?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
gpio: rcar: Fix runtime PM imbalance on error
[linux/fpc-iii.git] / fs / ext4 / super.c
blob9728e7b0e84fcae5b7f0d4753c1572dc08f813bb
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3 * linux/fs/ext4/super.c
4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/fs/minix/inode.c
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 *
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
18 */
19
20 #include <linux/module.h>
21 #include <linux/string.h>
22 #include <linux/fs.h>
23 #include <linux/time.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/dax.h>
42 #include <linux/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
45 #include <linux/unicode.h>
46 #include <linux/part_stat.h>
47 #include <linux/kthread.h>
48 #include <linux/freezer.h>
49
50 #include "ext4.h"
51 #include "ext4_extents.h" /* Needed for trace points definition */
52 #include "ext4_jbd2.h"
53 #include "xattr.h"
54 #include "acl.h"
55 #include "mballoc.h"
56 #include "fsmap.h"
57
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/ext4.h>
60
61 static struct ext4_lazy_init *ext4_li_info;
62 static struct mutex ext4_li_mtx;
63 static struct ratelimit_state ext4_mount_msg_ratelimit;
64
65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
66 unsigned long journal_devnum);
67 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
68 static int ext4_commit_super(struct super_block *sb, int sync);
69 static void ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71 static void ext4_clear_journal_err(struct super_block *sb,
72 struct ext4_super_block *es);
73 static int ext4_sync_fs(struct super_block *sb, int wait);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static struct inode *ext4_get_journal_inode(struct super_block *sb,
87 unsigned int journal_inum);
88
89 /*
90 * Lock ordering
91 *
92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
93 * i_mmap_rwsem (inode->i_mmap_rwsem)!
94 *
95 * page fault path:
96 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
97 * page lock -> i_data_sem (rw)
98 *
99 * buffered write path:
100 * sb_start_write -> i_mutex -> mmap_sem
101 * sb_start_write -> i_mutex -> transaction start -> page lock ->
102 * i_data_sem (rw)
103 *
104 * truncate:
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
107 * i_data_sem (rw)
108 *
109 * direct IO:
110 * sb_start_write -> i_mutex -> mmap_sem
111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
112 *
113 * writepages:
114 * transaction start -> page lock(s) -> i_data_sem (rw)
115 */
116
117 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
118 static struct file_system_type ext2_fs_type = {
119 .owner = THIS_MODULE,
120 .name = "ext2",
121 .mount = ext4_mount,
122 .kill_sb = kill_block_super,
123 .fs_flags = FS_REQUIRES_DEV,
124 };
125 MODULE_ALIAS_FS("ext2");
126 MODULE_ALIAS("ext2");
127 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
128 #else
129 #define IS_EXT2_SB(sb) (0)
130 #endif
131
132
133 static struct file_system_type ext3_fs_type = {
134 .owner = THIS_MODULE,
135 .name = "ext3",
136 .mount = ext4_mount,
137 .kill_sb = kill_block_super,
138 .fs_flags = FS_REQUIRES_DEV,
139 };
140 MODULE_ALIAS_FS("ext3");
141 MODULE_ALIAS("ext3");
142 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
143
144 /*
145 * This works like sb_bread() except it uses ERR_PTR for error
146 * returns. Currently with sb_bread it's impossible to distinguish
147 * between ENOMEM and EIO situations (since both result in a NULL
148 * return.
149 */
150 struct buffer_head *
151 ext4_sb_bread(struct super_block *sb, sector_t block, int op_flags)
152 {
153 struct buffer_head *bh = sb_getblk(sb, block);
154
155 if (bh == NULL)
156 return ERR_PTR(-ENOMEM);
157 if (ext4_buffer_uptodate(bh))
158 return bh;
159 ll_rw_block(REQ_OP_READ, REQ_META | op_flags, 1, &bh);
160 wait_on_buffer(bh);
161 if (buffer_uptodate(bh))
162 return bh;
163 put_bh(bh);
164 return ERR_PTR(-EIO);
165 }
166
167 static int ext4_verify_csum_type(struct super_block *sb,
168 struct ext4_super_block *es)
169 {
170 if (!ext4_has_feature_metadata_csum(sb))
171 return 1;
172
173 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
174 }
175
176 static __le32 ext4_superblock_csum(struct super_block *sb,
177 struct ext4_super_block *es)
178 {
179 struct ext4_sb_info *sbi = EXT4_SB(sb);
180 int offset = offsetof(struct ext4_super_block, s_checksum);
181 __u32 csum;
182
183 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
184
185 return cpu_to_le32(csum);
186 }
187
188 static int ext4_superblock_csum_verify(struct super_block *sb,
189 struct ext4_super_block *es)
190 {
191 if (!ext4_has_metadata_csum(sb))
192 return 1;
193
194 return es->s_checksum == ext4_superblock_csum(sb, es);
195 }
196
197 void ext4_superblock_csum_set(struct super_block *sb)
198 {
199 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
200
201 if (!ext4_has_metadata_csum(sb))
202 return;
203
204 es->s_checksum = ext4_superblock_csum(sb, es);
205 }
206
207 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
208 struct ext4_group_desc *bg)
209 {
210 return le32_to_cpu(bg->bg_block_bitmap_lo) |
211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
212 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
213 }
214
215 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
216 struct ext4_group_desc *bg)
217 {
218 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
220 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
221 }
222
223 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
224 struct ext4_group_desc *bg)
225 {
226 return le32_to_cpu(bg->bg_inode_table_lo) |
227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
228 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
229 }
230
231 __u32 ext4_free_group_clusters(struct super_block *sb,
232 struct ext4_group_desc *bg)
233 {
234 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
236 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
237 }
238
239 __u32 ext4_free_inodes_count(struct super_block *sb,
240 struct ext4_group_desc *bg)
241 {
242 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
244 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
245 }
246
247 __u32 ext4_used_dirs_count(struct super_block *sb,
248 struct ext4_group_desc *bg)
249 {
250 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
251 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
252 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
253 }
254
255 __u32 ext4_itable_unused_count(struct super_block *sb,
256 struct ext4_group_desc *bg)
257 {
258 return le16_to_cpu(bg->bg_itable_unused_lo) |
259 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
260 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
261 }
262
263 void ext4_block_bitmap_set(struct super_block *sb,
264 struct ext4_group_desc *bg, ext4_fsblk_t blk)
265 {
266 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
269 }
270
271 void ext4_inode_bitmap_set(struct super_block *sb,
272 struct ext4_group_desc *bg, ext4_fsblk_t blk)
273 {
274 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
277 }
278
279 void ext4_inode_table_set(struct super_block *sb,
280 struct ext4_group_desc *bg, ext4_fsblk_t blk)
281 {
282 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
285 }
286
287 void ext4_free_group_clusters_set(struct super_block *sb,
288 struct ext4_group_desc *bg, __u32 count)
289 {
290 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
293 }
294
295 void ext4_free_inodes_set(struct super_block *sb,
296 struct ext4_group_desc *bg, __u32 count)
297 {
298 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
301 }
302
303 void ext4_used_dirs_set(struct super_block *sb,
304 struct ext4_group_desc *bg, __u32 count)
305 {
306 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
307 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
308 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
309 }
310
311 void ext4_itable_unused_set(struct super_block *sb,
312 struct ext4_group_desc *bg, __u32 count)
313 {
314 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
315 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
316 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
317 }
318
319 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi)
320 {
321 time64_t now = ktime_get_real_seconds();
322
323 now = clamp_val(now, 0, (1ull << 40) - 1);
324
325 *lo = cpu_to_le32(lower_32_bits(now));
326 *hi = upper_32_bits(now);
327 }
328
329 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
330 {
331 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
332 }
333 #define ext4_update_tstamp(es, tstamp) \
334 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
335 #define ext4_get_tstamp(es, tstamp) \
336 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
337
338 static void __save_error_info(struct super_block *sb, int error,
339 __u32 ino, __u64 block,
340 const char *func, unsigned int line)
341 {
342 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
343 int err;
344
345 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
346 if (bdev_read_only(sb->s_bdev))
347 return;
348 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
349 ext4_update_tstamp(es, s_last_error_time);
350 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
351 es->s_last_error_line = cpu_to_le32(line);
352 es->s_last_error_ino = cpu_to_le32(ino);
353 es->s_last_error_block = cpu_to_le64(block);
354 switch (error) {
355 case EIO:
356 err = EXT4_ERR_EIO;
357 break;
358 case ENOMEM:
359 err = EXT4_ERR_ENOMEM;
360 break;
361 case EFSBADCRC:
362 err = EXT4_ERR_EFSBADCRC;
363 break;
364 case 0:
365 case EFSCORRUPTED:
366 err = EXT4_ERR_EFSCORRUPTED;
367 break;
368 case ENOSPC:
369 err = EXT4_ERR_ENOSPC;
370 break;
371 case ENOKEY:
372 err = EXT4_ERR_ENOKEY;
373 break;
374 case EROFS:
375 err = EXT4_ERR_EROFS;
376 break;
377 case EFBIG:
378 err = EXT4_ERR_EFBIG;
379 break;
380 case EEXIST:
381 err = EXT4_ERR_EEXIST;
382 break;
383 case ERANGE:
384 err = EXT4_ERR_ERANGE;
385 break;
386 case EOVERFLOW:
387 err = EXT4_ERR_EOVERFLOW;
388 break;
389 case EBUSY:
390 err = EXT4_ERR_EBUSY;
391 break;
392 case ENOTDIR:
393 err = EXT4_ERR_ENOTDIR;
394 break;
395 case ENOTEMPTY:
396 err = EXT4_ERR_ENOTEMPTY;
397 break;
398 case ESHUTDOWN:
399 err = EXT4_ERR_ESHUTDOWN;
400 break;
401 case EFAULT:
402 err = EXT4_ERR_EFAULT;
403 break;
404 default:
405 err = EXT4_ERR_UNKNOWN;
406 }
407 es->s_last_error_errcode = err;
408 if (!es->s_first_error_time) {
409 es->s_first_error_time = es->s_last_error_time;
410 es->s_first_error_time_hi = es->s_last_error_time_hi;
411 strncpy(es->s_first_error_func, func,
412 sizeof(es->s_first_error_func));
413 es->s_first_error_line = cpu_to_le32(line);
414 es->s_first_error_ino = es->s_last_error_ino;
415 es->s_first_error_block = es->s_last_error_block;
416 es->s_first_error_errcode = es->s_last_error_errcode;
417 }
418 /*
419 * Start the daily error reporting function if it hasn't been
420 * started already
421 */
422 if (!es->s_error_count)
423 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
424 le32_add_cpu(&es->s_error_count, 1);
425 }
426
427 static void save_error_info(struct super_block *sb, int error,
428 __u32 ino, __u64 block,
429 const char *func, unsigned int line)
430 {
431 __save_error_info(sb, error, ino, block, func, line);
432 if (!bdev_read_only(sb->s_bdev))
433 ext4_commit_super(sb, 1);
434 }
435
436 /*
437 * The del_gendisk() function uninitializes the disk-specific data
438 * structures, including the bdi structure, without telling anyone
439 * else. Once this happens, any attempt to call mark_buffer_dirty()
440 * (for example, by ext4_commit_super), will cause a kernel OOPS.
441 * This is a kludge to prevent these oops until we can put in a proper
442 * hook in del_gendisk() to inform the VFS and file system layers.
443 */
444 static int block_device_ejected(struct super_block *sb)
445 {
446 struct inode *bd_inode = sb->s_bdev->bd_inode;
447 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
448
449 return bdi->dev == NULL;
450 }
451
452 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
453 {
454 struct super_block *sb = journal->j_private;
455 struct ext4_sb_info *sbi = EXT4_SB(sb);
456 int error = is_journal_aborted(journal);
457 struct ext4_journal_cb_entry *jce;
458
459 BUG_ON(txn->t_state == T_FINISHED);
460
461 ext4_process_freed_data(sb, txn->t_tid);
462
463 spin_lock(&sbi->s_md_lock);
464 while (!list_empty(&txn->t_private_list)) {
465 jce = list_entry(txn->t_private_list.next,
466 struct ext4_journal_cb_entry, jce_list);
467 list_del_init(&jce->jce_list);
468 spin_unlock(&sbi->s_md_lock);
469 jce->jce_func(sb, jce, error);
470 spin_lock(&sbi->s_md_lock);
471 }
472 spin_unlock(&sbi->s_md_lock);
473 }
474
475 static bool system_going_down(void)
476 {
477 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
478 || system_state == SYSTEM_RESTART;
479 }
480
481 /* Deal with the reporting of failure conditions on a filesystem such as
482 * inconsistencies detected or read IO failures.
483 *
484 * On ext2, we can store the error state of the filesystem in the
485 * superblock. That is not possible on ext4, because we may have other
486 * write ordering constraints on the superblock which prevent us from
487 * writing it out straight away; and given that the journal is about to
488 * be aborted, we can't rely on the current, or future, transactions to
489 * write out the superblock safely.
490 *
491 * We'll just use the jbd2_journal_abort() error code to record an error in
492 * the journal instead. On recovery, the journal will complain about
493 * that error until we've noted it down and cleared it.
494 */
495
496 static void ext4_handle_error(struct super_block *sb)
497 {
498 if (test_opt(sb, WARN_ON_ERROR))
499 WARN_ON_ONCE(1);
500
501 if (sb_rdonly(sb))
502 return;
503
504 if (!test_opt(sb, ERRORS_CONT)) {
505 journal_t *journal = EXT4_SB(sb)->s_journal;
506
507 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
508 if (journal)
509 jbd2_journal_abort(journal, -EIO);
510 }
511 /*
512 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
513 * could panic during 'reboot -f' as the underlying device got already
514 * disabled.
515 */
516 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
517 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
518 /*
519 * Make sure updated value of ->s_mount_flags will be visible
520 * before ->s_flags update
521 */
522 smp_wmb();
523 sb->s_flags |= SB_RDONLY;
524 } else if (test_opt(sb, ERRORS_PANIC)) {
525 if (EXT4_SB(sb)->s_journal &&
526 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
527 return;
528 panic("EXT4-fs (device %s): panic forced after error\n",
529 sb->s_id);
530 }
531 }
532
533 #define ext4_error_ratelimit(sb) \
534 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
535 "EXT4-fs error")
536
537 void __ext4_error(struct super_block *sb, const char *function,
538 unsigned int line, int error, __u64 block,
539 const char *fmt, ...)
540 {
541 struct va_format vaf;
542 va_list args;
543
544 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
545 return;
546
547 trace_ext4_error(sb, function, line);
548 if (ext4_error_ratelimit(sb)) {
549 va_start(args, fmt);
550 vaf.fmt = fmt;
551 vaf.va = &args;
552 printk(KERN_CRIT
553 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
554 sb->s_id, function, line, current->comm, &vaf);
555 va_end(args);
556 }
557 save_error_info(sb, error, 0, block, function, line);
558 ext4_handle_error(sb);
559 }
560
561 void __ext4_error_inode(struct inode *inode, const char *function,
562 unsigned int line, ext4_fsblk_t block, int error,
563 const char *fmt, ...)
564 {
565 va_list args;
566 struct va_format vaf;
567
568 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
569 return;
570
571 trace_ext4_error(inode->i_sb, function, line);
572 if (ext4_error_ratelimit(inode->i_sb)) {
573 va_start(args, fmt);
574 vaf.fmt = fmt;
575 vaf.va = &args;
576 if (block)
577 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
578 "inode #%lu: block %llu: comm %s: %pV\n",
579 inode->i_sb->s_id, function, line, inode->i_ino,
580 block, current->comm, &vaf);
581 else
582 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
583 "inode #%lu: comm %s: %pV\n",
584 inode->i_sb->s_id, function, line, inode->i_ino,
585 current->comm, &vaf);
586 va_end(args);
587 }
588 save_error_info(inode->i_sb, error, inode->i_ino, block,
589 function, line);
590 ext4_handle_error(inode->i_sb);
591 }
592
593 void __ext4_error_file(struct file *file, const char *function,
594 unsigned int line, ext4_fsblk_t block,
595 const char *fmt, ...)
596 {
597 va_list args;
598 struct va_format vaf;
599 struct ext4_super_block *es;
600 struct inode *inode = file_inode(file);
601 char pathname[80], *path;
602
603 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
604 return;
605
606 trace_ext4_error(inode->i_sb, function, line);
607 es = EXT4_SB(inode->i_sb)->s_es;
608 if (ext4_error_ratelimit(inode->i_sb)) {
609 path = file_path(file, pathname, sizeof(pathname));
610 if (IS_ERR(path))
611 path = "(unknown)";
612 va_start(args, fmt);
613 vaf.fmt = fmt;
614 vaf.va = &args;
615 if (block)
616 printk(KERN_CRIT
617 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
618 "block %llu: comm %s: path %s: %pV\n",
619 inode->i_sb->s_id, function, line, inode->i_ino,
620 block, current->comm, path, &vaf);
621 else
622 printk(KERN_CRIT
623 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
624 "comm %s: path %s: %pV\n",
625 inode->i_sb->s_id, function, line, inode->i_ino,
626 current->comm, path, &vaf);
627 va_end(args);
628 }
629 save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block,
630 function, line);
631 ext4_handle_error(inode->i_sb);
632 }
633
634 const char *ext4_decode_error(struct super_block *sb, int errno,
635 char nbuf[16])
636 {
637 char *errstr = NULL;
638
639 switch (errno) {
640 case -EFSCORRUPTED:
641 errstr = "Corrupt filesystem";
642 break;
643 case -EFSBADCRC:
644 errstr = "Filesystem failed CRC";
645 break;
646 case -EIO:
647 errstr = "IO failure";
648 break;
649 case -ENOMEM:
650 errstr = "Out of memory";
651 break;
652 case -EROFS:
653 if (!sb || (EXT4_SB(sb)->s_journal &&
654 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
655 errstr = "Journal has aborted";
656 else
657 errstr = "Readonly filesystem";
658 break;
659 default:
660 /* If the caller passed in an extra buffer for unknown
661 * errors, textualise them now. Else we just return
662 * NULL. */
663 if (nbuf) {
664 /* Check for truncated error codes... */
665 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
666 errstr = nbuf;
667 }
668 break;
669 }
670
671 return errstr;
672 }
673
674 /* __ext4_std_error decodes expected errors from journaling functions
675 * automatically and invokes the appropriate error response. */
676
677 void __ext4_std_error(struct super_block *sb, const char *function,
678 unsigned int line, int errno)
679 {
680 char nbuf[16];
681 const char *errstr;
682
683 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
684 return;
685
686 /* Special case: if the error is EROFS, and we're not already
687 * inside a transaction, then there's really no point in logging
688 * an error. */
689 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
690 return;
691
692 if (ext4_error_ratelimit(sb)) {
693 errstr = ext4_decode_error(sb, errno, nbuf);
694 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
695 sb->s_id, function, line, errstr);
696 }
697
698 save_error_info(sb, -errno, 0, 0, function, line);
699 ext4_handle_error(sb);
700 }
701
702 /*
703 * ext4_abort is a much stronger failure handler than ext4_error. The
704 * abort function may be used to deal with unrecoverable failures such
705 * as journal IO errors or ENOMEM at a critical moment in log management.
706 *
707 * We unconditionally force the filesystem into an ABORT|READONLY state,
708 * unless the error response on the fs has been set to panic in which
709 * case we take the easy way out and panic immediately.
710 */
711
712 void __ext4_abort(struct super_block *sb, const char *function,
713 unsigned int line, int error, const char *fmt, ...)
714 {
715 struct va_format vaf;
716 va_list args;
717
718 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
719 return;
720
721 save_error_info(sb, error, 0, 0, function, line);
722 va_start(args, fmt);
723 vaf.fmt = fmt;
724 vaf.va = &args;
725 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
726 sb->s_id, function, line, &vaf);
727 va_end(args);
728
729 if (sb_rdonly(sb) == 0) {
730 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
731 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
732 /*
733 * Make sure updated value of ->s_mount_flags will be visible
734 * before ->s_flags update
735 */
736 smp_wmb();
737 sb->s_flags |= SB_RDONLY;
738 if (EXT4_SB(sb)->s_journal)
739 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
740 }
741 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
742 if (EXT4_SB(sb)->s_journal &&
743 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
744 return;
745 panic("EXT4-fs panic from previous error\n");
746 }
747 }
748
749 void __ext4_msg(struct super_block *sb,
750 const char *prefix, const char *fmt, ...)
751 {
752 struct va_format vaf;
753 va_list args;
754
755 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
756 return;
757
758 va_start(args, fmt);
759 vaf.fmt = fmt;
760 vaf.va = &args;
761 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
762 va_end(args);
763 }
764
765 #define ext4_warning_ratelimit(sb) \
766 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
767 "EXT4-fs warning")
768
769 void __ext4_warning(struct super_block *sb, const char *function,
770 unsigned int line, const char *fmt, ...)
771 {
772 struct va_format vaf;
773 va_list args;
774
775 if (!ext4_warning_ratelimit(sb))
776 return;
777
778 va_start(args, fmt);
779 vaf.fmt = fmt;
780 vaf.va = &args;
781 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
782 sb->s_id, function, line, &vaf);
783 va_end(args);
784 }
785
786 void __ext4_warning_inode(const struct inode *inode, const char *function,
787 unsigned int line, const char *fmt, ...)
788 {
789 struct va_format vaf;
790 va_list args;
791
792 if (!ext4_warning_ratelimit(inode->i_sb))
793 return;
794
795 va_start(args, fmt);
796 vaf.fmt = fmt;
797 vaf.va = &args;
798 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
799 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
800 function, line, inode->i_ino, current->comm, &vaf);
801 va_end(args);
802 }
803
804 void __ext4_grp_locked_error(const char *function, unsigned int line,
805 struct super_block *sb, ext4_group_t grp,
806 unsigned long ino, ext4_fsblk_t block,
807 const char *fmt, ...)
808 __releases(bitlock)
809 __acquires(bitlock)
810 {
811 struct va_format vaf;
812 va_list args;
813
814 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
815 return;
816
817 trace_ext4_error(sb, function, line);
818 __save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
819
820 if (ext4_error_ratelimit(sb)) {
821 va_start(args, fmt);
822 vaf.fmt = fmt;
823 vaf.va = &args;
824 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
825 sb->s_id, function, line, grp);
826 if (ino)
827 printk(KERN_CONT "inode %lu: ", ino);
828 if (block)
829 printk(KERN_CONT "block %llu:",
830 (unsigned long long) block);
831 printk(KERN_CONT "%pV\n", &vaf);
832 va_end(args);
833 }
834
835 if (test_opt(sb, WARN_ON_ERROR))
836 WARN_ON_ONCE(1);
837
838 if (test_opt(sb, ERRORS_CONT)) {
839 ext4_commit_super(sb, 0);
840 return;
841 }
842
843 ext4_unlock_group(sb, grp);
844 ext4_commit_super(sb, 1);
845 ext4_handle_error(sb);
846 /*
847 * We only get here in the ERRORS_RO case; relocking the group
848 * may be dangerous, but nothing bad will happen since the
849 * filesystem will have already been marked read/only and the
850 * journal has been aborted. We return 1 as a hint to callers
851 * who might what to use the return value from
852 * ext4_grp_locked_error() to distinguish between the
853 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
854 * aggressively from the ext4 function in question, with a
855 * more appropriate error code.
