projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: fix potential error in ext4_do_update_inode
[linux-2.6-block.git] / fs / ext4 / super.c
CommitLineData
f5166768 1// SPDX-License-Identifier: GPL-2.0
ac27a0ec 2/*
617ba13b 3 * linux/fs/ext4/super.c
ac27a0ec
DK		 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>
c5ca7c76 24#include <linux/vmalloc.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
66114cad 28#include <linux/backing-dev.h>
ac27a0ec 29#include <linux/parser.h>
ac27a0ec 30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 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>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ef510424 41#include <linux/dax.h>
7abc52c2 42#include <linux/cleancache.h>
7c0f6ba6 43#include <linux/uaccess.h>
ee73f9a5 44#include <linux/iversion.h>
c83ad55e 45#include <linux/unicode.h>
c6a564ff 46#include <linux/part_stat.h>
bfff6873
LC		 47#include <linux/kthread.h>
48#include <linux/freezer.h>
49
3dcf5451 50#include "ext4.h"
4a092d73 51#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 52#include "ext4_jbd2.h"
ac27a0ec
DK		 53#include "xattr.h"
54#include "acl.h"
3661d286 55#include "mballoc.h"
0c9ec4be 56#include "fsmap.h"
ac27a0ec 57
9bffad1e
TT		 58#define CREATE_TRACE_POINTS
59#include <trace/events/ext4.h>
60
0b75a840 61static struct ext4_lazy_init *ext4_li_info;
59ebc7fd 62static DEFINE_MUTEX(ext4_li_mtx);
e294a537 63static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 64
617ba13b 65static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 66 unsigned long journal_devnum);
2adf6da8 67static int ext4_show_options(struct seq_file *seq, struct dentry *root);
2d01ddc8 68static void ext4_update_super(struct super_block *sb);
4392fbc4 69static int ext4_commit_super(struct super_block *sb);
11215630 70static int ext4_mark_recovery_complete(struct super_block *sb,
2b2d6d01 71 struct ext4_super_block *es);
11215630
JK		 72static int ext4_clear_journal_err(struct super_block *sb,
73 struct ext4_super_block *es);
617ba13b 74static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01
TT		 75static int ext4_remount(struct super_block *sb, int *flags, char *data);
76static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 77static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 78static int ext4_freeze(struct super_block *sb);
152a0836
AV		 79static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 const char *dev_name, void *data);
2035e776
TT		 81static inline int ext2_feature_set_ok(struct super_block *sb);
82static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 83static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 84static void ext4_destroy_lazyinit_thread(void);
85static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 86static void ext4_clear_request_list(void);
c6cb7e77
EW		 87static struct inode *ext4_get_journal_inode(struct super_block *sb,
88 unsigned int journal_inum);
ac27a0ec 89
e74031fd
JK		 90/*
91 * Lock ordering
92 *
93 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
94 * i_mmap_rwsem (inode->i_mmap_rwsem)!
95 *
96 * page fault path:
c1e8d7c6 97 * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
e74031fd
JK		 98 * page lock -> i_data_sem (rw)
99 *
100 * buffered write path:
c1e8d7c6 101 * sb_start_write -> i_mutex -> mmap_lock
e74031fd
JK		 102 * sb_start_write -> i_mutex -> transaction start -> page lock ->
103 * i_data_sem (rw)
104 *
105 * truncate:
1d39834f
NB		 106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
107 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
108 * i_data_sem (rw)
e74031fd
JK		 109 *
110 * direct IO:
c1e8d7c6 111 * sb_start_write -> i_mutex -> mmap_lock
1d39834f 112 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
e74031fd
JK		 113 *
114 * writepages:
115 * transaction start -> page lock(s) -> i_data_sem (rw)
116 */
117
c290ea01 118#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776
TT		 119static struct file_system_type ext2_fs_type = {
120 .owner = THIS_MODULE,
121 .name = "ext2",
122 .mount = ext4_mount,
123 .kill_sb = kill_block_super,
124 .fs_flags = FS_REQUIRES_DEV,
125};
7f78e035 126MODULE_ALIAS_FS("ext2");
fa7614dd 127MODULE_ALIAS("ext2");
2035e776
TT		 128#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
129#else
130#define IS_EXT2_SB(sb) (0)
131#endif
132
133
ba69f9ab
JK		 134static struct file_system_type ext3_fs_type = {
135 .owner = THIS_MODULE,
136 .name = "ext3",
152a0836 137 .mount = ext4_mount,
ba69f9ab
JK		 138 .kill_sb = kill_block_super,
139 .fs_flags = FS_REQUIRES_DEV,
140};
7f78e035 141MODULE_ALIAS_FS("ext3");
fa7614dd 142MODULE_ALIAS("ext3");
ba69f9ab 143#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 144
fa491b14 145
146static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
147 bh_end_io_t *end_io)
148{
149 /*
150 * buffer's verified bit is no longer valid after reading from
151 * disk again due to write out error, clear it to make sure we
152 * recheck the buffer contents.
153 */
154 clear_buffer_verified(bh);
155
156 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
157 get_bh(bh);
158 submit_bh(REQ_OP_READ, op_flags, bh);
159}
160
161void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
162 bh_end_io_t *end_io)
163{
164 BUG_ON(!buffer_locked(bh));
165
166 if (ext4_buffer_uptodate(bh)) {
167 unlock_buffer(bh);
168 return;
169 }
170 __ext4_read_bh(bh, op_flags, end_io);
171}
172
173int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
174{
175 BUG_ON(!buffer_locked(bh));
176
177 if (ext4_buffer_uptodate(bh)) {
178 unlock_buffer(bh);
179 return 0;
180 }
181
182 __ext4_read_bh(bh, op_flags, end_io);
183
184 wait_on_buffer(bh);
185 if (buffer_uptodate(bh))
186 return 0;
187 return -EIO;
188}
189
190int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
191{
192 if (trylock_buffer(bh)) {
193 if (wait)
194 return ext4_read_bh(bh, op_flags, NULL);
195 ext4_read_bh_nowait(bh, op_flags, NULL);
196 return 0;
197 }
198 if (wait) {
199 wait_on_buffer(bh);
200 if (buffer_uptodate(bh))
201 return 0;
202 return -EIO;
203 }
204 return 0;
205}
206
fb265c9c 207/*
8394a6ab 208 * This works like __bread_gfp() except it uses ERR_PTR for error
fb265c9c
TT		 209 * returns. Currently with sb_bread it's impossible to distinguish
210 * between ENOMEM and EIO situations (since both result in a NULL
211 * return.
212 */
8394a6ab 213static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
214 sector_t block, int op_flags,
215 gfp_t gfp)
fb265c9c 216{
2d069c08 217 struct buffer_head *bh;
218 int ret;
fb265c9c 219
8394a6ab 220 bh = sb_getblk_gfp(sb, block, gfp);
fb265c9c
TT		 221 if (bh == NULL)
222 return ERR_PTR(-ENOMEM);
cf2834a5 223 if (ext4_buffer_uptodate(bh))
fb265c9c 224 return bh;
2d069c08 225
226 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
227 if (ret) {
228 put_bh(bh);
229 return ERR_PTR(ret);
230 }
231 return bh;
fb265c9c
TT		 232}
233
8394a6ab 234struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
235 int op_flags)
236{
237 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
238}
239
240struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
241 sector_t block)
242{
243 return __ext4_sb_bread_gfp(sb, block, 0, 0);
244}
245
5df1d412 246void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
247{
248 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
249
250 if (likely(bh)) {
251 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
252 brelse(bh);
253 }
fb265c9c
TT		 254}
255
d25425f8
DW		 256static int ext4_verify_csum_type(struct super_block *sb,
257 struct ext4_super_block *es)
258{
e2b911c5 259 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 260 return 1;
261
262 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
263}
264
a9c47317
DW		 265static __le32 ext4_superblock_csum(struct super_block *sb,
266 struct ext4_super_block *es)
267{
268 struct ext4_sb_info *sbi = EXT4_SB(sb);
269 int offset = offsetof(struct ext4_super_block, s_checksum);
270 __u32 csum;
271
272 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
273
274 return cpu_to_le32(csum);
275}
276
c197855e
SH		 277static int ext4_superblock_csum_verify(struct super_block *sb,
278 struct ext4_super_block *es)
a9c47317 279{
9aa5d32b 280 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 281 return 1;
282
283 return es->s_checksum == ext4_superblock_csum(sb, es);
284}
285
06db49e6 286void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 287{
06db49e6
TT		 288 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
289
9aa5d32b 290 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 291 return;
292
293 es->s_checksum = ext4_superblock_csum(sb, es);
294}
295
8fadc143
AR		 296ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
297 struct ext4_group_desc *bg)
bd81d8ee 298{
3a14589c 299 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 300 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 301 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 302}
303
8fadc143
AR		 304ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
305 struct ext4_group_desc *bg)
bd81d8ee 306{
5272f837 307 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 308 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 309 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 310}
311
8fadc143
AR		 312ext4_fsblk_t ext4_inode_table(struct super_block *sb,
313 struct ext4_group_desc *bg)
bd81d8ee 314{
5272f837 315 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 316 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 317 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 318}
319
021b65bb
TT		 320__u32 ext4_free_group_clusters(struct super_block *sb,
321 struct ext4_group_desc *bg)
560671a0
AK		 322{
323 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
324 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 325 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 326}
327
328__u32 ext4_free_inodes_count(struct super_block *sb,
329 struct ext4_group_desc *bg)
330{
331 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
332 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 333 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 334}
335
336__u32 ext4_used_dirs_count(struct super_block *sb,
337 struct ext4_group_desc *bg)
338{
339 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
340 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 341 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 342}
343
344__u32 ext4_itable_unused_count(struct super_block *sb,
345 struct ext4_group_desc *bg)
346{
347 return le16_to_cpu(bg->bg_itable_unused_lo) |
348 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 349 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 350}
351
8fadc143
AR		 352void ext4_block_bitmap_set(struct super_block *sb,
353 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 354{
3a14589c 355 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 356 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
357 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 358}
359
8fadc143
AR		 360void ext4_inode_bitmap_set(struct super_block *sb,
361 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 362{
5272f837 363 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 364 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
365 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 366}
367
8fadc143
AR		 368void ext4_inode_table_set(struct super_block *sb,
369 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 370{
5272f837 371 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 372 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
373 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 374}
375
021b65bb
TT		 376void ext4_free_group_clusters_set(struct super_block *sb,
377 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 378{
379 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
380 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
381 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
382}
383
384void ext4_free_inodes_set(struct super_block *sb,
385 struct ext4_group_desc *bg, __u32 count)
386{
387 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
388 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
389 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
390}
391
392void ext4_used_dirs_set(struct super_block *sb,
393 struct ext4_group_desc *bg, __u32 count)
394{
395 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
396 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
397 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
398}
399
400void ext4_itable_unused_set(struct super_block *sb,
401 struct ext4_group_desc *bg, __u32 count)
402{
403 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
404 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
405 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
406}
407
c92dc856 408static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
6a0678a7 409{
6a0678a7
AB		 410 now = clamp_val(now, 0, (1ull << 40) - 1);
411
412 *lo = cpu_to_le32(lower_32_bits(now));
413 *hi = upper_32_bits(now);
414}
415
416static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
417{
418 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
419}
420#define ext4_update_tstamp(es, tstamp) \
c92dc856
JK		 421 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
422 ktime_get_real_seconds())
6a0678a7
AB		 423#define ext4_get_tstamp(es, tstamp) \
424 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
d3d1faf6 425
bdfe0cbd
TT		 426/*
427 * The del_gendisk() function uninitializes the disk-specific data
428 * structures, including the bdi structure, without telling anyone
429 * else. Once this happens, any attempt to call mark_buffer_dirty()
430 * (for example, by ext4_commit_super), will cause a kernel OOPS.
431 * This is a kludge to prevent these oops until we can put in a proper
432 * hook in del_gendisk() to inform the VFS and file system layers.
433 */
434static int block_device_ejected(struct super_block *sb)
435{
436 struct inode *bd_inode = sb->s_bdev->bd_inode;
437 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
438
439 return bdi->dev == NULL;
440}
441
18aadd47
BJ		 442static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
443{
444 struct super_block *sb = journal->j_private;
445 struct ext4_sb_info *sbi = EXT4_SB(sb);
446 int error = is_journal_aborted(journal);
5d3ee208 447 struct ext4_journal_cb_entry *jce;
18aadd47 448
5d3ee208 449 BUG_ON(txn->t_state == T_FINISHED);
a0154344
DJ		 450
451 ext4_process_freed_data(sb, txn->t_tid);
452
18aadd47 453 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 454 while (!list_empty(&txn->t_private_list)) {
455 jce = list_entry(txn->t_private_list.next,
456 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 457 list_del_init(&jce->jce_list);
458 spin_unlock(&sbi->s_md_lock);
459 jce->jce_func(sb, jce, error);
460 spin_lock(&sbi->s_md_lock);
461 }
462 spin_unlock(&sbi->s_md_lock);
463}
1c13d5c0 464
afb585a9
MFO		 465/*
466 * This writepage callback for write_cache_pages()
467 * takes care of a few cases after page cleaning.
468 *
469 * write_cache_pages() already checks for dirty pages
470 * and calls clear_page_dirty_for_io(), which we want,
471 * to write protect the pages.
472 *
473 * However, we may have to redirty a page (see below.)
474 */
475static int ext4_journalled_writepage_callback(struct page *page,
476 struct writeback_control *wbc,
477 void *data)
478{
479 transaction_t *transaction = (transaction_t *) data;
480 struct buffer_head *bh, *head;
481 struct journal_head *jh;
482
483 bh = head = page_buffers(page);
484 do {
485 /*
486 * We have to redirty a page in these cases:
487 * 1) If buffer is dirty, it means the page was dirty because it
488 * contains a buffer that needs checkpointing. So the dirty bit
489 * needs to be preserved so that checkpointing writes the buffer
490 * properly.
491 * 2) If buffer is not part of the committing transaction
492 * (we may have just accidentally come across this buffer because
493 * inode range tracking is not exact) or if the currently running
494 * transaction already contains this buffer as well, dirty bit
495 * needs to be preserved so that the buffer gets writeprotected
496 * properly on running transaction's commit.
497 */
498 jh = bh2jh(bh);
499 if (buffer_dirty(bh) ||
500 (jh && (jh->b_transaction != transaction ||
501 jh->b_next_transaction))) {
502 redirty_page_for_writepage(wbc, page);
503 goto out;
504 }
505 } while ((bh = bh->b_this_page) != head);
506
507out:
508 return AOP_WRITEPAGE_ACTIVATE;
509}
510
511static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
512{
513 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
514 struct writeback_control wbc = {
515 .sync_mode = WB_SYNC_ALL,
516 .nr_to_write = LONG_MAX,
517 .range_start = jinode->i_dirty_start,
518 .range_end = jinode->i_dirty_end,
519 };
520
521 return write_cache_pages(mapping, &wbc,
522 ext4_journalled_writepage_callback,
523 jinode->i_transaction);
524}
525
526static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
527{
528 int ret;
529
530 if (ext4_should_journal_data(jinode->i_vfs_inode))
531 ret = ext4_journalled_submit_inode_data_buffers(jinode);
532 else
533 ret = jbd2_journal_submit_inode_data_buffers(jinode);
534
535 return ret;
536}
537
538static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
539{
540 int ret = 0;
541
542 if (!ext4_should_journal_data(jinode->i_vfs_inode))
543 ret = jbd2_journal_finish_inode_data_buffers(jinode);
544
545 return ret;
546}
547
1dc1097f
JK		 548static bool system_going_down(void)
549{
550 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
551 || system_state == SYSTEM_RESTART;
552}
553
02a7780e
JK		 554struct ext4_err_translation {
555 int code;
556 int errno;
557};
558
559#define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
560
561static struct ext4_err_translation err_translation[] = {
562 EXT4_ERR_TRANSLATE(EIO),
563 EXT4_ERR_TRANSLATE(ENOMEM),
564 EXT4_ERR_TRANSLATE(EFSBADCRC),
565 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
566 EXT4_ERR_TRANSLATE(ENOSPC),
567 EXT4_ERR_TRANSLATE(ENOKEY),
568 EXT4_ERR_TRANSLATE(EROFS),
569 EXT4_ERR_TRANSLATE(EFBIG),
570 EXT4_ERR_TRANSLATE(EEXIST),
571 EXT4_ERR_TRANSLATE(ERANGE),
572 EXT4_ERR_TRANSLATE(EOVERFLOW),
573 EXT4_ERR_TRANSLATE(EBUSY),
574 EXT4_ERR_TRANSLATE(ENOTDIR),
575 EXT4_ERR_TRANSLATE(ENOTEMPTY),
576 EXT4_ERR_TRANSLATE(ESHUTDOWN),
577 EXT4_ERR_TRANSLATE(EFAULT),
578};
579
580static int ext4_errno_to_code(int errno)
581{
582 int i;
583
584 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
585 if (err_translation[i].errno == errno)
586 return err_translation[i].code;
587 return EXT4_ERR_UNKNOWN;
588}
589
2d01ddc8
JK		 590static void save_error_info(struct super_block *sb, int error,
591 __u32 ino, __u64 block,
592 const char *func, unsigned int line)
40676623 593{
c92dc856 594 struct ext4_sb_info *sbi = EXT4_SB(sb);
40676623 595
02a7780e
JK		 596 /* We default to EFSCORRUPTED error... */
597 if (error == 0)
598 error = EFSCORRUPTED;
c92dc856
JK		 599
600 spin_lock(&sbi->s_error_lock);
601 sbi->s_add_error_count++;
602 sbi->s_last_error_code = error;
603 sbi->s_last_error_line = line;
604 sbi->s_last_error_ino = ino;
605 sbi->s_last_error_block = block;
606 sbi->s_last_error_func = func;
607 sbi->s_last_error_time = ktime_get_real_seconds();
608 if (!sbi->s_first_error_time) {
609 sbi->s_first_error_code = error;
610 sbi->s_first_error_line = line;
611 sbi->s_first_error_ino = ino;
612 sbi->s_first_error_block = block;
613 sbi->s_first_error_func = func;
614 sbi->s_first_error_time = sbi->s_last_error_time;
615 }
616 spin_unlock(&sbi->s_error_lock);
40676623
JK		 617}
618
ac27a0ec
DK		 619/* Deal with the reporting of failure conditions on a filesystem such as
620 * inconsistencies detected or read IO failures.
621 *
622 * On ext2, we can store the error state of the filesystem in the
617ba13b 623 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 624 * write ordering constraints on the superblock which prevent us from
625 * writing it out straight away; and given that the journal is about to
626 * be aborted, we can't rely on the current, or future, transactions to
627 * write out the superblock safely.
628 *
dab291af 629 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 630 * the journal instead. On recovery, the journal will complain about
ac27a0ec 631 * that error until we've noted it down and cleared it.
014c9caa
JK		 632 *
633 * If force_ro is set, we unconditionally force the filesystem into an
634 * ABORT|READONLY state, unless the error response on the fs has been set to
635 * panic in which case we take the easy way out and panic immediately. This is
636 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
637 * at a critical moment in log management.
ac27a0ec 638 */
e789ca0c
JK		 639static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
640 __u32 ino, __u64 block,
641 const char *func, unsigned int line)
ac27a0ec 642{
b08070ec 643 journal_t *journal = EXT4_SB(sb)->s_journal;
2d01ddc8 644 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
b08070ec 645
e789ca0c 646 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
327eaf73
TT		 647 if (test_opt(sb, WARN_ON_ERROR))
648 WARN_ON_ONCE(1);
649
2d01ddc8
JK		 650 if (!continue_fs && !sb_rdonly(sb)) {
651 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
652 if (journal)
653 jbd2_journal_abort(journal, -EIO);
654 }
655
656 if (!bdev_read_only(sb->s_bdev)) {
e789ca0c 657 save_error_info(sb, error, ino, block, func, line);
2d01ddc8
JK		 658 /*
659 * In case the fs should keep running, we need to writeout
660 * superblock through the journal. Due to lock ordering
661 * constraints, it may not be safe to do it right here so we
662 * defer superblock flushing to a workqueue.
663 */
664 if (continue_fs)
665 schedule_work(&EXT4_SB(sb)->s_error_work);
666 else
667 ext4_commit_super(sb);
668 }
e789ca0c 669
2d01ddc8 670 if (sb_rdonly(sb) || continue_fs)
ac27a0ec
DK		 671 return;
672
1dc1097f
JK		 673 /*
674 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
675 * could panic during 'reboot -f' as the underlying device got already
676 * disabled.
677 */
014c9caa 678 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
617ba13b 679 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 680 sb->s_id);
4327ba52 681 }
014c9caa
JK		 682 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
683 /*
684 * Make sure updated value of ->s_mount_flags will be visible before
685 * ->s_flags update
686 */
687 smp_wmb();
688 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 689}
690
c92dc856
JK		 691static void flush_stashed_error_work(struct work_struct *work)
692{
693 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
694 s_error_work);
2d01ddc8
JK		 695 journal_t *journal = sbi->s_journal;
696 handle_t *handle;
c92dc856 697
2d01ddc8
JK		 698 /*
699 * If the journal is still running, we have to write out superblock
700 * through the journal to avoid collisions of other journalled sb
701 * updates.
702 *
703 * We use directly jbd2 functions here to avoid recursing back into
704 * ext4 error handling code during handling of previous errors.
705 */
706 if (!sb_rdonly(sbi->s_sb) && journal) {
707 handle = jbd2_journal_start(journal, 1);
708 if (IS_ERR(handle))
709 goto write_directly;
710 if (jbd2_journal_get_write_access(handle, sbi->s_sbh)) {
711 jbd2_journal_stop(handle);
712 goto write_directly;
713 }
714 ext4_update_super(sbi->s_sb);
715 if (jbd2_journal_dirty_metadata(handle, sbi->s_sbh)) {
716 jbd2_journal_stop(handle);
717 goto write_directly;
718 }
719 jbd2_journal_stop(handle);
720 return;
721 }
722write_directly:
723 /*
724 * Write through journal failed. Write sb directly to get error info
725 * out and hope for the best.
726 */
4392fbc4 727 ext4_commit_super(sbi->s_sb);
ac27a0ec
DK		 728}
729
efbed4dc
TT		 730#define ext4_error_ratelimit(sb) \
731 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
732 "EXT4-fs error")
733
12062ddd 734void __ext4_error(struct super_block *sb, const char *function,
014c9caa 735 unsigned int line, bool force_ro, int error, __u64 block,
54d3adbc 736 const char *fmt, ...)
ac27a0ec 737{
0ff2ea7d 738 struct va_format vaf;
ac27a0ec
DK		 739 va_list args;
740
0db1ff22
TT		 741 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
742 return;
743
ccf0f32a 744 trace_ext4_error(sb, function, line);
efbed4dc
TT		 745 if (ext4_error_ratelimit(sb)) {
746 va_start(args, fmt);
747 vaf.fmt = fmt;
748 vaf.va = &args;
749 printk(KERN_CRIT
750 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
751 sb->s_id, function, line, current->comm, &vaf);
752 va_end(args);
753 }
e789ca0c 754 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
ac27a0ec
DK		 755}
756
e7c96e8e 757void __ext4_error_inode(struct inode *inode, const char *function,
54d3adbc 758 unsigned int line, ext4_fsblk_t block, int error,
e7c96e8e 759 const char *fmt, ...)
273df556
FM		 760{
761 va_list args;
f7c21177 762 struct va_format vaf;
273df556 763
0db1ff22
TT		 764 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
765 return;
766
ccf0f32a 767 trace_ext4_error(inode->i_sb, function, line);
efbed4dc
TT		 768 if (ext4_error_ratelimit(inode->i_sb)) {
769 va_start(args, fmt);
770 vaf.fmt = fmt;
771 vaf.va = &args;
772 if (block)
773 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
774 "inode #%lu: block %llu: comm %s: %pV\n",
775 inode->i_sb->s_id, function, line, inode->i_ino,
776 block, current->comm, &vaf);
777 else
778 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
779 "inode #%lu: comm %s: %pV\n",
780 inode->i_sb->s_id, function, line, inode->i_ino,
781 current->comm, &vaf);
782 va_end(args);
783 }
e789ca0c
JK		 784 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
785 function, line);
273df556
FM		 786}
787
e7c96e8e
JP		 788void __ext4_error_file(struct file *file, const char *function,
789 unsigned int line, ext4_fsblk_t block,
790 const char *fmt, ...)
273df556
FM		 791{
792 va_list args;
f7c21177 793 struct va_format vaf;
496ad9aa 794 struct inode *inode = file_inode(file);
273df556
FM		 795 char pathname[80], *path;
796
0db1ff22
TT		 797 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
798 return;
799
ccf0f32a 800 trace_ext4_error(inode->i_sb, function, line);
efbed4dc 801 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 802 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 803 if (IS_ERR(path))
804 path = "(unknown)";
805 va_start(args, fmt);
806 vaf.fmt = fmt;
807 vaf.va = &args;
808 if (block)
809 printk(KERN_CRIT
810 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
811 "block %llu: comm %s: path %s: %pV\n",
812 inode->i_sb->s_id, function, line, inode->i_ino,
813 block, current->comm, path, &vaf);
814 else
815 printk(KERN_CRIT
816 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
817 "comm %s: path %s: %pV\n",
818 inode->i_sb->s_id, function, line, inode->i_ino,
819 current->comm, path, &vaf);
820 va_end(args);
821 }
e789ca0c
JK		 822 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
823 function, line);
273df556
FM		 824}
825
722887dd
TT		 826const char *ext4_decode_error(struct super_block *sb, int errno,
827 char nbuf[16])
ac27a0ec
DK		 828{
829 char *errstr = NULL;
830
831 switch (errno) {
6a797d27
DW		 832 case -EFSCORRUPTED:
833 errstr = "Corrupt filesystem";
834 break;
835 case -EFSBADCRC:
836 errstr = "Filesystem failed CRC";
837 break;
ac27a0ec
DK		 838 case -EIO:
839 errstr = "IO failure";
840 break;
841 case -ENOMEM:
842 errstr = "Out of memory";
843 break;
844 case -EROFS:
78f1ddbb
TT		 845 if (!sb || (EXT4_SB(sb)->s_journal &&
846 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 847 errstr = "Journal has aborted";
848 else
849 errstr = "Readonly filesystem";
850 break;
851 default:
852 /* If the caller passed in an extra buffer for unknown
853 * errors, textualise them now. Else we just return
854 * NULL. */
855 if (nbuf) {
856 /* Check for truncated error codes... */
857 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
858 errstr = nbuf;
859 }
860 break;
861 }
862
863 return errstr;
864}
865
617ba13b 866/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 867 * automatically and invokes the appropriate error response. */
868
c398eda0
TT		 869void __ext4_std_error(struct super_block *sb, const char *function,
870 unsigned int line, int errno)
ac27a0ec
DK		 871{
872 char nbuf[16];
873 const char *errstr;
874
0db1ff22
TT		 875 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
876 return;
877
ac27a0ec
DK		 878 /* Special case: if the error is EROFS, and we're not already
879 * inside a transaction, then there's really no point in logging
880 * an error. */
bc98a42c 881 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
ac27a0ec
DK		 882 return;
883
efbed4dc
TT		 884 if (ext4_error_ratelimit(sb)) {
885 errstr = ext4_decode_error(sb, errno, nbuf);
886 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
887 sb->s_id, function, line, errstr);
888 }
ac27a0ec 889
e789ca0c 890 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
ac27a0ec
DK		 891}
892
e7c96e8e
JP		 893void __ext4_msg(struct super_block *sb,
894 const char *prefix, const char *fmt, ...)
b31e1552 895{
0ff2ea7d 896 struct va_format vaf;
b31e1552
ES		 897 va_list args;
898
1cf006ed 899 atomic_inc(&EXT4_SB(sb)->s_msg_count);
efbed4dc
TT		 900 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
901 return;
902
b31e1552 903 va_start(args, fmt);
0ff2ea7d
JP		 904 vaf.fmt = fmt;
905 vaf.va = &args;
906 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 907 va_end(args);
908}
909
1cf006ed
DM		 910static int ext4_warning_ratelimit(struct super_block *sb)
911{
912 atomic_inc(&EXT4_SB(sb)->s_warning_count);
913 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
914 "EXT4-fs warning");
915}
b03a2f7e 916
12062ddd 917void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 918 unsigned int line, const char *fmt, ...)
ac27a0ec 919{
0ff2ea7d 920 struct va_format vaf;
ac27a0ec
DK		 921 va_list args;
922
b03a2f7e 923 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 924 return;
925
ac27a0ec 926 va_start(args, fmt);
0ff2ea7d
JP		 927 vaf.fmt = fmt;
928 vaf.va = &args;
929 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
930 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 931 va_end(args);
932}
933
b03a2f7e
AD		 934void __ext4_warning_inode(const struct inode *inode, const char *function,
935 unsigned int line, const char *fmt, ...)
936{
937 struct va_format vaf;
938 va_list args;
939
940 if (!ext4_warning_ratelimit(inode->i_sb))
941 return;
942
943 va_start(args, fmt);
944 vaf.fmt = fmt;
945 vaf.va = &args;
946 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
947 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
948 function, line, inode->i_ino, current->comm, &vaf);
949 va_end(args);
950}
951
e29136f8
TT		 952void __ext4_grp_locked_error(const char *function, unsigned int line,
953 struct super_block *sb, ext4_group_t grp,
954 unsigned long ino, ext4_fsblk_t block,
955 const char *fmt, ...)
