projects / linux-block.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
ext4: check for non-zero journal inum in ext4_calculate_overhead
[linux-block.git] / fs / ext4 / super.c
... / ...
CommitLineData
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
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
61static struct ext4_lazy_init *ext4_li_info;
62static struct mutex ext4_li_mtx;
63static struct ratelimit_state ext4_mount_msg_ratelimit;
64
65static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
66 unsigned long journal_devnum);
67static int ext4_show_options(struct seq_file *seq, struct dentry *root);
68static int ext4_commit_super(struct super_block *sb, int sync);
69static void ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71static void ext4_clear_journal_err(struct super_block *sb,
72 struct ext4_super_block *es);
73static int ext4_sync_fs(struct super_block *sb, int wait);
74static int ext4_remount(struct super_block *sb, int *flags, char *data);
75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76static int ext4_unfreeze(struct super_block *sb);
77static int ext4_freeze(struct super_block *sb);
78static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
82static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83static void ext4_destroy_lazyinit_thread(void);
84static void ext4_unregister_li_request(struct super_block *sb);
85static void ext4_clear_request_list(void);
86static 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)
118static 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};
125MODULE_ALIAS_FS("ext2");
126MODULE_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
133static 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};
140MODULE_ALIAS_FS("ext3");
141MODULE_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 */
150struct buffer_head *
151ext4_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
167static 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
176static __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
188static 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
197void 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
207ext4_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
215ext4_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
223ext4_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
263void 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
271void 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
279void 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
287void 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
295void 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
303void 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
311void 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
319static 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
329static 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
338static void __save_error_info(struct super_block *sb, const char *func,
339 unsigned int line)
340{
341 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
342
343 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
344 if (bdev_read_only(sb->s_bdev))
345 return;
346 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
347 ext4_update_tstamp(es, s_last_error_time);
348 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
349 es->s_last_error_line = cpu_to_le32(line);
350 if (es->s_last_error_errcode == 0)
351 es->s_last_error_errcode = EXT4_ERR_EFSCORRUPTED;
352 if (!es->s_first_error_time) {
353 es->s_first_error_time = es->s_last_error_time;
354 es->s_first_error_time_hi = es->s_last_error_time_hi;
355 strncpy(es->s_first_error_func, func,
356 sizeof(es->s_first_error_func));
357 es->s_first_error_line = cpu_to_le32(line);
358 es->s_first_error_ino = es->s_last_error_ino;
359 es->s_first_error_block = es->s_last_error_block;
360 es->s_first_error_errcode = es->s_last_error_errcode;
361 }
362 /*
363 * Start the daily error reporting function if it hasn't been
364 * started already
365 */
366 if (!es->s_error_count)
367 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
368 le32_add_cpu(&es->s_error_count, 1);
369}
370
371static void save_error_info(struct super_block *sb, const char *func,
372 unsigned int line)
373{
374 __save_error_info(sb, func, line);
375 ext4_commit_super(sb, 1);
376}
377
378/*
379 * The del_gendisk() function uninitializes the disk-specific data
380 * structures, including the bdi structure, without telling anyone
381 * else. Once this happens, any attempt to call mark_buffer_dirty()
382 * (for example, by ext4_commit_super), will cause a kernel OOPS.
383 * This is a kludge to prevent these oops until we can put in a proper
384 * hook in del_gendisk() to inform the VFS and file system layers.
385 */
386static int block_device_ejected(struct super_block *sb)
387{
388 struct inode *bd_inode = sb->s_bdev->bd_inode;
389 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
390
391 return bdi->dev == NULL;
392}
393
394static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
395{
396 struct super_block *sb = journal->j_private;
397 struct ext4_sb_info *sbi = EXT4_SB(sb);
398 int error = is_journal_aborted(journal);
399 struct ext4_journal_cb_entry *jce;
400
401 BUG_ON(txn->t_state == T_FINISHED);
402
403 ext4_process_freed_data(sb, txn->t_tid);
404
405 spin_lock(&sbi->s_md_lock);
406 while (!list_empty(&txn->t_private_list)) {
407 jce = list_entry(txn->t_private_list.next,
408 struct ext4_journal_cb_entry, jce_list);
409 list_del_init(&jce->jce_list);
410 spin_unlock(&sbi->s_md_lock);
411 jce->jce_func(sb, jce, error);
412 spin_lock(&sbi->s_md_lock);
413 }
414 spin_unlock(&sbi->s_md_lock);
415}
416
417static bool system_going_down(void)
418{
419 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
420 || system_state == SYSTEM_RESTART;
421}
422
423/* Deal with the reporting of failure conditions on a filesystem such as
424 * inconsistencies detected or read IO failures.
425 *
426 * On ext2, we can store the error state of the filesystem in the
427 * superblock. That is not possible on ext4, because we may have other
428 * write ordering constraints on the superblock which prevent us from
429 * writing it out straight away; and given that the journal is about to
430 * be aborted, we can't rely on the current, or future, transactions to
431 * write out the superblock safely.
432 *
433 * We'll just use the jbd2_journal_abort() error code to record an error in
434 * the journal instead. On recovery, the journal will complain about
435 * that error until we've noted it down and cleared it.
436 */
437
438static void ext4_handle_error(struct super_block *sb)
439{
440 if (test_opt(sb, WARN_ON_ERROR))
441 WARN_ON_ONCE(1);
442
443 if (sb_rdonly(sb))
444 return;
445
446 if (!test_opt(sb, ERRORS_CONT)) {
447 journal_t *journal = EXT4_SB(sb)->s_journal;
448
449 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
450 if (journal)
451 jbd2_journal_abort(journal, -EIO);
452 }
453 /*
454 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
455 * could panic during 'reboot -f' as the underlying device got already
456 * disabled.
457 */
458 if (test_opt(sb, ERRORS_RO) || system_going_down()) {
459 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
460 /*
461 * Make sure updated value of ->s_mount_flags will be visible
462 * before ->s_flags update
463 */
464 smp_wmb();
465 sb->s_flags |= SB_RDONLY;
466 } else if (test_opt(sb, ERRORS_PANIC)) {
467 if (EXT4_SB(sb)->s_journal &&
468 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
469 return;
470 panic("EXT4-fs (device %s): panic forced after error\n",
471 sb->s_id);
472 }
473}
474
475#define ext4_error_ratelimit(sb) \
476 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
477 "EXT4-fs error")
478
479void __ext4_error(struct super_block *sb, const char *function,
480 unsigned int line, const char *fmt, ...)
481{
482 struct va_format vaf;
483 va_list args;
484
485 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
486 return;
487
488 trace_ext4_error(sb, function, line);
489 if (ext4_error_ratelimit(sb)) {
490 va_start(args, fmt);
491 vaf.fmt = fmt;
492 vaf.va = &args;
493 printk(KERN_CRIT
494 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
495 sb->s_id, function, line, current->comm, &vaf);
496 va_end(args);
497 }
498 save_error_info(sb, function, line);
499 ext4_handle_error(sb);
500}
501
502void __ext4_error_inode(struct inode *inode, const char *function,
503 unsigned int line, ext4_fsblk_t block,
504 const char *fmt, ...)
505{
506 va_list args;
507 struct va_format vaf;
508 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
509
510 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
511 return;
512
513 trace_ext4_error(inode->i_sb, function, line);
514 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
515 es->s_last_error_block = cpu_to_le64(block);
516 if (ext4_error_ratelimit(inode->i_sb)) {
517 va_start(args, fmt);
518 vaf.fmt = fmt;
519 vaf.va = &args;
520 if (block)
521 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
522 "inode #%lu: block %llu: comm %s: %pV\n",
523 inode->i_sb->s_id, function, line, inode->i_ino,
524 block, current->comm, &vaf);
525 else
526 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
527 "inode #%lu: comm %s: %pV\n",
528 inode->i_sb->s_id, function, line, inode->i_ino,
529 current->comm, &vaf);
530 va_end(args);
531 }
532 save_error_info(inode->i_sb, function, line);
533 ext4_handle_error(inode->i_sb);
534}
535
536void __ext4_error_file(struct file *file, const char *function,
537 unsigned int line, ext4_fsblk_t block,
538 const char *fmt, ...)
539{
540 va_list args;
541 struct va_format vaf;
542 struct ext4_super_block *es;
543 struct inode *inode = file_inode(file);
544 char pathname[80], *path;
545
546 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
547 return;
548
549 trace_ext4_error(inode->i_sb, function, line);
550 es = EXT4_SB(inode->i_sb)->s_es;
551 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
552 if (ext4_error_ratelimit(inode->i_sb)) {
553 path = file_path(file, pathname, sizeof(pathname));
554 if (IS_ERR(path))
555 path = "(unknown)";
556 va_start(args, fmt);
557 vaf.fmt = fmt;
558 vaf.va = &args;
559 if (block)
560 printk(KERN_CRIT
561 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
562 "block %llu: comm %s: path %s: %pV\n",
563 inode->i_sb->s_id, function, line, inode->i_ino,
564 block, current->comm, path, &vaf);
565 else
566 printk(KERN_CRIT
567 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
568 "comm %s: path %s: %pV\n",
569 inode->i_sb->s_id, function, line, inode->i_ino,
570 current->comm, path, &vaf);
571 va_end(args);
572 }
573 save_error_info(inode->i_sb, function, line);
574 ext4_handle_error(inode->i_sb);
575}
576
577const char *ext4_decode_error(struct super_block *sb, int errno,
578 char nbuf[16])
579{
580 char *errstr = NULL;
581
582 switch (errno) {
583 case -EFSCORRUPTED:
584 errstr = "Corrupt filesystem";
585 break;
586 case -EFSBADCRC:
587 errstr = "Filesystem failed CRC";
588 break;
589 case -EIO:
590 errstr = "IO failure";
591 break;
592 case -ENOMEM:
593 errstr = "Out of memory";
594 break;
595 case -EROFS:
596 if (!sb || (EXT4_SB(sb)->s_journal &&
597 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
598 errstr = "Journal has aborted";
599 else
600 errstr = "Readonly filesystem";
601 break;
602 default:
603 /* If the caller passed in an extra buffer for unknown
604 * errors, textualise them now. Else we just return
605 * NULL. */
606 if (nbuf) {
607 /* Check for truncated error codes... */
608 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
609 errstr = nbuf;
610 }
611 break;
612 }
613
614 return errstr;
615}
616
617void ext4_set_errno(struct super_block *sb, int err)
618{
619 if (err < 0)
620 err = -err;
621
622 switch (err) {
623 case EIO:
624 err = EXT4_ERR_EIO;
625 break;
626 case ENOMEM:
627 err = EXT4_ERR_ENOMEM;
628 break;
629 case EFSBADCRC:
630 err = EXT4_ERR_EFSBADCRC;
631 break;
632 case EFSCORRUPTED:
633 err = EXT4_ERR_EFSCORRUPTED;
634 break;
635 case ENOSPC:
636 err = EXT4_ERR_ENOSPC;
637 break;
638 case ENOKEY:
639 err = EXT4_ERR_ENOKEY;
640 break;
641 case EROFS:
642 err = EXT4_ERR_EROFS;
643 break;
644 case EFBIG:
645 err = EXT4_ERR_EFBIG;
646 break;
647 case EEXIST:
648 err = EXT4_ERR_EEXIST;
649 break;
650 case ERANGE:
651 err = EXT4_ERR_ERANGE;
652 break;
653 case EOVERFLOW:
654 err = EXT4_ERR_EOVERFLOW;
655 break;
656 case EBUSY:
657 err = EXT4_ERR_EBUSY;
658 break;
659 case ENOTDIR:
660 err = EXT4_ERR_ENOTDIR;
661 break;
662 case ENOTEMPTY:
663 err = EXT4_ERR_ENOTEMPTY;
664 break;
665 case ESHUTDOWN:
666 err = EXT4_ERR_ESHUTDOWN;
667 break;
668 case EFAULT:
669 err = EXT4_ERR_EFAULT;
670 break;
671 default:
672 err = EXT4_ERR_UNKNOWN;
673 }
674 EXT4_SB(sb)->s_es->s_last_error_errcode = err;
675}
676
677/* __ext4_std_error decodes expected errors from journaling functions
678 * automatically and invokes the appropriate error response. */
679
680void __ext4_std_error(struct super_block *sb, const char *function,
681 unsigned int line, int errno)
682{
683 char nbuf[16];
684 const char *errstr;
685
686 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
687 return;
688
689 /* Special case: if the error is EROFS, and we're not already
690 * inside a transaction, then there's really no point in logging
691 * an error. */
692 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
693 return;
694
695 if (ext4_error_ratelimit(sb)) {
696 errstr = ext4_decode_error(sb, errno, nbuf);
697 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
698 sb->s_id, function, line, errstr);
699 }
700
701 ext4_set_errno(sb, -errno);
702 save_error_info(sb, function, line);
703 ext4_handle_error(sb);
704}
705
706/*
707 * ext4_abort is a much stronger failure handler than ext4_error. The
708 * abort function may be used to deal with unrecoverable failures such
709 * as journal IO errors or ENOMEM at a critical moment in log management.
710 *
711 * We unconditionally force the filesystem into an ABORT|READONLY state,
712 * unless the error response on the fs has been set to panic in which
713 * case we take the easy way out and panic immediately.
714 */
715
716void __ext4_abort(struct super_block *sb, const char *function,
717 unsigned int line, const char *fmt, ...)
718{
719 struct va_format vaf;
720 va_list args;
721
722 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
723 return;
724
725 save_error_info(sb, function, line);
726 va_start(args, fmt);
727 vaf.fmt = fmt;
728 vaf.va = &args;
729 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
730 sb->s_id, function, line, &vaf);
731 va_end(args);
732
733 if (sb_rdonly(sb) == 0) {
734 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
735 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
736 /*
737 * Make sure updated value of ->s_mount_flags will be visible
738 * before ->s_flags update
739 */
740 smp_wmb();
741 sb->s_flags |= SB_RDONLY;
742 if (EXT4_SB(sb)->s_journal)
743 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
744 save_error_info(sb, function, line);
745 }
746 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
747 if (EXT4_SB(sb)->s_journal &&
748 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
749 return;
750 panic("EXT4-fs panic from previous error\n");
751 }
752}
753
754void __ext4_msg(struct super_block *sb,
755 const char *prefix, const char *fmt, ...)
756{
757 struct va_format vaf;
758 va_list args;
759
760 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
761 return;
762
763 va_start(args, fmt);
764 vaf.fmt = fmt;
765 vaf.va = &args;
766 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
767 va_end(args);
768}
769
770#define ext4_warning_ratelimit(sb) \
771 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
772 "EXT4-fs warning")
773
774void __ext4_warning(struct super_block *sb, const char *function,
775 unsigned int line, const char *fmt, ...)
776{
777 struct va_format vaf;
778 va_list args;
779
780 if (!ext4_warning_ratelimit(sb))
781 return;
782
783 va_start(args, fmt);
784 vaf.fmt = fmt;
785 vaf.va = &args;
786 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
787 sb->s_id, function, line, &vaf);
788 va_end(args);
789}
790
791void __ext4_warning_inode(const struct inode *inode, const char *function,
792 unsigned int line, const char *fmt, ...)
793{
794 struct va_format vaf;
795 va_list args;
796
797 if (!ext4_warning_ratelimit(inode->i_sb))
798 return;
799
800 va_start(args, fmt);
801 vaf.fmt = fmt;
802 vaf.va = &args;
803 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
804 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
805 function, line, inode->i_ino, current->comm, &vaf);
806 va_end(args);
807}
808
809void __ext4_grp_locked_error(const char *function, unsigned int line,
810 struct super_block *sb, ext4_group_t grp,
811 unsigned long ino, ext4_fsblk_t block,
812 const char *fmt, ...)
813__releases(bitlock)
814__acquires(bitlock)
815{
816 struct va_format vaf;
817 va_list args;
818 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
819
820 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
821 return;
822
823 trace_ext4_error(sb, function, line);
824 es->s_last_error_ino = cpu_to_le32(ino);
825 es->s_last_error_block = cpu_to_le64(block);
826 __save_error_info(sb, function, line);
827
828 if (ext4_error_ratelimit(sb)) {
829 va_start(args, fmt);
830 vaf.fmt = fmt;
831 vaf.va = &args;
832 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
833 sb->s_id, function, line, grp);
834 if (ino)
835 printk(KERN_CONT "inode %lu: ", ino);
836 if (block)
837 printk(KERN_CONT "block %llu:",
838 (unsigned long long) block);
839 printk(KERN_CONT "%pV\n", &vaf);
840 va_end(args);
841 }
842
843 if (test_opt(sb, WARN_ON_ERROR))
844 WARN_ON_ONCE(1);
845
846 if (test_opt(sb, ERRORS_CONT)) {
847 ext4_commit_super(sb, 0);
848 return;
849 }
850
851 ext4_unlock_group(sb, grp);
852 ext4_commit_super(sb, 1);
853 ext4_handle_error(sb);
854 /*
855 * We only get here in the ERRORS_RO case; relocking the group
856 * may be dangerous, but nothing bad will happen since the
857 * filesystem will have already been marked read/only and the
858 * journal has been aborted. We return 1 as a hint to callers
859 * who might what to use the return value from
860 * ext4_grp_locked_error() to distinguish between the
861 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
862 * aggressively from the ext4 function in question, with a
863 * more appropriate error code.
864 */
865 ext4_lock_group(sb, grp);
866 return;
867}
868
869void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
870 ext4_group_t group,
871 unsigned int flags)
872{
873 struct ext4_sb_info *sbi = EXT4_SB(sb);
874 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
875 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
876 int ret;
877
878 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
879 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
880 &grp->bb_state);
881 if (!ret)
882 percpu_counter_sub(&sbi->s_freeclusters_counter,
883 grp->bb_free);
884 }
885
886 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
887 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
888 &grp->bb_state);
889 if (!ret && gdp) {
890 int count;
891
892 count = ext4_free_inodes_count(sb, gdp);
893 percpu_counter_sub(&sbi->s_freeinodes_counter,
894 count);
895 }
896 }
897}
898
899void ext4_update_dynamic_rev(struct super_block *sb)
900{
901 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
902
903 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
904 return;
905
906 ext4_warning(sb,
907 "updating to rev %d because of new feature flag, "
908 "running e2fsck is recommended",
909 EXT4_DYNAMIC_REV);
910
911 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
912 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
913 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
914 /* leave es->s_feature_*compat flags alone */
915 /* es->s_uuid will be set by e2fsck if empty */
916
917 /*
918 * The rest of the superblock fields should be zero, and if not it
919 * means they are likely already in use, so leave them alone. We
920 * can leave it up to e2fsck to clean up any inconsistencies there.
921 */
922}
923
924/*
925 * Open the external journal device
926 */
927static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
928{
929 struct block_device *bdev;
930 char b[BDEVNAME_SIZE];
931
932 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
933 if (IS_ERR(bdev))
934 goto fail;
935 return bdev;
936
937fail:
938 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
939 __bdevname(dev, b), PTR_ERR(bdev));
940 return NULL;
941}
942
943/*
944 * Release the journal device
945 */
946static void ext4_blkdev_put(struct block_device *bdev)
947{
948 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
949}
950
951static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
952{
953 struct block_device *bdev;
954 bdev = sbi->journal_bdev;
955 if (bdev) {
956 ext4_blkdev_put(bdev);
957 sbi->journal_bdev = NULL;
958 }
959}
960
961static inline struct inode *orphan_list_entry(struct list_head *l)
962{
963 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
964}
965
966static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
967{
968 struct list_head *l;
969
970 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
971 le32_to_cpu(sbi->s_es->s_last_orphan));
972
973 printk(KERN_ERR "sb_info orphan list:\n");
974 list_for_each(l, &sbi->s_orphan) {
975 struct inode *inode = orphan_list_entry(l);
976 printk(KERN_ERR " "
977 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
978 inode->i_sb->s_id, inode->i_ino, inode,
979 inode->i_mode, inode->i_nlink,
980 NEXT_ORPHAN(inode));
981 }
982}
983
984#ifdef CONFIG_QUOTA
985static int ext4_quota_off(struct super_block *sb, int type);
986
987static inline void ext4_quota_off_umount(struct super_block *sb)
988{
989 int type;
990
991 /* Use our quota_off function to clear inode flags etc. */
992 for (type = 0; type < EXT4_MAXQUOTAS; type++)
993 ext4_quota_off(sb, type);
994}
995
996/*
997 * This is a helper function which is used in the mount/remount
998 * codepaths (which holds s_umount) to fetch the quota file name.
999 */
1000static inline char *get_qf_name(struct super_block *sb,
1001 struct ext4_sb_info *sbi,
1002 int type)
1003{
1004 return rcu_dereference_protected(sbi->s_qf_names[type],
1005 lockdep_is_held(&sb->s_umount));
1006}
1007#else
1008static inline void ext4_quota_off_umount(struct super_block *sb)
1009{
1010}
1011#endif
1012
1013static void ext4_put_super(struct super_block *sb)
1014{
1015 struct ext4_sb_info *sbi = EXT4_SB(sb);
1016 struct ext4_super_block *es = sbi->s_es;
1017 struct buffer_head **group_desc;
1018 struct flex_groups **flex_groups;
1019 int aborted = 0;
1020 int i, err;
1021
1022 ext4_unregister_li_request(sb);
1023 ext4_quota_off_umount(sb);
1024
1025 destroy_workqueue(sbi->rsv_conversion_wq);
1026
1027 if (sbi->s_journal) {
1028 aborted = is_journal_aborted(sbi->s_journal);
1029 err = jbd2_journal_destroy(sbi->s_journal);
1030 sbi->s_journal = NULL;
1031 if ((err < 0) && !aborted) {
1032 ext4_set_errno(sb, -err);
1033 ext4_abort(sb, "Couldn't clean up the journal");
1034 }
1035 }
1036
1037 ext4_unregister_sysfs(sb);
1038 ext4_es_unregister_shrinker(sbi);
1039 del_timer_sync(&sbi->s_err_report);
1040 ext4_release_system_zone(sb);
1041 ext4_mb_release(sb);
1042 ext4_ext_release(sb);
1043
1044 if (!sb_rdonly(sb) && !aborted) {
1045 ext4_clear_feature_journal_needs_recovery(sb);
1046 es->s_state = cpu_to_le16(sbi->s_mount_state);
1047 }
1048 if (!sb_rdonly(sb))
1049 ext4_commit_super(sb, 1);
1050
1051 rcu_read_lock();
1052 group_desc = rcu_dereference(sbi->s_group_desc);
1053 for (i = 0; i < sbi->s_gdb_count; i++)
1054 brelse(group_desc[i]);
1055 kvfree(group_desc);
1056 flex_groups = rcu_dereference(sbi->s_flex_groups);
1057 if (flex_groups) {
1058 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1059 kvfree(flex_groups[i]);
1060 kvfree(flex_groups);
1061 }
1062 rcu_read_unlock();
1063 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1064 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1065 percpu_counter_destroy(&sbi->s_dirs_counter);
1066 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1067 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1068#ifdef CONFIG_QUOTA
1069 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1070 kfree(get_qf_name(sb, sbi, i));
1071#endif
1072
1073 /* Debugging code just in case the in-memory inode orphan list
1074 * isn't empty. The on-disk one can be non-empty if we've
1075 * detected an error and taken the fs readonly, but the
1076 * in-memory list had better be clean by this point. */
1077 if (!list_empty(&sbi->s_orphan))
1078 dump_orphan_list(sb, sbi);
1079 J_ASSERT(list_empty(&sbi->s_orphan));
1080
1081 sync_blockdev(sb->s_bdev);
1082 invalidate_bdev(sb->s_bdev);
1083 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
1084 /*
1085 * Invalidate the journal device's buffers. We don't want them
1086 * floating about in memory - the physical journal device may
1087 * hotswapped, and it breaks the `ro-after' testing code.