856 */
857 ext4_lock_group(sb, grp);
858 return;
859 }
860
861 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
862 ext4_group_t group,
863 unsigned int flags)
864 {
865 struct ext4_sb_info *sbi = EXT4_SB(sb);
866 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
867 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
868 int ret;
869
870 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
871 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
872 &grp->bb_state);
873 if (!ret)
874 percpu_counter_sub(&sbi->s_freeclusters_counter,
875 grp->bb_free);
876 }
877
878 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
879 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
880 &grp->bb_state);
881 if (!ret && gdp) {
882 int count;
883
884 count = ext4_free_inodes_count(sb, gdp);
885 percpu_counter_sub(&sbi->s_freeinodes_counter,
886 count);
887 }
888 }
889 }
890
891 void ext4_update_dynamic_rev(struct super_block *sb)
892 {
893 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
894
895 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
896 return;
897
898 ext4_warning(sb,
899 "updating to rev %d because of new feature flag, "
900 "running e2fsck is recommended",
901 EXT4_DYNAMIC_REV);
902
903 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
904 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
905 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
906 /* leave es->s_feature_*compat flags alone */
907 /* es->s_uuid will be set by e2fsck if empty */
908
909 /*
910 * The rest of the superblock fields should be zero, and if not it
911 * means they are likely already in use, so leave them alone. We
912 * can leave it up to e2fsck to clean up any inconsistencies there.
913 */
914 }
915
916 /*
917 * Open the external journal device
918 */
919 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
920 {
921 struct block_device *bdev;
922
923 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
924 if (IS_ERR(bdev))
925 goto fail;
926 return bdev;
927
928 fail:
929 ext4_msg(sb, KERN_ERR,
930 "failed to open journal device unknown-block(%u,%u) %ld",
931 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
932 return NULL;
933 }
934
935 /*
936 * Release the journal device
937 */
938 static void ext4_blkdev_put(struct block_device *bdev)
939 {
940 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
941 }
942
943 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
944 {
945 struct block_device *bdev;
946 bdev = sbi->journal_bdev;
947 if (bdev) {
948 ext4_blkdev_put(bdev);
949 sbi->journal_bdev = NULL;
950 }
951 }
952
953 static inline struct inode *orphan_list_entry(struct list_head *l)
954 {
955 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
956 }
957
958 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
959 {
960 struct list_head *l;
961
962 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
963 le32_to_cpu(sbi->s_es->s_last_orphan));
964
965 printk(KERN_ERR "sb_info orphan list:\n");
966 list_for_each(l, &sbi->s_orphan) {
967 struct inode *inode = orphan_list_entry(l);
968 printk(KERN_ERR " "
969 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
970 inode->i_sb->s_id, inode->i_ino, inode,
971 inode->i_mode, inode->i_nlink,
972 NEXT_ORPHAN(inode));
973 }
974 }
975
976 #ifdef CONFIG_QUOTA
977 static int ext4_quota_off(struct super_block *sb, int type);
978
979 static inline void ext4_quota_off_umount(struct super_block *sb)
980 {
981 int type;
982
983 /* Use our quota_off function to clear inode flags etc. */
984 for (type = 0; type < EXT4_MAXQUOTAS; type++)
985 ext4_quota_off(sb, type);
986 }
987
988 /*
989 * This is a helper function which is used in the mount/remount
990 * codepaths (which holds s_umount) to fetch the quota file name.
991 */
992 static inline char *get_qf_name(struct super_block *sb,
993 struct ext4_sb_info *sbi,
994 int type)
995 {
996 return rcu_dereference_protected(sbi->s_qf_names[type],
997 lockdep_is_held(&sb->s_umount));
998 }
999 #else
1000 static inline void ext4_quota_off_umount(struct super_block *sb)
1001 {
1002 }
1003 #endif
1004
1005 static void ext4_put_super(struct super_block *sb)
1006 {
1007 struct ext4_sb_info *sbi = EXT4_SB(sb);
1008 struct ext4_super_block *es = sbi->s_es;
1009 struct buffer_head **group_desc;
1010 struct flex_groups **flex_groups;
1011 int aborted = 0;
1012 int i, err;
1013
1014 ext4_unregister_li_request(sb);
1015 ext4_quota_off_umount(sb);
1016
1017 destroy_workqueue(sbi->rsv_conversion_wq);
1018
1019 /*
1020 * Unregister sysfs before destroying jbd2 journal.
1021 * Since we could still access attr_journal_task attribute via sysfs
1022 * path which could have sbi->s_journal->j_task as NULL
1023 */
1024 ext4_unregister_sysfs(sb);
1025
1026 if (sbi->s_journal) {
1027 aborted = is_journal_aborted(sbi->s_journal);
1028 err = jbd2_journal_destroy(sbi->s_journal);
1029 sbi->s_journal = NULL;
1030 if ((err < 0) && !aborted) {
1031 ext4_abort(sb, -err, "Couldn't clean up the journal");
1032 }
1033 }
1034
1035 ext4_es_unregister_shrinker(sbi);
1036 del_timer_sync(&sbi->s_err_report);
1037 ext4_release_system_zone(sb);
1038 ext4_mb_release(sb);
1039 ext4_ext_release(sb);
1040
1041 if (!sb_rdonly(sb) && !aborted) {
1042 ext4_clear_feature_journal_needs_recovery(sb);
1043 es->s_state = cpu_to_le16(sbi->s_mount_state);
1044 }
1045 if (!sb_rdonly(sb))
1046 ext4_commit_super(sb, 1);
1047
1048 rcu_read_lock();
1049 group_desc = rcu_dereference(sbi->s_group_desc);
1050 for (i = 0; i < sbi->s_gdb_count; i++)
1051 brelse(group_desc[i]);
1052 kvfree(group_desc);
1053 flex_groups = rcu_dereference(sbi->s_flex_groups);
1054 if (flex_groups) {
1055 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1056 kvfree(flex_groups[i]);
1057 kvfree(flex_groups);
1058 }
1059 rcu_read_unlock();
1060 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1061 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1062 percpu_counter_destroy(&sbi->s_dirs_counter);
1063 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1064 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1065 #ifdef CONFIG_QUOTA
1066 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1067 kfree(get_qf_name(sb, sbi, i));
1068 #endif
1069
1070 /* Debugging code just in case the in-memory inode orphan list
1071 * isn't empty. The on-disk one can be non-empty if we've
1072 * detected an error and taken the fs readonly, but the
1073 * in-memory list had better be clean by this point. */
1074 if (!list_empty(&sbi->s_orphan))
1075 dump_orphan_list(sb, sbi);
1076 J_ASSERT(list_empty(&sbi->s_orphan));
1077
1078 sync_blockdev(sb->s_bdev);
1079 invalidate_bdev(sb->s_bdev);
1080 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
1081 /*
1082 * Invalidate the journal device's buffers. We don't want them
1083 * floating about in memory - the physical journal device may
1084 * hotswapped, and it breaks the `ro-after' testing code.
1085 */
1086 sync_blockdev(sbi->journal_bdev);
1087 invalidate_bdev(sbi->journal_bdev);
1088 ext4_blkdev_remove(sbi);
1089 }
1090
1091 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1092 sbi->s_ea_inode_cache = NULL;
1093
1094 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1095 sbi->s_ea_block_cache = NULL;
1096
1097 if (sbi->s_mmp_tsk)
1098 kthread_stop(sbi->s_mmp_tsk);
1099 brelse(sbi->s_sbh);
1100 sb->s_fs_info = NULL;
1101 /*
1102 * Now that we are completely done shutting down the
1103 * superblock, we need to actually destroy the kobject.
1104 */
1105 kobject_put(&sbi->s_kobj);
1106 wait_for_completion(&sbi->s_kobj_unregister);
1107 if (sbi->s_chksum_driver)
1108 crypto_free_shash(sbi->s_chksum_driver);
1109 kfree(sbi->s_blockgroup_lock);
1110 fs_put_dax(sbi->s_daxdev);
1111 #ifdef CONFIG_UNICODE
1112 utf8_unload(sbi->s_encoding);
1113 #endif
1114 kfree(sbi);
1115 }
1116
1117 static struct kmem_cache *ext4_inode_cachep;
1118
1119 /*
1120 * Called inside transaction, so use GFP_NOFS
1121 */
1122 static struct inode *ext4_alloc_inode(struct super_block *sb)
1123 {
1124 struct ext4_inode_info *ei;
1125
1126 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1127 if (!ei)
1128 return NULL;
1129
1130 inode_set_iversion(&ei->vfs_inode, 1);
1131 spin_lock_init(&ei->i_raw_lock);
1132 INIT_LIST_HEAD(&ei->i_prealloc_list);
1133 spin_lock_init(&ei->i_prealloc_lock);
1134 ext4_es_init_tree(&ei->i_es_tree);
1135 rwlock_init(&ei->i_es_lock);
1136 INIT_LIST_HEAD(&ei->i_es_list);
1137 ei->i_es_all_nr = 0;
1138 ei->i_es_shk_nr = 0;
1139 ei->i_es_shrink_lblk = 0;
1140 ei->i_reserved_data_blocks = 0;
1141 spin_lock_init(&(ei->i_block_reservation_lock));
1142 ext4_init_pending_tree(&ei->i_pending_tree);
1143 #ifdef CONFIG_QUOTA
1144 ei->i_reserved_quota = 0;
1145 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1146 #endif
1147 ei->jinode = NULL;
1148 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1149 spin_lock_init(&ei->i_completed_io_lock);
1150 ei->i_sync_tid = 0;
1151 ei->i_datasync_tid = 0;
1152 atomic_set(&ei->i_unwritten, 0);
1153 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1154 return &ei->vfs_inode;
1155 }
1156
1157 static int ext4_drop_inode(struct inode *inode)
1158 {
1159 int drop = generic_drop_inode(inode);
1160
1161 if (!drop)
1162 drop = fscrypt_drop_inode(inode);
1163
1164 trace_ext4_drop_inode(inode, drop);
1165 return drop;
1166 }
1167
1168 static void ext4_free_in_core_inode(struct inode *inode)
1169 {
1170 fscrypt_free_inode(inode);
1171 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1172 }
1173
1174 static void ext4_destroy_inode(struct inode *inode)
1175 {
1176 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1177 ext4_msg(inode->i_sb, KERN_ERR,
1178 "Inode %lu (%p): orphan list check failed!",
1179 inode->i_ino, EXT4_I(inode));
1180 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1181 EXT4_I(inode), sizeof(struct ext4_inode_info),
1182 true);
1183 dump_stack();
1184 }
1185 }
1186
1187 static void init_once(void *foo)
1188 {
1189 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1190
1191 INIT_LIST_HEAD(&ei->i_orphan);
1192 init_rwsem(&ei->xattr_sem);
1193 init_rwsem(&ei->i_data_sem);
1194 init_rwsem(&ei->i_mmap_sem);
1195 inode_init_once(&ei->vfs_inode);
1196 }
1197
1198 static int __init init_inodecache(void)
1199 {
1200 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1201 sizeof(struct ext4_inode_info), 0,
1202 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1203 SLAB_ACCOUNT),
1204 offsetof(struct ext4_inode_info, i_data),
1205 sizeof_field(struct ext4_inode_info, i_data),
1206 init_once);
1207 if (ext4_inode_cachep == NULL)
1208 return -ENOMEM;
1209 return 0;
1210 }
1211
1212 static void destroy_inodecache(void)
1213 {
1214 /*
1215 * Make sure all delayed rcu free inodes are flushed before we
1216 * destroy cache.
1217 */
1218 rcu_barrier();
1219 kmem_cache_destroy(ext4_inode_cachep);
1220 }
1221
1222 void ext4_clear_inode(struct inode *inode)
1223 {
1224 invalidate_inode_buffers(inode);
1225 clear_inode(inode);
1226 ext4_discard_preallocations(inode);
1227 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1228 dquot_drop(inode);
1229 if (EXT4_I(inode)->jinode) {
1230 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1231 EXT4_I(inode)->jinode);
1232 jbd2_free_inode(EXT4_I(inode)->jinode);
1233 EXT4_I(inode)->jinode = NULL;
1234 }
1235 fscrypt_put_encryption_info(inode);
1236 fsverity_cleanup_inode(inode);
1237 }
1238
1239 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1240 u64 ino, u32 generation)
1241 {
1242 struct inode *inode;
1243
1244 /*
1245 * Currently we don't know the generation for parent directory, so
1246 * a generation of 0 means "accept any"
1247 */
1248 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1249 if (IS_ERR(inode))
1250 return ERR_CAST(inode);
1251 if (generation && inode->i_generation != generation) {
1252 iput(inode);
1253 return ERR_PTR(-ESTALE);
1254 }
1255
1256 return inode;
1257 }
1258
1259 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1260 int fh_len, int fh_type)
1261 {
1262 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1263 ext4_nfs_get_inode);
1264 }
1265
1266 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1267 int fh_len, int fh_type)
1268 {
1269 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1270 ext4_nfs_get_inode);
1271 }
1272
1273 static int ext4_nfs_commit_metadata(struct inode *inode)
1274 {
1275 struct writeback_control wbc = {
1276 .sync_mode = WB_SYNC_ALL
1277 };
1278
1279 trace_ext4_nfs_commit_metadata(inode);
1280 return ext4_write_inode(inode, &wbc);
1281 }
1282
1283 /*
1284 * Try to release metadata pages (indirect blocks, directories) which are
1285 * mapped via the block device. Since these pages could have journal heads
1286 * which would prevent try_to_free_buffers() from freeing them, we must use
1287 * jbd2 layer's try_to_free_buffers() function to release them.
1288 */
1289 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1290 gfp_t wait)
1291 {
1292 journal_t *journal = EXT4_SB(sb)->s_journal;
1293
1294 WARN_ON(PageChecked(page));
1295 if (!page_has_buffers(page))
1296 return 0;
1297 if (journal)
1298 return jbd2_journal_try_to_free_buffers(journal, page,
1299 wait & ~__GFP_DIRECT_RECLAIM);
1300 return try_to_free_buffers(page);
1301 }
1302
1303 #ifdef CONFIG_FS_ENCRYPTION
1304 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1305 {
1306 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1307 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1308 }
1309
1310 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1311 void *fs_data)
1312 {
1313 handle_t *handle = fs_data;
1314 int res, res2, credits, retries = 0;
1315
1316 /*
1317 * Encrypting the root directory is not allowed because e2fsck expects
1318 * lost+found to exist and be unencrypted, and encrypting the root
1319 * directory would imply encrypting the lost+found directory as well as
1320 * the filename "lost+found" itself.
1321 */
1322 if (inode->i_ino == EXT4_ROOT_INO)
1323 return -EPERM;
1324
1325 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1326 return -EINVAL;
1327
1328 res = ext4_convert_inline_data(inode);
1329 if (res)
1330 return res;
1331
1332 /*
1333 * If a journal handle was specified, then the encryption context is
1334 * being set on a new inode via inheritance and is part of a larger
1335 * transaction to create the inode. Otherwise the encryption context is
1336 * being set on an existing inode in its own transaction. Only in the
1337 * latter case should the "retry on ENOSPC" logic be used.
1338 */
1339
1340 if (handle) {
1341 res = ext4_xattr_set_handle(handle, inode,
1342 EXT4_XATTR_INDEX_ENCRYPTION,
1343 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1344 ctx, len, 0);
1345 if (!res) {
1346 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1347 ext4_clear_inode_state(inode,
1348 EXT4_STATE_MAY_INLINE_DATA);
1349 /*
1350 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1351 * S_DAX may be disabled
1352 */
1353 ext4_set_inode_flags(inode);
1354 }
1355 return res;
1356 }
1357
1358 res = dquot_initialize(inode);
1359 if (res)
1360 return res;
1361 retry:
1362 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1363 &credits);
1364 if (res)
1365 return res;
1366
1367 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1368 if (IS_ERR(handle))
1369 return PTR_ERR(handle);
1370
1371 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1372 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1373 ctx, len, 0);
1374 if (!res) {
1375 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1376 /*
1377 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1378 * S_DAX may be disabled
1379 */
1380 ext4_set_inode_flags(inode);
1381 res = ext4_mark_inode_dirty(handle, inode);
1382 if (res)
1383 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1384 }
1385 res2 = ext4_journal_stop(handle);
1386
1387 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1388 goto retry;
1389 if (!res)
1390 res = res2;
1391 return res;
1392 }
1393
1394 static bool ext4_dummy_context(struct inode *inode)
1395 {
1396 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1397 }
1398
1399 static bool ext4_has_stable_inodes(struct super_block *sb)
1400 {
1401 return ext4_has_feature_stable_inodes(sb);
1402 }
1403
1404 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1405 int *ino_bits_ret, int *lblk_bits_ret)
1406 {
1407 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1408 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1409 }
1410
1411 static const struct fscrypt_operations ext4_cryptops = {
1412 .key_prefix = "ext4:",
1413 .get_context = ext4_get_context,
1414 .set_context = ext4_set_context,
1415 .dummy_context = ext4_dummy_context,
1416 .empty_dir = ext4_empty_dir,
1417 .max_namelen = EXT4_NAME_LEN,
1418 .has_stable_inodes = ext4_has_stable_inodes,
1419 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1420 };
1421 #endif
1422
1423 #ifdef CONFIG_QUOTA
1424 static const char * const quotatypes[] = INITQFNAMES;
1425 #define QTYPE2NAME(t) (quotatypes[t])
1426
1427 static int ext4_write_dquot(struct dquot *dquot);
1428 static int ext4_acquire_dquot(struct dquot *dquot);
1429 static int ext4_release_dquot(struct dquot *dquot);
1430 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1431 static int ext4_write_info(struct super_block *sb, int type);
1432 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1433 const struct path *path);
1434 static int ext4_quota_on_mount(struct super_block *sb, int type);
1435 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1436 size_t len, loff_t off);
1437 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1438 const char *data, size_t len, loff_t off);
1439 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1440 unsigned int flags);
1441 static int ext4_enable_quotas(struct super_block *sb);
1442
1443 static struct dquot **ext4_get_dquots(struct inode *inode)
1444 {
1445 return EXT4_I(inode)->i_dquot;
1446 }
1447
1448 static const struct dquot_operations ext4_quota_operations = {
1449 .get_reserved_space = ext4_get_reserved_space,
1450 .write_dquot = ext4_write_dquot,
1451 .acquire_dquot = ext4_acquire_dquot,
1452 .release_dquot = ext4_release_dquot,
1453 .mark_dirty = ext4_mark_dquot_dirty,
1454 .write_info = ext4_write_info,
1455 .alloc_dquot = dquot_alloc,
1456 .destroy_dquot = dquot_destroy,
1457 .get_projid = ext4_get_projid,
1458 .get_inode_usage = ext4_get_inode_usage,
1459 .get_next_id = dquot_get_next_id,
1460 };
1461
1462 static const struct quotactl_ops ext4_qctl_operations = {
1463 .quota_on = ext4_quota_on,
1464 .quota_off = ext4_quota_off,
1465 .quota_sync = dquot_quota_sync,
1466 .get_state = dquot_get_state,
1467 .set_info = dquot_set_dqinfo,
1468 .get_dqblk = dquot_get_dqblk,
1469 .set_dqblk = dquot_set_dqblk,
1470 .get_nextdqblk = dquot_get_next_dqblk,
1471 };
1472 #endif
1473
1474 static const struct super_operations ext4_sops = {
1475 .alloc_inode = ext4_alloc_inode,
1476 .free_inode = ext4_free_in_core_inode,
1477 .destroy_inode = ext4_destroy_inode,
1478 .write_inode = ext4_write_inode,
1479 .dirty_inode = ext4_dirty_inode,
1480 .drop_inode = ext4_drop_inode,
1481 .evict_inode = ext4_evict_inode,
1482 .put_super = ext4_put_super,
1483 .sync_fs = ext4_sync_fs,
1484 .freeze_fs = ext4_freeze,
1485 .unfreeze_fs = ext4_unfreeze,
1486 .statfs = ext4_statfs,
1487 .remount_fs = ext4_remount,
1488 .show_options = ext4_show_options,
1489 #ifdef CONFIG_QUOTA
1490 .quota_read = ext4_quota_read,
1491 .quota_write = ext4_quota_write,
1492 .get_dquots = ext4_get_dquots,
1493 #endif
1494 .bdev_try_to_free_page = bdev_try_to_free_page,
1495 };
1496
1497 static const struct export_operations ext4_export_ops = {
1498 .fh_to_dentry = ext4_fh_to_dentry,
1499 .fh_to_parent = ext4_fh_to_parent,
1500 .get_parent = ext4_get_parent,
1501 .commit_metadata = ext4_nfs_commit_metadata,
1502 };
1503
1504 enum {
1505 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1506 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1507 Opt_nouid32, Opt_debug, Opt_removed,
1508 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1509 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1510 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1511 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1512 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1513 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1514 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1515 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1516 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1517 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1518 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1519 Opt_nowarn_on_error, Opt_mblk_io_submit,
1520 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1521 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1522 Opt_inode_readahead_blks, Opt_journal_ioprio,
1523 Opt_dioread_nolock, Opt_dioread_lock,
1524 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1525 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1526 };