5d1b1b3f
AK		 956__releases(bitlock)
957__acquires(bitlock)
958{
0ff2ea7d 959 struct va_format vaf;
5d1b1b3f 960 va_list args;
5d1b1b3f 961
0db1ff22
TT		 962 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
963 return;
964
ccf0f32a 965 trace_ext4_error(sb, function, line);
efbed4dc
TT		 966 if (ext4_error_ratelimit(sb)) {
967 va_start(args, fmt);
968 vaf.fmt = fmt;
969 vaf.va = &args;
970 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
971 sb->s_id, function, line, grp);
972 if (ino)
973 printk(KERN_CONT "inode %lu: ", ino);
974 if (block)
975 printk(KERN_CONT "block %llu:",
976 (unsigned long long) block);
977 printk(KERN_CONT "%pV\n", &vaf);
978 va_end(args);
979 }
5d1b1b3f
AK		 980
981 if (test_opt(sb, ERRORS_CONT)) {
c92dc856
JK		 982 if (test_opt(sb, WARN_ON_ERROR))
983 WARN_ON_ONCE(1);
e789ca0c 984 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2d01ddc8
JK		 985 if (!bdev_read_only(sb->s_bdev)) {
986 save_error_info(sb, EFSCORRUPTED, ino, block, function,
987 line);
e789ca0c 988 schedule_work(&EXT4_SB(sb)->s_error_work);
2d01ddc8 989 }
5d1b1b3f
AK		 990 return;
991 }
992 ext4_unlock_group(sb, grp);
e789ca0c 993 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
5d1b1b3f
AK		 994 /*
995 * We only get here in the ERRORS_RO case; relocking the group
996 * may be dangerous, but nothing bad will happen since the
997 * filesystem will have already been marked read/only and the
998 * journal has been aborted. We return 1 as a hint to callers
999 * who might what to use the return value from
25985edc 1000 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 1001 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1002 * aggressively from the ext4 function in question, with a
1003 * more appropriate error code.
1004 */
1005 ext4_lock_group(sb, grp);
1006 return;
1007}
1008
db79e6d1
WS		 1009void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1010 ext4_group_t group,
1011 unsigned int flags)
1012{
1013 struct ext4_sb_info *sbi = EXT4_SB(sb);
1014 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1015 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
9af0b3d1
WS		 1016 int ret;
1017
1018 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1019 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1020 &grp->bb_state);
1021 if (!ret)
1022 percpu_counter_sub(&sbi->s_freeclusters_counter,
1023 grp->bb_free);
db79e6d1
WS		 1024 }
1025
9af0b3d1
WS		 1026 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1027 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1028 &grp->bb_state);
1029 if (!ret && gdp) {
db79e6d1
WS		 1030 int count;
1031
1032 count = ext4_free_inodes_count(sb, gdp);
1033 percpu_counter_sub(&sbi->s_freeinodes_counter,
1034 count);
1035 }
db79e6d1
WS		 1036 }
1037}
1038
617ba13b 1039void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 1040{
617ba13b 1041 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 1042
617ba13b 1043 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 1044 return;
1045
12062ddd 1046 ext4_warning(sb,
ac27a0ec
DK		 1047 "updating to rev %d because of new feature flag, "
1048 "running e2fsck is recommended",
617ba13b 1049 EXT4_DYNAMIC_REV);
ac27a0ec 1050
617ba13b
MC		 1051 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1052 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1053 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 1054 /* leave es->s_feature_*compat flags alone */
1055 /* es->s_uuid will be set by e2fsck if empty */
1056
1057 /*
1058 * The rest of the superblock fields should be zero, and if not it
1059 * means they are likely already in use, so leave them alone. We
1060 * can leave it up to e2fsck to clean up any inconsistencies there.
1061 */
1062}
1063
1064/*
1065 * Open the external journal device
1066 */
b31e1552 1067static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 1068{
1069 struct block_device *bdev;
ac27a0ec 1070
d4d77629 1071 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 1072 if (IS_ERR(bdev))
1073 goto fail;
1074 return bdev;
1075
1076fail:
ea3edd4d
CH		 1077 ext4_msg(sb, KERN_ERR,
1078 "failed to open journal device unknown-block(%u,%u) %ld",
1079 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
ac27a0ec
DK		 1080 return NULL;
1081}
1082
1083/*
1084 * Release the journal device
1085 */
4385bab1 1086static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 1087{
4385bab1 1088 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 1089}
1090
4385bab1 1091static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 1092{
1093 struct block_device *bdev;
ee7ed3aa 1094 bdev = sbi->s_journal_bdev;
ac27a0ec 1095 if (bdev) {
4385bab1 1096 ext4_blkdev_put(bdev);
ee7ed3aa 1097 sbi->s_journal_bdev = NULL;
ac27a0ec 1098 }
ac27a0ec
DK		 1099}
1100
1101static inline struct inode *orphan_list_entry(struct list_head *l)
1102{
617ba13b 1103 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 1104}
1105
617ba13b 1106static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 1107{
1108 struct list_head *l;
1109
b31e1552
ES		 1110 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1111 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 1112
1113 printk(KERN_ERR "sb_info orphan list:\n");
1114 list_for_each(l, &sbi->s_orphan) {
1115 struct inode *inode = orphan_list_entry(l);
1116 printk(KERN_ERR " "
1117 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1118 inode->i_sb->s_id, inode->i_ino, inode,
1119 inode->i_mode, inode->i_nlink,
1120 NEXT_ORPHAN(inode));
1121 }
1122}
1123
957153fc
JK		 1124#ifdef CONFIG_QUOTA
1125static int ext4_quota_off(struct super_block *sb, int type);
1126
1127static inline void ext4_quota_off_umount(struct super_block *sb)
1128{
1129 int type;
1130
964edf66
JK		 1131 /* Use our quota_off function to clear inode flags etc. */
1132 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1133 ext4_quota_off(sb, type);
957153fc 1134}
33458eab
TT		 1135
1136/*
1137 * This is a helper function which is used in the mount/remount
1138 * codepaths (which holds s_umount) to fetch the quota file name.
1139 */
1140static inline char *get_qf_name(struct super_block *sb,
1141 struct ext4_sb_info *sbi,
1142 int type)
1143{
1144 return rcu_dereference_protected(sbi->s_qf_names[type],
1145 lockdep_is_held(&sb->s_umount));
1146}
957153fc
JK		 1147#else
1148static inline void ext4_quota_off_umount(struct super_block *sb)
1149{
1150}
1151#endif
1152
2b2d6d01 1153static void ext4_put_super(struct super_block *sb)
ac27a0ec 1154{
617ba13b
MC		 1155 struct ext4_sb_info *sbi = EXT4_SB(sb);
1156 struct ext4_super_block *es = sbi->s_es;
1d0c3924 1157 struct buffer_head **group_desc;
7c990728 1158 struct flex_groups **flex_groups;
97abd7d4 1159 int aborted = 0;
ef2cabf7 1160 int i, err;
ac27a0ec 1161
857ac889 1162 ext4_unregister_li_request(sb);
957153fc 1163 ext4_quota_off_umount(sb);
e0ccfd95 1164
c92dc856 1165 flush_work(&sbi->s_error_work);
2e8fa54e 1166 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 1167
5e47868f
RH		 1168 /*
1169 * Unregister sysfs before destroying jbd2 journal.
1170 * Since we could still access attr_journal_task attribute via sysfs
1171 * path which could have sbi->s_journal->j_task as NULL
1172 */
1173 ext4_unregister_sysfs(sb);
1174
0390131b 1175 if (sbi->s_journal) {
97abd7d4 1176 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 1177 err = jbd2_journal_destroy(sbi->s_journal);
1178 sbi->s_journal = NULL;
878520ac 1179 if ((err < 0) && !aborted) {
54d3adbc 1180 ext4_abort(sb, -err, "Couldn't clean up the journal");
878520ac 1181 }
0390131b 1182 }
d4edac31 1183
d3922a77 1184 ext4_es_unregister_shrinker(sbi);
9105bb14 1185 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 1186 ext4_release_system_zone(sb);
1187 ext4_mb_release(sb);
1188 ext4_ext_release(sb);
d4edac31 1189
bc98a42c 1190 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 1191 ext4_clear_feature_journal_needs_recovery(sb);
ac27a0ec 1192 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 1193 }
bc98a42c 1194 if (!sb_rdonly(sb))
4392fbc4 1195 ext4_commit_super(sb);
a8e25a83 1196
1d0c3924
TT		 1197 rcu_read_lock();
1198 group_desc = rcu_dereference(sbi->s_group_desc);
ac27a0ec 1199 for (i = 0; i < sbi->s_gdb_count; i++)
1d0c3924
TT		 1200 brelse(group_desc[i]);
1201 kvfree(group_desc);
7c990728
SJS		 1202 flex_groups = rcu_dereference(sbi->s_flex_groups);
1203 if (flex_groups) {
1204 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1205 kvfree(flex_groups[i]);
1206 kvfree(flex_groups);
1207 }
1d0c3924 1208 rcu_read_unlock();
57042651 1209 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 1210 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1211 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 1212 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
efc61345 1213 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
bbd55937 1214 percpu_free_rwsem(&sbi->s_writepages_rwsem);
ac27a0ec 1215#ifdef CONFIG_QUOTA
a2d4a646 1216 for (i = 0; i < EXT4_MAXQUOTAS; i++)
33458eab 1217 kfree(get_qf_name(sb, sbi, i));
ac27a0ec
DK		 1218#endif
1219
1220 /* Debugging code just in case the in-memory inode orphan list
1221 * isn't empty. The on-disk one can be non-empty if we've
1222 * detected an error and taken the fs readonly, but the
1223 * in-memory list had better be clean by this point. */
1224 if (!list_empty(&sbi->s_orphan))
1225 dump_orphan_list(sb, sbi);
837c23fb 1226 ASSERT(list_empty(&sbi->s_orphan));
ac27a0ec 1227
89d96a6f 1228 sync_blockdev(sb->s_bdev);
f98393a6 1229 invalidate_bdev(sb->s_bdev);
ee7ed3aa 1230 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
ac27a0ec
DK		 1231 /*
1232 * Invalidate the journal device's buffers. We don't want them
1233 * floating about in memory - the physical journal device may
1234 * hotswapped, and it breaks the `ro-after' testing code.
1235 */
ee7ed3aa
CX		 1236 sync_blockdev(sbi->s_journal_bdev);
1237 invalidate_bdev(sbi->s_journal_bdev);
617ba13b 1238 ext4_blkdev_remove(sbi);
ac27a0ec 1239 }
50c15df6
CX		 1240
1241 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1242 sbi->s_ea_inode_cache = NULL;
1243
1244 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1245 sbi->s_ea_block_cache = NULL;
1246
c5e06d10
JL		 1247 if (sbi->s_mmp_tsk)
1248 kthread_stop(sbi->s_mmp_tsk);
9060dd2c 1249 brelse(sbi->s_sbh);
ac27a0ec 1250 sb->s_fs_info = NULL;
3197ebdb
TT		 1251 /*
1252 * Now that we are completely done shutting down the
1253 * superblock, we need to actually destroy the kobject.
1254 */
3197ebdb
TT		 1255 kobject_put(&sbi->s_kobj);
1256 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 1257 if (sbi->s_chksum_driver)
1258 crypto_free_shash(sbi->s_chksum_driver);
705895b6 1259 kfree(sbi->s_blockgroup_lock);
5e405595 1260 fs_put_dax(sbi->s_daxdev);
ac4acb1f 1261 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
c83ad55e 1262#ifdef CONFIG_UNICODE
f8f4acb6 1263 utf8_unload(sb->s_encoding);
c83ad55e 1264#endif
ac27a0ec 1265 kfree(sbi);
ac27a0ec
DK		 1266}
1267
e18b890b 1268static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 1269
1270/*
1271 * Called inside transaction, so use GFP_NOFS
1272 */
617ba13b 1273static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 1274{
617ba13b 1275 struct ext4_inode_info *ei;
ac27a0ec 1276
e6b4f8da 1277 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 1278 if (!ei)
1279 return NULL;
0b8e58a1 1280
ee73f9a5 1281 inode_set_iversion(&ei->vfs_inode, 1);
202ee5df 1282 spin_lock_init(&ei->i_raw_lock);
c9de560d 1283 INIT_LIST_HEAD(&ei->i_prealloc_list);
27bc446e 1284 atomic_set(&ei->i_prealloc_active, 0);
c9de560d 1285 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 1286 ext4_es_init_tree(&ei->i_es_tree);
1287 rwlock_init(&ei->i_es_lock);
edaa53ca 1288 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 1289 ei->i_es_all_nr = 0;
edaa53ca 1290 ei->i_es_shk_nr = 0;
dd475925 1291 ei->i_es_shrink_lblk = 0;
d2a17637 1292 ei->i_reserved_data_blocks = 0;
d2a17637 1293 spin_lock_init(&(ei->i_block_reservation_lock));
1dc0aa46 1294 ext4_init_pending_tree(&ei->i_pending_tree);
a9e7f447
DM		 1295#ifdef CONFIG_QUOTA
1296 ei->i_reserved_quota = 0;
96c7e0d9 1297 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 1298#endif
8aefcd55 1299 ei->jinode = NULL;
2e8fa54e 1300 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 1301 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 1302 ei->i_sync_tid = 0;
1303 ei->i_datasync_tid = 0;
e27f41e1 1304 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 1305 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
aa75f4d3
HS		 1306 ext4_fc_init_inode(&ei->vfs_inode);
1307 mutex_init(&ei->i_fc_lock);
ac27a0ec
DK		 1308 return &ei->vfs_inode;
1309}
1310
7ff9c073
TT		 1311static int ext4_drop_inode(struct inode *inode)
1312{
1313 int drop = generic_drop_inode(inode);
1314
29b3692e
EB		 1315 if (!drop)
1316 drop = fscrypt_drop_inode(inode);
1317
7ff9c073
TT		 1318 trace_ext4_drop_inode(inode, drop);
1319 return drop;
1320}
1321
94053139 1322static void ext4_free_in_core_inode(struct inode *inode)
fa0d7e3d 1323{
2c58d548 1324 fscrypt_free_inode(inode);
aa75f4d3
HS		 1325 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1326 pr_warn("%s: inode %ld still in fc list",
1327 __func__, inode->i_ino);
1328 }
fa0d7e3d
NP		 1329 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1330}
1331
617ba13b 1332static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1333{
9f7dd93d 1334 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1335 ext4_msg(inode->i_sb, KERN_ERR,
1336 "Inode %lu (%p): orphan list check failed!",
1337 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1338 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1339 EXT4_I(inode), sizeof(struct ext4_inode_info),
1340 true);
1341 dump_stack();
1342 }
ac27a0ec
DK		 1343}
1344
51cc5068 1345static void init_once(void *foo)
ac27a0ec 1346{
617ba13b 1347 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 1348
a35afb83 1349 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1350 init_rwsem(&ei->xattr_sem);
0e855ac8 1351 init_rwsem(&ei->i_data_sem);
ea3d7209 1352 init_rwsem(&ei->i_mmap_sem);
a35afb83 1353 inode_init_once(&ei->vfs_inode);
aa75f4d3 1354 ext4_fc_init_inode(&ei->vfs_inode);
ac27a0ec
DK		 1355}
1356
e67bc2b3 1357static int __init init_inodecache(void)
ac27a0ec 1358{
f8dd7c70
DW		 1359 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1360 sizeof(struct ext4_inode_info), 0,
1361 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1362 SLAB_ACCOUNT),
1363 offsetof(struct ext4_inode_info, i_data),
1364 sizeof_field(struct ext4_inode_info, i_data),
1365 init_once);
617ba13b 1366 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1367 return -ENOMEM;
1368 return 0;
1369}
1370
1371static void destroy_inodecache(void)
1372{
8c0a8537
KS		 1373 /*
1374 * Make sure all delayed rcu free inodes are flushed before we
1375 * destroy cache.
1376 */
1377 rcu_barrier();
617ba13b 1378 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1379}
1380
0930fcc1 1381void ext4_clear_inode(struct inode *inode)
ac27a0ec 1382{
aa75f4d3 1383 ext4_fc_del(inode);
0930fcc1 1384 invalidate_inode_buffers(inode);
dbd5768f 1385 clear_inode(inode);
27bc446e 1386 ext4_discard_preallocations(inode, 0);
51865fda 1387 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
f4c2d372 1388 dquot_drop(inode);
8aefcd55
TT		 1389 if (EXT4_I(inode)->jinode) {
1390 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1391 EXT4_I(inode)->jinode);
1392 jbd2_free_inode(EXT4_I(inode)->jinode);
1393 EXT4_I(inode)->jinode = NULL;
1394 }
3d204e24 1395 fscrypt_put_encryption_info(inode);
c93d8f88 1396 fsverity_cleanup_inode(inode);
ac27a0ec
DK		 1397}
1398
1b961ac0 1399static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1400 u64 ino, u32 generation)
ac27a0ec 1401{
ac27a0ec 1402 struct inode *inode;
ac27a0ec 1403
8a363970 1404 /*
ac27a0ec
DK		 1405 * Currently we don't know the generation for parent directory, so
1406 * a generation of 0 means "accept any"
1407 */
8a363970 1408 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1d1fe1ee
DH		 1409 if (IS_ERR(inode))
1410 return ERR_CAST(inode);
1411 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1412 iput(inode);
1413 return ERR_PTR(-ESTALE);
1414 }
1b961ac0
CH		 1415
1416 return inode;
1417}
1418
1419static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1420 int fh_len, int fh_type)
1b961ac0
CH		 1421{
1422 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1423 ext4_nfs_get_inode);
1424}
1425
1426static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1427 int fh_len, int fh_type)
1b961ac0
CH		 1428{
1429 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1430 ext4_nfs_get_inode);
ac27a0ec
DK		 1431}
1432
fde87268
TT		 1433static int ext4_nfs_commit_metadata(struct inode *inode)
1434{
1435 struct writeback_control wbc = {
1436 .sync_mode = WB_SYNC_ALL
1437 };
1438
1439 trace_ext4_nfs_commit_metadata(inode);
1440 return ext4_write_inode(inode, &wbc);
1441}
1442
c39a7f84
TO		 1443/*
1444 * Try to release metadata pages (indirect blocks, directories) which are
1445 * mapped via the block device. Since these pages could have journal heads
1446 * which would prevent try_to_free_buffers() from freeing them, we must use
1447 * jbd2 layer's try_to_free_buffers() function to release them.
1448 */
0b8e58a1
AD		 1449static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1450 gfp_t wait)
c39a7f84
TO		 1451{
1452 journal_t *journal = EXT4_SB(sb)->s_journal;
1453
1454 WARN_ON(PageChecked(page));
1455 if (!page_has_buffers(page))
1456 return 0;
1457 if (journal)
529a781e 1458 return jbd2_journal_try_to_free_buffers(journal, page);
1459
c39a7f84
TO		 1460 return try_to_free_buffers(page);
1461}
1462
643fa961 1463#ifdef CONFIG_FS_ENCRYPTION
a7550b30
JK		 1464static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1465{
1466 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1467 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1468}
1469
a7550b30
JK		 1470static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1471 void *fs_data)
1472{
2f8f5e76 1473 handle_t *handle = fs_data;
c1a5d5f6 1474 int res, res2, credits, retries = 0;
2f8f5e76 1475
9ce0151a
EB		 1476 /*
1477 * Encrypting the root directory is not allowed because e2fsck expects
1478 * lost+found to exist and be unencrypted, and encrypting the root
1479 * directory would imply encrypting the lost+found directory as well as
1480 * the filename "lost+found" itself.
1481 */
1482 if (inode->i_ino == EXT4_ROOT_INO)
1483 return -EPERM;
2f8f5e76 1484
7d3e06a8
RZ		 1485 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1486 return -EINVAL;
1487
b383a73f
IW		 1488 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1489 return -EOPNOTSUPP;
1490
94840e3c
EB		 1491 res = ext4_convert_inline_data(inode);
1492 if (res)
1493 return res;
1494
2f8f5e76
EB		 1495 /*
1496 * If a journal handle was specified, then the encryption context is
1497 * being set on a new inode via inheritance and is part of a larger
1498 * transaction to create the inode. Otherwise the encryption context is
1499 * being set on an existing inode in its own transaction. Only in the
1500 * latter case should the "retry on ENOSPC" logic be used.