1088 */
1089 sync_blockdev(sbi->journal_bdev);
1090 invalidate_bdev(sbi->journal_bdev);
1091 ext4_blkdev_remove(sbi);
1092 }
1093
1094 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1095 sbi->s_ea_inode_cache = NULL;
1096
1097 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1098 sbi->s_ea_block_cache = NULL;
1099
1100 if (sbi->s_mmp_tsk)
1101 kthread_stop(sbi->s_mmp_tsk);
1102 brelse(sbi->s_sbh);
1103 sb->s_fs_info = NULL;
1104 /*
1105 * Now that we are completely done shutting down the
1106 * superblock, we need to actually destroy the kobject.
1107 */
1108 kobject_put(&sbi->s_kobj);
1109 wait_for_completion(&sbi->s_kobj_unregister);
1110 if (sbi->s_chksum_driver)
1111 crypto_free_shash(sbi->s_chksum_driver);
1112 kfree(sbi->s_blockgroup_lock);
1113 fs_put_dax(sbi->s_daxdev);
1114#ifdef CONFIG_UNICODE
1115 utf8_unload(sbi->s_encoding);
1116#endif
1117 kfree(sbi);
1118}
1119
1120static struct kmem_cache *ext4_inode_cachep;
1121
1122/*
1123 * Called inside transaction, so use GFP_NOFS
1124 */
1125static struct inode *ext4_alloc_inode(struct super_block *sb)
1126{
1127 struct ext4_inode_info *ei;
1128
1129 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1130 if (!ei)
1131 return NULL;
1132
1133 inode_set_iversion(&ei->vfs_inode, 1);
1134 spin_lock_init(&ei->i_raw_lock);
1135 INIT_LIST_HEAD(&ei->i_prealloc_list);
1136 spin_lock_init(&ei->i_prealloc_lock);
1137 ext4_es_init_tree(&ei->i_es_tree);
1138 rwlock_init(&ei->i_es_lock);
1139 INIT_LIST_HEAD(&ei->i_es_list);
1140 ei->i_es_all_nr = 0;
1141 ei->i_es_shk_nr = 0;
1142 ei->i_es_shrink_lblk = 0;
1143 ei->i_reserved_data_blocks = 0;
1144 spin_lock_init(&(ei->i_block_reservation_lock));
1145 ext4_init_pending_tree(&ei->i_pending_tree);
1146#ifdef CONFIG_QUOTA
1147 ei->i_reserved_quota = 0;
1148 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1149#endif
1150 ei->jinode = NULL;
1151 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1152 spin_lock_init(&ei->i_completed_io_lock);
1153 ei->i_sync_tid = 0;
1154 ei->i_datasync_tid = 0;
1155 atomic_set(&ei->i_unwritten, 0);
1156 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1157 return &ei->vfs_inode;
1158}
1159
1160static int ext4_drop_inode(struct inode *inode)
1161{
1162 int drop = generic_drop_inode(inode);
1163
1164 if (!drop)
1165 drop = fscrypt_drop_inode(inode);
1166
1167 trace_ext4_drop_inode(inode, drop);
1168 return drop;
1169}
1170
1171static void ext4_free_in_core_inode(struct inode *inode)
1172{
1173 fscrypt_free_inode(inode);
1174 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1175}
1176
1177static void ext4_destroy_inode(struct inode *inode)
1178{
1179 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1180 ext4_msg(inode->i_sb, KERN_ERR,
1181 "Inode %lu (%p): orphan list check failed!",
1182 inode->i_ino, EXT4_I(inode));
1183 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1184 EXT4_I(inode), sizeof(struct ext4_inode_info),
1185 true);
1186 dump_stack();
1187 }
1188}
1189
1190static void init_once(void *foo)
1191{
1192 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1193
1194 INIT_LIST_HEAD(&ei->i_orphan);
1195 init_rwsem(&ei->xattr_sem);
1196 init_rwsem(&ei->i_data_sem);
1197 init_rwsem(&ei->i_mmap_sem);
1198 inode_init_once(&ei->vfs_inode);
1199}
1200
1201static int __init init_inodecache(void)
1202{
1203 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1204 sizeof(struct ext4_inode_info), 0,
1205 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1206 SLAB_ACCOUNT),
1207 offsetof(struct ext4_inode_info, i_data),
1208 sizeof_field(struct ext4_inode_info, i_data),
1209 init_once);
1210 if (ext4_inode_cachep == NULL)
1211 return -ENOMEM;
1212 return 0;
1213}
1214
1215static void destroy_inodecache(void)
1216{
1217 /*
1218 * Make sure all delayed rcu free inodes are flushed before we
1219 * destroy cache.
1220 */
1221 rcu_barrier();
1222 kmem_cache_destroy(ext4_inode_cachep);
1223}
1224
1225void ext4_clear_inode(struct inode *inode)
1226{
1227 invalidate_inode_buffers(inode);
1228 clear_inode(inode);
1229 ext4_discard_preallocations(inode);
1230 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1231 dquot_drop(inode);
1232 if (EXT4_I(inode)->jinode) {
1233 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1234 EXT4_I(inode)->jinode);
1235 jbd2_free_inode(EXT4_I(inode)->jinode);
1236 EXT4_I(inode)->jinode = NULL;
1237 }
1238 fscrypt_put_encryption_info(inode);
1239 fsverity_cleanup_inode(inode);
1240}
1241
1242static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1243 u64 ino, u32 generation)
1244{
1245 struct inode *inode;
1246
1247 /*
1248 * Currently we don't know the generation for parent directory, so
1249 * a generation of 0 means "accept any"
1250 */
1251 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1252 if (IS_ERR(inode))
1253 return ERR_CAST(inode);
1254 if (generation && inode->i_generation != generation) {
1255 iput(inode);
1256 return ERR_PTR(-ESTALE);
1257 }
1258
1259 return inode;
1260}
1261
1262static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1263 int fh_len, int fh_type)
1264{
1265 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1266 ext4_nfs_get_inode);
1267}
1268
1269static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1270 int fh_len, int fh_type)
1271{
1272 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1273 ext4_nfs_get_inode);
1274}
1275
1276static int ext4_nfs_commit_metadata(struct inode *inode)
1277{
1278 struct writeback_control wbc = {
1279 .sync_mode = WB_SYNC_ALL
1280 };
1281
1282 trace_ext4_nfs_commit_metadata(inode);
1283 return ext4_write_inode(inode, &wbc);
1284}
1285
1286/*
1287 * Try to release metadata pages (indirect blocks, directories) which are
1288 * mapped via the block device. Since these pages could have journal heads
1289 * which would prevent try_to_free_buffers() from freeing them, we must use
1290 * jbd2 layer's try_to_free_buffers() function to release them.
1291 */
1292static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1293 gfp_t wait)
1294{
1295 journal_t *journal = EXT4_SB(sb)->s_journal;
1296
1297 WARN_ON(PageChecked(page));
1298 if (!page_has_buffers(page))
1299 return 0;
1300 if (journal)
1301 return jbd2_journal_try_to_free_buffers(journal, page,
1302 wait & ~__GFP_DIRECT_RECLAIM);
1303 return try_to_free_buffers(page);
1304}
1305
1306#ifdef CONFIG_FS_ENCRYPTION
1307static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1308{
1309 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1310 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1311}
1312
1313static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1314 void *fs_data)
1315{
1316 handle_t *handle = fs_data;
1317 int res, res2, credits, retries = 0;
1318
1319 /*
1320 * Encrypting the root directory is not allowed because e2fsck expects
1321 * lost+found to exist and be unencrypted, and encrypting the root
1322 * directory would imply encrypting the lost+found directory as well as
1323 * the filename "lost+found" itself.
1324 */
1325 if (inode->i_ino == EXT4_ROOT_INO)
1326 return -EPERM;
1327
1328 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1329 return -EINVAL;
1330
1331 res = ext4_convert_inline_data(inode);
1332 if (res)
1333 return res;
1334
1335 /*
1336 * If a journal handle was specified, then the encryption context is
1337 * being set on a new inode via inheritance and is part of a larger
1338 * transaction to create the inode. Otherwise the encryption context is
1339 * being set on an existing inode in its own transaction. Only in the
1340 * latter case should the "retry on ENOSPC" logic be used.
1341 */
1342
1343 if (handle) {
1344 res = ext4_xattr_set_handle(handle, inode,
1345 EXT4_XATTR_INDEX_ENCRYPTION,
1346 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1347 ctx, len, 0);
1348 if (!res) {
1349 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1350 ext4_clear_inode_state(inode,
1351 EXT4_STATE_MAY_INLINE_DATA);
1352 /*
1353 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1354 * S_DAX may be disabled
1355 */
1356 ext4_set_inode_flags(inode);
1357 }
1358 return res;
1359 }
1360
1361 res = dquot_initialize(inode);
1362 if (res)
1363 return res;
1364retry:
1365 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1366 &credits);
1367 if (res)
1368 return res;
1369
1370 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1371 if (IS_ERR(handle))
1372 return PTR_ERR(handle);
1373
1374 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1375 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1376 ctx, len, 0);
1377 if (!res) {
1378 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1379 /*
1380 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1381 * S_DAX may be disabled
1382 */
1383 ext4_set_inode_flags(inode);
1384 res = ext4_mark_inode_dirty(handle, inode);
1385 if (res)
1386 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1387 }
1388 res2 = ext4_journal_stop(handle);
1389
1390 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1391 goto retry;
1392 if (!res)
1393 res = res2;
1394 return res;
1395}
1396
1397static bool ext4_dummy_context(struct inode *inode)
1398{
1399 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1400}
1401
1402static bool ext4_has_stable_inodes(struct super_block *sb)
1403{
1404 return ext4_has_feature_stable_inodes(sb);
1405}
1406
1407static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1408 int *ino_bits_ret, int *lblk_bits_ret)
1409{
1410 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1411 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1412}
1413
1414static const struct fscrypt_operations ext4_cryptops = {
1415 .key_prefix = "ext4:",
1416 .get_context = ext4_get_context,
1417 .set_context = ext4_set_context,
1418 .dummy_context = ext4_dummy_context,
1419 .empty_dir = ext4_empty_dir,
1420 .max_namelen = EXT4_NAME_LEN,
1421 .has_stable_inodes = ext4_has_stable_inodes,
1422 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1423};
1424#endif
1425
1426#ifdef CONFIG_QUOTA
1427static const char * const quotatypes[] = INITQFNAMES;
1428#define QTYPE2NAME(t) (quotatypes[t])
1429
1430static int ext4_write_dquot(struct dquot *dquot);
1431static int ext4_acquire_dquot(struct dquot *dquot);
1432static int ext4_release_dquot(struct dquot *dquot);
1433static int ext4_mark_dquot_dirty(struct dquot *dquot);
1434static int ext4_write_info(struct super_block *sb, int type);
1435static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1436 const struct path *path);
1437static int ext4_quota_on_mount(struct super_block *sb, int type);
1438static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1439 size_t len, loff_t off);
1440static ssize_t ext4_quota_write(struct super_block *sb, int type,
1441 const char *data, size_t len, loff_t off);
1442static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1443 unsigned int flags);
1444static int ext4_enable_quotas(struct super_block *sb);
1445
1446static struct dquot **ext4_get_dquots(struct inode *inode)
1447{
1448 return EXT4_I(inode)->i_dquot;
1449}
1450
1451static const struct dquot_operations ext4_quota_operations = {
1452 .get_reserved_space = ext4_get_reserved_space,
1453 .write_dquot = ext4_write_dquot,
1454 .acquire_dquot = ext4_acquire_dquot,
1455 .release_dquot = ext4_release_dquot,
1456 .mark_dirty = ext4_mark_dquot_dirty,
1457 .write_info = ext4_write_info,
1458 .alloc_dquot = dquot_alloc,
1459 .destroy_dquot = dquot_destroy,
1460 .get_projid = ext4_get_projid,
1461 .get_inode_usage = ext4_get_inode_usage,
1462 .get_next_id = dquot_get_next_id,
1463};
1464
1465static const struct quotactl_ops ext4_qctl_operations = {
1466 .quota_on = ext4_quota_on,
1467 .quota_off = ext4_quota_off,
1468 .quota_sync = dquot_quota_sync,
1469 .get_state = dquot_get_state,
1470 .set_info = dquot_set_dqinfo,
1471 .get_dqblk = dquot_get_dqblk,
1472 .set_dqblk = dquot_set_dqblk,
1473 .get_nextdqblk = dquot_get_next_dqblk,
1474};
1475#endif
1476
1477static const struct super_operations ext4_sops = {
1478 .alloc_inode = ext4_alloc_inode,
1479 .free_inode = ext4_free_in_core_inode,
1480 .destroy_inode = ext4_destroy_inode,
1481 .write_inode = ext4_write_inode,
1482 .dirty_inode = ext4_dirty_inode,
1483 .drop_inode = ext4_drop_inode,
1484 .evict_inode = ext4_evict_inode,
1485 .put_super = ext4_put_super,
1486 .sync_fs = ext4_sync_fs,
1487 .freeze_fs = ext4_freeze,
1488 .unfreeze_fs = ext4_unfreeze,
1489 .statfs = ext4_statfs,
1490 .remount_fs = ext4_remount,
1491 .show_options = ext4_show_options,
1492#ifdef CONFIG_QUOTA
1493 .quota_read = ext4_quota_read,
1494 .quota_write = ext4_quota_write,
1495 .get_dquots = ext4_get_dquots,
1496#endif
1497 .bdev_try_to_free_page = bdev_try_to_free_page,
1498};
1499
1500static const struct export_operations ext4_export_ops = {
1501 .fh_to_dentry = ext4_fh_to_dentry,
1502 .fh_to_parent = ext4_fh_to_parent,
1503 .get_parent = ext4_get_parent,
1504 .commit_metadata = ext4_nfs_commit_metadata,
1505};
1506
1507enum {
1508 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1509 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1510 Opt_nouid32, Opt_debug, Opt_removed,
1511 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1512 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1513 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1514 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1515 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1516 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1517 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1518 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1519 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1520 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1521 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1522 Opt_nowarn_on_error, Opt_mblk_io_submit,
1523 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1524 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1525 Opt_inode_readahead_blks, Opt_journal_ioprio,
1526 Opt_dioread_nolock, Opt_dioread_lock,
1527 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1528 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1529};
1530
1531static const match_table_t tokens = {
1532 {Opt_bsd_df, "bsddf"},
1533 {Opt_minix_df, "minixdf"},
1534 {Opt_grpid, "grpid"},
1535 {Opt_grpid, "bsdgroups"},
1536 {Opt_nogrpid, "nogrpid"},
1537 {Opt_nogrpid, "sysvgroups"},
1538 {Opt_resgid, "resgid=%u"},
1539 {Opt_resuid, "resuid=%u"},
1540 {Opt_sb, "sb=%u"},
1541 {Opt_err_cont, "errors=continue"},
1542 {Opt_err_panic, "errors=panic"},
1543 {Opt_err_ro, "errors=remount-ro"},
1544 {Opt_nouid32, "nouid32"},
1545 {Opt_debug, "debug"},
1546 {Opt_removed, "oldalloc"},
1547 {Opt_removed, "orlov"},
1548 {Opt_user_xattr, "user_xattr"},
1549 {Opt_nouser_xattr, "nouser_xattr"},
1550 {Opt_acl, "acl"},
1551 {Opt_noacl, "noacl"},
1552 {Opt_noload, "norecovery"},
1553 {Opt_noload, "noload"},
1554 {Opt_removed, "nobh"},
1555 {Opt_removed, "bh"},
1556 {Opt_commit, "commit=%u"},
1557 {Opt_min_batch_time, "min_batch_time=%u"},
1558 {Opt_max_batch_time, "max_batch_time=%u"},
1559 {Opt_journal_dev, "journal_dev=%u"},
1560 {Opt_journal_path, "journal_path=%s"},
1561 {Opt_journal_checksum, "journal_checksum"},
1562 {Opt_nojournal_checksum, "nojournal_checksum"},
1563 {Opt_journal_async_commit, "journal_async_commit"},
1564 {Opt_abort, "abort"},
1565 {Opt_data_journal, "data=journal"},
1566 {Opt_data_ordered, "data=ordered"},
1567 {Opt_data_writeback, "data=writeback"},
1568 {Opt_data_err_abort, "data_err=abort"},
1569 {Opt_data_err_ignore, "data_err=ignore"},
1570 {Opt_offusrjquota, "usrjquota="},
1571 {Opt_usrjquota, "usrjquota=%s"},
1572 {Opt_offgrpjquota, "grpjquota="},
1573 {Opt_grpjquota, "grpjquota=%s"},
1574 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1575 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1576 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1577 {Opt_grpquota, "grpquota"},
1578 {Opt_noquota, "noquota"},
1579 {Opt_quota, "quota"},
1580 {Opt_usrquota, "usrquota"},
1581 {Opt_prjquota, "prjquota"},
1582 {Opt_barrier, "barrier=%u"},
1583 {Opt_barrier, "barrier"},
1584 {Opt_nobarrier, "nobarrier"},
1585 {Opt_i_version, "i_version"},
1586 {Opt_dax, "dax"},
1587 {Opt_stripe, "stripe=%u"},
1588 {Opt_delalloc, "delalloc"},
1589 {Opt_warn_on_error, "warn_on_error"},
1590 {Opt_nowarn_on_error, "nowarn_on_error"},
1591 {Opt_lazytime, "lazytime"},
1592 {Opt_nolazytime, "nolazytime"},
1593 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1594 {Opt_nodelalloc, "nodelalloc"},
1595 {Opt_removed, "mblk_io_submit"},
1596 {Opt_removed, "nomblk_io_submit"},
1597 {Opt_block_validity, "block_validity"},
1598 {Opt_noblock_validity, "noblock_validity"},
1599 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1600 {Opt_journal_ioprio, "journal_ioprio=%u"},
1601 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1602 {Opt_auto_da_alloc, "auto_da_alloc"},
1603 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1604 {Opt_dioread_nolock, "dioread_nolock"},
1605 {Opt_dioread_lock, "nodioread_nolock"},
1606 {Opt_dioread_lock, "dioread_lock"},
1607 {Opt_discard, "discard"},
1608 {Opt_nodiscard, "nodiscard"},
1609 {Opt_init_itable, "init_itable=%u"},
1610 {Opt_init_itable, "init_itable"},
1611 {Opt_noinit_itable, "noinit_itable"},
1612 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1613 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1614 {Opt_nombcache, "nombcache"},
1615 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1616 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1617 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1618 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1619 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1620 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1621 {Opt_err, NULL},
1622};
1623
1624static ext4_fsblk_t get_sb_block(void **data)
1625{
1626 ext4_fsblk_t sb_block;
1627 char *options = (char *) *data;
1628
1629 if (!options || strncmp(options, "sb=", 3) != 0)
1630 return 1; /* Default location */
1631
1632 options += 3;
1633 /* TODO: use simple_strtoll with >32bit ext4 */
1634 sb_block = simple_strtoul(options, &options, 0);
1635 if (*options && *options != ',') {
1636 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1637 (char *) *data);
1638 return 1;
1639 }
1640 if (*options == ',')
1641 options++;
1642 *data = (void *) options;
1643
1644 return sb_block;
1645}
1646
1647#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1648static const char deprecated_msg[] =
1649 "Mount option \"%s\" will be removed by %s\n"
1650 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1651
1652#ifdef CONFIG_QUOTA
1653static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1654{
1655 struct ext4_sb_info *sbi = EXT4_SB(sb);
1656 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1657 int ret = -1;
1658
1659 if (sb_any_quota_loaded(sb) && !old_qname) {
1660 ext4_msg(sb, KERN_ERR,
1661 "Cannot change journaled "
1662 "quota options when quota turned on");
1663 return -1;
1664 }
1665 if (ext4_has_feature_quota(sb)) {
1666 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1667 "ignored when QUOTA feature is enabled");
1668 return 1;
1669 }
1670 qname = match_strdup(args);
1671 if (!qname) {
1672 ext4_msg(sb, KERN_ERR,
1673 "Not enough memory for storing quotafile name");
1674 return -1;
1675 }
1676 if (old_qname) {
1677 if (strcmp(old_qname, qname) == 0)
1678 ret = 1;
1679 else
1680 ext4_msg(sb, KERN_ERR,
1681 "%s quota file already specified",
1682 QTYPE2NAME(qtype));
1683 goto errout;
1684 }
1685 if (strchr(qname, '/')) {
1686 ext4_msg(sb, KERN_ERR,
1687 "quotafile must be on filesystem root");
1688 goto errout;
1689 }
1690 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1691 set_opt(sb, QUOTA);
1692 return 1;
1693errout:
1694 kfree(qname);
1695 return ret;
1696}
1697
1698static int clear_qf_name(struct super_block *sb, int qtype)
1699{
1700
1701 struct ext4_sb_info *sbi = EXT4_SB(sb);
1702 char *old_qname = get_qf_name(sb, sbi, qtype);
1703
1704 if (sb_any_quota_loaded(sb) && old_qname) {
1705 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1706 " when quota turned on");
1707 return -1;
1708 }
1709 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1710 synchronize_rcu();
1711 kfree(old_qname);
1712 return 1;
1713}
1714#endif
1715
1716#define MOPT_SET 0x0001
1717#define MOPT_CLEAR 0x0002
1718#define MOPT_NOSUPPORT 0x0004
1719#define MOPT_EXPLICIT 0x0008
1720#define MOPT_CLEAR_ERR 0x0010
1721#define MOPT_GTE0 0x0020
1722#ifdef CONFIG_QUOTA
1723#define MOPT_Q 0
1724#define MOPT_QFMT 0x0040
1725#else
1726#define MOPT_Q MOPT_NOSUPPORT
1727#define MOPT_QFMT MOPT_NOSUPPORT
1728#endif
1729#define MOPT_DATAJ 0x0080
1730#define MOPT_NO_EXT2 0x0100
1731#define MOPT_NO_EXT3 0x0200
1732#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1733#define MOPT_STRING 0x0400
1734
1735static const struct mount_opts {
1736 int token;
1737 int mount_opt;
1738 int flags;
1739} ext4_mount_opts[] = {
1740 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1741 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1742 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1743 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1744 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1745 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1746 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1747 MOPT_EXT4_ONLY | MOPT_SET},
1748 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1749 MOPT_EXT4_ONLY | MOPT_CLEAR},
1750 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1751 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1752 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1753 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1754 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1755 MOPT_EXT4_ONLY | MOPT_CLEAR},
1756 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1757 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1758 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1759 MOPT_EXT4_ONLY | MOPT_CLEAR},
1760 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1761 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1762 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1763 EXT4_MOUNT_JOURNAL_CHECKSUM),
1764 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1765 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1766 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1767 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1768 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1769 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1770 MOPT_NO_EXT2},
1771 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1772 MOPT_NO_EXT2},
1773 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1774 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1775 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1776 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1777 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1778 {Opt_commit, 0, MOPT_GTE0},
1779 {Opt_max_batch_time, 0, MOPT_GTE0},
1780 {Opt_min_batch_time, 0, MOPT_GTE0},
1781 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1782 {Opt_init_itable, 0, MOPT_GTE0},
1783 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1784 {Opt_stripe, 0, MOPT_GTE0},
1785 {Opt_resuid, 0, MOPT_GTE0},
1786 {Opt_resgid, 0, MOPT_GTE0},
1787 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1788 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1789 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1790 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1791 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1792 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1793 MOPT_NO_EXT2 | MOPT_DATAJ},
1794 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1795 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1796#ifdef CONFIG_EXT4_FS_POSIX_ACL
1797 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1798 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1799#else
1800 {Opt_acl, 0, MOPT_NOSUPPORT},
1801 {Opt_noacl, 0, MOPT_NOSUPPORT},
1802#endif
1803 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1804 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1805 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1806 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1807 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1808 MOPT_SET | MOPT_Q},
1809 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1810 MOPT_SET | MOPT_Q},
1811 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1812 MOPT_SET | MOPT_Q},
1813 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1814 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1815 MOPT_CLEAR | MOPT_Q},
1816 {Opt_usrjquota, 0, MOPT_Q},
1817 {Opt_grpjquota, 0, MOPT_Q},
1818 {Opt_offusrjquota, 0, MOPT_Q},
1819 {Opt_offgrpjquota, 0, MOPT_Q},
1820 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1821 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1822 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1823 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1824 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1825 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1826 {Opt_err, 0, 0}
1827};
1828
1829#ifdef CONFIG_UNICODE
1830static const struct ext4_sb_encodings {
1831 __u16 magic;
1832 char *name;
1833 char *version;
1834} ext4_sb_encoding_map[] = {
1835 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1836};
1837
1838static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1839 const struct ext4_sb_encodings **encoding,
1840 __u16 *flags)
1841{
1842 __u16 magic = le16_to_cpu(es->s_encoding);
1843 int i;
1844
1845 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1846 if (magic == ext4_sb_encoding_map[i].magic)
1847 break;
1848
1849 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
1850 return -EINVAL;