1527
1528 static const match_table_t tokens = {
1529 {Opt_bsd_df, "bsddf"},
1530 {Opt_minix_df, "minixdf"},
1531 {Opt_grpid, "grpid"},
1532 {Opt_grpid, "bsdgroups"},
1533 {Opt_nogrpid, "nogrpid"},
1534 {Opt_nogrpid, "sysvgroups"},
1535 {Opt_resgid, "resgid=%u"},
1536 {Opt_resuid, "resuid=%u"},
1537 {Opt_sb, "sb=%u"},
1538 {Opt_err_cont, "errors=continue"},
1539 {Opt_err_panic, "errors=panic"},
1540 {Opt_err_ro, "errors=remount-ro"},
1541 {Opt_nouid32, "nouid32"},
1542 {Opt_debug, "debug"},
1543 {Opt_removed, "oldalloc"},
1544 {Opt_removed, "orlov"},
1545 {Opt_user_xattr, "user_xattr"},
1546 {Opt_nouser_xattr, "nouser_xattr"},
1547 {Opt_acl, "acl"},
1548 {Opt_noacl, "noacl"},
1549 {Opt_noload, "norecovery"},
1550 {Opt_noload, "noload"},
1551 {Opt_removed, "nobh"},
1552 {Opt_removed, "bh"},
1553 {Opt_commit, "commit=%u"},
1554 {Opt_min_batch_time, "min_batch_time=%u"},
1555 {Opt_max_batch_time, "max_batch_time=%u"},
1556 {Opt_journal_dev, "journal_dev=%u"},
1557 {Opt_journal_path, "journal_path=%s"},
1558 {Opt_journal_checksum, "journal_checksum"},
1559 {Opt_nojournal_checksum, "nojournal_checksum"},
1560 {Opt_journal_async_commit, "journal_async_commit"},
1561 {Opt_abort, "abort"},
1562 {Opt_data_journal, "data=journal"},
1563 {Opt_data_ordered, "data=ordered"},
1564 {Opt_data_writeback, "data=writeback"},
1565 {Opt_data_err_abort, "data_err=abort"},
1566 {Opt_data_err_ignore, "data_err=ignore"},
1567 {Opt_offusrjquota, "usrjquota="},
1568 {Opt_usrjquota, "usrjquota=%s"},
1569 {Opt_offgrpjquota, "grpjquota="},
1570 {Opt_grpjquota, "grpjquota=%s"},
1571 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1572 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1573 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1574 {Opt_grpquota, "grpquota"},
1575 {Opt_noquota, "noquota"},
1576 {Opt_quota, "quota"},
1577 {Opt_usrquota, "usrquota"},
1578 {Opt_prjquota, "prjquota"},
1579 {Opt_barrier, "barrier=%u"},
1580 {Opt_barrier, "barrier"},
1581 {Opt_nobarrier, "nobarrier"},
1582 {Opt_i_version, "i_version"},
1583 {Opt_dax, "dax"},
1584 {Opt_stripe, "stripe=%u"},
1585 {Opt_delalloc, "delalloc"},
1586 {Opt_warn_on_error, "warn_on_error"},
1587 {Opt_nowarn_on_error, "nowarn_on_error"},
1588 {Opt_lazytime, "lazytime"},
1589 {Opt_nolazytime, "nolazytime"},
1590 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1591 {Opt_nodelalloc, "nodelalloc"},
1592 {Opt_removed, "mblk_io_submit"},
1593 {Opt_removed, "nomblk_io_submit"},
1594 {Opt_block_validity, "block_validity"},
1595 {Opt_noblock_validity, "noblock_validity"},
1596 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1597 {Opt_journal_ioprio, "journal_ioprio=%u"},
1598 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1599 {Opt_auto_da_alloc, "auto_da_alloc"},
1600 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1601 {Opt_dioread_nolock, "dioread_nolock"},
1602 {Opt_dioread_lock, "nodioread_nolock"},
1603 {Opt_dioread_lock, "dioread_lock"},
1604 {Opt_discard, "discard"},
1605 {Opt_nodiscard, "nodiscard"},
1606 {Opt_init_itable, "init_itable=%u"},
1607 {Opt_init_itable, "init_itable"},
1608 {Opt_noinit_itable, "noinit_itable"},
1609 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1610 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1611 {Opt_nombcache, "nombcache"},
1612 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1613 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1614 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1615 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1616 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1617 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1618 {Opt_err, NULL},
1619 };
1620
1621 static ext4_fsblk_t get_sb_block(void **data)
1622 {
1623 ext4_fsblk_t sb_block;
1624 char *options = (char *) *data;
1625
1626 if (!options || strncmp(options, "sb=", 3) != 0)
1627 return 1; /* Default location */
1628
1629 options += 3;
1630 /* TODO: use simple_strtoll with >32bit ext4 */
1631 sb_block = simple_strtoul(options, &options, 0);
1632 if (*options && *options != ',') {
1633 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1634 (char *) *data);
1635 return 1;
1636 }
1637 if (*options == ',')
1638 options++;
1639 *data = (void *) options;
1640
1641 return sb_block;
1642 }
1643
1644 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1645 static const char deprecated_msg[] =
1646 "Mount option \"%s\" will be removed by %s\n"
1647 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1648
1649 #ifdef CONFIG_QUOTA
1650 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1651 {
1652 struct ext4_sb_info *sbi = EXT4_SB(sb);
1653 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1654 int ret = -1;
1655
1656 if (sb_any_quota_loaded(sb) && !old_qname) {
1657 ext4_msg(sb, KERN_ERR,
1658 "Cannot change journaled "
1659 "quota options when quota turned on");
1660 return -1;
1661 }
1662 if (ext4_has_feature_quota(sb)) {
1663 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1664 "ignored when QUOTA feature is enabled");
1665 return 1;
1666 }
1667 qname = match_strdup(args);
1668 if (!qname) {
1669 ext4_msg(sb, KERN_ERR,
1670 "Not enough memory for storing quotafile name");
1671 return -1;
1672 }
1673 if (old_qname) {
1674 if (strcmp(old_qname, qname) == 0)
1675 ret = 1;
1676 else
1677 ext4_msg(sb, KERN_ERR,
1678 "%s quota file already specified",
1679 QTYPE2NAME(qtype));
1680 goto errout;
1681 }
1682 if (strchr(qname, '/')) {
1683 ext4_msg(sb, KERN_ERR,
1684 "quotafile must be on filesystem root");
1685 goto errout;
1686 }
1687 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1688 set_opt(sb, QUOTA);
1689 return 1;
1690 errout:
1691 kfree(qname);
1692 return ret;
1693 }
1694
1695 static int clear_qf_name(struct super_block *sb, int qtype)
1696 {
1697
1698 struct ext4_sb_info *sbi = EXT4_SB(sb);
1699 char *old_qname = get_qf_name(sb, sbi, qtype);
1700
1701 if (sb_any_quota_loaded(sb) && old_qname) {
1702 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1703 " when quota turned on");
1704 return -1;
1705 }
1706 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1707 synchronize_rcu();
1708 kfree(old_qname);
1709 return 1;
1710 }
1711 #endif
1712
1713 #define MOPT_SET 0x0001
1714 #define MOPT_CLEAR 0x0002
1715 #define MOPT_NOSUPPORT 0x0004
1716 #define MOPT_EXPLICIT 0x0008
1717 #define MOPT_CLEAR_ERR 0x0010
1718 #define MOPT_GTE0 0x0020
1719 #ifdef CONFIG_QUOTA
1720 #define MOPT_Q 0
1721 #define MOPT_QFMT 0x0040
1722 #else
1723 #define MOPT_Q MOPT_NOSUPPORT
1724 #define MOPT_QFMT MOPT_NOSUPPORT
1725 #endif
1726 #define MOPT_DATAJ 0x0080
1727 #define MOPT_NO_EXT2 0x0100
1728 #define MOPT_NO_EXT3 0x0200
1729 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1730 #define MOPT_STRING 0x0400
1731
1732 static const struct mount_opts {
1733 int token;
1734 int mount_opt;
1735 int flags;
1736 } ext4_mount_opts[] = {
1737 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1738 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1739 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1740 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1741 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1742 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1743 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1744 MOPT_EXT4_ONLY | MOPT_SET},
1745 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1746 MOPT_EXT4_ONLY | MOPT_CLEAR},
1747 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1748 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1749 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1750 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1751 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1752 MOPT_EXT4_ONLY | MOPT_CLEAR},
1753 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1754 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1755 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1756 MOPT_EXT4_ONLY | MOPT_CLEAR},
1757 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1758 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1759 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1760 EXT4_MOUNT_JOURNAL_CHECKSUM),
1761 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1762 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1763 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1764 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1765 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1766 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1767 MOPT_NO_EXT2},
1768 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1769 MOPT_NO_EXT2},
1770 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1771 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1772 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1773 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1774 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1775 {Opt_commit, 0, MOPT_GTE0},
1776 {Opt_max_batch_time, 0, MOPT_GTE0},
1777 {Opt_min_batch_time, 0, MOPT_GTE0},
1778 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1779 {Opt_init_itable, 0, MOPT_GTE0},
1780 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1781 {Opt_stripe, 0, MOPT_GTE0},
1782 {Opt_resuid, 0, MOPT_GTE0},
1783 {Opt_resgid, 0, MOPT_GTE0},
1784 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1785 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1786 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1787 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1788 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1789 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1790 MOPT_NO_EXT2 | MOPT_DATAJ},
1791 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1792 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1793 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1794 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1795 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1796 #else
1797 {Opt_acl, 0, MOPT_NOSUPPORT},
1798 {Opt_noacl, 0, MOPT_NOSUPPORT},
1799 #endif
1800 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1801 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1802 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1803 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1804 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1805 MOPT_SET | MOPT_Q},
1806 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1807 MOPT_SET | MOPT_Q},
1808 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1809 MOPT_SET | MOPT_Q},
1810 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1811 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1812 MOPT_CLEAR | MOPT_Q},
1813 {Opt_usrjquota, 0, MOPT_Q},
1814 {Opt_grpjquota, 0, MOPT_Q},
1815 {Opt_offusrjquota, 0, MOPT_Q},
1816 {Opt_offgrpjquota, 0, MOPT_Q},
1817 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1818 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1819 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1820 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1821 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1822 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1823 {Opt_err, 0, 0}
1824 };
1825
1826 #ifdef CONFIG_UNICODE
1827 static const struct ext4_sb_encodings {
1828 __u16 magic;
1829 char *name;
1830 char *version;
1831 } ext4_sb_encoding_map[] = {
1832 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1833 };
1834
1835 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1836 const struct ext4_sb_encodings **encoding,
1837 __u16 *flags)
1838 {
1839 __u16 magic = le16_to_cpu(es->s_encoding);
1840 int i;
1841
1842 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1843 if (magic == ext4_sb_encoding_map[i].magic)
1844 break;
1845
1846 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1847 return -EINVAL;
1848
1849 *encoding = &ext4_sb_encoding_map[i];
1850 *flags = le16_to_cpu(es->s_encoding_flags);
1851
1852 return 0;
1853 }
1854 #endif
1855
1856 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1857 substring_t *args, unsigned long *journal_devnum,
1858 unsigned int *journal_ioprio, int is_remount)
1859 {
1860 struct ext4_sb_info *sbi = EXT4_SB(sb);
1861 const struct mount_opts *m;
1862 kuid_t uid;
1863 kgid_t gid;
1864 int arg = 0;
1865
1866 #ifdef CONFIG_QUOTA
1867 if (token == Opt_usrjquota)
1868 return set_qf_name(sb, USRQUOTA, &args[0]);
1869 else if (token == Opt_grpjquota)
1870 return set_qf_name(sb, GRPQUOTA, &args[0]);
1871 else if (token == Opt_offusrjquota)
1872 return clear_qf_name(sb, USRQUOTA);
1873 else if (token == Opt_offgrpjquota)
1874 return clear_qf_name(sb, GRPQUOTA);
1875 #endif
1876 switch (token) {
1877 case Opt_noacl:
1878 case Opt_nouser_xattr:
1879 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1880 break;
1881 case Opt_sb:
1882 return 1; /* handled by get_sb_block() */
1883 case Opt_removed:
1884 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1885 return 1;
1886 case Opt_abort:
1887 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1888 return 1;
1889 case Opt_i_version:
1890 sb->s_flags |= SB_I_VERSION;
1891 return 1;
1892 case Opt_lazytime:
1893 sb->s_flags |= SB_LAZYTIME;
1894 return 1;
1895 case Opt_nolazytime:
1896 sb->s_flags &= ~SB_LAZYTIME;
1897 return 1;
1898 }
1899
1900 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1901 if (token == m->token)
1902 break;
1903
1904 if (m->token == Opt_err) {
1905 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1906 "or missing value", opt);
1907 return -1;
1908 }
1909
1910 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1911 ext4_msg(sb, KERN_ERR,
1912 "Mount option \"%s\" incompatible with ext2", opt);
1913 return -1;
1914 }
1915 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1916 ext4_msg(sb, KERN_ERR,
1917 "Mount option \"%s\" incompatible with ext3", opt);
1918 return -1;
1919 }
1920
1921 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1922 return -1;
1923 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1924 return -1;
1925 if (m->flags & MOPT_EXPLICIT) {
1926 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1927 set_opt2(sb, EXPLICIT_DELALLOC);
1928 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1929 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1930 } else
1931 return -1;
1932 }
1933 if (m->flags & MOPT_CLEAR_ERR)
1934 clear_opt(sb, ERRORS_MASK);
1935 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1936 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1937 "options when quota turned on");
1938 return -1;
1939 }
1940
1941 if (m->flags & MOPT_NOSUPPORT) {
1942 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1943 } else if (token == Opt_commit) {
1944 if (arg == 0)
1945 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1946 else if (arg > INT_MAX / HZ) {
1947 ext4_msg(sb, KERN_ERR,
1948 "Invalid commit interval %d, "
1949 "must be smaller than %d",
1950 arg, INT_MAX / HZ);
1951 return -1;
1952 }
1953 sbi->s_commit_interval = HZ * arg;
1954 } else if (token == Opt_debug_want_extra_isize) {
1955 if ((arg & 1) ||
1956 (arg < 4) ||
1957 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
1958 ext4_msg(sb, KERN_ERR,
1959 "Invalid want_extra_isize %d", arg);
1960 return -1;
1961 }
1962 sbi->s_want_extra_isize = arg;
1963 } else if (token == Opt_max_batch_time) {
1964 sbi->s_max_batch_time = arg;
1965 } else if (token == Opt_min_batch_time) {
1966 sbi->s_min_batch_time = arg;
1967 } else if (token == Opt_inode_readahead_blks) {
1968 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1969 ext4_msg(sb, KERN_ERR,
1970 "EXT4-fs: inode_readahead_blks must be "
1971 "0 or a power of 2 smaller than 2^31");
1972 return -1;
1973 }
1974 sbi->s_inode_readahead_blks = arg;
1975 } else if (token == Opt_init_itable) {
1976 set_opt(sb, INIT_INODE_TABLE);
1977 if (!args->from)
1978 arg = EXT4_DEF_LI_WAIT_MULT;
1979 sbi->s_li_wait_mult = arg;
1980 } else if (token == Opt_max_dir_size_kb) {
1981 sbi->s_max_dir_size_kb = arg;
1982 } else if (token == Opt_stripe) {
1983 sbi->s_stripe = arg;
1984 } else if (token == Opt_resuid) {
1985 uid = make_kuid(current_user_ns(), arg);
1986 if (!uid_valid(uid)) {
1987 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1988 return -1;
1989 }
1990 sbi->s_resuid = uid;
1991 } else if (token == Opt_resgid) {
1992 gid = make_kgid(current_user_ns(), arg);
1993 if (!gid_valid(gid)) {
1994 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1995 return -1;
1996 }
1997 sbi->s_resgid = gid;
1998 } else if (token == Opt_journal_dev) {
1999 if (is_remount) {
2000 ext4_msg(sb, KERN_ERR,
2001 "Cannot specify journal on remount");
2002 return -1;
2003 }
2004 *journal_devnum = arg;
2005 } else if (token == Opt_journal_path) {
2006 char *journal_path;
2007 struct inode *journal_inode;
2008 struct path path;
2009 int error;
2010
2011 if (is_remount) {
2012 ext4_msg(sb, KERN_ERR,
2013 "Cannot specify journal on remount");
2014 return -1;
2015 }
2016 journal_path = match_strdup(&args[0]);
2017 if (!journal_path) {
2018 ext4_msg(sb, KERN_ERR, "error: could not dup "
2019 "journal device string");
2020 return -1;
2021 }
2022
2023 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2024 if (error) {
2025 ext4_msg(sb, KERN_ERR, "error: could not find "
2026 "journal device path: error %d", error);
2027 kfree(journal_path);
2028 return -1;
2029 }
2030
2031 journal_inode = d_inode(path.dentry);
2032 if (!S_ISBLK(journal_inode->i_mode)) {
2033 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2034 "is not a block device", journal_path);
2035 path_put(&path);
2036 kfree(journal_path);
2037 return -1;
2038 }
2039
2040 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2041 path_put(&path);
2042 kfree(journal_path);
2043 } else if (token == Opt_journal_ioprio) {
2044 if (arg > 7) {
2045 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2046 " (must be 0-7)");
2047 return -1;
2048 }
2049 *journal_ioprio =
2050 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2051 } else if (token == Opt_test_dummy_encryption) {
2052 #ifdef CONFIG_FS_ENCRYPTION
2053 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
2054 ext4_msg(sb, KERN_WARNING,
2055 "Test dummy encryption mode enabled");
2056 #else
2057 ext4_msg(sb, KERN_WARNING,
2058 "Test dummy encryption mount option ignored");
2059 #endif
2060 } else if (m->flags & MOPT_DATAJ) {
2061 if (is_remount) {
2062 if (!sbi->s_journal)
2063 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2064 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2065 ext4_msg(sb, KERN_ERR,
2066 "Cannot change data mode on remount");
2067 return -1;
2068 }
2069 } else {
2070 clear_opt(sb, DATA_FLAGS);
2071 sbi->s_mount_opt |= m->mount_opt;
2072 }
2073 #ifdef CONFIG_QUOTA
2074 } else if (m->flags & MOPT_QFMT) {
2075 if (sb_any_quota_loaded(sb) &&
2076 sbi->s_jquota_fmt != m->mount_opt) {
2077 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2078 "quota options when quota turned on");
2079 return -1;
2080 }
2081 if (ext4_has_feature_quota(sb)) {
2082 ext4_msg(sb, KERN_INFO,
2083 "Quota format mount options ignored "
2084 "when QUOTA feature is enabled");
2085 return 1;
2086 }
2087 sbi->s_jquota_fmt = m->mount_opt;
2088 #endif
2089 } else if (token == Opt_dax) {
2090 #ifdef CONFIG_FS_DAX
2091 ext4_msg(sb, KERN_WARNING,
2092 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2093 sbi->s_mount_opt |= m->mount_opt;
2094 #else
2095 ext4_msg(sb, KERN_INFO, "dax option not supported");
2096 return -1;
2097 #endif
2098 } else if (token == Opt_data_err_abort) {
2099 sbi->s_mount_opt |= m->mount_opt;
2100 } else if (token == Opt_data_err_ignore) {
2101 sbi->s_mount_opt &= ~m->mount_opt;
2102 } else {
2103 if (!args->from)
2104 arg = 1;
2105 if (m->flags & MOPT_CLEAR)
2106 arg = !arg;
2107 else if (unlikely(!(m->flags & MOPT_SET))) {
2108 ext4_msg(sb, KERN_WARNING,
2109 "buggy handling of option %s", opt);
2110 WARN_ON(1);
2111 return -1;
2112 }
2113 if (arg != 0)
2114 sbi->s_mount_opt |= m->mount_opt;
2115 else
2116 sbi->s_mount_opt &= ~m->mount_opt;
2117 }
2118 return 1;
2119 }
2120
2121 static int parse_options(char *options, struct super_block *sb,
2122 unsigned long *journal_devnum,
2123 unsigned int *journal_ioprio,
2124 int is_remount)
2125 {
2126 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2127 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2128 substring_t args[MAX_OPT_ARGS];
2129 int token;
2130
2131 if (!options)
2132 return 1;
2133
2134 while ((p = strsep(&options, ",")) != NULL) {
2135 if (!*p)
2136 continue;
2137 /*
2138 * Initialize args struct so we know whether arg was
2139 * found; some options take optional arguments.
2140 */
2141 args[0].to = args[0].from = NULL;
2142 token = match_token(p, tokens, args);
2143 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2144 journal_ioprio, is_remount) < 0)
2145 return 0;
2146 }
2147 #ifdef CONFIG_QUOTA
2148 /*
2149 * We do the test below only for project quotas. 'usrquota' and
2150 * 'grpquota' mount options are allowed even without quota feature
2151 * to support legacy quotas in quota files.
