]> git.ipfire.org Git - people/ms/linux.git/blame - fs/ext4/super.c
projects / people / ms / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'for-6.0/dax' into libnvdimm-fixes
[people/ms/linux.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>
7c0f6ba6 42#include <linux/uaccess.h>
ee73f9a5 43#include <linux/iversion.h>
c83ad55e 44#include <linux/unicode.h>
c6a564ff 45#include <linux/part_stat.h>
bfff6873
LC		 46#include <linux/kthread.h>
47#include <linux/freezer.h>
9a089b21 48#include <linux/fsnotify.h>
e5a185c2
LC		 49#include <linux/fs_context.h>
50#include <linux/fs_parser.h>
bfff6873 51
3dcf5451 52#include "ext4.h"
4a092d73 53#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 54#include "ext4_jbd2.h"
ac27a0ec
DK		 55#include "xattr.h"
56#include "acl.h"
3661d286 57#include "mballoc.h"
0c9ec4be 58#include "fsmap.h"
ac27a0ec 59
9bffad1e
TT		 60#define CREATE_TRACE_POINTS
61#include <trace/events/ext4.h>
62
0b75a840 63static struct ext4_lazy_init *ext4_li_info;
59ebc7fd 64static DEFINE_MUTEX(ext4_li_mtx);
e294a537 65static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 66
617ba13b 67static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 68 unsigned long journal_devnum);
2adf6da8 69static int ext4_show_options(struct seq_file *seq, struct dentry *root);
2d01ddc8 70static void ext4_update_super(struct super_block *sb);
4392fbc4 71static int ext4_commit_super(struct super_block *sb);
11215630 72static int ext4_mark_recovery_complete(struct super_block *sb,
2b2d6d01 73 struct ext4_super_block *es);
11215630
JK		 74static int ext4_clear_journal_err(struct super_block *sb,
75 struct ext4_super_block *es);
617ba13b 76static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 77static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 78static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 79static int ext4_freeze(struct super_block *sb);
2035e776
TT		 80static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
bfff6873
LC		 82static void ext4_destroy_lazyinit_thread(void);
83static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 84static void ext4_clear_request_list(void);
c6cb7e77
EW		 85static struct inode *ext4_get_journal_inode(struct super_block *sb,
86 unsigned int journal_inum);
da812f61 87static int ext4_validate_options(struct fs_context *fc);
b6bd2435
LC		 88static int ext4_check_opt_consistency(struct fs_context *fc,
89 struct super_block *sb);
85456054 90static void ext4_apply_options(struct fs_context *fc, struct super_block *sb);
02f960f8 91static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param);
cebe85d5
LC		 92static int ext4_get_tree(struct fs_context *fc);
93static int ext4_reconfigure(struct fs_context *fc);
94static void ext4_fc_free(struct fs_context *fc);
95static int ext4_init_fs_context(struct fs_context *fc);
96static const struct fs_parameter_spec ext4_param_specs[];
ac27a0ec 97
e74031fd
JK		 98/*
99 * Lock ordering
100 *
e74031fd 101 * page fault path:
d4f5258e
JK		 102 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
103 * -> page lock -> i_data_sem (rw)
e74031fd
JK		 104 *
105 * buffered write path:
c1e8d7c6 106 * sb_start_write -> i_mutex -> mmap_lock
e74031fd
JK		 107 * sb_start_write -> i_mutex -> transaction start -> page lock ->
108 * i_data_sem (rw)
109 *
110 * truncate:
d4f5258e
JK		 111 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
112 * page lock
113 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
1d39834f 114 * i_data_sem (rw)
e74031fd
JK		 115 *
116 * direct IO:
c1e8d7c6 117 * sb_start_write -> i_mutex -> mmap_lock
1d39834f 118 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
e74031fd
JK		 119 *
120 * writepages:
121 * transaction start -> page lock(s) -> i_data_sem (rw)
122 */
123
02f960f8
LC		 124static const struct fs_context_operations ext4_context_ops = {
125 .parse_param = ext4_parse_param,
cebe85d5
LC		 126 .get_tree = ext4_get_tree,
127 .reconfigure = ext4_reconfigure,
128 .free = ext4_fc_free,
02f960f8
LC		 129};
130
131
c290ea01 132#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776 133static struct file_system_type ext2_fs_type = {
cebe85d5
LC		 134 .owner = THIS_MODULE,
135 .name = "ext2",
136 .init_fs_context = ext4_init_fs_context,
137 .parameters = ext4_param_specs,
138 .kill_sb = kill_block_super,
139 .fs_flags = FS_REQUIRES_DEV,
2035e776 140};
7f78e035 141MODULE_ALIAS_FS("ext2");
fa7614dd 142MODULE_ALIAS("ext2");
2035e776
TT		 143#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
144#else
145#define IS_EXT2_SB(sb) (0)
146#endif
147
148
ba69f9ab 149static struct file_system_type ext3_fs_type = {
cebe85d5
LC		 150 .owner = THIS_MODULE,
151 .name = "ext3",
152 .init_fs_context = ext4_init_fs_context,
153 .parameters = ext4_param_specs,
154 .kill_sb = kill_block_super,
155 .fs_flags = FS_REQUIRES_DEV,
ba69f9ab 156};
7f78e035 157MODULE_ALIAS_FS("ext3");
fa7614dd 158MODULE_ALIAS("ext3");
ba69f9ab 159#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
bd81d8ee 160
fa491b14 161
67c0f556 162static inline void __ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags,
fa491b14 163 bh_end_io_t *end_io)
164{
165 /*
166 * buffer's verified bit is no longer valid after reading from
167 * disk again due to write out error, clear it to make sure we
168 * recheck the buffer contents.
169 */
170 clear_buffer_verified(bh);
171
172 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
173 get_bh(bh);
1420c4a5 174 submit_bh(REQ_OP_READ | op_flags, bh);
fa491b14 175}
176
67c0f556 177void ext4_read_bh_nowait(struct buffer_head *bh, blk_opf_t op_flags,
fa491b14 178 bh_end_io_t *end_io)
179{
180 BUG_ON(!buffer_locked(bh));
181
182 if (ext4_buffer_uptodate(bh)) {
183 unlock_buffer(bh);
184 return;
185 }
186 __ext4_read_bh(bh, op_flags, end_io);
187}
188
67c0f556 189int ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags, bh_end_io_t *end_io)
fa491b14 190{
191 BUG_ON(!buffer_locked(bh));
192
193 if (ext4_buffer_uptodate(bh)) {
194 unlock_buffer(bh);
195 return 0;
196 }
197
198 __ext4_read_bh(bh, op_flags, end_io);
199
200 wait_on_buffer(bh);
201 if (buffer_uptodate(bh))
202 return 0;
203 return -EIO;
204}
205
67c0f556 206int ext4_read_bh_lock(struct buffer_head *bh, blk_opf_t op_flags, bool wait)
fa491b14 207{
208 if (trylock_buffer(bh)) {
209 if (wait)
210 return ext4_read_bh(bh, op_flags, NULL);
211 ext4_read_bh_nowait(bh, op_flags, NULL);
212 return 0;
213 }
214 if (wait) {
215 wait_on_buffer(bh);
216 if (buffer_uptodate(bh))
217 return 0;
218 return -EIO;
219 }
220 return 0;
221}
222
fb265c9c 223/*
8394a6ab 224 * This works like __bread_gfp() except it uses ERR_PTR for error
fb265c9c
TT		 225 * returns. Currently with sb_bread it's impossible to distinguish
226 * between ENOMEM and EIO situations (since both result in a NULL
227 * return.
228 */
8394a6ab 229static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
67c0f556
BVA		 230 sector_t block,
231 blk_opf_t op_flags, gfp_t gfp)
fb265c9c 232{
2d069c08 233 struct buffer_head *bh;
234 int ret;
fb265c9c 235
8394a6ab 236 bh = sb_getblk_gfp(sb, block, gfp);
fb265c9c
TT		 237 if (bh == NULL)
238 return ERR_PTR(-ENOMEM);
cf2834a5 239 if (ext4_buffer_uptodate(bh))
fb265c9c 240 return bh;
2d069c08 241
242 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
243 if (ret) {
244 put_bh(bh);
245 return ERR_PTR(ret);
246 }
247 return bh;
fb265c9c
TT		 248}
249
8394a6ab 250struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
67c0f556 251 blk_opf_t op_flags)
8394a6ab 252{
253 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
254}
255
256struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
257 sector_t block)
258{
259 return __ext4_sb_bread_gfp(sb, block, 0, 0);
260}
261
5df1d412 262void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
263{
264 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
265
266 if (likely(bh)) {
267 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
268 brelse(bh);
269 }
fb265c9c
TT		 270}
271
d25425f8
DW		 272static int ext4_verify_csum_type(struct super_block *sb,
273 struct ext4_super_block *es)
274{
e2b911c5 275 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 276 return 1;
277
278 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
279}
280
bbc605cd
LC		 281__le32 ext4_superblock_csum(struct super_block *sb,
282 struct ext4_super_block *es)
a9c47317
DW		 283{
284 struct ext4_sb_info *sbi = EXT4_SB(sb);
285 int offset = offsetof(struct ext4_super_block, s_checksum);
286 __u32 csum;
287
288 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
289
290 return cpu_to_le32(csum);
291}
292
c197855e
SH		 293static int ext4_superblock_csum_verify(struct super_block *sb,
294 struct ext4_super_block *es)
a9c47317 295{
9aa5d32b 296 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 297 return 1;
298
299 return es->s_checksum == ext4_superblock_csum(sb, es);
300}
301
06db49e6 302void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 303{
06db49e6
TT		 304 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
305
9aa5d32b 306 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 307 return;
308
309 es->s_checksum = ext4_superblock_csum(sb, es);
310}
311
8fadc143
AR		 312ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
313 struct ext4_group_desc *bg)
bd81d8ee 314{
3a14589c 315 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 316 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 317 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 318}
319
8fadc143
AR		 320ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
321 struct ext4_group_desc *bg)
bd81d8ee 322{
5272f837 323 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 324 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 325 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 326}
327
8fadc143
AR		 328ext4_fsblk_t ext4_inode_table(struct super_block *sb,
329 struct ext4_group_desc *bg)
bd81d8ee 330{
5272f837 331 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 332 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 333 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 334}
335
021b65bb
TT		 336__u32 ext4_free_group_clusters(struct super_block *sb,
337 struct ext4_group_desc *bg)
560671a0
AK		 338{
339 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
340 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 341 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 342}
343
344__u32 ext4_free_inodes_count(struct super_block *sb,
345 struct ext4_group_desc *bg)
346{
347 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
348 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 349 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 350}
351
352__u32 ext4_used_dirs_count(struct super_block *sb,
353 struct ext4_group_desc *bg)
354{
355 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
356 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 357 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 358}
359
360__u32 ext4_itable_unused_count(struct super_block *sb,
361 struct ext4_group_desc *bg)
362{
363 return le16_to_cpu(bg->bg_itable_unused_lo) |
364 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 365 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 366}
367
8fadc143
AR		 368void ext4_block_bitmap_set(struct super_block *sb,
369 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 370{
3a14589c 371 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 372 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
373 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 374}
375
8fadc143
AR		 376void ext4_inode_bitmap_set(struct super_block *sb,
377 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 378{
5272f837 379 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 380 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
381 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 382}
383
8fadc143
AR		 384void ext4_inode_table_set(struct super_block *sb,
385 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 386{
5272f837 387 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 388 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
389 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 390}
391
021b65bb
TT		 392void ext4_free_group_clusters_set(struct super_block *sb,
393 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 394{
395 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
396 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
397 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
398}
399
400void ext4_free_inodes_set(struct super_block *sb,
401 struct ext4_group_desc *bg, __u32 count)
402{
403 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
404 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
405 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
406}
407
408void ext4_used_dirs_set(struct super_block *sb,
409 struct ext4_group_desc *bg, __u32 count)
410{
411 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
412 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
413 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
414}
415
416void ext4_itable_unused_set(struct super_block *sb,
417 struct ext4_group_desc *bg, __u32 count)
418{
419 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
420 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
421 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
422}
423
c92dc856 424static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
6a0678a7 425{
6a0678a7
AB		 426 now = clamp_val(now, 0, (1ull << 40) - 1);
427
428 *lo = cpu_to_le32(lower_32_bits(now));
429 *hi = upper_32_bits(now);
430}
431
432static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
433{
434 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
435}
436#define ext4_update_tstamp(es, tstamp) \
c92dc856
JK		 437 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
438 ktime_get_real_seconds())
6a0678a7
AB		 439#define ext4_get_tstamp(es, tstamp) \
440 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
d3d1faf6 441
bdfe0cbd
TT		 442/*
443 * The del_gendisk() function uninitializes the disk-specific data
444 * structures, including the bdi structure, without telling anyone
445 * else. Once this happens, any attempt to call mark_buffer_dirty()
446 * (for example, by ext4_commit_super), will cause a kernel OOPS.