1851
1852 *encoding = &ext4_sb_encoding_map[i];
1853 *flags = le16_to_cpu(es->s_encoding_flags);
1854
1855 return 0;
1856}
1857#endif
1858
1859static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1860 substring_t *args, unsigned long *journal_devnum,
1861 unsigned int *journal_ioprio, int is_remount)
1862{
1863 struct ext4_sb_info *sbi = EXT4_SB(sb);
1864 const struct mount_opts *m;
1865 kuid_t uid;
1866 kgid_t gid;
1867 int arg = 0;
1868
1869#ifdef CONFIG_QUOTA
1870 if (token == Opt_usrjquota)
1871 return set_qf_name(sb, USRQUOTA, &args[0]);
1872 else if (token == Opt_grpjquota)
1873 return set_qf_name(sb, GRPQUOTA, &args[0]);
1874 else if (token == Opt_offusrjquota)
1875 return clear_qf_name(sb, USRQUOTA);
1876 else if (token == Opt_offgrpjquota)
1877 return clear_qf_name(sb, GRPQUOTA);
1878#endif
1879 switch (token) {
1880 case Opt_noacl:
1881 case Opt_nouser_xattr:
1882 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1883 break;
1884 case Opt_sb:
1885 return 1; /* handled by get_sb_block() */
1886 case Opt_removed:
1887 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1888 return 1;
1889 case Opt_abort:
1890 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1891 return 1;
1892 case Opt_i_version:
1893 sb->s_flags |= SB_I_VERSION;
1894 return 1;
1895 case Opt_lazytime:
1896 sb->s_flags |= SB_LAZYTIME;
1897 return 1;
1898 case Opt_nolazytime:
1899 sb->s_flags &= ~SB_LAZYTIME;
1900 return 1;
1901 }
1902
1903 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1904 if (token == m->token)
1905 break;
1906
1907 if (m->token == Opt_err) {
1908 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1909 "or missing value", opt);
1910 return -1;
1911 }
1912
1913 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1914 ext4_msg(sb, KERN_ERR,
1915 "Mount option \"%s\" incompatible with ext2", opt);
1916 return -1;
1917 }
1918 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1919 ext4_msg(sb, KERN_ERR,
1920 "Mount option \"%s\" incompatible with ext3", opt);
1921 return -1;
1922 }
1923
1924 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1925 return -1;
1926 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1927 return -1;
1928 if (m->flags & MOPT_EXPLICIT) {
1929 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1930 set_opt2(sb, EXPLICIT_DELALLOC);
1931 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1932 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1933 } else
1934 return -1;
1935 }
1936 if (m->flags & MOPT_CLEAR_ERR)
1937 clear_opt(sb, ERRORS_MASK);
1938 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1939 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1940 "options when quota turned on");
1941 return -1;
1942 }
1943
1944 if (m->flags & MOPT_NOSUPPORT) {
1945 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1946 } else if (token == Opt_commit) {
1947 if (arg == 0)
1948 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1949 else if (arg > INT_MAX / HZ) {
1950 ext4_msg(sb, KERN_ERR,
1951 "Invalid commit interval %d, "
1952 "must be smaller than %d",
1953 arg, INT_MAX / HZ);
1954 return -1;
1955 }
1956 sbi->s_commit_interval = HZ * arg;
1957 } else if (token == Opt_debug_want_extra_isize) {
1958 if ((arg & 1) ||
1959 (arg < 4) ||
1960 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
1961 ext4_msg(sb, KERN_ERR,
1962 "Invalid want_extra_isize %d", arg);
1963 return -1;
1964 }
1965 sbi->s_want_extra_isize = arg;
1966 } else if (token == Opt_max_batch_time) {
1967 sbi->s_max_batch_time = arg;
1968 } else if (token == Opt_min_batch_time) {
1969 sbi->s_min_batch_time = arg;
1970 } else if (token == Opt_inode_readahead_blks) {
1971 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1972 ext4_msg(sb, KERN_ERR,
1973 "EXT4-fs: inode_readahead_blks must be "
1974 "0 or a power of 2 smaller than 2^31");
1975 return -1;
1976 }
1977 sbi->s_inode_readahead_blks = arg;
1978 } else if (token == Opt_init_itable) {
1979 set_opt(sb, INIT_INODE_TABLE);
1980 if (!args->from)
1981 arg = EXT4_DEF_LI_WAIT_MULT;
1982 sbi->s_li_wait_mult = arg;
1983 } else if (token == Opt_max_dir_size_kb) {
1984 sbi->s_max_dir_size_kb = arg;
1985 } else if (token == Opt_stripe) {
1986 sbi->s_stripe = arg;
1987 } else if (token == Opt_resuid) {
1988 uid = make_kuid(current_user_ns(), arg);
1989 if (!uid_valid(uid)) {
1990 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1991 return -1;
1992 }
1993 sbi->s_resuid = uid;
1994 } else if (token == Opt_resgid) {
1995 gid = make_kgid(current_user_ns(), arg);
1996 if (!gid_valid(gid)) {
1997 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1998 return -1;
1999 }
2000 sbi->s_resgid = gid;
2001 } else if (token == Opt_journal_dev) {
2002 if (is_remount) {
2003 ext4_msg(sb, KERN_ERR,
2004 "Cannot specify journal on remount");
2005 return -1;
2006 }
2007 *journal_devnum = arg;
2008 } else if (token == Opt_journal_path) {
2009 char *journal_path;
2010 struct inode *journal_inode;
2011 struct path path;
2012 int error;
2013
2014 if (is_remount) {
2015 ext4_msg(sb, KERN_ERR,
2016 "Cannot specify journal on remount");
2017 return -1;
2018 }
2019 journal_path = match_strdup(&args[0]);
2020 if (!journal_path) {
2021 ext4_msg(sb, KERN_ERR, "error: could not dup "
2022 "journal device string");
2023 return -1;
2024 }
2025
2026 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2027 if (error) {
2028 ext4_msg(sb, KERN_ERR, "error: could not find "
2029 "journal device path: error %d", error);
2030 kfree(journal_path);
2031 return -1;
2032 }
2033
2034 journal_inode = d_inode(path.dentry);
2035 if (!S_ISBLK(journal_inode->i_mode)) {
2036 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2037 "is not a block device", journal_path);
2038 path_put(&path);
2039 kfree(journal_path);
2040 return -1;
2041 }
2042
2043 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2044 path_put(&path);
2045 kfree(journal_path);
2046 } else if (token == Opt_journal_ioprio) {
2047 if (arg > 7) {
2048 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2049 " (must be 0-7)");
2050 return -1;
2051 }
2052 *journal_ioprio =
2053 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2054 } else if (token == Opt_test_dummy_encryption) {
2055#ifdef CONFIG_FS_ENCRYPTION
2056 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
2057 ext4_msg(sb, KERN_WARNING,
2058 "Test dummy encryption mode enabled");
2059#else
2060 ext4_msg(sb, KERN_WARNING,
2061 "Test dummy encryption mount option ignored");
2062#endif
2063 } else if (m->flags & MOPT_DATAJ) {
2064 if (is_remount) {
2065 if (!sbi->s_journal)
2066 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2067 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2068 ext4_msg(sb, KERN_ERR,
2069 "Cannot change data mode on remount");
2070 return -1;
2071 }
2072 } else {
2073 clear_opt(sb, DATA_FLAGS);
2074 sbi->s_mount_opt |= m->mount_opt;
2075 }
2076#ifdef CONFIG_QUOTA
2077 } else if (m->flags & MOPT_QFMT) {
2078 if (sb_any_quota_loaded(sb) &&
2079 sbi->s_jquota_fmt != m->mount_opt) {
2080 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2081 "quota options when quota turned on");
2082 return -1;
2083 }
2084 if (ext4_has_feature_quota(sb)) {
2085 ext4_msg(sb, KERN_INFO,
2086 "Quota format mount options ignored "
2087 "when QUOTA feature is enabled");
2088 return 1;
2089 }
2090 sbi->s_jquota_fmt = m->mount_opt;
2091#endif
2092 } else if (token == Opt_dax) {
2093#ifdef CONFIG_FS_DAX
2094 ext4_msg(sb, KERN_WARNING,
2095 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2096 sbi->s_mount_opt |= m->mount_opt;
2097#else
2098 ext4_msg(sb, KERN_INFO, "dax option not supported");
2099 return -1;
2100#endif
2101 } else if (token == Opt_data_err_abort) {
2102 sbi->s_mount_opt |= m->mount_opt;
2103 } else if (token == Opt_data_err_ignore) {
2104 sbi->s_mount_opt &= ~m->mount_opt;
2105 } else {
2106 if (!args->from)
2107 arg = 1;
2108 if (m->flags & MOPT_CLEAR)
2109 arg = !arg;
2110 else if (unlikely(!(m->flags & MOPT_SET))) {
2111 ext4_msg(sb, KERN_WARNING,
2112 "buggy handling of option %s", opt);
2113 WARN_ON(1);
2114 return -1;
2115 }
2116 if (arg != 0)
2117 sbi->s_mount_opt |= m->mount_opt;
2118 else
2119 sbi->s_mount_opt &= ~m->mount_opt;
2120 }
2121 return 1;
2122}
2123
2124static int parse_options(char *options, struct super_block *sb,
2125 unsigned long *journal_devnum,
2126 unsigned int *journal_ioprio,
2127 int is_remount)
2128{
2129 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2130 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2131 substring_t args[MAX_OPT_ARGS];
2132 int token;
2133
2134 if (!options)
2135 return 1;
2136
2137 while ((p = strsep(&options, ",")) != NULL) {
2138 if (!*p)
2139 continue;
2140 /*
2141 * Initialize args struct so we know whether arg was
2142 * found; some options take optional arguments.
2143 */
2144 args[0].to = args[0].from = NULL;
2145 token = match_token(p, tokens, args);
2146 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2147 journal_ioprio, is_remount) < 0)
2148 return 0;
2149 }
2150#ifdef CONFIG_QUOTA
2151 /*
2152 * We do the test below only for project quotas. 'usrquota' and
2153 * 'grpquota' mount options are allowed even without quota feature
2154 * to support legacy quotas in quota files.
2155 */
2156 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2157 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2158 "Cannot enable project quota enforcement.");
2159 return 0;
2160 }
2161 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2162 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2163 if (usr_qf_name || grp_qf_name) {
2164 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2165 clear_opt(sb, USRQUOTA);
2166
2167 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2168 clear_opt(sb, GRPQUOTA);
2169
2170 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2171 ext4_msg(sb, KERN_ERR, "old and new quota "
2172 "format mixing");
2173 return 0;
2174 }
2175
2176 if (!sbi->s_jquota_fmt) {
2177 ext4_msg(sb, KERN_ERR, "journaled quota format "
2178 "not specified");
2179 return 0;
2180 }
2181 }
2182#endif
2183 return 1;
2184}
2185
2186static inline void ext4_show_quota_options(struct seq_file *seq,
2187 struct super_block *sb)
2188{
2189#if defined(CONFIG_QUOTA)
2190 struct ext4_sb_info *sbi = EXT4_SB(sb);
2191 char *usr_qf_name, *grp_qf_name;
2192
2193 if (sbi->s_jquota_fmt) {
2194 char *fmtname = "";
2195
2196 switch (sbi->s_jquota_fmt) {
2197 case QFMT_VFS_OLD:
2198 fmtname = "vfsold";
2199 break;
2200 case QFMT_VFS_V0:
2201 fmtname = "vfsv0";
2202 break;
2203 case QFMT_VFS_V1:
2204 fmtname = "vfsv1";
2205 break;
2206 }
2207 seq_printf(seq, ",jqfmt=%s", fmtname);
2208 }
2209
2210 rcu_read_lock();
2211 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2212 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2213 if (usr_qf_name)
2214 seq_show_option(seq, "usrjquota", usr_qf_name);
2215 if (grp_qf_name)
2216 seq_show_option(seq, "grpjquota", grp_qf_name);
2217 rcu_read_unlock();
2218#endif
2219}
2220
2221static const char *token2str(int token)
2222{
2223 const struct match_token *t;
2224
2225 for (t = tokens; t->token != Opt_err; t++)
2226 if (t->token == token && !strchr(t->pattern, '='))
2227 break;
2228 return t->pattern;
2229}
2230
2231/*
2232 * Show an option if
2233 * - it's set to a non-default value OR
2234 * - if the per-sb default is different from the global default
2235 */
2236static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2237 int nodefs)
2238{
2239 struct ext4_sb_info *sbi = EXT4_SB(sb);
2240 struct ext4_super_block *es = sbi->s_es;
2241 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2242 const struct mount_opts *m;
2243 char sep = nodefs ? '\n' : ',';
2244
2245#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2246#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2247
2248 if (sbi->s_sb_block != 1)
2249 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2250
2251 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2252 int want_set = m->flags & MOPT_SET;
2253 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2254 (m->flags & MOPT_CLEAR_ERR))
2255 continue;
2256 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2257 continue; /* skip if same as the default */
2258 if ((want_set &&
2259 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2260 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2261 continue; /* select Opt_noFoo vs Opt_Foo */
2262 SEQ_OPTS_PRINT("%s", token2str(m->token));
2263 }
2264
2265 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2266 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2267 SEQ_OPTS_PRINT("resuid=%u",
2268 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2269 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2270 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2271 SEQ_OPTS_PRINT("resgid=%u",
2272 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2273 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2274 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2275 SEQ_OPTS_PUTS("errors=remount-ro");
2276 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2277 SEQ_OPTS_PUTS("errors=continue");
2278 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2279 SEQ_OPTS_PUTS("errors=panic");
2280 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2281 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2282 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2283 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2284 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2285 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2286 if (sb->s_flags & SB_I_VERSION)
2287 SEQ_OPTS_PUTS("i_version");
2288 if (nodefs || sbi->s_stripe)
2289 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2290 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2291 (sbi->s_mount_opt ^ def_mount_opt)) {
2292 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2293 SEQ_OPTS_PUTS("data=journal");
2294 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2295 SEQ_OPTS_PUTS("data=ordered");
2296 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2297 SEQ_OPTS_PUTS("data=writeback");
2298 }
2299 if (nodefs ||
2300 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2301 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2302 sbi->s_inode_readahead_blks);
2303
2304 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2305 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2306 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2307 if (nodefs || sbi->s_max_dir_size_kb)
2308 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2309 if (test_opt(sb, DATA_ERR_ABORT))
2310 SEQ_OPTS_PUTS("data_err=abort");
2311 if (DUMMY_ENCRYPTION_ENABLED(sbi))
2312 SEQ_OPTS_PUTS("test_dummy_encryption");
2313
2314 ext4_show_quota_options(seq, sb);
2315 return 0;
2316}
2317
2318static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2319{
2320 return _ext4_show_options(seq, root->d_sb, 0);
2321}
2322
2323int ext4_seq_options_show(struct seq_file *seq, void *offset)
2324{
2325 struct super_block *sb = seq->private;
2326 int rc;
2327
2328 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2329 rc = _ext4_show_options(seq, sb, 1);
2330 seq_puts(seq, "\n");
2331 return rc;
2332}
2333
2334static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2335 int read_only)
2336{
2337 struct ext4_sb_info *sbi = EXT4_SB(sb);
2338 int err = 0;
2339
2340 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2341 ext4_msg(sb, KERN_ERR, "revision level too high, "
2342 "forcing read-only mode");
2343 err = -EROFS;
2344 }
2345 if (read_only)
2346 goto done;
2347 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2348 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2349 "running e2fsck is recommended");
2350 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2351 ext4_msg(sb, KERN_WARNING,
2352 "warning: mounting fs with errors, "
2353 "running e2fsck is recommended");
2354 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2355 le16_to_cpu(es->s_mnt_count) >=
2356 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2357 ext4_msg(sb, KERN_WARNING,
2358 "warning: maximal mount count reached, "
2359 "running e2fsck is recommended");
2360 else if (le32_to_cpu(es->s_checkinterval) &&
2361 (ext4_get_tstamp(es, s_lastcheck) +
2362 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2363 ext4_msg(sb, KERN_WARNING,
2364 "warning: checktime reached, "
2365 "running e2fsck is recommended");
2366 if (!sbi->s_journal)
2367 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2368 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2369 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2370 le16_add_cpu(&es->s_mnt_count, 1);
2371 ext4_update_tstamp(es, s_mtime);
2372 if (sbi->s_journal)
2373 ext4_set_feature_journal_needs_recovery(sb);
2374
2375 err = ext4_commit_super(sb, 1);
2376done:
2377 if (test_opt(sb, DEBUG))
2378 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2379 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2380 sb->s_blocksize,
2381 sbi->s_groups_count,
2382 EXT4_BLOCKS_PER_GROUP(sb),
2383 EXT4_INODES_PER_GROUP(sb),
2384 sbi->s_mount_opt, sbi->s_mount_opt2);
2385
2386 cleancache_init_fs(sb);
2387 return err;
2388}
2389
2390int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2391{
2392 struct ext4_sb_info *sbi = EXT4_SB(sb);
2393 struct flex_groups **old_groups, **new_groups;
2394 int size, i, j;
2395
2396 if (!sbi->s_log_groups_per_flex)
2397 return 0;
2398
2399 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2400 if (size <= sbi->s_flex_groups_allocated)
2401 return 0;
2402
2403 new_groups = kvzalloc(roundup_pow_of_two(size *
2404 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2405 if (!new_groups) {
2406 ext4_msg(sb, KERN_ERR,
2407 "not enough memory for %d flex group pointers", size);
2408 return -ENOMEM;
2409 }
2410 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2411 new_groups[i] = kvzalloc(roundup_pow_of_two(
2412 sizeof(struct flex_groups)),
2413 GFP_KERNEL);
2414 if (!new_groups[i]) {
2415 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2416 kvfree(new_groups[j]);
2417 kvfree(new_groups);
2418 ext4_msg(sb, KERN_ERR,
2419 "not enough memory for %d flex groups", size);
2420 return -ENOMEM;
2421 }
2422 }
2423 rcu_read_lock();
2424 old_groups = rcu_dereference(sbi->s_flex_groups);
2425 if (old_groups)
2426 memcpy(new_groups, old_groups,
2427 (sbi->s_flex_groups_allocated *
2428 sizeof(struct flex_groups *)));
2429 rcu_read_unlock();
2430 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2431 sbi->s_flex_groups_allocated = size;
2432 if (old_groups)
2433 ext4_kvfree_array_rcu(old_groups);
2434 return 0;
2435}
2436
2437static int ext4_fill_flex_info(struct super_block *sb)
2438{
2439 struct ext4_sb_info *sbi = EXT4_SB(sb);
2440 struct ext4_group_desc *gdp = NULL;
2441 struct flex_groups *fg;
2442 ext4_group_t flex_group;
2443 int i, err;
2444
2445 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2446 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2447 sbi->s_log_groups_per_flex = 0;
2448 return 1;
2449 }
2450
2451 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2452 if (err)
2453 goto failed;
2454
2455 for (i = 0; i < sbi->s_groups_count; i++) {
2456 gdp = ext4_get_group_desc(sb, i, NULL);
2457
2458 flex_group = ext4_flex_group(sbi, i);
2459 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2460 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2461 atomic64_add(ext4_free_group_clusters(sb, gdp),
2462 &fg->free_clusters);
2463 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2464 }
2465
2466 return 1;
2467failed:
2468 return 0;
2469}
2470
2471static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2472 struct ext4_group_desc *gdp)
2473{
2474 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2475 __u16 crc = 0;
2476 __le32 le_group = cpu_to_le32(block_group);
2477 struct ext4_sb_info *sbi = EXT4_SB(sb);
2478
2479 if (ext4_has_metadata_csum(sbi->s_sb)) {
2480 /* Use new metadata_csum algorithm */
2481 __u32 csum32;
2482 __u16 dummy_csum = 0;
2483
2484 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2485 sizeof(le_group));
2486 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2487 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2488 sizeof(dummy_csum));
2489 offset += sizeof(dummy_csum);
2490 if (offset < sbi->s_desc_size)
2491 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2492 sbi->s_desc_size - offset);
2493
2494 crc = csum32 & 0xFFFF;
2495 goto out;
2496 }
2497
2498 /* old crc16 code */
2499 if (!ext4_has_feature_gdt_csum(sb))
2500 return 0;
2501
2502 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2503 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2504 crc = crc16(crc, (__u8 *)gdp, offset);
2505 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2506 /* for checksum of struct ext4_group_desc do the rest...*/
2507 if (ext4_has_feature_64bit(sb) &&
2508 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2509 crc = crc16(crc, (__u8 *)gdp + offset,
2510 le16_to_cpu(sbi->s_es->s_desc_size) -
2511 offset);
2512
2513out:
2514 return cpu_to_le16(crc);
2515}
2516
2517int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2518 struct ext4_group_desc *gdp)
2519{
2520 if (ext4_has_group_desc_csum(sb) &&
2521 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2522 return 0;
2523
2524 return 1;
2525}
2526
2527void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2528 struct ext4_group_desc *gdp)
2529{
2530 if (!ext4_has_group_desc_csum(sb))
2531 return;
2532 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2533}
2534
2535/* Called at mount-time, super-block is locked */
2536static int ext4_check_descriptors(struct super_block *sb,
2537 ext4_fsblk_t sb_block,
2538 ext4_group_t *first_not_zeroed)
2539{
2540 struct ext4_sb_info *sbi = EXT4_SB(sb);
2541 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2542 ext4_fsblk_t last_block;
2543 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2544 ext4_fsblk_t block_bitmap;
2545 ext4_fsblk_t inode_bitmap;
2546 ext4_fsblk_t inode_table;
2547 int flexbg_flag = 0;
2548 ext4_group_t i, grp = sbi->s_groups_count;
2549
2550 if (ext4_has_feature_flex_bg(sb))
2551 flexbg_flag = 1;
2552
2553 ext4_debug("Checking group descriptors");
2554
2555 for (i = 0; i < sbi->s_groups_count; i++) {
2556 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2557
2558 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2559 last_block = ext4_blocks_count(sbi->s_es) - 1;
2560 else
2561 last_block = first_block +
2562 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2563
2564 if ((grp == sbi->s_groups_count) &&
2565 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2566 grp = i;
2567
2568 block_bitmap = ext4_block_bitmap(sb, gdp);
2569 if (block_bitmap == sb_block) {
2570 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2571 "Block bitmap for group %u overlaps "
2572 "superblock", i);
2573 if (!sb_rdonly(sb))
2574 return 0;
2575 }
2576 if (block_bitmap >= sb_block + 1 &&
2577 block_bitmap <= last_bg_block) {
2578 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2579 "Block bitmap for group %u overlaps "
2580 "block group descriptors", i);
2581 if (!sb_rdonly(sb))
2582 return 0;
2583 }
2584 if (block_bitmap < first_block || block_bitmap > last_block) {
2585 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2586 "Block bitmap for group %u not in group "
2587 "(block %llu)!", i, block_bitmap);
2588 return 0;
2589 }
2590 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2591 if (inode_bitmap == sb_block) {
2592 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2593 "Inode bitmap for group %u overlaps "
2594 "superblock", i);
2595 if (!sb_rdonly(sb))
2596 return 0;
2597 }
2598 if (inode_bitmap >= sb_block + 1 &&
2599 inode_bitmap <= last_bg_block) {
2600 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2601 "Inode bitmap for group %u overlaps "
2602 "block group descriptors", i);
2603 if (!sb_rdonly(sb))
2604 return 0;
2605 }
2606 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2607 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2608 "Inode bitmap for group %u not in group "
2609 "(block %llu)!", i, inode_bitmap);
2610 return 0;
2611 }
2612 inode_table = ext4_inode_table(sb, gdp);
2613 if (inode_table == sb_block) {
2614 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2615 "Inode table for group %u overlaps "
2616 "superblock", i);
2617 if (!sb_rdonly(sb))
2618 return 0;
2619 }
2620 if (inode_table >= sb_block + 1 &&
2621 inode_table <= last_bg_block) {
2622 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2623 "Inode table for group %u overlaps "
2624 "block group descriptors", i);
2625 if (!sb_rdonly(sb))
2626 return 0;
2627 }
2628 if (inode_table < first_block ||
2629 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2630 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2631 "Inode table for group %u not in group "
2632 "(block %llu)!", i, inode_table);
2633 return 0;
2634 }
2635 ext4_lock_group(sb, i);
2636 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2637 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2638 "Checksum for group %u failed (%u!=%u)",
2639 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2640 gdp)), le16_to_cpu(gdp->bg_checksum));
2641 if (!sb_rdonly(sb)) {
2642 ext4_unlock_group(sb, i);
2643 return 0;
2644 }
2645 }
2646 ext4_unlock_group(sb, i);
2647 if (!flexbg_flag)
2648 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2649 }
2650 if (NULL != first_not_zeroed)
2651 *first_not_zeroed = grp;
2652 return 1;
2653}
2654
2655/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2656 * the superblock) which were deleted from all directories, but held open by
2657 * a process at the time of a crash. We walk the list and try to delete these
2658 * inodes at recovery time (only with a read-write filesystem).