2152 */
2153 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2154 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2155 "Cannot enable project quota enforcement.");
2156 return 0;
2157 }
2158 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2159 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2160 if (usr_qf_name || grp_qf_name) {
2161 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2162 clear_opt(sb, USRQUOTA);
2163
2164 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2165 clear_opt(sb, GRPQUOTA);
2166
2167 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2168 ext4_msg(sb, KERN_ERR, "old and new quota "
2169 "format mixing");
2170 return 0;
2171 }
2172
2173 if (!sbi->s_jquota_fmt) {
2174 ext4_msg(sb, KERN_ERR, "journaled quota format "
2175 "not specified");
2176 return 0;
2177 }
2178 }
2179 #endif
2180 if (test_opt(sb, DIOREAD_NOLOCK)) {
2181 int blocksize =
2182 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2183 if (blocksize < PAGE_SIZE)
2184 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2185 "experimental mount option 'dioread_nolock' "
2186 "for blocksize < PAGE_SIZE");
2187 }
2188 return 1;
2189 }
2190
2191 static inline void ext4_show_quota_options(struct seq_file *seq,
2192 struct super_block *sb)
2193 {
2194 #if defined(CONFIG_QUOTA)
2195 struct ext4_sb_info *sbi = EXT4_SB(sb);
2196 char *usr_qf_name, *grp_qf_name;
2197
2198 if (sbi->s_jquota_fmt) {
2199 char *fmtname = "";
2200
2201 switch (sbi->s_jquota_fmt) {
2202 case QFMT_VFS_OLD:
2203 fmtname = "vfsold";
2204 break;
2205 case QFMT_VFS_V0:
2206 fmtname = "vfsv0";
2207 break;
2208 case QFMT_VFS_V1:
2209 fmtname = "vfsv1";
2210 break;
2211 }
2212 seq_printf(seq, ",jqfmt=%s", fmtname);
2213 }
2214
2215 rcu_read_lock();
2216 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2217 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2218 if (usr_qf_name)
2219 seq_show_option(seq, "usrjquota", usr_qf_name);
2220 if (grp_qf_name)
2221 seq_show_option(seq, "grpjquota", grp_qf_name);
2222 rcu_read_unlock();
2223 #endif
2224 }
2225
2226 static const char *token2str(int token)
2227 {
2228 const struct match_token *t;
2229
2230 for (t = tokens; t->token != Opt_err; t++)
2231 if (t->token == token && !strchr(t->pattern, '='))
2232 break;
2233 return t->pattern;
2234 }
2235
2236 /*
2237 * Show an option if
2238 * - it's set to a non-default value OR
2239 * - if the per-sb default is different from the global default
2240 */
2241 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2242 int nodefs)
2243 {
2244 struct ext4_sb_info *sbi = EXT4_SB(sb);
2245 struct ext4_super_block *es = sbi->s_es;
2246 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2247 const struct mount_opts *m;
2248 char sep = nodefs ? '\n' : ',';
2249
2250 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2251 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2252
2253 if (sbi->s_sb_block != 1)
2254 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2255
2256 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2257 int want_set = m->flags & MOPT_SET;
2258 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2259 (m->flags & MOPT_CLEAR_ERR))
2260 continue;
2261 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2262 continue; /* skip if same as the default */
2263 if ((want_set &&
2264 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2265 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2266 continue; /* select Opt_noFoo vs Opt_Foo */
2267 SEQ_OPTS_PRINT("%s", token2str(m->token));
2268 }
2269
2270 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2271 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2272 SEQ_OPTS_PRINT("resuid=%u",
2273 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2274 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2275 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2276 SEQ_OPTS_PRINT("resgid=%u",
2277 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2278 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2279 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2280 SEQ_OPTS_PUTS("errors=remount-ro");
2281 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2282 SEQ_OPTS_PUTS("errors=continue");
2283 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2284 SEQ_OPTS_PUTS("errors=panic");
2285 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2286 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2287 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2288 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2289 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2290 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2291 if (sb->s_flags & SB_I_VERSION)
2292 SEQ_OPTS_PUTS("i_version");
2293 if (nodefs || sbi->s_stripe)
2294 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2295 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2296 (sbi->s_mount_opt ^ def_mount_opt)) {
2297 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2298 SEQ_OPTS_PUTS("data=journal");
2299 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2300 SEQ_OPTS_PUTS("data=ordered");
2301 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2302 SEQ_OPTS_PUTS("data=writeback");
2303 }
2304 if (nodefs ||
2305 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2306 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2307 sbi->s_inode_readahead_blks);
2308
2309 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2310 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2311 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2312 if (nodefs || sbi->s_max_dir_size_kb)
2313 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2314 if (test_opt(sb, DATA_ERR_ABORT))
2315 SEQ_OPTS_PUTS("data_err=abort");
2316 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2317 SEQ_OPTS_PUTS("test_dummy_encryption");
2318
2319 ext4_show_quota_options(seq, sb);
2320 return 0;
2321 }
2322
2323 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2324 {
2325 return _ext4_show_options(seq, root->d_sb, 0);
2326 }
2327
2328 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2329 {
2330 struct super_block *sb = seq->private;
2331 int rc;
2332
2333 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2334 rc = _ext4_show_options(seq, sb, 1);
2335 seq_puts(seq, "\n");
2336 return rc;
2337 }
2338
2339 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2340 int read_only)
2341 {
2342 struct ext4_sb_info *sbi = EXT4_SB(sb);
2343 int err = 0;
2344
2345 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2346 ext4_msg(sb, KERN_ERR, "revision level too high, "
2347 "forcing read-only mode");
2348 err = -EROFS;
2349 }
2350 if (read_only)
2351 goto done;
2352 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2353 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2354 "running e2fsck is recommended");
2355 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2356 ext4_msg(sb, KERN_WARNING,
2357 "warning: mounting fs with errors, "
2358 "running e2fsck is recommended");
2359 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2360 le16_to_cpu(es->s_mnt_count) >=
2361 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2362 ext4_msg(sb, KERN_WARNING,
2363 "warning: maximal mount count reached, "
2364 "running e2fsck is recommended");
2365 else if (le32_to_cpu(es->s_checkinterval) &&
2366 (ext4_get_tstamp(es, s_lastcheck) +
2367 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2368 ext4_msg(sb, KERN_WARNING,
2369 "warning: checktime reached, "
2370 "running e2fsck is recommended");
2371 if (!sbi->s_journal)
2372 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2373 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2374 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2375 le16_add_cpu(&es->s_mnt_count, 1);
2376 ext4_update_tstamp(es, s_mtime);
2377 if (sbi->s_journal)
2378 ext4_set_feature_journal_needs_recovery(sb);
2379
2380 err = ext4_commit_super(sb, 1);
2381 done:
2382 if (test_opt(sb, DEBUG))
2383 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2384 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2385 sb->s_blocksize,
2386 sbi->s_groups_count,
2387 EXT4_BLOCKS_PER_GROUP(sb),
2388 EXT4_INODES_PER_GROUP(sb),
2389 sbi->s_mount_opt, sbi->s_mount_opt2);
2390
2391 cleancache_init_fs(sb);
2392 return err;
2393 }
2394
2395 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2396 {
2397 struct ext4_sb_info *sbi = EXT4_SB(sb);
2398 struct flex_groups **old_groups, **new_groups;
2399 int size, i, j;
2400
2401 if (!sbi->s_log_groups_per_flex)
2402 return 0;
2403
2404 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2405 if (size <= sbi->s_flex_groups_allocated)
2406 return 0;
2407
2408 new_groups = kvzalloc(roundup_pow_of_two(size *
2409 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2410 if (!new_groups) {
2411 ext4_msg(sb, KERN_ERR,
2412 "not enough memory for %d flex group pointers", size);
2413 return -ENOMEM;
2414 }
2415 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2416 new_groups[i] = kvzalloc(roundup_pow_of_two(
2417 sizeof(struct flex_groups)),
2418 GFP_KERNEL);
2419 if (!new_groups[i]) {
2420 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2421 kvfree(new_groups[j]);
2422 kvfree(new_groups);
2423 ext4_msg(sb, KERN_ERR,
2424 "not enough memory for %d flex groups", size);
2425 return -ENOMEM;
2426 }
2427 }
2428 rcu_read_lock();
2429 old_groups = rcu_dereference(sbi->s_flex_groups);
2430 if (old_groups)
2431 memcpy(new_groups, old_groups,
2432 (sbi->s_flex_groups_allocated *
2433 sizeof(struct flex_groups *)));
2434 rcu_read_unlock();
2435 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2436 sbi->s_flex_groups_allocated = size;
2437 if (old_groups)
2438 ext4_kvfree_array_rcu(old_groups);
2439 return 0;
2440 }
2441
2442 static int ext4_fill_flex_info(struct super_block *sb)
2443 {
2444 struct ext4_sb_info *sbi = EXT4_SB(sb);
2445 struct ext4_group_desc *gdp = NULL;
2446 struct flex_groups *fg;
2447 ext4_group_t flex_group;
2448 int i, err;
2449
2450 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2451 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2452 sbi->s_log_groups_per_flex = 0;
2453 return 1;
2454 }
2455
2456 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2457 if (err)
2458 goto failed;
2459
2460 for (i = 0; i < sbi->s_groups_count; i++) {
2461 gdp = ext4_get_group_desc(sb, i, NULL);
2462
2463 flex_group = ext4_flex_group(sbi, i);
2464 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2465 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2466 atomic64_add(ext4_free_group_clusters(sb, gdp),
2467 &fg->free_clusters);
2468 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2469 }
2470
2471 return 1;
2472 failed:
2473 return 0;
2474 }
2475
2476 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2477 struct ext4_group_desc *gdp)
2478 {
2479 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2480 __u16 crc = 0;
2481 __le32 le_group = cpu_to_le32(block_group);
2482 struct ext4_sb_info *sbi = EXT4_SB(sb);
2483
2484 if (ext4_has_metadata_csum(sbi->s_sb)) {
2485 /* Use new metadata_csum algorithm */
2486 __u32 csum32;
2487 __u16 dummy_csum = 0;
2488
2489 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2490 sizeof(le_group));
2491 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2492 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2493 sizeof(dummy_csum));
2494 offset += sizeof(dummy_csum);
2495 if (offset < sbi->s_desc_size)
2496 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2497 sbi->s_desc_size - offset);
2498
2499 crc = csum32 & 0xFFFF;
2500 goto out;
2501 }
2502
2503 /* old crc16 code */
2504 if (!ext4_has_feature_gdt_csum(sb))
2505 return 0;
2506
2507 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2508 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2509 crc = crc16(crc, (__u8 *)gdp, offset);
2510 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2511 /* for checksum of struct ext4_group_desc do the rest...*/
2512 if (ext4_has_feature_64bit(sb) &&
2513 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2514 crc = crc16(crc, (__u8 *)gdp + offset,
2515 le16_to_cpu(sbi->s_es->s_desc_size) -
2516 offset);
2517
2518 out:
2519 return cpu_to_le16(crc);
2520 }
2521
2522 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2523 struct ext4_group_desc *gdp)
2524 {
2525 if (ext4_has_group_desc_csum(sb) &&
2526 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2527 return 0;
2528
2529 return 1;
2530 }
2531
2532 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2533 struct ext4_group_desc *gdp)
2534 {
2535 if (!ext4_has_group_desc_csum(sb))
2536 return;
2537 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2538 }
2539
2540 /* Called at mount-time, super-block is locked */
2541 static int ext4_check_descriptors(struct super_block *sb,
2542 ext4_fsblk_t sb_block,
2543 ext4_group_t *first_not_zeroed)
2544 {
2545 struct ext4_sb_info *sbi = EXT4_SB(sb);
2546 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2547 ext4_fsblk_t last_block;
2548 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2549 ext4_fsblk_t block_bitmap;
2550 ext4_fsblk_t inode_bitmap;
2551 ext4_fsblk_t inode_table;
2552 int flexbg_flag = 0;
2553 ext4_group_t i, grp = sbi->s_groups_count;
2554
2555 if (ext4_has_feature_flex_bg(sb))
2556 flexbg_flag = 1;
2557
2558 ext4_debug("Checking group descriptors");
2559
2560 for (i = 0; i < sbi->s_groups_count; i++) {
2561 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2562
2563 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2564 last_block = ext4_blocks_count(sbi->s_es) - 1;
2565 else
2566 last_block = first_block +
2567 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2568
2569 if ((grp == sbi->s_groups_count) &&
2570 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2571 grp = i;
2572
2573 block_bitmap = ext4_block_bitmap(sb, gdp);
2574 if (block_bitmap == sb_block) {
2575 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2576 "Block bitmap for group %u overlaps "
2577 "superblock", i);
2578 if (!sb_rdonly(sb))
2579 return 0;
2580 }
2581 if (block_bitmap >= sb_block + 1 &&
2582 block_bitmap <= last_bg_block) {
2583 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2584 "Block bitmap for group %u overlaps "
2585 "block group descriptors", i);
2586 if (!sb_rdonly(sb))
2587 return 0;
2588 }
2589 if (block_bitmap < first_block || block_bitmap > last_block) {
2590 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2591 "Block bitmap for group %u not in group "
2592 "(block %llu)!", i, block_bitmap);
2593 return 0;
2594 }
2595 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2596 if (inode_bitmap == sb_block) {
2597 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2598 "Inode bitmap for group %u overlaps "
2599 "superblock", i);
2600 if (!sb_rdonly(sb))
2601 return 0;
2602 }
2603 if (inode_bitmap >= sb_block + 1 &&
2604 inode_bitmap <= last_bg_block) {
2605 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2606 "Inode bitmap for group %u overlaps "
2607 "block group descriptors", i);
2608 if (!sb_rdonly(sb))
2609 return 0;
2610 }
2611 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2612 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2613 "Inode bitmap for group %u not in group "
2614 "(block %llu)!", i, inode_bitmap);
2615 return 0;
2616 }
2617 inode_table = ext4_inode_table(sb, gdp);
2618 if (inode_table == sb_block) {
2619 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2620 "Inode table for group %u overlaps "
2621 "superblock", i);
2622 if (!sb_rdonly(sb))
2623 return 0;
2624 }
2625 if (inode_table >= sb_block + 1 &&
2626 inode_table <= last_bg_block) {
2627 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2628 "Inode table for group %u overlaps "
2629 "block group descriptors", i);
2630 if (!sb_rdonly(sb))
2631 return 0;
2632 }
2633 if (inode_table < first_block ||
2634 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2635 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2636 "Inode table for group %u not in group "
2637 "(block %llu)!", i, inode_table);
2638 return 0;
2639 }
2640 ext4_lock_group(sb, i);
2641 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2642 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2643 "Checksum for group %u failed (%u!=%u)",
2644 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2645 gdp)), le16_to_cpu(gdp->bg_checksum));
2646 if (!sb_rdonly(sb)) {
2647 ext4_unlock_group(sb, i);
2648 return 0;
2649 }
2650 }
2651 ext4_unlock_group(sb, i);
2652 if (!flexbg_flag)
2653 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2654 }
2655 if (NULL != first_not_zeroed)
2656 *first_not_zeroed = grp;
2657 return 1;
2658 }
2659
2660 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2661 * the superblock) which were deleted from all directories, but held open by
2662 * a process at the time of a crash. We walk the list and try to delete these
2663 * inodes at recovery time (only with a read-write filesystem).
2664 *
2665 * In order to keep the orphan inode chain consistent during traversal (in
2666 * case of crash during recovery), we link each inode into the superblock
2667 * orphan list_head and handle it the same way as an inode deletion during
2668 * normal operation (which journals the operations for us).
2669 *
2670 * We only do an iget() and an iput() on each inode, which is very safe if we
2671 * accidentally point at an in-use or already deleted inode. The worst that
2672 * can happen in this case is that we get a "bit already cleared" message from
2673 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2674 * e2fsck was run on this filesystem, and it must have already done the orphan
2675 * inode cleanup for us, so we can safely abort without any further action.
2676 */
2677 static void ext4_orphan_cleanup(struct super_block *sb,
2678 struct ext4_super_block *es)
2679 {
2680 unsigned int s_flags = sb->s_flags;
2681 int ret, nr_orphans = 0, nr_truncates = 0;
2682 #ifdef CONFIG_QUOTA
2683 int quota_update = 0;
2684 int i;
2685 #endif
2686 if (!es->s_last_orphan) {
2687 jbd_debug(4, "no orphan inodes to clean up\n");
2688 return;
2689 }
2690
2691 if (bdev_read_only(sb->s_bdev)) {
2692 ext4_msg(sb, KERN_ERR, "write access "
2693 "unavailable, skipping orphan cleanup");
2694 return;
2695 }
2696
2697 /* Check if feature set would not allow a r/w mount */
2698 if (!ext4_feature_set_ok(sb, 0)) {
2699 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2700 "unknown ROCOMPAT features");
2701 return;
2702 }
2703
2704 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2705 /* don't clear list on RO mount w/ errors */
2706 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2707 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2708 "clearing orphan list.\n");
2709 es->s_last_orphan = 0;
2710 }
2711 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2712 return;
2713 }
2714
2715 if (s_flags & SB_RDONLY) {
2716 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2717 sb->s_flags &= ~SB_RDONLY;
2718 }
2719 #ifdef CONFIG_QUOTA
2720 /* Needed for iput() to work correctly and not trash data */
2721 sb->s_flags |= SB_ACTIVE;
2722
2723 /*
2724 * Turn on quotas which were not enabled for read-only mounts if
2725 * filesystem has quota feature, so that they are updated correctly.
2726 */
2727 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2728 int ret = ext4_enable_quotas(sb);
2729
2730 if (!ret)
2731 quota_update = 1;
2732 else
2733 ext4_msg(sb, KERN_ERR,
2734 "Cannot turn on quotas: error %d", ret);
2735 }
2736
2737 /* Turn on journaled quotas used for old sytle */
2738 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2739 if (EXT4_SB(sb)->s_qf_names[i]) {
2740 int ret = ext4_quota_on_mount(sb, i);
2741
2742 if (!ret)
2743 quota_update = 1;
2744 else
2745 ext4_msg(sb, KERN_ERR,
2746 "Cannot turn on journaled "
2747 "quota: type %d: error %d", i, ret);
2748 }
2749 }
2750 #endif
2751
2752 while (es->s_last_orphan) {
2753 struct inode *inode;
2754
2755 /*
2756 * We may have encountered an error during cleanup; if
2757 * so, skip the rest.
2758 */
2759 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2760 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2761 es->s_last_orphan = 0;
2762 break;
2763 }
2764
2765 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2766 if (IS_ERR(inode)) {
2767 es->s_last_orphan = 0;
2768 break;
2769 }
2770
2771 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2772 dquot_initialize(inode);
2773 if (inode->i_nlink) {
2774 if (test_opt(sb, DEBUG))
2775 ext4_msg(sb, KERN_DEBUG,
2776 "%s: truncating inode %lu to %lld bytes",
2777 __func__, inode->i_ino, inode->i_size);
2778 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2779 inode->i_ino, inode->i_size);
2780 inode_lock(inode);
2781 truncate_inode_pages(inode->i_mapping, inode->i_size);
2782 ret = ext4_truncate(inode);
2783 if (ret)
2784 ext4_std_error(inode->i_sb, ret);
2785 inode_unlock(inode);
2786 nr_truncates++;
2787 } else {
2788 if (test_opt(sb, DEBUG))
2789 ext4_msg(sb, KERN_DEBUG,
2790 "%s: deleting unreferenced inode %lu",
2791 __func__, inode->i_ino);
2792 jbd_debug(2, "deleting unreferenced inode %lu\n",
2793 inode->i_ino);
2794 nr_orphans++;
2795 }
2796 iput(inode); /* The delete magic happens here! */
2797 }
2798
2799 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2800
2801 if (nr_orphans)
2802 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2803 PLURAL(nr_orphans));
2804 if (nr_truncates)
2805 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2806 PLURAL(nr_truncates));
2807 #ifdef CONFIG_QUOTA
2808 /* Turn off quotas if they were enabled for orphan cleanup */
2809 if (quota_update) {
2810 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2811 if (sb_dqopt(sb)->files[i])
2812 dquot_quota_off(sb, i);
2813 }
2814 }
2815 #endif
2816 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2817 }
2818
2819 /*
2820 * Maximal extent format file size.
2821 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2822 * extent format containers, within a sector_t, and within i_blocks
2823 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2824 * so that won't be a limiting factor.
2825 *
2826 * However there is other limiting factor. We do store extents in the form
2827 * of starting block and length, hence the resulting length of the extent
2828 * covering maximum file size must fit into on-disk format containers as
2829 * well. Given that length is always by 1 unit bigger than max unit (because
2830 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2831 *
2832 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2833 */
2834 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2835 {
2836 loff_t res;
2837 loff_t upper_limit = MAX_LFS_FILESIZE;
2838
2839 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2840
2841 if (!has_huge_files) {
2842 upper_limit = (1LL << 32) - 1;
2843
2844 /* total blocks in file system block size */
2845 upper_limit >>= (blkbits - 9);
2846 upper_limit <<= blkbits;
2847 }
2848
2849 /*
2850 * 32-bit extent-start container, ee_block. We lower the maxbytes
2851 * by one fs block, so ee_len can cover the extent of maximum file
2852 * size
2853 */
2854 res = (1LL << 32) - 1;
2855 res <<= blkbits;
2856
2857 /* Sanity check against vm- & vfs- imposed limits */
2858 if (res > upper_limit)
2859 res = upper_limit;
2860
2861 return res;
2862 }
2863
2864 /*
2865 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2866 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2867 * We need to be 1 filesystem block less than the 2^48 sector limit.
2868 */
2869 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2870 {
2871 loff_t res = EXT4_NDIR_BLOCKS;
2872 int meta_blocks;
2873 loff_t upper_limit;
2874 /* This is calculated to be the largest file size for a dense, block
2875 * mapped file such that the file's total number of 512-byte sectors,
2876 * including data and all indirect blocks, does not exceed (2^48 - 1).
2877 *
2878 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2879 * number of 512-byte sectors of the file.
2880 */
2881
2882 if (!has_huge_files) {
2883 /*
2884 * !has_huge_files or implies that the inode i_block field
2885 * represents total file blocks in 2^32 512-byte sectors ==
2886 * size of vfs inode i_blocks * 8
2887 */
2888 upper_limit = (1LL << 32) - 1;
2889
2890 /* total blocks in file system block size */
2891 upper_limit >>= (bits - 9);
2892
2893 } else {
2894 /*
2895 * We use 48 bit ext4_inode i_blocks
2896 * With EXT4_HUGE_FILE_FL set the i_blocks
2897 * represent total number of blocks in
2898 * file system block size
2899 */
2900 upper_limit = (1LL << 48) - 1;
2901
2902 }
2903
2904 /* indirect blocks */
2905 meta_blocks = 1;
2906 /* double indirect blocks */
2907 meta_blocks += 1 + (1LL << (bits-2));
2908 /* tripple indirect blocks */
2909 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2910
2911 upper_limit -= meta_blocks;
2912 upper_limit <<= bits;
2913
2914 res += 1LL << (bits-2);
2915 res += 1LL << (2*(bits-2));
2916 res += 1LL << (3*(bits-2));
2917 res <<= bits;
2918 if (res > upper_limit)
2919 res = upper_limit;
2920
2921 if (res > MAX_LFS_FILESIZE)
2922 res = MAX_LFS_FILESIZE;
2923
2924 return res;
2925 }
2926
2927 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2928 ext4_fsblk_t logical_sb_block, int nr)
2929 {
2930 struct ext4_sb_info *sbi = EXT4_SB(sb);
2931 ext4_group_t bg, first_meta_bg;
2932 int has_super = 0;
2933
2934 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2935
2936 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2937 return logical_sb_block + nr + 1;
2938 bg = sbi->s_desc_per_block * nr;
2939 if (ext4_bg_has_super(sb, bg))
2940 has_super = 1;
2941
2942 /*
2943 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2944 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2945 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2946 * compensate.
2947 */
2948 if (sb->s_blocksize == 1024 && nr == 0 &&
2949 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
2950 has_super++;
2951
2952 return (has_super + ext4_group_first_block_no(sb, bg));
2953 }
2954
2955 /**
2956 * ext4_get_stripe_size: Get the stripe size.
2957 * @sbi: In memory super block info
2958 *
2959 * If we have specified it via mount option, then
2960 * use the mount option value. If the value specified at mount time is
2961 * greater than the blocks per group use the super block value.
2962 * If the super block value is greater than blocks per group return 0.
2963 * Allocator needs it be less than blocks per group.
2964 *
2965 */
2966 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2967 {
2968 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2969 unsigned long stripe_width =
2970 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2971 int ret;
2972
2973 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2974 ret = sbi->s_stripe;
2975 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2976 ret = stripe_width;
2977 else if (stride && stride <= sbi->s_blocks_per_group)
2978 ret = stride;
2979 else
2980 ret = 0;
2981
2982 /*
2983 * If the stripe width is 1, this makes no sense and
2984 * we set it to 0 to turn off stripe handling code.
2985 */
2986 if (ret <= 1)
2987 ret = 0;
2988
2989 return ret;
2990 }
2991
2992 /*
2993 * Check whether this filesystem can be mounted based on
2994 * the features present and the RDONLY/RDWR mount requested.
2995 * Returns 1 if this filesystem can be mounted as requested,
2996 * 0 if it cannot be.
2997 */
2998 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2999 {
3000 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3001 ext4_msg(sb, KERN_ERR,
3002 "Couldn't mount because of "
3003 "unsupported optional features (%x)",
3004 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3005 ~EXT4_FEATURE_INCOMPAT_SUPP));
3006 return 0;
3007 }
3008
3009 #ifndef CONFIG_UNICODE
3010 if (ext4_has_feature_casefold(sb)) {
3011 ext4_msg(sb, KERN_ERR,
3012 "Filesystem with casefold feature cannot be "
3013 "mounted without CONFIG_UNICODE");
3014 return 0;
3015 }
3016 #endif
3017
3018 if (readonly)
3019 return 1;
3020
3021 if (ext4_has_feature_readonly(sb)) {
3022 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3023 sb->s_flags |= SB_RDONLY;
3024 return 1;
3025 }
3026
3027 /* Check that feature set is OK for a read-write mount */
3028 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3029 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3030 "unsupported optional features (%x)",
3031 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3032 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3033 return 0;
3034 }
3035 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3036 ext4_msg(sb, KERN_ERR,
3037 "Can't support bigalloc feature without "
3038 "extents feature\n");
3039 return 0;
3040 }
3041
3042 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3043 if (!readonly && (ext4_has_feature_quota(sb) ||
3044 ext4_has_feature_project(sb))) {
3045 ext4_msg(sb, KERN_ERR,
3046 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3047 return 0;
3048 }
3049 #endif /* CONFIG_QUOTA */
3050 return 1;
3051 }
3052
3053 /*
3054 * This function is called once a day if we have errors logged
3055 * on the file system
3056 */
3057 static void print_daily_error_info(struct timer_list *t)
3058 {
3059 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3060 struct super_block *sb = sbi->s_sb;
3061 struct ext4_super_block *es = sbi->s_es;
3062
3063 if (es->s_error_count)
3064 /* fsck newer than v1.41.13 is needed to clean this condition. */
3065 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3066 le32_to_cpu(es->s_error_count));
3067 if (es->s_first_error_time) {
3068 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3069 sb->s_id,
3070 ext4_get_tstamp(es, s_first_error_time),
3071 (int) sizeof(es->s_first_error_func),
3072 es->s_first_error_func,
3073 le32_to_cpu(es->s_first_error_line));
3074 if (es->s_first_error_ino)
3075 printk(KERN_CONT ": inode %u",
3076 le32_to_cpu(es->s_first_error_ino));
3077 if (es->s_first_error_block)
3078 printk(KERN_CONT ": block %llu", (unsigned long long)
3079 le64_to_cpu(es->s_first_error_block));
3080 printk(KERN_CONT "\n");
3081 }
3082 if (es->s_last_error_time) {
3083 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3084 sb->s_id,
3085 ext4_get_tstamp(es, s_last_error_time),
3086 (int) sizeof(es->s_last_error_func),
3087 es->s_last_error_func,
3088 le32_to_cpu(es->s_last_error_line));
3089 if (es->s_last_error_ino)
3090 printk(KERN_CONT ": inode %u",
3091 le32_to_cpu(es->s_last_error_ino));
3092 if (es->s_last_error_block)
3093 printk(KERN_CONT ": block %llu", (unsigned long long)
3094 le64_to_cpu(es->s_last_error_block));
3095 printk(KERN_CONT "\n");
3096 }
3097 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3098 }
3099
3100 /* Find next suitable group and run ext4_init_inode_table */
3101 static int ext4_run_li_request(struct ext4_li_request *elr)
3102 {
3103 struct ext4_group_desc *gdp = NULL;
3104 ext4_group_t group, ngroups;
3105 struct super_block *sb;
3106 unsigned long timeout = 0;
3107 int ret = 0;
3108
3109 sb = elr->lr_super;
3110 ngroups = EXT4_SB(sb)->s_groups_count;
3111
3112 for (group = elr->lr_next_group; group < ngroups; group++) {
3113 gdp = ext4_get_group_desc(sb, group, NULL);
3114 if (!gdp) {
3115 ret = 1;
3116 break;
3117 }
3118
3119 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3120 break;
3121 }
3122
3123 if (group >= ngroups)
3124 ret = 1;
3125
3126 if (!ret) {
3127 timeout = jiffies;
3128 ret = ext4_init_inode_table(sb, group,
3129 elr->lr_timeout ? 0 : 1);
3130 if (elr->lr_timeout == 0) {
3131 timeout = (jiffies - timeout) *
3132 elr->lr_sbi->s_li_wait_mult;
3133 elr->lr_timeout = timeout;
3134 }
3135 elr->lr_next_sched = jiffies + elr->lr_timeout;
3136 elr->lr_next_group = group + 1;
3137 }
3138 return ret;
3139 }
3140
3141 /*
3142 * Remove lr_request from the list_request and free the
3143 * request structure. Should be called with li_list_mtx held
3144 */
3145 static void ext4_remove_li_request(struct ext4_li_request *elr)
3146 {
3147 struct ext4_sb_info *sbi;
3148
3149 if (!elr)
3150 return;
3151
3152 sbi = elr->lr_sbi;
3153
3154 list_del(&elr->lr_request);
3155 sbi->s_li_request = NULL;
3156 kfree(elr);
3157 }
3158
3159 static void ext4_unregister_li_request(struct super_block *sb)
3160 {
3161 mutex_lock(&ext4_li_mtx);
3162 if (!ext4_li_info) {
3163 mutex_unlock(&ext4_li_mtx);
3164 return;
3165 }
3166
3167 mutex_lock(&ext4_li_info->li_list_mtx);
3168 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3169 mutex_unlock(&ext4_li_info->li_list_mtx);
3170 mutex_unlock(&ext4_li_mtx);
3171 }
3172
3173 static struct task_struct *ext4_lazyinit_task;
3174
3175 /*
3176 * This is the function where ext4lazyinit thread lives. It walks
3177 * through the request list searching for next scheduled filesystem.