1501 */
a7550b30 1502
2f8f5e76
EB		 1503 if (handle) {
1504 res = ext4_xattr_set_handle(handle, inode,
1505 EXT4_XATTR_INDEX_ENCRYPTION,
1506 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1507 ctx, len, 0);
a7550b30
JK		 1508 if (!res) {
1509 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1510 ext4_clear_inode_state(inode,
1511 EXT4_STATE_MAY_INLINE_DATA);
a3caa24b 1512 /*
2ee6a576
EB		 1513 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1514 * S_DAX may be disabled
a3caa24b 1515 */
043546e4 1516 ext4_set_inode_flags(inode, false);
a7550b30
JK		 1517 }
1518 return res;
1519 }
1520
b8cb5a54
TE		 1521 res = dquot_initialize(inode);
1522 if (res)
1523 return res;
2f8f5e76 1524retry:
af65207c
TE		 1525 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1526 &credits);
dec214d0
TE		 1527 if (res)
1528 return res;
1529
c1a5d5f6 1530 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
a7550b30
JK		 1531 if (IS_ERR(handle))
1532 return PTR_ERR(handle);
1533
2f8f5e76
EB		 1534 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1535 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1536 ctx, len, 0);
a7550b30
JK		 1537 if (!res) {
1538 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
2ee6a576
EB		 1539 /*
1540 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1541 * S_DAX may be disabled
1542 */
043546e4 1543 ext4_set_inode_flags(inode, false);
a7550b30
JK		 1544 res = ext4_mark_inode_dirty(handle, inode);
1545 if (res)
1546 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1547 }
1548 res2 = ext4_journal_stop(handle);
2f8f5e76
EB		 1549
1550 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1551 goto retry;
a7550b30
JK		 1552 if (!res)
1553 res = res2;
1554 return res;
1555}
1556
ac4acb1f 1557static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
a7550b30 1558{
ac4acb1f 1559 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
a7550b30
JK		 1560}
1561
b925acb8
EB		 1562static bool ext4_has_stable_inodes(struct super_block *sb)
1563{
1564 return ext4_has_feature_stable_inodes(sb);
1565}
1566
1567static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1568 int *ino_bits_ret, int *lblk_bits_ret)
1569{
1570 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1571 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1572}
1573
6f69f0ed 1574static const struct fscrypt_operations ext4_cryptops = {
a5d431ef 1575 .key_prefix = "ext4:",
a7550b30 1576 .get_context = ext4_get_context,
a7550b30 1577 .set_context = ext4_set_context,
ac4acb1f 1578 .get_dummy_policy = ext4_get_dummy_policy,
a7550b30 1579 .empty_dir = ext4_empty_dir,
e12ee683 1580 .max_namelen = EXT4_NAME_LEN,
b925acb8
EB		 1581 .has_stable_inodes = ext4_has_stable_inodes,
1582 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
a7550b30 1583};
a7550b30
JK		 1584#endif
1585
ac27a0ec 1586#ifdef CONFIG_QUOTA
d6006186 1587static const char * const quotatypes[] = INITQFNAMES;
689c958c 1588#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1589
617ba13b
MC		 1590static int ext4_write_dquot(struct dquot *dquot);
1591static int ext4_acquire_dquot(struct dquot *dquot);
1592static int ext4_release_dquot(struct dquot *dquot);
1593static int ext4_mark_dquot_dirty(struct dquot *dquot);
1594static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1595static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1596 const struct path *path);
617ba13b
MC		 1597static int ext4_quota_on_mount(struct super_block *sb, int type);
1598static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1599 size_t len, loff_t off);
617ba13b 1600static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1601 const char *data, size_t len, loff_t off);
7c319d32
AK		 1602static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1603 unsigned int flags);
1604static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1605
96c7e0d9
JK		 1606static struct dquot **ext4_get_dquots(struct inode *inode)
1607{
1608 return EXT4_I(inode)->i_dquot;
1609}
1610
61e225dc 1611static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1612 .get_reserved_space = ext4_get_reserved_space,
1613 .write_dquot = ext4_write_dquot,
1614 .acquire_dquot = ext4_acquire_dquot,
1615 .release_dquot = ext4_release_dquot,
1616 .mark_dirty = ext4_mark_dquot_dirty,
1617 .write_info = ext4_write_info,
1618 .alloc_dquot = dquot_alloc,
1619 .destroy_dquot = dquot_destroy,
1620 .get_projid = ext4_get_projid,
1621 .get_inode_usage = ext4_get_inode_usage,
ebc11f7b 1622 .get_next_id = dquot_get_next_id,
ac27a0ec
DK		 1623};
1624
0d54b217 1625static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1626 .quota_on = ext4_quota_on,
ca0e05e4 1627 .quota_off = ext4_quota_off,
287a8095 1628 .quota_sync = dquot_quota_sync,
0a240339 1629 .get_state = dquot_get_state,
287a8095
CH		 1630 .set_info = dquot_set_dqinfo,
1631 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1632 .set_dqblk = dquot_set_dqblk,
1633 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1634};
1635#endif
1636
ee9b6d61 1637static const struct super_operations ext4_sops = {
617ba13b 1638 .alloc_inode = ext4_alloc_inode,
94053139 1639 .free_inode = ext4_free_in_core_inode,
617ba13b 1640 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1641 .write_inode = ext4_write_inode,
1642 .dirty_inode = ext4_dirty_inode,
7ff9c073 1643 .drop_inode = ext4_drop_inode,
0930fcc1 1644 .evict_inode = ext4_evict_inode,
617ba13b 1645 .put_super = ext4_put_super,
617ba13b 1646 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1647 .freeze_fs = ext4_freeze,
1648 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1649 .statfs = ext4_statfs,
1650 .remount_fs = ext4_remount,
617ba13b 1651 .show_options = ext4_show_options,
ac27a0ec 1652#ifdef CONFIG_QUOTA
617ba13b
MC		 1653 .quota_read = ext4_quota_read,
1654 .quota_write = ext4_quota_write,
96c7e0d9 1655 .get_dquots = ext4_get_dquots,
ac27a0ec 1656#endif
c39a7f84 1657 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1658};
1659
39655164 1660static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1661 .fh_to_dentry = ext4_fh_to_dentry,
1662 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1663 .get_parent = ext4_get_parent,
fde87268 1664 .commit_metadata = ext4_nfs_commit_metadata,
ac27a0ec
DK		 1665};
1666
1667enum {
1668 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1669 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1670 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1671 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1672 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1673 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1674 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1675 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1676 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
4f74d15f 1677 Opt_inlinecrypt,
ac27a0ec 1678 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1679 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1680 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
9cb20f94
IW		 1681 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1682 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
327eaf73
TT		 1683 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1684 Opt_nowarn_on_error, Opt_mblk_io_submit,
670e9875 1685 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1449032b 1686 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1687 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1688 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1689 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1690 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
99c880de 1691 Opt_prefetch_block_bitmaps,
8016e29f 1692#ifdef CONFIG_EXT4_DEBUG
99c880de 1693 Opt_fc_debug_max_replay, Opt_fc_debug_force
8016e29f 1694#endif
ac27a0ec
DK		 1695};
1696
a447c093 1697static const match_table_t tokens = {
ac27a0ec
DK		 1698 {Opt_bsd_df, "bsddf"},
1699 {Opt_minix_df, "minixdf"},
1700 {Opt_grpid, "grpid"},
1701 {Opt_grpid, "bsdgroups"},
1702 {Opt_nogrpid, "nogrpid"},
1703 {Opt_nogrpid, "sysvgroups"},
1704 {Opt_resgid, "resgid=%u"},
1705 {Opt_resuid, "resuid=%u"},
1706 {Opt_sb, "sb=%u"},
1707 {Opt_err_cont, "errors=continue"},
1708 {Opt_err_panic, "errors=panic"},
1709 {Opt_err_ro, "errors=remount-ro"},
1710 {Opt_nouid32, "nouid32"},
ac27a0ec 1711 {Opt_debug, "debug"},
72578c33
TT		 1712 {Opt_removed, "oldalloc"},
1713 {Opt_removed, "orlov"},
ac27a0ec
DK		 1714 {Opt_user_xattr, "user_xattr"},
1715 {Opt_nouser_xattr, "nouser_xattr"},
1716 {Opt_acl, "acl"},
1717 {Opt_noacl, "noacl"},
e3bb52ae 1718 {Opt_noload, "norecovery"},
5a916be1 1719 {Opt_noload, "noload"},
72578c33
TT		 1720 {Opt_removed, "nobh"},
1721 {Opt_removed, "bh"},
ac27a0ec 1722 {Opt_commit, "commit=%u"},
30773840
TT		 1723 {Opt_min_batch_time, "min_batch_time=%u"},
1724 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1725 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1726 {Opt_journal_path, "journal_path=%s"},
818d276c 1727 {Opt_journal_checksum, "journal_checksum"},
c6d3d56d 1728 {Opt_nojournal_checksum, "nojournal_checksum"},
818d276c 1729 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1730 {Opt_abort, "abort"},
1731 {Opt_data_journal, "data=journal"},
1732 {Opt_data_ordered, "data=ordered"},
1733 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1734 {Opt_data_err_abort, "data_err=abort"},
1735 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1736 {Opt_offusrjquota, "usrjquota="},
1737 {Opt_usrjquota, "usrjquota=%s"},
1738 {Opt_offgrpjquota, "grpjquota="},
1739 {Opt_grpjquota, "grpjquota=%s"},
1740 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1741 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1742 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1743 {Opt_grpquota, "grpquota"},
1744 {Opt_noquota, "noquota"},
1745 {Opt_quota, "quota"},
1746 {Opt_usrquota, "usrquota"},
49da9392 1747 {Opt_prjquota, "prjquota"},
ac27a0ec 1748 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1749 {Opt_barrier, "barrier"},
1750 {Opt_nobarrier, "nobarrier"},
25ec56b5 1751 {Opt_i_version, "i_version"},
923ae0ff 1752 {Opt_dax, "dax"},
9cb20f94
IW		 1753 {Opt_dax_always, "dax=always"},
1754 {Opt_dax_inode, "dax=inode"},
1755 {Opt_dax_never, "dax=never"},
c9de560d 1756 {Opt_stripe, "stripe=%u"},
64769240 1757 {Opt_delalloc, "delalloc"},
327eaf73
TT		 1758 {Opt_warn_on_error, "warn_on_error"},
1759 {Opt_nowarn_on_error, "nowarn_on_error"},
a26f4992
TT		 1760 {Opt_lazytime, "lazytime"},
1761 {Opt_nolazytime, "nolazytime"},
670e9875 1762 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
dd919b98 1763 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1764 {Opt_removed, "mblk_io_submit"},
1765 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1766 {Opt_block_validity, "block_validity"},
1767 {Opt_noblock_validity, "noblock_validity"},
240799cd 1768 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1769 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1770 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1771 {Opt_auto_da_alloc, "auto_da_alloc"},
1772 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc 1773 {Opt_dioread_nolock, "dioread_nolock"},
244adf64 1774 {Opt_dioread_lock, "nodioread_nolock"},
744692dc 1775 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1776 {Opt_discard, "discard"},
1777 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1778 {Opt_init_itable, "init_itable=%u"},
1779 {Opt_init_itable, "init_itable"},
1780 {Opt_noinit_itable, "noinit_itable"},
8016e29f 1781#ifdef CONFIG_EXT4_DEBUG
99c880de 1782 {Opt_fc_debug_force, "fc_debug_force"},
8016e29f
HS		 1783 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1784#endif
df981d03 1785 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
ed318a6c 1786 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
6ddb2447 1787 {Opt_test_dummy_encryption, "test_dummy_encryption"},
4f74d15f 1788 {Opt_inlinecrypt, "inlinecrypt"},
cdb7ee4c
TE		 1789 {Opt_nombcache, "nombcache"},
1790 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
3d392b26 1791 {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"},
c7198b9c
TT		 1792 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1793 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1794 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1795 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1796 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1797 {Opt_err, NULL},
ac27a0ec
DK		 1798};
1799
617ba13b 1800static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1801{
617ba13b 1802 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1803 char *options = (char *) *data;
1804
1805 if (!options || strncmp(options, "sb=", 3) != 0)
1806 return 1; /* Default location */
0b8e58a1 1807
ac27a0ec 1808 options += 3;
0b8e58a1 1809 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1810 sb_block = simple_strtoul(options, &options, 0);
1811 if (*options && *options != ',') {
4776004f 1812 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1813 (char *) *data);
1814 return 1;
1815 }
1816 if (*options == ',')
1817 options++;
1818 *data = (void *) options;
0b8e58a1 1819
ac27a0ec
DK		 1820 return sb_block;
1821}
1822
b3881f74 1823#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
d6006186
EB		 1824static const char deprecated_msg[] =
1825 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1826 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1827
56c50f11
DM		 1828#ifdef CONFIG_QUOTA
1829static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1830{
1831 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 1832 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
03dafb5f 1833 int ret = -1;
56c50f11 1834
33458eab 1835 if (sb_any_quota_loaded(sb) && !old_qname) {
56c50f11
DM		 1836 ext4_msg(sb, KERN_ERR,
1837 "Cannot change journaled "
1838 "quota options when quota turned on");
57f73c2c 1839 return -1;
56c50f11 1840 }
e2b911c5 1841 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 1842 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1843 "ignored when QUOTA feature is enabled");
1844 return 1;
262b4662 1845 }
56c50f11
DM		 1846 qname = match_strdup(args);
1847 if (!qname) {
1848 ext4_msg(sb, KERN_ERR,
1849 "Not enough memory for storing quotafile name");
57f73c2c 1850 return -1;
56c50f11 1851 }
33458eab
TT		 1852 if (old_qname) {
1853 if (strcmp(old_qname, qname) == 0)
03dafb5f
CG		 1854 ret = 1;
1855 else
1856 ext4_msg(sb, KERN_ERR,
1857 "%s quota file already specified",
1858 QTYPE2NAME(qtype));
1859 goto errout;
56c50f11 1860 }
03dafb5f 1861 if (strchr(qname, '/')) {
56c50f11
DM		 1862 ext4_msg(sb, KERN_ERR,
1863 "quotafile must be on filesystem root");
03dafb5f 1864 goto errout;
56c50f11 1865 }
33458eab 1866 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
fd8c37ec 1867 set_opt(sb, QUOTA);
56c50f11 1868 return 1;
03dafb5f
CG		 1869errout:
1870 kfree(qname);
1871 return ret;
56c50f11
DM		 1872}
1873
1874static int clear_qf_name(struct super_block *sb, int qtype)
1875{
1876
1877 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 1878 char *old_qname = get_qf_name(sb, sbi, qtype);
56c50f11 1879
33458eab 1880 if (sb_any_quota_loaded(sb) && old_qname) {
56c50f11
DM		 1881 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1882 " when quota turned on");
57f73c2c 1883 return -1;
56c50f11 1884 }
33458eab
TT		 1885 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1886 synchronize_rcu();
1887 kfree(old_qname);
56c50f11
DM		 1888 return 1;
1889}
1890#endif
1891
26092bf5
TT		 1892#define MOPT_SET 0x0001
1893#define MOPT_CLEAR 0x0002
1894#define MOPT_NOSUPPORT 0x0004
1895#define MOPT_EXPLICIT 0x0008
1896#define MOPT_CLEAR_ERR 0x0010
1897#define MOPT_GTE0 0x0020
ac27a0ec 1898#ifdef CONFIG_QUOTA
26092bf5
TT		 1899#define MOPT_Q 0
1900#define MOPT_QFMT 0x0040
1901#else
1902#define MOPT_Q MOPT_NOSUPPORT
1903#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1904#endif
26092bf5 1905#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1906#define MOPT_NO_EXT2 0x0100
1907#define MOPT_NO_EXT3 0x0200
1908#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1909#define MOPT_STRING 0x0400
9cb20f94 1910#define MOPT_SKIP 0x0800
995a3ed6 1911#define MOPT_2 0x1000
26092bf5
TT		 1912
1913static const struct mount_opts {
1914 int token;
1915 int mount_opt;
1916 int flags;
1917} ext4_mount_opts[] = {
1918 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1919 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1920 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1921 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1922 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1923 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1924 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1925 MOPT_EXT4_ONLY | MOPT_SET},
1926 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1927 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1928 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1929 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1930 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1931 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1932 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1933 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1934 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1935 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
c6d3d56d
DW		 1936 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1937 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1938 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1939 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1940 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1941 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1942 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1943 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1944 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1945 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1946 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c 1947 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1948 MOPT_NO_EXT2},
8dc0aa8c 1949 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
7915a861 1950 MOPT_NO_EXT2},
26092bf5
TT		 1951 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1952 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1953 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1954 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1955 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1956 {Opt_commit, 0, MOPT_GTE0},
1957 {Opt_max_batch_time, 0, MOPT_GTE0},
1958 {Opt_min_batch_time, 0, MOPT_GTE0},
1959 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1960 {Opt_init_itable, 0, MOPT_GTE0},
9cb20f94
IW		 1961 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1962 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1963 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1964 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1965 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1966 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1967 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
26092bf5 1968 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1969 {Opt_resuid, 0, MOPT_GTE0},
1970 {Opt_resgid, 0, MOPT_GTE0},
5ba92bcf
CM		 1971 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1972 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1973 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
8dc0aa8c
TT		 1974 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1975 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1976 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1977 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1978 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1979 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1980#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1981 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1982 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1983#else
26092bf5
TT		 1984 {Opt_acl, 0, MOPT_NOSUPPORT},
1985 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1986#endif
26092bf5
TT		 1987 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1988 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
670e9875 1989 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
26092bf5
TT		 1990 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1991 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1992 MOPT_SET | MOPT_Q},
1993 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1994 MOPT_SET | MOPT_Q},
49da9392
JK		 1995 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1996 MOPT_SET | MOPT_Q},
26092bf5 1997 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1998 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1999 MOPT_CLEAR | MOPT_Q},
174fe5ba
KX		 2000 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
2001 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
26092bf5
TT		 2002 {Opt_offusrjquota, 0, MOPT_Q},
2003 {Opt_offgrpjquota, 0, MOPT_Q},
2004 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
2005 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
2006 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 2007 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
ed318a6c 2008 {Opt_test_dummy_encryption, 0, MOPT_STRING},
cdb7ee4c 2009 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
3d392b26
TT		 2010 {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS,
2011 MOPT_SET},
99c880de 2012#ifdef CONFIG_EXT4_DEBUG
0f0672ff
HS		 2013 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2014 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
8016e29f
HS		 2015 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2016#endif
26092bf5
TT		 2017 {Opt_err, 0, 0}
2018};
2019
c83ad55e
GKB		 2020#ifdef CONFIG_UNICODE
2021static const struct ext4_sb_encodings {
2022 __u16 magic;
2023 char *name;
2024 char *version;
2025} ext4_sb_encoding_map[] = {
2026 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2027};
2028
2029static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2030 const struct ext4_sb_encodings **encoding,
2031 __u16 *flags)
2032{
2033 __u16 magic = le16_to_cpu(es->s_encoding);
2034 int i;
2035
2036 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2037 if (magic == ext4_sb_encoding_map[i].magic)
2038 break;
2039
2040 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2041 return -EINVAL;
2042
2043 *encoding = &ext4_sb_encoding_map[i];
2044 *flags = le16_to_cpu(es->s_encoding_flags);
2045
2046 return 0;
2047}
2048#endif
2049
ed318a6c
EB		 2050static int ext4_set_test_dummy_encryption(struct super_block *sb,
2051 const char *opt,
2052 const substring_t *arg,
2053 bool is_remount)
2054{
2055#ifdef CONFIG_FS_ENCRYPTION
2056 struct ext4_sb_info *sbi = EXT4_SB(sb);
2057 int err;
2058
2059 /*
2060 * This mount option is just for testing, and it's not worthwhile to
2061 * implement the extra complexity (e.g. RCU protection) that would be
2062 * needed to allow it to be set or changed during remount. We do allow
2063 * it to be specified during remount, but only if there is no change.
2064 */
ac4acb1f 2065 if (is_remount && !sbi->s_dummy_enc_policy.policy) {
ed318a6c
EB		 2066 ext4_msg(sb, KERN_WARNING,
2067 "Can't set test_dummy_encryption on remount");
2068 return -1;
2069 }
c8c868ab 2070 err = fscrypt_set_test_dummy_encryption(sb, arg->from,
ac4acb1f 2071 &sbi->s_dummy_enc_policy);
ed318a6c
EB		 2072 if (err) {
2073 if (err == -EEXIST)
2074 ext4_msg(sb, KERN_WARNING,
2075 "Can't change test_dummy_encryption on remount");
2076 else if (err == -EINVAL)
2077 ext4_msg(sb, KERN_WARNING,
2078 "Value of option \"%s\" is unrecognized", opt);
2079 else
2080 ext4_msg(sb, KERN_WARNING,
2081 "Error processing option \"%s\" [%d]",
2082 opt, err);
2083 return -1;
2084 }
2085 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2086#else
2087 ext4_msg(sb, KERN_WARNING,
2088 "Test dummy encryption mount option ignored");
2089#endif
2090 return 1;
2091}
2092
26092bf5
TT		 2093static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2094 substring_t *args, unsigned long *journal_devnum,
2095 unsigned int *journal_ioprio, int is_remount)
2096{
2097 struct ext4_sb_info *sbi = EXT4_SB(sb);
2098 const struct mount_opts *m;
08cefc7a
EB		 2099 kuid_t uid;
2100 kgid_t gid;
26092bf5
TT		 2101 int arg = 0;
2102
57f73c2c
TT		 2103#ifdef CONFIG_QUOTA
2104 if (token == Opt_usrjquota)
2105 return set_qf_name(sb, USRQUOTA, &args[0]);
2106 else if (token == Opt_grpjquota)
2107 return set_qf_name(sb, GRPQUOTA, &args[0]);
2108 else if (token == Opt_offusrjquota)
2109 return clear_qf_name(sb, USRQUOTA);
2110 else if (token == Opt_offgrpjquota)
2111 return clear_qf_name(sb, GRPQUOTA);
2112#endif
26092bf5 2113 switch (token) {
f7048605
TT		 2114 case Opt_noacl:
2115 case Opt_nouser_xattr:
2116 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2117 break;
26092bf5
TT		 2118 case Opt_sb:
2119 return 1; /* handled by get_sb_block() */
2120 case Opt_removed:
5f3633e3 2121 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 2122 return 1;
26092bf5 2123 case Opt_abort:
9b5f6c9b 2124 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
26092bf5
TT		 2125 return 1;
2126 case Opt_i_version:
357fdad0 2127 sb->s_flags |= SB_I_VERSION;
26092bf5 2128 return 1;
a26f4992 2129 case Opt_lazytime:
1751e8a6 2130 sb->s_flags |= SB_LAZYTIME;
a26f4992
TT		 2131 return 1;
2132 case Opt_nolazytime:
1751e8a6 2133 sb->s_flags &= ~SB_LAZYTIME;
a26f4992 2134 return 1;
4f74d15f
EB		 2135 case Opt_inlinecrypt:
2136#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2137 sb->s_flags |= SB_INLINECRYPT;
2138#else
2139 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2140#endif
2141 return 1;
26092bf5
TT		 2142 }
2143
5f3633e3
JK		 2144 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2145 if (token == m->token)
2146 break;
2147
2148 if (m->token == Opt_err) {
2149 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2150 "or missing value", opt);
2151 return -1;
2152 }
2153
8dc0aa8c
TT		 2154 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2155 ext4_msg(sb, KERN_ERR,
2156 "Mount option \"%s\" incompatible with ext2", opt);
2157 return -1;
2158 }
2159 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2160 ext4_msg(sb, KERN_ERR,
2161 "Mount option \"%s\" incompatible with ext3", opt);
2162 return -1;
2163 }
2164
ad4eec61 2165 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 2166 return -1;
2167 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2168 return -1;
c93cf2d7
DM		 2169 if (m->flags & MOPT_EXPLICIT) {
2170 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2171 set_opt2(sb, EXPLICIT_DELALLOC);
1e381f60
DM		 2172 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2173 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
c93cf2d7
DM		 2174 } else
2175 return -1;
2176 }
5f3633e3
JK		 2177 if (m->flags & MOPT_CLEAR_ERR)
2178 clear_opt(sb, ERRORS_MASK);
2179 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2180 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2181 "options when quota turned on");
2182 return -1;
2183 }
2184
2185 if (m->flags & MOPT_NOSUPPORT) {
2186 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2187 } else if (token == Opt_commit) {
2188 if (arg == 0)
2189 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
9ba55543 2190 else if (arg > INT_MAX / HZ) {
2191 ext4_msg(sb, KERN_ERR,
2192 "Invalid commit interval %d, "
2193 "must be smaller than %d",
2194 arg, INT_MAX / HZ);
2195 return -1;
2196 }
5f3633e3 2197 sbi->s_commit_interval = HZ * arg;
670e9875 2198 } else if (token == Opt_debug_want_extra_isize) {
9803387c
TT		 2199 if ((arg & 1) ||
2200 (arg < 4) ||
2201 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2202 ext4_msg(sb, KERN_ERR,
2203 "Invalid want_extra_isize %d", arg);
2204 return -1;
2205 }
670e9875 2206 sbi->s_want_extra_isize = arg;
5f3633e3 2207 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 2208 sbi->s_max_batch_time = arg;
2209 } else if (token == Opt_min_batch_time) {
2210 sbi->s_min_batch_time = arg;
2211 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 2212 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2213 ext4_msg(sb, KERN_ERR,
2214 "EXT4-fs: inode_readahead_blks must be "
2215 "0 or a power of 2 smaller than 2^31");
26092bf5 2216 return -1;
5f3633e3
JK		 2217 }
2218 sbi->s_inode_readahead_blks = arg;
2219 } else if (token == Opt_init_itable) {
2220 set_opt(sb, INIT_INODE_TABLE);
2221 if (!args->from)
2222 arg = EXT4_DEF_LI_WAIT_MULT;
2223 sbi->s_li_wait_mult = arg;
2224 } else if (token == Opt_max_dir_size_kb) {
2225 sbi->s_max_dir_size_kb = arg;
8016e29f
HS		 2226#ifdef CONFIG_EXT4_DEBUG
2227 } else if (token == Opt_fc_debug_max_replay) {
2228 sbi->s_fc_debug_max_replay = arg;
2229#endif
5f3633e3
JK		 2230 } else if (token == Opt_stripe) {
2231 sbi->s_stripe = arg;
2232 } else if (token == Opt_resuid) {
2233 uid = make_kuid(current_user_ns(), arg);
2234 if (!uid_valid(uid)) {
2235 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 2236 return -1;
2237 }
5f3633e3
JK		 2238 sbi->s_resuid = uid;
2239 } else if (token == Opt_resgid) {
2240 gid = make_kgid(current_user_ns(), arg);
2241 if (!gid_valid(gid)) {
2242 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2243 return -1;
2244 }
2245 sbi->s_resgid = gid;
2246 } else if (token == Opt_journal_dev) {
2247 if (is_remount) {
2248 ext4_msg(sb, KERN_ERR,
2249 "Cannot specify journal on remount");
2250 return -1;
2251 }
2252 *journal_devnum = arg;
ad4eec61
ES		 2253 } else if (token == Opt_journal_path) {
2254 char *journal_path;
2255 struct inode *journal_inode;
2256 struct path path;
2257 int error;
2258
2259 if (is_remount) {
2260 ext4_msg(sb, KERN_ERR,
2261 "Cannot specify journal on remount");
2262 return -1;
2263 }
2264 journal_path = match_strdup(&args[0]);
2265 if (!journal_path) {
2266 ext4_msg(sb, KERN_ERR, "error: could not dup "
2267 "journal device string");
2268 return -1;
2269 }
2270
2271 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2272 if (error) {
2273 ext4_msg(sb, KERN_ERR, "error: could not find "
2274 "journal device path: error %d", error);
2275 kfree(journal_path);
2276 return -1;
2277 }
2278
2b0143b5 2279 journal_inode = d_inode(path.dentry);
ad4eec61
ES		 2280 if (!S_ISBLK(journal_inode->i_mode)) {
2281 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2282 "is not a block device", journal_path);
2283 path_put(&path);
2284 kfree(journal_path);
2285 return -1;
2286 }
2287
2288 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2289 path_put(&path);
2290 kfree(journal_path);
5f3633e3
JK		 2291 } else if (token == Opt_journal_ioprio) {
2292 if (arg > 7) {
2293 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2294 " (must be 0-7)");
2295 return -1;
2296 }
2297 *journal_ioprio =
2298 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
6ddb2447 2299 } else if (token == Opt_test_dummy_encryption) {
ed318a6c
EB		 2300 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2301 is_remount);
5f3633e3
JK		 2302 } else if (m->flags & MOPT_DATAJ) {
2303 if (is_remount) {
2304 if (!sbi->s_journal)
2305 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2306 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 2307 ext4_msg(sb, KERN_ERR,
26092bf5 2308 "Cannot change data mode on remount");
26092bf5 2309 return -1;
ac27a0ec 2310 }
26092bf5 2311 } else {
5f3633e3
JK		 2312 clear_opt(sb, DATA_FLAGS);
2313 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 2314 }
5f3633e3
JK		 2315#ifdef CONFIG_QUOTA
2316 } else if (m->flags & MOPT_QFMT) {
2317 if (sb_any_quota_loaded(sb) &&
2318 sbi->s_jquota_fmt != m->mount_opt) {
2319 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2320 "quota options when quota turned on");
2321 return -1;
2322 }
e2b911c5 2323 if (ext4_has_feature_quota(sb)) {
c325a67c
TT		 2324 ext4_msg(sb, KERN_INFO,
2325 "Quota format mount options ignored "
262b4662 2326 "when QUOTA feature is enabled");
c325a67c 2327 return 1;
262b4662 2328 }
5f3633e3 2329 sbi->s_jquota_fmt = m->mount_opt;
923ae0ff 2330#endif
9cb20f94
IW		 2331 } else if (token == Opt_dax || token == Opt_dax_always ||
2332 token == Opt_dax_inode || token == Opt_dax_never) {
ef83b6e8 2333#ifdef CONFIG_FS_DAX
9cb20f94
IW		 2334 switch (token) {
2335 case Opt_dax:
2336 case Opt_dax_always:
829b37b8
TT		 2337 if (is_remount &&
2338 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2339 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2340 fail_dax_change_remount:
2341 ext4_msg(sb, KERN_ERR, "can't change "
2342 "dax mount option while remounting");
2343 return -1;
2344 }
2345 if (is_remount &&
2346 (test_opt(sb, DATA_FLAGS) ==
2347 EXT4_MOUNT_JOURNAL_DATA)) {
2348 ext4_msg(sb, KERN_ERR, "can't mount with "
2349 "both data=journal and dax");
2350 return -1;
2351 }
9cb20f94
IW		 2352 ext4_msg(sb, KERN_WARNING,
2353 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2354 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2355 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2356 break;
2357 case Opt_dax_never:
829b37b8
TT		 2358 if (is_remount &&
2359 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2360 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2361 goto fail_dax_change_remount;
9cb20f94
IW		 2362 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2363 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2364 break;
2365 case Opt_dax_inode:
829b37b8
TT		 2366 if (is_remount &&
2367 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2368 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2369 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2370 goto fail_dax_change_remount;
9cb20f94
IW		 2371 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2372 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2373 /* Strictly for printing options */
2374 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2375 break;
2376 }
ef83b6e8 2377#else
923ae0ff 2378 ext4_msg(sb, KERN_INFO, "dax option not supported");
9cb20f94
IW		 2379 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2380 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
923ae0ff 2381 return -1;
5f3633e3 2382#endif
7915a861
AN		 2383 } else if (token == Opt_data_err_abort) {
2384 sbi->s_mount_opt |= m->mount_opt;
2385 } else if (token == Opt_data_err_ignore) {
2386 sbi->s_mount_opt &= ~m->mount_opt;
5f3633e3
JK		 2387 } else {
2388 if (!args->from)
2389 arg = 1;
2390 if (m->flags & MOPT_CLEAR)
2391 arg = !arg;
2392 else if (unlikely(!(m->flags & MOPT_SET))) {
2393 ext4_msg(sb, KERN_WARNING,
2394 "buggy handling of option %s", opt);
2395 WARN_ON(1);
2396 return -1;
2397 }
995a3ed6
HS		 2398 if (m->flags & MOPT_2) {
2399 if (arg != 0)
2400 sbi->s_mount_opt2 |= m->mount_opt;
2401 else
2402 sbi->s_mount_opt2 &= ~m->mount_opt;
2403 } else {
2404 if (arg != 0)
2405 sbi->s_mount_opt |= m->mount_opt;
2406 else
2407 sbi->s_mount_opt &= ~m->mount_opt;
2408 }
26092bf5 2409 }
5f3633e3 2410 return 1;
26092bf5
TT		 2411}
2412
2413static int parse_options(char *options, struct super_block *sb,
2414 unsigned long *journal_devnum,
2415 unsigned int *journal_ioprio,
2416 int is_remount)
2417{
1e1a76ed 2418 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
33458eab 2419 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
26092bf5
TT		 2420 substring_t args[MAX_OPT_ARGS];
2421 int token;
2422
2423 if (!options)
2424 return 1;
2425
2426 while ((p = strsep(&options, ",")) != NULL) {
2427 if (!*p)
2428 continue;
2429 /*
2430 * Initialize args struct so we know whether arg was
2431 * found; some options take optional arguments.
2432 */
caecd0af 2433 args[0].to = args[0].from = NULL;
26092bf5
TT		 2434 token = match_token(p, tokens, args);
2435 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2436 journal_ioprio, is_remount) < 0)
2437 return 0;
ac27a0ec
DK		 2438 }
2439#ifdef CONFIG_QUOTA
49da9392
JK		 2440 /*
2441 * We do the test below only for project quotas. 'usrquota' and
2442 * 'grpquota' mount options are allowed even without quota feature
2443 * to support legacy quotas in quota files.