2659 *
2660 * In order to keep the orphan inode chain consistent during traversal (in
2661 * case of crash during recovery), we link each inode into the superblock
2662 * orphan list_head and handle it the same way as an inode deletion during
2663 * normal operation (which journals the operations for us).
2664 *
2665 * We only do an iget() and an iput() on each inode, which is very safe if we
2666 * accidentally point at an in-use or already deleted inode. The worst that
2667 * can happen in this case is that we get a "bit already cleared" message from
2668 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2669 * e2fsck was run on this filesystem, and it must have already done the orphan
2670 * inode cleanup for us, so we can safely abort without any further action.
2671 */
2672static void ext4_orphan_cleanup(struct super_block *sb,
2673 struct ext4_super_block *es)
2674{
2675 unsigned int s_flags = sb->s_flags;
2676 int ret, nr_orphans = 0, nr_truncates = 0;
2677#ifdef CONFIG_QUOTA
2678 int quota_update = 0;
2679 int i;
2680#endif
2681 if (!es->s_last_orphan) {
2682 jbd_debug(4, "no orphan inodes to clean up\n");
2683 return;
2684 }
2685
2686 if (bdev_read_only(sb->s_bdev)) {
2687 ext4_msg(sb, KERN_ERR, "write access "
2688 "unavailable, skipping orphan cleanup");
2689 return;
2690 }
2691
2692 /* Check if feature set would not allow a r/w mount */
2693 if (!ext4_feature_set_ok(sb, 0)) {
2694 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2695 "unknown ROCOMPAT features");
2696 return;
2697 }
2698
2699 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2700 /* don't clear list on RO mount w/ errors */
2701 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2702 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2703 "clearing orphan list.\n");
2704 es->s_last_orphan = 0;
2705 }
2706 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2707 return;
2708 }
2709
2710 if (s_flags & SB_RDONLY) {
2711 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2712 sb->s_flags &= ~SB_RDONLY;
2713 }
2714#ifdef CONFIG_QUOTA
2715 /* Needed for iput() to work correctly and not trash data */
2716 sb->s_flags |= SB_ACTIVE;
2717
2718 /*
2719 * Turn on quotas which were not enabled for read-only mounts if
2720 * filesystem has quota feature, so that they are updated correctly.
2721 */
2722 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2723 int ret = ext4_enable_quotas(sb);
2724
2725 if (!ret)
2726 quota_update = 1;
2727 else
2728 ext4_msg(sb, KERN_ERR,
2729 "Cannot turn on quotas: error %d", ret);
2730 }
2731
2732 /* Turn on journaled quotas used for old sytle */
2733 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2734 if (EXT4_SB(sb)->s_qf_names[i]) {
2735 int ret = ext4_quota_on_mount(sb, i);
2736
2737 if (!ret)
2738 quota_update = 1;
2739 else
2740 ext4_msg(sb, KERN_ERR,
2741 "Cannot turn on journaled "
2742 "quota: type %d: error %d", i, ret);
2743 }
2744 }
2745#endif
2746
2747 while (es->s_last_orphan) {
2748 struct inode *inode;
2749
2750 /*
2751 * We may have encountered an error during cleanup; if
2752 * so, skip the rest.
2753 */
2754 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2755 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2756 es->s_last_orphan = 0;
2757 break;
2758 }
2759
2760 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2761 if (IS_ERR(inode)) {
2762 es->s_last_orphan = 0;
2763 break;
2764 }
2765
2766 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2767 dquot_initialize(inode);
2768 if (inode->i_nlink) {
2769 if (test_opt(sb, DEBUG))
2770 ext4_msg(sb, KERN_DEBUG,
2771 "%s: truncating inode %lu to %lld bytes",
2772 __func__, inode->i_ino, inode->i_size);
2773 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2774 inode->i_ino, inode->i_size);
2775 inode_lock(inode);
2776 truncate_inode_pages(inode->i_mapping, inode->i_size);
2777 ret = ext4_truncate(inode);
2778 if (ret)
2779 ext4_std_error(inode->i_sb, ret);
2780 inode_unlock(inode);
2781 nr_truncates++;
2782 } else {
2783 if (test_opt(sb, DEBUG))
2784 ext4_msg(sb, KERN_DEBUG,
2785 "%s: deleting unreferenced inode %lu",
2786 __func__, inode->i_ino);
2787 jbd_debug(2, "deleting unreferenced inode %lu\n",
2788 inode->i_ino);
2789 nr_orphans++;
2790 }
2791 iput(inode); /* The delete magic happens here! */
2792 }
2793
2794#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2795
2796 if (nr_orphans)
2797 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2798 PLURAL(nr_orphans));
2799 if (nr_truncates)
2800 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2801 PLURAL(nr_truncates));
2802#ifdef CONFIG_QUOTA
2803 /* Turn off quotas if they were enabled for orphan cleanup */
2804 if (quota_update) {
2805 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2806 if (sb_dqopt(sb)->files[i])
2807 dquot_quota_off(sb, i);
2808 }
2809 }
2810#endif
2811 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2812}
2813
2814/*
2815 * Maximal extent format file size.
2816 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2817 * extent format containers, within a sector_t, and within i_blocks
2818 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2819 * so that won't be a limiting factor.
2820 *
2821 * However there is other limiting factor. We do store extents in the form
2822 * of starting block and length, hence the resulting length of the extent
2823 * covering maximum file size must fit into on-disk format containers as
2824 * well. Given that length is always by 1 unit bigger than max unit (because
2825 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2826 *
2827 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2828 */
2829static loff_t ext4_max_size(int blkbits, int has_huge_files)
2830{
2831 loff_t res;
2832 loff_t upper_limit = MAX_LFS_FILESIZE;
2833
2834 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2835
2836 if (!has_huge_files) {
2837 upper_limit = (1LL << 32) - 1;
2838
2839 /* total blocks in file system block size */
2840 upper_limit >>= (blkbits - 9);
2841 upper_limit <<= blkbits;
2842 }
2843
2844 /*
2845 * 32-bit extent-start container, ee_block. We lower the maxbytes
2846 * by one fs block, so ee_len can cover the extent of maximum file
2847 * size
2848 */
2849 res = (1LL << 32) - 1;
2850 res <<= blkbits;
2851
2852 /* Sanity check against vm- & vfs- imposed limits */
2853 if (res > upper_limit)
2854 res = upper_limit;
2855
2856 return res;
2857}
2858
2859/*
2860 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2861 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2862 * We need to be 1 filesystem block less than the 2^48 sector limit.
2863 */
2864static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2865{
2866 loff_t res = EXT4_NDIR_BLOCKS;
2867 int meta_blocks;
2868 loff_t upper_limit;
2869 /* This is calculated to be the largest file size for a dense, block
2870 * mapped file such that the file's total number of 512-byte sectors,
2871 * including data and all indirect blocks, does not exceed (2^48 - 1).
2872 *
2873 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2874 * number of 512-byte sectors of the file.
2875 */
2876
2877 if (!has_huge_files) {
2878 /*
2879 * !has_huge_files or implies that the inode i_block field
2880 * represents total file blocks in 2^32 512-byte sectors ==
2881 * size of vfs inode i_blocks * 8
2882 */
2883 upper_limit = (1LL << 32) - 1;
2884
2885 /* total blocks in file system block size */
2886 upper_limit >>= (bits - 9);
2887
2888 } else {
2889 /*
2890 * We use 48 bit ext4_inode i_blocks
2891 * With EXT4_HUGE_FILE_FL set the i_blocks
2892 * represent total number of blocks in
2893 * file system block size
2894 */
2895 upper_limit = (1LL << 48) - 1;
2896
2897 }
2898
2899 /* indirect blocks */
2900 meta_blocks = 1;
2901 /* double indirect blocks */
2902 meta_blocks += 1 + (1LL << (bits-2));
2903 /* tripple indirect blocks */
2904 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2905
2906 upper_limit -= meta_blocks;
2907 upper_limit <<= bits;
2908
2909 res += 1LL << (bits-2);
2910 res += 1LL << (2*(bits-2));
2911 res += 1LL << (3*(bits-2));
2912 res <<= bits;
2913 if (res > upper_limit)
2914 res = upper_limit;
2915
2916 if (res > MAX_LFS_FILESIZE)
2917 res = MAX_LFS_FILESIZE;
2918
2919 return res;
2920}
2921
2922static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2923 ext4_fsblk_t logical_sb_block, int nr)
2924{
2925 struct ext4_sb_info *sbi = EXT4_SB(sb);
2926 ext4_group_t bg, first_meta_bg;
2927 int has_super = 0;
2928
2929 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2930
2931 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2932 return logical_sb_block + nr + 1;
2933 bg = sbi->s_desc_per_block * nr;
2934 if (ext4_bg_has_super(sb, bg))
2935 has_super = 1;
2936
2937 /*
2938 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2939 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2940 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2941 * compensate.
2942 */
2943 if (sb->s_blocksize == 1024 && nr == 0 &&
2944 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
2945 has_super++;
2946
2947 return (has_super + ext4_group_first_block_no(sb, bg));
2948}
2949
2950/**
2951 * ext4_get_stripe_size: Get the stripe size.
2952 * @sbi: In memory super block info
2953 *
2954 * If we have specified it via mount option, then
2955 * use the mount option value. If the value specified at mount time is
2956 * greater than the blocks per group use the super block value.
2957 * If the super block value is greater than blocks per group return 0.
2958 * Allocator needs it be less than blocks per group.
2959 *
2960 */
2961static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2962{
2963 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2964 unsigned long stripe_width =
2965 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2966 int ret;
2967
2968 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2969 ret = sbi->s_stripe;
2970 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2971 ret = stripe_width;
2972 else if (stride && stride <= sbi->s_blocks_per_group)
2973 ret = stride;
2974 else
2975 ret = 0;
2976
2977 /*
2978 * If the stripe width is 1, this makes no sense and
2979 * we set it to 0 to turn off stripe handling code.
2980 */
2981 if (ret <= 1)
2982 ret = 0;
2983
2984 return ret;
2985}
2986
2987/*
2988 * Check whether this filesystem can be mounted based on
2989 * the features present and the RDONLY/RDWR mount requested.
2990 * Returns 1 if this filesystem can be mounted as requested,
2991 * 0 if it cannot be.
2992 */
2993static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2994{
2995 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2996 ext4_msg(sb, KERN_ERR,
2997 "Couldn't mount because of "
2998 "unsupported optional features (%x)",
2999 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3000 ~EXT4_FEATURE_INCOMPAT_SUPP));
3001 return 0;
3002 }
3003
3004#ifndef CONFIG_UNICODE
3005 if (ext4_has_feature_casefold(sb)) {
3006 ext4_msg(sb, KERN_ERR,
3007 "Filesystem with casefold feature cannot be "
3008 "mounted without CONFIG_UNICODE");
3009 return 0;
3010 }
3011#endif
3012
3013 if (readonly)
3014 return 1;
3015
3016 if (ext4_has_feature_readonly(sb)) {
3017 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3018 sb->s_flags |= SB_RDONLY;
3019 return 1;
3020 }
3021
3022 /* Check that feature set is OK for a read-write mount */
3023 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3024 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3025 "unsupported optional features (%x)",
3026 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3027 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3028 return 0;
3029 }
3030 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3031 ext4_msg(sb, KERN_ERR,
3032 "Can't support bigalloc feature without "
3033 "extents feature\n");
3034 return 0;
3035 }
3036
3037#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3038 if (!readonly && (ext4_has_feature_quota(sb) ||
3039 ext4_has_feature_project(sb))) {
3040 ext4_msg(sb, KERN_ERR,
3041 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3042 return 0;
3043 }
3044#endif /* CONFIG_QUOTA */
3045 return 1;
3046}
3047
3048/*
3049 * This function is called once a day if we have errors logged
3050 * on the file system
3051 */
3052static void print_daily_error_info(struct timer_list *t)
3053{
3054 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3055 struct super_block *sb = sbi->s_sb;
3056 struct ext4_super_block *es = sbi->s_es;
3057
3058 if (es->s_error_count)
3059 /* fsck newer than v1.41.13 is needed to clean this condition. */
3060 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3061 le32_to_cpu(es->s_error_count));
3062 if (es->s_first_error_time) {
3063 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3064 sb->s_id,
3065 ext4_get_tstamp(es, s_first_error_time),
3066 (int) sizeof(es->s_first_error_func),
3067 es->s_first_error_func,
3068 le32_to_cpu(es->s_first_error_line));
3069 if (es->s_first_error_ino)
3070 printk(KERN_CONT ": inode %u",
3071 le32_to_cpu(es->s_first_error_ino));
3072 if (es->s_first_error_block)
3073 printk(KERN_CONT ": block %llu", (unsigned long long)
3074 le64_to_cpu(es->s_first_error_block));
3075 printk(KERN_CONT "\n");
3076 }
3077 if (es->s_last_error_time) {
3078 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3079 sb->s_id,
3080 ext4_get_tstamp(es, s_last_error_time),
3081 (int) sizeof(es->s_last_error_func),
3082 es->s_last_error_func,
3083 le32_to_cpu(es->s_last_error_line));
3084 if (es->s_last_error_ino)
3085 printk(KERN_CONT ": inode %u",
3086 le32_to_cpu(es->s_last_error_ino));
3087 if (es->s_last_error_block)
3088 printk(KERN_CONT ": block %llu", (unsigned long long)
3089 le64_to_cpu(es->s_last_error_block));
3090 printk(KERN_CONT "\n");
3091 }
3092 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3093}
3094
3095/* Find next suitable group and run ext4_init_inode_table */
3096static int ext4_run_li_request(struct ext4_li_request *elr)
3097{
3098 struct ext4_group_desc *gdp = NULL;
3099 ext4_group_t group, ngroups;
3100 struct super_block *sb;
3101 unsigned long timeout = 0;
3102 int ret = 0;
3103
3104 sb = elr->lr_super;
3105 ngroups = EXT4_SB(sb)->s_groups_count;
3106
3107 for (group = elr->lr_next_group; group < ngroups; group++) {
3108 gdp = ext4_get_group_desc(sb, group, NULL);
3109 if (!gdp) {
3110 ret = 1;
3111 break;
3112 }
3113
3114 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3115 break;
3116 }
3117
3118 if (group >= ngroups)
3119 ret = 1;
3120
3121 if (!ret) {
3122 timeout = jiffies;
3123 ret = ext4_init_inode_table(sb, group,
3124 elr->lr_timeout ? 0 : 1);
3125 if (elr->lr_timeout == 0) {
3126 timeout = (jiffies - timeout) *
3127 elr->lr_sbi->s_li_wait_mult;
3128 elr->lr_timeout = timeout;
3129 }
3130 elr->lr_next_sched = jiffies + elr->lr_timeout;
3131 elr->lr_next_group = group + 1;
3132 }
3133 return ret;
3134}
3135
3136/*
3137 * Remove lr_request from the list_request and free the
3138 * request structure. Should be called with li_list_mtx held
3139 */
3140static void ext4_remove_li_request(struct ext4_li_request *elr)
3141{
3142 struct ext4_sb_info *sbi;
3143
3144 if (!elr)
3145 return;
3146
3147 sbi = elr->lr_sbi;
3148
3149 list_del(&elr->lr_request);
3150 sbi->s_li_request = NULL;
3151 kfree(elr);
3152}
3153
3154static void ext4_unregister_li_request(struct super_block *sb)
3155{
3156 mutex_lock(&ext4_li_mtx);
3157 if (!ext4_li_info) {
3158 mutex_unlock(&ext4_li_mtx);
3159 return;
3160 }
3161
3162 mutex_lock(&ext4_li_info->li_list_mtx);
3163 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3164 mutex_unlock(&ext4_li_info->li_list_mtx);
3165 mutex_unlock(&ext4_li_mtx);
3166}
3167
3168static struct task_struct *ext4_lazyinit_task;
3169
3170/*
3171 * This is the function where ext4lazyinit thread lives. It walks
3172 * through the request list searching for next scheduled filesystem.
3173 * When such a fs is found, run the lazy initialization request
3174 * (ext4_rn_li_request) and keep track of the time spend in this
3175 * function. Based on that time we compute next schedule time of
3176 * the request. When walking through the list is complete, compute
3177 * next waking time and put itself into sleep.
3178 */
3179static int ext4_lazyinit_thread(void *arg)
3180{
3181 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3182 struct list_head *pos, *n;
3183 struct ext4_li_request *elr;
3184 unsigned long next_wakeup, cur;
3185
3186 BUG_ON(NULL == eli);
3187
3188cont_thread:
3189 while (true) {
3190 next_wakeup = MAX_JIFFY_OFFSET;
3191
3192 mutex_lock(&eli->li_list_mtx);
3193 if (list_empty(&eli->li_request_list)) {
3194 mutex_unlock(&eli->li_list_mtx);
3195 goto exit_thread;
3196 }
3197 list_for_each_safe(pos, n, &eli->li_request_list) {
3198 int err = 0;
3199 int progress = 0;
3200 elr = list_entry(pos, struct ext4_li_request,
3201 lr_request);
3202
3203 if (time_before(jiffies, elr->lr_next_sched)) {
3204 if (time_before(elr->lr_next_sched, next_wakeup))
3205 next_wakeup = elr->lr_next_sched;
3206 continue;
3207 }
3208 if (down_read_trylock(&elr->lr_super->s_umount)) {
3209 if (sb_start_write_trylock(elr->lr_super)) {
3210 progress = 1;
3211 /*
3212 * We hold sb->s_umount, sb can not
3213 * be removed from the list, it is
3214 * now safe to drop li_list_mtx
3215 */
3216 mutex_unlock(&eli->li_list_mtx);
3217 err = ext4_run_li_request(elr);
3218 sb_end_write(elr->lr_super);
3219 mutex_lock(&eli->li_list_mtx);
3220 n = pos->next;
3221 }
3222 up_read((&elr->lr_super->s_umount));
3223 }
3224 /* error, remove the lazy_init job */
3225 if (err) {
3226 ext4_remove_li_request(elr);
3227 continue;
3228 }
3229 if (!progress) {
3230 elr->lr_next_sched = jiffies +
3231 (prandom_u32()
3232 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3233 }
3234 if (time_before(elr->lr_next_sched, next_wakeup))
3235 next_wakeup = elr->lr_next_sched;
3236 }
3237 mutex_unlock(&eli->li_list_mtx);
3238
3239 try_to_freeze();
3240
3241 cur = jiffies;
3242 if ((time_after_eq(cur, next_wakeup)) ||
3243 (MAX_JIFFY_OFFSET == next_wakeup)) {
3244 cond_resched();
3245 continue;
3246 }
3247
3248 schedule_timeout_interruptible(next_wakeup - cur);
3249
3250 if (kthread_should_stop()) {
3251 ext4_clear_request_list();
3252 goto exit_thread;
3253 }
3254 }
3255
3256exit_thread:
3257 /*
3258 * It looks like the request list is empty, but we need
3259 * to check it under the li_list_mtx lock, to prevent any
3260 * additions into it, and of course we should lock ext4_li_mtx
3261 * to atomically free the list and ext4_li_info, because at
3262 * this point another ext4 filesystem could be registering
3263 * new one.