3178 * When such a fs is found, run the lazy initialization request
3179 * (ext4_rn_li_request) and keep track of the time spend in this
3180 * function. Based on that time we compute next schedule time of
3181 * the request. When walking through the list is complete, compute
3182 * next waking time and put itself into sleep.
3183 */
3184 static int ext4_lazyinit_thread(void *arg)
3185 {
3186 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3187 struct list_head *pos, *n;
3188 struct ext4_li_request *elr;
3189 unsigned long next_wakeup, cur;
3190
3191 BUG_ON(NULL == eli);
3192
3193 cont_thread:
3194 while (true) {
3195 next_wakeup = MAX_JIFFY_OFFSET;
3196
3197 mutex_lock(&eli->li_list_mtx);
3198 if (list_empty(&eli->li_request_list)) {
3199 mutex_unlock(&eli->li_list_mtx);
3200 goto exit_thread;
3201 }
3202 list_for_each_safe(pos, n, &eli->li_request_list) {
3203 int err = 0;
3204 int progress = 0;
3205 elr = list_entry(pos, struct ext4_li_request,
3206 lr_request);
3207
3208 if (time_before(jiffies, elr->lr_next_sched)) {
3209 if (time_before(elr->lr_next_sched, next_wakeup))
3210 next_wakeup = elr->lr_next_sched;
3211 continue;
3212 }
3213 if (down_read_trylock(&elr->lr_super->s_umount)) {
3214 if (sb_start_write_trylock(elr->lr_super)) {
3215 progress = 1;
3216 /*
3217 * We hold sb->s_umount, sb can not
3218 * be removed from the list, it is
3219 * now safe to drop li_list_mtx
3220 */
3221 mutex_unlock(&eli->li_list_mtx);
3222 err = ext4_run_li_request(elr);
3223 sb_end_write(elr->lr_super);
3224 mutex_lock(&eli->li_list_mtx);
3225 n = pos->next;
3226 }
3227 up_read((&elr->lr_super->s_umount));
3228 }
3229 /* error, remove the lazy_init job */
3230 if (err) {
3231 ext4_remove_li_request(elr);
3232 continue;
3233 }
3234 if (!progress) {
3235 elr->lr_next_sched = jiffies +
3236 (prandom_u32()
3237 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3238 }
3239 if (time_before(elr->lr_next_sched, next_wakeup))
3240 next_wakeup = elr->lr_next_sched;
3241 }
3242 mutex_unlock(&eli->li_list_mtx);
3243
3244 try_to_freeze();
3245
3246 cur = jiffies;
3247 if ((time_after_eq(cur, next_wakeup)) ||
3248 (MAX_JIFFY_OFFSET == next_wakeup)) {
3249 cond_resched();
3250 continue;
3251 }
3252
3253 schedule_timeout_interruptible(next_wakeup - cur);
3254
3255 if (kthread_should_stop()) {
3256 ext4_clear_request_list();
3257 goto exit_thread;
3258 }
3259 }
3260
3261 exit_thread:
3262 /*
3263 * It looks like the request list is empty, but we need
3264 * to check it under the li_list_mtx lock, to prevent any
3265 * additions into it, and of course we should lock ext4_li_mtx
3266 * to atomically free the list and ext4_li_info, because at
3267 * this point another ext4 filesystem could be registering
3268 * new one.
3269 */
3270 mutex_lock(&ext4_li_mtx);
3271 mutex_lock(&eli->li_list_mtx);
3272 if (!list_empty(&eli->li_request_list)) {
3273 mutex_unlock(&eli->li_list_mtx);
3274 mutex_unlock(&ext4_li_mtx);
3275 goto cont_thread;
3276 }
3277 mutex_unlock(&eli->li_list_mtx);
3278 kfree(ext4_li_info);
3279 ext4_li_info = NULL;
3280 mutex_unlock(&ext4_li_mtx);
3281
3282 return 0;
3283 }
3284
3285 static void ext4_clear_request_list(void)
3286 {
3287 struct list_head *pos, *n;
3288 struct ext4_li_request *elr;
3289
3290 mutex_lock(&ext4_li_info->li_list_mtx);
3291 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3292 elr = list_entry(pos, struct ext4_li_request,
3293 lr_request);
3294 ext4_remove_li_request(elr);
3295 }
3296 mutex_unlock(&ext4_li_info->li_list_mtx);
3297 }
3298
3299 static int ext4_run_lazyinit_thread(void)
3300 {
3301 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3302 ext4_li_info, "ext4lazyinit");
3303 if (IS_ERR(ext4_lazyinit_task)) {
3304 int err = PTR_ERR(ext4_lazyinit_task);
3305 ext4_clear_request_list();
3306 kfree(ext4_li_info);
3307 ext4_li_info = NULL;
3308 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3309 "initialization thread\n",
3310 err);
3311 return err;
3312 }
3313 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3314 return 0;
3315 }
3316
3317 /*
3318 * Check whether it make sense to run itable init. thread or not.
3319 * If there is at least one uninitialized inode table, return
3320 * corresponding group number, else the loop goes through all
3321 * groups and return total number of groups.
3322 */
3323 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3324 {
3325 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3326 struct ext4_group_desc *gdp = NULL;
3327
3328 if (!ext4_has_group_desc_csum(sb))
3329 return ngroups;
3330
3331 for (group = 0; group < ngroups; group++) {
3332 gdp = ext4_get_group_desc(sb, group, NULL);
3333 if (!gdp)
3334 continue;
3335
3336 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3337 break;
3338 }
3339
3340 return group;
3341 }
3342
3343 static int ext4_li_info_new(void)
3344 {
3345 struct ext4_lazy_init *eli = NULL;
3346
3347 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3348 if (!eli)
3349 return -ENOMEM;
3350
3351 INIT_LIST_HEAD(&eli->li_request_list);
3352 mutex_init(&eli->li_list_mtx);
3353
3354 eli->li_state |= EXT4_LAZYINIT_QUIT;
3355
3356 ext4_li_info = eli;
3357
3358 return 0;
3359 }
3360
3361 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3362 ext4_group_t start)
3363 {
3364 struct ext4_sb_info *sbi = EXT4_SB(sb);
3365 struct ext4_li_request *elr;
3366
3367 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3368 if (!elr)
3369 return NULL;
3370
3371 elr->lr_super = sb;
3372 elr->lr_sbi = sbi;
3373 elr->lr_next_group = start;
3374
3375 /*
3376 * Randomize first schedule time of the request to
3377 * spread the inode table initialization requests
3378 * better.
3379 */
3380 elr->lr_next_sched = jiffies + (prandom_u32() %
3381 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3382 return elr;
3383 }
3384
3385 int ext4_register_li_request(struct super_block *sb,
3386 ext4_group_t first_not_zeroed)
3387 {
3388 struct ext4_sb_info *sbi = EXT4_SB(sb);
3389 struct ext4_li_request *elr = NULL;
3390 ext4_group_t ngroups = sbi->s_groups_count;
3391 int ret = 0;
3392
3393 mutex_lock(&ext4_li_mtx);
3394 if (sbi->s_li_request != NULL) {
3395 /*
3396 * Reset timeout so it can be computed again, because
3397 * s_li_wait_mult might have changed.
3398 */
3399 sbi->s_li_request->lr_timeout = 0;
3400 goto out;
3401 }
3402
3403 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3404 !test_opt(sb, INIT_INODE_TABLE))
3405 goto out;
3406
3407 elr = ext4_li_request_new(sb, first_not_zeroed);
3408 if (!elr) {
3409 ret = -ENOMEM;
3410 goto out;
3411 }
3412
3413 if (NULL == ext4_li_info) {
3414 ret = ext4_li_info_new();
3415 if (ret)
3416 goto out;
3417 }
3418
3419 mutex_lock(&ext4_li_info->li_list_mtx);
3420 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3421 mutex_unlock(&ext4_li_info->li_list_mtx);
3422
3423 sbi->s_li_request = elr;
3424 /*
3425 * set elr to NULL here since it has been inserted to
3426 * the request_list and the removal and free of it is
3427 * handled by ext4_clear_request_list from now on.
3428 */
3429 elr = NULL;
3430
3431 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3432 ret = ext4_run_lazyinit_thread();
3433 if (ret)
3434 goto out;
3435 }
3436 out:
3437 mutex_unlock(&ext4_li_mtx);
3438 if (ret)
3439 kfree(elr);
3440 return ret;
3441 }
3442
3443 /*
3444 * We do not need to lock anything since this is called on
3445 * module unload.
3446 */
3447 static void ext4_destroy_lazyinit_thread(void)
3448 {
3449 /*
3450 * If thread exited earlier
3451 * there's nothing to be done.
3452 */
3453 if (!ext4_li_info || !ext4_lazyinit_task)
3454 return;
3455
3456 kthread_stop(ext4_lazyinit_task);
3457 }
3458
3459 static int set_journal_csum_feature_set(struct super_block *sb)
3460 {
3461 int ret = 1;
3462 int compat, incompat;
3463 struct ext4_sb_info *sbi = EXT4_SB(sb);
3464
3465 if (ext4_has_metadata_csum(sb)) {
3466 /* journal checksum v3 */
3467 compat = 0;
3468 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3469 } else {
3470 /* journal checksum v1 */
3471 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3472 incompat = 0;
3473 }
3474
3475 jbd2_journal_clear_features(sbi->s_journal,
3476 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3477 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3478 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3479 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3480 ret = jbd2_journal_set_features(sbi->s_journal,
3481 compat, 0,
3482 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3483 incompat);
3484 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3485 ret = jbd2_journal_set_features(sbi->s_journal,
3486 compat, 0,
3487 incompat);
3488 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3489 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3490 } else {
3491 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3492 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3493 }
3494
3495 return ret;
3496 }
3497
3498 /*
3499 * Note: calculating the overhead so we can be compatible with
3500 * historical BSD practice is quite difficult in the face of
3501 * clusters/bigalloc. This is because multiple metadata blocks from
3502 * different block group can end up in the same allocation cluster.
3503 * Calculating the exact overhead in the face of clustered allocation
3504 * requires either O(all block bitmaps) in memory or O(number of block
3505 * groups**2) in time. We will still calculate the superblock for
3506 * older file systems --- and if we come across with a bigalloc file
3507 * system with zero in s_overhead_clusters the estimate will be close to
3508 * correct especially for very large cluster sizes --- but for newer
3509 * file systems, it's better to calculate this figure once at mkfs
3510 * time, and store it in the superblock. If the superblock value is
3511 * present (even for non-bigalloc file systems), we will use it.
3512 */
3513 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3514 char *buf)
3515 {
3516 struct ext4_sb_info *sbi = EXT4_SB(sb);
3517 struct ext4_group_desc *gdp;
3518 ext4_fsblk_t first_block, last_block, b;
3519 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3520 int s, j, count = 0;
3521
3522 if (!ext4_has_feature_bigalloc(sb))
3523 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3524 sbi->s_itb_per_group + 2);
3525
3526 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3527 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3528 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3529 for (i = 0; i < ngroups; i++) {
3530 gdp = ext4_get_group_desc(sb, i, NULL);
3531 b = ext4_block_bitmap(sb, gdp);
3532 if (b >= first_block && b <= last_block) {
3533 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3534 count++;
3535 }
3536 b = ext4_inode_bitmap(sb, gdp);
3537 if (b >= first_block && b <= last_block) {
3538 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3539 count++;
3540 }
3541 b = ext4_inode_table(sb, gdp);
3542 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3543 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3544 int c = EXT4_B2C(sbi, b - first_block);
3545 ext4_set_bit(c, buf);
3546 count++;
3547 }
3548 if (i != grp)
3549 continue;
3550 s = 0;
3551 if (ext4_bg_has_super(sb, grp)) {
3552 ext4_set_bit(s++, buf);
3553 count++;
3554 }
3555 j = ext4_bg_num_gdb(sb, grp);
3556 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3557 ext4_error(sb, "Invalid number of block group "
3558 "descriptor blocks: %d", j);
3559 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3560 }
3561 count += j;
3562 for (; j > 0; j--)
3563 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3564 }
3565 if (!count)
3566 return 0;
3567 return EXT4_CLUSTERS_PER_GROUP(sb) -
3568 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3569 }
3570
3571 /*
3572 * Compute the overhead and stash it in sbi->s_overhead
3573 */
3574 int ext4_calculate_overhead(struct super_block *sb)
3575 {
3576 struct ext4_sb_info *sbi = EXT4_SB(sb);
3577 struct ext4_super_block *es = sbi->s_es;
3578 struct inode *j_inode;
3579 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3580 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3581 ext4_fsblk_t overhead = 0;
3582 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3583
3584 if (!buf)
3585 return -ENOMEM;
3586
3587 /*
3588 * Compute the overhead (FS structures). This is constant
3589 * for a given filesystem unless the number of block groups
3590 * changes so we cache the previous value until it does.
3591 */
3592
3593 /*
3594 * All of the blocks before first_data_block are overhead
3595 */
3596 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3597
3598 /*
3599 * Add the overhead found in each block group
3600 */
3601 for (i = 0; i < ngroups; i++) {
3602 int blks;
3603
3604 blks = count_overhead(sb, i, buf);
3605 overhead += blks;
3606 if (blks)
3607 memset(buf, 0, PAGE_SIZE);
3608 cond_resched();
3609 }
3610
3611 /*
3612 * Add the internal journal blocks whether the journal has been
3613 * loaded or not
3614 */
3615 if (sbi->s_journal && !sbi->journal_bdev)
3616 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3617 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3618 /* j_inum for internal journal is non-zero */
3619 j_inode = ext4_get_journal_inode(sb, j_inum);
3620 if (j_inode) {
3621 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3622 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3623 iput(j_inode);
3624 } else {
3625 ext4_msg(sb, KERN_ERR, "can't get journal size");
3626 }
3627 }
3628 sbi->s_overhead = overhead;
3629 smp_wmb();
3630 free_page((unsigned long) buf);
3631 return 0;
3632 }
3633
3634 static void ext4_set_resv_clusters(struct super_block *sb)
3635 {
3636 ext4_fsblk_t resv_clusters;
3637 struct ext4_sb_info *sbi = EXT4_SB(sb);
3638
3639 /*
3640 * There's no need to reserve anything when we aren't using extents.
3641 * The space estimates are exact, there are no unwritten extents,
3642 * hole punching doesn't need new metadata... This is needed especially
3643 * to keep ext2/3 backward compatibility.
3644 */
3645 if (!ext4_has_feature_extents(sb))
3646 return;
3647 /*
3648 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3649 * This should cover the situations where we can not afford to run
3650 * out of space like for example punch hole, or converting
3651 * unwritten extents in delalloc path. In most cases such
3652 * allocation would require 1, or 2 blocks, higher numbers are
3653 * very rare.
3654 */
3655 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3656 sbi->s_cluster_bits);
3657
3658 do_div(resv_clusters, 50);
3659 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3660
3661 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3662 }
3663
3664 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3665 {
3666 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3667 char *orig_data = kstrdup(data, GFP_KERNEL);
3668 struct buffer_head *bh, **group_desc;
3669 struct ext4_super_block *es = NULL;
3670 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3671 struct flex_groups **flex_groups;
3672 ext4_fsblk_t block;
3673 ext4_fsblk_t sb_block = get_sb_block(&data);
3674 ext4_fsblk_t logical_sb_block;
3675 unsigned long offset = 0;
3676 unsigned long journal_devnum = 0;
3677 unsigned long def_mount_opts;
3678 struct inode *root;
3679 const char *descr;
3680 int ret = -ENOMEM;
3681 int blocksize, clustersize;
3682 unsigned int db_count;
3683 unsigned int i;
3684 int needs_recovery, has_huge_files, has_bigalloc;
3685 __u64 blocks_count;
3686 int err = 0;
3687 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3688 ext4_group_t first_not_zeroed;
3689
3690 if ((data && !orig_data) || !sbi)
3691 goto out_free_base;
3692
3693 sbi->s_daxdev = dax_dev;
3694 sbi->s_blockgroup_lock =
3695 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3696 if (!sbi->s_blockgroup_lock)
3697 goto out_free_base;
3698
3699 sb->s_fs_info = sbi;
3700 sbi->s_sb = sb;
3701 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3702 sbi->s_sb_block = sb_block;
3703 if (sb->s_bdev->bd_part)
3704 sbi->s_sectors_written_start =
3705 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
3706
3707 /* Cleanup superblock name */
3708 strreplace(sb->s_id, '/', '!');
3709
3710 /* -EINVAL is default */
3711 ret = -EINVAL;
3712 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3713 if (!blocksize) {
3714 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3715 goto out_fail;
3716 }
3717
3718 /*
3719 * The ext4 superblock will not be buffer aligned for other than 1kB
3720 * block sizes. We need to calculate the offset from buffer start.
3721 */
3722 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3723 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3724 offset = do_div(logical_sb_block, blocksize);
3725 } else {
3726 logical_sb_block = sb_block;
3727 }
3728
3729 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3730 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3731 goto out_fail;
3732 }
3733 /*
3734 * Note: s_es must be initialized as soon as possible because
3735 * some ext4 macro-instructions depend on its value
3736 */
3737 es = (struct ext4_super_block *) (bh->b_data + offset);
3738 sbi->s_es = es;
3739 sb->s_magic = le16_to_cpu(es->s_magic);
3740 if (sb->s_magic != EXT4_SUPER_MAGIC)
3741 goto cantfind_ext4;
3742 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3743
3744 /* Warn if metadata_csum and gdt_csum are both set. */
3745 if (ext4_has_feature_metadata_csum(sb) &&
3746 ext4_has_feature_gdt_csum(sb))
3747 ext4_warning(sb, "metadata_csum and uninit_bg are "
3748 "redundant flags; please run fsck.");
3749
3750 /* Check for a known checksum algorithm */
3751 if (!ext4_verify_csum_type(sb, es)) {
3752 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3753 "unknown checksum algorithm.");
3754 silent = 1;
3755 goto cantfind_ext4;
3756 }
3757
3758 /* Load the checksum driver */
3759 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3760 if (IS_ERR(sbi->s_chksum_driver)) {
3761 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3762 ret = PTR_ERR(sbi->s_chksum_driver);
3763 sbi->s_chksum_driver = NULL;
3764 goto failed_mount;
3765 }
3766
3767 /* Check superblock checksum */
3768 if (!ext4_superblock_csum_verify(sb, es)) {
3769 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3770 "invalid superblock checksum. Run e2fsck?");
3771 silent = 1;
3772 ret = -EFSBADCRC;
3773 goto cantfind_ext4;
3774 }
3775
3776 /* Precompute checksum seed for all metadata */
3777 if (ext4_has_feature_csum_seed(sb))
3778 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3779 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3780 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3781 sizeof(es->s_uuid));
3782
3783 /* Set defaults before we parse the mount options */
3784 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3785 set_opt(sb, INIT_INODE_TABLE);
3786 if (def_mount_opts & EXT4_DEFM_DEBUG)
3787 set_opt(sb, DEBUG);
3788 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3789 set_opt(sb, GRPID);
3790 if (def_mount_opts & EXT4_DEFM_UID16)
3791 set_opt(sb, NO_UID32);
3792 /* xattr user namespace & acls are now defaulted on */
3793 set_opt(sb, XATTR_USER);
3794 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3795 set_opt(sb, POSIX_ACL);
3796 #endif
3797 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3798 if (ext4_has_metadata_csum(sb))
3799 set_opt(sb, JOURNAL_CHECKSUM);
3800
3801 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3802 set_opt(sb, JOURNAL_DATA);
3803 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3804 set_opt(sb, ORDERED_DATA);
3805 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3806 set_opt(sb, WRITEBACK_DATA);
3807
3808 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3809 set_opt(sb, ERRORS_PANIC);
3810 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3811 set_opt(sb, ERRORS_CONT);
3812 else
3813 set_opt(sb, ERRORS_RO);
3814 /* block_validity enabled by default; disable with noblock_validity */
3815 set_opt(sb, BLOCK_VALIDITY);
3816 if (def_mount_opts & EXT4_DEFM_DISCARD)
3817 set_opt(sb, DISCARD);
3818
3819 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3820 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3821 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3822 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3823 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3824
3825 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3826 set_opt(sb, BARRIER);
3827
3828 /*
3829 * enable delayed allocation by default
3830 * Use -o nodelalloc to turn it off
3831 */
3832 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3833 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3834 set_opt(sb, DELALLOC);
3835
3836 /*
3837 * set default s_li_wait_mult for lazyinit, for the case there is
3838 * no mount option specified.
3839 */
3840 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3841
3842 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3843
3844 if (blocksize == PAGE_SIZE)
3845 set_opt(sb, DIOREAD_NOLOCK);
3846
3847 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3848 blocksize > EXT4_MAX_BLOCK_SIZE) {
3849 ext4_msg(sb, KERN_ERR,
3850 "Unsupported filesystem blocksize %d (%d log_block_size)",
3851 blocksize, le32_to_cpu(es->s_log_block_size));
3852 goto failed_mount;
3853 }
3854
3855 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3856 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3857 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3858 } else {
3859 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3860 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3861 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3862 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3863 sbi->s_first_ino);
3864 goto failed_mount;
3865 }
3866 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3867 (!is_power_of_2(sbi->s_inode_size)) ||
3868 (sbi->s_inode_size > blocksize)) {
3869 ext4_msg(sb, KERN_ERR,
3870 "unsupported inode size: %d",
3871 sbi->s_inode_size);
3872 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
3873 goto failed_mount;
3874 }
3875 /*
3876 * i_atime_extra is the last extra field available for
3877 * [acm]times in struct ext4_inode. Checking for that
3878 * field should suffice to ensure we have extra space
3879 * for all three.