2444 */
2445 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2446 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2447 "Cannot enable project quota enforcement.");
2448 return 0;
2449 }
33458eab
TT		 2450 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2451 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2452 if (usr_qf_name || grp_qf_name) {
2453 if (test_opt(sb, USRQUOTA) && usr_qf_name)
fd8c37ec 2454 clear_opt(sb, USRQUOTA);
ac27a0ec 2455
33458eab 2456 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
fd8c37ec 2457 clear_opt(sb, GRPQUOTA);
ac27a0ec 2458
56c50f11 2459 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 2460 ext4_msg(sb, KERN_ERR, "old and new quota "
2461 "format mixing");
ac27a0ec
DK		 2462 return 0;
2463 }
2464
2465 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 2466 ext4_msg(sb, KERN_ERR, "journaled quota format "
2467 "not specified");
ac27a0ec
DK		 2468 return 0;
2469 }
ac27a0ec
DK		 2470 }
2471#endif
626b035b
RH		 2472 if (test_opt(sb, DIOREAD_NOLOCK)) {
2473 int blocksize =
2474 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2475 if (blocksize < PAGE_SIZE)
2476 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2477 "experimental mount option 'dioread_nolock' "
2478 "for blocksize < PAGE_SIZE");
2479 }
ac27a0ec
DK		 2480 return 1;
2481}
2482
2adf6da8
TT		 2483static inline void ext4_show_quota_options(struct seq_file *seq,
2484 struct super_block *sb)
2485{
2486#if defined(CONFIG_QUOTA)
2487 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 2488 char *usr_qf_name, *grp_qf_name;
2adf6da8
TT		 2489
2490 if (sbi->s_jquota_fmt) {
2491 char *fmtname = "";
2492
2493 switch (sbi->s_jquota_fmt) {
2494 case QFMT_VFS_OLD:
2495 fmtname = "vfsold";
2496 break;
2497 case QFMT_VFS_V0:
2498 fmtname = "vfsv0";
2499 break;
2500 case QFMT_VFS_V1:
2501 fmtname = "vfsv1";
2502 break;
2503 }
2504 seq_printf(seq, ",jqfmt=%s", fmtname);
2505 }
2506
33458eab
TT		 2507 rcu_read_lock();
2508 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2509 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2510 if (usr_qf_name)
2511 seq_show_option(seq, "usrjquota", usr_qf_name);
2512 if (grp_qf_name)
2513 seq_show_option(seq, "grpjquota", grp_qf_name);
2514 rcu_read_unlock();
2adf6da8
TT		 2515#endif
2516}
2517
5a916be1
TT		 2518static const char *token2str(int token)
2519{
50df9fd5 2520 const struct match_token *t;
5a916be1
TT		 2521
2522 for (t = tokens; t->token != Opt_err; t++)
2523 if (t->token == token && !strchr(t->pattern, '='))
2524 break;
2525 return t->pattern;
2526}
2527
2adf6da8
TT		 2528/*
2529 * Show an option if
2530 * - it's set to a non-default value OR
2531 * - if the per-sb default is different from the global default
2532 */
66acdcf4
TT		 2533static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2534 int nodefs)
2adf6da8 2535{
2adf6da8
TT		 2536 struct ext4_sb_info *sbi = EXT4_SB(sb);
2537 struct ext4_super_block *es = sbi->s_es;
68afa7e0 2538 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
5a916be1 2539 const struct mount_opts *m;
66acdcf4 2540 char sep = nodefs ? '\n' : ',';
2adf6da8 2541
66acdcf4
TT		 2542#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2543#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2544
2545 if (sbi->s_sb_block != 1)
5a916be1
TT		 2546 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2547
2548 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2549 int want_set = m->flags & MOPT_SET;
2550 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
9cb20f94 2551 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
5a916be1 2552 continue;
68afa7e0 2553 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2554 continue; /* skip if same as the default */
2555 if ((want_set &&
2556 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2557 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2558 continue; /* select Opt_noFoo vs Opt_Foo */
2559 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2560 }
5a916be1 2561
08cefc7a 2562 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2563 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2564 SEQ_OPTS_PRINT("resuid=%u",
2565 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2566 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2567 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2568 SEQ_OPTS_PRINT("resgid=%u",
2569 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2570 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2571 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2572 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2573 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2574 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2575 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2576 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2577 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2578 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2579 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2580 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2581 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2582 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
357fdad0 2583 if (sb->s_flags & SB_I_VERSION)
5a916be1 2584 SEQ_OPTS_PUTS("i_version");
66acdcf4 2585 if (nodefs || sbi->s_stripe)
5a916be1 2586 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0
TN		 2587 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2588 (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2589 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2590 SEQ_OPTS_PUTS("data=journal");
2591 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2592 SEQ_OPTS_PUTS("data=ordered");
2593 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2594 SEQ_OPTS_PUTS("data=writeback");
2595 }
66acdcf4
TT		 2596 if (nodefs ||
2597 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2598 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2599 sbi->s_inode_readahead_blks);
2adf6da8 2600
ceec0376 2601 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 2602 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2603 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2604 if (nodefs || sbi->s_max_dir_size_kb)
2605 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2606 if (test_opt(sb, DATA_ERR_ABORT))
2607 SEQ_OPTS_PUTS("data_err=abort");
ed318a6c
EB		 2608
2609 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2adf6da8 2610
4f74d15f
EB		 2611 if (sb->s_flags & SB_INLINECRYPT)
2612 SEQ_OPTS_PUTS("inlinecrypt");
2613
9cb20f94
IW		 2614 if (test_opt(sb, DAX_ALWAYS)) {
2615 if (IS_EXT2_SB(sb))
2616 SEQ_OPTS_PUTS("dax");
2617 else
2618 SEQ_OPTS_PUTS("dax=always");
2619 } else if (test_opt2(sb, DAX_NEVER)) {
2620 SEQ_OPTS_PUTS("dax=never");
2621 } else if (test_opt2(sb, DAX_INODE)) {
2622 SEQ_OPTS_PUTS("dax=inode");
2623 }
2adf6da8 2624 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2625 return 0;
2626}
2627
66acdcf4
TT		 2628static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2629{
2630 return _ext4_show_options(seq, root->d_sb, 0);
2631}
2632
ebd173be 2633int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 2634{
2635 struct super_block *sb = seq->private;
2636 int rc;
2637
bc98a42c 2638 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 2639 rc = _ext4_show_options(seq, sb, 1);
2640 seq_puts(seq, "\n");
2641 return rc;
2642}
2643
617ba13b 2644static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 2645 int read_only)
2646{
617ba13b 2647 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 2648 int err = 0;
ac27a0ec 2649
617ba13b 2650 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 2651 ext4_msg(sb, KERN_ERR, "revision level too high, "
2652 "forcing read-only mode");
c89128a0 2653 err = -EROFS;
5adaccac 2654 goto done;
ac27a0ec
DK		 2655 }
2656 if (read_only)
281b5995 2657 goto done;
617ba13b 2658 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 2659 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2660 "running e2fsck is recommended");
c8b459f4 2661 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 2662 ext4_msg(sb, KERN_WARNING,
2663 "warning: mounting fs with errors, "
2664 "running e2fsck is recommended");
ed3ce80a 2665 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 2666 le16_to_cpu(es->s_mnt_count) >=
2667 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 2668 ext4_msg(sb, KERN_WARNING,
2669 "warning: maximal mount count reached, "
2670 "running e2fsck is recommended");
ac27a0ec 2671 else if (le32_to_cpu(es->s_checkinterval) &&
6a0678a7
AB		 2672 (ext4_get_tstamp(es, s_lastcheck) +
2673 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
b31e1552
ES		 2674 ext4_msg(sb, KERN_WARNING,
2675 "warning: checktime reached, "
2676 "running e2fsck is recommended");
0b8e58a1 2677 if (!sbi->s_journal)
0390131b 2678 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 2679 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 2680 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 2681 le16_add_cpu(&es->s_mnt_count, 1);
6a0678a7 2682 ext4_update_tstamp(es, s_mtime);
0390131b 2683 if (sbi->s_journal)
e2b911c5 2684 ext4_set_feature_journal_needs_recovery(sb);
ac27a0ec 2685
4392fbc4 2686 err = ext4_commit_super(sb);
281b5995 2687done:
ac27a0ec 2688 if (test_opt(sb, DEBUG))
a9df9a49 2689 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 2690 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 2691 sb->s_blocksize,
2692 sbi->s_groups_count,
617ba13b
MC		 2693 EXT4_BLOCKS_PER_GROUP(sb),
2694 EXT4_INODES_PER_GROUP(sb),
a2595b8a 2695 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 2696
7abc52c2 2697 cleancache_init_fs(sb);
c89128a0 2698 return err;
ac27a0ec
DK		 2699}
2700
117fff10
TT		 2701int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2702{
2703 struct ext4_sb_info *sbi = EXT4_SB(sb);
7c990728 2704 struct flex_groups **old_groups, **new_groups;
37b0b6b8 2705 int size, i, j;
117fff10
TT		 2706
2707 if (!sbi->s_log_groups_per_flex)
2708 return 0;
2709
2710 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2711 if (size <= sbi->s_flex_groups_allocated)
2712 return 0;
2713
7c990728
SJS		 2714 new_groups = kvzalloc(roundup_pow_of_two(size *
2715 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
117fff10 2716 if (!new_groups) {
7c990728
SJS		 2717 ext4_msg(sb, KERN_ERR,
2718 "not enough memory for %d flex group pointers", size);
117fff10
TT		 2719 return -ENOMEM;
2720 }
7c990728
SJS		 2721 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2722 new_groups[i] = kvzalloc(roundup_pow_of_two(
2723 sizeof(struct flex_groups)),
2724 GFP_KERNEL);
2725 if (!new_groups[i]) {
37b0b6b8
DC		 2726 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2727 kvfree(new_groups[j]);
7c990728
SJS		 2728 kvfree(new_groups);
2729 ext4_msg(sb, KERN_ERR,
2730 "not enough memory for %d flex groups", size);
2731 return -ENOMEM;
2732 }
117fff10 2733 }
7c990728
SJS		 2734 rcu_read_lock();
2735 old_groups = rcu_dereference(sbi->s_flex_groups);
2736 if (old_groups)
2737 memcpy(new_groups, old_groups,
2738 (sbi->s_flex_groups_allocated *
2739 sizeof(struct flex_groups *)));
2740 rcu_read_unlock();
2741 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2742 sbi->s_flex_groups_allocated = size;
2743 if (old_groups)
2744 ext4_kvfree_array_rcu(old_groups);
117fff10
TT		 2745 return 0;
2746}
2747
772cb7c8
JS		 2748static int ext4_fill_flex_info(struct super_block *sb)
2749{
2750 struct ext4_sb_info *sbi = EXT4_SB(sb);
2751 struct ext4_group_desc *gdp = NULL;
7c990728 2752 struct flex_groups *fg;
772cb7c8 2753 ext4_group_t flex_group;
117fff10 2754 int i, err;
772cb7c8 2755
503358ae 2756 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 2757 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 2758 sbi->s_log_groups_per_flex = 0;
2759 return 1;
2760 }
2761
117fff10
TT		 2762 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2763 if (err)
9933fc0a 2764 goto failed;
772cb7c8 2765
772cb7c8 2766 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2767 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2768
2769 flex_group = ext4_flex_group(sbi, i);
7c990728
SJS		 2770 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2771 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
90ba983f 2772 atomic64_add(ext4_free_group_clusters(sb, gdp),
7c990728
SJS		 2773 &fg->free_clusters);
2774 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
772cb7c8
JS		 2775 }
2776
2777 return 1;
2778failed:
2779 return 0;
2780}
2781
e2b911c5 2782static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 2783 struct ext4_group_desc *gdp)
717d50e4 2784{
b47820ed 2785 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 2786 __u16 crc = 0;
feb0ab32 2787 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 2788 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 2789
9aa5d32b 2790 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 2791 /* Use new metadata_csum algorithm */
feb0ab32 2792 __u32 csum32;
b47820ed 2793 __u16 dummy_csum = 0;
feb0ab32 2794
feb0ab32
DW		 2795 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2796 sizeof(le_group));
b47820ed
DJ		 2797 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2798 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2799 sizeof(dummy_csum));
2800 offset += sizeof(dummy_csum);
2801 if (offset < sbi->s_desc_size)
2802 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2803 sbi->s_desc_size - offset);
feb0ab32
DW		 2804
2805 crc = csum32 & 0xFFFF;
2806 goto out;
717d50e4
AD		 2807 }
2808
feb0ab32 2809 /* old crc16 code */
e2b911c5 2810 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 2811 return 0;
2812
feb0ab32
DW		 2813 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2814 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2815 crc = crc16(crc, (__u8 *)gdp, offset);
2816 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2817 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 2818 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 2819 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2820 crc = crc16(crc, (__u8 *)gdp + offset,
2821 le16_to_cpu(sbi->s_es->s_desc_size) -
2822 offset);
2823
2824out:
717d50e4
AD		 2825 return cpu_to_le16(crc);
2826}
2827
feb0ab32 2828int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2829 struct ext4_group_desc *gdp)
2830{
feb0ab32 2831 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 2832 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 2833 return 0;
2834
2835 return 1;
2836}
2837
feb0ab32
DW		 2838void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2839 struct ext4_group_desc *gdp)
2840{
2841 if (!ext4_has_group_desc_csum(sb))
2842 return;
e2b911c5 2843 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 2844}
2845
ac27a0ec 2846/* Called at mount-time, super-block is locked */
bfff6873 2847static int ext4_check_descriptors(struct super_block *sb,
829fa70d 2848 ext4_fsblk_t sb_block,
bfff6873 2849 ext4_group_t *first_not_zeroed)
ac27a0ec 2850{
617ba13b
MC		 2851 struct ext4_sb_info *sbi = EXT4_SB(sb);
2852 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2853 ext4_fsblk_t last_block;
44de022c 2854 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
bd81d8ee
LV		 2855 ext4_fsblk_t block_bitmap;
2856 ext4_fsblk_t inode_bitmap;
2857 ext4_fsblk_t inode_table;
ce421581 2858 int flexbg_flag = 0;
bfff6873 2859 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2860
e2b911c5 2861 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 2862 flexbg_flag = 1;
2863
af5bc92d 2864 ext4_debug("Checking group descriptors");
ac27a0ec 2865
197cd65a
AM		 2866 for (i = 0; i < sbi->s_groups_count; i++) {
2867 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2868
ce421581 2869 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2870 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2871 else
2872 last_block = first_block +
617ba13b 2873 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2874
bfff6873
LC		 2875 if ((grp == sbi->s_groups_count) &&
2876 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2877 grp = i;
2878
8fadc143 2879 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 2880 if (block_bitmap == sb_block) {
2881 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2882 "Block bitmap for group %u overlaps "
2883 "superblock", i);
18db4b4e
TT		 2884 if (!sb_rdonly(sb))
2885 return 0;
829fa70d 2886 }
77260807
TT		 2887 if (block_bitmap >= sb_block + 1 &&
2888 block_bitmap <= last_bg_block) {
2889 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2890 "Block bitmap for group %u overlaps "
2891 "block group descriptors", i);
2892 if (!sb_rdonly(sb))
2893 return 0;
2894 }
2b2d6d01 2895 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2896 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2897 "Block bitmap for group %u not in group "
b31e1552 2898 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2899 return 0;
2900 }
8fadc143 2901 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 2902 if (inode_bitmap == sb_block) {
2903 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2904 "Inode bitmap for group %u overlaps "
2905 "superblock", i);
18db4b4e
TT		 2906 if (!sb_rdonly(sb))
2907 return 0;
829fa70d 2908 }
77260807
TT		 2909 if (inode_bitmap >= sb_block + 1 &&
2910 inode_bitmap <= last_bg_block) {
2911 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2912 "Inode bitmap for group %u overlaps "
2913 "block group descriptors", i);
2914 if (!sb_rdonly(sb))
2915 return 0;
2916 }
2b2d6d01 2917 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2918 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2919 "Inode bitmap for group %u not in group "
b31e1552 2920 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2921 return 0;
2922 }
8fadc143 2923 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 2924 if (inode_table == sb_block) {
2925 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2926 "Inode table for group %u overlaps "
2927 "superblock", i);
18db4b4e
TT		 2928 if (!sb_rdonly(sb))
2929 return 0;
829fa70d 2930 }
77260807
TT		 2931 if (inode_table >= sb_block + 1 &&
2932 inode_table <= last_bg_block) {
2933 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2934 "Inode table for group %u overlaps "
2935 "block group descriptors", i);
2936 if (!sb_rdonly(sb))
2937 return 0;
2938 }
bd81d8ee 2939 if (inode_table < first_block ||
2b2d6d01 2940 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2941 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2942 "Inode table for group %u not in group "
b31e1552 2943 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2944 return 0;
2945 }
955ce5f5 2946 ext4_lock_group(sb, i);
feb0ab32 2947 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2948 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2949 "Checksum for group %u failed (%u!=%u)",
e2b911c5 2950 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 2951 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 2952 if (!sb_rdonly(sb)) {
955ce5f5 2953 ext4_unlock_group(sb, i);
8a266467 2954 return 0;
7ee1ec4c 2955 }
717d50e4 2956 }
955ce5f5 2957 ext4_unlock_group(sb, i);
ce421581
JS		 2958 if (!flexbg_flag)
2959 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2960 }
bfff6873
LC		 2961 if (NULL != first_not_zeroed)
2962 *first_not_zeroed = grp;
ac27a0ec
DK		 2963 return 1;
2964}
2965
617ba13b 2966/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2967 * the superblock) which were deleted from all directories, but held open by
2968 * a process at the time of a crash. We walk the list and try to delete these
2969 * inodes at recovery time (only with a read-write filesystem).
2970 *
2971 * In order to keep the orphan inode chain consistent during traversal (in
2972 * case of crash during recovery), we link each inode into the superblock
2973 * orphan list_head and handle it the same way as an inode deletion during
2974 * normal operation (which journals the operations for us).
2975 *
2976 * We only do an iget() and an iput() on each inode, which is very safe if we
2977 * accidentally point at an in-use or already deleted inode. The worst that
2978 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2979 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2980 * e2fsck was run on this filesystem, and it must have already done the orphan
2981 * inode cleanup for us, so we can safely abort without any further action.
2982 */
2b2d6d01
TT		 2983static void ext4_orphan_cleanup(struct super_block *sb,
2984 struct ext4_super_block *es)
ac27a0ec
DK		 2985{
2986 unsigned int s_flags = sb->s_flags;
2c98eb5e 2987 int ret, nr_orphans = 0, nr_truncates = 0;
ac27a0ec 2988#ifdef CONFIG_QUOTA
95f1fda4 2989 int quota_update = 0;
ac27a0ec
DK		 2990 int i;
2991#endif
2992 if (!es->s_last_orphan) {
2993 jbd_debug(4, "no orphan inodes to clean up\n");
2994 return;
2995 }
2996
a8f48a95 2997 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2998 ext4_msg(sb, KERN_ERR, "write access "
2999 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 3000 return;
3001 }
3002
d39195c3
AG		 3003 /* Check if feature set would not allow a r/w mount */
3004 if (!ext4_feature_set_ok(sb, 0)) {
3005 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
3006 "unknown ROCOMPAT features");
3007 return;
3008 }
3009
617ba13b 3010 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6 3011 /* don't clear list on RO mount w/ errors */
1751e8a6 3012 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
84474976 3013 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
ac27a0ec 3014 "clearing orphan list.\n");
c25f9bc6
ES		 3015 es->s_last_orphan = 0;
3016 }
ac27a0ec
DK		 3017 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3018 return;
3019 }
3020
1751e8a6 3021 if (s_flags & SB_RDONLY) {
b31e1552 3022 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
1751e8a6 3023 sb->s_flags &= ~SB_RDONLY;
ac27a0ec
DK		 3024 }
3025#ifdef CONFIG_QUOTA
3026 /* Needed for iput() to work correctly and not trash data */
1751e8a6 3027 sb->s_flags |= SB_ACTIVE;
95f1fda4 3028
3029 /*
3030 * Turn on quotas which were not enabled for read-only mounts if
3031 * filesystem has quota feature, so that they are updated correctly.
3032 */
1751e8a6 3033 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
95f1fda4 3034 int ret = ext4_enable_quotas(sb);
3035
3036 if (!ret)
3037 quota_update = 1;
3038 else
3039 ext4_msg(sb, KERN_ERR,
3040 "Cannot turn on quotas: error %d", ret);
3041 }
3042
3043 /* Turn on journaled quotas used for old sytle */
a2d4a646 3044 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
617ba13b
MC		 3045 if (EXT4_SB(sb)->s_qf_names[i]) {
3046 int ret = ext4_quota_on_mount(sb, i);
95f1fda4 3047
3048 if (!ret)
3049 quota_update = 1;
3050 else
b31e1552
ES		 3051 ext4_msg(sb, KERN_ERR,
3052 "Cannot turn on journaled "
95f1fda4 3053 "quota: type %d: error %d", i, ret);
ac27a0ec
DK		 3054 }
3055 }
3056#endif
3057
3058 while (es->s_last_orphan) {
3059 struct inode *inode;
3060
c65d5c6c
VN		 3061 /*
3062 * We may have encountered an error during cleanup; if
3063 * so, skip the rest.
3064 */
3065 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3066 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3067 es->s_last_orphan = 0;
3068 break;
3069 }
3070
97bd42b9
JB		 3071 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
3072 if (IS_ERR(inode)) {
ac27a0ec
DK		 3073 es->s_last_orphan = 0;
3074 break;
3075 }
3076
617ba13b 3077 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 3078 dquot_initialize(inode);
ac27a0ec 3079 if (inode->i_nlink) {
566370a2
PT		 3080 if (test_opt(sb, DEBUG))
3081 ext4_msg(sb, KERN_DEBUG,
3082 "%s: truncating inode %lu to %lld bytes",
3083 __func__, inode->i_ino, inode->i_size);
e5f8eab8 3084 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 3085 inode->i_ino, inode->i_size);
5955102c 3086 inode_lock(inode);
55f252c9 3087 truncate_inode_pages(inode->i_mapping, inode->i_size);
2c98eb5e
TT		 3088 ret = ext4_truncate(inode);
3089 if (ret)
3090 ext4_std_error(inode->i_sb, ret);
5955102c 3091 inode_unlock(inode);
ac27a0ec
DK		 3092 nr_truncates++;
3093 } else {
566370a2
PT		 3094 if (test_opt(sb, DEBUG))
3095 ext4_msg(sb, KERN_DEBUG,
3096 "%s: deleting unreferenced inode %lu",
3097 __func__, inode->i_ino);
ac27a0ec
DK		 3098 jbd_debug(2, "deleting unreferenced inode %lu\n",
3099 inode->i_ino);
3100 nr_orphans++;
3101 }
3102 iput(inode); /* The delete magic happens here! */
3103 }
3104
2b2d6d01 3105#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 3106
3107 if (nr_orphans)
b31e1552
ES		 3108 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
3109 PLURAL(nr_orphans));
ac27a0ec 3110 if (nr_truncates)
b31e1552
ES		 3111 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3112 PLURAL(nr_truncates));
ac27a0ec 3113#ifdef CONFIG_QUOTA
95f1fda4 3114 /* Turn off quotas if they were enabled for orphan cleanup */
3115 if (quota_update) {
3116 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3117 if (sb_dqopt(sb)->files[i])
3118 dquot_quota_off(sb, i);
3119 }
ac27a0ec
DK		 3120 }
3121#endif
1751e8a6 3122 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
ac27a0ec 3123}
0b8e58a1 3124
cd2291a4
ES		 3125/*
3126 * Maximal extent format file size.
3127 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3128 * extent format containers, within a sector_t, and within i_blocks
3129 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3130 * so that won't be a limiting factor.
3131 *
f17722f9
LC		 3132 * However there is other limiting factor. We do store extents in the form
3133 * of starting block and length, hence the resulting length of the extent
3134 * covering maximum file size must fit into on-disk format containers as
3135 * well. Given that length is always by 1 unit bigger than max unit (because
3136 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3137 *
cd2291a4
ES		 3138 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3139 */
f287a1a5 3140static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 3141{
3142 loff_t res;
3143 loff_t upper_limit = MAX_LFS_FILESIZE;
3144
72deb455
CH		 3145 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3146
3147 if (!has_huge_files) {
cd2291a4
ES		 3148 upper_limit = (1LL << 32) - 1;
3149
3150 /* total blocks in file system block size */
3151 upper_limit >>= (blkbits - 9);
3152 upper_limit <<= blkbits;
3153 }
3154
f17722f9
LC		 3155 /*
3156 * 32-bit extent-start container, ee_block. We lower the maxbytes
3157 * by one fs block, so ee_len can cover the extent of maximum file
3158 * size
3159 */
3160 res = (1LL << 32) - 1;
cd2291a4 3161 res <<= blkbits;
cd2291a4
ES		 3162
3163 /* Sanity check against vm- & vfs- imposed limits */
3164 if (res > upper_limit)
3165 res = upper_limit;
3166
3167 return res;
3168}
ac27a0ec 3169
ac27a0ec 3170/*
cd2291a4 3171 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 3172 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3173 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 3174 */
f287a1a5 3175static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 3176{
617ba13b 3177 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 3178 int meta_blocks;
3179 loff_t upper_limit;
0b8e58a1
AD		 3180 /* This is calculated to be the largest file size for a dense, block
3181 * mapped file such that the file's total number of 512-byte sectors,
3182 * including data and all indirect blocks, does not exceed (2^48 - 1).
3183 *
3184 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3185 * number of 512-byte sectors of the file.
0fc1b451
AK		 3186 */
3187
72deb455 3188 if (!has_huge_files) {
0fc1b451 3189 /*
72deb455
CH		 3190 * !has_huge_files or implies that the inode i_block field
3191 * represents total file blocks in 2^32 512-byte sectors ==
3192 * size of vfs inode i_blocks * 8
0fc1b451
AK		 3193 */
3194 upper_limit = (1LL << 32) - 1;
3195
3196 /* total blocks in file system block size */
3197 upper_limit >>= (bits - 9);
3198
3199 } else {
8180a562
AK		 3200 /*
3201 * We use 48 bit ext4_inode i_blocks
3202 * With EXT4_HUGE_FILE_FL set the i_blocks
3203 * represent total number of blocks in
3204 * file system block size
3205 */
0fc1b451
AK		 3206 upper_limit = (1LL << 48) - 1;
3207
0fc1b451
AK		 3208 }
3209
3210 /* indirect blocks */
3211 meta_blocks = 1;
3212 /* double indirect blocks */
3213 meta_blocks += 1 + (1LL << (bits-2));
3214 /* tripple indirect blocks */
3215 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3216
3217 upper_limit -= meta_blocks;
3218 upper_limit <<= bits;
ac27a0ec
DK		 3219
3220 res += 1LL << (bits-2);
3221 res += 1LL << (2*(bits-2));
3222 res += 1LL << (3*(bits-2));
3223 res <<= bits;
3224 if (res > upper_limit)
3225 res = upper_limit;
0fc1b451
AK		 3226
3227 if (res > MAX_LFS_FILESIZE)
3228 res = MAX_LFS_FILESIZE;
3229
ac27a0ec
DK		 3230 return res;
3231}
3232
617ba13b 3233static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 3234 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 3235{
617ba13b 3236 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 3237 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 3238 int has_super = 0;
3239
3240 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3241
e2b911c5 3242 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 3243 return logical_sb_block + nr + 1;
ac27a0ec 3244 bg = sbi->s_desc_per_block * nr;
617ba13b 3245 if (ext4_bg_has_super(sb, bg))
ac27a0ec 3246 has_super = 1;
0b8e58a1 3247
bd63f6b0
DW		 3248 /*
3249 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3250 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3251 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3252 * compensate.
3253 */
3254 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 3255 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 3256 has_super++;
3257
617ba13b 3258 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 3259}
3260
c9de560d
AT		 3261/**
3262 * ext4_get_stripe_size: Get the stripe size.
3263 * @sbi: In memory super block info
3264 *
3265 * If we have specified it via mount option, then
3266 * use the mount option value. If the value specified at mount time is
3267 * greater than the blocks per group use the super block value.
3268 * If the super block value is greater than blocks per group return 0.
3269 * Allocator needs it be less than blocks per group.
3270 *
3271 */
3272static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3273{
3274 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3275 unsigned long stripe_width =
3276 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 3277 int ret;
c9de560d
AT		 3278
3279 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 3280 ret = sbi->s_stripe;
5469d7c3 3281 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 3282 ret = stripe_width;
5469d7c3 3283 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 3284 ret = stride;
3285 else
3286 ret = 0;
c9de560d 3287
3eb08658
DE		 3288 /*
3289 * If the stripe width is 1, this makes no sense and
3290 * we set it to 0 to turn off stripe handling code.
3291 */
3292 if (ret <= 1)
3293 ret = 0;
c9de560d 3294
3eb08658 3295 return ret;
c9de560d 3296}
ac27a0ec 3297
a13fb1a4
ES		 3298/*
3299 * Check whether this filesystem can be mounted based on
3300 * the features present and the RDONLY/RDWR mount requested.
3301 * Returns 1 if this filesystem can be mounted as requested,
3302 * 0 if it cannot be.
3303 */
3304static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3305{
e2b911c5 3306 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 3307 ext4_msg(sb, KERN_ERR,
3308 "Couldn't mount because of "
3309 "unsupported optional features (%x)",
3310 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3311 ~EXT4_FEATURE_INCOMPAT_SUPP));
3312 return 0;
3313 }
3314
c83ad55e
GKB		 3315#ifndef CONFIG_UNICODE
3316 if (ext4_has_feature_casefold(sb)) {
3317 ext4_msg(sb, KERN_ERR,
3318 "Filesystem with casefold feature cannot be "
3319 "mounted without CONFIG_UNICODE");
3320 return 0;
3321 }
3322#endif
3323
a13fb1a4
ES		 3324 if (readonly)
3325 return 1;
3326
e2b911c5 3327 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 3328 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 3329 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 3330 return 1;
3331 }
3332
a13fb1a4 3333 /* Check that feature set is OK for a read-write mount */
e2b911c5 3334 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 3335 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3336 "unsupported optional features (%x)",
3337 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3338 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3339 return 0;
3340 }
e2b911c5 3341 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 3342 ext4_msg(sb, KERN_ERR,
3343 "Can't support bigalloc feature without "
3344 "extents feature\n");
3345 return 0;
3346 }
7c319d32 3347
9db176bc 3348#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
d65d87a0
TT		 3349 if (!readonly && (ext4_has_feature_quota(sb) ||
3350 ext4_has_feature_project(sb))) {
7c319d32 3351 ext4_msg(sb, KERN_ERR,
d65d87a0 3352 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
689c958c
LX		 3353 return 0;
3354 }
7c319d32 3355#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 3356 return 1;
3357}
3358
66e61a9e
TT		 3359/*
3360 * This function is called once a day if we have errors logged
3361 * on the file system
3362 */
235699a8 3363static void print_daily_error_info(struct timer_list *t)
66e61a9e 3364{
235699a8
KC		 3365 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3366 struct super_block *sb = sbi->s_sb;
3367 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 3368
3369 if (es->s_error_count)
ae0f78de
TT		 3370 /* fsck newer than v1.41.13 is needed to clean this condition. */
3371 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 3372 le32_to_cpu(es->s_error_count));
3373 if (es->s_first_error_time) {
6a0678a7
AB		 3374 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3375 sb->s_id,
3376 ext4_get_tstamp(es, s_first_error_time),
66e61a9e
TT		 3377 (int) sizeof(es->s_first_error_func),
3378 es->s_first_error_func,
3379 le32_to_cpu(es->s_first_error_line));
3380 if (es->s_first_error_ino)
651e1c3b 3381 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3382 le32_to_cpu(es->s_first_error_ino));
3383 if (es->s_first_error_block)
651e1c3b 3384 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3385 le64_to_cpu(es->s_first_error_block));
651e1c3b 3386 printk(KERN_CONT "\n");
66e61a9e
TT		 3387 }
3388 if (es->s_last_error_time) {
6a0678a7
AB		 3389 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3390 sb->s_id,
3391 ext4_get_tstamp(es, s_last_error_time),
66e61a9e
TT		 3392 (int) sizeof(es->s_last_error_func),
3393 es->s_last_error_func,
3394 le32_to_cpu(es->s_last_error_line));
3395 if (es->s_last_error_ino)
651e1c3b 3396 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3397 le32_to_cpu(es->s_last_error_ino));
3398 if (es->s_last_error_block)
651e1c3b 3399 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3400 le64_to_cpu(es->s_last_error_block));
651e1c3b 3401 printk(KERN_CONT "\n");
66e61a9e
TT		 3402 }
3403 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3404}
3405
bfff6873
LC		 3406/* Find next suitable group and run ext4_init_inode_table */
3407static int ext4_run_li_request(struct ext4_li_request *elr)
3408{
3409 struct ext4_group_desc *gdp = NULL;
3d392b26
TT		 3410 struct super_block *sb = elr->lr_super;
3411 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3412 ext4_group_t group = elr->lr_next_group;
bfff6873 3413 unsigned long timeout = 0;
3d392b26 3414 unsigned int prefetch_ios = 0;
bfff6873
LC		 3415 int ret = 0;
3416
3d392b26
TT		 3417 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3418 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3419 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3420 if (prefetch_ios)
3421 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3422 prefetch_ios);
3423 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3424 prefetch_ios);
3425 if (group >= elr->lr_next_group) {
3426 ret = 1;
3427 if (elr->lr_first_not_zeroed != ngroups &&
3428 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3429 elr->lr_next_group = elr->lr_first_not_zeroed;
3430 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3431 ret = 0;
3432 }
3433 }
3434 return ret;
3435 }
bfff6873 3436
3d392b26 3437 for (; group < ngroups; group++) {
bfff6873
LC		 3438 gdp = ext4_get_group_desc(sb, group, NULL);
3439 if (!gdp) {
3440 ret = 1;
3441 break;
3442 }
3443
3444 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3445 break;
3446 }
3447
7f511862 3448 if (group >= ngroups)
bfff6873
LC		 3449 ret = 1;
3450
3451 if (!ret) {
3452 timeout = jiffies;
3453 ret = ext4_init_inode_table(sb, group,
3454 elr->lr_timeout ? 0 : 1);
3d392b26 3455 trace_ext4_lazy_itable_init(sb, group);
bfff6873 3456 if (elr->lr_timeout == 0) {
51ce6511 3457 timeout = (jiffies - timeout) *
3d392b26 3458 EXT4_SB(elr->lr_super)->s_li_wait_mult;
bfff6873
LC		 3459 elr->lr_timeout = timeout;
3460 }
3461 elr->lr_next_sched = jiffies + elr->lr_timeout;
3462 elr->lr_next_group = group + 1;
3463 }
bfff6873
LC		 3464 return ret;
3465}
3466
3467/*
3468 * Remove lr_request from the list_request and free the
4ed5c033 3469 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 3470 */
3471static void ext4_remove_li_request(struct ext4_li_request *elr)
3472{
bfff6873
LC		 3473 if (!elr)
3474 return;
3475
bfff6873 3476 list_del(&elr->lr_request);
3d392b26 3477 EXT4_SB(elr->lr_super)->s_li_request = NULL;
bfff6873
LC		 3478 kfree(elr);
3479}
3480
3481static void ext4_unregister_li_request(struct super_block *sb)
3482{
1bb933fb
LC		 3483 mutex_lock(&ext4_li_mtx);
3484 if (!ext4_li_info) {
3485 mutex_unlock(&ext4_li_mtx);
bfff6873 3486 return;
1bb933fb 3487 }
bfff6873
LC		 3488
3489 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 3490 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 3491 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 3492 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 3493}
3494
8f1f7453
ES		 3495static struct task_struct *ext4_lazyinit_task;
3496
bfff6873
LC		 3497/*
3498 * This is the function where ext4lazyinit thread lives. It walks
3499 * through the request list searching for next scheduled filesystem.