447 * This is a kludge to prevent these oops until we can put in a proper
448 * hook in del_gendisk() to inform the VFS and file system layers.
449 */
450static int block_device_ejected(struct super_block *sb)
451{
452 struct inode *bd_inode = sb->s_bdev->bd_inode;
453 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
454
455 return bdi->dev == NULL;
456}
457
18aadd47
BJ		 458static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
459{
460 struct super_block *sb = journal->j_private;
461 struct ext4_sb_info *sbi = EXT4_SB(sb);
462 int error = is_journal_aborted(journal);
5d3ee208 463 struct ext4_journal_cb_entry *jce;
18aadd47 464
5d3ee208 465 BUG_ON(txn->t_state == T_FINISHED);
a0154344
DJ		 466
467 ext4_process_freed_data(sb, txn->t_tid);
468
18aadd47 469 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 470 while (!list_empty(&txn->t_private_list)) {
471 jce = list_entry(txn->t_private_list.next,
472 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 473 list_del_init(&jce->jce_list);
474 spin_unlock(&sbi->s_md_lock);
475 jce->jce_func(sb, jce, error);
476 spin_lock(&sbi->s_md_lock);
477 }
478 spin_unlock(&sbi->s_md_lock);
479}
1c13d5c0 480
afb585a9
MFO		 481/*
482 * This writepage callback for write_cache_pages()
483 * takes care of a few cases after page cleaning.
484 *
485 * write_cache_pages() already checks for dirty pages
486 * and calls clear_page_dirty_for_io(), which we want,
487 * to write protect the pages.
488 *
489 * However, we may have to redirty a page (see below.)
490 */
491static int ext4_journalled_writepage_callback(struct page *page,
492 struct writeback_control *wbc,
493 void *data)
494{
495 transaction_t *transaction = (transaction_t *) data;
496 struct buffer_head *bh, *head;
497 struct journal_head *jh;
498
499 bh = head = page_buffers(page);
500 do {
501 /*
502 * We have to redirty a page in these cases:
503 * 1) If buffer is dirty, it means the page was dirty because it
504 * contains a buffer that needs checkpointing. So the dirty bit
505 * needs to be preserved so that checkpointing writes the buffer
506 * properly.
507 * 2) If buffer is not part of the committing transaction
508 * (we may have just accidentally come across this buffer because
509 * inode range tracking is not exact) or if the currently running
510 * transaction already contains this buffer as well, dirty bit
511 * needs to be preserved so that the buffer gets writeprotected
512 * properly on running transaction's commit.
513 */
514 jh = bh2jh(bh);
515 if (buffer_dirty(bh) ||
516 (jh && (jh->b_transaction != transaction ||
517 jh->b_next_transaction))) {
518 redirty_page_for_writepage(wbc, page);
519 goto out;
520 }
521 } while ((bh = bh->b_this_page) != head);
522
523out:
524 return AOP_WRITEPAGE_ACTIVATE;
525}
526
527static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
528{
529 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
530 struct writeback_control wbc = {
531 .sync_mode = WB_SYNC_ALL,
532 .nr_to_write = LONG_MAX,
533 .range_start = jinode->i_dirty_start,
534 .range_end = jinode->i_dirty_end,
535 };
536
537 return write_cache_pages(mapping, &wbc,
538 ext4_journalled_writepage_callback,
539 jinode->i_transaction);
540}
541
542static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
543{
544 int ret;
545
546 if (ext4_should_journal_data(jinode->i_vfs_inode))
547 ret = ext4_journalled_submit_inode_data_buffers(jinode);
548 else
549 ret = jbd2_journal_submit_inode_data_buffers(jinode);
550
551 return ret;
552}
553
554static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
555{
556 int ret = 0;
557
558 if (!ext4_should_journal_data(jinode->i_vfs_inode))
559 ret = jbd2_journal_finish_inode_data_buffers(jinode);
560
561 return ret;
562}
563
1dc1097f
JK		 564static bool system_going_down(void)
565{
566 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
567 || system_state == SYSTEM_RESTART;
568}
569
02a7780e
JK		 570struct ext4_err_translation {
571 int code;
572 int errno;
573};
574
575#define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
576
577static struct ext4_err_translation err_translation[] = {
578 EXT4_ERR_TRANSLATE(EIO),
579 EXT4_ERR_TRANSLATE(ENOMEM),
580 EXT4_ERR_TRANSLATE(EFSBADCRC),
581 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
582 EXT4_ERR_TRANSLATE(ENOSPC),
583 EXT4_ERR_TRANSLATE(ENOKEY),
584 EXT4_ERR_TRANSLATE(EROFS),
585 EXT4_ERR_TRANSLATE(EFBIG),
586 EXT4_ERR_TRANSLATE(EEXIST),
587 EXT4_ERR_TRANSLATE(ERANGE),
588 EXT4_ERR_TRANSLATE(EOVERFLOW),
589 EXT4_ERR_TRANSLATE(EBUSY),
590 EXT4_ERR_TRANSLATE(ENOTDIR),
591 EXT4_ERR_TRANSLATE(ENOTEMPTY),
592 EXT4_ERR_TRANSLATE(ESHUTDOWN),
593 EXT4_ERR_TRANSLATE(EFAULT),
594};
595
596static int ext4_errno_to_code(int errno)
597{
598 int i;
599
600 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
601 if (err_translation[i].errno == errno)
602 return err_translation[i].code;
603 return EXT4_ERR_UNKNOWN;
604}
605
2d01ddc8
JK		 606static void save_error_info(struct super_block *sb, int error,
607 __u32 ino, __u64 block,
608 const char *func, unsigned int line)
40676623 609{
c92dc856 610 struct ext4_sb_info *sbi = EXT4_SB(sb);
40676623 611
02a7780e
JK		 612 /* We default to EFSCORRUPTED error... */
613 if (error == 0)
614 error = EFSCORRUPTED;
c92dc856
JK		 615
616 spin_lock(&sbi->s_error_lock);
617 sbi->s_add_error_count++;
618 sbi->s_last_error_code = error;
619 sbi->s_last_error_line = line;
620 sbi->s_last_error_ino = ino;
621 sbi->s_last_error_block = block;
622 sbi->s_last_error_func = func;
623 sbi->s_last_error_time = ktime_get_real_seconds();
624 if (!sbi->s_first_error_time) {
625 sbi->s_first_error_code = error;
626 sbi->s_first_error_line = line;
627 sbi->s_first_error_ino = ino;
628 sbi->s_first_error_block = block;
629 sbi->s_first_error_func = func;
630 sbi->s_first_error_time = sbi->s_last_error_time;
631 }
632 spin_unlock(&sbi->s_error_lock);
40676623
JK		 633}
634
ac27a0ec
DK		 635/* Deal with the reporting of failure conditions on a filesystem such as
636 * inconsistencies detected or read IO failures.
637 *
638 * On ext2, we can store the error state of the filesystem in the
617ba13b 639 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 640 * write ordering constraints on the superblock which prevent us from
641 * writing it out straight away; and given that the journal is about to
642 * be aborted, we can't rely on the current, or future, transactions to
643 * write out the superblock safely.
644 *
dab291af 645 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 646 * the journal instead. On recovery, the journal will complain about
ac27a0ec 647 * that error until we've noted it down and cleared it.
014c9caa
JK		 648 *
649 * If force_ro is set, we unconditionally force the filesystem into an
650 * ABORT|READONLY state, unless the error response on the fs has been set to
651 * panic in which case we take the easy way out and panic immediately. This is
652 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
653 * at a critical moment in log management.
ac27a0ec 654 */
e789ca0c
JK		 655static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
656 __u32 ino, __u64 block,
657 const char *func, unsigned int line)
ac27a0ec 658{
b08070ec 659 journal_t *journal = EXT4_SB(sb)->s_journal;
2d01ddc8 660 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
b08070ec 661
e789ca0c 662 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
327eaf73
TT		 663 if (test_opt(sb, WARN_ON_ERROR))
664 WARN_ON_ONCE(1);
665
2d01ddc8
JK		 666 if (!continue_fs && !sb_rdonly(sb)) {
667 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
668 if (journal)
669 jbd2_journal_abort(journal, -EIO);
670 }
671
672 if (!bdev_read_only(sb->s_bdev)) {
e789ca0c 673 save_error_info(sb, error, ino, block, func, line);
2d01ddc8
JK		 674 /*
675 * In case the fs should keep running, we need to writeout
676 * superblock through the journal. Due to lock ordering
677 * constraints, it may not be safe to do it right here so we
678 * defer superblock flushing to a workqueue.
679 */
bb9464e0 680 if (continue_fs && journal)
2d01ddc8
JK		 681 schedule_work(&EXT4_SB(sb)->s_error_work);
682 else
683 ext4_commit_super(sb);
684 }
e789ca0c 685
1dc1097f
JK		 686 /*
687 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
688 * could panic during 'reboot -f' as the underlying device got already
689 * disabled.
690 */
014c9caa 691 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
617ba13b 692 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 693 sb->s_id);
4327ba52 694 }
ac2f7ca5
YB		 695
696 if (sb_rdonly(sb) || continue_fs)
697 return;
698
014c9caa
JK		 699 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
700 /*
701 * Make sure updated value of ->s_mount_flags will be visible before
702 * ->s_flags update
703 */
704 smp_wmb();
705 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 706}
707
c92dc856
JK		 708static void flush_stashed_error_work(struct work_struct *work)
709{
710 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
711 s_error_work);
2d01ddc8
JK		 712 journal_t *journal = sbi->s_journal;
713 handle_t *handle;
c92dc856 714
2d01ddc8
JK		 715 /*
716 * If the journal is still running, we have to write out superblock
717 * through the journal to avoid collisions of other journalled sb
718 * updates.
719 *
720 * We use directly jbd2 functions here to avoid recursing back into
721 * ext4 error handling code during handling of previous errors.
722 */
723 if (!sb_rdonly(sbi->s_sb) && journal) {
558d6450 724 struct buffer_head *sbh = sbi->s_sbh;
2d01ddc8
JK		 725 handle = jbd2_journal_start(journal, 1);
726 if (IS_ERR(handle))
727 goto write_directly;
558d6450 728 if (jbd2_journal_get_write_access(handle, sbh)) {
2d01ddc8
JK		 729 jbd2_journal_stop(handle);
730 goto write_directly;
731 }
732 ext4_update_super(sbi->s_sb);
558d6450
YB		 733 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
734 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
735 "superblock detected");
736 clear_buffer_write_io_error(sbh);
737 set_buffer_uptodate(sbh);
738 }
739
740 if (jbd2_journal_dirty_metadata(handle, sbh)) {
2d01ddc8
JK		 741 jbd2_journal_stop(handle);
742 goto write_directly;
743 }
744 jbd2_journal_stop(handle);
d578b994 745 ext4_notify_error_sysfs(sbi);
2d01ddc8
JK		 746 return;
747 }
748write_directly:
749 /*
750 * Write through journal failed. Write sb directly to get error info
751 * out and hope for the best.
752 */
4392fbc4 753 ext4_commit_super(sbi->s_sb);
d578b994 754 ext4_notify_error_sysfs(sbi);
ac27a0ec
DK		 755}
756
efbed4dc
TT		 757#define ext4_error_ratelimit(sb) \
758 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
759 "EXT4-fs error")
760
12062ddd 761void __ext4_error(struct super_block *sb, const char *function,
014c9caa 762 unsigned int line, bool force_ro, int error, __u64 block,
54d3adbc 763 const char *fmt, ...)
ac27a0ec 764{
0ff2ea7d 765 struct va_format vaf;
ac27a0ec
DK		 766 va_list args;
767
0db1ff22
TT		 768 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
769 return;
770
ccf0f32a 771 trace_ext4_error(sb, function, line);
efbed4dc
TT		 772 if (ext4_error_ratelimit(sb)) {
773 va_start(args, fmt);
774 vaf.fmt = fmt;
775 vaf.va = &args;
776 printk(KERN_CRIT
777 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
778 sb->s_id, function, line, current->comm, &vaf);
779 va_end(args);
780 }
9a089b21
GKB		 781 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
782
e789ca0c 783 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
ac27a0ec
DK		 784}
785
e7c96e8e 786void __ext4_error_inode(struct inode *inode, const char *function,
54d3adbc 787 unsigned int line, ext4_fsblk_t block, int error,
e7c96e8e 788 const char *fmt, ...)
273df556
FM		 789{
790 va_list args;
f7c21177 791 struct va_format vaf;
273df556 792
0db1ff22
TT		 793 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
794 return;
795
ccf0f32a 796 trace_ext4_error(inode->i_sb, function, line);
efbed4dc
TT		 797 if (ext4_error_ratelimit(inode->i_sb)) {
798 va_start(args, fmt);
799 vaf.fmt = fmt;
800 vaf.va = &args;
801 if (block)
802 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
803 "inode #%lu: block %llu: comm %s: %pV\n",
804 inode->i_sb->s_id, function, line, inode->i_ino,
805 block, current->comm, &vaf);
806 else
807 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
808 "inode #%lu: comm %s: %pV\n",
809 inode->i_sb->s_id, function, line, inode->i_ino,
810 current->comm, &vaf);
811 va_end(args);
812 }
9a089b21
GKB		 813 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
814
e789ca0c
JK		 815 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
816 function, line);
273df556
FM		 817}
818
e7c96e8e
JP		 819void __ext4_error_file(struct file *file, const char *function,
820 unsigned int line, ext4_fsblk_t block,
821 const char *fmt, ...)