3880 */
3881 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
3882 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
3883 sb->s_time_gran = 1;
3884 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
3885 } else {
3886 sb->s_time_gran = NSEC_PER_SEC;
3887 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
3888 }
3889 sb->s_time_min = EXT4_TIMESTAMP_MIN;
3890 }
3891 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3892 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3893 EXT4_GOOD_OLD_INODE_SIZE;
3894 if (ext4_has_feature_extra_isize(sb)) {
3895 unsigned v, max = (sbi->s_inode_size -
3896 EXT4_GOOD_OLD_INODE_SIZE);
3897
3898 v = le16_to_cpu(es->s_want_extra_isize);
3899 if (v > max) {
3900 ext4_msg(sb, KERN_ERR,
3901 "bad s_want_extra_isize: %d", v);
3902 goto failed_mount;
3903 }
3904 if (sbi->s_want_extra_isize < v)
3905 sbi->s_want_extra_isize = v;
3906
3907 v = le16_to_cpu(es->s_min_extra_isize);
3908 if (v > max) {
3909 ext4_msg(sb, KERN_ERR,
3910 "bad s_min_extra_isize: %d", v);
3911 goto failed_mount;
3912 }
3913 if (sbi->s_want_extra_isize < v)
3914 sbi->s_want_extra_isize = v;
3915 }
3916 }
3917
3918 if (sbi->s_es->s_mount_opts[0]) {
3919 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3920 sizeof(sbi->s_es->s_mount_opts),
3921 GFP_KERNEL);
3922 if (!s_mount_opts)
3923 goto failed_mount;
3924 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3925 &journal_ioprio, 0)) {
3926 ext4_msg(sb, KERN_WARNING,
3927 "failed to parse options in superblock: %s",
3928 s_mount_opts);
3929 }
3930 kfree(s_mount_opts);
3931 }
3932 sbi->s_def_mount_opt = sbi->s_mount_opt;
3933 if (!parse_options((char *) data, sb, &journal_devnum,
3934 &journal_ioprio, 0))
3935 goto failed_mount;
3936
3937 #ifdef CONFIG_UNICODE
3938 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
3939 const struct ext4_sb_encodings *encoding_info;
3940 struct unicode_map *encoding;
3941 __u16 encoding_flags;
3942
3943 if (ext4_has_feature_encrypt(sb)) {
3944 ext4_msg(sb, KERN_ERR,
3945 "Can't mount with encoding and encryption");
3946 goto failed_mount;
3947 }
3948
3949 if (ext4_sb_read_encoding(es, &encoding_info,
3950 &encoding_flags)) {
3951 ext4_msg(sb, KERN_ERR,
3952 "Encoding requested by superblock is unknown");
3953 goto failed_mount;
3954 }
3955
3956 encoding = utf8_load(encoding_info->version);
3957 if (IS_ERR(encoding)) {
3958 ext4_msg(sb, KERN_ERR,
3959 "can't mount with superblock charset: %s-%s "
3960 "not supported by the kernel. flags: 0x%x.",
3961 encoding_info->name, encoding_info->version,
3962 encoding_flags);
3963 goto failed_mount;
3964 }
3965 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
3966 "%s-%s with flags 0x%hx", encoding_info->name,
3967 encoding_info->version?:"\b", encoding_flags);
3968
3969 sbi->s_encoding = encoding;
3970 sbi->s_encoding_flags = encoding_flags;
3971 }
3972 #endif
3973
3974 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3975 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n");
3976 clear_opt(sb, DIOREAD_NOLOCK);
3977 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3978 ext4_msg(sb, KERN_ERR, "can't mount with "
3979 "both data=journal and delalloc");
3980 goto failed_mount;
3981 }
3982 if (test_opt(sb, DIOREAD_NOLOCK)) {
3983 ext4_msg(sb, KERN_ERR, "can't mount with "
3984 "both data=journal and dioread_nolock");
3985 goto failed_mount;
3986 }
3987 if (test_opt(sb, DAX)) {
3988 ext4_msg(sb, KERN_ERR, "can't mount with "
3989 "both data=journal and dax");
3990 goto failed_mount;
3991 }
3992 if (ext4_has_feature_encrypt(sb)) {
3993 ext4_msg(sb, KERN_WARNING,
3994 "encrypted files will use data=ordered "
3995 "instead of data journaling mode");
3996 }
3997 if (test_opt(sb, DELALLOC))
3998 clear_opt(sb, DELALLOC);
3999 } else {
4000 sb->s_iflags |= SB_I_CGROUPWB;
4001 }
4002
4003 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4004 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4005
4006 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4007 (ext4_has_compat_features(sb) ||
4008 ext4_has_ro_compat_features(sb) ||
4009 ext4_has_incompat_features(sb)))
4010 ext4_msg(sb, KERN_WARNING,
4011 "feature flags set on rev 0 fs, "
4012 "running e2fsck is recommended");
4013
4014 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4015 set_opt2(sb, HURD_COMPAT);
4016 if (ext4_has_feature_64bit(sb)) {
4017 ext4_msg(sb, KERN_ERR,
4018 "The Hurd can't support 64-bit file systems");
4019 goto failed_mount;
4020 }
4021
4022 /*
4023 * ea_inode feature uses l_i_version field which is not
4024 * available in HURD_COMPAT mode.
4025 */
4026 if (ext4_has_feature_ea_inode(sb)) {
4027 ext4_msg(sb, KERN_ERR,
4028 "ea_inode feature is not supported for Hurd");
4029 goto failed_mount;
4030 }
4031 }
4032
4033 if (IS_EXT2_SB(sb)) {
4034 if (ext2_feature_set_ok(sb))
4035 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4036 "using the ext4 subsystem");
4037 else {
4038 /*
4039 * If we're probing be silent, if this looks like
4040 * it's actually an ext[34] filesystem.
4041 */
4042 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4043 goto failed_mount;
4044 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4045 "to feature incompatibilities");
4046 goto failed_mount;
4047 }
4048 }
4049
4050 if (IS_EXT3_SB(sb)) {
4051 if (ext3_feature_set_ok(sb))
4052 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4053 "using the ext4 subsystem");
4054 else {
4055 /*
4056 * If we're probing be silent, if this looks like
4057 * it's actually an ext4 filesystem.
4058 */
4059 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4060 goto failed_mount;
4061 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4062 "to feature incompatibilities");
4063 goto failed_mount;
4064 }
4065 }
4066
4067 /*
4068 * Check feature flags regardless of the revision level, since we
4069 * previously didn't change the revision level when setting the flags,
4070 * so there is a chance incompat flags are set on a rev 0 filesystem.
4071 */
4072 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4073 goto failed_mount;
4074
4075 if (le32_to_cpu(es->s_log_block_size) >
4076 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4077 ext4_msg(sb, KERN_ERR,
4078 "Invalid log block size: %u",
4079 le32_to_cpu(es->s_log_block_size));
4080 goto failed_mount;
4081 }
4082 if (le32_to_cpu(es->s_log_cluster_size) >
4083 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4084 ext4_msg(sb, KERN_ERR,
4085 "Invalid log cluster size: %u",
4086 le32_to_cpu(es->s_log_cluster_size));
4087 goto failed_mount;
4088 }
4089
4090 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4091 ext4_msg(sb, KERN_ERR,
4092 "Number of reserved GDT blocks insanely large: %d",
4093 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4094 goto failed_mount;
4095 }
4096
4097 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
4098 if (ext4_has_feature_inline_data(sb)) {
4099 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4100 " that may contain inline data");
4101 goto failed_mount;
4102 }
4103 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
4104 ext4_msg(sb, KERN_ERR,
4105 "DAX unsupported by block device.");
4106 goto failed_mount;
4107 }
4108 }
4109
4110 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4111 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4112 es->s_encryption_level);
4113 goto failed_mount;
4114 }
4115
4116 if (sb->s_blocksize != blocksize) {
4117 /* Validate the filesystem blocksize */
4118 if (!sb_set_blocksize(sb, blocksize)) {
4119 ext4_msg(sb, KERN_ERR, "bad block size %d",
4120 blocksize);
4121 goto failed_mount;
4122 }
4123
4124 brelse(bh);
4125 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4126 offset = do_div(logical_sb_block, blocksize);
4127 bh = sb_bread_unmovable(sb, logical_sb_block);
4128 if (!bh) {
4129 ext4_msg(sb, KERN_ERR,
4130 "Can't read superblock on 2nd try");
4131 goto failed_mount;
4132 }
4133 es = (struct ext4_super_block *)(bh->b_data + offset);
4134 sbi->s_es = es;
4135 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4136 ext4_msg(sb, KERN_ERR,
4137 "Magic mismatch, very weird!");
4138 goto failed_mount;
4139 }
4140 }
4141
4142 has_huge_files = ext4_has_feature_huge_file(sb);
4143 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4144 has_huge_files);
4145 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4146
4147 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4148 if (ext4_has_feature_64bit(sb)) {
4149 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4150 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4151 !is_power_of_2(sbi->s_desc_size)) {
4152 ext4_msg(sb, KERN_ERR,
4153 "unsupported descriptor size %lu",
4154 sbi->s_desc_size);
4155 goto failed_mount;
4156 }
4157 } else
4158 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4159
4160 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4161 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4162
4163 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4164 if (sbi->s_inodes_per_block == 0)
4165 goto cantfind_ext4;
4166 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4167 sbi->s_inodes_per_group > blocksize * 8) {
4168 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4169 sbi->s_inodes_per_group);
4170 goto failed_mount;
4171 }
4172 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4173 sbi->s_inodes_per_block;
4174 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4175 sbi->s_sbh = bh;
4176 sbi->s_mount_state = le16_to_cpu(es->s_state);
4177 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4178 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4179
4180 for (i = 0; i < 4; i++)
4181 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4182 sbi->s_def_hash_version = es->s_def_hash_version;
4183 if (ext4_has_feature_dir_index(sb)) {
4184 i = le32_to_cpu(es->s_flags);
4185 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4186 sbi->s_hash_unsigned = 3;
4187 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4188 #ifdef __CHAR_UNSIGNED__
4189 if (!sb_rdonly(sb))
4190 es->s_flags |=
4191 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4192 sbi->s_hash_unsigned = 3;
4193 #else
4194 if (!sb_rdonly(sb))
4195 es->s_flags |=
4196 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4197 #endif
4198 }
4199 }
4200
4201 /* Handle clustersize */
4202 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4203 has_bigalloc = ext4_has_feature_bigalloc(sb);
4204 if (has_bigalloc) {
4205 if (clustersize < blocksize) {
4206 ext4_msg(sb, KERN_ERR,
4207 "cluster size (%d) smaller than "
4208 "block size (%d)", clustersize, blocksize);
4209 goto failed_mount;
4210 }
4211 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4212 le32_to_cpu(es->s_log_block_size);
4213 sbi->s_clusters_per_group =
4214 le32_to_cpu(es->s_clusters_per_group);
4215 if (sbi->s_clusters_per_group > blocksize * 8) {
4216 ext4_msg(sb, KERN_ERR,
4217 "#clusters per group too big: %lu",
4218 sbi->s_clusters_per_group);
4219 goto failed_mount;
4220 }
4221 if (sbi->s_blocks_per_group !=
4222 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4223 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4224 "clusters per group (%lu) inconsistent",
4225 sbi->s_blocks_per_group,
4226 sbi->s_clusters_per_group);
4227 goto failed_mount;
4228 }
4229 } else {
4230 if (clustersize != blocksize) {
4231 ext4_msg(sb, KERN_ERR,
4232 "fragment/cluster size (%d) != "
4233 "block size (%d)", clustersize, blocksize);
4234 goto failed_mount;
4235 }
4236 if (sbi->s_blocks_per_group > blocksize * 8) {
4237 ext4_msg(sb, KERN_ERR,
4238 "#blocks per group too big: %lu",
4239 sbi->s_blocks_per_group);
4240 goto failed_mount;
4241 }
4242 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4243 sbi->s_cluster_bits = 0;
4244 }
4245 sbi->s_cluster_ratio = clustersize / blocksize;
4246
4247 /* Do we have standard group size of clustersize * 8 blocks ? */
4248 if (sbi->s_blocks_per_group == clustersize << 3)
4249 set_opt2(sb, STD_GROUP_SIZE);
4250
4251 /*
4252 * Test whether we have more sectors than will fit in sector_t,
4253 * and whether the max offset is addressable by the page cache.
4254 */
4255 err = generic_check_addressable(sb->s_blocksize_bits,
4256 ext4_blocks_count(es));
4257 if (err) {
4258 ext4_msg(sb, KERN_ERR, "filesystem"
4259 " too large to mount safely on this system");
4260 goto failed_mount;
4261 }
4262
4263 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4264 goto cantfind_ext4;
4265
4266 /* check blocks count against device size */
4267 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4268 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4269 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4270 "exceeds size of device (%llu blocks)",
4271 ext4_blocks_count(es), blocks_count);
4272 goto failed_mount;
4273 }
4274
4275 /*
4276 * It makes no sense for the first data block to be beyond the end
4277 * of the filesystem.
4278 */
4279 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4280 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4281 "block %u is beyond end of filesystem (%llu)",
4282 le32_to_cpu(es->s_first_data_block),
4283 ext4_blocks_count(es));
4284 goto failed_mount;
4285 }
4286 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4287 (sbi->s_cluster_ratio == 1)) {
4288 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4289 "block is 0 with a 1k block and cluster size");
4290 goto failed_mount;
4291 }
4292
4293 blocks_count = (ext4_blocks_count(es) -
4294 le32_to_cpu(es->s_first_data_block) +
4295 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4296 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4297 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4298 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4299 "(block count %llu, first data block %u, "
4300 "blocks per group %lu)", blocks_count,
4301 ext4_blocks_count(es),
4302 le32_to_cpu(es->s_first_data_block),
4303 EXT4_BLOCKS_PER_GROUP(sb));
4304 goto failed_mount;
4305 }
4306 sbi->s_groups_count = blocks_count;
4307 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4308 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4309 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4310 le32_to_cpu(es->s_inodes_count)) {
4311 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4312 le32_to_cpu(es->s_inodes_count),
4313 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4314 ret = -EINVAL;
4315 goto failed_mount;
4316 }
4317 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4318 EXT4_DESC_PER_BLOCK(sb);
4319 if (ext4_has_feature_meta_bg(sb)) {
4320 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4321 ext4_msg(sb, KERN_WARNING,
4322 "first meta block group too large: %u "
4323 "(group descriptor block count %u)",
4324 le32_to_cpu(es->s_first_meta_bg), db_count);
4325 goto failed_mount;
4326 }
4327 }
4328 rcu_assign_pointer(sbi->s_group_desc,
4329 kvmalloc_array(db_count,
4330 sizeof(struct buffer_head *),
4331 GFP_KERNEL));
4332 if (sbi->s_group_desc == NULL) {
4333 ext4_msg(sb, KERN_ERR, "not enough memory");
4334 ret = -ENOMEM;
4335 goto failed_mount;
4336 }
4337
4338 bgl_lock_init(sbi->s_blockgroup_lock);
4339
4340 /* Pre-read the descriptors into the buffer cache */
4341 for (i = 0; i < db_count; i++) {
4342 block = descriptor_loc(sb, logical_sb_block, i);
4343 sb_breadahead(sb, block);
4344 }
4345
4346 for (i = 0; i < db_count; i++) {
4347 struct buffer_head *bh;
4348
4349 block = descriptor_loc(sb, logical_sb_block, i);
4350 bh = sb_bread_unmovable(sb, block);
4351 if (!bh) {
4352 ext4_msg(sb, KERN_ERR,
4353 "can't read group descriptor %d", i);
4354 db_count = i;
4355 goto failed_mount2;
4356 }
4357 rcu_read_lock();
4358 rcu_dereference(sbi->s_group_desc)[i] = bh;
4359 rcu_read_unlock();
4360 }
4361 sbi->s_gdb_count = db_count;
4362 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4363 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4364 ret = -EFSCORRUPTED;
4365 goto failed_mount2;
4366 }
4367
4368 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4369
4370 /* Register extent status tree shrinker */
4371 if (ext4_es_register_shrinker(sbi))
4372 goto failed_mount3;
4373
4374 sbi->s_stripe = ext4_get_stripe_size(sbi);
4375 sbi->s_extent_max_zeroout_kb = 32;
4376
4377 /*
4378 * set up enough so that it can read an inode
4379 */
4380 sb->s_op = &ext4_sops;
4381 sb->s_export_op = &ext4_export_ops;
4382 sb->s_xattr = ext4_xattr_handlers;
4383 #ifdef CONFIG_FS_ENCRYPTION
4384 sb->s_cop = &ext4_cryptops;
4385 #endif
4386 #ifdef CONFIG_FS_VERITY
4387 sb->s_vop = &ext4_verityops;
4388 #endif
4389 #ifdef CONFIG_QUOTA
4390 sb->dq_op = &ext4_quota_operations;
4391 if (ext4_has_feature_quota(sb))
4392 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4393 else
4394 sb->s_qcop = &ext4_qctl_operations;
4395 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4396 #endif
4397 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4398
4399 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4400 mutex_init(&sbi->s_orphan_lock);
4401
4402 sb->s_root = NULL;
4403
4404 needs_recovery = (es->s_last_orphan != 0 ||
4405 ext4_has_feature_journal_needs_recovery(sb));
4406
4407 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4408 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4409 goto failed_mount3a;
4410
4411 /*
4412 * The first inode we look at is the journal inode. Don't try
4413 * root first: it may be modified in the journal!
4414 */
4415 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4416 err = ext4_load_journal(sb, es, journal_devnum);
4417 if (err)
4418 goto failed_mount3a;
4419 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4420 ext4_has_feature_journal_needs_recovery(sb)) {
4421 ext4_msg(sb, KERN_ERR, "required journal recovery "
4422 "suppressed and not mounted read-only");
4423 goto failed_mount_wq;
4424 } else {
4425 /* Nojournal mode, all journal mount options are illegal */
4426 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4427 ext4_msg(sb, KERN_ERR, "can't mount with "
4428 "journal_checksum, fs mounted w/o journal");
4429 goto failed_mount_wq;
4430 }
4431 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4432 ext4_msg(sb, KERN_ERR, "can't mount with "
4433 "journal_async_commit, fs mounted w/o journal");
4434 goto failed_mount_wq;
4435 }
4436 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4437 ext4_msg(sb, KERN_ERR, "can't mount with "
4438 "commit=%lu, fs mounted w/o journal",
4439 sbi->s_commit_interval / HZ);
4440 goto failed_mount_wq;
4441 }
4442 if (EXT4_MOUNT_DATA_FLAGS &
4443 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4444 ext4_msg(sb, KERN_ERR, "can't mount with "
4445 "data=, fs mounted w/o journal");
4446 goto failed_mount_wq;
4447 }
4448 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4449 clear_opt(sb, JOURNAL_CHECKSUM);
4450 clear_opt(sb, DATA_FLAGS);
4451 sbi->s_journal = NULL;
4452 needs_recovery = 0;
4453 goto no_journal;
4454 }
4455
4456 if (ext4_has_feature_64bit(sb) &&
4457 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4458 JBD2_FEATURE_INCOMPAT_64BIT)) {
4459 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4460 goto failed_mount_wq;
4461 }
4462
4463 if (!set_journal_csum_feature_set(sb)) {
4464 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4465 "feature set");
4466 goto failed_mount_wq;
4467 }
4468
4469 /* We have now updated the journal if required, so we can
4470 * validate the data journaling mode. */
4471 switch (test_opt(sb, DATA_FLAGS)) {
4472 case 0:
4473 /* No mode set, assume a default based on the journal
4474 * capabilities: ORDERED_DATA if the journal can
4475 * cope, else JOURNAL_DATA
4476 */
4477 if (jbd2_journal_check_available_features
4478 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4479 set_opt(sb, ORDERED_DATA);
4480 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4481 } else {
4482 set_opt(sb, JOURNAL_DATA);
4483 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4484 }
4485 break;
4486
4487 case EXT4_MOUNT_ORDERED_DATA:
4488 case EXT4_MOUNT_WRITEBACK_DATA:
4489 if (!jbd2_journal_check_available_features
4490 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4491 ext4_msg(sb, KERN_ERR, "Journal does not support "
4492 "requested data journaling mode");
4493 goto failed_mount_wq;
4494 }
4495 default:
4496 break;
4497 }
4498
4499 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4500 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4501 ext4_msg(sb, KERN_ERR, "can't mount with "
4502 "journal_async_commit in data=ordered mode");
4503 goto failed_mount_wq;
4504 }
4505
4506 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4507
4508 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4509
4510 no_journal:
4511 if (!test_opt(sb, NO_MBCACHE)) {
4512 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4513 if (!sbi->s_ea_block_cache) {
4514 ext4_msg(sb, KERN_ERR,
4515 "Failed to create ea_block_cache");
4516 goto failed_mount_wq;
4517 }
4518
4519 if (ext4_has_feature_ea_inode(sb)) {
4520 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4521 if (!sbi->s_ea_inode_cache) {
4522 ext4_msg(sb, KERN_ERR,
4523 "Failed to create ea_inode_cache");
4524 goto failed_mount_wq;
4525 }
4526 }
4527 }
4528
4529 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4530 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4531 goto failed_mount_wq;
4532 }
4533
4534 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4535 !ext4_has_feature_encrypt(sb)) {
4536 ext4_set_feature_encrypt(sb);
4537 ext4_commit_super(sb, 1);
4538 }
4539
4540 /*
4541 * Get the # of file system overhead blocks from the
4542 * superblock if present.
4543 */
4544 if (es->s_overhead_clusters)
4545 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4546 else {
4547 err = ext4_calculate_overhead(sb);
4548 if (err)
4549 goto failed_mount_wq;
4550 }
4551
4552 /*
4553 * The maximum number of concurrent works can be high and
4554 * concurrency isn't really necessary. Limit it to 1.
4555 */
4556 EXT4_SB(sb)->rsv_conversion_wq =
4557 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4558 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4559 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4560 ret = -ENOMEM;
4561 goto failed_mount4;
4562 }
4563
4564 /*
4565 * The jbd2_journal_load will have done any necessary log recovery,
4566 * so we can safely mount the rest of the filesystem now.