3500 * When such a fs is found, run the lazy initialization request
3501 * (ext4_rn_li_request) and keep track of the time spend in this
3502 * function. Based on that time we compute next schedule time of
3503 * the request. When walking through the list is complete, compute
3504 * next waking time and put itself into sleep.
3505 */
3506static int ext4_lazyinit_thread(void *arg)
3507{
3508 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3509 struct list_head *pos, *n;
3510 struct ext4_li_request *elr;
4ed5c033 3511 unsigned long next_wakeup, cur;
bfff6873
LC		 3512
3513 BUG_ON(NULL == eli);
3514
bfff6873
LC		 3515cont_thread:
3516 while (true) {
3517 next_wakeup = MAX_JIFFY_OFFSET;
3518
3519 mutex_lock(&eli->li_list_mtx);
3520 if (list_empty(&eli->li_request_list)) {
3521 mutex_unlock(&eli->li_list_mtx);
3522 goto exit_thread;
3523 }
bfff6873 3524 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3525 int err = 0;
3526 int progress = 0;
bfff6873
LC		 3527 elr = list_entry(pos, struct ext4_li_request,
3528 lr_request);
3529
e22834f0
DM		 3530 if (time_before(jiffies, elr->lr_next_sched)) {
3531 if (time_before(elr->lr_next_sched, next_wakeup))
3532 next_wakeup = elr->lr_next_sched;
3533 continue;
3534 }
3535 if (down_read_trylock(&elr->lr_super->s_umount)) {
3536 if (sb_start_write_trylock(elr->lr_super)) {
3537 progress = 1;
3538 /*
3539 * We hold sb->s_umount, sb can not
3540 * be removed from the list, it is
3541 * now safe to drop li_list_mtx
3542 */
3543 mutex_unlock(&eli->li_list_mtx);
3544 err = ext4_run_li_request(elr);
3545 sb_end_write(elr->lr_super);
3546 mutex_lock(&eli->li_list_mtx);
3547 n = pos->next;
b2c78cd0 3548 }
e22834f0
DM		 3549 up_read((&elr->lr_super->s_umount));
3550 }
3551 /* error, remove the lazy_init job */
3552 if (err) {
3553 ext4_remove_li_request(elr);
3554 continue;
3555 }
3556 if (!progress) {
3557 elr->lr_next_sched = jiffies +
3558 (prandom_u32()
3559 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 3560 }
bfff6873
LC		 3561 if (time_before(elr->lr_next_sched, next_wakeup))
3562 next_wakeup = elr->lr_next_sched;
3563 }
3564 mutex_unlock(&eli->li_list_mtx);
3565
a0acae0e 3566 try_to_freeze();
bfff6873 3567
4ed5c033
LC		 3568 cur = jiffies;
3569 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3570 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3571 cond_resched();
3572 continue;
3573 }
3574
4ed5c033
LC		 3575 schedule_timeout_interruptible(next_wakeup - cur);
3576
8f1f7453
ES		 3577 if (kthread_should_stop()) {
3578 ext4_clear_request_list();
3579 goto exit_thread;
3580 }
bfff6873
LC		 3581 }
3582
3583exit_thread:
3584 /*
3585 * It looks like the request list is empty, but we need
3586 * to check it under the li_list_mtx lock, to prevent any
3587 * additions into it, and of course we should lock ext4_li_mtx
3588 * to atomically free the list and ext4_li_info, because at
3589 * this point another ext4 filesystem could be registering
3590 * new one.
3591 */
3592 mutex_lock(&ext4_li_mtx);
3593 mutex_lock(&eli->li_list_mtx);
3594 if (!list_empty(&eli->li_request_list)) {
3595 mutex_unlock(&eli->li_list_mtx);
3596 mutex_unlock(&ext4_li_mtx);
3597 goto cont_thread;
3598 }
3599 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3600 kfree(ext4_li_info);
3601 ext4_li_info = NULL;
3602 mutex_unlock(&ext4_li_mtx);
3603
3604 return 0;
3605}
3606
3607static void ext4_clear_request_list(void)
3608{
3609 struct list_head *pos, *n;
3610 struct ext4_li_request *elr;
3611
3612 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3613 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3614 elr = list_entry(pos, struct ext4_li_request,
3615 lr_request);
3616 ext4_remove_li_request(elr);
3617 }
3618 mutex_unlock(&ext4_li_info->li_list_mtx);
3619}
3620
3621static int ext4_run_lazyinit_thread(void)
3622{
8f1f7453
ES		 3623 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3624 ext4_li_info, "ext4lazyinit");
3625 if (IS_ERR(ext4_lazyinit_task)) {
3626 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3627 ext4_clear_request_list();
bfff6873
LC		 3628 kfree(ext4_li_info);
3629 ext4_li_info = NULL;
92b97816 3630 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3631 "initialization thread\n",
3632 err);
3633 return err;
3634 }
3635 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3636 return 0;
3637}
3638
3639/*
3640 * Check whether it make sense to run itable init. thread or not.
3641 * If there is at least one uninitialized inode table, return
3642 * corresponding group number, else the loop goes through all
3643 * groups and return total number of groups.
3644 */
3645static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3646{
3647 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3648 struct ext4_group_desc *gdp = NULL;
3649
8844618d
TT		 3650 if (!ext4_has_group_desc_csum(sb))
3651 return ngroups;
3652
bfff6873
LC		 3653 for (group = 0; group < ngroups; group++) {
3654 gdp = ext4_get_group_desc(sb, group, NULL);
3655 if (!gdp)
3656 continue;
3657
50122847 3658 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
bfff6873
LC		 3659 break;
3660 }
3661
3662 return group;
3663}
3664
3665static int ext4_li_info_new(void)
3666{
3667 struct ext4_lazy_init *eli = NULL;
3668
3669 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3670 if (!eli)
3671 return -ENOMEM;
3672
bfff6873
LC		 3673 INIT_LIST_HEAD(&eli->li_request_list);
3674 mutex_init(&eli->li_list_mtx);
3675
bfff6873
LC		 3676 eli->li_state |= EXT4_LAZYINIT_QUIT;
3677
3678 ext4_li_info = eli;
3679
3680 return 0;
3681}
3682
3683static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3684 ext4_group_t start)
3685{
bfff6873 3686 struct ext4_li_request *elr;
bfff6873
LC		 3687
3688 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3689 if (!elr)
3690 return NULL;
3691
3692 elr->lr_super = sb;
3d392b26
TT		 3693 elr->lr_first_not_zeroed = start;
3694 if (test_opt(sb, PREFETCH_BLOCK_BITMAPS))
3695 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3696 else {
3697 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3698 elr->lr_next_group = start;
3699 }
bfff6873
LC		 3700
3701 /*
3702 * Randomize first schedule time of the request to
3703 * spread the inode table initialization requests
3704 * better.
3705 */
dd1f723b
TT		 3706 elr->lr_next_sched = jiffies + (prandom_u32() %
3707 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3708 return elr;
3709}
3710
7f511862
TT		 3711int ext4_register_li_request(struct super_block *sb,
3712 ext4_group_t first_not_zeroed)
bfff6873
LC		 3713{
3714 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3715 struct ext4_li_request *elr = NULL;
49598e04 3716 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 3717 int ret = 0;
bfff6873 3718
7f511862 3719 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3720 if (sbi->s_li_request != NULL) {
3721 /*
3722 * Reset timeout so it can be computed again, because
3723 * s_li_wait_mult might have changed.
3724 */
3725 sbi->s_li_request->lr_timeout = 0;
7f511862 3726 goto out;
51ce6511 3727 }
bfff6873 3728
3d392b26
TT		 3729 if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) &&
3730 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3731 !test_opt(sb, INIT_INODE_TABLE)))
7f511862 3732 goto out;
bfff6873
LC		 3733
3734 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3735 if (!elr) {
3736 ret = -ENOMEM;
3737 goto out;
3738 }
bfff6873
LC		 3739
3740 if (NULL == ext4_li_info) {
3741 ret = ext4_li_info_new();
3742 if (ret)
3743 goto out;
3744 }
3745
3746 mutex_lock(&ext4_li_info->li_list_mtx);
3747 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3748 mutex_unlock(&ext4_li_info->li_list_mtx);
3749
3750 sbi->s_li_request = elr;
46e4690b
TM		 3751 /*
3752 * set elr to NULL here since it has been inserted to
3753 * the request_list and the removal and free of it is
3754 * handled by ext4_clear_request_list from now on.
3755 */
3756 elr = NULL;
bfff6873
LC		 3757
3758 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3759 ret = ext4_run_lazyinit_thread();
3760 if (ret)
3761 goto out;
3762 }
bfff6873 3763out:
beed5ecb
NK		 3764 mutex_unlock(&ext4_li_mtx);
3765 if (ret)
bfff6873 3766 kfree(elr);
bfff6873
LC		 3767 return ret;
3768}
3769
3770/*
3771 * We do not need to lock anything since this is called on
3772 * module unload.
3773 */
3774static void ext4_destroy_lazyinit_thread(void)
3775{
3776 /*
3777 * If thread exited earlier
3778 * there's nothing to be done.
3779 */
8f1f7453 3780 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3781 return;
3782
8f1f7453 3783 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3784}
3785
25ed6e8a
DW		 3786static int set_journal_csum_feature_set(struct super_block *sb)
3787{
3788 int ret = 1;
3789 int compat, incompat;
3790 struct ext4_sb_info *sbi = EXT4_SB(sb);
3791
9aa5d32b 3792 if (ext4_has_metadata_csum(sb)) {
db9ee220 3793 /* journal checksum v3 */
25ed6e8a 3794 compat = 0;
db9ee220 3795 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 3796 } else {
3797 /* journal checksum v1 */
3798 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3799 incompat = 0;
3800 }
3801
feb8c6d3
DW		 3802 jbd2_journal_clear_features(sbi->s_journal,
3803 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3804 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3805 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 3806 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3807 ret = jbd2_journal_set_features(sbi->s_journal,
3808 compat, 0,
3809 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3810 incompat);
3811 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3812 ret = jbd2_journal_set_features(sbi->s_journal,
3813 compat, 0,
3814 incompat);
3815 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3816 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3817 } else {
feb8c6d3
DW		 3818 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3819 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 3820 }
3821
3822 return ret;
3823}
3824
952fc18e
TT		 3825/*
3826 * Note: calculating the overhead so we can be compatible with
3827 * historical BSD practice is quite difficult in the face of
3828 * clusters/bigalloc. This is because multiple metadata blocks from
3829 * different block group can end up in the same allocation cluster.
3830 * Calculating the exact overhead in the face of clustered allocation
3831 * requires either O(all block bitmaps) in memory or O(number of block
3832 * groups**2) in time. We will still calculate the superblock for
3833 * older file systems --- and if we come across with a bigalloc file
3834 * system with zero in s_overhead_clusters the estimate will be close to
3835 * correct especially for very large cluster sizes --- but for newer
3836 * file systems, it's better to calculate this figure once at mkfs
3837 * time, and store it in the superblock. If the superblock value is
3838 * present (even for non-bigalloc file systems), we will use it.
3839 */
3840static int count_overhead(struct super_block *sb, ext4_group_t grp,
3841 char *buf)
3842{
3843 struct ext4_sb_info *sbi = EXT4_SB(sb);
3844 struct ext4_group_desc *gdp;
3845 ext4_fsblk_t first_block, last_block, b;
3846 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3847 int s, j, count = 0;
3848
e2b911c5 3849 if (!ext4_has_feature_bigalloc(sb))
0548bbb8
TT		 3850 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3851 sbi->s_itb_per_group + 2);
3852
952fc18e
TT		 3853 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3854 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3855 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3856 for (i = 0; i < ngroups; i++) {
3857 gdp = ext4_get_group_desc(sb, i, NULL);
3858 b = ext4_block_bitmap(sb, gdp);
3859 if (b >= first_block && b <= last_block) {
3860 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3861 count++;
3862 }
3863 b = ext4_inode_bitmap(sb, gdp);
3864 if (b >= first_block && b <= last_block) {
3865 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3866 count++;
3867 }
3868 b = ext4_inode_table(sb, gdp);
3869 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3870 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3871 int c = EXT4_B2C(sbi, b - first_block);
3872 ext4_set_bit(c, buf);
3873 count++;
3874 }
3875 if (i != grp)
3876 continue;
3877 s = 0;
3878 if (ext4_bg_has_super(sb, grp)) {
3879 ext4_set_bit(s++, buf);
3880 count++;
3881 }
c48ae41b
TT		 3882 j = ext4_bg_num_gdb(sb, grp);
3883 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3884 ext4_error(sb, "Invalid number of block group "
3885 "descriptor blocks: %d", j);
3886 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 3887 }
c48ae41b
TT		 3888 count += j;
3889 for (; j > 0; j--)
3890 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 3891 }
3892 if (!count)
3893 return 0;
3894 return EXT4_CLUSTERS_PER_GROUP(sb) -
3895 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3896}
3897
3898/*
3899 * Compute the overhead and stash it in sbi->s_overhead
3900 */
3901int ext4_calculate_overhead(struct super_block *sb)
3902{
3903 struct ext4_sb_info *sbi = EXT4_SB(sb);
3904 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 3905 struct inode *j_inode;
3906 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 3907 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3908 ext4_fsblk_t overhead = 0;
4fdb5543 3909 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 3910
952fc18e
TT		 3911 if (!buf)
3912 return -ENOMEM;
3913
3914 /*
3915 * Compute the overhead (FS structures). This is constant
3916 * for a given filesystem unless the number of block groups
3917 * changes so we cache the previous value until it does.
3918 */
3919
3920 /*
3921 * All of the blocks before first_data_block are overhead
3922 */
3923 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3924
3925 /*
3926 * Add the overhead found in each block group
3927 */
3928 for (i = 0; i < ngroups; i++) {
3929 int blks;
3930
3931 blks = count_overhead(sb, i, buf);
3932 overhead += blks;
3933 if (blks)
3934 memset(buf, 0, PAGE_SIZE);
3935 cond_resched();
3936 }
3c816ded
EW		 3937
3938 /*
3939 * Add the internal journal blocks whether the journal has been
3940 * loaded or not
3941 */
ee7ed3aa 3942 if (sbi->s_journal && !sbi->s_journal_bdev)
ede7dc7f 3943 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
f1eec3b0
RH		 3944 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3945 /* j_inum for internal journal is non-zero */
3c816ded
EW		 3946 j_inode = ext4_get_journal_inode(sb, j_inum);
3947 if (j_inode) {
3948 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3949 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3950 iput(j_inode);
3951 } else {
3952 ext4_msg(sb, KERN_ERR, "can't get journal size");
3953 }
3954 }
952fc18e
TT		 3955 sbi->s_overhead = overhead;
3956 smp_wmb();
3957 free_page((unsigned long) buf);
3958 return 0;
3959}
3960
b5799018 3961static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 3962{
3963 ext4_fsblk_t resv_clusters;
b5799018 3964 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 3965
30fac0f7
JK		 3966 /*
3967 * There's no need to reserve anything when we aren't using extents.
3968 * The space estimates are exact, there are no unwritten extents,
3969 * hole punching doesn't need new metadata... This is needed especially
3970 * to keep ext2/3 backward compatibility.
3971 */
e2b911c5 3972 if (!ext4_has_feature_extents(sb))
b5799018 3973 return;
27dd4385
LC		 3974 /*
3975 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3976 * This should cover the situations where we can not afford to run
3977 * out of space like for example punch hole, or converting
556615dc 3978 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3979 * allocation would require 1, or 2 blocks, higher numbers are
3980 * very rare.
3981 */
b5799018
TT		 3982 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3983 sbi->s_cluster_bits);
27dd4385
LC		 3984
3985 do_div(resv_clusters, 50);
3986 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3987
b5799018 3988 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 3989}
3990
ca9b404f
RA		 3991static const char *ext4_quota_mode(struct super_block *sb)
3992{
3993#ifdef CONFIG_QUOTA
3994 if (!ext4_quota_capable(sb))
3995 return "none";
3996
3997 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
3998 return "journalled";
3999 else
4000 return "writeback";
4001#else
4002 return "disabled";
4003#endif
4004}
4005
2b2d6d01 4006static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 4007{
5e405595 4008 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
d4c402d9 4009 char *orig_data = kstrdup(data, GFP_KERNEL);
1d0c3924 4010 struct buffer_head *bh, **group_desc;
617ba13b 4011 struct ext4_super_block *es = NULL;
5aee0f8a 4012 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
7c990728 4013 struct flex_groups **flex_groups;
617ba13b
MC		 4014 ext4_fsblk_t block;
4015 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 4016 ext4_fsblk_t logical_sb_block;
ac27a0ec 4017 unsigned long offset = 0;
ac27a0ec
DK		 4018 unsigned long journal_devnum = 0;
4019 unsigned long def_mount_opts;
4020 struct inode *root;
0390131b 4021 const char *descr;
dcc7dae3 4022 int ret = -ENOMEM;
281b5995 4023 int blocksize, clustersize;
4ec11028
TT		 4024 unsigned int db_count;
4025 unsigned int i;
ef5fd681 4026 int needs_recovery, has_huge_files;
bd81d8ee 4027 __u64 blocks_count;
07aa2ea1 4028 int err = 0;
b3881f74 4029 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 4030 ext4_group_t first_not_zeroed;
ac27a0ec 4031
5aee0f8a
TT		 4032 if ((data && !orig_data) || !sbi)
4033 goto out_free_base;
705895b6 4034
aed9eb1b 4035 sbi->s_daxdev = dax_dev;
705895b6
PE		 4036 sbi->s_blockgroup_lock =
4037 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
5aee0f8a
TT		 4038 if (!sbi->s_blockgroup_lock)
4039 goto out_free_base;
4040
ac27a0ec 4041 sb->s_fs_info = sbi;
2c0544b2 4042 sbi->s_sb = sb;
240799cd 4043 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 4044 sbi->s_sb_block = sb_block;
8446fe92
CH		 4045 sbi->s_sectors_written_start =
4046 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
ac27a0ec 4047
9f6200bb 4048 /* Cleanup superblock name */
ec3904dc 4049 strreplace(sb->s_id, '/', '!');
9f6200bb 4050
07aa2ea1 4051 /* -EINVAL is default */
dcc7dae3 4052 ret = -EINVAL;
617ba13b 4053 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 4054 if (!blocksize) {
b31e1552 4055 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 4056 goto out_fail;
4057 }
4058
4059 /*
617ba13b 4060 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 4061 * block sizes. We need to calculate the offset from buffer start.
4062 */
617ba13b 4063 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 4064 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4065 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 4066 } else {
70bbb3e0 4067 logical_sb_block = sb_block;
ac27a0ec
DK		 4068 }
4069
8394a6ab 4070 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4071 if (IS_ERR(bh)) {
b31e1552 4072 ext4_msg(sb, KERN_ERR, "unable to read superblock");
8394a6ab 4073 ret = PTR_ERR(bh);
ac27a0ec
DK		 4074 goto out_fail;
4075 }
4076 /*
4077 * Note: s_es must be initialized as soon as possible because
617ba13b 4078 * some ext4 macro-instructions depend on its value
ac27a0ec 4079 */
2716b802 4080 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 4081 sbi->s_es = es;
4082 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 4083 if (sb->s_magic != EXT4_SUPER_MAGIC)
4084 goto cantfind_ext4;
afc32f7e 4085 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 4086
feb0ab32 4087 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 4088 if (ext4_has_feature_metadata_csum(sb) &&
4089 ext4_has_feature_gdt_csum(sb))
363307e6 4090 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 4091 "redundant flags; please run fsck.");
4092
d25425f8
DW		 4093 /* Check for a known checksum algorithm */
4094 if (!ext4_verify_csum_type(sb, es)) {
4095 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4096 "unknown checksum algorithm.");
4097 silent = 1;
4098 goto cantfind_ext4;
4099 }
4100
0441984a 4101 /* Load the checksum driver */
a45403b5
TT		 4102 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4103 if (IS_ERR(sbi->s_chksum_driver)) {
4104 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4105 ret = PTR_ERR(sbi->s_chksum_driver);
4106 sbi->s_chksum_driver = NULL;
4107 goto failed_mount;
0441984a
DW		 4108 }
4109
a9c47317
DW		 4110 /* Check superblock checksum */
4111 if (!ext4_superblock_csum_verify(sb, es)) {
4112 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4113 "invalid superblock checksum. Run e2fsck?");
4114 silent = 1;
6a797d27 4115 ret = -EFSBADCRC;
a9c47317
DW		 4116 goto cantfind_ext4;
4117 }
4118
4119 /* Precompute checksum seed for all metadata */
e2b911c5 4120 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 4121 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 4122 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 4123 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4124 sizeof(es->s_uuid));
4125
ac27a0ec
DK		 4126 /* Set defaults before we parse the mount options */
4127 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 4128 set_opt(sb, INIT_INODE_TABLE);
617ba13b 4129 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 4130 set_opt(sb, DEBUG);
87f26807 4131 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 4132 set_opt(sb, GRPID);
617ba13b 4133 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 4134 set_opt(sb, NO_UID32);
ea663336 4135 /* xattr user namespace & acls are now defaulted on */
ea663336 4136 set_opt(sb, XATTR_USER);
03010a33 4137#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 4138 set_opt(sb, POSIX_ACL);
2e7842b8 4139#endif
995a3ed6
HS		 4140 if (ext4_has_feature_fast_commit(sb))
4141 set_opt2(sb, JOURNAL_FAST_COMMIT);
98c1a759
DW		 4142 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4143 if (ext4_has_metadata_csum(sb))
4144 set_opt(sb, JOURNAL_CHECKSUM);
4145
617ba13b 4146 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 4147 set_opt(sb, JOURNAL_DATA);
617ba13b 4148 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 4149 set_opt(sb, ORDERED_DATA);
617ba13b 4150 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 4151 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 4152
4153 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 4154 set_opt(sb, ERRORS_PANIC);
bb4f397a 4155 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 4156 set_opt(sb, ERRORS_CONT);
bb4f397a 4157 else
fd8c37ec 4158 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 4159 /* block_validity enabled by default; disable with noblock_validity */
4160 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 4161 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 4162 set_opt(sb, DISCARD);
ac27a0ec 4163
08cefc7a
EB		 4164 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4165 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 4166 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4167 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4168 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 4169
8b67f04a 4170 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 4171 set_opt(sb, BARRIER);
ac27a0ec 4172
dd919b98
AK		 4173 /*
4174 * enable delayed allocation by default
4175 * Use -o nodelalloc to turn it off
4176 */
bc0b75f7 4177 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 4178 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 4179 set_opt(sb, DELALLOC);
dd919b98 4180
51ce6511
LC		 4181 /*
4182 * set default s_li_wait_mult for lazyinit, for the case there is
4183 * no mount option specified.
4184 */
4185 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4186
c9200760
TT		 4187 if (le32_to_cpu(es->s_log_block_size) >
4188 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4f97a681 4189 ext4_msg(sb, KERN_ERR,
c9200760
TT		 4190 "Invalid log block size: %u",
4191 le32_to_cpu(es->s_log_block_size));
4f97a681
TT		 4192 goto failed_mount;
4193 }
c9200760
TT		 4194 if (le32_to_cpu(es->s_log_cluster_size) >
4195 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4f97a681 4196 ext4_msg(sb, KERN_ERR,
c9200760
TT		 4197 "Invalid log cluster size: %u",
4198 le32_to_cpu(es->s_log_cluster_size));
4f97a681
TT		 4199 goto failed_mount;
4200 }
4201
c9200760
TT		 4202 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4203
4204 if (blocksize == PAGE_SIZE)
4205 set_opt(sb, DIOREAD_NOLOCK);
4f97a681 4206
9803387c
TT		 4207 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4208 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4209 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4210 } else {
4211 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4212 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4213 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4214 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4215 sbi->s_first_ino);
4216 goto failed_mount;
4217 }
4218 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4219 (!is_power_of_2(sbi->s_inode_size)) ||
4220 (sbi->s_inode_size > blocksize)) {
4221 ext4_msg(sb, KERN_ERR,
4222 "unsupported inode size: %d",
4223 sbi->s_inode_size);
4f97a681 4224 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
9803387c
TT		 4225 goto failed_mount;
4226 }
4227 /*
4228 * i_atime_extra is the last extra field available for
4229 * [acm]times in struct ext4_inode. Checking for that
4230 * field should suffice to ensure we have extra space
4231 * for all three.