273df556
FM		 822{
823 va_list args;
f7c21177 824 struct va_format vaf;
496ad9aa 825 struct inode *inode = file_inode(file);
273df556
FM		 826 char pathname[80], *path;
827
0db1ff22
TT		 828 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
829 return;
830
ccf0f32a 831 trace_ext4_error(inode->i_sb, function, line);
efbed4dc 832 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 833 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 834 if (IS_ERR(path))
835 path = "(unknown)";
836 va_start(args, fmt);
837 vaf.fmt = fmt;
838 vaf.va = &args;
839 if (block)
840 printk(KERN_CRIT
841 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
842 "block %llu: comm %s: path %s: %pV\n",
843 inode->i_sb->s_id, function, line, inode->i_ino,
844 block, current->comm, path, &vaf);
845 else
846 printk(KERN_CRIT
847 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
848 "comm %s: path %s: %pV\n",
849 inode->i_sb->s_id, function, line, inode->i_ino,
850 current->comm, path, &vaf);
851 va_end(args);
852 }
9a089b21
GKB		 853 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
854
e789ca0c
JK		 855 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
856 function, line);
273df556
FM		 857}
858
722887dd
TT		 859const char *ext4_decode_error(struct super_block *sb, int errno,
860 char nbuf[16])
ac27a0ec
DK		 861{
862 char *errstr = NULL;
863
864 switch (errno) {
6a797d27
DW		 865 case -EFSCORRUPTED:
866 errstr = "Corrupt filesystem";
867 break;
868 case -EFSBADCRC:
869 errstr = "Filesystem failed CRC";
870 break;
ac27a0ec
DK		 871 case -EIO:
872 errstr = "IO failure";
873 break;
874 case -ENOMEM:
875 errstr = "Out of memory";
876 break;
877 case -EROFS:
78f1ddbb
TT		 878 if (!sb || (EXT4_SB(sb)->s_journal &&
879 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 880 errstr = "Journal has aborted";
881 else
882 errstr = "Readonly filesystem";
883 break;
884 default:
885 /* If the caller passed in an extra buffer for unknown
886 * errors, textualise them now. Else we just return
887 * NULL. */
888 if (nbuf) {
889 /* Check for truncated error codes... */
890 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
891 errstr = nbuf;
892 }
893 break;
894 }
895
896 return errstr;
897}
898
617ba13b 899/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 900 * automatically and invokes the appropriate error response. */
901
c398eda0
TT		 902void __ext4_std_error(struct super_block *sb, const char *function,
903 unsigned int line, int errno)
ac27a0ec
DK		 904{
905 char nbuf[16];
906 const char *errstr;
907
0db1ff22
TT		 908 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
909 return;
910
ac27a0ec
DK		 911 /* Special case: if the error is EROFS, and we're not already
912 * inside a transaction, then there's really no point in logging
913 * an error. */
bc98a42c 914 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
ac27a0ec
DK		 915 return;
916
efbed4dc
TT		 917 if (ext4_error_ratelimit(sb)) {
918 errstr = ext4_decode_error(sb, errno, nbuf);
919 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
920 sb->s_id, function, line, errstr);
921 }
9a089b21 922 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
ac27a0ec 923
e789ca0c 924 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
ac27a0ec
DK		 925}
926
e7c96e8e
JP		 927void __ext4_msg(struct super_block *sb,
928 const char *prefix, const char *fmt, ...)
b31e1552 929{
0ff2ea7d 930 struct va_format vaf;
b31e1552
ES		 931 va_list args;
932
da812f61
LC		 933 if (sb) {
934 atomic_inc(&EXT4_SB(sb)->s_msg_count);
935 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
936 "EXT4-fs"))
937 return;
938 }
efbed4dc 939
b31e1552 940 va_start(args, fmt);
0ff2ea7d
JP		 941 vaf.fmt = fmt;
942 vaf.va = &args;
da812f61
LC		 943 if (sb)
944 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
945 else
946 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
b31e1552
ES		 947 va_end(args);
948}
949
1cf006ed
DM		 950static int ext4_warning_ratelimit(struct super_block *sb)
951{
952 atomic_inc(&EXT4_SB(sb)->s_warning_count);
953 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
954 "EXT4-fs warning");
955}
b03a2f7e 956
12062ddd 957void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 958 unsigned int line, const char *fmt, ...)
ac27a0ec 959{
0ff2ea7d 960 struct va_format vaf;
ac27a0ec
DK		 961 va_list args;
962
b03a2f7e 963 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 964 return;
965
ac27a0ec 966 va_start(args, fmt);
0ff2ea7d
JP		 967 vaf.fmt = fmt;
968 vaf.va = &args;
969 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
970 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 971 va_end(args);
972}
973
b03a2f7e
AD		 974void __ext4_warning_inode(const struct inode *inode, const char *function,
975 unsigned int line, const char *fmt, ...)
976{
977 struct va_format vaf;
978 va_list args;
979
980 if (!ext4_warning_ratelimit(inode->i_sb))
981 return;
982
983 va_start(args, fmt);
984 vaf.fmt = fmt;
985 vaf.va = &args;
986 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
987 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
988 function, line, inode->i_ino, current->comm, &vaf);
989 va_end(args);
990}
991
e29136f8
TT		 992void __ext4_grp_locked_error(const char *function, unsigned int line,
993 struct super_block *sb, ext4_group_t grp,
994 unsigned long ino, ext4_fsblk_t block,
995 const char *fmt, ...)
5d1b1b3f
AK		 996__releases(bitlock)
997__acquires(bitlock)
998{
0ff2ea7d 999 struct va_format vaf;
5d1b1b3f 1000 va_list args;
5d1b1b3f 1001
0db1ff22
TT		 1002 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
1003 return;
1004
ccf0f32a 1005 trace_ext4_error(sb, function, line);
efbed4dc
TT		 1006 if (ext4_error_ratelimit(sb)) {
1007 va_start(args, fmt);
1008 vaf.fmt = fmt;
1009 vaf.va = &args;
1010 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1011 sb->s_id, function, line, grp);
1012 if (ino)
1013 printk(KERN_CONT "inode %lu: ", ino);
1014 if (block)
1015 printk(KERN_CONT "block %llu:",
1016 (unsigned long long) block);
1017 printk(KERN_CONT "%pV\n", &vaf);
1018 va_end(args);
1019 }
5d1b1b3f
AK		 1020
1021 if (test_opt(sb, ERRORS_CONT)) {
c92dc856
JK		 1022 if (test_opt(sb, WARN_ON_ERROR))
1023 WARN_ON_ONCE(1);
e789ca0c 1024 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2d01ddc8
JK		 1025 if (!bdev_read_only(sb->s_bdev)) {
1026 save_error_info(sb, EFSCORRUPTED, ino, block, function,
1027 line);
e789ca0c 1028 schedule_work(&EXT4_SB(sb)->s_error_work);
2d01ddc8 1029 }
5d1b1b3f
AK		 1030 return;
1031 }
1032 ext4_unlock_group(sb, grp);
e789ca0c 1033 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
5d1b1b3f
AK		 1034 /*
1035 * We only get here in the ERRORS_RO case; relocking the group
1036 * may be dangerous, but nothing bad will happen since the
1037 * filesystem will have already been marked read/only and the
1038 * journal has been aborted. We return 1 as a hint to callers
1039 * who might what to use the return value from
25985edc 1040 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 1041 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1042 * aggressively from the ext4 function in question, with a
1043 * more appropriate error code.
1044 */
1045 ext4_lock_group(sb, grp);
1046 return;
1047}
1048
db79e6d1
WS		 1049void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1050 ext4_group_t group,
1051 unsigned int flags)
1052{
1053 struct ext4_sb_info *sbi = EXT4_SB(sb);
1054 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1055 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
9af0b3d1
WS		 1056 int ret;
1057
1058 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1059 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1060 &grp->bb_state);
1061 if (!ret)
1062 percpu_counter_sub(&sbi->s_freeclusters_counter,
1063 grp->bb_free);
db79e6d1
WS		 1064 }
1065
9af0b3d1
WS		 1066 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1067 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1068 &grp->bb_state);
1069 if (!ret && gdp) {
db79e6d1
WS		 1070 int count;
1071
1072 count = ext4_free_inodes_count(sb, gdp);
1073 percpu_counter_sub(&sbi->s_freeinodes_counter,
1074 count);
1075 }
db79e6d1
WS		 1076 }
1077}
1078
617ba13b 1079void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 1080{
617ba13b 1081 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 1082
617ba13b 1083 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 1084 return;
1085
12062ddd 1086 ext4_warning(sb,
ac27a0ec
DK		 1087 "updating to rev %d because of new feature flag, "
1088 "running e2fsck is recommended",
617ba13b 1089 EXT4_DYNAMIC_REV);
ac27a0ec 1090
617ba13b
MC		 1091 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1092 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1093 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 1094 /* leave es->s_feature_*compat flags alone */
1095 /* es->s_uuid will be set by e2fsck if empty */
1096
1097 /*
1098 * The rest of the superblock fields should be zero, and if not it
1099 * means they are likely already in use, so leave them alone. We
1100 * can leave it up to e2fsck to clean up any inconsistencies there.
1101 */
1102}
1103
1104/*
1105 * Open the external journal device
1106 */
b31e1552 1107static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 1108{
1109 struct block_device *bdev;
ac27a0ec 1110
d4d77629 1111 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 1112 if (IS_ERR(bdev))
1113 goto fail;
1114 return bdev;
1115
1116fail:
ea3edd4d
CH		 1117 ext4_msg(sb, KERN_ERR,
1118 "failed to open journal device unknown-block(%u,%u) %ld",
1119 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
ac27a0ec
DK		 1120 return NULL;
1121}
1122
1123/*
1124 * Release the journal device
1125 */
4385bab1 1126static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 1127{
4385bab1 1128 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 1129}
1130
4385bab1 1131static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 1132{
1133 struct block_device *bdev;
ee7ed3aa 1134 bdev = sbi->s_journal_bdev;
ac27a0ec 1135 if (bdev) {
4385bab1 1136 ext4_blkdev_put(bdev);
ee7ed3aa 1137 sbi->s_journal_bdev = NULL;
ac27a0ec 1138 }
ac27a0ec
DK		 1139}
1140
1141static inline struct inode *orphan_list_entry(struct list_head *l)
1142{
617ba13b 1143 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 1144}
1145
617ba13b 1146static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 1147{
1148 struct list_head *l;
1149
b31e1552
ES		 1150 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1151 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 1152
1153 printk(KERN_ERR "sb_info orphan list:\n");
1154 list_for_each(l, &sbi->s_orphan) {
1155 struct inode *inode = orphan_list_entry(l);
1156 printk(KERN_ERR " "
1157 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1158 inode->i_sb->s_id, inode->i_ino, inode,
1159 inode->i_mode, inode->i_nlink,
1160 NEXT_ORPHAN(inode));
1161 }
1162}
1163
957153fc
JK		 1164#ifdef CONFIG_QUOTA
1165static int ext4_quota_off(struct super_block *sb, int type);
1166
1167static inline void ext4_quota_off_umount(struct super_block *sb)
1168{
1169 int type;
1170
964edf66
JK		 1171 /* Use our quota_off function to clear inode flags etc. */
1172 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1173 ext4_quota_off(sb, type);
957153fc 1174}
33458eab
TT		 1175
1176/*
1177 * This is a helper function which is used in the mount/remount
1178 * codepaths (which holds s_umount) to fetch the quota file name.
1179 */
1180static inline char *get_qf_name(struct super_block *sb,
1181 struct ext4_sb_info *sbi,
1182 int type)
1183{
1184 return rcu_dereference_protected(sbi->s_qf_names[type],
1185 lockdep_is_held(&sb->s_umount));
1186}
957153fc
JK		 1187#else
1188static inline void ext4_quota_off_umount(struct super_block *sb)
1189{
1190}
1191#endif
1192
2b2d6d01 1193static void ext4_put_super(struct super_block *sb)
ac27a0ec 1194{
617ba13b
MC		 1195 struct ext4_sb_info *sbi = EXT4_SB(sb);
1196 struct ext4_super_block *es = sbi->s_es;
1d0c3924 1197 struct buffer_head **group_desc;
7c990728 1198 struct flex_groups **flex_groups;
97abd7d4 1199 int aborted = 0;
ef2cabf7 1200 int i, err;
ac27a0ec 1201
5e47868f
RH		 1202 /*
1203 * Unregister sysfs before destroying jbd2 journal.
