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4938 }
4939
4940 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4941 ext4_abort(sb, "Abort forced by user");
4942
4943 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4944 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4945
4946 es = sbi->s_es;
4947
4948 if (sbi->s_journal) {
4949 ext4_init_journal_params(sb, sbi->s_journal);
4950 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4951 }
4952
4953 if (*flags & MS_LAZYTIME)
4954 sb->s_flags |= MS_LAZYTIME;
4955
4956 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4957 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4958 err = -EROFS;
4959 goto restore_opts;
4960 }
4961
4962 if (*flags & MS_RDONLY) {
4963 err = sync_filesystem(sb);
4964 if (err < 0)
4965 goto restore_opts;
4966 err = dquot_suspend(sb, -1);
4967 if (err < 0)
4968 goto restore_opts;
4969
4970 /*
4971 * First of all, the unconditional stuff we have to do
4972 * to disable replay of the journal when we next remount
4973 */
4974 sb->s_flags |= MS_RDONLY;
4975
4976 /*
4977 * OK, test if we are remounting a valid rw partition
4978 * readonly, and if so set the rdonly flag and then
4979 * mark the partition as valid again.
4980 */
4981 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4982 (sbi->s_mount_state & EXT4_VALID_FS))
4983 es->s_state = cpu_to_le16(sbi->s_mount_state);
4984
4985 if (sbi->s_journal)
4986 ext4_mark_recovery_complete(sb, es);
4987 } else {
4988 /* Make sure we can mount this feature set readwrite */
4989 if (ext4_has_feature_readonly(sb) ||
4990 !ext4_feature_set_ok(sb, 0)) {
4991 err = -EROFS;
4992 goto restore_opts;
4993 }
4994 /*
4995 * Make sure the group descriptor checksums
4996 * are sane. If they aren't, refuse to remount r/w.
4997 */
4998 for (g = 0; g < sbi->s_groups_count; g++) {
4999 struct ext4_group_desc *gdp =
5000 ext4_get_group_desc(sb, g, NULL);
5001
5002 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5003 ext4_msg(sb, KERN_ERR,
5004 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5005 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5006 le16_to_cpu(gdp->bg_checksum));
5007 err = -EFSBADCRC;
5008 goto restore_opts;
5009 }
5010 }
5011
5012 /*
5013 * If we have an unprocessed orphan list hanging
5014 * around from a previously readonly bdev mount,
5015 * require a full umount/remount for now.
5016 */
5017 if (es->s_last_orphan) {
5018 ext4_msg(sb, KERN_WARNING, "Couldn't "
5019 "remount RDWR because of unprocessed "
5020 "orphan inode list. Please "
5021 "umount/remount instead");
5022 err = -EINVAL;
5023 goto restore_opts;
5024 }
5025
5026 /*
5027 * Mounting a RDONLY partition read-write, so reread
5028 * and store the current valid flag. (It may have
5029 * been changed by e2fsck since we originally mounted
5030 * the partition.)