4232 */
4233 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4234 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4235 sb->s_time_gran = 1;
4236 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4237 } else {
4238 sb->s_time_gran = NSEC_PER_SEC;
4239 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4240 }
4241 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4242 }
4243 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4244 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4245 EXT4_GOOD_OLD_INODE_SIZE;
4246 if (ext4_has_feature_extra_isize(sb)) {
4247 unsigned v, max = (sbi->s_inode_size -
4248 EXT4_GOOD_OLD_INODE_SIZE);
4249
4250 v = le16_to_cpu(es->s_want_extra_isize);
4251 if (v > max) {
4252 ext4_msg(sb, KERN_ERR,
4253 "bad s_want_extra_isize: %d", v);
4254 goto failed_mount;
4255 }
4256 if (sbi->s_want_extra_isize < v)
4257 sbi->s_want_extra_isize = v;
4258
4259 v = le16_to_cpu(es->s_min_extra_isize);
4260 if (v > max) {
4261 ext4_msg(sb, KERN_ERR,
4262 "bad s_min_extra_isize: %d", v);
4263 goto failed_mount;
4264 }
4265 if (sbi->s_want_extra_isize < v)
4266 sbi->s_want_extra_isize = v;
4267 }
4268 }
4269
5aee0f8a
TT		 4270 if (sbi->s_es->s_mount_opts[0]) {
4271 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4272 sizeof(sbi->s_es->s_mount_opts),
4273 GFP_KERNEL);
4274 if (!s_mount_opts)
4275 goto failed_mount;
4276 if (!parse_options(s_mount_opts, sb, &journal_devnum,
4277 &journal_ioprio, 0)) {
4278 ext4_msg(sb, KERN_WARNING,
4279 "failed to parse options in superblock: %s",
4280 s_mount_opts);
4281 }
4282 kfree(s_mount_opts);
8b67f04a 4283 }
5a916be1 4284 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 4285 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 4286 &journal_ioprio, 0))
ac27a0ec
DK		 4287 goto failed_mount;
4288
c83ad55e 4289#ifdef CONFIG_UNICODE
f8f4acb6 4290 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
c83ad55e
GKB		 4291 const struct ext4_sb_encodings *encoding_info;
4292 struct unicode_map *encoding;
4293 __u16 encoding_flags;
4294
4295 if (ext4_has_feature_encrypt(sb)) {
4296 ext4_msg(sb, KERN_ERR,
4297 "Can't mount with encoding and encryption");
4298 goto failed_mount;
4299 }
4300
4301 if (ext4_sb_read_encoding(es, &encoding_info,
4302 &encoding_flags)) {
4303 ext4_msg(sb, KERN_ERR,
4304 "Encoding requested by superblock is unknown");
4305 goto failed_mount;
4306 }
4307
4308 encoding = utf8_load(encoding_info->version);
4309 if (IS_ERR(encoding)) {
4310 ext4_msg(sb, KERN_ERR,
4311 "can't mount with superblock charset: %s-%s "
4312 "not supported by the kernel. flags: 0x%x.",
4313 encoding_info->name, encoding_info->version,
4314 encoding_flags);
4315 goto failed_mount;
4316 }
4317 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4318 "%s-%s with flags 0x%hx", encoding_info->name,
4319 encoding_info->version?:"\b", encoding_flags);
4320
f8f4acb6
DR		 4321 sb->s_encoding = encoding;
4322 sb->s_encoding_flags = encoding_flags;
c83ad55e
GKB		 4323 }
4324#endif
4325
56889787 4326 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
556e0319 4327 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
781c036b 4328 /* can't mount with both data=journal and dioread_nolock. */
244adf64 4329 clear_opt(sb, DIOREAD_NOLOCK);
556e0319 4330 clear_opt2(sb, JOURNAL_FAST_COMMIT);
56889787
TT		 4331 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4332 ext4_msg(sb, KERN_ERR, "can't mount with "
4333 "both data=journal and delalloc");
4334 goto failed_mount;
4335 }
fc626fe3 4336 if (test_opt(sb, DAX_ALWAYS)) {
923ae0ff
RZ		 4337 ext4_msg(sb, KERN_ERR, "can't mount with "
4338 "both data=journal and dax");
4339 goto failed_mount;
4340 }
73b92a2a
SK		 4341 if (ext4_has_feature_encrypt(sb)) {
4342 ext4_msg(sb, KERN_WARNING,
4343 "encrypted files will use data=ordered "
4344 "instead of data journaling mode");
4345 }
56889787
TT		 4346 if (test_opt(sb, DELALLOC))
4347 clear_opt(sb, DELALLOC);
001e4a87
TH		 4348 } else {
4349 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 4350 }
4351
1751e8a6
LT		 4352 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4353 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 4354
617ba13b 4355 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 4356 (ext4_has_compat_features(sb) ||
4357 ext4_has_ro_compat_features(sb) ||
4358 ext4_has_incompat_features(sb)))
b31e1552
ES		 4359 ext4_msg(sb, KERN_WARNING,
4360 "feature flags set on rev 0 fs, "
4361 "running e2fsck is recommended");
469108ff 4362
ed3654eb
TT		 4363 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4364 set_opt2(sb, HURD_COMPAT);
e2b911c5 4365 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 4366 ext4_msg(sb, KERN_ERR,
4367 "The Hurd can't support 64-bit file systems");
4368 goto failed_mount;
4369 }
dec214d0
TE		 4370
4371 /*
4372 * ea_inode feature uses l_i_version field which is not
4373 * available in HURD_COMPAT mode.
4374 */
4375 if (ext4_has_feature_ea_inode(sb)) {
4376 ext4_msg(sb, KERN_ERR,
4377 "ea_inode feature is not supported for Hurd");
4378 goto failed_mount;
4379 }
ed3654eb
TT		 4380 }
4381
2035e776
TT		 4382 if (IS_EXT2_SB(sb)) {
4383 if (ext2_feature_set_ok(sb))
4384 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4385 "using the ext4 subsystem");
4386 else {
0d9366d6
ES		 4387 /*
4388 * If we're probing be silent, if this looks like
4389 * it's actually an ext[34] filesystem.
4390 */
4391 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4392 goto failed_mount;
2035e776
TT		 4393 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4394 "to feature incompatibilities");
4395 goto failed_mount;
4396 }
4397 }
4398
4399 if (IS_EXT3_SB(sb)) {
4400 if (ext3_feature_set_ok(sb))
4401 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4402 "using the ext4 subsystem");
4403 else {
0d9366d6
ES		 4404 /*
4405 * If we're probing be silent, if this looks like
4406 * it's actually an ext4 filesystem.
4407 */
4408 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4409 goto failed_mount;
2035e776
TT		 4410 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4411 "to feature incompatibilities");
4412 goto failed_mount;
4413 }
4414 }
4415
ac27a0ec
DK		 4416 /*
4417 * Check feature flags regardless of the revision level, since we
4418 * previously didn't change the revision level when setting the flags,
4419 * so there is a chance incompat flags are set on a rev 0 filesystem.
4420 */
bc98a42c 4421 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
ac27a0ec 4422 goto failed_mount;
a13fb1a4 4423
5b9554dc
TT		 4424 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4425 ext4_msg(sb, KERN_ERR,
4426 "Number of reserved GDT blocks insanely large: %d",
4427 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4428 goto failed_mount;
4429 }
4430
a8ab6d38
IW		 4431 if (bdev_dax_supported(sb->s_bdev, blocksize))
4432 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4433
fc626fe3 4434 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
559db4c6
RZ		 4435 if (ext4_has_feature_inline_data(sb)) {
4436 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4437 " that may contain inline data");
361d24d4 4438 goto failed_mount;
559db4c6 4439 }
a8ab6d38 4440 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
24f3478d 4441 ext4_msg(sb, KERN_ERR,
361d24d4
ES		 4442 "DAX unsupported by block device.");
4443 goto failed_mount;
24f3478d 4444 }
923ae0ff
RZ		 4445 }
4446
e2b911c5 4447 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 4448 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4449 es->s_encryption_level);
4450 goto failed_mount;
4451 }
4452
ac27a0ec 4453 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 4454 /* Validate the filesystem blocksize */
4455 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 4456 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 4457 blocksize);
ac27a0ec
DK		 4458 goto failed_mount;
4459 }
4460
2b2d6d01 4461 brelse(bh);
70bbb3e0
AM		 4462 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4463 offset = do_div(logical_sb_block, blocksize);
8394a6ab 4464 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4465 if (IS_ERR(bh)) {
b31e1552
ES		 4466 ext4_msg(sb, KERN_ERR,
4467 "Can't read superblock on 2nd try");
8394a6ab 4468 ret = PTR_ERR(bh);
4469 bh = NULL;
ac27a0ec
DK		 4470 goto failed_mount;
4471 }
2716b802 4472 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 4473 sbi->s_es = es;
617ba13b 4474 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 4475 ext4_msg(sb, KERN_ERR,
4476 "Magic mismatch, very weird!");
ac27a0ec
DK		 4477 goto failed_mount;
4478 }
4479 }
4480
e2b911c5 4481 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 4482 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4483 has_huge_files);
4484 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 4485
0d1ee42f 4486 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 4487 if (ext4_has_feature_64bit(sb)) {
8fadc143 4488 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 4489 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 4490 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 4491 ext4_msg(sb, KERN_ERR,
4492 "unsupported descriptor size %lu",
0d1ee42f
AR		 4493 sbi->s_desc_size);
4494 goto failed_mount;
4495 }
4496 } else
4497 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 4498
ac27a0ec 4499 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 4500 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 4501
617ba13b 4502 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 4503 if (sbi->s_inodes_per_block == 0)
617ba13b 4504 goto cantfind_ext4;
cd6bb35b
TT		 4505 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4506 sbi->s_inodes_per_group > blocksize * 8) {
4507 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
b9c538da 4508 sbi->s_inodes_per_group);
cd6bb35b
TT		 4509 goto failed_mount;
4510 }
ac27a0ec
DK		 4511 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4512 sbi->s_inodes_per_block;
0d1ee42f 4513 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 4514 sbi->s_sbh = bh;
4515 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 4516 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4517 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 4518
2b2d6d01 4519 for (i = 0; i < 4; i++)
ac27a0ec
DK		 4520 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4521 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 4522 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 4523 i = le32_to_cpu(es->s_flags);
4524 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4525 sbi->s_hash_unsigned = 3;
4526 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 4527#ifdef __CHAR_UNSIGNED__
bc98a42c 4528 if (!sb_rdonly(sb))
23301410
TT		 4529 es->s_flags |=
4530 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4531 sbi->s_hash_unsigned = 3;
f99b2589 4532#else
bc98a42c 4533 if (!sb_rdonly(sb))
23301410
TT		 4534 es->s_flags |=
4535 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 4536#endif
23301410 4537 }
f99b2589 4538 }
ac27a0ec 4539
281b5995
TT		 4540 /* Handle clustersize */
4541 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
ef5fd681 4542 if (ext4_has_feature_bigalloc(sb)) {
281b5995
TT		 4543 if (clustersize < blocksize) {
4544 ext4_msg(sb, KERN_ERR,
4545 "cluster size (%d) smaller than "
4546 "block size (%d)", clustersize, blocksize);
4547 goto failed_mount;
4548 }
4549 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4550 le32_to_cpu(es->s_log_block_size);
4551 sbi->s_clusters_per_group =
4552 le32_to_cpu(es->s_clusters_per_group);
4553 if (sbi->s_clusters_per_group > blocksize * 8) {
4554 ext4_msg(sb, KERN_ERR,
4555 "#clusters per group too big: %lu",
4556 sbi->s_clusters_per_group);
4557 goto failed_mount;
4558 }
4559 if (sbi->s_blocks_per_group !=
4560 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4561 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4562 "clusters per group (%lu) inconsistent",
4563 sbi->s_blocks_per_group,
4564 sbi->s_clusters_per_group);
4565 goto failed_mount;
4566 }
4567 } else {
4568 if (clustersize != blocksize) {
bfe0a5f4
TT		 4569 ext4_msg(sb, KERN_ERR,
4570 "fragment/cluster size (%d) != "
4571 "block size (%d)", clustersize, blocksize);
4572 goto failed_mount;
281b5995
TT		 4573 }
4574 if (sbi->s_blocks_per_group > blocksize * 8) {
4575 ext4_msg(sb, KERN_ERR,
4576 "#blocks per group too big: %lu",
4577 sbi->s_blocks_per_group);
4578 goto failed_mount;
4579 }
4580 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4581 sbi->s_cluster_bits = 0;
ac27a0ec 4582 }
281b5995
TT		 4583 sbi->s_cluster_ratio = clustersize / blocksize;
4584
960fd856
TT		 4585 /* Do we have standard group size of clustersize * 8 blocks ? */
4586 if (sbi->s_blocks_per_group == clustersize << 3)
4587 set_opt2(sb, STD_GROUP_SIZE);
4588
bf43d84b
ES		 4589 /*
4590 * Test whether we have more sectors than will fit in sector_t,
4591 * and whether the max offset is addressable by the page cache.
4592 */
5a9ae68a 4593 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 4594 ext4_blocks_count(es));
5a9ae68a 4595 if (err) {
b31e1552 4596 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 4597 " too large to mount safely on this system");
ac27a0ec
DK		 4598 goto failed_mount;
4599 }
4600
617ba13b
MC		 4601 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4602 goto cantfind_ext4;
e7c95593 4603
0f2ddca6
FTN		 4604 /* check blocks count against device size */
4605 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4606 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 4607 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4608 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 4609 ext4_blocks_count(es), blocks_count);
4610 goto failed_mount;
4611 }
4612
0b8e58a1
AD		 4613 /*
4614 * It makes no sense for the first data block to be beyond the end
4615 * of the filesystem.
4616 */
4617 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 4618 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 4619 "block %u is beyond end of filesystem (%llu)",
4620 le32_to_cpu(es->s_first_data_block),
4621 ext4_blocks_count(es));
e7c95593
ES		 4622 goto failed_mount;
4623 }
bfe0a5f4
TT		 4624 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4625 (sbi->s_cluster_ratio == 1)) {
4626 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4627 "block is 0 with a 1k block and cluster size");
4628 goto failed_mount;
4629 }
4630
bd81d8ee
LV		 4631 blocks_count = (ext4_blocks_count(es) -
4632 le32_to_cpu(es->s_first_data_block) +
4633 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4634 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 4635 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
df41460a 4636 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4ec11028 4637 "(block count %llu, first data block %u, "
df41460a 4638 "blocks per group %lu)", blocks_count,
4ec11028
TT		 4639 ext4_blocks_count(es),
4640 le32_to_cpu(es->s_first_data_block),
4641 EXT4_BLOCKS_PER_GROUP(sb));
4642 goto failed_mount;
4643 }
bd81d8ee 4644 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 4645 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4646 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
9e463084
TT		 4647 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4648 le32_to_cpu(es->s_inodes_count)) {
4649 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4650 le32_to_cpu(es->s_inodes_count),
4651 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4652 ret = -EINVAL;
4653 goto failed_mount;
4654 }
617ba13b
MC		 4655 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4656 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 4657 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 4658 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 4659 ext4_msg(sb, KERN_WARNING,
4660 "first meta block group too large: %u "
4661 "(group descriptor block count %u)",
4662 le32_to_cpu(es->s_first_meta_bg), db_count);
4663 goto failed_mount;
4664 }
4665 }
1d0c3924
TT		 4666 rcu_assign_pointer(sbi->s_group_desc,
4667 kvmalloc_array(db_count,
4668 sizeof(struct buffer_head *),
4669 GFP_KERNEL));
ac27a0ec 4670 if (sbi->s_group_desc == NULL) {
b31e1552 4671 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 4672 ret = -ENOMEM;
ac27a0ec
DK		 4673 goto failed_mount;
4674 }
4675
705895b6 4676 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 4677
85c8f176
AP		 4678 /* Pre-read the descriptors into the buffer cache */
4679 for (i = 0; i < db_count; i++) {
4680 block = descriptor_loc(sb, logical_sb_block, i);
5df1d412 4681 ext4_sb_breadahead_unmovable(sb, block);
85c8f176
AP		 4682 }
4683
ac27a0ec 4684 for (i = 0; i < db_count; i++) {
1d0c3924
TT		 4685 struct buffer_head *bh;
4686
70bbb3e0 4687 block = descriptor_loc(sb, logical_sb_block, i);
8394a6ab 4688 bh = ext4_sb_bread_unmovable(sb, block);
4689 if (IS_ERR(bh)) {
b31e1552
ES		 4690 ext4_msg(sb, KERN_ERR,
4691 "can't read group descriptor %d", i);
ac27a0ec 4692 db_count = i;
8394a6ab 4693 ret = PTR_ERR(bh);
ac27a0ec
DK		 4694 goto failed_mount2;
4695 }
1d0c3924
TT		 4696 rcu_read_lock();
4697 rcu_dereference(sbi->s_group_desc)[i] = bh;
4698 rcu_read_unlock();
ac27a0ec 4699 }
44de022c 4700 sbi->s_gdb_count = db_count;
829fa70d 4701 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 4702 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 4703 ret = -EFSCORRUPTED;
f9ae9cf5 4704 goto failed_mount2;
ac27a0ec 4705 }
772cb7c8 4706
235699a8 4707 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
c92dc856
JK		 4708 spin_lock_init(&sbi->s_error_lock);
4709 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
04496411 4710
a75ae78f 4711 /* Register extent status tree shrinker */
eb68d0e2 4712 if (ext4_es_register_shrinker(sbi))
ce7e010a 4713 goto failed_mount3;
ce7e010a 4714
c9de560d 4715 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 4716 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 4717
f9ae9cf5
TT		 4718 /*
4719 * set up enough so that it can read an inode
4720 */
f6e63f90 4721 sb->s_op = &ext4_sops;
617ba13b
MC		 4722 sb->s_export_op = &ext4_export_ops;
4723 sb->s_xattr = ext4_xattr_handlers;
643fa961 4724#ifdef CONFIG_FS_ENCRYPTION
a7550b30 4725 sb->s_cop = &ext4_cryptops;
ffcc4182 4726#endif
c93d8f88
EB		 4727#ifdef CONFIG_FS_VERITY
4728 sb->s_vop = &ext4_verityops;
4729#endif
ac27a0ec 4730#ifdef CONFIG_QUOTA
617ba13b 4731 sb->dq_op = &ext4_quota_operations;
e2b911c5 4732 if (ext4_has_feature_quota(sb))
1fa5efe3 4733 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 4734 else
4735 sb->s_qcop = &ext4_qctl_operations;
689c958c 4736 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 4737#endif
85787090 4738 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 4739
ac27a0ec 4740 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 4741 mutex_init(&sbi->s_orphan_lock);
ac27a0ec 4742
aa75f4d3
HS		 4743 /* Initialize fast commit stuff */
4744 atomic_set(&sbi->s_fc_subtid, 0);
4745 atomic_set(&sbi->s_fc_ineligible_updates, 0);
4746 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4747 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4748 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4749 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4750 sbi->s_fc_bytes = 0;
9b5f6c9b
HS		 4751 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4752 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
aa75f4d3
HS		 4753 spin_lock_init(&sbi->s_fc_lock);
4754 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
8016e29f
HS		 4755 sbi->s_fc_replay_state.fc_regions = NULL;
4756 sbi->s_fc_replay_state.fc_regions_size = 0;
4757 sbi->s_fc_replay_state.fc_regions_used = 0;
4758 sbi->s_fc_replay_state.fc_regions_valid = 0;
4759 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4760 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4761 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
aa75f4d3 4762
ac27a0ec
DK		 4763 sb->s_root = NULL;
4764
4765 needs_recovery = (es->s_last_orphan != 0 ||
e2b911c5 4766 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 4767
bc98a42c 4768 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 4769 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 4770 goto failed_mount3a;
c5e06d10 4771
ac27a0ec
DK		 4772 /*
4773 * The first inode we look at is the journal inode. Don't try
4774 * root first: it may be modified in the journal!
4775 */
e2b911c5 4776 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4753d8a2
TT		 4777 err = ext4_load_journal(sb, es, journal_devnum);
4778 if (err)
50460fe8 4779 goto failed_mount3a;
bc98a42c 4780 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 4781 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 4782 ext4_msg(sb, KERN_ERR, "required journal recovery "
4783 "suppressed and not mounted read-only");
744692dc 4784 goto failed_mount_wq;
ac27a0ec 4785 } else {
1e381f60
DM		 4786 /* Nojournal mode, all journal mount options are illegal */
4787 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4788 ext4_msg(sb, KERN_ERR, "can't mount with "
4789 "journal_checksum, fs mounted w/o journal");
4790 goto failed_mount_wq;
4791 }
4792 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4793 ext4_msg(sb, KERN_ERR, "can't mount with "
4794 "journal_async_commit, fs mounted w/o journal");
4795 goto failed_mount_wq;
4796 }
4797 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4798 ext4_msg(sb, KERN_ERR, "can't mount with "
4799 "commit=%lu, fs mounted w/o journal",
4800 sbi->s_commit_interval / HZ);
4801 goto failed_mount_wq;
4802 }
4803 if (EXT4_MOUNT_DATA_FLAGS &
4804 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4805 ext4_msg(sb, KERN_ERR, "can't mount with "
4806 "data=, fs mounted w/o journal");
4807 goto failed_mount_wq;
4808 }
50b29d8f 4809 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
1e381f60 4810 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 4811 clear_opt(sb, DATA_FLAGS);
995a3ed6 4812 clear_opt2(sb, JOURNAL_FAST_COMMIT);
0390131b
FM		 4813 sbi->s_journal = NULL;
4814 needs_recovery = 0;
4815 goto no_journal;
ac27a0ec
DK		 4816 }
4817
e2b911c5 4818 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 4819 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4820 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 4821 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 4822 goto failed_mount_wq;
eb40a09c
JS		 4823 }
4824
25ed6e8a
DW		 4825 if (!set_journal_csum_feature_set(sb)) {
4826 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4827 "feature set");
4828 goto failed_mount_wq;
d4da6c9c 4829 }
818d276c 4830
a1e5e465
HS		 4831 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4832 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4833 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4834 ext4_msg(sb, KERN_ERR,
4835 "Failed to set fast commit journal feature");
4836 goto failed_mount_wq;
4837 }
4838
ac27a0ec
DK		 4839 /* We have now updated the journal if required, so we can
4840 * validate the data journaling mode. */
4841 switch (test_opt(sb, DATA_FLAGS)) {
4842 case 0:
4843 /* No mode set, assume a default based on the journal
63f57933
AM		 4844 * capabilities: ORDERED_DATA if the journal can
4845 * cope, else JOURNAL_DATA
4846 */
dab291af 4847 if (jbd2_journal_check_available_features
27f394a7 4848 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
fd8c37ec 4849 set_opt(sb, ORDERED_DATA);
27f394a7
TN		 4850 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4851 } else {
fd8c37ec 4852 set_opt(sb, JOURNAL_DATA);
27f394a7
TN		 4853 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4854 }
ac27a0ec
DK		 4855 break;
4856
617ba13b
MC		 4857 case EXT4_MOUNT_ORDERED_DATA:
4858 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4859 if (!jbd2_journal_check_available_features
4860 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4861 ext4_msg(sb, KERN_ERR, "Journal does not support "
4862 "requested data journaling mode");
744692dc 4863 goto failed_mount_wq;
ac27a0ec 4864 }
5a150bde 4865 break;
ac27a0ec
DK		 4866 default:
4867 break;
4868 }
ab04df78
JK		 4869
4870 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4871 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4872 ext4_msg(sb, KERN_ERR, "can't mount with "
4873 "journal_async_commit in data=ordered mode");
4874 goto failed_mount_wq;
4875 }
4876
b3881f74 4877 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4878
342af94e 4879 sbi->s_journal->j_submit_inode_data_buffers =
afb585a9 4880 ext4_journal_submit_inode_data_buffers;
342af94e 4881 sbi->s_journal->j_finish_inode_data_buffers =
afb585a9 4882 ext4_journal_finish_inode_data_buffers;
18aadd47 4883
ce7e010a 4884no_journal:
cdb7ee4c
TE		 4885 if (!test_opt(sb, NO_MBCACHE)) {
4886 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4887 if (!sbi->s_ea_block_cache) {
dec214d0 4888 ext4_msg(sb, KERN_ERR,
cdb7ee4c 4889 "Failed to create ea_block_cache");
dec214d0
TE		 4890 goto failed_mount_wq;
4891 }
cdb7ee4c
TE		 4892
4893 if (ext4_has_feature_ea_inode(sb)) {
4894 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4895 if (!sbi->s_ea_inode_cache) {
4896 ext4_msg(sb, KERN_ERR,
4897 "Failed to create ea_inode_cache");
4898 goto failed_mount_wq;
4899 }
4900 }
9c191f70
M		 4901 }
4902
c93d8f88
EB		 4903 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4904 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4905 goto failed_mount_wq;
4906 }
4907
bc98a42c 4908 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
e2b911c5
DW		 4909 !ext4_has_feature_encrypt(sb)) {
4910 ext4_set_feature_encrypt(sb);
4392fbc4 4911 ext4_commit_super(sb);
6ddb2447
TT		 4912 }
4913
952fc18e
TT		 4914 /*
4915 * Get the # of file system overhead blocks from the
4916 * superblock if present.
4917 */
4918 if (es->s_overhead_clusters)
4919 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4920 else {
07aa2ea1
LC		 4921 err = ext4_calculate_overhead(sb);
4922 if (err)
952fc18e
TT		 4923 goto failed_mount_wq;
4924 }
4925
fd89d5f2
TH		 4926 /*
4927 * The maximum number of concurrent works can be high and
4928 * concurrency isn't really necessary. Limit it to 1.
4929 */
2e8fa54e
JK		 4930 EXT4_SB(sb)->rsv_conversion_wq =
4931 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4932 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4933 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4934 ret = -ENOMEM;
2e8fa54e
JK		 4935 goto failed_mount4;
4936 }
4937
ac27a0ec 4938 /*
dab291af 4939 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4940 * so we can safely mount the rest of the filesystem now.
4941 */
4942
8a363970 4943 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
1d1fe1ee 4944 if (IS_ERR(root)) {
b31e1552 4945 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4946 ret = PTR_ERR(root);
32a9bb57 4947 root = NULL;
ac27a0ec
DK		 4948 goto failed_mount4;
4949 }
4950 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4951 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4952 iput(root);
ac27a0ec
DK		 4953 goto failed_mount4;
4954 }
b886ee3e 4955
48fde701 4956 sb->s_root = d_make_root(root);
1d1fe1ee 4957 if (!sb->s_root) {
b31e1552 4958 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4959 ret = -ENOMEM;
4960 goto failed_mount4;
4961 }
ac27a0ec 4962
c89128a0
JK		 4963 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4964 if (ret == -EROFS) {
1751e8a6 4965 sb->s_flags |= SB_RDONLY;
c89128a0
JK		 4966 ret = 0;
4967 } else if (ret)
4968 goto failed_mount4a;
ef7f3835 4969
b5799018 4970 ext4_set_resv_clusters(sb);
27dd4385 4971
0f5bde1d
JK		 4972 if (test_opt(sb, BLOCK_VALIDITY)) {
4973 err = ext4_setup_system_zone(sb);
4974 if (err) {
4975 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4976 "zone (%d)", err);
4977 goto failed_mount4a;
4978 }
f9ae9cf5 4979 }
8016e29f 4980 ext4_fc_replay_cleanup(sb);
f9ae9cf5
TT		 4981
4982 ext4_ext_init(sb);
4983 err = ext4_mb_init(sb);
4984 if (err) {
4985 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4986 err);
dcf2d804 4987 goto failed_mount5;
c2774d84
AK		 4988 }
4989
027f14f5
TT		 4990 /*
4991 * We can only set up the journal commit callback once
4992 * mballoc is initialized
4993 */
4994 if (sbi->s_journal)
4995 sbi->s_journal->j_commit_callback =
4996 ext4_journal_commit_callback;
4997
d5e03cbb
TT		 4998 block = ext4_count_free_clusters(sb);
4999 ext4_free_blocks_count_set(sbi->s_es,
5000 EXT4_C2B(sbi, block));
908c7f19
TH		 5001 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5002 GFP_KERNEL);
d5e03cbb
TT		 5003 if (!err) {
5004 unsigned long freei = ext4_count_free_inodes(sb);
5005 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 5006 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5007 GFP_KERNEL);
d5e03cbb
TT		 5008 }
5009 if (!err)
5010 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 5011 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 5012 if (!err)
908c7f19
TH		 5013 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5014 GFP_KERNEL);
efc61345
EW		 5015 if (!err)
5016 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
5017 GFP_KERNEL);
c8585c6f 5018 if (!err)
bbd55937 5019 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
c8585c6f 5020
d5e03cbb
TT		 5021 if (err) {
5022 ext4_msg(sb, KERN_ERR, "insufficient memory");
5023 goto failed_mount6;
5024 }
5025
e2b911c5 5026 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 5027 if (!ext4_fill_flex_info(sb)) {
5028 ext4_msg(sb, KERN_ERR,
5029 "unable to initialize "
5030 "flex_bg meta info!");
5031 goto failed_mount6;
5032 }
5033
bfff6873
LC		 5034 err = ext4_register_li_request(sb, first_not_zeroed);
5035 if (err)
dcf2d804 5036 goto failed_mount6;
bfff6873 5037
b5799018 5038 err = ext4_register_sysfs(sb);
dcf2d804
TM		 5039 if (err)
5040 goto failed_mount7;
3197ebdb 5041
9b2ff357
JK		 5042#ifdef CONFIG_QUOTA
5043 /* Enable quota usage during mount. */
bc98a42c 5044 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 5045 err = ext4_enable_quotas(sb);
5046 if (err)
5047 goto failed_mount8;
5048 }
5049#endif /* CONFIG_QUOTA */
5050
bc71726c 5051 /*
5052 * Save the original bdev mapping's wb_err value which could be
5053 * used to detect the metadata async write error.