1204 * Since we could still access attr_journal_task attribute via sysfs
1205 * path which could have sbi->s_journal->j_task as NULL
b98535d0
YB		 1206 * Unregister sysfs before flush sbi->s_error_work.
1207 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1208 * read metadata verify failed then will queue error work.
1209 * flush_stashed_error_work will call start_this_handle may trigger
1210 * BUG_ON.
5e47868f
RH		 1211 */
1212 ext4_unregister_sysfs(sb);
1213
4808cb5b
ZY		 1214 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
1215 ext4_msg(sb, KERN_INFO, "unmounting filesystem.");
1216
b98535d0
YB		 1217 ext4_unregister_li_request(sb);
1218 ext4_quota_off_umount(sb);
1219
1220 flush_work(&sbi->s_error_work);
1221 destroy_workqueue(sbi->rsv_conversion_wq);
1222 ext4_release_orphan_info(sb);
1223
0390131b 1224 if (sbi->s_journal) {
97abd7d4 1225 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 1226 err = jbd2_journal_destroy(sbi->s_journal);
1227 sbi->s_journal = NULL;
878520ac 1228 if ((err < 0) && !aborted) {
54d3adbc 1229 ext4_abort(sb, -err, "Couldn't clean up the journal");
878520ac 1230 }
0390131b 1231 }
d4edac31 1232
d3922a77 1233 ext4_es_unregister_shrinker(sbi);
9105bb14 1234 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 1235 ext4_release_system_zone(sb);
1236 ext4_mb_release(sb);
1237 ext4_ext_release(sb);
d4edac31 1238
bc98a42c 1239 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 1240 ext4_clear_feature_journal_needs_recovery(sb);
02f310fc 1241 ext4_clear_feature_orphan_present(sb);
ac27a0ec 1242 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 1243 }
bc98a42c 1244 if (!sb_rdonly(sb))
4392fbc4 1245 ext4_commit_super(sb);
a8e25a83 1246
1d0c3924
TT		 1247 rcu_read_lock();
1248 group_desc = rcu_dereference(sbi->s_group_desc);
ac27a0ec 1249 for (i = 0; i < sbi->s_gdb_count; i++)
1d0c3924
TT		 1250 brelse(group_desc[i]);
1251 kvfree(group_desc);
7c990728
SJS		 1252 flex_groups = rcu_dereference(sbi->s_flex_groups);
1253 if (flex_groups) {
1254 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1255 kvfree(flex_groups[i]);
1256 kvfree(flex_groups);
1257 }
1d0c3924 1258 rcu_read_unlock();
57042651 1259 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 1260 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1261 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 1262 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
efc61345 1263 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
bbd55937 1264 percpu_free_rwsem(&sbi->s_writepages_rwsem);
ac27a0ec 1265#ifdef CONFIG_QUOTA
a2d4a646 1266 for (i = 0; i < EXT4_MAXQUOTAS; i++)
33458eab 1267 kfree(get_qf_name(sb, sbi, i));
ac27a0ec
DK		 1268#endif
1269
1270 /* Debugging code just in case the in-memory inode orphan list
1271 * isn't empty. The on-disk one can be non-empty if we've
1272 * detected an error and taken the fs readonly, but the
1273 * in-memory list had better be clean by this point. */
1274 if (!list_empty(&sbi->s_orphan))
1275 dump_orphan_list(sb, sbi);
837c23fb 1276 ASSERT(list_empty(&sbi->s_orphan));
ac27a0ec 1277
89d96a6f 1278 sync_blockdev(sb->s_bdev);
f98393a6 1279 invalidate_bdev(sb->s_bdev);
ee7ed3aa 1280 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
ac27a0ec
DK		 1281 /*
1282 * Invalidate the journal device's buffers. We don't want them
1283 * floating about in memory - the physical journal device may
1284 * hotswapped, and it breaks the `ro-after' testing code.
1285 */
ee7ed3aa
CX		 1286 sync_blockdev(sbi->s_journal_bdev);
1287 invalidate_bdev(sbi->s_journal_bdev);
617ba13b 1288 ext4_blkdev_remove(sbi);
ac27a0ec 1289 }
50c15df6
CX		 1290
1291 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1292 sbi->s_ea_inode_cache = NULL;
1293
1294 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1295 sbi->s_ea_block_cache = NULL;
1296
618f0031
PS		 1297 ext4_stop_mmpd(sbi);
1298
9060dd2c 1299 brelse(sbi->s_sbh);
ac27a0ec 1300 sb->s_fs_info = NULL;
3197ebdb
TT		 1301 /*
1302 * Now that we are completely done shutting down the
1303 * superblock, we need to actually destroy the kobject.
1304 */
3197ebdb
TT		 1305 kobject_put(&sbi->s_kobj);
1306 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 1307 if (sbi->s_chksum_driver)
1308 crypto_free_shash(sbi->s_chksum_driver);
705895b6 1309 kfree(sbi->s_blockgroup_lock);
8012b866 1310 fs_put_dax(sbi->s_daxdev, NULL);
ac4acb1f 1311 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5298d4bf 1312#if IS_ENABLED(CONFIG_UNICODE)
f8f4acb6 1313 utf8_unload(sb->s_encoding);
c83ad55e 1314#endif
ac27a0ec 1315 kfree(sbi);
ac27a0ec
DK		 1316}
1317
e18b890b 1318static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 1319
1320/*
1321 * Called inside transaction, so use GFP_NOFS
1322 */
617ba13b 1323static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 1324{
617ba13b 1325 struct ext4_inode_info *ei;
ac27a0ec 1326
fd60b288 1327 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 1328 if (!ei)
1329 return NULL;
0b8e58a1 1330
ee73f9a5 1331 inode_set_iversion(&ei->vfs_inode, 1);
202ee5df 1332 spin_lock_init(&ei->i_raw_lock);
c9de560d 1333 INIT_LIST_HEAD(&ei->i_prealloc_list);
27bc446e 1334 atomic_set(&ei->i_prealloc_active, 0);
c9de560d 1335 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 1336 ext4_es_init_tree(&ei->i_es_tree);
1337 rwlock_init(&ei->i_es_lock);
edaa53ca 1338 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 1339 ei->i_es_all_nr = 0;
edaa53ca 1340 ei->i_es_shk_nr = 0;
dd475925 1341 ei->i_es_shrink_lblk = 0;
d2a17637 1342 ei->i_reserved_data_blocks = 0;
d2a17637 1343 spin_lock_init(&(ei->i_block_reservation_lock));
1dc0aa46 1344 ext4_init_pending_tree(&ei->i_pending_tree);
a9e7f447
DM		 1345#ifdef CONFIG_QUOTA
1346 ei->i_reserved_quota = 0;
96c7e0d9 1347 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 1348#endif
8aefcd55 1349 ei->jinode = NULL;
2e8fa54e 1350 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 1351 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 1352 ei->i_sync_tid = 0;
1353 ei->i_datasync_tid = 0;
e27f41e1 1354 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 1355 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
aa75f4d3
HS		 1356 ext4_fc_init_inode(&ei->vfs_inode);
1357 mutex_init(&ei->i_fc_lock);
ac27a0ec
DK		 1358 return &ei->vfs_inode;
1359}
1360
7ff9c073
TT		 1361static int ext4_drop_inode(struct inode *inode)
1362{
1363 int drop = generic_drop_inode(inode);
1364
29b3692e
EB		 1365 if (!drop)
1366 drop = fscrypt_drop_inode(inode);
1367
7ff9c073
TT		 1368 trace_ext4_drop_inode(inode, drop);
1369 return drop;
1370}
1371
94053139 1372static void ext4_free_in_core_inode(struct inode *inode)
fa0d7e3d 1373{
2c58d548 1374 fscrypt_free_inode(inode);
aa75f4d3
HS		 1375 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1376 pr_warn("%s: inode %ld still in fc list",
1377 __func__, inode->i_ino);
1378 }
fa0d7e3d
NP		 1379 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1380}
1381
617ba13b 1382static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1383{
9f7dd93d 1384 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1385 ext4_msg(inode->i_sb, KERN_ERR,
1386 "Inode %lu (%p): orphan list check failed!",
1387 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1388 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1389 EXT4_I(inode), sizeof(struct ext4_inode_info),
1390 true);
1391 dump_stack();
1392 }
6fed8395
JX		 1393
1394 if (EXT4_I(inode)->i_reserved_data_blocks)
1395 ext4_msg(inode->i_sb, KERN_ERR,
1396 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1397 inode->i_ino, EXT4_I(inode),
1398 EXT4_I(inode)->i_reserved_data_blocks);
ac27a0ec
DK		 1399}
1400
51cc5068 1401static void init_once(void *foo)
ac27a0ec 1402{
c30365b9 1403 struct ext4_inode_info *ei = foo;
ac27a0ec 1404
a35afb83 1405 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1406 init_rwsem(&ei->xattr_sem);
0e855ac8 1407 init_rwsem(&ei->i_data_sem);
a35afb83 1408 inode_init_once(&ei->vfs_inode);
aa75f4d3 1409 ext4_fc_init_inode(&ei->vfs_inode);
ac27a0ec
DK		 1410}
1411
e67bc2b3 1412static int __init init_inodecache(void)
ac27a0ec 1413{
f8dd7c70
DW		 1414 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1415 sizeof(struct ext4_inode_info), 0,
1416 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1417 SLAB_ACCOUNT),
1418 offsetof(struct ext4_inode_info, i_data),
1419 sizeof_field(struct ext4_inode_info, i_data),
1420 init_once);
617ba13b 1421 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1422 return -ENOMEM;
1423 return 0;
1424}
1425
1426static void destroy_inodecache(void)
1427{
8c0a8537
KS		 1428 /*
1429 * Make sure all delayed rcu free inodes are flushed before we
1430 * destroy cache.