4567 */
4568
4569 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4570 if (IS_ERR(root)) {
4571 ext4_msg(sb, KERN_ERR, "get root inode failed");
4572 ret = PTR_ERR(root);
4573 root = NULL;
4574 goto failed_mount4;
4575 }
4576 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4577 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4578 iput(root);
4579 goto failed_mount4;
4580 }
4581
4582 #ifdef CONFIG_UNICODE
4583 if (sbi->s_encoding)
4584 sb->s_d_op = &ext4_dentry_ops;
4585 #endif
4586
4587 sb->s_root = d_make_root(root);
4588 if (!sb->s_root) {
4589 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4590 ret = -ENOMEM;
4591 goto failed_mount4;
4592 }
4593
4594 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4595 if (ret == -EROFS) {
4596 sb->s_flags |= SB_RDONLY;
4597 ret = 0;
4598 } else if (ret)
4599 goto failed_mount4a;
4600
4601 ext4_set_resv_clusters(sb);
4602
4603 err = ext4_setup_system_zone(sb);
4604 if (err) {
4605 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4606 "zone (%d)", err);
4607 goto failed_mount4a;
4608 }
4609
4610 ext4_ext_init(sb);
4611 err = ext4_mb_init(sb);
4612 if (err) {
4613 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4614 err);
4615 goto failed_mount5;
4616 }
4617
4618 block = ext4_count_free_clusters(sb);
4619 ext4_free_blocks_count_set(sbi->s_es,
4620 EXT4_C2B(sbi, block));
4621 ext4_superblock_csum_set(sb);
4622 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4623 GFP_KERNEL);
4624 if (!err) {
4625 unsigned long freei = ext4_count_free_inodes(sb);
4626 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4627 ext4_superblock_csum_set(sb);
4628 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4629 GFP_KERNEL);
4630 }
4631 if (!err)
4632 err = percpu_counter_init(&sbi->s_dirs_counter,
4633 ext4_count_dirs(sb), GFP_KERNEL);
4634 if (!err)
4635 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4636 GFP_KERNEL);
4637 if (!err)
4638 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4639
4640 if (err) {
4641 ext4_msg(sb, KERN_ERR, "insufficient memory");
4642 goto failed_mount6;
4643 }
4644
4645 if (ext4_has_feature_flex_bg(sb))
4646 if (!ext4_fill_flex_info(sb)) {
4647 ext4_msg(sb, KERN_ERR,
4648 "unable to initialize "
4649 "flex_bg meta info!");
4650 goto failed_mount6;
4651 }
4652
4653 err = ext4_register_li_request(sb, first_not_zeroed);
4654 if (err)
4655 goto failed_mount6;
4656
4657 err = ext4_register_sysfs(sb);
4658 if (err)
4659 goto failed_mount7;
4660
4661 #ifdef CONFIG_QUOTA
4662 /* Enable quota usage during mount. */
4663 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4664 err = ext4_enable_quotas(sb);
4665 if (err)
4666 goto failed_mount8;
4667 }
4668 #endif /* CONFIG_QUOTA */
4669
4670 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4671 ext4_orphan_cleanup(sb, es);
4672 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4673 if (needs_recovery) {
4674 ext4_msg(sb, KERN_INFO, "recovery complete");
4675 ext4_mark_recovery_complete(sb, es);
4676 }
4677 if (EXT4_SB(sb)->s_journal) {
4678 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4679 descr = " journalled data mode";
4680 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4681 descr = " ordered data mode";
4682 else
4683 descr = " writeback data mode";
4684 } else
4685 descr = "out journal";
4686
4687 if (test_opt(sb, DISCARD)) {
4688 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4689 if (!blk_queue_discard(q))
4690 ext4_msg(sb, KERN_WARNING,
4691 "mounting with \"discard\" option, but "
4692 "the device does not support discard");
4693 }
4694
4695 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4696 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4697 "Opts: %.*s%s%s", descr,
4698 (int) sizeof(sbi->s_es->s_mount_opts),
4699 sbi->s_es->s_mount_opts,
4700 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4701
4702 if (es->s_error_count)
4703 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4704
4705 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4706 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4707 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4708 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4709
4710 kfree(orig_data);
4711 return 0;
4712
4713 cantfind_ext4:
4714 if (!silent)
4715 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4716 goto failed_mount;
4717
4718 #ifdef CONFIG_QUOTA
4719 failed_mount8:
4720 ext4_unregister_sysfs(sb);
4721 #endif
4722 failed_mount7:
4723 ext4_unregister_li_request(sb);
4724 failed_mount6:
4725 ext4_mb_release(sb);
4726 rcu_read_lock();
4727 flex_groups = rcu_dereference(sbi->s_flex_groups);
4728 if (flex_groups) {
4729 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4730 kvfree(flex_groups[i]);
4731 kvfree(flex_groups);
4732 }
4733 rcu_read_unlock();
4734 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4735 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4736 percpu_counter_destroy(&sbi->s_dirs_counter);
4737 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4738 percpu_free_rwsem(&sbi->s_writepages_rwsem);
4739 failed_mount5:
4740 ext4_ext_release(sb);
4741 ext4_release_system_zone(sb);
4742 failed_mount4a:
4743 dput(sb->s_root);
4744 sb->s_root = NULL;
4745 failed_mount4:
4746 ext4_msg(sb, KERN_ERR, "mount failed");
4747 if (EXT4_SB(sb)->rsv_conversion_wq)
4748 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4749 failed_mount_wq:
4750 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4751 sbi->s_ea_inode_cache = NULL;
4752
4753 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4754 sbi->s_ea_block_cache = NULL;
4755
4756 if (sbi->s_journal) {
4757 jbd2_journal_destroy(sbi->s_journal);
4758 sbi->s_journal = NULL;
4759 }
4760 failed_mount3a:
4761 ext4_es_unregister_shrinker(sbi);
4762 failed_mount3:
4763 del_timer_sync(&sbi->s_err_report);
4764 if (sbi->s_mmp_tsk)
4765 kthread_stop(sbi->s_mmp_tsk);
4766 failed_mount2:
4767 rcu_read_lock();
4768 group_desc = rcu_dereference(sbi->s_group_desc);
4769 for (i = 0; i < db_count; i++)
4770 brelse(group_desc[i]);
4771 kvfree(group_desc);
4772 rcu_read_unlock();
4773 failed_mount:
4774 if (sbi->s_chksum_driver)
4775 crypto_free_shash(sbi->s_chksum_driver);
4776
4777 #ifdef CONFIG_UNICODE
4778 utf8_unload(sbi->s_encoding);
4779 #endif
4780
4781 #ifdef CONFIG_QUOTA
4782 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4783 kfree(get_qf_name(sb, sbi, i));
4784 #endif
4785 ext4_blkdev_remove(sbi);
4786 brelse(bh);
4787 out_fail:
4788 sb->s_fs_info = NULL;
4789 kfree(sbi->s_blockgroup_lock);
4790 out_free_base:
4791 kfree(sbi);
4792 kfree(orig_data);
4793 fs_put_dax(dax_dev);
4794 return err ? err : ret;
4795 }
4796
4797 /*
4798 * Setup any per-fs journal parameters now. We'll do this both on
4799 * initial mount, once the journal has been initialised but before we've
4800 * done any recovery; and again on any subsequent remount.
4801 */
4802 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4803 {
4804 struct ext4_sb_info *sbi = EXT4_SB(sb);
4805
4806 journal->j_commit_interval = sbi->s_commit_interval;
4807 journal->j_min_batch_time = sbi->s_min_batch_time;
4808 journal->j_max_batch_time = sbi->s_max_batch_time;
4809
4810 write_lock(&journal->j_state_lock);
4811 if (test_opt(sb, BARRIER))
4812 journal->j_flags |= JBD2_BARRIER;
4813 else
4814 journal->j_flags &= ~JBD2_BARRIER;
4815 if (test_opt(sb, DATA_ERR_ABORT))
4816 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4817 else
4818 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4819 write_unlock(&journal->j_state_lock);
4820 }
4821
4822 static struct inode *ext4_get_journal_inode(struct super_block *sb,
4823 unsigned int journal_inum)
4824 {
4825 struct inode *journal_inode;
4826
4827 /*
4828 * Test for the existence of a valid inode on disk. Bad things
4829 * happen if we iget() an unused inode, as the subsequent iput()
4830 * will try to delete it.
4831 */
4832 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
4833 if (IS_ERR(journal_inode)) {
4834 ext4_msg(sb, KERN_ERR, "no journal found");
4835 return NULL;
4836 }
4837 if (!journal_inode->i_nlink) {
4838 make_bad_inode(journal_inode);
4839 iput(journal_inode);
4840 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4841 return NULL;
4842 }
4843
4844 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4845 journal_inode, journal_inode->i_size);
4846 if (!S_ISREG(journal_inode->i_mode)) {
4847 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4848 iput(journal_inode);
4849 return NULL;
4850 }
4851 return journal_inode;
4852 }
4853
4854 static journal_t *ext4_get_journal(struct super_block *sb,
4855 unsigned int journal_inum)
4856 {
4857 struct inode *journal_inode;
4858 journal_t *journal;
4859
4860 BUG_ON(!ext4_has_feature_journal(sb));
4861
4862 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4863 if (!journal_inode)
4864 return NULL;
4865
4866 journal = jbd2_journal_init_inode(journal_inode);
4867 if (!journal) {
4868 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4869 iput(journal_inode);
4870 return NULL;
4871 }
4872 journal->j_private = sb;
4873 ext4_init_journal_params(sb, journal);
4874 return journal;
4875 }
4876
4877 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4878 dev_t j_dev)
4879 {
4880 struct buffer_head *bh;
4881 journal_t *journal;
4882 ext4_fsblk_t start;
4883 ext4_fsblk_t len;
4884 int hblock, blocksize;
4885 ext4_fsblk_t sb_block;
4886 unsigned long offset;
4887 struct ext4_super_block *es;
4888 struct block_device *bdev;
4889
4890 BUG_ON(!ext4_has_feature_journal(sb));
4891
4892 bdev = ext4_blkdev_get(j_dev, sb);
4893 if (bdev == NULL)
4894 return NULL;
4895
4896 blocksize = sb->s_blocksize;
4897 hblock = bdev_logical_block_size(bdev);
4898 if (blocksize < hblock) {
4899 ext4_msg(sb, KERN_ERR,
4900 "blocksize too small for journal device");
4901 goto out_bdev;
4902 }
4903
4904 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4905 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4906 set_blocksize(bdev, blocksize);
4907 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4908 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4909 "external journal");
4910 goto out_bdev;
4911 }
4912
4913 es = (struct ext4_super_block *) (bh->b_data + offset);
4914 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4915 !(le32_to_cpu(es->s_feature_incompat) &
4916 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4917 ext4_msg(sb, KERN_ERR, "external journal has "
4918 "bad superblock");
4919 brelse(bh);
4920 goto out_bdev;
4921 }
4922
4923 if ((le32_to_cpu(es->s_feature_ro_compat) &
4924 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4925 es->s_checksum != ext4_superblock_csum(sb, es)) {
4926 ext4_msg(sb, KERN_ERR, "external journal has "
4927 "corrupt superblock");
4928 brelse(bh);
4929 goto out_bdev;
4930 }
4931
4932 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4933 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4934 brelse(bh);
4935 goto out_bdev;
4936 }
4937
4938 len = ext4_blocks_count(es);
4939 start = sb_block + 1;
4940 brelse(bh); /* we're done with the superblock */
4941
4942 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4943 start, len, blocksize);
4944 if (!journal) {
4945 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4946 goto out_bdev;
4947 }
4948 journal->j_private = sb;
4949 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4950 wait_on_buffer(journal->j_sb_buffer);
4951 if (!buffer_uptodate(journal->j_sb_buffer)) {
4952 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4953 goto out_journal;
4954 }
4955 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4956 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4957 "user (unsupported) - %d",
4958 be32_to_cpu(journal->j_superblock->s_nr_users));
4959 goto out_journal;
4960 }
4961 EXT4_SB(sb)->journal_bdev = bdev;
4962 ext4_init_journal_params(sb, journal);
4963 return journal;
4964
4965 out_journal:
4966 jbd2_journal_destroy(journal);
4967 out_bdev:
4968 ext4_blkdev_put(bdev);
4969 return NULL;
4970 }
4971
4972 static int ext4_load_journal(struct super_block *sb,
4973 struct ext4_super_block *es,
4974 unsigned long journal_devnum)
4975 {
4976 journal_t *journal;
4977 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4978 dev_t journal_dev;
4979 int err = 0;
4980 int really_read_only;
4981
4982 BUG_ON(!ext4_has_feature_journal(sb));
4983
4984 if (journal_devnum &&
4985 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4986 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4987 "numbers have changed");
4988 journal_dev = new_decode_dev(journal_devnum);
4989 } else
4990 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4991
4992 really_read_only = bdev_read_only(sb->s_bdev);
4993
4994 /*
4995 * Are we loading a blank journal or performing recovery after a
4996 * crash? For recovery, we need to check in advance whether we
4997 * can get read-write access to the device.
4998 */
4999 if (ext4_has_feature_journal_needs_recovery(sb)) {
5000 if (sb_rdonly(sb)) {
5001 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5002 "required on readonly filesystem");
5003 if (really_read_only) {
5004 ext4_msg(sb, KERN_ERR, "write access "
5005 "unavailable, cannot proceed "
5006 "(try mounting with noload)");
5007 return -EROFS;
5008 }
5009 ext4_msg(sb, KERN_INFO, "write access will "
5010 "be enabled during recovery");
5011 }
5012 }
5013
5014 if (journal_inum && journal_dev) {
5015 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
5016 "and inode journals!");
5017 return -EINVAL;
5018 }
5019
5020 if (journal_inum) {
5021 if (!(journal = ext4_get_journal(sb, journal_inum)))
5022 return -EINVAL;
5023 } else {
5024 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
5025 return -EINVAL;
5026 }
5027
5028 if (!(journal->j_flags & JBD2_BARRIER))
5029 ext4_msg(sb, KERN_INFO, "barriers disabled");
5030
5031 if (!ext4_has_feature_journal_needs_recovery(sb))
5032 err = jbd2_journal_wipe(journal, !really_read_only);
5033 if (!err) {
5034 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5035 if (save)
5036 memcpy(save, ((char *) es) +
5037 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5038 err = jbd2_journal_load(journal);
5039 if (save)
5040 memcpy(((char *) es) + EXT4_S_ERR_START,
5041 save, EXT4_S_ERR_LEN);
5042 kfree(save);
5043 }
5044
5045 if (err) {
5046 ext4_msg(sb, KERN_ERR, "error loading journal");
5047 jbd2_journal_destroy(journal);
5048 return err;
5049 }
5050
5051 EXT4_SB(sb)->s_journal = journal;
5052 ext4_clear_journal_err(sb, es);
5053
5054 if (!really_read_only && journal_devnum &&
5055 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5056 es->s_journal_dev = cpu_to_le32(journal_devnum);
5057
5058 /* Make sure we flush the recovery flag to disk. */
5059 ext4_commit_super(sb, 1);
5060 }
5061
5062 return 0;
5063 }
5064
5065 static int ext4_commit_super(struct super_block *sb, int sync)
5066 {
5067 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5068 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5069 int error = 0;
5070
5071 if (!sbh || block_device_ejected(sb))
5072 return error;
5073
5074 /*
5075 * The superblock bh should be mapped, but it might not be if the
5076 * device was hot-removed. Not much we can do but fail the I/O.
5077 */
5078 if (!buffer_mapped(sbh))
5079 return error;
5080
5081 /*
5082 * If the file system is mounted read-only, don't update the
5083 * superblock write time. This avoids updating the superblock
5084 * write time when we are mounting the root file system
5085 * read/only but we need to replay the journal; at that point,
5086 * for people who are east of GMT and who make their clock
5087 * tick in localtime for Windows bug-for-bug compatibility,
5088 * the clock is set in the future, and this will cause e2fsck
5089 * to complain and force a full file system check.
5090 */
5091 if (!(sb->s_flags & SB_RDONLY))
5092 ext4_update_tstamp(es, s_wtime);
5093 if (sb->s_bdev->bd_part)
5094 es->s_kbytes_written =
5095 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5096 ((part_stat_read(sb->s_bdev->bd_part,
5097 sectors[STAT_WRITE]) -
5098 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5099 else
5100 es->s_kbytes_written =
5101 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5102 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5103 ext4_free_blocks_count_set(es,
5104 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5105 &EXT4_SB(sb)->s_freeclusters_counter)));
5106 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5107 es->s_free_inodes_count =
5108 cpu_to_le32(percpu_counter_sum_positive(
5109 &EXT4_SB(sb)->s_freeinodes_counter));
5110 BUFFER_TRACE(sbh, "marking dirty");
5111 ext4_superblock_csum_set(sb);
5112 if (sync)
5113 lock_buffer(sbh);
5114 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5115 /*
5116 * Oh, dear. A previous attempt to write the
5117 * superblock failed. This could happen because the
5118 * USB device was yanked out. Or it could happen to
5119 * be a transient write error and maybe the block will
5120 * be remapped. Nothing we can do but to retry the
5121 * write and hope for the best.
5122 */
5123 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5124 "superblock detected");
5125 clear_buffer_write_io_error(sbh);
5126 set_buffer_uptodate(sbh);
5127 }
5128 mark_buffer_dirty(sbh);
5129 if (sync) {
5130 unlock_buffer(sbh);
5131 error = __sync_dirty_buffer(sbh,
5132 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5133 if (buffer_write_io_error(sbh)) {
5134 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5135 "superblock");
5136 clear_buffer_write_io_error(sbh);
5137 set_buffer_uptodate(sbh);
5138 }
5139 }
5140 return error;
5141 }
5142
5143 /*
5144 * Have we just finished recovery? If so, and if we are mounting (or
5145 * remounting) the filesystem readonly, then we will end up with a
5146 * consistent fs on disk. Record that fact.
5147 */
5148 static void ext4_mark_recovery_complete(struct super_block *sb,
5149 struct ext4_super_block *es)
5150 {
5151 journal_t *journal = EXT4_SB(sb)->s_journal;
5152
5153 if (!ext4_has_feature_journal(sb)) {
5154 BUG_ON(journal != NULL);
5155 return;
5156 }
5157 jbd2_journal_lock_updates(journal);
5158 if (jbd2_journal_flush(journal) < 0)
5159 goto out;
5160
5161 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5162 ext4_clear_feature_journal_needs_recovery(sb);
5163 ext4_commit_super(sb, 1);
5164 }
5165
5166 out:
5167 jbd2_journal_unlock_updates(journal);
5168 }
5169
5170 /*
5171 * If we are mounting (or read-write remounting) a filesystem whose journal
5172 * has recorded an error from a previous lifetime, move that error to the
5173 * main filesystem now.
5174 */
5175 static void ext4_clear_journal_err(struct super_block *sb,
5176 struct ext4_super_block *es)
5177 {
5178 journal_t *journal;
5179 int j_errno;
5180 const char *errstr;
5181
5182 BUG_ON(!ext4_has_feature_journal(sb));
5183
5184 journal = EXT4_SB(sb)->s_journal;
5185
5186 /*
5187 * Now check for any error status which may have been recorded in the
5188 * journal by a prior ext4_error() or ext4_abort()
5189 */
5190
5191 j_errno = jbd2_journal_errno(journal);
5192 if (j_errno) {
5193 char nbuf[16];
5194
5195 errstr = ext4_decode_error(sb, j_errno, nbuf);
5196 ext4_warning(sb, "Filesystem error recorded "
5197 "from previous mount: %s", errstr);
5198 ext4_warning(sb, "Marking fs in need of filesystem check.");
5199
5200 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5201 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5202 ext4_commit_super(sb, 1);
5203
5204 jbd2_journal_clear_err(journal);
5205 jbd2_journal_update_sb_errno(journal);
5206 }
5207 }
5208
5209 /*
5210 * Force the running and committing transactions to commit,
5211 * and wait on the commit.
5212 */
5213 int ext4_force_commit(struct super_block *sb)
5214 {
5215 journal_t *journal;
5216
5217 if (sb_rdonly(sb))
5218 return 0;
5219
5220 journal = EXT4_SB(sb)->s_journal;
5221 return ext4_journal_force_commit(journal);
5222 }
5223
5224 static int ext4_sync_fs(struct super_block *sb, int wait)
5225 {
5226 int ret = 0;
5227 tid_t target;
5228 bool needs_barrier = false;
5229 struct ext4_sb_info *sbi = EXT4_SB(sb);
5230
5231 if (unlikely(ext4_forced_shutdown(sbi)))
5232 return 0;
5233
5234 trace_ext4_sync_fs(sb, wait);
5235 flush_workqueue(sbi->rsv_conversion_wq);
5236 /*
5237 * Writeback quota in non-journalled quota case - journalled quota has
5238 * no dirty dquots
5239 */
5240 dquot_writeback_dquots(sb, -1);
5241 /*
5242 * Data writeback is possible w/o journal transaction, so barrier must
5243 * being sent at the end of the function. But we can skip it if
5244 * transaction_commit will do it for us.
5245 */
5246 if (sbi->s_journal) {
5247 target = jbd2_get_latest_transaction(sbi->s_journal);
5248 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5249 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5250 needs_barrier = true;
5251
5252 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5253 if (wait)
5254 ret = jbd2_log_wait_commit(sbi->s_journal,
5255 target);
5256 }
5257 } else if (wait && test_opt(sb, BARRIER))
5258 needs_barrier = true;
5259 if (needs_barrier) {
5260 int err;
5261 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5262 if (!ret)
5263 ret = err;
5264 }
5265
5266 return ret;
5267 }
5268
5269 /*
5270 * LVM calls this function before a (read-only) snapshot is created. This
5271 * gives us a chance to flush the journal completely and mark the fs clean.
5272 *
5273 * Note that only this function cannot bring a filesystem to be in a clean
5274 * state independently. It relies on upper layer to stop all data & metadata
5275 * modifications.
5276 */
5277 static int ext4_freeze(struct super_block *sb)
5278 {
5279 int error = 0;
5280 journal_t *journal;
5281
5282 if (sb_rdonly(sb))
5283 return 0;
5284
5285 journal = EXT4_SB(sb)->s_journal;
5286
5287 if (journal) {
5288 /* Now we set up the journal barrier. */
5289 jbd2_journal_lock_updates(journal);
5290
5291 /*
5292 * Don't clear the needs_recovery flag if we failed to
5293 * flush the journal.
5294 */
5295 error = jbd2_journal_flush(journal);
5296 if (error < 0)
5297 goto out;
5298
5299 /* Journal blocked and flushed, clear needs_recovery flag. */
5300 ext4_clear_feature_journal_needs_recovery(sb);
5301 }
5302
5303 error = ext4_commit_super(sb, 1);
5304 out:
5305 if (journal)
5306 /* we rely on upper layer to stop further updates */
5307 jbd2_journal_unlock_updates(journal);
5308 return error;
5309 }
5310
5311 /*
5312 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5313 * flag here, even though the filesystem is not technically dirty yet.
5314 */
5315 static int ext4_unfreeze(struct super_block *sb)
5316 {
5317 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5318 return 0;
5319
5320 if (EXT4_SB(sb)->s_journal) {
5321 /* Reset the needs_recovery flag before the fs is unlocked. */
5322 ext4_set_feature_journal_needs_recovery(sb);
5323 }
5324
5325 ext4_commit_super(sb, 1);
5326 return 0;
5327 }
5328
5329 /*
5330 * Structure to save mount options for ext4_remount's benefit
5331 */
5332 struct ext4_mount_options {
5333 unsigned long s_mount_opt;
5334 unsigned long s_mount_opt2;
5335 kuid_t s_resuid;
5336 kgid_t s_resgid;
5337 unsigned long s_commit_interval;
5338 u32 s_min_batch_time, s_max_batch_time;
5339 #ifdef CONFIG_QUOTA
5340 int s_jquota_fmt;
5341 char *s_qf_names[EXT4_MAXQUOTAS];
5342 #endif
5343 };
5344
5345 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5346 {
5347 struct ext4_super_block *es;
5348 struct ext4_sb_info *sbi = EXT4_SB(sb);
5349 unsigned long old_sb_flags;
5350 struct ext4_mount_options old_opts;
5351 int enable_quota = 0;
5352 ext4_group_t g;
5353 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5354 int err = 0;
5355 #ifdef CONFIG_QUOTA
5356 int i, j;
5357 char *to_free[EXT4_MAXQUOTAS];
5358 #endif
5359 char *orig_data = kstrdup(data, GFP_KERNEL);
5360
5361 if (data && !orig_data)
5362 return -ENOMEM;
5363
5364 /* Store the original options */
5365 old_sb_flags = sb->s_flags;
5366 old_opts.s_mount_opt = sbi->s_mount_opt;
5367 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5368 old_opts.s_resuid = sbi->s_resuid;
5369 old_opts.s_resgid = sbi->s_resgid;
5370 old_opts.s_commit_interval = sbi->s_commit_interval;
5371 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5372 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5373 #ifdef CONFIG_QUOTA
5374 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5375 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5376 if (sbi->s_qf_names[i]) {
5377 char *qf_name = get_qf_name(sb, sbi, i);
5378
5379 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5380 if (!old_opts.s_qf_names[i]) {
5381 for (j = 0; j < i; j++)
5382 kfree(old_opts.s_qf_names[j]);
5383 kfree(orig_data);
5384 return -ENOMEM;
5385 }
5386 } else
5387 old_opts.s_qf_names[i] = NULL;
5388 #endif
5389 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5390 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5391
5392 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5393 err = -EINVAL;
5394 goto restore_opts;
5395 }
5396
5397 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5398 test_opt(sb, JOURNAL_CHECKSUM)) {
5399 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5400 "during remount not supported; ignoring");
5401 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5402 }
5403
5404 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5405 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5406 ext4_msg(sb, KERN_ERR, "can't mount with "
5407 "both data=journal and delalloc");
5408 err = -EINVAL;
5409 goto restore_opts;
5410 }
5411 if (test_opt(sb, DIOREAD_NOLOCK)) {
5412 ext4_msg(sb, KERN_ERR, "can't mount with "
5413 "both data=journal and dioread_nolock");
5414 err = -EINVAL;
5415 goto restore_opts;
5416 }
5417 if (test_opt(sb, DAX)) {
5418 ext4_msg(sb, KERN_ERR, "can't mount with "
5419 "both data=journal and dax");
5420 err = -EINVAL;
5421 goto restore_opts;
5422 }
5423 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5424 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5425 ext4_msg(sb, KERN_ERR, "can't mount with "
5426 "journal_async_commit in data=ordered mode");
5427 err = -EINVAL;
5428 goto restore_opts;
5429 }
5430 }
5431
5432 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5433 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5434 err = -EINVAL;
5435 goto restore_opts;
5436 }
5437
5438 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5439 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5440 "dax flag with busy inodes while remounting");
5441 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5442 }
5443
5444 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5445 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5446
5447 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5448 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5449
5450 es = sbi->s_es;
5451
5452 if (sbi->s_journal) {
5453 ext4_init_journal_params(sb, sbi->s_journal);
5454 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5455 }
5456
5457 if (*flags & SB_LAZYTIME)
5458 sb->s_flags |= SB_LAZYTIME;
5459
5460 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5461 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5462 err = -EROFS;
5463 goto restore_opts;
5464 }
5465
5466 if (*flags & SB_RDONLY) {
5467 err = sync_filesystem(sb);
5468 if (err < 0)
5469 goto restore_opts;
5470 err = dquot_suspend(sb, -1);
5471 if (err < 0)
5472 goto restore_opts;
5473
5474 /*
5475 * First of all, the unconditional stuff we have to do
5476 * to disable replay of the journal when we next remount
5477 */
5478 sb->s_flags |= SB_RDONLY;
5479
5480 /*
5481 * OK, test if we are remounting a valid rw partition
5482 * readonly, and if so set the rdonly flag and then
5483 * mark the partition as valid again.