3264 */
3265 mutex_lock(&ext4_li_mtx);
3266 mutex_lock(&eli->li_list_mtx);
3267 if (!list_empty(&eli->li_request_list)) {
3268 mutex_unlock(&eli->li_list_mtx);
3269 mutex_unlock(&ext4_li_mtx);
3270 goto cont_thread;
3271 }
3272 mutex_unlock(&eli->li_list_mtx);
3273 kfree(ext4_li_info);
3274 ext4_li_info = NULL;
3275 mutex_unlock(&ext4_li_mtx);
3276
3277 return 0;
3278}
3279
3280static void ext4_clear_request_list(void)
3281{
3282 struct list_head *pos, *n;
3283 struct ext4_li_request *elr;
3284
3285 mutex_lock(&ext4_li_info->li_list_mtx);
3286 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3287 elr = list_entry(pos, struct ext4_li_request,
3288 lr_request);
3289 ext4_remove_li_request(elr);
3290 }
3291 mutex_unlock(&ext4_li_info->li_list_mtx);
3292}
3293
3294static int ext4_run_lazyinit_thread(void)
3295{
3296 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3297 ext4_li_info, "ext4lazyinit");
3298 if (IS_ERR(ext4_lazyinit_task)) {
3299 int err = PTR_ERR(ext4_lazyinit_task);
3300 ext4_clear_request_list();
3301 kfree(ext4_li_info);
3302 ext4_li_info = NULL;
3303 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3304 "initialization thread\n",
3305 err);
3306 return err;
3307 }
3308 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3309 return 0;
3310}
3311
3312/*
3313 * Check whether it make sense to run itable init. thread or not.
3314 * If there is at least one uninitialized inode table, return
3315 * corresponding group number, else the loop goes through all
3316 * groups and return total number of groups.
3317 */
3318static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3319{
3320 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3321 struct ext4_group_desc *gdp = NULL;
3322
3323 if (!ext4_has_group_desc_csum(sb))
3324 return ngroups;
3325
3326 for (group = 0; group < ngroups; group++) {
3327 gdp = ext4_get_group_desc(sb, group, NULL);
3328 if (!gdp)
3329 continue;
3330
3331 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3332 break;
3333 }
3334
3335 return group;
3336}
3337
3338static int ext4_li_info_new(void)
3339{
3340 struct ext4_lazy_init *eli = NULL;
3341
3342 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3343 if (!eli)
3344 return -ENOMEM;
3345
3346 INIT_LIST_HEAD(&eli->li_request_list);
3347 mutex_init(&eli->li_list_mtx);
3348
3349 eli->li_state |= EXT4_LAZYINIT_QUIT;
3350
3351 ext4_li_info = eli;
3352
3353 return 0;
3354}
3355
3356static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3357 ext4_group_t start)
3358{
3359 struct ext4_sb_info *sbi = EXT4_SB(sb);
3360 struct ext4_li_request *elr;
3361
3362 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3363 if (!elr)
3364 return NULL;
3365
3366 elr->lr_super = sb;
3367 elr->lr_sbi = sbi;
3368 elr->lr_next_group = start;
3369
3370 /*
3371 * Randomize first schedule time of the request to
3372 * spread the inode table initialization requests
3373 * better.
3374 */
3375 elr->lr_next_sched = jiffies + (prandom_u32() %
3376 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3377 return elr;
3378}
3379
3380int ext4_register_li_request(struct super_block *sb,
3381 ext4_group_t first_not_zeroed)
3382{
3383 struct ext4_sb_info *sbi = EXT4_SB(sb);
3384 struct ext4_li_request *elr = NULL;
3385 ext4_group_t ngroups = sbi->s_groups_count;
3386 int ret = 0;
3387
3388 mutex_lock(&ext4_li_mtx);
3389 if (sbi->s_li_request != NULL) {
3390 /*
3391 * Reset timeout so it can be computed again, because
3392 * s_li_wait_mult might have changed.
3393 */
3394 sbi->s_li_request->lr_timeout = 0;
3395 goto out;
3396 }
3397
3398 if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3399 !test_opt(sb, INIT_INODE_TABLE))
3400 goto out;
3401
3402 elr = ext4_li_request_new(sb, first_not_zeroed);
3403 if (!elr) {
3404 ret = -ENOMEM;
3405 goto out;
3406 }
3407
3408 if (NULL == ext4_li_info) {
3409 ret = ext4_li_info_new();
3410 if (ret)
3411 goto out;
3412 }
3413
3414 mutex_lock(&ext4_li_info->li_list_mtx);
3415 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3416 mutex_unlock(&ext4_li_info->li_list_mtx);
3417
3418 sbi->s_li_request = elr;
3419 /*
3420 * set elr to NULL here since it has been inserted to
3421 * the request_list and the removal and free of it is
3422 * handled by ext4_clear_request_list from now on.
3423 */
3424 elr = NULL;
3425
3426 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3427 ret = ext4_run_lazyinit_thread();
3428 if (ret)
3429 goto out;
3430 }
3431out:
3432 mutex_unlock(&ext4_li_mtx);
3433 if (ret)
3434 kfree(elr);
3435 return ret;
3436}
3437
3438/*
3439 * We do not need to lock anything since this is called on
3440 * module unload.
3441 */
3442static void ext4_destroy_lazyinit_thread(void)
3443{
3444 /*
3445 * If thread exited earlier
3446 * there's nothing to be done.
3447 */
3448 if (!ext4_li_info || !ext4_lazyinit_task)
3449 return;
3450
3451 kthread_stop(ext4_lazyinit_task);
3452}
3453
3454static int set_journal_csum_feature_set(struct super_block *sb)
3455{
3456 int ret = 1;
3457 int compat, incompat;
3458 struct ext4_sb_info *sbi = EXT4_SB(sb);
3459
3460 if (ext4_has_metadata_csum(sb)) {
3461 /* journal checksum v3 */
3462 compat = 0;
3463 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3464 } else {
3465 /* journal checksum v1 */
3466 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3467 incompat = 0;
3468 }
3469
3470 jbd2_journal_clear_features(sbi->s_journal,
3471 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3472 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3473 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3474 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3475 ret = jbd2_journal_set_features(sbi->s_journal,
3476 compat, 0,
3477 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3478 incompat);
3479 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3480 ret = jbd2_journal_set_features(sbi->s_journal,
3481 compat, 0,
3482 incompat);
3483 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3484 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3485 } else {
3486 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3487 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3488 }
3489
3490 return ret;
3491}
3492
3493/*
3494 * Note: calculating the overhead so we can be compatible with
3495 * historical BSD practice is quite difficult in the face of
3496 * clusters/bigalloc. This is because multiple metadata blocks from
3497 * different block group can end up in the same allocation cluster.
3498 * Calculating the exact overhead in the face of clustered allocation
3499 * requires either O(all block bitmaps) in memory or O(number of block
3500 * groups**2) in time. We will still calculate the superblock for
3501 * older file systems --- and if we come across with a bigalloc file
3502 * system with zero in s_overhead_clusters the estimate will be close to
3503 * correct especially for very large cluster sizes --- but for newer
3504 * file systems, it's better to calculate this figure once at mkfs
3505 * time, and store it in the superblock. If the superblock value is
3506 * present (even for non-bigalloc file systems), we will use it.
3507 */
3508static int count_overhead(struct super_block *sb, ext4_group_t grp,
3509 char *buf)
3510{
3511 struct ext4_sb_info *sbi = EXT4_SB(sb);
3512 struct ext4_group_desc *gdp;
3513 ext4_fsblk_t first_block, last_block, b;
3514 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3515 int s, j, count = 0;
3516
3517 if (!ext4_has_feature_bigalloc(sb))
3518 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3519 sbi->s_itb_per_group + 2);
3520
3521 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3522 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3523 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3524 for (i = 0; i < ngroups; i++) {
3525 gdp = ext4_get_group_desc(sb, i, NULL);
3526 b = ext4_block_bitmap(sb, gdp);
3527 if (b >= first_block && b <= last_block) {
3528 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3529 count++;
3530 }
3531 b = ext4_inode_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_table(sb, gdp);
3537 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3538 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3539 int c = EXT4_B2C(sbi, b - first_block);
3540 ext4_set_bit(c, buf);
3541 count++;
3542 }
3543 if (i != grp)
3544 continue;
3545 s = 0;
3546 if (ext4_bg_has_super(sb, grp)) {
3547 ext4_set_bit(s++, buf);
3548 count++;
3549 }
3550 j = ext4_bg_num_gdb(sb, grp);
3551 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3552 ext4_error(sb, "Invalid number of block group "
3553 "descriptor blocks: %d", j);
3554 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3555 }
3556 count += j;
3557 for (; j > 0; j--)
3558 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3559 }
3560 if (!count)
3561 return 0;
3562 return EXT4_CLUSTERS_PER_GROUP(sb) -
3563 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3564}
3565
3566/*
3567 * Compute the overhead and stash it in sbi->s_overhead
3568 */
3569int ext4_calculate_overhead(struct super_block *sb)
3570{
3571 struct ext4_sb_info *sbi = EXT4_SB(sb);
3572 struct ext4_super_block *es = sbi->s_es;
3573 struct inode *j_inode;
3574 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3575 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3576 ext4_fsblk_t overhead = 0;
3577 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3578
3579 if (!buf)
3580 return -ENOMEM;
3581
3582 /*
3583 * Compute the overhead (FS structures). This is constant
3584 * for a given filesystem unless the number of block groups
3585 * changes so we cache the previous value until it does.
3586 */
3587
3588 /*
3589 * All of the blocks before first_data_block are overhead
3590 */
3591 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3592
3593 /*
3594 * Add the overhead found in each block group
3595 */
3596 for (i = 0; i < ngroups; i++) {
3597 int blks;
3598
3599 blks = count_overhead(sb, i, buf);
3600 overhead += blks;
3601 if (blks)
3602 memset(buf, 0, PAGE_SIZE);
3603 cond_resched();
3604 }
3605
3606 /*
3607 * Add the internal journal blocks whether the journal has been
3608 * loaded or not
3609 */
3610 if (sbi->s_journal && !sbi->journal_bdev)
3611 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3612 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3613 /* j_inum for internal journal is non-zero */
3614 j_inode = ext4_get_journal_inode(sb, j_inum);
3615 if (j_inode) {
3616 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3617 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3618 iput(j_inode);
3619 } else {
3620 ext4_msg(sb, KERN_ERR, "can't get journal size");
3621 }
3622 }
3623 sbi->s_overhead = overhead;
3624 smp_wmb();
3625 free_page((unsigned long) buf);
3626 return 0;
3627}
3628
3629static void ext4_set_resv_clusters(struct super_block *sb)
3630{
3631 ext4_fsblk_t resv_clusters;
3632 struct ext4_sb_info *sbi = EXT4_SB(sb);
3633
3634 /*
3635 * There's no need to reserve anything when we aren't using extents.
3636 * The space estimates are exact, there are no unwritten extents,
3637 * hole punching doesn't need new metadata... This is needed especially
3638 * to keep ext2/3 backward compatibility.
3639 */
3640 if (!ext4_has_feature_extents(sb))
3641 return;
3642 /*
3643 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3644 * This should cover the situations where we can not afford to run
3645 * out of space like for example punch hole, or converting
3646 * unwritten extents in delalloc path. In most cases such
3647 * allocation would require 1, or 2 blocks, higher numbers are
3648 * very rare.
3649 */
3650 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3651 sbi->s_cluster_bits);
3652
3653 do_div(resv_clusters, 50);
3654 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3655
3656 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3657}
3658
3659static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3660{
3661 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3662 char *orig_data = kstrdup(data, GFP_KERNEL);
3663 struct buffer_head *bh, **group_desc;
3664 struct ext4_super_block *es = NULL;
3665 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3666 struct flex_groups **flex_groups;
3667 ext4_fsblk_t block;
3668 ext4_fsblk_t sb_block = get_sb_block(&data);
3669 ext4_fsblk_t logical_sb_block;
3670 unsigned long offset = 0;
3671 unsigned long journal_devnum = 0;
3672 unsigned long def_mount_opts;
3673 struct inode *root;
3674 const char *descr;
3675 int ret = -ENOMEM;
3676 int blocksize, clustersize;
3677 unsigned int db_count;
3678 unsigned int i;
3679 int needs_recovery, has_huge_files, has_bigalloc;
3680 __u64 blocks_count;
3681 int err = 0;
3682 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3683 ext4_group_t first_not_zeroed;
3684
3685 if ((data && !orig_data) || !sbi)
3686 goto out_free_base;
3687
3688 sbi->s_daxdev = dax_dev;
3689 sbi->s_blockgroup_lock =
3690 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3691 if (!sbi->s_blockgroup_lock)
3692 goto out_free_base;
3693
3694 sb->s_fs_info = sbi;
3695 sbi->s_sb = sb;
3696 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3697 sbi->s_sb_block = sb_block;
3698 if (sb->s_bdev->bd_part)
3699 sbi->s_sectors_written_start =
3700 part_stat_read(sb->s_bdev->bd_part, sectors[STAT_WRITE]);
3701
3702 /* Cleanup superblock name */
3703 strreplace(sb->s_id, '/', '!');
3704
3705 /* -EINVAL is default */
3706 ret = -EINVAL;
3707 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3708 if (!blocksize) {
3709 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3710 goto out_fail;
3711 }
3712
3713 /*
3714 * The ext4 superblock will not be buffer aligned for other than 1kB
3715 * block sizes. We need to calculate the offset from buffer start.
3716 */
3717 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3718 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3719 offset = do_div(logical_sb_block, blocksize);
3720 } else {
3721 logical_sb_block = sb_block;
3722 }
3723
3724 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3725 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3726 goto out_fail;
3727 }
3728 /*
3729 * Note: s_es must be initialized as soon as possible because
3730 * some ext4 macro-instructions depend on its value
3731 */
3732 es = (struct ext4_super_block *) (bh->b_data + offset);
3733 sbi->s_es = es;
3734 sb->s_magic = le16_to_cpu(es->s_magic);
3735 if (sb->s_magic != EXT4_SUPER_MAGIC)
3736 goto cantfind_ext4;
3737 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3738
3739 /* Warn if metadata_csum and gdt_csum are both set. */
3740 if (ext4_has_feature_metadata_csum(sb) &&
3741 ext4_has_feature_gdt_csum(sb))
3742 ext4_warning(sb, "metadata_csum and uninit_bg are "
3743 "redundant flags; please run fsck.");
3744
3745 /* Check for a known checksum algorithm */
3746 if (!ext4_verify_csum_type(sb, es)) {
3747 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3748 "unknown checksum algorithm.");
3749 silent = 1;
3750 goto cantfind_ext4;
3751 }
3752
3753 /* Load the checksum driver */
3754 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3755 if (IS_ERR(sbi->s_chksum_driver)) {
3756 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3757 ret = PTR_ERR(sbi->s_chksum_driver);
3758 sbi->s_chksum_driver = NULL;
3759 goto failed_mount;
3760 }
3761
3762 /* Check superblock checksum */
3763 if (!ext4_superblock_csum_verify(sb, es)) {
3764 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3765 "invalid superblock checksum. Run e2fsck?");
3766 silent = 1;
3767 ret = -EFSBADCRC;
3768 goto cantfind_ext4;
3769 }
3770
3771 /* Precompute checksum seed for all metadata */
3772 if (ext4_has_feature_csum_seed(sb))
3773 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3774 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3775 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3776 sizeof(es->s_uuid));
3777
3778 /* Set defaults before we parse the mount options */
3779 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3780 set_opt(sb, INIT_INODE_TABLE);
3781 if (def_mount_opts & EXT4_DEFM_DEBUG)
3782 set_opt(sb, DEBUG);
3783 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3784 set_opt(sb, GRPID);
3785 if (def_mount_opts & EXT4_DEFM_UID16)
3786 set_opt(sb, NO_UID32);
3787 /* xattr user namespace & acls are now defaulted on */
3788 set_opt(sb, XATTR_USER);
3789 set_opt(sb, DIOREAD_NOLOCK);
3790#ifdef CONFIG_EXT4_FS_POSIX_ACL
3791 set_opt(sb, POSIX_ACL);
3792#endif
3793 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3794 if (ext4_has_metadata_csum(sb))
3795 set_opt(sb, JOURNAL_CHECKSUM);
3796
3797 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3798 set_opt(sb, JOURNAL_DATA);
3799 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3800 set_opt(sb, ORDERED_DATA);
3801 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3802 set_opt(sb, WRITEBACK_DATA);
3803
3804 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3805 set_opt(sb, ERRORS_PANIC);
3806 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3807 set_opt(sb, ERRORS_CONT);
3808 else
3809 set_opt(sb, ERRORS_RO);
3810 /* block_validity enabled by default; disable with noblock_validity */
3811 set_opt(sb, BLOCK_VALIDITY);
3812 if (def_mount_opts & EXT4_DEFM_DISCARD)
3813 set_opt(sb, DISCARD);
3814
3815 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3816 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3817 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3818 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3819 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3820
3821 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3822 set_opt(sb, BARRIER);
3823
3824 /*
3825 * enable delayed allocation by default
3826 * Use -o nodelalloc to turn it off
3827 */
3828 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3829 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3830 set_opt(sb, DELALLOC);
3831
3832 /*
3833 * set default s_li_wait_mult for lazyinit, for the case there is
3834 * no mount option specified.
3835 */
3836 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3837
3838 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3839 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3840 blocksize > EXT4_MAX_BLOCK_SIZE) {
3841 ext4_msg(sb, KERN_ERR,
3842 "Unsupported filesystem blocksize %d (%d log_block_size)",
3843 blocksize, le32_to_cpu(es->s_log_block_size));
3844 goto failed_mount;
3845 }
3846
3847 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3848 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3849 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3850 } else {
3851 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3852 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3853 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
3854 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
3855 sbi->s_first_ino);
3856 goto failed_mount;
3857 }
3858 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3859 (!is_power_of_2(sbi->s_inode_size)) ||
3860 (sbi->s_inode_size > blocksize)) {
3861 ext4_msg(sb, KERN_ERR,
3862 "unsupported inode size: %d",
3863 sbi->s_inode_size);
3864 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
3865 goto failed_mount;
3866 }
3867 /*
3868 * i_atime_extra is the last extra field available for
3869 * [acm]times in struct ext4_inode. Checking for that
3870 * field should suffice to ensure we have extra space
3871 * for all three.
3872 */
3873 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
3874 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
3875 sb->s_time_gran = 1;
3876 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
3877 } else {
3878 sb->s_time_gran = NSEC_PER_SEC;
3879 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
3880 }
3881 sb->s_time_min = EXT4_TIMESTAMP_MIN;
3882 }
3883 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3884 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3885 EXT4_GOOD_OLD_INODE_SIZE;
3886 if (ext4_has_feature_extra_isize(sb)) {
3887 unsigned v, max = (sbi->s_inode_size -
3888 EXT4_GOOD_OLD_INODE_SIZE);
3889
3890 v = le16_to_cpu(es->s_want_extra_isize);
3891 if (v > max) {
3892 ext4_msg(sb, KERN_ERR,
3893 "bad s_want_extra_isize: %d", v);
3894 goto failed_mount;
3895 }
3896 if (sbi->s_want_extra_isize < v)
3897 sbi->s_want_extra_isize = v;
3898
3899 v = le16_to_cpu(es->s_min_extra_isize);
3900 if (v > max) {
3901 ext4_msg(sb, KERN_ERR,
3902 "bad s_min_extra_isize: %d", v);
3903 goto failed_mount;
3904 }
3905 if (sbi->s_want_extra_isize < v)
3906 sbi->s_want_extra_isize = v;
3907 }
3908 }
3909
3910 if (sbi->s_es->s_mount_opts[0]) {
3911 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3912 sizeof(sbi->s_es->s_mount_opts),
3913 GFP_KERNEL);
3914 if (!s_mount_opts)
3915 goto failed_mount;
3916 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3917 &journal_ioprio, 0)) {
3918 ext4_msg(sb, KERN_WARNING,
3919 "failed to parse options in superblock: %s",
3920 s_mount_opts);
3921 }
3922 kfree(s_mount_opts);
3923 }
3924 sbi->s_def_mount_opt = sbi->s_mount_opt;
3925 if (!parse_options((char *) data, sb, &journal_devnum,
3926 &journal_ioprio, 0))
3927 goto failed_mount;
3928
3929#ifdef CONFIG_UNICODE
3930 if (ext4_has_feature_casefold(sb) && !sbi->s_encoding) {
3931 const struct ext4_sb_encodings *encoding_info;
3932 struct unicode_map *encoding;
3933 __u16 encoding_flags;
3934
3935 if (ext4_has_feature_encrypt(sb)) {
3936 ext4_msg(sb, KERN_ERR,
3937 "Can't mount with encoding and encryption");
3938 goto failed_mount;
3939 }
3940
3941 if (ext4_sb_read_encoding(es, &encoding_info,
3942 &encoding_flags)) {
3943 ext4_msg(sb, KERN_ERR,
3944 "Encoding requested by superblock is unknown");
3945 goto failed_mount;
3946 }
3947
3948 encoding = utf8_load(encoding_info->version);
3949 if (IS_ERR(encoding)) {
3950 ext4_msg(sb, KERN_ERR,
3951 "can't mount with superblock charset: %s-%s "
3952 "not supported by the kernel. flags: 0x%x.",
3953 encoding_info->name, encoding_info->version,
3954 encoding_flags);
3955 goto failed_mount;
3956 }
3957 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
3958 "%s-%s with flags 0x%hx", encoding_info->name,
3959 encoding_info->version?:"\b", encoding_flags);
3960
3961 sbi->s_encoding = encoding;
3962 sbi->s_encoding_flags = encoding_flags;
3963 }
3964#endif
3965
3966 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3967 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, and O_DIRECT support!\n");
3968 clear_opt(sb, DIOREAD_NOLOCK);
3969 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3970 ext4_msg(sb, KERN_ERR, "can't mount with "
3971 "both data=journal and delalloc");
3972 goto failed_mount;
3973 }
3974 if (test_opt(sb, DIOREAD_NOLOCK)) {
3975 ext4_msg(sb, KERN_ERR, "can't mount with "
3976 "both data=journal and dioread_nolock");
3977 goto failed_mount;
3978 }
3979 if (test_opt(sb, DAX)) {
3980 ext4_msg(sb, KERN_ERR, "can't mount with "
3981 "both data=journal and dax");
3982 goto failed_mount;
3983 }
3984 if (ext4_has_feature_encrypt(sb)) {
3985 ext4_msg(sb, KERN_WARNING,
3986 "encrypted files will use data=ordered "
3987 "instead of data journaling mode");
3988 }
3989 if (test_opt(sb, DELALLOC))
3990 clear_opt(sb, DELALLOC);
3991 } else {
3992 sb->s_iflags |= SB_I_CGROUPWB;
3993 }
3994
3995 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
3996 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
3997
3998 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3999 (ext4_has_compat_features(sb) ||
4000 ext4_has_ro_compat_features(sb) ||
4001 ext4_has_incompat_features(sb)))
4002 ext4_msg(sb, KERN_WARNING,
4003 "feature flags set on rev 0 fs, "
4004 "running e2fsck is recommended");
4005
4006 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4007 set_opt2(sb, HURD_COMPAT);
4008 if (ext4_has_feature_64bit(sb)) {
4009 ext4_msg(sb, KERN_ERR,
4010 "The Hurd can't support 64-bit file systems");
4011 goto failed_mount;
4012 }
4013
4014 /*
4015 * ea_inode feature uses l_i_version field which is not
4016 * available in HURD_COMPAT mode.