1431 */
1432 rcu_barrier();
617ba13b 1433 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1434}
1435
0930fcc1 1436void ext4_clear_inode(struct inode *inode)
ac27a0ec 1437{
aa75f4d3 1438 ext4_fc_del(inode);
0930fcc1 1439 invalidate_inode_buffers(inode);
dbd5768f 1440 clear_inode(inode);
27bc446e 1441 ext4_discard_preallocations(inode, 0);
51865fda 1442 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
f4c2d372 1443 dquot_drop(inode);
8aefcd55
TT		 1444 if (EXT4_I(inode)->jinode) {
1445 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1446 EXT4_I(inode)->jinode);
1447 jbd2_free_inode(EXT4_I(inode)->jinode);
1448 EXT4_I(inode)->jinode = NULL;
1449 }
3d204e24 1450 fscrypt_put_encryption_info(inode);
c93d8f88 1451 fsverity_cleanup_inode(inode);
ac27a0ec
DK		 1452}
1453
1b961ac0 1454static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1455 u64 ino, u32 generation)
ac27a0ec 1456{
ac27a0ec 1457 struct inode *inode;
ac27a0ec 1458
8a363970 1459 /*
ac27a0ec
DK		 1460 * Currently we don't know the generation for parent directory, so
1461 * a generation of 0 means "accept any"
1462 */
8a363970 1463 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1d1fe1ee
DH		 1464 if (IS_ERR(inode))
1465 return ERR_CAST(inode);
1466 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1467 iput(inode);
1468 return ERR_PTR(-ESTALE);
1469 }
1b961ac0
CH		 1470
1471 return inode;
1472}
1473
1474static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1475 int fh_len, int fh_type)
1b961ac0
CH		 1476{
1477 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1478 ext4_nfs_get_inode);
1479}
1480
1481static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1482 int fh_len, int fh_type)
1b961ac0
CH		 1483{
1484 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1485 ext4_nfs_get_inode);
ac27a0ec
DK		 1486}
1487
fde87268
TT		 1488static int ext4_nfs_commit_metadata(struct inode *inode)
1489{
1490 struct writeback_control wbc = {
1491 .sync_mode = WB_SYNC_ALL
1492 };
1493
1494 trace_ext4_nfs_commit_metadata(inode);
1495 return ext4_write_inode(inode, &wbc);
1496}
1497
ac27a0ec 1498#ifdef CONFIG_QUOTA
d6006186 1499static const char * const quotatypes[] = INITQFNAMES;
689c958c 1500#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1501
617ba13b
MC		 1502static int ext4_write_dquot(struct dquot *dquot);
1503static int ext4_acquire_dquot(struct dquot *dquot);
1504static int ext4_release_dquot(struct dquot *dquot);
1505static int ext4_mark_dquot_dirty(struct dquot *dquot);
1506static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1507static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1508 const struct path *path);
617ba13b 1509static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1510 size_t len, loff_t off);
617ba13b 1511static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1512 const char *data, size_t len, loff_t off);
7c319d32
AK		 1513static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1514 unsigned int flags);
ac27a0ec 1515
96c7e0d9
JK		 1516static struct dquot **ext4_get_dquots(struct inode *inode)
1517{
1518 return EXT4_I(inode)->i_dquot;
1519}
1520
61e225dc 1521static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1522 .get_reserved_space = ext4_get_reserved_space,
1523 .write_dquot = ext4_write_dquot,
1524 .acquire_dquot = ext4_acquire_dquot,
1525 .release_dquot = ext4_release_dquot,
1526 .mark_dirty = ext4_mark_dquot_dirty,
1527 .write_info = ext4_write_info,
1528 .alloc_dquot = dquot_alloc,
1529 .destroy_dquot = dquot_destroy,
1530 .get_projid = ext4_get_projid,
1531 .get_inode_usage = ext4_get_inode_usage,
ebc11f7b 1532 .get_next_id = dquot_get_next_id,
ac27a0ec
DK		 1533};
1534
0d54b217 1535static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1536 .quota_on = ext4_quota_on,
ca0e05e4 1537 .quota_off = ext4_quota_off,
287a8095 1538 .quota_sync = dquot_quota_sync,
0a240339 1539 .get_state = dquot_get_state,
287a8095
CH		 1540 .set_info = dquot_set_dqinfo,
1541 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1542 .set_dqblk = dquot_set_dqblk,
1543 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1544};
1545#endif
1546
ee9b6d61 1547static const struct super_operations ext4_sops = {
617ba13b 1548 .alloc_inode = ext4_alloc_inode,
94053139 1549 .free_inode = ext4_free_in_core_inode,
617ba13b 1550 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1551 .write_inode = ext4_write_inode,
1552 .dirty_inode = ext4_dirty_inode,
7ff9c073 1553 .drop_inode = ext4_drop_inode,
0930fcc1 1554 .evict_inode = ext4_evict_inode,
617ba13b 1555 .put_super = ext4_put_super,
617ba13b 1556 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1557 .freeze_fs = ext4_freeze,
1558 .unfreeze_fs = ext4_unfreeze,
617ba13b 1559 .statfs = ext4_statfs,
617ba13b 1560 .show_options = ext4_show_options,
ac27a0ec 1561#ifdef CONFIG_QUOTA
617ba13b
MC		 1562 .quota_read = ext4_quota_read,
1563 .quota_write = ext4_quota_write,
96c7e0d9 1564 .get_dquots = ext4_get_dquots,
ac27a0ec
DK		 1565#endif
1566};
1567
39655164 1568static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1569 .fh_to_dentry = ext4_fh_to_dentry,
1570 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1571 .get_parent = ext4_get_parent,
fde87268 1572 .commit_metadata = ext4_nfs_commit_metadata,
ac27a0ec
DK		 1573};
1574
1575enum {
1576 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
ba2e524d 1577 Opt_resgid, Opt_resuid, Opt_sb,
72578c33 1578 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1579 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1580 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1581 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1582 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1583 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1584 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
4f74d15f 1585 Opt_inlinecrypt,
ba2e524d 1586 Opt_usrjquota, Opt_grpjquota, Opt_quota,
ee4a3fcd 1587 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
9cb20f94
IW		 1588 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1589 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
327eaf73 1590 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
4437992b 1591 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1449032b 1592 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1593 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1594 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1595 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1596 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
21175ca4 1597 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
e5a185c2 1598 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
8016e29f 1599#ifdef CONFIG_EXT4_DEBUG
99c880de 1600 Opt_fc_debug_max_replay, Opt_fc_debug_force
8016e29f 1601#endif
ac27a0ec
DK		 1602};
1603
e5a185c2 1604static const struct constant_table ext4_param_errors[] = {
ba2e524d
LC		 1605 {"continue", EXT4_MOUNT_ERRORS_CONT},
1606 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1607 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
e5a185c2
LC		 1608 {}
1609};
1610
1611static const struct constant_table ext4_param_data[] = {
ba2e524d
LC		 1612 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1613 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1614 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
e5a185c2
LC		 1615 {}
1616};
1617
1618static const struct constant_table ext4_param_data_err[] = {
1619 {"abort", Opt_data_err_abort},
1620 {"ignore", Opt_data_err_ignore},
1621 {}
1622};
1623
1624static const struct constant_table ext4_param_jqfmt[] = {
ba2e524d
LC		 1625 {"vfsold", QFMT_VFS_OLD},
1626 {"vfsv0", QFMT_VFS_V0},
1627 {"vfsv1", QFMT_VFS_V1},
e5a185c2
LC		 1628 {}
1629};
1630
1631static const struct constant_table ext4_param_dax[] = {
1632 {"always", Opt_dax_always},
1633 {"inode", Opt_dax_inode},
1634 {"never", Opt_dax_never},
1635 {}
1636};
1637
1638/* String parameter that allows empty argument */
1639#define fsparam_string_empty(NAME, OPT) \
1640 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1641
1642/*
1643 * Mount option specification
1644 * We don't use fsparam_flag_no because of the way we set the
1645 * options and the way we show them in _ext4_show_options(). To
1646 * keep the changes to a minimum, let's keep the negative options
1647 * separate for now.
1648 */
1649static const struct fs_parameter_spec ext4_param_specs[] = {
1650 fsparam_flag ("bsddf", Opt_bsd_df),
1651 fsparam_flag ("minixdf", Opt_minix_df),
1652 fsparam_flag ("grpid", Opt_grpid),
1653 fsparam_flag ("bsdgroups", Opt_grpid),
1654 fsparam_flag ("nogrpid", Opt_nogrpid),
1655 fsparam_flag ("sysvgroups", Opt_nogrpid),
1656 fsparam_u32 ("resgid", Opt_resgid),
1657 fsparam_u32 ("resuid", Opt_resuid),
1658 fsparam_u32 ("sb", Opt_sb),
1659 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1660 fsparam_flag ("nouid32", Opt_nouid32),
1661 fsparam_flag ("debug", Opt_debug),
1662 fsparam_flag ("oldalloc", Opt_removed),
1663 fsparam_flag ("orlov", Opt_removed),
1664 fsparam_flag ("user_xattr", Opt_user_xattr),
1665 fsparam_flag ("nouser_xattr", Opt_nouser_xattr),
1666 fsparam_flag ("acl", Opt_acl),
1667 fsparam_flag ("noacl", Opt_noacl),
1668 fsparam_flag ("norecovery", Opt_noload),
1669 fsparam_flag ("noload", Opt_noload),
1670 fsparam_flag ("bh", Opt_removed),
1671 fsparam_flag ("nobh", Opt_removed),
1672 fsparam_u32 ("commit", Opt_commit),
1673 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1674 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1675 fsparam_u32 ("journal_dev", Opt_journal_dev),
1676 fsparam_bdev ("journal_path", Opt_journal_path),
1677 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1678 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1679 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1680 fsparam_flag ("abort", Opt_abort),
1681 fsparam_enum ("data", Opt_data, ext4_param_data),
1682 fsparam_enum ("data_err", Opt_data_err,
1683 ext4_param_data_err),
1684 fsparam_string_empty
1685 ("usrjquota", Opt_usrjquota),
1686 fsparam_string_empty
1687 ("grpjquota", Opt_grpjquota),
1688 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1689 fsparam_flag ("grpquota", Opt_grpquota),
1690 fsparam_flag ("quota", Opt_quota),
1691 fsparam_flag ("noquota", Opt_noquota),
1692 fsparam_flag ("usrquota", Opt_usrquota),
1693 fsparam_flag ("prjquota", Opt_prjquota),
1694 fsparam_flag ("barrier", Opt_barrier),
1695 fsparam_u32 ("barrier", Opt_barrier),
1696 fsparam_flag ("nobarrier", Opt_nobarrier),
1697 fsparam_flag ("i_version", Opt_i_version),
1698 fsparam_flag ("dax", Opt_dax),
1699 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1700 fsparam_u32 ("stripe", Opt_stripe),
1701 fsparam_flag ("delalloc", Opt_delalloc),
1702 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1703 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1704 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
e5a185c2
LC		 1705 fsparam_u32 ("debug_want_extra_isize",
1706 Opt_debug_want_extra_isize),
1707 fsparam_flag ("mblk_io_submit", Opt_removed),
1708 fsparam_flag ("nomblk_io_submit", Opt_removed),
1709 fsparam_flag ("block_validity", Opt_block_validity),
1710 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1711 fsparam_u32 ("inode_readahead_blks",
1712 Opt_inode_readahead_blks),
1713 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1714 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1715 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1716 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1717 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1718 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1719 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1720 fsparam_flag ("discard", Opt_discard),
1721 fsparam_flag ("nodiscard", Opt_nodiscard),
1722 fsparam_u32 ("init_itable", Opt_init_itable),
1723 fsparam_flag ("init_itable", Opt_init_itable),
1724 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1725#ifdef CONFIG_EXT4_DEBUG
1726 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1727 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1728#endif
1729 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1730 fsparam_flag ("test_dummy_encryption",
1731 Opt_test_dummy_encryption),
1732 fsparam_string ("test_dummy_encryption",
1733 Opt_test_dummy_encryption),
1734 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1735 fsparam_flag ("nombcache", Opt_nombcache),
1736 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1737 fsparam_flag ("prefetch_block_bitmaps",
1738 Opt_removed),
1739 fsparam_flag ("no_prefetch_block_bitmaps",
1740 Opt_no_prefetch_block_bitmaps),
1741 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1742 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1743 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1744 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1745 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1746 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1747 {}
1748};
1749
b3881f74 1750#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
196e402a 1751
d6006186
EB		 1752static const char deprecated_msg[] =
1753 "Mount option \"%s\" will be removed by %s\n"
437ca0fd 1754 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1755
26092bf5
TT		 1756#define MOPT_SET 0x0001
1757#define MOPT_CLEAR 0x0002
1758#define MOPT_NOSUPPORT 0x0004
1759#define MOPT_EXPLICIT 0x0008
ac27a0ec 1760#ifdef CONFIG_QUOTA
26092bf5 1761#define MOPT_Q 0