5031 */
5032 if (sbi->s_journal)
5033 ext4_clear_journal_err(sb, es);
5034 sbi->s_mount_state = le16_to_cpu(es->s_state);
5035 if (!ext4_setup_super(sb, es, 0))
5036 sb->s_flags &= ~MS_RDONLY;
5037 if (ext4_has_feature_mmp(sb))
5038 if (ext4_multi_mount_protect(sb,
5039 le64_to_cpu(es->s_mmp_block))) {
5040 err = -EROFS;
5041 goto restore_opts;
5042 }
5043 enable_quota = 1;
5044 }
5045 }
5046
5047 /*
5048 * Reinitialize lazy itable initialization thread based on
5049 * current settings
5050 */
5051 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5052 ext4_unregister_li_request(sb);
5053 else {
5054 ext4_group_t first_not_zeroed;
5055 first_not_zeroed = ext4_has_uninit_itable(sb);
5056 ext4_register_li_request(sb, first_not_zeroed);
5057 }
5058
5059 ext4_setup_system_zone(sb);
5060 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5061 ext4_commit_super(sb, 1);
5062
5063 #ifdef CONFIG_QUOTA
5064 /* Release old quota file names */
5065 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5066 kfree(old_opts.s_qf_names[i]);
5067 if (enable_quota) {
5068 if (sb_any_quota_suspended(sb))
5069 dquot_resume(sb, -1);
5070 else if (ext4_has_feature_quota(sb)) {
5071 err = ext4_enable_quotas(sb);
5072 if (err)
5073 goto restore_opts;
5074 }
5075 }
5076 #endif
5077
5078 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5079 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5080 kfree(orig_data);
5081 return 0;
5082
5083 restore_opts:
5084 sb->s_flags = old_sb_flags;
5085 sbi->s_mount_opt = old_opts.s_mount_opt;
5086 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5087 sbi->s_resuid = old_opts.s_resuid;
5088 sbi->s_resgid = old_opts.s_resgid;
5089 sbi->s_commit_interval = old_opts.s_commit_interval;
5090 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5091 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5092 #ifdef CONFIG_QUOTA
5093 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5094 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5095 kfree(sbi->s_qf_names[i]);
5096 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5097 }
5098 #endif
5099 kfree(orig_data);
5100 return err;
5101 }
5102
5103 #ifdef CONFIG_QUOTA
5104 static int ext4_statfs_project(struct super_block *sb,
5105 kprojid_t projid, struct kstatfs *buf)
5106 {
5107 struct kqid qid;
5108 struct dquot *dquot;
5109 u64 limit;
5110 u64 curblock;
5111
5112 qid = make_kqid_projid(projid);
5113 dquot = dqget(sb, qid);
5114 if (IS_ERR(dquot))
5115 return PTR_ERR(dquot);
5116 spin_lock(&dq_data_lock);
5117
5118 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5119 dquot->dq_dqb.dqb_bsoftlimit :
5120 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5121 if (limit && buf->f_blocks > limit) {
5122 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5123 buf->f_blocks = limit;
5124 buf->f_bfree = buf->f_bavail =
5125 (buf->f_blocks > curblock) ?
5126 (buf->f_blocks - curblock) : 0;
5127 }
5128
5129 limit = dquot->dq_dqb.dqb_isoftlimit ?
5130 dquot->dq_dqb.dqb_isoftlimit :
5131 dquot->dq_dqb.dqb_ihardlimit;
5132 if (limit && buf->f_files > limit) {
5133 buf->f_files = limit;
5134 buf->f_ffree =
5135 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5136 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5137 }
5138
5139 spin_unlock(&dq_data_lock);
5140 dqput(dquot);
5141 return 0;
5142 }
5143 #endif
5144
5145 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5146 {
5147 struct super_block *sb = dentry->d_sb;
5148 struct ext4_sb_info *sbi = EXT4_SB(sb);
5149 struct ext4_super_block *es = sbi->s_es;
5150 ext4_fsblk_t overhead = 0, resv_blocks;
5151 u64 fsid;
5152 s64 bfree;
5153 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5154
5155 if (!test_opt(sb, MINIX_DF))
5156 overhead = sbi->s_overhead;
5157
5158 buf->f_type = EXT4_SUPER_MAGIC;
5159 buf->f_bsize = sb->s_blocksize;
5160 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5161 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5162 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5163 /* prevent underflow in case that few free space is available */
5164 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5165 buf->f_bavail = buf->f_bfree -
5166 (ext4_r_blocks_count(es) + resv_blocks);