4017 */
4018 if (ext4_has_feature_ea_inode(sb)) {
4019 ext4_msg(sb, KERN_ERR,
4020 "ea_inode feature is not supported for Hurd");
4021 goto failed_mount;
4022 }
4023 }
4024
4025 if (IS_EXT2_SB(sb)) {
4026 if (ext2_feature_set_ok(sb))
4027 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4028 "using the ext4 subsystem");
4029 else {
4030 /*
4031 * If we're probing be silent, if this looks like
4032 * it's actually an ext[34] filesystem.
4033 */
4034 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4035 goto failed_mount;
4036 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4037 "to feature incompatibilities");
4038 goto failed_mount;
4039 }
4040 }
4041
4042 if (IS_EXT3_SB(sb)) {
4043 if (ext3_feature_set_ok(sb))
4044 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4045 "using the ext4 subsystem");
4046 else {
4047 /*
4048 * If we're probing be silent, if this looks like
4049 * it's actually an ext4 filesystem.
4050 */
4051 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4052 goto failed_mount;
4053 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4054 "to feature incompatibilities");
4055 goto failed_mount;
4056 }
4057 }
4058
4059 /*
4060 * Check feature flags regardless of the revision level, since we
4061 * previously didn't change the revision level when setting the flags,
4062 * so there is a chance incompat flags are set on a rev 0 filesystem.
4063 */
4064 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4065 goto failed_mount;
4066
4067 if (le32_to_cpu(es->s_log_block_size) >
4068 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4069 ext4_msg(sb, KERN_ERR,
4070 "Invalid log block size: %u",
4071 le32_to_cpu(es->s_log_block_size));
4072 goto failed_mount;
4073 }
4074 if (le32_to_cpu(es->s_log_cluster_size) >
4075 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4076 ext4_msg(sb, KERN_ERR,
4077 "Invalid log cluster size: %u",
4078 le32_to_cpu(es->s_log_cluster_size));
4079 goto failed_mount;
4080 }
4081
4082 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4083 ext4_msg(sb, KERN_ERR,
4084 "Number of reserved GDT blocks insanely large: %d",
4085 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4086 goto failed_mount;
4087 }
4088
4089 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
4090 if (ext4_has_feature_inline_data(sb)) {
4091 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4092 " that may contain inline data");
4093 goto failed_mount;
4094 }
4095 if (!bdev_dax_supported(sb->s_bdev, blocksize)) {
4096 ext4_msg(sb, KERN_ERR,
4097 "DAX unsupported by block device.");
4098 goto failed_mount;
4099 }
4100 }
4101
4102 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4103 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4104 es->s_encryption_level);
4105 goto failed_mount;
4106 }
4107
4108 if (sb->s_blocksize != blocksize) {
4109 /* Validate the filesystem blocksize */
4110 if (!sb_set_blocksize(sb, blocksize)) {
4111 ext4_msg(sb, KERN_ERR, "bad block size %d",
4112 blocksize);
4113 goto failed_mount;
4114 }
4115
4116 brelse(bh);
4117 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4118 offset = do_div(logical_sb_block, blocksize);
4119 bh = sb_bread_unmovable(sb, logical_sb_block);
4120 if (!bh) {
4121 ext4_msg(sb, KERN_ERR,
4122 "Can't read superblock on 2nd try");
4123 goto failed_mount;
4124 }
4125 es = (struct ext4_super_block *)(bh->b_data + offset);
4126 sbi->s_es = es;
4127 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4128 ext4_msg(sb, KERN_ERR,
4129 "Magic mismatch, very weird!");
4130 goto failed_mount;
4131 }
4132 }
4133
4134 has_huge_files = ext4_has_feature_huge_file(sb);
4135 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4136 has_huge_files);
4137 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4138
4139 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4140 if (ext4_has_feature_64bit(sb)) {
4141 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4142 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4143 !is_power_of_2(sbi->s_desc_size)) {
4144 ext4_msg(sb, KERN_ERR,
4145 "unsupported descriptor size %lu",
4146 sbi->s_desc_size);
4147 goto failed_mount;
4148 }
4149 } else
4150 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4151
4152 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4153 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4154
4155 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4156 if (sbi->s_inodes_per_block == 0)
4157 goto cantfind_ext4;
4158 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4159 sbi->s_inodes_per_group > blocksize * 8) {
4160 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4161 sbi->s_blocks_per_group);
4162 goto failed_mount;
4163 }
4164 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4165 sbi->s_inodes_per_block;
4166 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4167 sbi->s_sbh = bh;
4168 sbi->s_mount_state = le16_to_cpu(es->s_state);
4169 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4170 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4171
4172 for (i = 0; i < 4; i++)
4173 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4174 sbi->s_def_hash_version = es->s_def_hash_version;
4175 if (ext4_has_feature_dir_index(sb)) {
4176 i = le32_to_cpu(es->s_flags);
4177 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4178 sbi->s_hash_unsigned = 3;
4179 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4180#ifdef __CHAR_UNSIGNED__
4181 if (!sb_rdonly(sb))
4182 es->s_flags |=
4183 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4184 sbi->s_hash_unsigned = 3;
4185#else
4186 if (!sb_rdonly(sb))
4187 es->s_flags |=
4188 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4189#endif
4190 }
4191 }
4192
4193 /* Handle clustersize */
4194 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4195 has_bigalloc = ext4_has_feature_bigalloc(sb);
4196 if (has_bigalloc) {
4197 if (clustersize < blocksize) {
4198 ext4_msg(sb, KERN_ERR,
4199 "cluster size (%d) smaller than "
4200 "block size (%d)", clustersize, blocksize);
4201 goto failed_mount;
4202 }
4203 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4204 le32_to_cpu(es->s_log_block_size);
4205 sbi->s_clusters_per_group =
4206 le32_to_cpu(es->s_clusters_per_group);
4207 if (sbi->s_clusters_per_group > blocksize * 8) {
4208 ext4_msg(sb, KERN_ERR,
4209 "#clusters per group too big: %lu",
4210 sbi->s_clusters_per_group);
4211 goto failed_mount;
4212 }
4213 if (sbi->s_blocks_per_group !=
4214 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4215 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4216 "clusters per group (%lu) inconsistent",
4217 sbi->s_blocks_per_group,
4218 sbi->s_clusters_per_group);
4219 goto failed_mount;
4220 }
4221 } else {
4222 if (clustersize != blocksize) {
4223 ext4_msg(sb, KERN_ERR,
4224 "fragment/cluster size (%d) != "
4225 "block size (%d)", clustersize, blocksize);
4226 goto failed_mount;
4227 }
4228 if (sbi->s_blocks_per_group > blocksize * 8) {
4229 ext4_msg(sb, KERN_ERR,
4230 "#blocks per group too big: %lu",
4231 sbi->s_blocks_per_group);
4232 goto failed_mount;
4233 }
4234 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4235 sbi->s_cluster_bits = 0;
4236 }
4237 sbi->s_cluster_ratio = clustersize / blocksize;
4238
4239 /* Do we have standard group size of clustersize * 8 blocks ? */
4240 if (sbi->s_blocks_per_group == clustersize << 3)
4241 set_opt2(sb, STD_GROUP_SIZE);
4242
4243 /*
4244 * Test whether we have more sectors than will fit in sector_t,
4245 * and whether the max offset is addressable by the page cache.
4246 */
4247 err = generic_check_addressable(sb->s_blocksize_bits,
4248 ext4_blocks_count(es));
4249 if (err) {
4250 ext4_msg(sb, KERN_ERR, "filesystem"
4251 " too large to mount safely on this system");
4252 goto failed_mount;
4253 }
4254
4255 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4256 goto cantfind_ext4;
4257
4258 /* check blocks count against device size */
4259 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4260 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4261 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4262 "exceeds size of device (%llu blocks)",
4263 ext4_blocks_count(es), blocks_count);
4264 goto failed_mount;
4265 }
4266
4267 /*
4268 * It makes no sense for the first data block to be beyond the end
4269 * of the filesystem.
4270 */
4271 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4272 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4273 "block %u is beyond end of filesystem (%llu)",
4274 le32_to_cpu(es->s_first_data_block),
4275 ext4_blocks_count(es));
4276 goto failed_mount;
4277 }
4278 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4279 (sbi->s_cluster_ratio == 1)) {
4280 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4281 "block is 0 with a 1k block and cluster size");
4282 goto failed_mount;
4283 }
4284
4285 blocks_count = (ext4_blocks_count(es) -
4286 le32_to_cpu(es->s_first_data_block) +
4287 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4288 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4289 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4290 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4291 "(block count %llu, first data block %u, "
4292 "blocks per group %lu)", sbi->s_groups_count,
4293 ext4_blocks_count(es),
4294 le32_to_cpu(es->s_first_data_block),
4295 EXT4_BLOCKS_PER_GROUP(sb));
4296 goto failed_mount;
4297 }
4298 sbi->s_groups_count = blocks_count;
4299 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4300 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4301 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4302 le32_to_cpu(es->s_inodes_count)) {
4303 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4304 le32_to_cpu(es->s_inodes_count),
4305 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4306 ret = -EINVAL;
4307 goto failed_mount;
4308 }
4309 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4310 EXT4_DESC_PER_BLOCK(sb);
4311 if (ext4_has_feature_meta_bg(sb)) {
4312 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4313 ext4_msg(sb, KERN_WARNING,
4314 "first meta block group too large: %u "
4315 "(group descriptor block count %u)",
4316 le32_to_cpu(es->s_first_meta_bg), db_count);
4317 goto failed_mount;
4318 }
4319 }
4320 rcu_assign_pointer(sbi->s_group_desc,
4321 kvmalloc_array(db_count,
4322 sizeof(struct buffer_head *),
4323 GFP_KERNEL));
4324 if (sbi->s_group_desc == NULL) {
4325 ext4_msg(sb, KERN_ERR, "not enough memory");
4326 ret = -ENOMEM;
4327 goto failed_mount;
4328 }
4329
4330 bgl_lock_init(sbi->s_blockgroup_lock);
4331
4332 /* Pre-read the descriptors into the buffer cache */
4333 for (i = 0; i < db_count; i++) {
4334 block = descriptor_loc(sb, logical_sb_block, i);
4335 sb_breadahead(sb, block);
4336 }
4337
4338 for (i = 0; i < db_count; i++) {
4339 struct buffer_head *bh;
4340
4341 block = descriptor_loc(sb, logical_sb_block, i);
4342 bh = sb_bread_unmovable(sb, block);
4343 if (!bh) {
4344 ext4_msg(sb, KERN_ERR,
4345 "can't read group descriptor %d", i);
4346 db_count = i;
4347 goto failed_mount2;
4348 }
4349 rcu_read_lock();
4350 rcu_dereference(sbi->s_group_desc)[i] = bh;
4351 rcu_read_unlock();
4352 }
4353 sbi->s_gdb_count = db_count;
4354 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4355 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4356 ret = -EFSCORRUPTED;
4357 goto failed_mount2;
4358 }
4359
4360 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4361
4362 /* Register extent status tree shrinker */
4363 if (ext4_es_register_shrinker(sbi))
4364 goto failed_mount3;
4365
4366 sbi->s_stripe = ext4_get_stripe_size(sbi);
4367 sbi->s_extent_max_zeroout_kb = 32;
4368
4369 /*
4370 * set up enough so that it can read an inode
4371 */
4372 sb->s_op = &ext4_sops;
4373 sb->s_export_op = &ext4_export_ops;
4374 sb->s_xattr = ext4_xattr_handlers;
4375#ifdef CONFIG_FS_ENCRYPTION
4376 sb->s_cop = &ext4_cryptops;
4377#endif
4378#ifdef CONFIG_FS_VERITY
4379 sb->s_vop = &ext4_verityops;
4380#endif
4381#ifdef CONFIG_QUOTA
4382 sb->dq_op = &ext4_quota_operations;
4383 if (ext4_has_feature_quota(sb))
4384 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4385 else
4386 sb->s_qcop = &ext4_qctl_operations;
4387 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4388#endif
4389 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4390
4391 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4392 mutex_init(&sbi->s_orphan_lock);
4393
4394 sb->s_root = NULL;
4395
4396 needs_recovery = (es->s_last_orphan != 0 ||
4397 ext4_has_feature_journal_needs_recovery(sb));
4398
4399 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4400 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4401 goto failed_mount3a;
4402
4403 /*
4404 * The first inode we look at is the journal inode. Don't try
4405 * root first: it may be modified in the journal!
4406 */
4407 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4408 err = ext4_load_journal(sb, es, journal_devnum);
4409 if (err)
4410 goto failed_mount3a;
4411 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4412 ext4_has_feature_journal_needs_recovery(sb)) {
4413 ext4_msg(sb, KERN_ERR, "required journal recovery "
4414 "suppressed and not mounted read-only");
4415 goto failed_mount_wq;
4416 } else {
4417 /* Nojournal mode, all journal mount options are illegal */
4418 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4419 ext4_msg(sb, KERN_ERR, "can't mount with "
4420 "journal_checksum, fs mounted w/o journal");
4421 goto failed_mount_wq;
4422 }
4423 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4424 ext4_msg(sb, KERN_ERR, "can't mount with "
4425 "journal_async_commit, fs mounted w/o journal");
4426 goto failed_mount_wq;
4427 }
4428 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4429 ext4_msg(sb, KERN_ERR, "can't mount with "
4430 "commit=%lu, fs mounted w/o journal",
4431 sbi->s_commit_interval / HZ);
4432 goto failed_mount_wq;
4433 }
4434 if (EXT4_MOUNT_DATA_FLAGS &
4435 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4436 ext4_msg(sb, KERN_ERR, "can't mount with "
4437 "data=, fs mounted w/o journal");
4438 goto failed_mount_wq;
4439 }
4440 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4441 clear_opt(sb, JOURNAL_CHECKSUM);
4442 clear_opt(sb, DATA_FLAGS);
4443 sbi->s_journal = NULL;
4444 needs_recovery = 0;
4445 goto no_journal;
4446 }
4447
4448 if (ext4_has_feature_64bit(sb) &&
4449 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4450 JBD2_FEATURE_INCOMPAT_64BIT)) {
4451 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4452 goto failed_mount_wq;
4453 }
4454
4455 if (!set_journal_csum_feature_set(sb)) {
4456 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4457 "feature set");
4458 goto failed_mount_wq;
4459 }
4460
4461 /* We have now updated the journal if required, so we can
4462 * validate the data journaling mode. */
4463 switch (test_opt(sb, DATA_FLAGS)) {
4464 case 0:
4465 /* No mode set, assume a default based on the journal
4466 * capabilities: ORDERED_DATA if the journal can
4467 * cope, else JOURNAL_DATA
4468 */
4469 if (jbd2_journal_check_available_features
4470 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4471 set_opt(sb, ORDERED_DATA);
4472 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4473 } else {
4474 set_opt(sb, JOURNAL_DATA);
4475 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4476 }
4477 break;
4478
4479 case EXT4_MOUNT_ORDERED_DATA:
4480 case EXT4_MOUNT_WRITEBACK_DATA:
4481 if (!jbd2_journal_check_available_features
4482 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4483 ext4_msg(sb, KERN_ERR, "Journal does not support "
4484 "requested data journaling mode");
4485 goto failed_mount_wq;
4486 }
4487 default:
4488 break;
4489 }
4490
4491 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4492 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4493 ext4_msg(sb, KERN_ERR, "can't mount with "
4494 "journal_async_commit in data=ordered mode");
4495 goto failed_mount_wq;
4496 }
4497
4498 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4499
4500 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4501
4502no_journal:
4503 if (!test_opt(sb, NO_MBCACHE)) {
4504 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4505 if (!sbi->s_ea_block_cache) {
4506 ext4_msg(sb, KERN_ERR,
4507 "Failed to create ea_block_cache");
4508 goto failed_mount_wq;
4509 }
4510
4511 if (ext4_has_feature_ea_inode(sb)) {
4512 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4513 if (!sbi->s_ea_inode_cache) {
4514 ext4_msg(sb, KERN_ERR,
4515 "Failed to create ea_inode_cache");
4516 goto failed_mount_wq;
4517 }
4518 }
4519 }
4520
4521 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4522 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4523 goto failed_mount_wq;
4524 }
4525
4526 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4527 !ext4_has_feature_encrypt(sb)) {
4528 ext4_set_feature_encrypt(sb);
4529 ext4_commit_super(sb, 1);
4530 }
4531
4532 /*
4533 * Get the # of file system overhead blocks from the
4534 * superblock if present.
4535 */
4536 if (es->s_overhead_clusters)
4537 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4538 else {
4539 err = ext4_calculate_overhead(sb);
4540 if (err)
4541 goto failed_mount_wq;
4542 }
4543
4544 /*
4545 * The maximum number of concurrent works can be high and
4546 * concurrency isn't really necessary. Limit it to 1.
4547 */
4548 EXT4_SB(sb)->rsv_conversion_wq =
4549 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4550 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4551 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4552 ret = -ENOMEM;
4553 goto failed_mount4;
4554 }
4555
4556 /*
4557 * The jbd2_journal_load will have done any necessary log recovery,
4558 * so we can safely mount the rest of the filesystem now.
4559 */
4560
4561 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4562 if (IS_ERR(root)) {
4563 ext4_msg(sb, KERN_ERR, "get root inode failed");
4564 ret = PTR_ERR(root);
4565 root = NULL;
4566 goto failed_mount4;
4567 }
4568 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4569 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4570 iput(root);
4571 goto failed_mount4;
4572 }
4573
4574#ifdef CONFIG_UNICODE
4575 if (sbi->s_encoding)
4576 sb->s_d_op = &ext4_dentry_ops;
4577#endif
4578
4579 sb->s_root = d_make_root(root);
4580 if (!sb->s_root) {
4581 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4582 ret = -ENOMEM;
4583 goto failed_mount4;
4584 }
4585
4586 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4587 if (ret == -EROFS) {
4588 sb->s_flags |= SB_RDONLY;
4589 ret = 0;
4590 } else if (ret)
4591 goto failed_mount4a;
4592
4593 ext4_set_resv_clusters(sb);
4594
4595 err = ext4_setup_system_zone(sb);
4596 if (err) {
4597 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4598 "zone (%d)", err);
4599 goto failed_mount4a;
4600 }
4601
4602 ext4_ext_init(sb);
4603 err = ext4_mb_init(sb);
4604 if (err) {
4605 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4606 err);
4607 goto failed_mount5;
4608 }
4609
4610 block = ext4_count_free_clusters(sb);
4611 ext4_free_blocks_count_set(sbi->s_es,
4612 EXT4_C2B(sbi, block));
4613 ext4_superblock_csum_set(sb);
4614 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4615 GFP_KERNEL);
4616 if (!err) {
4617 unsigned long freei = ext4_count_free_inodes(sb);
4618 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4619 ext4_superblock_csum_set(sb);
4620 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4621 GFP_KERNEL);
4622 }
4623 if (!err)
4624 err = percpu_counter_init(&sbi->s_dirs_counter,
4625 ext4_count_dirs(sb), GFP_KERNEL);
4626 if (!err)
4627 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4628 GFP_KERNEL);
4629 if (!err)
4630 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4631
4632 if (err) {
4633 ext4_msg(sb, KERN_ERR, "insufficient memory");
4634 goto failed_mount6;
4635 }
4636
4637 if (ext4_has_feature_flex_bg(sb))
4638 if (!ext4_fill_flex_info(sb)) {
4639 ext4_msg(sb, KERN_ERR,
4640 "unable to initialize "
4641 "flex_bg meta info!");
4642 goto failed_mount6;
4643 }
4644
4645 err = ext4_register_li_request(sb, first_not_zeroed);
4646 if (err)
4647 goto failed_mount6;
4648
4649 err = ext4_register_sysfs(sb);
4650 if (err)
4651 goto failed_mount7;
4652
4653#ifdef CONFIG_QUOTA
4654 /* Enable quota usage during mount. */
4655 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4656 err = ext4_enable_quotas(sb);
4657 if (err)
4658 goto failed_mount8;
4659 }
4660#endif /* CONFIG_QUOTA */
4661
4662 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4663 ext4_orphan_cleanup(sb, es);
4664 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4665 if (needs_recovery) {
4666 ext4_msg(sb, KERN_INFO, "recovery complete");
4667 ext4_mark_recovery_complete(sb, es);
4668 }
4669 if (EXT4_SB(sb)->s_journal) {
4670 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4671 descr = " journalled data mode";
4672 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4673 descr = " ordered data mode";
4674 else
4675 descr = " writeback data mode";
4676 } else
4677 descr = "out journal";
4678
4679 if (test_opt(sb, DISCARD)) {
4680 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4681 if (!blk_queue_discard(q))
4682 ext4_msg(sb, KERN_WARNING,
4683 "mounting with \"discard\" option, but "
4684 "the device does not support discard");
4685 }
4686
4687 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4688 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4689 "Opts: %.*s%s%s", descr,
4690 (int) sizeof(sbi->s_es->s_mount_opts),
4691 sbi->s_es->s_mount_opts,
4692 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4693
4694 if (es->s_error_count)
4695 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4696
4697 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4698 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4699 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4700 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4701
4702 kfree(orig_data);
4703 return 0;
4704
4705cantfind_ext4:
4706 if (!silent)
4707 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4708 goto failed_mount;
4709
4710#ifdef CONFIG_QUOTA
4711failed_mount8:
4712 ext4_unregister_sysfs(sb);
4713#endif
4714failed_mount7:
4715 ext4_unregister_li_request(sb);
4716failed_mount6:
4717 ext4_mb_release(sb);
4718 rcu_read_lock();
4719 flex_groups = rcu_dereference(sbi->s_flex_groups);
4720 if (flex_groups) {
4721 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
4722 kvfree(flex_groups[i]);
4723 kvfree(flex_groups);
4724 }
4725 rcu_read_unlock();
4726 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4727 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4728 percpu_counter_destroy(&sbi->s_dirs_counter);
4729 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4730 percpu_free_rwsem(&sbi->s_writepages_rwsem);
4731failed_mount5:
4732 ext4_ext_release(sb);
4733 ext4_release_system_zone(sb);
4734failed_mount4a:
4735 dput(sb->s_root);
4736 sb->s_root = NULL;
4737failed_mount4:
4738 ext4_msg(sb, KERN_ERR, "mount failed");
4739 if (EXT4_SB(sb)->rsv_conversion_wq)
4740 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4741failed_mount_wq:
4742 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4743 sbi->s_ea_inode_cache = NULL;
4744
4745 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4746 sbi->s_ea_block_cache = NULL;
4747
4748 if (sbi->s_journal) {
4749 jbd2_journal_destroy(sbi->s_journal);
4750 sbi->s_journal = NULL;
4751 }
4752failed_mount3a:
4753 ext4_es_unregister_shrinker(sbi);
4754failed_mount3:
4755 del_timer_sync(&sbi->s_err_report);
4756 if (sbi->s_mmp_tsk)
4757 kthread_stop(sbi->s_mmp_tsk);
4758failed_mount2:
4759 rcu_read_lock();
4760 group_desc = rcu_dereference(sbi->s_group_desc);
4761 for (i = 0; i < db_count; i++)
4762 brelse(group_desc[i]);
4763 kvfree(group_desc);
4764 rcu_read_unlock();
4765failed_mount:
4766 if (sbi->s_chksum_driver)
4767 crypto_free_shash(sbi->s_chksum_driver);
4768
4769#ifdef CONFIG_UNICODE
4770 utf8_unload(sbi->s_encoding);
4771#endif
4772
4773#ifdef CONFIG_QUOTA
4774 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4775 kfree(get_qf_name(sb, sbi, i));
4776#endif
4777 ext4_blkdev_remove(sbi);
4778 brelse(bh);
4779out_fail:
4780 sb->s_fs_info = NULL;
4781 kfree(sbi->s_blockgroup_lock);
4782out_free_base:
4783 kfree(sbi);
4784 kfree(orig_data);
4785 fs_put_dax(dax_dev);
4786 return err ? err : ret;
4787}
4788
4789/*
4790 * Setup any per-fs journal parameters now. We'll do this both on
4791 * initial mount, once the journal has been initialised but before we've
4792 * done any recovery; and again on any subsequent remount.