ba2e524d 1762#define MOPT_QFMT 0x0010
26092bf5
TT		 1763#else
1764#define MOPT_Q MOPT_NOSUPPORT
1765#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1766#endif
ba2e524d
LC		 1767#define MOPT_NO_EXT2 0x0020
1768#define MOPT_NO_EXT3 0x0040
8dc0aa8c 1769#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ba2e524d
LC		 1770#define MOPT_SKIP 0x0080
1771#define MOPT_2 0x0100
26092bf5
TT		 1772
1773static const struct mount_opts {
1774 int token;
1775 int mount_opt;
1776 int flags;
1777} ext4_mount_opts[] = {
1778 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1779 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1780 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1781 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1782 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1783 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1784 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1785 MOPT_EXT4_ONLY | MOPT_SET},
1786 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1787 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1788 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1789 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1790 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1791 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1792 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1793 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1794 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1795 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
cb8435dc 1796 {Opt_commit, 0, MOPT_NO_EXT2},
c6d3d56d
DW		 1797 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1798 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1799 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1800 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1801 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1802 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1803 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1804 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
ba2e524d 1805 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
26092bf5
TT		 1806 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1807 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1808 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1809 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1810 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
ba2e524d
LC		 1811 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1812 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1813 {Opt_journal_path, 0, MOPT_NO_EXT2},
1814 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1815 {Opt_data, 0, MOPT_NO_EXT2},
26092bf5
TT		 1816 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1817 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1818#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1819 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1820 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1821#else
26092bf5
TT		 1822 {Opt_acl, 0, MOPT_NOSUPPORT},
1823 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1824#endif
26092bf5
TT		 1825 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1826 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1827 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1828 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1829 MOPT_SET | MOPT_Q},
1830 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1831 MOPT_SET | MOPT_Q},
49da9392
JK		 1832 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1833 MOPT_SET | MOPT_Q},
26092bf5 1834 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1835 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1836 MOPT_CLEAR | MOPT_Q},
ba2e524d
LC		 1837 {Opt_usrjquota, 0, MOPT_Q},
1838 {Opt_grpjquota, 0, MOPT_Q},
1839 {Opt_jqfmt, 0, MOPT_QFMT},
cdb7ee4c 1840 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
21175ca4 1841 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
3d392b26 1842 MOPT_SET},
99c880de 1843#ifdef CONFIG_EXT4_DEBUG
0f0672ff
HS		 1844 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1845 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
8016e29f 1846#endif
26092bf5
TT		 1847 {Opt_err, 0, 0}
1848};
1849
5298d4bf 1850#if IS_ENABLED(CONFIG_UNICODE)
c83ad55e
GKB		 1851static const struct ext4_sb_encodings {
1852 __u16 magic;
1853 char *name;
49bd03cc 1854 unsigned int version;
c83ad55e 1855} ext4_sb_encoding_map[] = {
49bd03cc 1856 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
c83ad55e
GKB		 1857};
1858
aa8bf298
CH		 1859static const struct ext4_sb_encodings *
1860ext4_sb_read_encoding(const struct ext4_super_block *es)
c83ad55e
GKB		 1861{
1862 __u16 magic = le16_to_cpu(es->s_encoding);
1863 int i;
1864
1865 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1866 if (magic == ext4_sb_encoding_map[i].magic)
aa8bf298 1867 return &ext4_sb_encoding_map[i];
c83ad55e 1868
aa8bf298 1869 return NULL;
c83ad55e
GKB		 1870}
1871#endif
1872
6e47a3cc
LC		 1873#define EXT4_SPEC_JQUOTA (1 << 0)
1874#define EXT4_SPEC_JQFMT (1 << 1)
1875#define EXT4_SPEC_DATAJ (1 << 2)
1876#define EXT4_SPEC_SB_BLOCK (1 << 3)
1877#define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1878#define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
6e47a3cc
LC		 1879#define EXT4_SPEC_s_want_extra_isize (1 << 7)
1880#define EXT4_SPEC_s_max_batch_time (1 << 8)
1881#define EXT4_SPEC_s_min_batch_time (1 << 9)
1882#define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1883#define EXT4_SPEC_s_li_wait_mult (1 << 11)
1884#define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1885#define EXT4_SPEC_s_stripe (1 << 13)
1886#define EXT4_SPEC_s_resuid (1 << 14)
1887#define EXT4_SPEC_s_resgid (1 << 15)
1888#define EXT4_SPEC_s_commit_interval (1 << 16)
1889#define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
7edfd85b 1890#define EXT4_SPEC_s_sb_block (1 << 18)
27b38686 1891#define EXT4_SPEC_mb_optimize_scan (1 << 19)
6e47a3cc 1892
461c3af0 1893struct ext4_fs_context {
e6e268cb 1894 char *s_qf_names[EXT4_MAXQUOTAS];
85456054 1895 struct fscrypt_dummy_policy dummy_enc_policy;
e6e268cb 1896 int s_jquota_fmt; /* Format of quota to use */
6e47a3cc
LC		 1897#ifdef CONFIG_EXT4_DEBUG
1898 int s_fc_debug_max_replay;
1899#endif
e6e268cb 1900 unsigned short qname_spec;
6e47a3cc
LC		 1901 unsigned long vals_s_flags; /* Bits to set in s_flags */
1902 unsigned long mask_s_flags; /* Bits changed in s_flags */
e6e268cb 1903 unsigned long journal_devnum;
6e47a3cc
LC		 1904 unsigned long s_commit_interval;
1905 unsigned long s_stripe;
1906 unsigned int s_inode_readahead_blks;
1907 unsigned int s_want_extra_isize;
1908 unsigned int s_li_wait_mult;
1909 unsigned int s_max_dir_size_kb;
e6e268cb 1910 unsigned int journal_ioprio;
6e47a3cc
LC		 1911 unsigned int vals_s_mount_opt;
1912 unsigned int mask_s_mount_opt;
1913 unsigned int vals_s_mount_opt2;
1914 unsigned int mask_s_mount_opt2;
e3952fcc
LC		 1915 unsigned long vals_s_mount_flags;
1916 unsigned long mask_s_mount_flags;
b6bd2435 1917 unsigned int opt_flags; /* MOPT flags */
6e47a3cc
LC		 1918 unsigned int spec;
1919 u32 s_max_batch_time;
1920 u32 s_min_batch_time;
1921 kuid_t s_resuid;
1922 kgid_t s_resgid;
7edfd85b 1923 ext4_fsblk_t s_sb_block;
b237e304
HS		 1924};
1925
cebe85d5
LC		 1926static void ext4_fc_free(struct fs_context *fc)
1927{
1928 struct ext4_fs_context *ctx = fc->fs_private;
1929 int i;
1930
1931 if (!ctx)
1932 return;
1933
1934 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1935 kfree(ctx->s_qf_names[i]);
1936
85456054 1937 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
cebe85d5
LC		 1938 kfree(ctx);
1939}
1940
1941int ext4_init_fs_context(struct fs_context *fc)
1942{
da9e4802 1943 struct ext4_fs_context *ctx;
cebe85d5
LC		 1944
1945 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
1946 if (!ctx)
1947 return -ENOMEM;
1948
1949 fc->fs_private = ctx;
1950 fc->ops = &ext4_context_ops;
1951
1952 return 0;
1953}
1954
e6e268cb
LC		 1955#ifdef CONFIG_QUOTA
1956/*
1957 * Note the name of the specified quota file.
1958 */
1959static int note_qf_name(struct fs_context *fc, int qtype,
1960 struct fs_parameter *param)
1961{
1962 struct ext4_fs_context *ctx = fc->fs_private;
1963 char *qname;
1964
1965 if (param->size < 1) {
1966 ext4_msg(NULL, KERN_ERR, "Missing quota name");
1967 return -EINVAL;
1968 }
1969 if (strchr(param->string, '/')) {
1970 ext4_msg(NULL, KERN_ERR,
1971 "quotafile must be on filesystem root");
1972 return -EINVAL;
1973 }
1974 if (ctx->s_qf_names[qtype]) {
1975 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
1976 ext4_msg(NULL, KERN_ERR,
1977 "%s quota file already specified",
1978 QTYPE2NAME(qtype));
1979 return -EINVAL;
1980 }
1981 return 0;
1982 }
1983
1984 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
1985 if (!qname) {
1986 ext4_msg(NULL, KERN_ERR,
1987 "Not enough memory for storing quotafile name");
1988 return -ENOMEM;
1989 }
1990 ctx->s_qf_names[qtype] = qname;
1991 ctx->qname_spec |= 1 << qtype;
6e47a3cc 1992 ctx->spec |= EXT4_SPEC_JQUOTA;
e6e268cb
LC		 1993 return 0;
1994}
1995
1996/*
1997 * Clear the name of the specified quota file.
1998 */
1999static int unnote_qf_name(struct fs_context *fc, int qtype)
2000{
2001 struct ext4_fs_context *ctx = fc->fs_private;
2002
2003 if (ctx->s_qf_names[qtype])
2004 kfree(ctx->s_qf_names[qtype]);
2005
2006 ctx->s_qf_names[qtype] = NULL;
2007 ctx->qname_spec |= 1 << qtype;
6e47a3cc 2008 ctx->spec |= EXT4_SPEC_JQUOTA;
e6e268cb
LC		 2009 return 0;
2010}
2011#endif
2012
85456054
EB		 2013static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2014 struct ext4_fs_context *ctx)
2015{
2016 int err;
2017
2018 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2019 ext4_msg(NULL, KERN_WARNING,
2020 "test_dummy_encryption option not supported");
2021 return -EINVAL;
2022 }
2023 err = fscrypt_parse_test_dummy_encryption(param,
2024 &ctx->dummy_enc_policy);
2025 if (err == -EINVAL) {
2026 ext4_msg(NULL, KERN_WARNING,
2027 "Value of option \"%s\" is unrecognized", param->key);
2028 } else if (err == -EEXIST) {
2029 ext4_msg(NULL, KERN_WARNING,
2030 "Conflicting test_dummy_encryption options");
2031 return -EINVAL;
2032 }
2033 return err;
2034}
2035
6e47a3cc 2036#define EXT4_SET_CTX(name) \
4c246728
LC		 2037static inline void ctx_set_##name(struct ext4_fs_context *ctx, \
2038 unsigned long flag) \
6e47a3cc
LC		 2039{ \
2040 ctx->mask_s_##name |= flag; \
2041 ctx->vals_s_##name |= flag; \
e3952fcc
LC		 2042}
2043
2044#define EXT4_CLEAR_CTX(name) \
4c246728
LC		 2045static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \
2046 unsigned long flag) \
6e47a3cc
LC		 2047{ \
2048 ctx->mask_s_##name |= flag; \
2049 ctx->vals_s_##name &= ~flag; \
e3952fcc
LC		 2050}
2051
2052#define EXT4_TEST_CTX(name) \
4c246728
LC		 2053static inline unsigned long \
2054ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
6e47a3cc 2055{ \
4c246728 2056 return (ctx->vals_s_##name & flag); \
e3952fcc 2057}
6e47a3cc 2058
e3952fcc 2059EXT4_SET_CTX(flags); /* set only */
6e47a3cc 2060EXT4_SET_CTX(mount_opt);
e3952fcc
LC		 2061EXT4_CLEAR_CTX(mount_opt);
2062EXT4_TEST_CTX(mount_opt);
6e47a3cc 2063EXT4_SET_CTX(mount_opt2);
e3952fcc
LC		 2064EXT4_CLEAR_CTX(mount_opt2);
2065EXT4_TEST_CTX(mount_opt2);
2066
2067static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
2068{
2069 set_bit(bit, &ctx->mask_s_mount_flags);
2070 set_bit(bit, &ctx->vals_s_mount_flags);
2071}
6e47a3cc 2072
02f960f8 2073static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
26092bf5 2074{
461c3af0 2075 struct ext4_fs_context *ctx = fc->fs_private;
461c3af0 2076 struct fs_parse_result result;
26092bf5 2077 const struct mount_opts *m;
461c3af0 2078 int is_remount;
08cefc7a
EB		 2079 kuid_t uid;
2080 kgid_t gid;
461c3af0
LC		 2081 int token;
2082
2083 token = fs_parse(fc, ext4_param_specs, param, &result);
2084 if (token < 0)
2085 return token;
2086 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
26092bf5 2087
ba2e524d
LC		 2088 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2089 if (token == m->token)
2090 break;
2091
2092 ctx->opt_flags |= m->flags;
2093
2094 if (m->flags & MOPT_EXPLICIT) {
2095 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2096 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2097 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2098 ctx_set_mount_opt2(ctx,
2099 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2100 } else
2101 return -EINVAL;
2102 }
2103
2104 if (m->flags & MOPT_NOSUPPORT) {
2105 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2106 param->key);
2107 return 0;
2108 }
2109
2110 switch (token) {
57f73c2c 2111#ifdef CONFIG_QUOTA
ba2e524d 2112 case Opt_usrjquota:
461c3af0 2113 if (!*param->string)
e6e268cb 2114 return unnote_qf_name(fc, USRQUOTA);
461c3af0 2115 else
e6e268cb 2116 return note_qf_name(fc, USRQUOTA, param);
ba2e524d 2117 case Opt_grpjquota:
461c3af0 2118 if (!*param->string)
e6e268cb 2119 return unnote_qf_name(fc, GRPQUOTA);
461c3af0 2120 else
e6e268cb 2121 return note_qf_name(fc, GRPQUOTA, param);
57f73c2c 2122#endif
f7048605
TT		 2123 case Opt_noacl:
2124 case Opt_nouser_xattr:
da812f61 2125 ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "3.5");
f7048605 2126 break;
26092bf5 2127 case Opt_sb:
7edfd85b
LC		 2128 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2129 ext4_msg(NULL, KERN_WARNING,
2130 "Ignoring %s option on remount", param->key);
2131 } else {
2132 ctx->s_sb_block = result.uint_32;