5054 */
5055 spin_lock_init(&sbi->s_bdev_wb_lock);
9704a322
ZX		 5056 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5057 &sbi->s_bdev_wb_err);
bc71726c 5058 sb->s_bdev->bd_super = sb;
617ba13b
MC		 5059 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5060 ext4_orphan_cleanup(sb, es);
5061 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 5062 if (needs_recovery) {
b31e1552 5063 ext4_msg(sb, KERN_INFO, "recovery complete");
11215630
JK		 5064 err = ext4_mark_recovery_complete(sb, es);
5065 if (err)
5066 goto failed_mount8;
0390131b
FM		 5067 }
5068 if (EXT4_SB(sb)->s_journal) {
5069 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5070 descr = " journalled data mode";
5071 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5072 descr = " ordered data mode";
5073 else
5074 descr = " writeback data mode";
5075 } else
5076 descr = "out journal";
5077
79add3a3
LC		 5078 if (test_opt(sb, DISCARD)) {
5079 struct request_queue *q = bdev_get_queue(sb->s_bdev);
5080 if (!blk_queue_discard(q))
5081 ext4_msg(sb, KERN_WARNING,
5082 "mounting with \"discard\" option, but "
5083 "the device does not support discard");
5084 }
5085
e294a537
TT		 5086 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5087 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
ca9b404f 5088 "Opts: %.*s%s%s. Quota mode: %s.", descr,
5aee0f8a
TT		 5089 (int) sizeof(sbi->s_es->s_mount_opts),
5090 sbi->s_es->s_mount_opts,
ca9b404f
RA		 5091 *sbi->s_es->s_mount_opts ? "; " : "", orig_data,
5092 ext4_quota_mode(sb));
ac27a0ec 5093
66e61a9e
TT		 5094 if (es->s_error_count)
5095 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 5096
efbed4dc
TT		 5097 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5098 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5099 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5100 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
1cf006ed
DM		 5101 atomic_set(&sbi->s_warning_count, 0);
5102 atomic_set(&sbi->s_msg_count, 0);
efbed4dc 5103
d4c402d9 5104 kfree(orig_data);
ac27a0ec
DK		 5105 return 0;
5106
617ba13b 5107cantfind_ext4:
ac27a0ec 5108 if (!silent)
b31e1552 5109 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 5110 goto failed_mount;
5111
72ba7450 5112failed_mount8:
ebd173be 5113 ext4_unregister_sysfs(sb);
cb8d53d2 5114 kobject_put(&sbi->s_kobj);
dcf2d804
TM		 5115failed_mount7:
5116 ext4_unregister_li_request(sb);
5117failed_mount6:
f9ae9cf5 5118 ext4_mb_release(sb);
7c990728
SJS		 5119 rcu_read_lock();
5120 flex_groups = rcu_dereference(sbi->s_flex_groups);
5121 if (flex_groups) {
5122 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5123 kvfree(flex_groups[i]);
5124 kvfree(flex_groups);
5125 }
5126 rcu_read_unlock();
d5e03cbb
TT		 5127 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5128 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5129 percpu_counter_destroy(&sbi->s_dirs_counter);
5130 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
efc61345 5131 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
bbd55937 5132 percpu_free_rwsem(&sbi->s_writepages_rwsem);
00764937 5133failed_mount5:
f9ae9cf5
TT		 5134 ext4_ext_release(sb);
5135 ext4_release_system_zone(sb);
5136failed_mount4a:
94bf608a 5137 dput(sb->s_root);
32a9bb57 5138 sb->s_root = NULL;
94bf608a 5139failed_mount4:
b31e1552 5140 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 5141 if (EXT4_SB(sb)->rsv_conversion_wq)
5142 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 5143failed_mount_wq:
50c15df6
CX		 5144 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5145 sbi->s_ea_inode_cache = NULL;
5146
5147 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5148 sbi->s_ea_block_cache = NULL;
5149
0390131b
FM		 5150 if (sbi->s_journal) {
5151 jbd2_journal_destroy(sbi->s_journal);
5152 sbi->s_journal = NULL;
5153 }
50460fe8 5154failed_mount3a:
d3922a77 5155 ext4_es_unregister_shrinker(sbi);
eb68d0e2 5156failed_mount3:
9105bb14 5157 del_timer_sync(&sbi->s_err_report);
c92dc856 5158 flush_work(&sbi->s_error_work);
c5e06d10
JL		 5159 if (sbi->s_mmp_tsk)
5160 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 5161failed_mount2:
1d0c3924
TT		 5162 rcu_read_lock();
5163 group_desc = rcu_dereference(sbi->s_group_desc);
ac27a0ec 5164 for (i = 0; i < db_count; i++)
1d0c3924
TT		 5165 brelse(group_desc[i]);
5166 kvfree(group_desc);
5167 rcu_read_unlock();
ac27a0ec 5168failed_mount:
0441984a
DW		 5169 if (sbi->s_chksum_driver)
5170 crypto_free_shash(sbi->s_chksum_driver);
c83ad55e
GKB		 5171
5172#ifdef CONFIG_UNICODE
f8f4acb6 5173 utf8_unload(sb->s_encoding);
c83ad55e
GKB		 5174#endif
5175
ac27a0ec 5176#ifdef CONFIG_QUOTA
a2d4a646 5177 for (i = 0; i < EXT4_MAXQUOTAS; i++)
0ba33fac 5178 kfree(get_qf_name(sb, sbi, i));
ac27a0ec 5179#endif
ac4acb1f 5180 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
617ba13b 5181 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 5182 brelse(bh);
5183out_fail:
5184 sb->s_fs_info = NULL;
f6830165 5185 kfree(sbi->s_blockgroup_lock);
5aee0f8a 5186out_free_base:
ac27a0ec 5187 kfree(sbi);
d4c402d9 5188 kfree(orig_data);
5e405595 5189 fs_put_dax(dax_dev);
07aa2ea1 5190 return err ? err : ret;
ac27a0ec
DK		 5191}
5192
5193/*
5194 * Setup any per-fs journal parameters now. We'll do this both on
5195 * initial mount, once the journal has been initialised but before we've
5196 * done any recovery; and again on any subsequent remount.
5197 */
617ba13b 5198static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 5199{
617ba13b 5200 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5201
30773840
TT		 5202 journal->j_commit_interval = sbi->s_commit_interval;
5203 journal->j_min_batch_time = sbi->s_min_batch_time;
5204 journal->j_max_batch_time = sbi->s_max_batch_time;
6866d7b3 5205 ext4_fc_init(sb, journal);
ac27a0ec 5206
a931da6a 5207 write_lock(&journal->j_state_lock);
ac27a0ec 5208 if (test_opt(sb, BARRIER))
dab291af 5209 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 5210 else
dab291af 5211 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 5212 if (test_opt(sb, DATA_ERR_ABORT))
5213 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5214 else
5215 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 5216 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 5217}
5218
c6cb7e77
EW		 5219static struct inode *ext4_get_journal_inode(struct super_block *sb,
5220 unsigned int journal_inum)
ac27a0ec
DK		 5221{
5222 struct inode *journal_inode;
ac27a0ec 5223
c6cb7e77
EW		 5224 /*
5225 * Test for the existence of a valid inode on disk. Bad things
5226 * happen if we iget() an unused inode, as the subsequent iput()
5227 * will try to delete it.
5228 */
8a363970 5229 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
1d1fe1ee 5230 if (IS_ERR(journal_inode)) {
b31e1552 5231 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 5232 return NULL;
5233 }
5234 if (!journal_inode->i_nlink) {
5235 make_bad_inode(journal_inode);
5236 iput(journal_inode);
b31e1552 5237 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 5238 return NULL;
5239 }
5240
e5f8eab8 5241 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 5242 journal_inode, journal_inode->i_size);
1d1fe1ee 5243 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 5244 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 5245 iput(journal_inode);
5246 return NULL;
5247 }
c6cb7e77
EW		 5248 return journal_inode;
5249}
5250
5251static journal_t *ext4_get_journal(struct super_block *sb,
5252 unsigned int journal_inum)
5253{
5254 struct inode *journal_inode;
5255 journal_t *journal;
5256
11215630
JK		 5257 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5258 return NULL;
c6cb7e77
EW		 5259
5260 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5261 if (!journal_inode)
5262 return NULL;
ac27a0ec 5263
dab291af 5264 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 5265 if (!journal) {
b31e1552 5266 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 5267 iput(journal_inode);
5268 return NULL;
5269 }
5270 journal->j_private = sb;
617ba13b 5271 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 5272 return journal;
5273}
5274
617ba13b 5275static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 5276 dev_t j_dev)
5277{
2b2d6d01 5278 struct buffer_head *bh;
ac27a0ec 5279 journal_t *journal;
617ba13b
MC		 5280 ext4_fsblk_t start;
5281 ext4_fsblk_t len;
ac27a0ec 5282 int hblock, blocksize;
617ba13b 5283 ext4_fsblk_t sb_block;
ac27a0ec 5284 unsigned long offset;
2b2d6d01 5285 struct ext4_super_block *es;
ac27a0ec
DK		 5286 struct block_device *bdev;
5287
11215630
JK		 5288 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5289 return NULL;
0390131b 5290
b31e1552 5291 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 5292 if (bdev == NULL)
5293 return NULL;
5294
ac27a0ec 5295 blocksize = sb->s_blocksize;
e1defc4f 5296 hblock = bdev_logical_block_size(bdev);
ac27a0ec 5297 if (blocksize < hblock) {
b31e1552
ES		 5298 ext4_msg(sb, KERN_ERR,
5299 "blocksize too small for journal device");
ac27a0ec
DK		 5300 goto out_bdev;
5301 }
5302
617ba13b
MC		 5303 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5304 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 5305 set_blocksize(bdev, blocksize);
5306 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 5307 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5308 "external journal");
ac27a0ec
DK		 5309 goto out_bdev;
5310 }
5311
2716b802 5312 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 5313 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 5314 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 5315 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 5316 ext4_msg(sb, KERN_ERR, "external journal has "
5317 "bad superblock");
ac27a0ec
DK		 5318 brelse(bh);
5319 goto out_bdev;
5320 }
5321
df4763be
DW		 5322 if ((le32_to_cpu(es->s_feature_ro_compat) &
5323 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5324 es->s_checksum != ext4_superblock_csum(sb, es)) {
5325 ext4_msg(sb, KERN_ERR, "external journal has "
5326 "corrupt superblock");
5327 brelse(bh);
5328 goto out_bdev;
5329 }
5330
617ba13b 5331 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 5332 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 5333 brelse(bh);
5334 goto out_bdev;
5335 }
5336
bd81d8ee 5337 len = ext4_blocks_count(es);
ac27a0ec
DK		 5338 start = sb_block + 1;
5339 brelse(bh); /* we're done with the superblock */
5340
dab291af 5341 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 5342 start, len, blocksize);
5343 if (!journal) {
b31e1552 5344 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 5345 goto out_bdev;
5346 }
5347 journal->j_private = sb;
2d069c08 5348 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
b31e1552 5349 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 5350 goto out_journal;
5351 }
5352 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 5353 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5354 "user (unsupported) - %d",
ac27a0ec
DK		 5355 be32_to_cpu(journal->j_superblock->s_nr_users));
5356 goto out_journal;
5357 }
ee7ed3aa 5358 EXT4_SB(sb)->s_journal_bdev = bdev;
617ba13b 5359 ext4_init_journal_params(sb, journal);
ac27a0ec 5360 return journal;
0b8e58a1 5361
ac27a0ec 5362out_journal:
dab291af 5363 jbd2_journal_destroy(journal);
ac27a0ec 5364out_bdev:
617ba13b 5365 ext4_blkdev_put(bdev);
ac27a0ec
DK		 5366 return NULL;
5367}
5368
617ba13b
MC		 5369static int ext4_load_journal(struct super_block *sb,
5370 struct ext4_super_block *es,
ac27a0ec
DK		 5371 unsigned long journal_devnum)
5372{
5373 journal_t *journal;
5374 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5375 dev_t journal_dev;
5376 int err = 0;
5377 int really_read_only;
273108fa 5378 int journal_dev_ro;
ac27a0ec 5379
11215630
JK		 5380 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5381 return -EFSCORRUPTED;
0390131b 5382
ac27a0ec
DK		 5383 if (journal_devnum &&
5384 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 5385 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5386 "numbers have changed");
ac27a0ec
DK		 5387 journal_dev = new_decode_dev(journal_devnum);
5388 } else
5389 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5390
273108fa
LC		 5391 if (journal_inum && journal_dev) {
5392 ext4_msg(sb, KERN_ERR,
5393 "filesystem has both journal inode and journal device!");
5394 return -EINVAL;
5395 }
5396
5397 if (journal_inum) {
5398 journal = ext4_get_journal(sb, journal_inum);
5399 if (!journal)
5400 return -EINVAL;
5401 } else {
5402 journal = ext4_get_dev_journal(sb, journal_dev);
5403 if (!journal)
5404 return -EINVAL;
5405 }
5406
5407 journal_dev_ro = bdev_read_only(journal->j_dev);
5408 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5409
5410 if (journal_dev_ro && !sb_rdonly(sb)) {
5411 ext4_msg(sb, KERN_ERR,
5412 "journal device read-only, try mounting with '-o ro'");
5413 err = -EROFS;
5414 goto err_out;
5415 }
ac27a0ec
DK		 5416
5417 /*
5418 * Are we loading a blank journal or performing recovery after a
5419 * crash? For recovery, we need to check in advance whether we
5420 * can get read-write access to the device.
5421 */
e2b911c5 5422 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 5423 if (sb_rdonly(sb)) {
b31e1552
ES		 5424 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5425 "required on readonly filesystem");
ac27a0ec 5426 if (really_read_only) {
b31e1552 5427 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 5428 "unavailable, cannot proceed "
5429 "(try mounting with noload)");
273108fa
LC		 5430 err = -EROFS;
5431 goto err_out;
ac27a0ec 5432 }
b31e1552
ES		 5433 ext4_msg(sb, KERN_INFO, "write access will "
5434 "be enabled during recovery");
ac27a0ec
DK		 5435 }
5436 }
5437
90576c0b 5438 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 5439 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 5440
e2b911c5 5441 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 5442 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 5443 if (!err) {
5444 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5445 if (save)
5446 memcpy(save, ((char *) es) +
5447 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 5448 err = jbd2_journal_load(journal);
1c13d5c0
TT		 5449 if (save)
5450 memcpy(((char *) es) + EXT4_S_ERR_START,
5451 save, EXT4_S_ERR_LEN);
5452 kfree(save);
5453 }
ac27a0ec
DK		 5454
5455 if (err) {
b31e1552 5456 ext4_msg(sb, KERN_ERR, "error loading journal");
273108fa 5457 goto err_out;
ac27a0ec
DK		 5458 }
5459
617ba13b 5460 EXT4_SB(sb)->s_journal = journal;
11215630
JK		 5461 err = ext4_clear_journal_err(sb, es);
5462 if (err) {
5463 EXT4_SB(sb)->s_journal = NULL;
5464 jbd2_journal_destroy(journal);
5465 return err;
5466 }
ac27a0ec 5467
c41303ce 5468 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 5469 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5470 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 5471
5472 /* Make sure we flush the recovery flag to disk. */
4392fbc4 5473 ext4_commit_super(sb);
ac27a0ec
DK		 5474 }
5475
5476 return 0;
273108fa
LC		 5477
5478err_out:
5479 jbd2_journal_destroy(journal);
5480 return err;
ac27a0ec
DK		 5481}
5482
2d01ddc8
JK		 5483/* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5484static void ext4_update_super(struct super_block *sb)
ac27a0ec 5485{
c92dc856 5486 struct ext4_sb_info *sbi = EXT4_SB(sb);
e92ad03f
JK		 5487 struct ext4_super_block *es = sbi->s_es;
5488 struct buffer_head *sbh = sbi->s_sbh;
a17712c8 5489
05c2c00f 5490 lock_buffer(sbh);
71290b36
TT		 5491 /*
5492 * If the file system is mounted read-only, don't update the
5493 * superblock write time. This avoids updating the superblock
5494 * write time when we are mounting the root file system
5495 * read/only but we need to replay the journal; at that point,
5496 * for people who are east of GMT and who make their clock
5497 * tick in localtime for Windows bug-for-bug compatibility,
5498 * the clock is set in the future, and this will cause e2fsck
5499 * to complain and force a full file system check.
5500 */
1751e8a6 5501 if (!(sb->s_flags & SB_RDONLY))
6a0678a7 5502 ext4_update_tstamp(es, s_wtime);
8446fe92 5503 es->s_kbytes_written =
0bc9bc1d 5504 cpu_to_le64(sbi->s_kbytes_written +
8446fe92 5505 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
0bc9bc1d 5506 sbi->s_sectors_written_start) >> 1));
e92ad03f 5507 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
d5e03cbb 5508 ext4_free_blocks_count_set(es,
e92ad03f
JK		 5509 EXT4_C2B(sbi, percpu_counter_sum_positive(
5510 &sbi->s_freeclusters_counter)));
5511 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
d5e03cbb
TT		 5512 es->s_free_inodes_count =
5513 cpu_to_le32(percpu_counter_sum_positive(
e92ad03f 5514 &sbi->s_freeinodes_counter));
c92dc856
JK		 5515 /* Copy error information to the on-disk superblock */
5516 spin_lock(&sbi->s_error_lock);
5517 if (sbi->s_add_error_count > 0) {
5518 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5519 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
5520 __ext4_update_tstamp(&es->s_first_error_time,
5521 &es->s_first_error_time_hi,
5522 sbi->s_first_error_time);
5523 strncpy(es->s_first_error_func, sbi->s_first_error_func,
5524 sizeof(es->s_first_error_func));
5525 es->s_first_error_line =
5526 cpu_to_le32(sbi->s_first_error_line);
5527 es->s_first_error_ino =
5528 cpu_to_le32(sbi->s_first_error_ino);
5529 es->s_first_error_block =
5530 cpu_to_le64(sbi->s_first_error_block);
5531 es->s_first_error_errcode =
5532 ext4_errno_to_code(sbi->s_first_error_code);
5533 }
5534 __ext4_update_tstamp(&es->s_last_error_time,
5535 &es->s_last_error_time_hi,
5536 sbi->s_last_error_time);
5537 strncpy(es->s_last_error_func, sbi->s_last_error_func,
5538 sizeof(es->s_last_error_func));
5539 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
5540 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
5541 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
5542 es->s_last_error_errcode =
5543 ext4_errno_to_code(sbi->s_last_error_code);
5544 /*
5545 * Start the daily error reporting function if it hasn't been
5546 * started already
5547 */
5548 if (!es->s_error_count)
5549 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
5550 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
5551 sbi->s_add_error_count = 0;
5552 }
5553 spin_unlock(&sbi->s_error_lock);
5554
06db49e6 5555 ext4_superblock_csum_set(sb);
2d01ddc8
JK		 5556 unlock_buffer(sbh);
5557}
5558
5559static int ext4_commit_super(struct super_block *sb)
5560{
5561 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5562 int error = 0;
5563
5564 if (!sbh || block_device_ejected(sb))
5565 return error;
5566
5567 ext4_update_super(sb);
5568
e8680786 5569 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4743f839
PKS		 5570 /*
5571 * Oh, dear. A previous attempt to write the
5572 * superblock failed. This could happen because the
5573 * USB device was yanked out. Or it could happen to
5574 * be a transient write error and maybe the block will
5575 * be remapped. Nothing we can do but to retry the
5576 * write and hope for the best.
5577 */
5578 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5579 "superblock detected");
5580 clear_buffer_write_io_error(sbh);
5581 set_buffer_uptodate(sbh);
5582 }
2d01ddc8 5583 BUFFER_TRACE(sbh, "marking dirty");
ac27a0ec 5584 mark_buffer_dirty(sbh);
4392fbc4
JK		 5585 error = __sync_dirty_buffer(sbh,
5586 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5587 if (buffer_write_io_error(sbh)) {
5588 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5589 "superblock");
5590 clear_buffer_write_io_error(sbh);
5591 set_buffer_uptodate(sbh);
914258bf 5592 }
c4be0c1d 5593 return error;
ac27a0ec
DK		 5594}
5595
ac27a0ec
DK		 5596/*
5597 * Have we just finished recovery? If so, and if we are mounting (or
5598 * remounting) the filesystem readonly, then we will end up with a
5599 * consistent fs on disk. Record that fact.
5600 */
11215630
JK		 5601static int ext4_mark_recovery_complete(struct super_block *sb,
5602 struct ext4_super_block *es)
ac27a0ec 5603{
11215630 5604 int err;
617ba13b 5605 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 5606
e2b911c5 5607 if (!ext4_has_feature_journal(sb)) {
11215630
JK		 5608 if (journal != NULL) {
5609 ext4_error(sb, "Journal got removed while the fs was "
5610 "mounted!");
5611 return -EFSCORRUPTED;
5612 }
5613 return 0;
0390131b 5614 }
dab291af 5615 jbd2_journal_lock_updates(journal);
11215630
JK		 5616 err = jbd2_journal_flush(journal);
5617 if (err < 0)
7ffe1ea8
HK		 5618 goto out;
5619
bc98a42c 5620 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
e2b911c5 5621 ext4_clear_feature_journal_needs_recovery(sb);
4392fbc4 5622 ext4_commit_super(sb);
ac27a0ec 5623 }
7ffe1ea8 5624out:
dab291af 5625 jbd2_journal_unlock_updates(journal);
11215630 5626 return err;
ac27a0ec
DK		 5627}
5628
5629/*
5630 * If we are mounting (or read-write remounting) a filesystem whose journal
5631 * has recorded an error from a previous lifetime, move that error to the
5632 * main filesystem now.
5633 */
11215630 5634static int ext4_clear_journal_err(struct super_block *sb,
2b2d6d01 5635 struct ext4_super_block *es)
ac27a0ec
DK		 5636{
5637 journal_t *journal;
5638 int j_errno;
5639 const char *errstr;
5640
11215630
JK		 5641 if (!ext4_has_feature_journal(sb)) {
5642 ext4_error(sb, "Journal got removed while the fs was mounted!");
5643 return -EFSCORRUPTED;
5644 }
0390131b 5645
617ba13b 5646 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 5647
5648 /*
5649 * Now check for any error status which may have been recorded in the
617ba13b 5650 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 5651 */
5652
dab291af 5653 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 5654 if (j_errno) {
5655 char nbuf[16];
5656
617ba13b 5657 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 5658 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 5659 "from previous mount: %s", errstr);
12062ddd 5660 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 5661
617ba13b
MC		 5662 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5663 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4392fbc4 5664 ext4_commit_super(sb);
ac27a0ec 5665
dab291af 5666 jbd2_journal_clear_err(journal);
d796c52e 5667 jbd2_journal_update_sb_errno(journal);
ac27a0ec 5668 }
11215630 5669 return 0;
ac27a0ec
DK		 5670}
5671
5672/*
5673 * Force the running and committing transactions to commit,
5674 * and wait on the commit.
5675 */
617ba13b 5676int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 5677{
5678 journal_t *journal;
ac27a0ec 5679
bc98a42c 5680 if (sb_rdonly(sb))
ac27a0ec
DK		 5681 return 0;
5682
617ba13b 5683 journal = EXT4_SB(sb)->s_journal;
b1deefc9 5684 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 5685}
5686
617ba13b 5687static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 5688{
14ce0cb4 5689 int ret = 0;
9eddacf9 5690 tid_t target;
06a407f1 5691 bool needs_barrier = false;
8d5d02e6 5692 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5693
49598e04 5694 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 5695 return 0;
5696
9bffad1e 5697 trace_ext4_sync_fs(sb, wait);
2e8fa54e 5698 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 5699 /*
5700 * Writeback quota in non-journalled quota case - journalled quota has
5701 * no dirty dquots
5702 */
5703 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 5704 /*
5705 * Data writeback is possible w/o journal transaction, so barrier must
5706 * being sent at the end of the function. But we can skip it if
5707 * transaction_commit will do it for us.
5708 */
bda32530
TT		 5709 if (sbi->s_journal) {
5710 target = jbd2_get_latest_transaction(sbi->s_journal);
5711 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5712 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5713 needs_barrier = true;
5714
5715 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5716 if (wait)
5717 ret = jbd2_log_wait_commit(sbi->s_journal,
5718 target);
5719 }
5720 } else if (wait && test_opt(sb, BARRIER))
06a407f1 5721 needs_barrier = true;
06a407f1
DM		 5722 if (needs_barrier) {
5723 int err;
c6bf3f0e 5724 err = blkdev_issue_flush(sb->s_bdev);
06a407f1
DM		 5725 if (!ret)
5726 ret = err;
0390131b 5727 }
06a407f1
DM		 5728
5729 return ret;
5730}
5731
ac27a0ec
DK		 5732/*
5733 * LVM calls this function before a (read-only) snapshot is created. This
5734 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 5735 *
5736 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 5737 * state independently. It relies on upper layer to stop all data & metadata
5738 * modifications.
ac27a0ec 5739 */
c4be0c1d 5740static int ext4_freeze(struct super_block *sb)
ac27a0ec 5741{
c4be0c1d
TS		 5742 int error = 0;
5743 journal_t *journal;
ac27a0ec 5744
bc98a42c 5745 if (sb_rdonly(sb))
9ca92389 5746 return 0;
ac27a0ec 5747
9ca92389 5748 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 5749
bb044576
TT		 5750 if (journal) {
5751 /* Now we set up the journal barrier. */
5752 jbd2_journal_lock_updates(journal);
ac27a0ec 5753
bb044576
TT		 5754 /*
5755 * Don't clear the needs_recovery flag if we failed to
5756 * flush the journal.
5757 */
5758 error = jbd2_journal_flush(journal);
5759 if (error < 0)
5760 goto out;
c642dc9e
ES		 5761
5762 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 5763 ext4_clear_feature_journal_needs_recovery(sb);
bb044576 5764 }
9ca92389 5765
4392fbc4 5766 error = ext4_commit_super(sb);
6b0310fb 5767out:
bb044576
TT		 5768 if (journal)
5769 /* we rely on upper layer to stop further updates */
5770 jbd2_journal_unlock_updates(journal);
6b0310fb 5771 return error;
ac27a0ec
DK		 5772}
5773
5774/*
5775 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5776 * flag here, even though the filesystem is not technically dirty yet.
5777 */
c4be0c1d 5778static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 5779{
bc98a42c 5780 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 5781 return 0;
5782
c642dc9e
ES		 5783 if (EXT4_SB(sb)->s_journal) {
5784 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 5785 ext4_set_feature_journal_needs_recovery(sb);
c642dc9e
ES		 5786 }
5787
4392fbc4 5788 ext4_commit_super(sb);
c4be0c1d 5789 return 0;
ac27a0ec
DK		 5790}
5791
673c6100
TT		 5792/*
5793 * Structure to save mount options for ext4_remount's benefit
5794 */
5795struct ext4_mount_options {
5796 unsigned long s_mount_opt;
a2595b8a 5797 unsigned long s_mount_opt2;
08cefc7a
EB		 5798 kuid_t s_resuid;
5799 kgid_t s_resgid;
673c6100
TT		 5800 unsigned long s_commit_interval;
5801 u32 s_min_batch_time, s_max_batch_time;
5802#ifdef CONFIG_QUOTA
5803 int s_jquota_fmt;
a2d4a646 5804 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 5805#endif
5806};
5807
2b2d6d01 5808static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 5809{
2b2d6d01 5810 struct ext4_super_block *es;
617ba13b 5811 struct ext4_sb_info *sbi = EXT4_SB(sb);
f25391eb 5812 unsigned long old_sb_flags, vfs_flags;
617ba13b 5813 struct ext4_mount_options old_opts;
c79d967d 5814 int enable_quota = 0;
8a266467 5815 ext4_group_t g;
b3881f74 5816 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 5817 int err = 0;
ac27a0ec 5818#ifdef CONFIG_QUOTA
03dafb5f 5819 int i, j;
33458eab 5820 char *to_free[EXT4_MAXQUOTAS];
ac27a0ec 5821#endif
d4c402d9 5822 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec 5823
21ac738e
CX		 5824 if (data && !orig_data)
5825 return -ENOMEM;
5826
ac27a0ec
DK		 5827 /* Store the original options */
5828 old_sb_flags = sb->s_flags;
5829 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 5830 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 5831 old_opts.s_resuid = sbi->s_resuid;
5832 old_opts.s_resgid = sbi->s_resgid;
5833 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 5834 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5835 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 5836#ifdef CONFIG_QUOTA
5837 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 5838 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 5839 if (sbi->s_qf_names[i]) {
33458eab
TT		 5840 char *qf_name = get_qf_name(sb, sbi, i);
5841
5842 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
03dafb5f
CG		 5843 if (!old_opts.s_qf_names[i]) {
5844 for (j = 0; j < i; j++)
5845 kfree(old_opts.s_qf_names[j]);
3e36a163 5846 kfree(orig_data);
03dafb5f
CG		 5847 return -ENOMEM;
5848 }
5849 } else
5850 old_opts.s_qf_names[i] = NULL;
ac27a0ec 5851#endif
b3881f74
TT		 5852 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5853 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec 5854
f25391eb
LC		 5855 /*
5856 * Some options can be enabled by ext4 and/or by VFS mount flag
5857 * either way we need to make sure it matches in both *flags and
5858 * s_flags. Copy those selected flags from *flags to s_flags
5859 */
5860 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5861 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5862
661aa520 5863 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 5864 err = -EINVAL;
5865 goto restore_opts;
5866 }
5867
6b992ff2 5868 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 5869 test_opt(sb, JOURNAL_CHECKSUM)) {
5870 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 5871 "during remount not supported; ignoring");
5872 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 5873 }
5874
6ae6514b
PS		 5875 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5876 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5877 ext4_msg(sb, KERN_ERR, "can't mount with "
5878 "both data=journal and delalloc");
5879 err = -EINVAL;
5880 goto restore_opts;
5881 }
5882 if (test_opt(sb, DIOREAD_NOLOCK)) {
5883 ext4_msg(sb, KERN_ERR, "can't mount with "
5884 "both data=journal and dioread_nolock");
5885 err = -EINVAL;
5886 goto restore_opts;
5887 }
ab04df78
JK		 5888 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5889 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5890 ext4_msg(sb, KERN_ERR, "can't mount with "
5891 "journal_async_commit in data=ordered mode");
5892 err = -EINVAL;
5893 goto restore_opts;
5894 }
923ae0ff
RZ		 5895 }
5896
cdb7ee4c
TE		 5897 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5898 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5899 err = -EINVAL;
5900 goto restore_opts;
5901 }
5902
9b5f6c9b 5903 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
54d3adbc 5904 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
ac27a0ec 5905
1751e8a6
LT		 5906 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5907 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 5908
5909 es = sbi->s_es;
5910
b3881f74 5911 if (sbi->s_journal) {
0390131b 5912 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 5913 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5914 }
ac27a0ec 5915
c92dc856
JK		 5916 /* Flush outstanding errors before changing fs state */
5917 flush_work(&sbi->s_error_work);
5918
1751e8a6 5919 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
9b5f6c9b 5920 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
ac27a0ec
DK		 5921 err = -EROFS;
5922 goto restore_opts;
5923 }
5924
1751e8a6 5925 if (*flags & SB_RDONLY) {
38c03b34
TT		 5926 err = sync_filesystem(sb);
5927 if (err < 0)
5928 goto restore_opts;
0f0dd62f
CH		 5929 err = dquot_suspend(sb, -1);
5930 if (err < 0)
c79d967d 5931 goto restore_opts;
c79d967d 5932
ac27a0ec
DK		 5933 /*
5934 * First of all, the unconditional stuff we have to do
5935 * to disable replay of the journal when we next remount
5936 */
1751e8a6 5937 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 5938
5939 /*
5940 * OK, test if we are remounting a valid rw partition
5941 * readonly, and if so set the rdonly flag and then
5942 * mark the partition as valid again.