5167 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5168 buf->f_bavail = 0;
5169 buf->f_files = le32_to_cpu(es->s_inodes_count);
5170 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5171 buf->f_namelen = EXT4_NAME_LEN;
5172 fsid = le64_to_cpup((void *)es->s_uuid) ^
5173 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5174 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5175 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5176
5177 #ifdef CONFIG_QUOTA
5178 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5179 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5180 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5181 #endif
5182 return 0;
5183 }
5184
5185 /* Helper function for writing quotas on sync - we need to start transaction
5186 * before quota file is locked for write. Otherwise the are possible deadlocks:
5187 * Process 1 Process 2
5188 * ext4_create() quota_sync()
5189 * jbd2_journal_start() write_dquot()
5190 * dquot_initialize() down(dqio_mutex)
5191 * down(dqio_mutex) jbd2_journal_start()
5192 *
5193 */
5194
5195 #ifdef CONFIG_QUOTA
5196
5197 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5198 {
5199 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5200 }
5201
5202 static int ext4_write_dquot(struct dquot *dquot)
5203 {
5204 int ret, err;
5205 handle_t *handle;
5206 struct inode *inode;
5207
5208 inode = dquot_to_inode(dquot);
5209 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5210 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5211 if (IS_ERR(handle))
5212 return PTR_ERR(handle);
5213 ret = dquot_commit(dquot);
5214 err = ext4_journal_stop(handle);
5215 if (!ret)
5216 ret = err;
5217 return ret;
5218 }
5219
5220 static int ext4_acquire_dquot(struct dquot *dquot)
5221 {
5222 int ret, err;
5223 handle_t *handle;
5224
5225 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5226 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5227 if (IS_ERR(handle))
5228 return PTR_ERR(handle);
5229 ret = dquot_acquire(dquot);
5230 err = ext4_journal_stop(handle);
5231 if (!ret)
5232 ret = err;
5233 return ret;
5234 }
5235
5236 static int ext4_release_dquot(struct dquot *dquot)
5237 {
5238 int ret, err;
5239 handle_t *handle;
5240
5241 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5242 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5243 if (IS_ERR(handle)) {
5244 /* Release dquot anyway to avoid endless cycle in dqput() */
5245 dquot_release(dquot);
5246 return PTR_ERR(handle);
5247 }
5248 ret = dquot_release(dquot);
5249 err = ext4_journal_stop(handle);
5250 if (!ret)
5251 ret = err;
5252 return ret;
5253 }
5254
5255 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5256 {
5257 struct super_block *sb = dquot->dq_sb;
5258 struct ext4_sb_info *sbi = EXT4_SB(sb);
5259
5260 /* Are we journaling quotas? */
5261 if (ext4_has_feature_quota(sb) ||
5262 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5263 dquot_mark_dquot_dirty(dquot);
5264 return ext4_write_dquot(dquot);
5265 } else {
5266 return dquot_mark_dquot_dirty(dquot);
5267 }
5268 }
5269
5270 static int ext4_write_info(struct super_block *sb, int type)
5271 {
5272 int ret, err;
5273 handle_t *handle;
5274
5275 /* Data block + inode block */
5276 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5277 if (IS_ERR(handle))
5278 return PTR_ERR(handle);
5279 ret = dquot_commit_info(sb, type);
5280 err = ext4_journal_stop(handle);
5281 if (!ret)
5282 ret = err;
5283 return ret;
5284 }
5285
5286 /*
5287 * Turn on quotas during mount time - we need to find
5288 * the quota file and such...
5289 */
5290 static int ext4_quota_on_mount(struct super_block *sb, int type)
5291 {
5292 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5293 EXT4_SB(sb)->s_jquota_fmt, type);
5294 }
5295
5296 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5297 {
5298 struct ext4_inode_info *ei = EXT4_I(inode);
5299
5300 /* The first argument of lockdep_set_subclass has to be
5301 * *exactly* the same as the argument to init_rwsem() --- in
5302 * this case, in init_once() --- or lockdep gets unhappy
5303 * because the name of the lock is set using the
5304 * stringification of the argument to init_rwsem().