4793 */
4794static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4795{
4796 struct ext4_sb_info *sbi = EXT4_SB(sb);
4797
4798 journal->j_commit_interval = sbi->s_commit_interval;
4799 journal->j_min_batch_time = sbi->s_min_batch_time;
4800 journal->j_max_batch_time = sbi->s_max_batch_time;
4801
4802 write_lock(&journal->j_state_lock);
4803 if (test_opt(sb, BARRIER))
4804 journal->j_flags |= JBD2_BARRIER;
4805 else
4806 journal->j_flags &= ~JBD2_BARRIER;
4807 if (test_opt(sb, DATA_ERR_ABORT))
4808 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4809 else
4810 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4811 write_unlock(&journal->j_state_lock);
4812}
4813
4814static struct inode *ext4_get_journal_inode(struct super_block *sb,
4815 unsigned int journal_inum)
4816{
4817 struct inode *journal_inode;
4818
4819 /*
4820 * Test for the existence of a valid inode on disk. Bad things
4821 * happen if we iget() an unused inode, as the subsequent iput()
4822 * will try to delete it.
4823 */
4824 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
4825 if (IS_ERR(journal_inode)) {
4826 ext4_msg(sb, KERN_ERR, "no journal found");
4827 return NULL;
4828 }
4829 if (!journal_inode->i_nlink) {
4830 make_bad_inode(journal_inode);
4831 iput(journal_inode);
4832 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4833 return NULL;
4834 }
4835
4836 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4837 journal_inode, journal_inode->i_size);
4838 if (!S_ISREG(journal_inode->i_mode)) {
4839 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4840 iput(journal_inode);
4841 return NULL;
4842 }
4843 return journal_inode;
4844}
4845
4846static journal_t *ext4_get_journal(struct super_block *sb,
4847 unsigned int journal_inum)
4848{
4849 struct inode *journal_inode;
4850 journal_t *journal;
4851
4852 BUG_ON(!ext4_has_feature_journal(sb));
4853
4854 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4855 if (!journal_inode)
4856 return NULL;
4857
4858 journal = jbd2_journal_init_inode(journal_inode);
4859 if (!journal) {
4860 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4861 iput(journal_inode);
4862 return NULL;
4863 }
4864 journal->j_private = sb;
4865 ext4_init_journal_params(sb, journal);
4866 return journal;
4867}
4868
4869static journal_t *ext4_get_dev_journal(struct super_block *sb,
4870 dev_t j_dev)
4871{
4872 struct buffer_head *bh;
4873 journal_t *journal;
4874 ext4_fsblk_t start;
4875 ext4_fsblk_t len;
4876 int hblock, blocksize;
4877 ext4_fsblk_t sb_block;
4878 unsigned long offset;
4879 struct ext4_super_block *es;
4880 struct block_device *bdev;
4881
4882 BUG_ON(!ext4_has_feature_journal(sb));
4883
4884 bdev = ext4_blkdev_get(j_dev, sb);
4885 if (bdev == NULL)
4886 return NULL;
4887
4888 blocksize = sb->s_blocksize;
4889 hblock = bdev_logical_block_size(bdev);
4890 if (blocksize < hblock) {
4891 ext4_msg(sb, KERN_ERR,
4892 "blocksize too small for journal device");
4893 goto out_bdev;
4894 }
4895
4896 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4897 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4898 set_blocksize(bdev, blocksize);
4899 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4900 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4901 "external journal");
4902 goto out_bdev;
4903 }
4904
4905 es = (struct ext4_super_block *) (bh->b_data + offset);
4906 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4907 !(le32_to_cpu(es->s_feature_incompat) &
4908 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4909 ext4_msg(sb, KERN_ERR, "external journal has "
4910 "bad superblock");
4911 brelse(bh);
4912 goto out_bdev;
4913 }
4914
4915 if ((le32_to_cpu(es->s_feature_ro_compat) &
4916 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4917 es->s_checksum != ext4_superblock_csum(sb, es)) {
4918 ext4_msg(sb, KERN_ERR, "external journal has "
4919 "corrupt superblock");
4920 brelse(bh);
4921 goto out_bdev;
4922 }
4923
4924 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4925 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4926 brelse(bh);
4927 goto out_bdev;
4928 }
4929
4930 len = ext4_blocks_count(es);
4931 start = sb_block + 1;
4932 brelse(bh); /* we're done with the superblock */
4933
4934 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4935 start, len, blocksize);
4936 if (!journal) {
4937 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4938 goto out_bdev;
4939 }
4940 journal->j_private = sb;
4941 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4942 wait_on_buffer(journal->j_sb_buffer);
4943 if (!buffer_uptodate(journal->j_sb_buffer)) {
4944 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4945 goto out_journal;
4946 }
4947 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4948 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4949 "user (unsupported) - %d",
4950 be32_to_cpu(journal->j_superblock->s_nr_users));
4951 goto out_journal;
4952 }
4953 EXT4_SB(sb)->journal_bdev = bdev;
4954 ext4_init_journal_params(sb, journal);
4955 return journal;
4956
4957out_journal:
4958 jbd2_journal_destroy(journal);
4959out_bdev:
4960 ext4_blkdev_put(bdev);
4961 return NULL;
4962}
4963
4964static int ext4_load_journal(struct super_block *sb,
4965 struct ext4_super_block *es,
4966 unsigned long journal_devnum)
4967{
4968 journal_t *journal;
4969 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4970 dev_t journal_dev;
4971 int err = 0;
4972 int really_read_only;
4973
4974 BUG_ON(!ext4_has_feature_journal(sb));
4975
4976 if (journal_devnum &&
4977 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4978 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4979 "numbers have changed");
4980 journal_dev = new_decode_dev(journal_devnum);
4981 } else
4982 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4983
4984 really_read_only = bdev_read_only(sb->s_bdev);
4985
4986 /*
4987 * Are we loading a blank journal or performing recovery after a
4988 * crash? For recovery, we need to check in advance whether we
4989 * can get read-write access to the device.
4990 */
4991 if (ext4_has_feature_journal_needs_recovery(sb)) {
4992 if (sb_rdonly(sb)) {
4993 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4994 "required on readonly filesystem");
4995 if (really_read_only) {
4996 ext4_msg(sb, KERN_ERR, "write access "
4997 "unavailable, cannot proceed "
4998 "(try mounting with noload)");
4999 return -EROFS;
5000 }
5001 ext4_msg(sb, KERN_INFO, "write access will "
5002 "be enabled during recovery");
5003 }
5004 }
5005
5006 if (journal_inum && journal_dev) {
5007 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
5008 "and inode journals!");
5009 return -EINVAL;
5010 }
5011
5012 if (journal_inum) {
5013 if (!(journal = ext4_get_journal(sb, journal_inum)))
5014 return -EINVAL;
5015 } else {
5016 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
5017 return -EINVAL;
5018 }
5019
5020 if (!(journal->j_flags & JBD2_BARRIER))
5021 ext4_msg(sb, KERN_INFO, "barriers disabled");
5022
5023 if (!ext4_has_feature_journal_needs_recovery(sb))
5024 err = jbd2_journal_wipe(journal, !really_read_only);
5025 if (!err) {
5026 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5027 if (save)
5028 memcpy(save, ((char *) es) +
5029 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5030 err = jbd2_journal_load(journal);
5031 if (save)
5032 memcpy(((char *) es) + EXT4_S_ERR_START,
5033 save, EXT4_S_ERR_LEN);
5034 kfree(save);
5035 }
5036
5037 if (err) {
5038 ext4_msg(sb, KERN_ERR, "error loading journal");
5039 jbd2_journal_destroy(journal);
5040 return err;
5041 }
5042
5043 EXT4_SB(sb)->s_journal = journal;
5044 ext4_clear_journal_err(sb, es);
5045
5046 if (!really_read_only && journal_devnum &&
5047 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5048 es->s_journal_dev = cpu_to_le32(journal_devnum);
5049
5050 /* Make sure we flush the recovery flag to disk. */
5051 ext4_commit_super(sb, 1);
5052 }
5053
5054 return 0;
5055}
5056
5057static int ext4_commit_super(struct super_block *sb, int sync)
5058{
5059 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5060 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5061 int error = 0;
5062
5063 if (!sbh || block_device_ejected(sb))
5064 return error;
5065
5066 /*
5067 * The superblock bh should be mapped, but it might not be if the
5068 * device was hot-removed. Not much we can do but fail the I/O.
5069 */
5070 if (!buffer_mapped(sbh))
5071 return error;
5072
5073 /*
5074 * If the file system is mounted read-only, don't update the
5075 * superblock write time. This avoids updating the superblock
5076 * write time when we are mounting the root file system
5077 * read/only but we need to replay the journal; at that point,
5078 * for people who are east of GMT and who make their clock
5079 * tick in localtime for Windows bug-for-bug compatibility,
5080 * the clock is set in the future, and this will cause e2fsck
5081 * to complain and force a full file system check.
5082 */
5083 if (!(sb->s_flags & SB_RDONLY))
5084 ext4_update_tstamp(es, s_wtime);
5085 if (sb->s_bdev->bd_part)
5086 es->s_kbytes_written =
5087 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5088 ((part_stat_read(sb->s_bdev->bd_part,
5089 sectors[STAT_WRITE]) -
5090 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5091 else
5092 es->s_kbytes_written =
5093 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5094 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5095 ext4_free_blocks_count_set(es,
5096 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5097 &EXT4_SB(sb)->s_freeclusters_counter)));
5098 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5099 es->s_free_inodes_count =
5100 cpu_to_le32(percpu_counter_sum_positive(
5101 &EXT4_SB(sb)->s_freeinodes_counter));
5102 BUFFER_TRACE(sbh, "marking dirty");
5103 ext4_superblock_csum_set(sb);
5104 if (sync)
5105 lock_buffer(sbh);
5106 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5107 /*
5108 * Oh, dear. A previous attempt to write the
5109 * superblock failed. This could happen because the
5110 * USB device was yanked out. Or it could happen to
5111 * be a transient write error and maybe the block will
5112 * be remapped. Nothing we can do but to retry the
5113 * write and hope for the best.
5114 */
5115 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5116 "superblock detected");
5117 clear_buffer_write_io_error(sbh);
5118 set_buffer_uptodate(sbh);
5119 }
5120 mark_buffer_dirty(sbh);
5121 if (sync) {
5122 unlock_buffer(sbh);
5123 error = __sync_dirty_buffer(sbh,
5124 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5125 if (buffer_write_io_error(sbh)) {
5126 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5127 "superblock");
5128 clear_buffer_write_io_error(sbh);
5129 set_buffer_uptodate(sbh);
5130 }
5131 }
5132 return error;
5133}
5134
5135/*
5136 * Have we just finished recovery? If so, and if we are mounting (or
5137 * remounting) the filesystem readonly, then we will end up with a
5138 * consistent fs on disk. Record that fact.
5139 */
5140static void ext4_mark_recovery_complete(struct super_block *sb,
5141 struct ext4_super_block *es)
5142{
5143 journal_t *journal = EXT4_SB(sb)->s_journal;
5144
5145 if (!ext4_has_feature_journal(sb)) {
5146 BUG_ON(journal != NULL);
5147 return;
5148 }
5149 jbd2_journal_lock_updates(journal);
5150 if (jbd2_journal_flush(journal) < 0)
5151 goto out;
5152
5153 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5154 ext4_clear_feature_journal_needs_recovery(sb);
5155 ext4_commit_super(sb, 1);
5156 }
5157
5158out:
5159 jbd2_journal_unlock_updates(journal);
5160}
5161
5162/*
5163 * If we are mounting (or read-write remounting) a filesystem whose journal
5164 * has recorded an error from a previous lifetime, move that error to the
5165 * main filesystem now.
5166 */
5167static void ext4_clear_journal_err(struct super_block *sb,
5168 struct ext4_super_block *es)
5169{
5170 journal_t *journal;
5171 int j_errno;
5172 const char *errstr;
5173
5174 BUG_ON(!ext4_has_feature_journal(sb));
5175
5176 journal = EXT4_SB(sb)->s_journal;
5177
5178 /*
5179 * Now check for any error status which may have been recorded in the
5180 * journal by a prior ext4_error() or ext4_abort()
5181 */
5182
5183 j_errno = jbd2_journal_errno(journal);
5184 if (j_errno) {
5185 char nbuf[16];
5186
5187 errstr = ext4_decode_error(sb, j_errno, nbuf);
5188 ext4_warning(sb, "Filesystem error recorded "
5189 "from previous mount: %s", errstr);
5190 ext4_warning(sb, "Marking fs in need of filesystem check.");
5191
5192 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5193 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5194 ext4_commit_super(sb, 1);
5195
5196 jbd2_journal_clear_err(journal);
5197 jbd2_journal_update_sb_errno(journal);
5198 }
5199}
5200
5201/*
5202 * Force the running and committing transactions to commit,
5203 * and wait on the commit.
5204 */
5205int ext4_force_commit(struct super_block *sb)
5206{
5207 journal_t *journal;
5208
5209 if (sb_rdonly(sb))
5210 return 0;
5211
5212 journal = EXT4_SB(sb)->s_journal;
5213 return ext4_journal_force_commit(journal);
5214}
5215
5216static int ext4_sync_fs(struct super_block *sb, int wait)
5217{
5218 int ret = 0;
5219 tid_t target;
5220 bool needs_barrier = false;
5221 struct ext4_sb_info *sbi = EXT4_SB(sb);
5222
5223 if (unlikely(ext4_forced_shutdown(sbi)))
5224 return 0;
5225
5226 trace_ext4_sync_fs(sb, wait);
5227 flush_workqueue(sbi->rsv_conversion_wq);
5228 /*
5229 * Writeback quota in non-journalled quota case - journalled quota has
5230 * no dirty dquots
5231 */
5232 dquot_writeback_dquots(sb, -1);
5233 /*
5234 * Data writeback is possible w/o journal transaction, so barrier must
5235 * being sent at the end of the function. But we can skip it if
5236 * transaction_commit will do it for us.
5237 */
5238 if (sbi->s_journal) {
5239 target = jbd2_get_latest_transaction(sbi->s_journal);
5240 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5241 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5242 needs_barrier = true;
5243
5244 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5245 if (wait)
5246 ret = jbd2_log_wait_commit(sbi->s_journal,
5247 target);
5248 }
5249 } else if (wait && test_opt(sb, BARRIER))
5250 needs_barrier = true;
5251 if (needs_barrier) {
5252 int err;
5253 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
5254 if (!ret)
5255 ret = err;
5256 }
5257
5258 return ret;
5259}
5260
5261/*
5262 * LVM calls this function before a (read-only) snapshot is created. This
5263 * gives us a chance to flush the journal completely and mark the fs clean.
5264 *
5265 * Note that only this function cannot bring a filesystem to be in a clean
5266 * state independently. It relies on upper layer to stop all data & metadata
5267 * modifications.
5268 */
5269static int ext4_freeze(struct super_block *sb)
5270{
5271 int error = 0;
5272 journal_t *journal;
5273
5274 if (sb_rdonly(sb))
5275 return 0;
5276
5277 journal = EXT4_SB(sb)->s_journal;
5278
5279 if (journal) {
5280 /* Now we set up the journal barrier. */
5281 jbd2_journal_lock_updates(journal);
5282
5283 /*
5284 * Don't clear the needs_recovery flag if we failed to
5285 * flush the journal.
5286 */
5287 error = jbd2_journal_flush(journal);
5288 if (error < 0)
5289 goto out;
5290
5291 /* Journal blocked and flushed, clear needs_recovery flag. */
5292 ext4_clear_feature_journal_needs_recovery(sb);
5293 }
5294
5295 error = ext4_commit_super(sb, 1);
5296out:
5297 if (journal)
5298 /* we rely on upper layer to stop further updates */
5299 jbd2_journal_unlock_updates(journal);
5300 return error;
5301}
5302
5303/*
5304 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5305 * flag here, even though the filesystem is not technically dirty yet.
5306 */
5307static int ext4_unfreeze(struct super_block *sb)
5308{
5309 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5310 return 0;
5311
5312 if (EXT4_SB(sb)->s_journal) {
5313 /* Reset the needs_recovery flag before the fs is unlocked. */
5314 ext4_set_feature_journal_needs_recovery(sb);
5315 }
5316
5317 ext4_commit_super(sb, 1);
5318 return 0;
5319}
5320
5321/*
5322 * Structure to save mount options for ext4_remount's benefit
5323 */
5324struct ext4_mount_options {
5325 unsigned long s_mount_opt;
5326 unsigned long s_mount_opt2;
5327 kuid_t s_resuid;
5328 kgid_t s_resgid;
5329 unsigned long s_commit_interval;
5330 u32 s_min_batch_time, s_max_batch_time;
5331#ifdef CONFIG_QUOTA
5332 int s_jquota_fmt;
5333 char *s_qf_names[EXT4_MAXQUOTAS];
5334#endif
5335};
5336
5337static int ext4_remount(struct super_block *sb, int *flags, char *data)
5338{
5339 struct ext4_super_block *es;
5340 struct ext4_sb_info *sbi = EXT4_SB(sb);
5341 unsigned long old_sb_flags;
5342 struct ext4_mount_options old_opts;
5343 int enable_quota = 0;
5344 ext4_group_t g;
5345 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5346 int err = 0;
5347#ifdef CONFIG_QUOTA
5348 int i, j;
5349 char *to_free[EXT4_MAXQUOTAS];
5350#endif
5351 char *orig_data = kstrdup(data, GFP_KERNEL);
5352
5353 if (data && !orig_data)
5354 return -ENOMEM;
5355
5356 /* Store the original options */
5357 old_sb_flags = sb->s_flags;
5358 old_opts.s_mount_opt = sbi->s_mount_opt;
5359 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5360 old_opts.s_resuid = sbi->s_resuid;
5361 old_opts.s_resgid = sbi->s_resgid;
5362 old_opts.s_commit_interval = sbi->s_commit_interval;
5363 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5364 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5365#ifdef CONFIG_QUOTA
5366 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5367 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5368 if (sbi->s_qf_names[i]) {
5369 char *qf_name = get_qf_name(sb, sbi, i);
5370
5371 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5372 if (!old_opts.s_qf_names[i]) {
5373 for (j = 0; j < i; j++)
5374 kfree(old_opts.s_qf_names[j]);
5375 kfree(orig_data);
5376 return -ENOMEM;
5377 }
5378 } else
5379 old_opts.s_qf_names[i] = NULL;
5380#endif
5381 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5382 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5383
5384 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5385 err = -EINVAL;
5386 goto restore_opts;
5387 }
5388
5389 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5390 test_opt(sb, JOURNAL_CHECKSUM)) {
5391 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5392 "during remount not supported; ignoring");
5393 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5394 }
5395
5396 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5397 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5398 ext4_msg(sb, KERN_ERR, "can't mount with "
5399 "both data=journal and delalloc");
5400 err = -EINVAL;
5401 goto restore_opts;
5402 }
5403 if (test_opt(sb, DIOREAD_NOLOCK)) {
5404 ext4_msg(sb, KERN_ERR, "can't mount with "
5405 "both data=journal and dioread_nolock");
5406 err = -EINVAL;
5407 goto restore_opts;
5408 }
5409 if (test_opt(sb, DAX)) {
5410 ext4_msg(sb, KERN_ERR, "can't mount with "
5411 "both data=journal and dax");
5412 err = -EINVAL;
5413 goto restore_opts;
5414 }
5415 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5416 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5417 ext4_msg(sb, KERN_ERR, "can't mount with "
5418 "journal_async_commit in data=ordered mode");
5419 err = -EINVAL;
5420 goto restore_opts;
5421 }
5422 }
5423
5424 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5425 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5426 err = -EINVAL;
5427 goto restore_opts;
5428 }
5429
5430 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5431 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5432 "dax flag with busy inodes while remounting");
5433 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5434 }
5435
5436 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5437 ext4_abort(sb, "Abort forced by user");
5438
5439 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5440 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5441
5442 es = sbi->s_es;
5443
5444 if (sbi->s_journal) {
5445 ext4_init_journal_params(sb, sbi->s_journal);
5446 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5447 }
5448
5449 if (*flags & SB_LAZYTIME)
5450 sb->s_flags |= SB_LAZYTIME;
5451
5452 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5453 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5454 err = -EROFS;
5455 goto restore_opts;
5456 }
5457
5458 if (*flags & SB_RDONLY) {
5459 err = sync_filesystem(sb);
5460 if (err < 0)
5461 goto restore_opts;
5462 err = dquot_suspend(sb, -1);
5463 if (err < 0)
5464 goto restore_opts;
5465
5466 /*
5467 * First of all, the unconditional stuff we have to do
5468 * to disable replay of the journal when we next remount
5469 */
5470 sb->s_flags |= SB_RDONLY;
5471
5472 /*
5473 * OK, test if we are remounting a valid rw partition
5474 * readonly, and if so set the rdonly flag and then
5475 * mark the partition as valid again.
5476 */
5477 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5478 (sbi->s_mount_state & EXT4_VALID_FS))
5479 es->s_state = cpu_to_le16(sbi->s_mount_state);
5480
5481 if (sbi->s_journal)
5482 ext4_mark_recovery_complete(sb, es);
5483 if (sbi->s_mmp_tsk)
5484 kthread_stop(sbi->s_mmp_tsk);
5485 } else {
5486 /* Make sure we can mount this feature set readwrite */
5487 if (ext4_has_feature_readonly(sb) ||
5488 !ext4_feature_set_ok(sb, 0)) {
5489 err = -EROFS;
5490 goto restore_opts;
5491 }
5492 /*
5493 * Make sure the group descriptor checksums
5494 * are sane. If they aren't, refuse to remount r/w.