5484 */
5485 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5486 (sbi->s_mount_state & EXT4_VALID_FS))
5487 es->s_state = cpu_to_le16(sbi->s_mount_state);
5488
5489 if (sbi->s_journal)
5490 ext4_mark_recovery_complete(sb, es);
5491 if (sbi->s_mmp_tsk)
5492 kthread_stop(sbi->s_mmp_tsk);
5493 } else {
5494 /* Make sure we can mount this feature set readwrite */
5495 if (ext4_has_feature_readonly(sb) ||
5496 !ext4_feature_set_ok(sb, 0)) {
5497 err = -EROFS;
5498 goto restore_opts;
5499 }
5500 /*
5501 * Make sure the group descriptor checksums
5502 * are sane. If they aren't, refuse to remount r/w.
5503 */
5504 for (g = 0; g < sbi->s_groups_count; g++) {
5505 struct ext4_group_desc *gdp =
5506 ext4_get_group_desc(sb, g, NULL);
5507
5508 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5509 ext4_msg(sb, KERN_ERR,
5510 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5511 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5512 le16_to_cpu(gdp->bg_checksum));
5513 err = -EFSBADCRC;
5514 goto restore_opts;
5515 }
5516 }
5517
5518 /*
5519 * If we have an unprocessed orphan list hanging
5520 * around from a previously readonly bdev mount,
5521 * require a full umount/remount for now.
5522 */
5523 if (es->s_last_orphan) {
5524 ext4_msg(sb, KERN_WARNING, "Couldn't "
5525 "remount RDWR because of unprocessed "
5526 "orphan inode list. Please "
5527 "umount/remount instead");
5528 err = -EINVAL;
5529 goto restore_opts;
5530 }
5531
5532 /*
5533 * Mounting a RDONLY partition read-write, so reread
5534 * and store the current valid flag. (It may have
5535 * been changed by e2fsck since we originally mounted
5536 * the partition.)
5537 */
5538 if (sbi->s_journal)
5539 ext4_clear_journal_err(sb, es);
5540 sbi->s_mount_state = le16_to_cpu(es->s_state);
5541
5542 err = ext4_setup_super(sb, es, 0);
5543 if (err)
5544 goto restore_opts;
5545
5546 sb->s_flags &= ~SB_RDONLY;
5547 if (ext4_has_feature_mmp(sb))
5548 if (ext4_multi_mount_protect(sb,
5549 le64_to_cpu(es->s_mmp_block))) {
5550 err = -EROFS;
5551 goto restore_opts;
5552 }
5553 enable_quota = 1;
5554 }
5555 }
5556
5557 /*
5558 * Reinitialize lazy itable initialization thread based on
5559 * current settings
5560 */
5561 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5562 ext4_unregister_li_request(sb);
5563 else {
5564 ext4_group_t first_not_zeroed;
5565 first_not_zeroed = ext4_has_uninit_itable(sb);
5566 ext4_register_li_request(sb, first_not_zeroed);
5567 }
5568
5569 ext4_setup_system_zone(sb);
5570 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5571 err = ext4_commit_super(sb, 1);
5572 if (err)
5573 goto restore_opts;
5574 }
5575
5576 #ifdef CONFIG_QUOTA
5577 /* Release old quota file names */
5578 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5579 kfree(old_opts.s_qf_names[i]);
5580 if (enable_quota) {
5581 if (sb_any_quota_suspended(sb))
5582 dquot_resume(sb, -1);
5583 else if (ext4_has_feature_quota(sb)) {
5584 err = ext4_enable_quotas(sb);
5585 if (err)
5586 goto restore_opts;
5587 }
5588 }
5589 #endif
5590
5591 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
5592 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5593 kfree(orig_data);
5594 return 0;
5595
5596 restore_opts:
5597 sb->s_flags = old_sb_flags;
5598 sbi->s_mount_opt = old_opts.s_mount_opt;
5599 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5600 sbi->s_resuid = old_opts.s_resuid;
5601 sbi->s_resgid = old_opts.s_resgid;
5602 sbi->s_commit_interval = old_opts.s_commit_interval;
5603 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5604 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5605 #ifdef CONFIG_QUOTA
5606 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5607 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5608 to_free[i] = get_qf_name(sb, sbi, i);
5609 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5610 }
5611 synchronize_rcu();
5612 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5613 kfree(to_free[i]);
5614 #endif
5615 kfree(orig_data);
5616 return err;
5617 }
5618
5619 #ifdef CONFIG_QUOTA
5620 static int ext4_statfs_project(struct super_block *sb,
5621 kprojid_t projid, struct kstatfs *buf)
5622 {
5623 struct kqid qid;
5624 struct dquot *dquot;
5625 u64 limit;
5626 u64 curblock;
5627
5628 qid = make_kqid_projid(projid);
5629 dquot = dqget(sb, qid);
5630 if (IS_ERR(dquot))
5631 return PTR_ERR(dquot);
5632 spin_lock(&dquot->dq_dqb_lock);
5633
5634 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
5635 dquot->dq_dqb.dqb_bhardlimit);
5636 limit >>= sb->s_blocksize_bits;
5637
5638 if (limit && buf->f_blocks > limit) {
5639 curblock = (dquot->dq_dqb.dqb_curspace +
5640 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
5641 buf->f_blocks = limit;
5642 buf->f_bfree = buf->f_bavail =
5643 (buf->f_blocks > curblock) ?
5644 (buf->f_blocks - curblock) : 0;
5645 }
5646
5647 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
5648 dquot->dq_dqb.dqb_ihardlimit);
5649 if (limit && buf->f_files > limit) {
5650 buf->f_files = limit;
5651 buf->f_ffree =
5652 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5653 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5654 }
5655
5656 spin_unlock(&dquot->dq_dqb_lock);
5657 dqput(dquot);
5658 return 0;
5659 }
5660 #endif
5661
5662 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5663 {
5664 struct super_block *sb = dentry->d_sb;
5665 struct ext4_sb_info *sbi = EXT4_SB(sb);
5666 struct ext4_super_block *es = sbi->s_es;
5667 ext4_fsblk_t overhead = 0, resv_blocks;
5668 u64 fsid;
5669 s64 bfree;
5670 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5671
5672 if (!test_opt(sb, MINIX_DF))
5673 overhead = sbi->s_overhead;
5674
5675 buf->f_type = EXT4_SUPER_MAGIC;
5676 buf->f_bsize = sb->s_blocksize;
5677 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5678 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5679 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5680 /* prevent underflow in case that few free space is available */
5681 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5682 buf->f_bavail = buf->f_bfree -
5683 (ext4_r_blocks_count(es) + resv_blocks);
5684 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5685 buf->f_bavail = 0;
5686 buf->f_files = le32_to_cpu(es->s_inodes_count);
5687 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5688 buf->f_namelen = EXT4_NAME_LEN;
5689 fsid = le64_to_cpup((void *)es->s_uuid) ^
5690 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5691 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5692 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5693
5694 #ifdef CONFIG_QUOTA
5695 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5696 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5697 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5698 #endif
5699 return 0;
5700 }
5701
5702
5703 #ifdef CONFIG_QUOTA
5704
5705 /*
5706 * Helper functions so that transaction is started before we acquire dqio_sem
5707 * to keep correct lock ordering of transaction > dqio_sem
5708 */
5709 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5710 {
5711 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5712 }
5713
5714 static int ext4_write_dquot(struct dquot *dquot)
5715 {
5716 int ret, err;
5717 handle_t *handle;
5718 struct inode *inode;
5719
5720 inode = dquot_to_inode(dquot);
5721 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5722 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5723 if (IS_ERR(handle))
5724 return PTR_ERR(handle);
5725 ret = dquot_commit(dquot);
5726 err = ext4_journal_stop(handle);
5727 if (!ret)
5728 ret = err;
5729 return ret;
5730 }
5731
5732 static int ext4_acquire_dquot(struct dquot *dquot)
5733 {
5734 int ret, err;
5735 handle_t *handle;
5736
5737 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5738 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5739 if (IS_ERR(handle))
5740 return PTR_ERR(handle);
5741 ret = dquot_acquire(dquot);
5742 err = ext4_journal_stop(handle);
5743 if (!ret)
5744 ret = err;
5745 return ret;
5746 }
5747
5748 static int ext4_release_dquot(struct dquot *dquot)
5749 {
5750 int ret, err;
5751 handle_t *handle;
5752
5753 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5754 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5755 if (IS_ERR(handle)) {
5756 /* Release dquot anyway to avoid endless cycle in dqput() */
5757 dquot_release(dquot);
5758 return PTR_ERR(handle);
5759 }
5760 ret = dquot_release(dquot);
5761 err = ext4_journal_stop(handle);
5762 if (!ret)
5763 ret = err;
5764 return ret;
5765 }
5766
5767 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5768 {
5769 struct super_block *sb = dquot->dq_sb;
5770 struct ext4_sb_info *sbi = EXT4_SB(sb);
5771
5772 /* Are we journaling quotas? */
5773 if (ext4_has_feature_quota(sb) ||
5774 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5775 dquot_mark_dquot_dirty(dquot);
5776 return ext4_write_dquot(dquot);
5777 } else {
5778 return dquot_mark_dquot_dirty(dquot);
5779 }
5780 }
5781
5782 static int ext4_write_info(struct super_block *sb, int type)
5783 {
5784 int ret, err;
5785 handle_t *handle;
5786
5787 /* Data block + inode block */
5788 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5789 if (IS_ERR(handle))
5790 return PTR_ERR(handle);
5791 ret = dquot_commit_info(sb, type);
5792 err = ext4_journal_stop(handle);
5793 if (!ret)
5794 ret = err;
5795 return ret;
5796 }
5797
5798 /*
5799 * Turn on quotas during mount time - we need to find
5800 * the quota file and such...
5801 */
5802 static int ext4_quota_on_mount(struct super_block *sb, int type)
5803 {
5804 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
5805 EXT4_SB(sb)->s_jquota_fmt, type);
5806 }
5807
5808 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5809 {
5810 struct ext4_inode_info *ei = EXT4_I(inode);
5811
5812 /* The first argument of lockdep_set_subclass has to be
5813 * *exactly* the same as the argument to init_rwsem() --- in
5814 * this case, in init_once() --- or lockdep gets unhappy
5815 * because the name of the lock is set using the
5816 * stringification of the argument to init_rwsem().
5817 */
5818 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5819 lockdep_set_subclass(&ei->i_data_sem, subclass);
5820 }
5821
5822 /*
5823 * Standard function to be called on quota_on
5824 */
5825 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5826 const struct path *path)
5827 {
5828 int err;
5829
5830 if (!test_opt(sb, QUOTA))
5831 return -EINVAL;
5832
5833 /* Quotafile not on the same filesystem? */
5834 if (path->dentry->d_sb != sb)
5835 return -EXDEV;
5836 /* Journaling quota? */
5837 if (EXT4_SB(sb)->s_qf_names[type]) {
5838 /* Quotafile not in fs root? */
5839 if (path->dentry->d_parent != sb->s_root)
5840 ext4_msg(sb, KERN_WARNING,
5841 "Quota file not on filesystem root. "
5842 "Journaled quota will not work");
5843 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5844 } else {
5845 /*
5846 * Clear the flag just in case mount options changed since
5847 * last time.
5848 */
5849 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5850 }
5851
5852 /*
5853 * When we journal data on quota file, we have to flush journal to see
5854 * all updates to the file when we bypass pagecache...
5855 */
5856 if (EXT4_SB(sb)->s_journal &&
5857 ext4_should_journal_data(d_inode(path->dentry))) {
5858 /*
5859 * We don't need to lock updates but journal_flush() could
5860 * otherwise be livelocked...
5861 */
5862 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5863 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5864 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5865 if (err)
5866 return err;
5867 }
5868
5869 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5870 err = dquot_quota_on(sb, type, format_id, path);
5871 if (err) {
5872 lockdep_set_quota_inode(path->dentry->d_inode,
5873 I_DATA_SEM_NORMAL);
5874 } else {
5875 struct inode *inode = d_inode(path->dentry);
5876 handle_t *handle;
5877
5878 /*
5879 * Set inode flags to prevent userspace from messing with quota
5880 * files. If this fails, we return success anyway since quotas
5881 * are already enabled and this is not a hard failure.
5882 */
5883 inode_lock(inode);
5884 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5885 if (IS_ERR(handle))
5886 goto unlock_inode;
5887 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5888 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5889 S_NOATIME | S_IMMUTABLE);
5890 ext4_mark_inode_dirty(handle, inode);
5891 ext4_journal_stop(handle);
5892 unlock_inode:
5893 inode_unlock(inode);
5894 }
5895 return err;
5896 }
5897
5898 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5899 unsigned int flags)
5900 {
5901 int err;
5902 struct inode *qf_inode;
5903 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5904 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5905 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5906 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5907 };
5908
5909 BUG_ON(!ext4_has_feature_quota(sb));
5910
5911 if (!qf_inums[type])
5912 return -EPERM;
5913
5914 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
5915 if (IS_ERR(qf_inode)) {
5916 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5917 return PTR_ERR(qf_inode);
5918 }
5919
5920 /* Don't account quota for quota files to avoid recursion */
5921 qf_inode->i_flags |= S_NOQUOTA;
5922 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5923 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
5924 if (err)
5925 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5926 iput(qf_inode);
5927
5928 return err;
5929 }
5930
5931 /* Enable usage tracking for all quota types. */
5932 static int ext4_enable_quotas(struct super_block *sb)
5933 {
5934 int type, err = 0;
5935 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5936 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5937 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5938 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5939 };
5940 bool quota_mopt[EXT4_MAXQUOTAS] = {
5941 test_opt(sb, USRQUOTA),
5942 test_opt(sb, GRPQUOTA),
5943 test_opt(sb, PRJQUOTA),
5944 };
5945
5946 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5947 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5948 if (qf_inums[type]) {
5949 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5950 DQUOT_USAGE_ENABLED |
5951 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5952 if (err) {
5953 ext4_warning(sb,
5954 "Failed to enable quota tracking "
5955 "(type=%d, err=%d). Please run "
5956 "e2fsck to fix.", type, err);
5957 for (type--; type >= 0; type--)
5958 dquot_quota_off(sb, type);
5959
5960 return err;
5961 }
5962 }
5963 }
5964 return 0;
5965 }
5966
5967 static int ext4_quota_off(struct super_block *sb, int type)
5968 {
5969 struct inode *inode = sb_dqopt(sb)->files[type];
5970 handle_t *handle;
5971 int err;
5972
5973 /* Force all delayed allocation blocks to be allocated.
5974 * Caller already holds s_umount sem */
5975 if (test_opt(sb, DELALLOC))
5976 sync_filesystem(sb);
5977
5978 if (!inode || !igrab(inode))
5979 goto out;
5980
5981 err = dquot_quota_off(sb, type);
5982 if (err || ext4_has_feature_quota(sb))
5983 goto out_put;
5984
5985 inode_lock(inode);
5986 /*
5987 * Update modification times of quota files when userspace can
5988 * start looking at them. If we fail, we return success anyway since
5989 * this is not a hard failure and quotas are already disabled.
5990 */
5991 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5992 if (IS_ERR(handle))
5993 goto out_unlock;
5994 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5995 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5996 inode->i_mtime = inode->i_ctime = current_time(inode);
5997 ext4_mark_inode_dirty(handle, inode);
5998 ext4_journal_stop(handle);
5999 out_unlock:
6000 inode_unlock(inode);
6001 out_put:
6002 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6003 iput(inode);
6004 return err;
6005 out:
6006 return dquot_quota_off(sb, type);
6007 }
6008
6009 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6010 * acquiring the locks... As quota files are never truncated and quota code
6011 * itself serializes the operations (and no one else should touch the files)
6012 * we don't have to be afraid of races */
6013 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6014 size_t len, loff_t off)
6015 {
6016 struct inode *inode = sb_dqopt(sb)->files[type];
6017 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6018 int offset = off & (sb->s_blocksize - 1);
6019 int tocopy;
6020 size_t toread;
6021 struct buffer_head *bh;
6022 loff_t i_size = i_size_read(inode);
6023
6024 if (off > i_size)
6025 return 0;
6026 if (off+len > i_size)
6027 len = i_size-off;
6028 toread = len;
6029 while (toread > 0) {
6030 tocopy = sb->s_blocksize - offset < toread ?
6031 sb->s_blocksize - offset : toread;
6032 bh = ext4_bread(NULL, inode, blk, 0);
6033 if (IS_ERR(bh))
6034 return PTR_ERR(bh);
6035 if (!bh) /* A hole? */
6036 memset(data, 0, tocopy);
6037 else
6038 memcpy(data, bh->b_data+offset, tocopy);
6039 brelse(bh);
6040 offset = 0;
6041 toread -= tocopy;
6042 data += tocopy;
6043 blk++;
6044 }
6045 return len;
6046 }
6047
6048 /* Write to quotafile (we know the transaction is already started and has
6049 * enough credits) */
6050 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6051 const char *data, size_t len, loff_t off)
6052 {
6053 struct inode *inode = sb_dqopt(sb)->files[type];
6054 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6055 int err, offset = off & (sb->s_blocksize - 1);
6056 int retries = 0;
6057 struct buffer_head *bh;
6058 handle_t *handle = journal_current_handle();
6059
6060 if (EXT4_SB(sb)->s_journal && !handle) {
6061 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6062 " cancelled because transaction is not started",
6063 (unsigned long long)off, (unsigned long long)len);
6064 return -EIO;
6065 }
6066 /*
6067 * Since we account only one data block in transaction credits,
6068 * then it is impossible to cross a block boundary.
6069 */
6070 if (sb->s_blocksize - offset < len) {
6071 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6072 " cancelled because not block aligned",
6073 (unsigned long long)off, (unsigned long long)len);
6074 return -EIO;
6075 }
6076
6077 do {
6078 bh = ext4_bread(handle, inode, blk,
6079 EXT4_GET_BLOCKS_CREATE |
6080 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6081 } while (PTR_ERR(bh) == -ENOSPC &&
6082 ext4_should_retry_alloc(inode->i_sb, &retries));
6083 if (IS_ERR(bh))
6084 return PTR_ERR(bh);
6085 if (!bh)
6086 goto out;
6087 BUFFER_TRACE(bh, "get write access");
6088 err = ext4_journal_get_write_access(handle, bh);
6089 if (err) {
6090 brelse(bh);
6091 return err;
6092 }
6093 lock_buffer(bh);
6094 memcpy(bh->b_data+offset, data, len);
6095 flush_dcache_page(bh->b_page);
6096 unlock_buffer(bh);
6097 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6098 brelse(bh);
6099 out:
6100 if (inode->i_size < off + len) {
6101 i_size_write(inode, off + len);
6102 EXT4_I(inode)->i_disksize = inode->i_size;
6103 ext4_mark_inode_dirty(handle, inode);
6104 }
6105 return len;
6106 }
6107 #endif
6108
6109 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6110 const char *dev_name, void *data)
6111 {
6112 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6113 }
6114
6115 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6116 static inline void register_as_ext2(void)
6117 {
6118 int err = register_filesystem(&ext2_fs_type);
6119 if (err)
6120 printk(KERN_WARNING
6121 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6122 }
6123
6124 static inline void unregister_as_ext2(void)
6125 {
6126 unregister_filesystem(&ext2_fs_type);
6127 }
6128
6129 static inline int ext2_feature_set_ok(struct super_block *sb)
6130 {
6131 if (ext4_has_unknown_ext2_incompat_features(sb))
6132 return 0;
6133 if (sb_rdonly(sb))
6134 return 1;
6135 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6136 return 0;
6137 return 1;
6138 }
6139 #else
6140 static inline void register_as_ext2(void) { }
6141 static inline void unregister_as_ext2(void) { }
6142 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6143 #endif
6144
6145 static inline void register_as_ext3(void)
6146 {
6147 int err = register_filesystem(&ext3_fs_type);
6148 if (err)
6149 printk(KERN_WARNING
6150 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6151 }
6152
6153 static inline void unregister_as_ext3(void)
6154 {
6155 unregister_filesystem(&ext3_fs_type);
6156 }
6157
6158 static inline int ext3_feature_set_ok(struct super_block *sb)
6159 {
6160 if (ext4_has_unknown_ext3_incompat_features(sb))
6161 return 0;
6162 if (!ext4_has_feature_journal(sb))
6163 return 0;
6164 if (sb_rdonly(sb))
6165 return 1;
6166 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6167 return 0;
6168 return 1;
6169 }
6170
6171 static struct file_system_type ext4_fs_type = {
6172 .owner = THIS_MODULE,
6173 .name = "ext4",
6174 .mount = ext4_mount,
6175 .kill_sb = kill_block_super,
6176 .fs_flags = FS_REQUIRES_DEV,
6177 };
6178 MODULE_ALIAS_FS("ext4");
6179
6180 /* Shared across all ext4 file systems */
6181 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6182
6183 static int __init ext4_init_fs(void)
6184 {
6185 int i, err;
6186
6187 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6188 ext4_li_info = NULL;
6189 mutex_init(&ext4_li_mtx);
6190
6191 /* Build-time check for flags consistency */
6192 ext4_check_flag_values();
6193
6194 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6195 init_waitqueue_head(&ext4__ioend_wq[i]);
6196
6197 err = ext4_init_es();
6198 if (err)
6199 return err;
6200
6201 err = ext4_init_pending();
6202 if (err)
6203 goto out7;
6204
6205 err = ext4_init_post_read_processing();
6206 if (err)
6207 goto out6;
6208
6209 err = ext4_init_pageio();
6210 if (err)
6211 goto out5;
6212
6213 err = ext4_init_system_zone();
6214 if (err)
6215 goto out4;
6216
6217 err = ext4_init_sysfs();
6218 if (err)
6219 goto out3;
6220
6221 err = ext4_init_mballoc();
6222 if (err)
6223 goto out2;
6224 err = init_inodecache();
6225 if (err)
6226 goto out1;
6227 register_as_ext3();
6228 register_as_ext2();
6229 err = register_filesystem(&ext4_fs_type);
6230 if (err)
6231 goto out;
6232
6233 return 0;
6234 out:
6235 unregister_as_ext2();
6236 unregister_as_ext3();
6237 destroy_inodecache();
6238 out1:
6239 ext4_exit_mballoc();
6240 out2:
6241 ext4_exit_sysfs();
6242 out3:
6243 ext4_exit_system_zone();
6244 out4:
6245 ext4_exit_pageio();
6246 out5:
6247 ext4_exit_post_read_processing();
6248 out6:
6249 ext4_exit_pending();
6250 out7:
6251 ext4_exit_es();
6252
6253 return err;
6254 }
6255
6256 static void __exit ext4_exit_fs(void)
6257 {
6258 ext4_destroy_lazyinit_thread();
6259 unregister_as_ext2();
6260 unregister_as_ext3();
6261 unregister_filesystem(&ext4_fs_type);
6262 destroy_inodecache();
6263 ext4_exit_mballoc();
6264 ext4_exit_sysfs();
6265 ext4_exit_system_zone();
6266 ext4_exit_pageio();
6267 ext4_exit_post_read_processing();
6268 ext4_exit_es();
6269 ext4_exit_pending();
6270 }
6271
6272 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6273 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6274 MODULE_LICENSE("GPL");
6275 MODULE_SOFTDEP("pre: crc32c");
6276 module_init(ext4_init_fs)
6277 module_exit(ext4_exit_fs)
Linux with added FPC-III support