5305 */
5306 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5307 lockdep_set_subclass(&ei->i_data_sem, subclass);
5308 }
5309
5310 /*
5311 * Standard function to be called on quota_on
5312 */
5313 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5314 const struct path *path)
5315 {
5316 int err;
5317
5318 if (!test_opt(sb, QUOTA))
5319 return -EINVAL;
5320
5321 /* Quotafile not on the same filesystem? */
5322 if (path->dentry->d_sb != sb)
5323 return -EXDEV;
5324 /* Journaling quota? */
5325 if (EXT4_SB(sb)->s_qf_names[type]) {
5326 /* Quotafile not in fs root? */
5327 if (path->dentry->d_parent != sb->s_root)
5328 ext4_msg(sb, KERN_WARNING,
5329 "Quota file not on filesystem root. "
5330 "Journaled quota will not work");
5331 }
5332
5333 /*
5334 * When we journal data on quota file, we have to flush journal to see
5335 * all updates to the file when we bypass pagecache...
5336 */
5337 if (EXT4_SB(sb)->s_journal &&
5338 ext4_should_journal_data(d_inode(path->dentry))) {
5339 /*
5340 * We don't need to lock updates but journal_flush() could
5341 * otherwise be livelocked...
5342 */
5343 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5344 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5345 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5346 if (err)
5347 return err;
5348 }
5349 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5350 err = dquot_quota_on(sb, type, format_id, path);
5351 if (err)
5352 lockdep_set_quota_inode(path->dentry->d_inode,
5353 I_DATA_SEM_NORMAL);
5354 return err;
5355 }
5356
5357 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5358 unsigned int flags)
5359 {
5360 int err;
5361 struct inode *qf_inode;
5362 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5363 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5364 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5365 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5366 };
5367
5368 BUG_ON(!ext4_has_feature_quota(sb));
5369
5370 if (!qf_inums[type])
5371 return -EPERM;
5372
5373 qf_inode = ext4_iget(sb, qf_inums[type]);
5374 if (IS_ERR(qf_inode)) {
5375 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5376 return PTR_ERR(qf_inode);
5377 }
5378
5379 /* Don't account quota for quota files to avoid recursion */
5380 qf_inode->i_flags |= S_NOQUOTA;
5381 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5382 err = dquot_enable(qf_inode, type, format_id, flags);
5383 iput(qf_inode);
5384 if (err)
5385 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5386
5387 return err;
5388 }
5389
5390 /* Enable usage tracking for all quota types. */
5391 static int ext4_enable_quotas(struct super_block *sb)
5392 {
5393 int type, err = 0;
5394 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5395 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5396 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5397 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5398 };
5399 bool quota_mopt[EXT4_MAXQUOTAS] = {
5400 test_opt(sb, USRQUOTA),
5401 test_opt(sb, GRPQUOTA),
5402 test_opt(sb, PRJQUOTA),
5403 };
5404
5405 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5406 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5407 if (qf_inums[type]) {
5408 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5409 DQUOT_USAGE_ENABLED |
5410 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5411 if (err) {
5412 ext4_warning(sb,
5413 "Failed to enable quota tracking "
5414 "(type=%d, err=%d). Please run "
5415 "e2fsck to fix.", type, err);
5416 return err;
5417 }
5418 }
5419 }
5420 return 0;
5421 }
5422
5423 static int ext4_quota_off(struct super_block *sb, int type)
5424 {
5425 struct inode *inode = sb_dqopt(sb)->files[type];
5426 handle_t *handle;
5427
5428 /* Force all delayed allocation blocks to be allocated.
5429 * Caller already holds s_umount sem */
5430 if (test_opt(sb, DELALLOC))
5431 sync_filesystem(sb);
5432
5433 if (!inode)
5434 goto out;
5435
5436 /* Update modification times of quota files when userspace can
5437 * start looking at them */
5438 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5439 if (IS_ERR(handle))
5440 goto out;
5441 inode->i_mtime = inode->i_ctime = current_time(inode);
5442 ext4_mark_inode_dirty(handle, inode);
5443 ext4_journal_stop(handle);
5444
5445 out:
5446 return dquot_quota_off(sb, type);
5447 }
5448
5449 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5450 * acquiring the locks... As quota files are never truncated and quota code
5451 * itself serializes the operations (and no one else should touch the files)
5452 * we don't have to be afraid of races */
5453 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5454 size_t len, loff_t off)
5455 {
5456 struct inode *inode = sb_dqopt(sb)->files[type];
5457 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5458 int offset = off & (sb->s_blocksize - 1);
5459 int tocopy;
5460 size_t toread;
5461 struct buffer_head *bh;
5462 loff_t i_size = i_size_read(inode);
5463
5464 if (off > i_size)
5465 return 0;
5466 if (off+len > i_size)
5467 len = i_size-off;
5468 toread = len;
5469 while (toread > 0) {
5470 tocopy = sb->s_blocksize - offset < toread ?