5943 */
617ba13b
MC		 5944 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5945 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 5946 es->s_state = cpu_to_le16(sbi->s_mount_state);
5947
11215630
JK		 5948 if (sbi->s_journal) {
5949 /*
5950 * We let remount-ro finish even if marking fs
5951 * as clean failed...
5952 */
0390131b 5953 ext4_mark_recovery_complete(sb, es);
11215630 5954 }
2dca60d9
TT		 5955 if (sbi->s_mmp_tsk)
5956 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 5957 } else {
a13fb1a4 5958 /* Make sure we can mount this feature set readwrite */
e2b911c5 5959 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 5960 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 5961 err = -EROFS;
5962 goto restore_opts;
5963 }
8a266467
TT		 5964 /*
5965 * Make sure the group descriptor checksums
0b8e58a1 5966 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 5967 */
5968 for (g = 0; g < sbi->s_groups_count; g++) {
5969 struct ext4_group_desc *gdp =
5970 ext4_get_group_desc(sb, g, NULL);
5971
feb0ab32 5972 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 5973 ext4_msg(sb, KERN_ERR,
5974 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 5975 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 5976 le16_to_cpu(gdp->bg_checksum));
6a797d27 5977 err = -EFSBADCRC;
8a266467
TT		 5978 goto restore_opts;
5979 }
5980 }
5981
ead6596b
ES		 5982 /*
5983 * If we have an unprocessed orphan list hanging
5984 * around from a previously readonly bdev mount,
5985 * require a full umount/remount for now.
5986 */
5987 if (es->s_last_orphan) {
b31e1552 5988 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 5989 "remount RDWR because of unprocessed "
5990 "orphan inode list. Please "
b31e1552 5991 "umount/remount instead");
ead6596b
ES		 5992 err = -EINVAL;
5993 goto restore_opts;
5994 }
5995
ac27a0ec
DK		 5996 /*
5997 * Mounting a RDONLY partition read-write, so reread
5998 * and store the current valid flag. (It may have
5999 * been changed by e2fsck since we originally mounted
6000 * the partition.)
6001 */
11215630
JK		 6002 if (sbi->s_journal) {
6003 err = ext4_clear_journal_err(sb, es);
6004 if (err)
6005 goto restore_opts;
6006 }
ac27a0ec 6007 sbi->s_mount_state = le16_to_cpu(es->s_state);
c89128a0
JK		 6008
6009 err = ext4_setup_super(sb, es, 0);
6010 if (err)
6011 goto restore_opts;
6012
6013 sb->s_flags &= ~SB_RDONLY;
e2b911c5 6014 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 6015 if (ext4_multi_mount_protect(sb,
6016 le64_to_cpu(es->s_mmp_block))) {
6017 err = -EROFS;
6018 goto restore_opts;
6019 }
c79d967d 6020 enable_quota = 1;
ac27a0ec
DK		 6021 }
6022 }
bfff6873
LC		 6023
6024 /*
6025 * Reinitialize lazy itable initialization thread based on
6026 * current settings
6027 */
bc98a42c 6028 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 6029 ext4_unregister_li_request(sb);
6030 else {
6031 ext4_group_t first_not_zeroed;
6032 first_not_zeroed = ext4_has_uninit_itable(sb);
6033 ext4_register_li_request(sb, first_not_zeroed);
6034 }
6035
0f5bde1d
JK		 6036 /*
6037 * Handle creation of system zone data early because it can fail.
6038 * Releasing of existing data is done when we are sure remount will
6039 * succeed.
6040 */
dd0db94f 6041 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
0f5bde1d
JK		 6042 err = ext4_setup_system_zone(sb);
6043 if (err)
6044 goto restore_opts;
6045 }
d176b1f6 6046
c89128a0 6047 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
4392fbc4 6048 err = ext4_commit_super(sb);
c89128a0
JK		 6049 if (err)
6050 goto restore_opts;
6051 }
0390131b 6052
ac27a0ec
DK		 6053#ifdef CONFIG_QUOTA
6054 /* Release old quota file names */
a2d4a646 6055 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 6056 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 6057 if (enable_quota) {
6058 if (sb_any_quota_suspended(sb))
6059 dquot_resume(sb, -1);
e2b911c5 6060 else if (ext4_has_feature_quota(sb)) {
7c319d32 6061 err = ext4_enable_quotas(sb);
07724f98 6062 if (err)
7c319d32 6063 goto restore_opts;
7c319d32
AK		 6064 }
6065 }
ac27a0ec 6066#endif
dd0db94f 6067 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
0f5bde1d 6068 ext4_release_system_zone(sb);
d4c402d9 6069
f25391eb
LC		 6070 /*
6071 * Some options can be enabled by ext4 and/or by VFS mount flag
6072 * either way we need to make sure it matches in both *flags and
6073 * s_flags. Copy those selected flags from s_flags to *flags
6074 */
6075 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
d4c402d9 6076
ca9b404f
RA		 6077 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.",
6078 orig_data, ext4_quota_mode(sb));
d4c402d9 6079 kfree(orig_data);
ac27a0ec 6080 return 0;
0b8e58a1 6081
ac27a0ec
DK		 6082restore_opts:
6083 sb->s_flags = old_sb_flags;
6084 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 6085 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 6086 sbi->s_resuid = old_opts.s_resuid;
6087 sbi->s_resgid = old_opts.s_resgid;
6088 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 6089 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6090 sbi->s_max_batch_time = old_opts.s_max_batch_time;
dd0db94f 6091 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
0f5bde1d 6092 ext4_release_system_zone(sb);
ac27a0ec
DK		 6093#ifdef CONFIG_QUOTA
6094 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 6095 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
33458eab
TT		 6096 to_free[i] = get_qf_name(sb, sbi, i);
6097 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
ac27a0ec 6098 }
33458eab
TT		 6099 synchronize_rcu();
6100 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6101 kfree(to_free[i]);
ac27a0ec 6102#endif
d4c402d9 6103 kfree(orig_data);
ac27a0ec
DK		 6104 return err;
6105}
6106
689c958c
LX		 6107#ifdef CONFIG_QUOTA
6108static int ext4_statfs_project(struct super_block *sb,
6109 kprojid_t projid, struct kstatfs *buf)
6110{
6111 struct kqid qid;
6112 struct dquot *dquot;
6113 u64 limit;
6114 u64 curblock;
6115
6116 qid = make_kqid_projid(projid);
6117 dquot = dqget(sb, qid);
6118 if (IS_ERR(dquot))
6119 return PTR_ERR(dquot);
7b9ca4c6 6120 spin_lock(&dquot->dq_dqb_lock);
689c958c 6121
a08fe66e
CX		 6122 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6123 dquot->dq_dqb.dqb_bhardlimit);
57c32ea4
CX		 6124 limit >>= sb->s_blocksize_bits;
6125
689c958c 6126 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 6127 curblock = (dquot->dq_dqb.dqb_curspace +
6128 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 6129 buf->f_blocks = limit;
6130 buf->f_bfree = buf->f_bavail =
6131 (buf->f_blocks > curblock) ?
6132 (buf->f_blocks - curblock) : 0;
6133 }
6134
a08fe66e
CX		 6135 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6136 dquot->dq_dqb.dqb_ihardlimit);
689c958c
LX		 6137 if (limit && buf->f_files > limit) {
6138 buf->f_files = limit;
6139 buf->f_ffree =
6140 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6141 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6142 }
6143
7b9ca4c6 6144 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 6145 dqput(dquot);
6146 return 0;
6147}
6148#endif
6149
2b2d6d01 6150static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 6151{
6152 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 6153 struct ext4_sb_info *sbi = EXT4_SB(sb);
6154 struct ext4_super_block *es = sbi->s_es;
27dd4385 6155 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 6156 u64 fsid;
d02a9391 6157 s64 bfree;
27dd4385 6158 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 6159
952fc18e
TT		 6160 if (!test_opt(sb, MINIX_DF))
6161 overhead = sbi->s_overhead;
ac27a0ec 6162
617ba13b 6163 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 6164 buf->f_bsize = sb->s_blocksize;
b72f78cb 6165 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 6166 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6167 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 6168 /* prevent underflow in case that few free space is available */
57042651 6169 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 6170 buf->f_bavail = buf->f_bfree -
6171 (ext4_r_blocks_count(es) + resv_blocks);
6172 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 6173 buf->f_bavail = 0;
6174 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 6175 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 6176 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 6177 fsid = le64_to_cpup((void *)es->s_uuid) ^
6178 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
6d1349c7 6179 buf->f_fsid = u64_to_fsid(fsid);
0b8e58a1 6180
689c958c
LX		 6181#ifdef CONFIG_QUOTA
6182 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6183 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6184 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6185#endif
ac27a0ec
DK		 6186 return 0;
6187}
6188
ac27a0ec
DK		 6189
6190#ifdef CONFIG_QUOTA
6191
bc8230ee
JK		 6192/*
6193 * Helper functions so that transaction is started before we acquire dqio_sem
6194 * to keep correct lock ordering of transaction > dqio_sem
6195 */
ac27a0ec
DK		 6196static inline struct inode *dquot_to_inode(struct dquot *dquot)
6197{
4c376dca 6198 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 6199}
6200
617ba13b 6201static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 6202{
6203 int ret, err;
6204 handle_t *handle;
6205 struct inode *inode;
6206
6207 inode = dquot_to_inode(dquot);
9924a92a 6208 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 6209 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 6210 if (IS_ERR(handle))
6211 return PTR_ERR(handle);
6212 ret = dquot_commit(dquot);
617ba13b 6213 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6214 if (!ret)
6215 ret = err;
6216 return ret;
6217}
6218
617ba13b 6219static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 6220{
6221 int ret, err;
6222 handle_t *handle;
6223
9924a92a 6224 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 6225 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 6226 if (IS_ERR(handle))
6227 return PTR_ERR(handle);
6228 ret = dquot_acquire(dquot);
617ba13b 6229 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6230 if (!ret)
6231 ret = err;
6232 return ret;
6233}
6234
617ba13b 6235static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 6236{
6237 int ret, err;
6238 handle_t *handle;
6239
9924a92a 6240 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 6241 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 6242 if (IS_ERR(handle)) {
6243 /* Release dquot anyway to avoid endless cycle in dqput() */
6244 dquot_release(dquot);
ac27a0ec 6245 return PTR_ERR(handle);
9c3013e9 6246 }
ac27a0ec 6247 ret = dquot_release(dquot);
617ba13b 6248 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6249 if (!ret)
6250 ret = err;
6251 return ret;
6252}
6253
617ba13b 6254static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 6255{
262b4662 6256 struct super_block *sb = dquot->dq_sb;
262b4662 6257
f177ee08 6258 if (ext4_is_quota_journalled(sb)) {
ac27a0ec 6259 dquot_mark_dquot_dirty(dquot);
617ba13b 6260 return ext4_write_dquot(dquot);
ac27a0ec
DK		 6261 } else {
6262 return dquot_mark_dquot_dirty(dquot);
6263 }
6264}
6265
617ba13b 6266static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 6267{
6268 int ret, err;
6269 handle_t *handle;
6270
6271 /* Data block + inode block */
2b0143b5 6272 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 6273 if (IS_ERR(handle))
6274 return PTR_ERR(handle);
6275 ret = dquot_commit_info(sb, type);
617ba13b 6276 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6277 if (!ret)
6278 ret = err;
6279 return ret;
6280}
6281
6282/*
6283 * Turn on quotas during mount time - we need to find
6284 * the quota file and such...
6285 */
617ba13b 6286static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 6287{
33458eab 6288 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
287a8095 6289 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 6290}
6291
daf647d2
TT		 6292static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6293{
6294 struct ext4_inode_info *ei = EXT4_I(inode);
6295
6296 /* The first argument of lockdep_set_subclass has to be
6297 * *exactly* the same as the argument to init_rwsem() --- in
6298 * this case, in init_once() --- or lockdep gets unhappy
6299 * because the name of the lock is set using the
6300 * stringification of the argument to init_rwsem().
6301 */
6302 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6303 lockdep_set_subclass(&ei->i_data_sem, subclass);
6304}
6305
ac27a0ec
DK		 6306/*
6307 * Standard function to be called on quota_on
6308 */
617ba13b 6309static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 6310 const struct path *path)
ac27a0ec
DK		 6311{
6312 int err;
ac27a0ec
DK		 6313
6314 if (!test_opt(sb, QUOTA))
6315 return -EINVAL;
0623543b 6316
ac27a0ec 6317 /* Quotafile not on the same filesystem? */
d8c9584e 6318 if (path->dentry->d_sb != sb)
ac27a0ec 6319 return -EXDEV;
e0770e91
JK		 6320
6321 /* Quota already enabled for this file? */
6322 if (IS_NOQUOTA(d_inode(path->dentry)))
6323 return -EBUSY;
6324
0623543b
JK		 6325 /* Journaling quota? */
6326 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 6327 /* Quotafile not in fs root? */
f00c9e44 6328 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 6329 ext4_msg(sb, KERN_WARNING,
6330 "Quota file not on filesystem root. "
6331 "Journaled quota will not work");
91389240
JK		 6332 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6333 } else {
6334 /*
6335 * Clear the flag just in case mount options changed since
6336 * last time.
6337 */
6338 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 6339 }
0623543b
JK		 6340
6341 /*
6342 * When we journal data on quota file, we have to flush journal to see
6343 * all updates to the file when we bypass pagecache...
6344 */
0390131b 6345 if (EXT4_SB(sb)->s_journal &&
2b0143b5 6346 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 6347 /*
6348 * We don't need to lock updates but journal_flush() could
6349 * otherwise be livelocked...
6350 */
6351 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 6352 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 6353 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 6354 if (err)
7ffe1ea8 6355 return err;
0623543b 6356 }
957153fc 6357
daf647d2
TT		 6358 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6359 err = dquot_quota_on(sb, type, format_id, path);
957153fc 6360 if (err) {
daf647d2
TT		 6361 lockdep_set_quota_inode(path->dentry->d_inode,
6362 I_DATA_SEM_NORMAL);
957153fc
JK		 6363 } else {
6364 struct inode *inode = d_inode(path->dentry);
6365 handle_t *handle;
6366
61a92987
JK		 6367 /*
6368 * Set inode flags to prevent userspace from messing with quota
6369 * files. If this fails, we return success anyway since quotas
6370 * are already enabled and this is not a hard failure.
6371 */
957153fc
JK		 6372 inode_lock(inode);
6373 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6374 if (IS_ERR(handle))
6375 goto unlock_inode;
6376 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6377 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6378 S_NOATIME | S_IMMUTABLE);
4209ae12 6379 err = ext4_mark_inode_dirty(handle, inode);
957153fc
JK		 6380 ext4_journal_stop(handle);
6381 unlock_inode:
6382 inode_unlock(inode);
6383 }
daf647d2 6384 return err;
ac27a0ec
DK		 6385}
6386
7c319d32
AK		 6387static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6388 unsigned int flags)
6389{
6390 int err;
6391 struct inode *qf_inode;
a2d4a646 6392 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 6393 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 6394 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6395 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 6396 };
6397
e2b911c5 6398 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 6399
6400 if (!qf_inums[type])
6401 return -EPERM;
6402
8a363970 6403 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7c319d32
AK		 6404 if (IS_ERR(qf_inode)) {
6405 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6406 return PTR_ERR(qf_inode);
6407 }
6408
bcb13850
JK		 6409 /* Don't account quota for quota files to avoid recursion */
6410 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 6411 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7212b95e 6412 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
daf647d2
TT		 6413 if (err)
6414 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
61157b24 6415 iput(qf_inode);
7c319d32
AK		 6416
6417 return err;
6418}
6419
6420/* Enable usage tracking for all quota types. */
6421static int ext4_enable_quotas(struct super_block *sb)
6422{
6423 int type, err = 0;
a2d4a646 6424 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 6425 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 6426 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6427 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 6428 };
49da9392
JK		 6429 bool quota_mopt[EXT4_MAXQUOTAS] = {
6430 test_opt(sb, USRQUOTA),
6431 test_opt(sb, GRPQUOTA),
6432 test_opt(sb, PRJQUOTA),
6433 };
7c319d32 6434
91389240 6435 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 6436 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 6437 if (qf_inums[type]) {
6438 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 6439 DQUOT_USAGE_ENABLED |
6440 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 6441 if (err) {
6442 ext4_warning(sb,
72ba7450
TT		 6443 "Failed to enable quota tracking "
6444 "(type=%d, err=%d). Please run "
6445 "e2fsck to fix.", type, err);
7f144fd0
JU		 6446 for (type--; type >= 0; type--)
6447 dquot_quota_off(sb, type);
6448
7c319d32
AK		 6449 return err;
6450 }
6451 }
6452 }
6453 return 0;
6454}
6455
ca0e05e4
DM		 6456static int ext4_quota_off(struct super_block *sb, int type)
6457{
21f97697
JK		 6458 struct inode *inode = sb_dqopt(sb)->files[type];
6459 handle_t *handle;
957153fc 6460 int err;
21f97697 6461
87009d86
DM		 6462 /* Force all delayed allocation blocks to be allocated.
6463 * Caller already holds s_umount sem */
6464 if (test_opt(sb, DELALLOC))
ca0e05e4 6465 sync_filesystem(sb);
ca0e05e4 6466
957153fc 6467 if (!inode || !igrab(inode))
0b268590
AG		 6468 goto out;
6469
957153fc 6470 err = dquot_quota_off(sb, type);
964edf66 6471 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 6472 goto out_put;
6473
6474 inode_lock(inode);
61a92987
JK		 6475 /*
6476 * Update modification times of quota files when userspace can
6477 * start looking at them. If we fail, we return success anyway since
6478 * this is not a hard failure and quotas are already disabled.
6479 */
9924a92a 6480 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
4209ae12
HS		 6481 if (IS_ERR(handle)) {
6482 err = PTR_ERR(handle);
957153fc 6483 goto out_unlock;
4209ae12 6484 }
957153fc
JK		 6485 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6486 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 6487 inode->i_mtime = inode->i_ctime = current_time(inode);
4209ae12 6488 err = ext4_mark_inode_dirty(handle, inode);
21f97697 6489 ext4_journal_stop(handle);
957153fc
JK		 6490out_unlock:
6491 inode_unlock(inode);
6492out_put:
964edf66 6493 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 6494 iput(inode);
6495 return err;
21f97697 6496out:
ca0e05e4
DM		 6497 return dquot_quota_off(sb, type);
6498}
6499
ac27a0ec
DK		 6500/* Read data from quotafile - avoid pagecache and such because we cannot afford
6501 * acquiring the locks... As quota files are never truncated and quota code
25985edc 6502 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 6503 * we don't have to be afraid of races */
617ba13b 6504static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 6505 size_t len, loff_t off)
6506{
6507 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 6508 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 6509 int offset = off & (sb->s_blocksize - 1);
6510 int tocopy;
6511 size_t toread;
6512 struct buffer_head *bh;
6513 loff_t i_size = i_size_read(inode);
6514
6515 if (off > i_size)
6516 return 0;
6517 if (off+len > i_size)
6518 len = i_size-off;
6519 toread = len;
6520 while (toread > 0) {
6521 tocopy = sb->s_blocksize - offset < toread ?
6522 sb->s_blocksize - offset : toread;
1c215028
TT		 6523 bh = ext4_bread(NULL, inode, blk, 0);
6524 if (IS_ERR(bh))
6525 return PTR_ERR(bh);
ac27a0ec
DK		 6526 if (!bh) /* A hole? */
6527 memset(data, 0, tocopy);
6528 else
6529 memcpy(data, bh->b_data+offset, tocopy);
6530 brelse(bh);
6531 offset = 0;
6532 toread -= tocopy;
6533 data += tocopy;
6534 blk++;
6535 }
6536 return len;
6537}
6538
6539/* Write to quotafile (we know the transaction is already started and has
6540 * enough credits) */
617ba13b 6541static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 6542 const char *data, size_t len, loff_t off)
6543{
6544 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 6545 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4209ae12 6546 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
c5e298ae 6547 int retries = 0;
ac27a0ec
DK		 6548 struct buffer_head *bh;
6549 handle_t *handle = journal_current_handle();
6550
0390131b 6551 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 6552 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6553 " cancelled because transaction is not started",
9c3013e9
JK		 6554 (unsigned long long)off, (unsigned long long)len);
6555 return -EIO;
6556 }
67eeb568
DM		 6557 /*
6558 * Since we account only one data block in transaction credits,
6559 * then it is impossible to cross a block boundary.
6560 */
6561 if (sb->s_blocksize - offset < len) {
6562 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6563 " cancelled because not block aligned",
6564 (unsigned long long)off, (unsigned long long)len);
6565 return -EIO;
6566 }
6567
c5e298ae
TT		 6568 do {
6569 bh = ext4_bread(handle, inode, blk,
6570 EXT4_GET_BLOCKS_CREATE |
6571 EXT4_GET_BLOCKS_METADATA_NOFAIL);
45586c70 6572 } while (PTR_ERR(bh) == -ENOSPC &&
c5e298ae 6573 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 6574 if (IS_ERR(bh))
6575 return PTR_ERR(bh);
67eeb568
DM		 6576 if (!bh)
6577 goto out;
5d601255 6578 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 6579 err = ext4_journal_get_write_access(handle, bh);
6580 if (err) {
6581 brelse(bh);
1c215028 6582 return err;
ac27a0ec 6583 }
67eeb568
DM		 6584 lock_buffer(bh);
6585 memcpy(bh->b_data+offset, data, len);
6586 flush_dcache_page(bh->b_page);
6587 unlock_buffer(bh);
62d2b5f2 6588 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 6589 brelse(bh);
ac27a0ec 6590out:
67eeb568
DM		 6591 if (inode->i_size < off + len) {
6592 i_size_write(inode, off + len);
617ba13b 6593 EXT4_I(inode)->i_disksize = inode->i_size;
4209ae12
HS		 6594 err2 = ext4_mark_inode_dirty(handle, inode);
6595 if (unlikely(err2 && !err))
6596 err = err2;
ac27a0ec 6597 }
4209ae12 6598 return err ? err : len;
ac27a0ec 6599}
ac27a0ec
DK		 6600#endif
6601
152a0836
AV		 6602static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6603 const char *dev_name, void *data)
ac27a0ec 6604{
152a0836 6605 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 6606}
6607
c290ea01 6608#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 6609static inline void register_as_ext2(void)
6610{
6611 int err = register_filesystem(&ext2_fs_type);
6612 if (err)
6613 printk(KERN_WARNING
6614 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6615}
6616
6617static inline void unregister_as_ext2(void)
6618{
6619 unregister_filesystem(&ext2_fs_type);
6620}
2035e776
TT		 6621
6622static inline int ext2_feature_set_ok(struct super_block *sb)
6623{
e2b911c5 6624 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 6625 return 0;
bc98a42c 6626 if (sb_rdonly(sb))
2035e776 6627 return 1;
e2b911c5 6628 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 6629 return 0;
6630 return 1;
6631}
24b58424
TT		 6632#else
6633static inline void register_as_ext2(void) { }
6634static inline void unregister_as_ext2(void) { }
2035e776 6635static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 6636#endif
6637
24b58424
TT		 6638static inline void register_as_ext3(void)
6639{
6640 int err = register_filesystem(&ext3_fs_type);
6641 if (err)
6642 printk(KERN_WARNING
6643 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6644}
6645
6646static inline void unregister_as_ext3(void)
6647{
6648 unregister_filesystem(&ext3_fs_type);
6649}
2035e776
TT		 6650
6651static inline int ext3_feature_set_ok(struct super_block *sb)
6652{
e2b911c5 6653 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 6654 return 0;
e2b911c5 6655 if (!ext4_has_feature_journal(sb))
2035e776 6656 return 0;
bc98a42c 6657 if (sb_rdonly(sb))
2035e776 6658 return 1;
e2b911c5 6659 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 6660 return 0;
6661 return 1;
6662}
24b58424 6663
03010a33
TT		 6664static struct file_system_type ext4_fs_type = {
6665 .owner = THIS_MODULE,
6666 .name = "ext4",
152a0836 6667 .mount = ext4_mount,
03010a33 6668 .kill_sb = kill_block_super,
14f3db55 6669 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
03010a33 6670};
7f78e035 6671MODULE_ALIAS_FS("ext4");
03010a33 6672
e9e3bcec
ES		 6673/* Shared across all ext4 file systems */
6674wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 6675
5dabfc78 6676static int __init ext4_init_fs(void)
ac27a0ec 6677{
e9e3bcec 6678 int i, err;
c9de560d 6679
e294a537 6680 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8 6681 ext4_li_info = NULL;
07c0c5d8 6682
9a4c8019 6683 /* Build-time check for flags consistency */
12e9b892 6684 ext4_check_flag_values();
e9e3bcec 6685
e142d052 6686 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 6687 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 6688
51865fda 6689 err = ext4_init_es();
6fd058f7
TT		 6690 if (err)
6691 return err;
51865fda 6692
1dc0aa46 6693 err = ext4_init_pending();
22cfe4b4
EB		 6694 if (err)
6695 goto out7;
6696
6697 err = ext4_init_post_read_processing();
1dc0aa46
EW		 6698 if (err)
6699 goto out6;
6700
51865fda
ZL		 6701 err = ext4_init_pageio();
6702 if (err)
b5799018 6703 goto out5;
51865fda 6704
5dabfc78 6705 err = ext4_init_system_zone();
bd2d0210 6706 if (err)
b5799018 6707 goto out4;
857ac889 6708
b5799018 6709 err = ext4_init_sysfs();
dd68314c 6710 if (err)
b5799018 6711 goto out3;
857ac889 6712
5dabfc78 6713 err = ext4_init_mballoc();
c9de560d
AT		 6714 if (err)
6715 goto out2;
ac27a0ec
DK		 6716 err = init_inodecache();
6717 if (err)
6718 goto out1;
aa75f4d3
HS		 6719
6720 err = ext4_fc_init_dentry_cache();
6721 if (err)
6722 goto out05;
6723
24b58424 6724 register_as_ext3();
2035e776 6725 register_as_ext2();
03010a33 6726 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 6727 if (err)
6728 goto out;
bfff6873 6729
ac27a0ec
DK		 6730 return 0;
6731out:
24b58424
TT		 6732 unregister_as_ext2();
6733 unregister_as_ext3();
aa75f4d3 6734out05:
ac27a0ec
DK		 6735 destroy_inodecache();
6736out1:
5dabfc78 6737 ext4_exit_mballoc();
9c191f70 6738out2:
b5799018
TT		 6739 ext4_exit_sysfs();
6740out3:
5dabfc78 6741 ext4_exit_system_zone();
b5799018 6742out4:
5dabfc78 6743 ext4_exit_pageio();
b5799018 6744out5:
22cfe4b4 6745 ext4_exit_post_read_processing();
1dc0aa46 6746out6:
22cfe4b4
EB		 6747 ext4_exit_pending();
6748out7:
51865fda
ZL		 6749 ext4_exit_es();
6750
ac27a0ec
DK		 6751 return err;
6752}
6753
5dabfc78 6754static void __exit ext4_exit_fs(void)
ac27a0ec 6755{
bfff6873 6756 ext4_destroy_lazyinit_thread();
24b58424
TT		 6757 unregister_as_ext2();
6758 unregister_as_ext3();
03010a33 6759 unregister_filesystem(&ext4_fs_type);
ac27a0ec 6760 destroy_inodecache();
5dabfc78 6761 ext4_exit_mballoc();
b5799018 6762 ext4_exit_sysfs();
5dabfc78
TT		 6763 ext4_exit_system_zone();
6764 ext4_exit_pageio();
22cfe4b4 6765 ext4_exit_post_read_processing();
dd12ed14 6766 ext4_exit_es();
1dc0aa46 6767 ext4_exit_pending();
ac27a0ec
DK		 6768}
6769
6770MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 6771MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 6772MODULE_LICENSE("GPL");
7ef79ad5 6773MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 6774module_init(ext4_init_fs)
6775module_exit(ext4_exit_fs)
Linux 6.x block layer and io_uring tree(s)