2133 ctx->spec |= EXT4_SPEC_s_sb_block;
2134 }
02f960f8 2135 return 0;
26092bf5 2136 case Opt_removed:
da812f61 2137 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
461c3af0 2138 param->key);
02f960f8 2139 return 0;
26092bf5 2140 case Opt_abort:
e3952fcc 2141 ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
02f960f8 2142 return 0;
26092bf5 2143 case Opt_i_version:
960e0ab6
LC		 2144 ext4_msg(NULL, KERN_WARNING, deprecated_msg, param->key, "5.20");
2145 ext4_msg(NULL, KERN_WARNING, "Use iversion instead\n");
6e47a3cc 2146 ctx_set_flags(ctx, SB_I_VERSION);
02f960f8 2147 return 0;
4f74d15f
EB		 2148 case Opt_inlinecrypt:
2149#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
6e47a3cc 2150 ctx_set_flags(ctx, SB_INLINECRYPT);
4f74d15f 2151#else
da812f61 2152 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
4f74d15f 2153#endif
02f960f8 2154 return 0;
461c3af0 2155 case Opt_errors:
6e47a3cc 2156 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
ba2e524d
LC		 2157 ctx_set_mount_opt(ctx, result.uint_32);
2158 return 0;
2159#ifdef CONFIG_QUOTA
2160 case Opt_jqfmt:
2161 ctx->s_jquota_fmt = result.uint_32;
2162 ctx->spec |= EXT4_SPEC_JQFMT;
2163 return 0;
2164#endif
2165 case Opt_data:
2166 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2167 ctx_set_mount_opt(ctx, result.uint_32);
2168 ctx->spec |= EXT4_SPEC_DATAJ;
2169 return 0;
2170 case Opt_commit:
461c3af0 2171 if (result.uint_32 == 0)
6e47a3cc 2172 ctx->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE;
461c3af0 2173 else if (result.uint_32 > INT_MAX / HZ) {
da812f61 2174 ext4_msg(NULL, KERN_ERR,
9ba55543 2175 "Invalid commit interval %d, "
2176 "must be smaller than %d",
461c3af0 2177 result.uint_32, INT_MAX / HZ);
da812f61 2178 return -EINVAL;
9ba55543 2179 }
6e47a3cc
LC		 2180 ctx->s_commit_interval = HZ * result.uint_32;
2181 ctx->spec |= EXT4_SPEC_s_commit_interval;
ba2e524d
LC		 2182 return 0;
2183 case Opt_debug_want_extra_isize:
6e47a3cc 2184 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
da812f61 2185 ext4_msg(NULL, KERN_ERR,
461c3af0 2186 "Invalid want_extra_isize %d", result.uint_32);
da812f61 2187 return -EINVAL;
9803387c 2188 }
6e47a3cc
LC		 2189 ctx->s_want_extra_isize = result.uint_32;
2190 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
ba2e524d
LC		 2191 return 0;
2192 case Opt_max_batch_time:
6e47a3cc
LC		 2193 ctx->s_max_batch_time = result.uint_32;
2194 ctx->spec |= EXT4_SPEC_s_max_batch_time;
ba2e524d
LC		 2195 return 0;
2196 case Opt_min_batch_time:
6e47a3cc
LC		 2197 ctx->s_min_batch_time = result.uint_32;
2198 ctx->spec |= EXT4_SPEC_s_min_batch_time;
ba2e524d
LC		 2199 return 0;
2200 case Opt_inode_readahead_blks:
461c3af0
LC		 2201 if (result.uint_32 &&
2202 (result.uint_32 > (1 << 30) ||
2203 !is_power_of_2(result.uint_32))) {
da812f61 2204 ext4_msg(NULL, KERN_ERR,
e33e60ea
JK		 2205 "EXT4-fs: inode_readahead_blks must be "
2206 "0 or a power of 2 smaller than 2^31");
da812f61 2207 return -EINVAL;
5f3633e3 2208 }
6e47a3cc
LC		 2209 ctx->s_inode_readahead_blks = result.uint_32;
2210 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
ba2e524d
LC		 2211 return 0;
2212 case Opt_init_itable:
6e47a3cc
LC		 2213 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2214 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
461c3af0 2215 if (param->type == fs_value_is_string)
6e47a3cc
LC		 2216 ctx->s_li_wait_mult = result.uint_32;
2217 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
ba2e524d
LC		 2218 return 0;
2219 case Opt_max_dir_size_kb:
6e47a3cc
LC		 2220 ctx->s_max_dir_size_kb = result.uint_32;
2221 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
ba2e524d 2222 return 0;
8016e29f 2223#ifdef CONFIG_EXT4_DEBUG
ba2e524d 2224 case Opt_fc_debug_max_replay:
6e47a3cc
LC		 2225 ctx->s_fc_debug_max_replay = result.uint_32;
2226 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
ba2e524d 2227 return 0;
8016e29f 2228#endif
ba2e524d 2229 case Opt_stripe:
6e47a3cc
LC		 2230 ctx->s_stripe = result.uint_32;
2231 ctx->spec |= EXT4_SPEC_s_stripe;
ba2e524d
LC		 2232 return 0;
2233 case Opt_resuid:
461c3af0 2234 uid = make_kuid(current_user_ns(), result.uint_32);
5f3633e3 2235 if (!uid_valid(uid)) {
da812f61 2236 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
461c3af0 2237 result.uint_32);
da812f61 2238 return -EINVAL;
26092bf5 2239 }
6e47a3cc
LC		 2240 ctx->s_resuid = uid;
2241 ctx->spec |= EXT4_SPEC_s_resuid;
ba2e524d
LC		 2242 return 0;
2243 case Opt_resgid:
461c3af0 2244 gid = make_kgid(current_user_ns(), result.uint_32);
5f3633e3 2245 if (!gid_valid(gid)) {
da812f61 2246 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
461c3af0 2247 result.uint_32);
da812f61 2248 return -EINVAL;
5f3633e3 2249 }
6e47a3cc
LC		 2250 ctx->s_resgid = gid;
2251 ctx->spec |= EXT4_SPEC_s_resgid;
ba2e524d
LC		 2252 return 0;
2253 case Opt_journal_dev:
5f3633e3 2254 if (is_remount) {
da812f61 2255 ext4_msg(NULL, KERN_ERR,
5f3633e3 2256 "Cannot specify journal on remount");
da812f61 2257 return -EINVAL;
5f3633e3 2258 }
461c3af0 2259 ctx->journal_devnum = result.uint_32;
6e47a3cc 2260 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
ba2e524d
LC		 2261 return 0;
2262 case Opt_journal_path:
2263 {
ad4eec61
ES		 2264 struct inode *journal_inode;
2265 struct path path;
2266 int error;
2267
2268 if (is_remount) {
da812f61 2269 ext4_msg(NULL, KERN_ERR,
ad4eec61 2270 "Cannot specify journal on remount");
da812f61 2271 return -EINVAL;
ad4eec61 2272 }
ad4eec61 2273
461c3af0 2274 error = fs_lookup_param(fc, param, 1, &path);
ad4eec61 2275 if (error) {
da812f61 2276 ext4_msg(NULL, KERN_ERR, "error: could not find "
461c3af0 2277 "journal device path");
da812f61 2278 return -EINVAL;
ad4eec61
ES		 2279 }
2280
2b0143b5 2281 journal_inode = d_inode(path.dentry);
461c3af0 2282 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
6e47a3cc 2283 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
ad4eec61 2284 path_put(&path);
ba2e524d
LC		 2285 return 0;
2286 }
2287 case Opt_journal_ioprio:
461c3af0 2288 if (result.uint_32 > 7) {
da812f61 2289 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
5f3633e3 2290 " (must be 0-7)");
da812f61 2291 return -EINVAL;
5f3633e3 2292 }
461c3af0
LC		 2293 ctx->journal_ioprio =
2294 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
6e47a3cc 2295 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
ba2e524d
LC		 2296 return 0;
2297 case Opt_test_dummy_encryption:
85456054 2298 return ext4_parse_test_dummy_encryption(param, ctx);
ba2e524d
LC		 2299 case Opt_dax:
2300 case Opt_dax_type:
ef83b6e8 2301#ifdef CONFIG_FS_DAX
ba2e524d
LC		 2302 {
2303 int type = (token == Opt_dax) ?
2304 Opt_dax : result.uint_32;
2305
2306 switch (type) {
9cb20f94
IW		 2307 case Opt_dax:
2308 case Opt_dax_always:
ba2e524d 2309 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
6e47a3cc 2310 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
9cb20f94
IW		 2311 break;
2312 case Opt_dax_never:
ba2e524d 2313 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
6e47a3cc 2314 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
9cb20f94
IW		 2315 break;
2316 case Opt_dax_inode:
6e47a3cc
LC		 2317 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2318 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
9cb20f94 2319 /* Strictly for printing options */
ba2e524d 2320 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
9cb20f94
IW		 2321 break;
2322 }
ba2e524d
LC		 2323 return 0;
2324 }
ef83b6e8 2325#else
da812f61 2326 ext4_msg(NULL, KERN_INFO, "dax option not supported");
da812f61 2327 return -EINVAL;
5f3633e3 2328#endif
ba2e524d
LC		 2329 case Opt_data_err:
2330 if (result.uint_32 == Opt_data_err_abort)
2331 ctx_set_mount_opt(ctx, m->mount_opt);
2332 else if (result.uint_32 == Opt_data_err_ignore)
2333 ctx_clear_mount_opt(ctx, m->mount_opt);
2334 return 0;
2335 case Opt_mb_optimize_scan:
27b38686
OM		 2336 if (result.int_32 == 1) {
2337 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2338 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2339 } else if (result.int_32 == 0) {
2340 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2341 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2342 } else {
da812f61 2343 ext4_msg(NULL, KERN_WARNING,
196e402a 2344 "mb_optimize_scan should be set to 0 or 1.");
da812f61 2345 return -EINVAL;
196e402a 2346 }
ba2e524d
LC		 2347 return 0;
2348 }
2349
2350 /*
2351 * At this point we should only be getting options requiring MOPT_SET,
2352 * or MOPT_CLEAR. Anything else is a bug
2353 */
2354 if (m->token == Opt_err) {
2355 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2356 param->key);
2357 WARN_ON(1);
2358 return -EINVAL;
2359 }
2360
2361 else {
461c3af0
LC		 2362 unsigned int set = 0;
2363
2364 if ((param->type == fs_value_is_flag) ||
2365 result.uint_32 > 0)
2366 set = 1;
2367
5f3633e3 2368 if (m->flags & MOPT_CLEAR)
461c3af0 2369 set = !set;
5f3633e3 2370 else if (unlikely(!(m->flags & MOPT_SET))) {
da812f61 2371 ext4_msg(NULL, KERN_WARNING,
461c3af0
LC		 2372 "buggy handling of option %s",
2373 param->key);
5f3633e3 2374 WARN_ON(1);
da812f61 2375 return -EINVAL;
5f3633e3 2376 }
995a3ed6 2377 if (m->flags & MOPT_2) {
461c3af0 2378 if (set != 0)
6e47a3cc 2379 ctx_set_mount_opt2(ctx, m->mount_opt);
995a3ed6 2380 else
6e47a3cc 2381 ctx_clear_mount_opt2(ctx, m->mount_opt);
995a3ed6 2382 } else {
461c3af0 2383 if (set != 0)
6e47a3cc 2384 ctx_set_mount_opt(ctx, m->mount_opt);
995a3ed6 2385 else
6e47a3cc 2386 ctx_clear_mount_opt(ctx, m->mount_opt);
995a3ed6 2387 }
26092bf5 2388 }
ba2e524d 2389
02f960f8 2390 return 0;
26092bf5
TT		 2391}
2392
7edfd85b 2393static int parse_options(struct fs_context *fc, char *options)
26092bf5 2394{
461c3af0 2395 struct fs_parameter param;
461c3af0
LC		 2396 int ret;
2397 char *key;
26092bf5
TT		 2398
2399 if (!options)
7edfd85b 2400 return 0;
461c3af0
LC		 2401
2402 while ((key = strsep(&options, ",")) != NULL) {
2403 if (*key) {
2404 size_t v_len = 0;
2405 char *value = strchr(key, '=');
2406
2407 param.type = fs_value_is_flag;
2408 param.string = NULL;
2409
2410 if (value) {
2411 if (value == key)
2412 continue;
2413
2414 *value++ = 0;
2415 v_len = strlen(value);
2416 param.string = kmemdup_nul(value, v_len,
2417 GFP_KERNEL);
2418 if (!param.string)
7edfd85b 2419 return -ENOMEM;
461c3af0
LC		 2420 param.type = fs_value_is_string;
2421 }
2422
2423 param.key = key;
2424 param.size = v_len;
2425
02f960f8 2426 ret = ext4_parse_param(fc, &param);
461c3af0
LC		 2427 if (param.string)
2428 kfree(param.string);
2429 if (ret < 0)
7edfd85b 2430 return ret;
461c3af0 2431 }
ac27a0ec 2432 }
461c3af0 2433
7edfd85b 2434 ret = ext4_validate_options(fc);
da812f61 2435 if (ret < 0)
7edfd85b 2436 return ret;
da812f61 2437
7edfd85b
LC		 2438 return 0;
2439}
2440
2441static int parse_apply_sb_mount_options(struct super_block *sb,
2442 struct ext4_fs_context *m_ctx)
2443{
2444 struct ext4_sb_info *sbi = EXT4_SB(sb);
2445 char *s_mount_opts = NULL;
2446 struct ext4_fs_context *s_ctx = NULL;
2447 struct fs_context *fc = NULL;
2448 int ret = -ENOMEM;
2449
2450 if (!sbi->s_es->s_mount_opts[0])
e6e268cb
LC		 2451 return 0;
2452
7edfd85b
LC		 2453 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2454 sizeof(sbi->s_es->s_mount_opts),
2455 GFP_KERNEL);
2456 if (!s_mount_opts)
2457 return ret;
2458
2459 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2460 if (!fc)
2461 goto out_free;
2462
2463 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2464 if (!s_ctx)
2465 goto out_free;
2466
2467 fc->fs_private = s_ctx;
2468 fc->s_fs_info = sbi;
2469
2470 ret = parse_options(fc, s_mount_opts);
6e47a3cc 2471 if (ret < 0)
7edfd85b 2472 goto parse_failed;
e6e268cb 2473
7edfd85b
LC		 2474 ret = ext4_check_opt_consistency(fc, sb);
2475 if (ret < 0) {
2476parse_failed:
2477 ext4_msg(sb, KERN_WARNING,
2478 "failed to parse options in superblock: %s",
2479 s_mount_opts);
2480 ret = 0;
2481 goto out_free;
2482 }
2483
2484 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2485 m_ctx->journal_devnum = s_ctx->journal_devnum;
2486 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2487 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2488
85456054
EB		 2489 ext4_apply_options(fc, sb);
2490 ret = 0;
7edfd85b
LC		 2491
2492out_free:
c069db76
EB		 2493 if (fc) {
2494 ext4_fc_free(fc);
2495 kfree(fc);
2496 }
7edfd85b
LC		 2497 kfree(s_mount_opts);
2498 return ret;
4c94bff9
LC		 2499}
2500
e6e268cb
LC		 2501static void ext4_apply_quota_options(struct fs_context *fc,
2502 struct super_block *sb)
2503{
2504#ifdef CONFIG_QUOTA
6e47a3cc 2505 bool quota_feature = ext4_has_feature_quota(sb);
e6e268cb
LC		 2506 struct ext4_fs_context *ctx = fc->fs_private;
2507 struct ext4_sb_info *sbi = EXT4_SB(sb);
2508 char *qname;
2509 int i;
2510
6e47a3cc
LC		 2511 if (quota_feature)
2512 return;
2513
2514 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2515 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2516 if (!(ctx->qname_spec & (1 << i)))
2517 continue;
2518
2519 qname = ctx->s_qf_names[i]; /* May be NULL */
4c1bd5a9
LC		 2520 if (qname)
2521 set_opt(sb, QUOTA);
6e47a3cc 2522 ctx->s_qf_names[i] = NULL;
13b215a9
LC		 2523 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2524 lockdep_is_held(&sb->s_umount));
2525 if (qname)
2526 kfree_rcu(qname);
6e47a3cc 2527 }
e6e268cb 2528 }
6e47a3cc
LC		 2529
2530 if (ctx->spec & EXT4_SPEC_JQFMT)
2531 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
e6e268cb
LC		 2532#endif
2533}
2534
2535/*
2536 * Check quota settings consistency.