5471 sb->s_blocksize - offset : toread;
5472 bh = ext4_bread(NULL, inode, blk, 0);
5473 if (IS_ERR(bh))
5474 return PTR_ERR(bh);
5475 if (!bh) /* A hole? */
5476 memset(data, 0, tocopy);
5477 else
5478 memcpy(data, bh->b_data+offset, tocopy);
5479 brelse(bh);
5480 offset = 0;
5481 toread -= tocopy;
5482 data += tocopy;
5483 blk++;
5484 }
5485 return len;
5486 }
5487
5488 /* Write to quotafile (we know the transaction is already started and has
5489 * enough credits) */
5490 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5491 const char *data, size_t len, loff_t off)
5492 {
5493 struct inode *inode = sb_dqopt(sb)->files[type];
5494 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5495 int err, offset = off & (sb->s_blocksize - 1);
5496 int retries = 0;
5497 struct buffer_head *bh;
5498 handle_t *handle = journal_current_handle();
5499
5500 if (EXT4_SB(sb)->s_journal && !handle) {
5501 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5502 " cancelled because transaction is not started",
5503 (unsigned long long)off, (unsigned long long)len);
5504 return -EIO;
5505 }
5506 /*
5507 * Since we account only one data block in transaction credits,
5508 * then it is impossible to cross a block boundary.
5509 */
5510 if (sb->s_blocksize - offset < len) {
5511 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5512 " cancelled because not block aligned",
5513 (unsigned long long)off, (unsigned long long)len);
5514 return -EIO;
5515 }
5516
5517 do {
5518 bh = ext4_bread(handle, inode, blk,
5519 EXT4_GET_BLOCKS_CREATE |
5520 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5521 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5522 ext4_should_retry_alloc(inode->i_sb, &retries));
5523 if (IS_ERR(bh))
5524 return PTR_ERR(bh);
5525 if (!bh)
5526 goto out;
5527 BUFFER_TRACE(bh, "get write access");
5528 err = ext4_journal_get_write_access(handle, bh);
5529 if (err) {
5530 brelse(bh);
5531 return err;
5532 }
5533 lock_buffer(bh);
5534 memcpy(bh->b_data+offset, data, len);
5535 flush_dcache_page(bh->b_page);
5536 unlock_buffer(bh);
5537 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5538 brelse(bh);
5539 out:
5540 if (inode->i_size < off + len) {
5541 i_size_write(inode, off + len);
5542 EXT4_I(inode)->i_disksize = inode->i_size;
5543 ext4_mark_inode_dirty(handle, inode);
5544 }
5545 return len;
5546 }
5547
5548 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5549 {
5550 const struct quota_format_ops *ops;
5551
5552 if (!sb_has_quota_loaded(sb, qid->type))
5553 return -ESRCH;
5554 ops = sb_dqopt(sb)->ops[qid->type];
5555 if (!ops || !ops->get_next_id)
5556 return -ENOSYS;
5557 return dquot_get_next_id(sb, qid);
5558 }
5559 #endif
5560
5561 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5562 const char *dev_name, void *data)
5563 {
5564 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5565 }
5566
5567 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5568 static inline void register_as_ext2(void)
5569 {
5570 int err = register_filesystem(&ext2_fs_type);
5571 if (err)
5572 printk(KERN_WARNING
5573 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5574 }
5575
5576 static inline void unregister_as_ext2(void)
5577 {
5578 unregister_filesystem(&ext2_fs_type);
5579 }
5580