5495 */
5496 for (g = 0; g < sbi->s_groups_count; g++) {
5497 struct ext4_group_desc *gdp =
5498 ext4_get_group_desc(sb, g, NULL);
5499
5500 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5501 ext4_msg(sb, KERN_ERR,
5502 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5503 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5504 le16_to_cpu(gdp->bg_checksum));
5505 err = -EFSBADCRC;
5506 goto restore_opts;
5507 }
5508 }
5509
5510 /*
5511 * If we have an unprocessed orphan list hanging
5512 * around from a previously readonly bdev mount,
5513 * require a full umount/remount for now.
5514 */
5515 if (es->s_last_orphan) {
5516 ext4_msg(sb, KERN_WARNING, "Couldn't "
5517 "remount RDWR because of unprocessed "
5518 "orphan inode list. Please "
5519 "umount/remount instead");
5520 err = -EINVAL;
5521 goto restore_opts;
5522 }
5523
5524 /*
5525 * Mounting a RDONLY partition read-write, so reread
5526 * and store the current valid flag. (It may have
5527 * been changed by e2fsck since we originally mounted
5528 * the partition.)
5529 */
5530 if (sbi->s_journal)
5531 ext4_clear_journal_err(sb, es);
5532 sbi->s_mount_state = le16_to_cpu(es->s_state);
5533
5534 err = ext4_setup_super(sb, es, 0);
5535 if (err)
5536 goto restore_opts;
5537
5538 sb->s_flags &= ~SB_RDONLY;
5539 if (ext4_has_feature_mmp(sb))
5540 if (ext4_multi_mount_protect(sb,
5541 le64_to_cpu(es->s_mmp_block))) {
5542 err = -EROFS;
5543 goto restore_opts;
5544 }
5545 enable_quota = 1;
5546 }
5547 }
5548
5549 /*
5550 * Reinitialize lazy itable initialization thread based on
5551 * current settings
5552 */
5553 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5554 ext4_unregister_li_request(sb);
5555 else {
5556 ext4_group_t first_not_zeroed;
5557 first_not_zeroed = ext4_has_uninit_itable(sb);
5558 ext4_register_li_request(sb, first_not_zeroed);
5559 }
5560
5561 ext4_setup_system_zone(sb);
5562 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5563 err = ext4_commit_super(sb, 1);
5564 if (err)
5565 goto restore_opts;
5566 }
5567
5568#ifdef CONFIG_QUOTA
5569 /* Release old quota file names */
5570 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5571 kfree(old_opts.s_qf_names[i]);
5572 if (enable_quota) {
5573 if (sb_any_quota_suspended(sb))
5574 dquot_resume(sb, -1);
5575 else if (ext4_has_feature_quota(sb)) {
5576 err = ext4_enable_quotas(sb);
5577 if (err)
5578 goto restore_opts;
5579 }
5580 }
5581#endif
5582
5583 *flags = (*flags & ~SB_LAZYTIME) | (sb->s_flags & SB_LAZYTIME);
5584 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5585 kfree(orig_data);
5586 return 0;
5587
5588restore_opts:
5589 sb->s_flags = old_sb_flags;
5590 sbi->s_mount_opt = old_opts.s_mount_opt;
5591 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5592 sbi->s_resuid = old_opts.s_resuid;
5593 sbi->s_resgid = old_opts.s_resgid;
5594 sbi->s_commit_interval = old_opts.s_commit_interval;
5595 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5596 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5597#ifdef CONFIG_QUOTA
5598 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5599 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5600 to_free[i] = get_qf_name(sb, sbi, i);
5601 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
5602 }
5603 synchronize_rcu();
5604 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5605 kfree(to_free[i]);
5606#endif
5607 kfree(orig_data);
5608 return err;
5609}
5610
5611#ifdef CONFIG_QUOTA
5612static int ext4_statfs_project(struct super_block *sb,
5613 kprojid_t projid, struct kstatfs *buf)
5614{
5615 struct kqid qid;
5616 struct dquot *dquot;
5617 u64 limit;
5618 u64 curblock;
5619
5620 qid = make_kqid_projid(projid);
5621 dquot = dqget(sb, qid);
5622 if (IS_ERR(dquot))
5623 return PTR_ERR(dquot);
5624 spin_lock(&dquot->dq_dqb_lock);
5625
5626 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
5627 dquot->dq_dqb.dqb_bhardlimit);
5628 limit >>= sb->s_blocksize_bits;
5629
5630 if (limit && buf->f_blocks > limit) {
5631 curblock = (dquot->dq_dqb.dqb_curspace +
5632 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
5633 buf->f_blocks = limit;
5634 buf->f_bfree = buf->f_bavail =
5635 (buf->f_blocks > curblock) ?
5636 (buf->f_blocks - curblock) : 0;
5637 }
5638
5639 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
5640 dquot->dq_dqb.dqb_ihardlimit);
5641 if (limit && buf->f_files > limit) {
5642 buf->f_files = limit;
5643 buf->f_ffree =
5644 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5645 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5646 }
5647
5648 spin_unlock(&dquot->dq_dqb_lock);
5649 dqput(dquot);
5650 return 0;
5651}
5652#endif
5653
5654static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5655{
5656 struct super_block *sb = dentry->d_sb;
5657 struct ext4_sb_info *sbi = EXT4_SB(sb);
5658 struct ext4_super_block *es = sbi->s_es;
5659 ext4_fsblk_t overhead = 0, resv_blocks;
5660 u64 fsid;
5661 s64 bfree;
5662 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5663
5664 if (!test_opt(sb, MINIX_DF))
5665 overhead = sbi->s_overhead;
5666
5667 buf->f_type = EXT4_SUPER_MAGIC;
5668 buf->f_bsize = sb->s_blocksize;
5669 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5670 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5671 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5672 /* prevent underflow in case that few free space is available */
5673 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5674 buf->f_bavail = buf->f_bfree -
5675 (ext4_r_blocks_count(es) + resv_blocks);
5676 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5677 buf->f_bavail = 0;
5678 buf->f_files = le32_to_cpu(es->s_inodes_count);
5679 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5680 buf->f_namelen = EXT4_NAME_LEN;
5681 fsid = le64_to_cpup((void *)es->s_uuid) ^
5682 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5683 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5684 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5685
5686#ifdef CONFIG_QUOTA
5687 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5688 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5689 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5690#endif
5691 return 0;
5692}
5693
5694
5695#ifdef CONFIG_QUOTA
5696
5697/*
5698 * Helper functions so that transaction is started before we acquire dqio_sem
5699 * to keep correct lock ordering of transaction > dqio_sem
5700 */
5701static inline struct inode *dquot_to_inode(struct dquot *dquot)
5702{
5703 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5704}
5705
5706static int ext4_write_dquot(struct dquot *dquot)
5707{
5708 int ret, err;
5709 handle_t *handle;
5710 struct inode *inode;
5711
5712 inode = dquot_to_inode(dquot);
5713 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5714 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5715 if (IS_ERR(handle))
5716 return PTR_ERR(handle);
5717 ret = dquot_commit(dquot);
5718 err = ext4_journal_stop(handle);
5719 if (!ret)
5720 ret = err;
5721 return ret;
5722}
5723
5724static int ext4_acquire_dquot(struct dquot *dquot)
5725{
5726 int ret, err;
5727 handle_t *handle;
5728
5729 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5730 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5731 if (IS_ERR(handle))
5732 return PTR_ERR(handle);
5733 ret = dquot_acquire(dquot);
5734 err = ext4_journal_stop(handle);
5735 if (!ret)
5736 ret = err;
5737 return ret;
5738}
5739
5740static int ext4_release_dquot(struct dquot *dquot)
5741{
5742 int ret, err;
5743 handle_t *handle;
5744
5745 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5746 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5747 if (IS_ERR(handle)) {
5748 /* Release dquot anyway to avoid endless cycle in dqput() */
5749 dquot_release(dquot);
5750 return PTR_ERR(handle);
5751 }
5752 ret = dquot_release(dquot);
5753 err = ext4_journal_stop(handle);
5754 if (!ret)
5755 ret = err;
5756 return ret;
5757}
5758
5759static int ext4_mark_dquot_dirty(struct dquot *dquot)
5760{
5761 struct super_block *sb = dquot->dq_sb;
5762 struct ext4_sb_info *sbi = EXT4_SB(sb);
5763
5764 /* Are we journaling quotas? */
5765 if (ext4_has_feature_quota(sb) ||
5766 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5767 dquot_mark_dquot_dirty(dquot);
5768 return ext4_write_dquot(dquot);
5769 } else {
5770 return dquot_mark_dquot_dirty(dquot);
5771 }
5772}
5773
5774static int ext4_write_info(struct super_block *sb, int type)
5775{
5776 int ret, err;
5777 handle_t *handle;
5778
5779 /* Data block + inode block */
5780 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5781 if (IS_ERR(handle))
5782 return PTR_ERR(handle);
5783 ret = dquot_commit_info(sb, type);
5784 err = ext4_journal_stop(handle);
5785 if (!ret)
5786 ret = err;
5787 return ret;
5788}
5789
5790/*
5791 * Turn on quotas during mount time - we need to find
5792 * the quota file and such...
5793 */
5794static int ext4_quota_on_mount(struct super_block *sb, int type)
5795{
5796 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
5797 EXT4_SB(sb)->s_jquota_fmt, type);
5798}
5799
5800static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5801{
5802 struct ext4_inode_info *ei = EXT4_I(inode);
5803
5804 /* The first argument of lockdep_set_subclass has to be
5805 * *exactly* the same as the argument to init_rwsem() --- in
5806 * this case, in init_once() --- or lockdep gets unhappy
5807 * because the name of the lock is set using the
5808 * stringification of the argument to init_rwsem().
5809 */
5810 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5811 lockdep_set_subclass(&ei->i_data_sem, subclass);
5812}
5813
5814/*
5815 * Standard function to be called on quota_on
5816 */
5817static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5818 const struct path *path)
5819{
5820 int err;
5821
5822 if (!test_opt(sb, QUOTA))
5823 return -EINVAL;
5824
5825 /* Quotafile not on the same filesystem? */
5826 if (path->dentry->d_sb != sb)
5827 return -EXDEV;
5828 /* Journaling quota? */
5829 if (EXT4_SB(sb)->s_qf_names[type]) {
5830 /* Quotafile not in fs root? */
5831 if (path->dentry->d_parent != sb->s_root)
5832 ext4_msg(sb, KERN_WARNING,
5833 "Quota file not on filesystem root. "
5834 "Journaled quota will not work");
5835 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
5836 } else {
5837 /*
5838 * Clear the flag just in case mount options changed since
5839 * last time.
5840 */
5841 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
5842 }
5843
5844 /*
5845 * When we journal data on quota file, we have to flush journal to see
5846 * all updates to the file when we bypass pagecache...
5847 */
5848 if (EXT4_SB(sb)->s_journal &&
5849 ext4_should_journal_data(d_inode(path->dentry))) {
5850 /*
5851 * We don't need to lock updates but journal_flush() could
5852 * otherwise be livelocked...
5853 */
5854 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5855 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5856 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5857 if (err)
5858 return err;
5859 }
5860
5861 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5862 err = dquot_quota_on(sb, type, format_id, path);
5863 if (err) {
5864 lockdep_set_quota_inode(path->dentry->d_inode,
5865 I_DATA_SEM_NORMAL);
5866 } else {
5867 struct inode *inode = d_inode(path->dentry);
5868 handle_t *handle;
5869
5870 /*
5871 * Set inode flags to prevent userspace from messing with quota
5872 * files. If this fails, we return success anyway since quotas
5873 * are already enabled and this is not a hard failure.
5874 */
5875 inode_lock(inode);
5876 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5877 if (IS_ERR(handle))
5878 goto unlock_inode;
5879 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5880 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5881 S_NOATIME | S_IMMUTABLE);
5882 ext4_mark_inode_dirty(handle, inode);
5883 ext4_journal_stop(handle);
5884 unlock_inode:
5885 inode_unlock(inode);
5886 }
5887 return err;
5888}
5889
5890static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5891 unsigned int flags)
5892{
5893 int err;
5894 struct inode *qf_inode;
5895 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5896 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5897 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5898 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5899 };
5900
5901 BUG_ON(!ext4_has_feature_quota(sb));
5902
5903 if (!qf_inums[type])
5904 return -EPERM;
5905
5906 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
5907 if (IS_ERR(qf_inode)) {
5908 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5909 return PTR_ERR(qf_inode);
5910 }
5911
5912 /* Don't account quota for quota files to avoid recursion */
5913 qf_inode->i_flags |= S_NOQUOTA;
5914 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5915 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
5916 if (err)
5917 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5918 iput(qf_inode);
5919
5920 return err;
5921}
5922
5923/* Enable usage tracking for all quota types. */
5924static int ext4_enable_quotas(struct super_block *sb)
5925{
5926 int type, err = 0;
5927 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5928 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5929 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5930 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5931 };
5932 bool quota_mopt[EXT4_MAXQUOTAS] = {
5933 test_opt(sb, USRQUOTA),
5934 test_opt(sb, GRPQUOTA),
5935 test_opt(sb, PRJQUOTA),
5936 };
5937
5938 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
5939 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5940 if (qf_inums[type]) {
5941 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5942 DQUOT_USAGE_ENABLED |
5943 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5944 if (err) {
5945 ext4_warning(sb,
5946 "Failed to enable quota tracking "
5947 "(type=%d, err=%d). Please run "
5948 "e2fsck to fix.", type, err);
5949 for (type--; type >= 0; type--)
5950 dquot_quota_off(sb, type);
5951
5952 return err;
5953 }
5954 }
5955 }
5956 return 0;
5957}
5958
5959static int ext4_quota_off(struct super_block *sb, int type)
5960{
5961 struct inode *inode = sb_dqopt(sb)->files[type];
5962 handle_t *handle;
5963 int err;
5964
5965 /* Force all delayed allocation blocks to be allocated.
5966 * Caller already holds s_umount sem */
5967 if (test_opt(sb, DELALLOC))
5968 sync_filesystem(sb);
5969
5970 if (!inode || !igrab(inode))
5971 goto out;
5972
5973 err = dquot_quota_off(sb, type);
5974 if (err || ext4_has_feature_quota(sb))
5975 goto out_put;
5976
5977 inode_lock(inode);
5978 /*
5979 * Update modification times of quota files when userspace can
5980 * start looking at them. If we fail, we return success anyway since
5981 * this is not a hard failure and quotas are already disabled.
5982 */
5983 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5984 if (IS_ERR(handle))
5985 goto out_unlock;
5986 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5987 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5988 inode->i_mtime = inode->i_ctime = current_time(inode);
5989 ext4_mark_inode_dirty(handle, inode);
5990 ext4_journal_stop(handle);
5991out_unlock:
5992 inode_unlock(inode);
5993out_put:
5994 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5995 iput(inode);
5996 return err;
5997out:
5998 return dquot_quota_off(sb, type);
5999}
6000
6001/* Read data from quotafile - avoid pagecache and such because we cannot afford
6002 * acquiring the locks... As quota files are never truncated and quota code
6003 * itself serializes the operations (and no one else should touch the files)
6004 * we don't have to be afraid of races */
6005static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6006 size_t len, loff_t off)
6007{
6008 struct inode *inode = sb_dqopt(sb)->files[type];
6009 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6010 int offset = off & (sb->s_blocksize - 1);
6011 int tocopy;
6012 size_t toread;
6013 struct buffer_head *bh;
6014 loff_t i_size = i_size_read(inode);
6015
6016 if (off > i_size)
6017 return 0;
6018 if (off+len > i_size)
6019 len = i_size-off;
6020 toread = len;
6021 while (toread > 0) {
6022 tocopy = sb->s_blocksize - offset < toread ?
6023 sb->s_blocksize - offset : toread;
6024 bh = ext4_bread(NULL, inode, blk, 0);
6025 if (IS_ERR(bh))
6026 return PTR_ERR(bh);
6027 if (!bh) /* A hole? */
6028 memset(data, 0, tocopy);
6029 else
6030 memcpy(data, bh->b_data+offset, tocopy);
6031 brelse(bh);
6032 offset = 0;
6033 toread -= tocopy;
6034 data += tocopy;
6035 blk++;
6036 }
6037 return len;
6038}
6039
6040/* Write to quotafile (we know the transaction is already started and has
6041 * enough credits) */
6042static ssize_t ext4_quota_write(struct super_block *sb, int type,
6043 const char *data, size_t len, loff_t off)
6044{
6045 struct inode *inode = sb_dqopt(sb)->files[type];
6046 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6047 int err, offset = off & (sb->s_blocksize - 1);
6048 int retries = 0;
6049 struct buffer_head *bh;
6050 handle_t *handle = journal_current_handle();
6051
6052 if (EXT4_SB(sb)->s_journal && !handle) {
6053 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6054 " cancelled because transaction is not started",
6055 (unsigned long long)off, (unsigned long long)len);
6056 return -EIO;
6057 }
6058 /*
6059 * Since we account only one data block in transaction credits,
6060 * then it is impossible to cross a block boundary.
6061 */
6062 if (sb->s_blocksize - offset < len) {
6063 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6064 " cancelled because not block aligned",
6065 (unsigned long long)off, (unsigned long long)len);
6066 return -EIO;
6067 }
6068
6069 do {
6070 bh = ext4_bread(handle, inode, blk,
6071 EXT4_GET_BLOCKS_CREATE |
6072 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6073 } while (PTR_ERR(bh) == -ENOSPC &&
6074 ext4_should_retry_alloc(inode->i_sb, &retries));
6075 if (IS_ERR(bh))
6076 return PTR_ERR(bh);
6077 if (!bh)
6078 goto out;
6079 BUFFER_TRACE(bh, "get write access");
6080 err = ext4_journal_get_write_access(handle, bh);
6081 if (err) {
6082 brelse(bh);
6083 return err;
6084 }
6085 lock_buffer(bh);
6086 memcpy(bh->b_data+offset, data, len);
6087 flush_dcache_page(bh->b_page);
6088 unlock_buffer(bh);
6089 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6090 brelse(bh);
6091out:
6092 if (inode->i_size < off + len) {
6093 i_size_write(inode, off + len);
6094 EXT4_I(inode)->i_disksize = inode->i_size;
6095 ext4_mark_inode_dirty(handle, inode);
6096 }
6097 return len;
6098}
6099#endif
6100
6101static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6102 const char *dev_name, void *data)
6103{
6104 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6105}
6106
6107#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6108static inline void register_as_ext2(void)
6109{
6110 int err = register_filesystem(&ext2_fs_type);
6111 if (err)
6112 printk(KERN_WARNING
6113 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6114}
6115
6116static inline void unregister_as_ext2(void)
6117{
6118 unregister_filesystem(&ext2_fs_type);
6119}
6120
6121static inline int ext2_feature_set_ok(struct super_block *sb)
6122{
6123 if (ext4_has_unknown_ext2_incompat_features(sb))
6124 return 0;
6125 if (sb_rdonly(sb))
6126 return 1;
6127 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6128 return 0;
6129 return 1;
6130}
6131#else
6132static inline void register_as_ext2(void) { }
6133static inline void unregister_as_ext2(void) { }
6134static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6135#endif
6136
6137static inline void register_as_ext3(void)
6138{
6139 int err = register_filesystem(&ext3_fs_type);
6140 if (err)
6141 printk(KERN_WARNING
6142 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6143}
6144
6145static inline void unregister_as_ext3(void)
6146{
6147 unregister_filesystem(&ext3_fs_type);
6148}
6149
6150static inline int ext3_feature_set_ok(struct super_block *sb)
6151{
6152 if (ext4_has_unknown_ext3_incompat_features(sb))
6153 return 0;
6154 if (!ext4_has_feature_journal(sb))
6155 return 0;
6156 if (sb_rdonly(sb))
6157 return 1;
6158 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6159 return 0;
6160 return 1;
6161}
6162
6163static struct file_system_type ext4_fs_type = {
6164 .owner = THIS_MODULE,
6165 .name = "ext4",
6166 .mount = ext4_mount,
6167 .kill_sb = kill_block_super,
6168 .fs_flags = FS_REQUIRES_DEV,
6169};
6170MODULE_ALIAS_FS("ext4");
6171
6172/* Shared across all ext4 file systems */
6173wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6174
6175static int __init ext4_init_fs(void)
6176{
6177 int i, err;
6178
6179 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6180 ext4_li_info = NULL;
6181 mutex_init(&ext4_li_mtx);
6182
6183 /* Build-time check for flags consistency */
6184 ext4_check_flag_values();
6185
6186 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6187 init_waitqueue_head(&ext4__ioend_wq[i]);
6188
6189 err = ext4_init_es();
6190 if (err)
6191 return err;
6192
6193 err = ext4_init_pending();
6194 if (err)
6195 goto out7;
6196
6197 err = ext4_init_post_read_processing();
6198 if (err)
6199 goto out6;
6200
6201 err = ext4_init_pageio();
6202 if (err)
6203 goto out5;
6204
6205 err = ext4_init_system_zone();
6206 if (err)
6207 goto out4;
6208
6209 err = ext4_init_sysfs();
6210 if (err)
6211 goto out3;
6212
6213 err = ext4_init_mballoc();
6214 if (err)
6215 goto out2;
6216 err = init_inodecache();
6217 if (err)
6218 goto out1;
6219 register_as_ext3();
6220 register_as_ext2();
6221 err = register_filesystem(&ext4_fs_type);
6222 if (err)
6223 goto out;
6224
6225 return 0;
6226out:
6227 unregister_as_ext2();
6228 unregister_as_ext3();
6229 destroy_inodecache();
6230out1:
6231 ext4_exit_mballoc();
6232out2:
6233 ext4_exit_sysfs();
6234out3:
6235 ext4_exit_system_zone();
6236out4:
6237 ext4_exit_pageio();
6238out5:
6239 ext4_exit_post_read_processing();
6240out6:
6241 ext4_exit_pending();
6242out7:
6243 ext4_exit_es();
6244
6245 return err;
6246}
6247
6248static void __exit ext4_exit_fs(void)
6249{
6250 ext4_destroy_lazyinit_thread();
6251 unregister_as_ext2();
6252 unregister_as_ext3();
6253 unregister_filesystem(&ext4_fs_type);
6254 destroy_inodecache();
6255 ext4_exit_mballoc();
6256 ext4_exit_sysfs();
6257 ext4_exit_system_zone();
6258 ext4_exit_pageio();
6259 ext4_exit_post_read_processing();
6260 ext4_exit_es();
6261 ext4_exit_pending();
6262}
6263
6264MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6265MODULE_DESCRIPTION("Fourth Extended Filesystem");
6266MODULE_LICENSE("GPL");
6267MODULE_SOFTDEP("pre: crc32c");
6268module_init(ext4_init_fs)
6269module_exit(ext4_exit_fs)
Linux 6.x block layer and io_uring tree(s)