5581 static inline int ext2_feature_set_ok(struct super_block *sb)
5582 {
5583 if (ext4_has_unknown_ext2_incompat_features(sb))
5584 return 0;
5585 if (sb->s_flags & MS_RDONLY)
5586 return 1;
5587 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5588 return 0;
5589 return 1;
5590 }
5591 #else
5592 static inline void register_as_ext2(void) { }
5593 static inline void unregister_as_ext2(void) { }
5594 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5595 #endif
5596
5597 static inline void register_as_ext3(void)
5598 {
5599 int err = register_filesystem(&ext3_fs_type);
5600 if (err)
5601 printk(KERN_WARNING
5602 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5603 }
5604
5605 static inline void unregister_as_ext3(void)
5606 {
5607 unregister_filesystem(&ext3_fs_type);
5608 }
5609
5610 static inline int ext3_feature_set_ok(struct super_block *sb)
5611 {
5612 if (ext4_has_unknown_ext3_incompat_features(sb))
5613 return 0;
5614 if (!ext4_has_feature_journal(sb))
5615 return 0;
5616 if (sb->s_flags & MS_RDONLY)
5617 return 1;
5618 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5619 return 0;
5620 return 1;
5621 }
5622
5623 static struct file_system_type ext4_fs_type = {
5624 .owner = THIS_MODULE,
5625 .name = "ext4",
5626 .mount = ext4_mount,
5627 .kill_sb = kill_block_super,
5628 .fs_flags = FS_REQUIRES_DEV,
5629 };
5630 MODULE_ALIAS_FS("ext4");
5631
5632 /* Shared across all ext4 file systems */
5633 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5634
5635 static int __init ext4_init_fs(void)
5636 {
5637 int i, err;
5638
5639 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5640 ext4_li_info = NULL;
5641 mutex_init(&ext4_li_mtx);
5642
5643 /* Build-time check for flags consistency */
5644 ext4_check_flag_values();
5645
5646 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5647 init_waitqueue_head(&ext4__ioend_wq[i]);
5648
5649 err = ext4_init_es();
5650 if (err)
5651 return err;
5652
5653 err = ext4_init_pageio();
5654 if (err)
5655 goto out5;
5656
5657 err = ext4_init_system_zone();
5658 if (err)
5659 goto out4;
5660
5661 err = ext4_init_sysfs();
5662 if (err)
5663 goto out3;
5664
5665 err = ext4_init_mballoc();
5666 if (err)
5667 goto out2;
5668 err = init_inodecache();
5669 if (err)
5670 goto out1;
5671 register_as_ext3();
5672 register_as_ext2();
5673 err = register_filesystem(&ext4_fs_type);
5674 if (err)
5675 goto out;
5676
5677 return 0;
5678 out:
5679 unregister_as_ext2();
5680 unregister_as_ext3();
5681 destroy_inodecache();
5682 out1:
5683 ext4_exit_mballoc();
5684 out2:
5685 ext4_exit_sysfs();
5686 out3:
5687 ext4_exit_system_zone();
5688 out4:
5689 ext4_exit_pageio();
5690 out5:
5691 ext4_exit_es();
5692
5693 return err;
5694 }
5695
5696 static void __exit ext4_exit_fs(void)
5697 {
5698 ext4_destroy_lazyinit_thread();
5699 unregister_as_ext2();
5700 unregister_as_ext3();
5701 unregister_filesystem(&ext4_fs_type);
5702 destroy_inodecache();
5703 ext4_exit_mballoc();
5704 ext4_exit_sysfs();
5705 ext4_exit_system_zone();
5706 ext4_exit_pageio();
5707 ext4_exit_es();
5708 }
5709
5710 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5711 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5712 MODULE_LICENSE("GPL");
5713 module_init(ext4_init_fs)
5714 module_exit(ext4_exit_fs)
ubuntu-bionic-kernel mirror