2537 */
2538static int ext4_check_quota_consistency(struct fs_context *fc,
2539 struct super_block *sb)
2540{
2541#ifdef CONFIG_QUOTA
2542 struct ext4_fs_context *ctx = fc->fs_private;
2543 struct ext4_sb_info *sbi = EXT4_SB(sb);
2544 bool quota_feature = ext4_has_feature_quota(sb);
2545 bool quota_loaded = sb_any_quota_loaded(sb);
6e47a3cc
LC		 2546 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2547 int quota_flags, i;
2548
2549 /*
2550 * We do the test below only for project quotas. 'usrquota' and
2551 * 'grpquota' mount options are allowed even without quota feature
2552 * to support legacy quotas in quota files.
2553 */
2554 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2555 !ext4_has_feature_project(sb)) {
2556 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2557 "Cannot enable project quota enforcement.");
2558 return -EINVAL;
2559 }
e6e268cb 2560
6e47a3cc
LC		 2561 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2562 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2563 if (quota_loaded &&
2564 ctx->mask_s_mount_opt & quota_flags &&
2565 !ctx_test_mount_opt(ctx, quota_flags))
2566 goto err_quota_change;
2567
2568 if (ctx->spec & EXT4_SPEC_JQUOTA) {
e6e268cb
LC		 2569
2570 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2571 if (!(ctx->qname_spec & (1 << i)))
2572 continue;
2573
6e47a3cc
LC		 2574 if (quota_loaded &&
2575 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
e6e268cb
LC		 2576 goto err_jquota_change;
2577
2578 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
13b215a9 2579 strcmp(get_qf_name(sb, sbi, i),
e6e268cb
LC		 2580 ctx->s_qf_names[i]) != 0)
2581 goto err_jquota_specified;
2582 }
6e47a3cc
LC		 2583
2584 if (quota_feature) {
2585 ext4_msg(NULL, KERN_INFO,
2586 "Journaled quota options ignored when "
2587 "QUOTA feature is enabled");
2588 return 0;
2589 }
e6e268cb
LC		 2590 }
2591
6e47a3cc 2592 if (ctx->spec & EXT4_SPEC_JQFMT) {
e6e268cb 2593 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
6e47a3cc 2594 goto err_jquota_change;
e6e268cb
LC		 2595 if (quota_feature) {
2596 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2597 "ignored when QUOTA feature is enabled");
2598 return 0;
2599 }
2600 }
6e47a3cc
LC		 2601
2602 /* Make sure we don't mix old and new quota format */
2603 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2604 ctx->s_qf_names[USRQUOTA]);
2605 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2606 ctx->s_qf_names[GRPQUOTA]);
2607
2608 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2609 test_opt(sb, USRQUOTA));
2610
2611 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2612 test_opt(sb, GRPQUOTA));
2613
2614 if (usr_qf_name) {
2615 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2616 usrquota = false;
2617 }
2618 if (grp_qf_name) {
2619 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2620 grpquota = false;
2621 }
2622
2623 if (usr_qf_name || grp_qf_name) {
2624 if (usrquota || grpquota) {
2625 ext4_msg(NULL, KERN_ERR, "old and new quota "
2626 "format mixing");
2627 return -EINVAL;
2628 }
2629
2630 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2631 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2632 "not specified");
2633 return -EINVAL;
2634 }
2635 }
2636
e6e268cb
LC		 2637 return 0;
2638
2639err_quota_change:
2640 ext4_msg(NULL, KERN_ERR,
2641 "Cannot change quota options when quota turned on");
2642 return -EINVAL;
2643err_jquota_change:
2644 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2645 "options when quota turned on");
2646 return -EINVAL;
2647err_jquota_specified:
2648 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2649 QTYPE2NAME(i));
2650 return -EINVAL;
e6e268cb
LC		 2651#else
2652 return 0;
2653#endif
2654}
2655
5f41fdae
EB		 2656static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2657 struct super_block *sb)
2658{
5f41fdae
EB		 2659 const struct ext4_fs_context *ctx = fc->fs_private;
2660 const struct ext4_sb_info *sbi = EXT4_SB(sb);
85456054 2661 int err;
5f41fdae 2662
85456054 2663 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
5f41fdae
EB		 2664 return 0;
2665
2666 if (!ext4_has_feature_encrypt(sb)) {
2667 ext4_msg(NULL, KERN_WARNING,
2668 "test_dummy_encryption requires encrypt feature");
2669 return -EINVAL;
2670 }
2671 /*
2672 * This mount option is just for testing, and it's not worthwhile to
2673 * implement the extra complexity (e.g. RCU protection) that would be
2674 * needed to allow it to be set or changed during remount. We do allow
2675 * it to be specified during remount, but only if there is no change.
2676 */
85456054
EB		 2677 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2678 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2679 &ctx->dummy_enc_policy))
2680 return 0;
5f41fdae 2681 ext4_msg(NULL, KERN_WARNING,
85456054 2682 "Can't set or change test_dummy_encryption on remount");
5f41fdae
EB		 2683 return -EINVAL;
2684 }
85456054
EB		 2685 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2686 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2687 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2688 &ctx->dummy_enc_policy))
2689 return 0;
2690 ext4_msg(NULL, KERN_WARNING,
2691 "Conflicting test_dummy_encryption options");
2692 return -EINVAL;
2693 }
2694 /*
2695 * fscrypt_add_test_dummy_key() technically changes the super_block, so
2696 * technically it should be delayed until ext4_apply_options() like the
2697 * other changes. But since we never get here for remounts (see above),
2698 * and this is the last chance to report errors, we do it here.
2699 */
2700 err = fscrypt_add_test_dummy_key(sb, &ctx->dummy_enc_policy);
2701 if (err)
2702 ext4_msg(NULL, KERN_WARNING,
2703 "Error adding test dummy encryption key [%d]", err);
2704 return err;
2705}
2706
2707static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2708 struct super_block *sb)
2709{
2710 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2711 /* if already set, it was already verified to be the same */
2712 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2713 return;
2714 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2715 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2716 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
5f41fdae
EB		 2717}
2718
b6bd2435
LC		 2719static int ext4_check_opt_consistency(struct fs_context *fc,
2720 struct super_block *sb)
2721{
2722 struct ext4_fs_context *ctx = fc->fs_private;
6e47a3cc
LC		 2723 struct ext4_sb_info *sbi = fc->s_fs_info;
2724 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
5f41fdae 2725 int err;
b6bd2435
LC		 2726
2727 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2728 ext4_msg(NULL, KERN_ERR,
2729 "Mount option(s) incompatible with ext2");
2730 return -EINVAL;
2731 }
2732 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2733 ext4_msg(NULL, KERN_ERR,
2734 "Mount option(s) incompatible with ext3");
2735 return -EINVAL;
2736 }
2737
6e47a3cc
LC		 2738 if (ctx->s_want_extra_isize >
2739 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2740 ext4_msg(NULL, KERN_ERR,
2741 "Invalid want_extra_isize %d",
2742 ctx->s_want_extra_isize);
2743 return -EINVAL;
2744 }
2745
2746 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) {
2747 int blocksize =
2748 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2749 if (blocksize < PAGE_SIZE)
2750 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an "
2751 "experimental mount option 'dioread_nolock' "
2752 "for blocksize < PAGE_SIZE");
2753 }
2754
5f41fdae
EB		 2755 err = ext4_check_test_dummy_encryption(fc, sb);
2756 if (err)
2757 return err;
6e47a3cc
LC		 2758
2759 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2760 if (!sbi->s_journal) {
2761 ext4_msg(NULL, KERN_WARNING,
2762 "Remounting file system with no journal "
2763 "so ignoring journalled data option");
2764 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
4c246728
LC		 2765 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2766 test_opt(sb, DATA_FLAGS)) {
6e47a3cc
LC		 2767 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2768 "on remount");
2769 return -EINVAL;
2770 }
2771 }
2772
2773 if (is_remount) {
2774 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2775 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2776 ext4_msg(NULL, KERN_ERR, "can't mount with "
2777 "both data=journal and dax");
2778 return -EINVAL;
2779 }
2780
2781 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2782 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2783 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2784fail_dax_change_remount:
2785 ext4_msg(NULL, KERN_ERR, "can't change "
2786 "dax mount option while remounting");
2787 return -EINVAL;
2788 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2789 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2790 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2791 goto fail_dax_change_remount;
2792 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2793 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2794 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2795 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2796 goto fail_dax_change_remount;
2797 }
2798 }
2799
b6bd2435
LC		 2800 return ext4_check_quota_consistency(fc, sb);
2801}
2802
85456054 2803static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
4c94bff9 2804{
6e47a3cc 2805 struct ext4_fs_context *ctx = fc->fs_private;
da812f61 2806 struct ext4_sb_info *sbi = fc->s_fs_info;
6e47a3cc
LC		 2807
2808 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2809 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2810 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2811 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2812 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags;
2813 sbi->s_mount_flags |= ctx->vals_s_mount_flags;
2814 sb->s_flags &= ~ctx->mask_s_flags;
2815 sb->s_flags |= ctx->vals_s_flags;
2816
960e0ab6
LC		 2817 /*
2818 * i_version differs from common mount option iversion so we have
2819 * to let vfs know that it was set, otherwise it would get cleared
2820 * on remount
2821 */
2822 if (ctx->mask_s_flags & SB_I_VERSION)
2823 fc->sb_flags |= SB_I_VERSION;
2824
6e47a3cc
LC		 2825#define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2826 APPLY(s_commit_interval);
2827 APPLY(s_stripe);
2828 APPLY(s_max_batch_time);
2829 APPLY(s_min_batch_time);
2830 APPLY(s_want_extra_isize);
2831 APPLY(s_inode_readahead_blks);
2832 APPLY(s_max_dir_size_kb);
2833 APPLY(s_li_wait_mult);
2834 APPLY(s_resgid);
2835 APPLY(s_resuid);
2836
2837#ifdef CONFIG_EXT4_DEBUG
2838 APPLY(s_fc_debug_max_replay);
2839#endif
2840
2841 ext4_apply_quota_options(fc, sb);
85456054 2842 ext4_apply_test_dummy_encryption(ctx, sb);
6e47a3cc
LC		 2843}
2844
2845
2846static int ext4_validate_options(struct fs_context *fc)
2847{
ac27a0ec 2848#ifdef CONFIG_QUOTA
6e47a3cc 2849 struct ext4_fs_context *ctx = fc->fs_private;
4c94bff9 2850 char *usr_qf_name, *grp_qf_name;
6e47a3cc
LC		 2851
2852 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2853 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2854
33458eab 2855 if (usr_qf_name || grp_qf_name) {
6e47a3cc
LC		 2856 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2857 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
ac27a0ec 2858
6e47a3cc
LC		 2859 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2860 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
ac27a0ec 2861
6e47a3cc
LC		 2862 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2863 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
da812f61 2864 ext4_msg(NULL, KERN_ERR, "old and new quota "
6e47a3cc 2865 "format mixing");
da812f61 2866 return -EINVAL;
ac27a0ec 2867 }
ac27a0ec
DK		 2868 }
2869#endif
6e47a3cc 2870 return 1;
ac27a0ec
DK		 2871}
2872
2adf6da8
TT		 2873static inline void ext4_show_quota_options(struct seq_file *seq,
2874 struct super_block *sb)
2875{
2876#if defined(CONFIG_QUOTA)
2877 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 2878 char *usr_qf_name, *grp_qf_name;
2adf6da8
TT		 2879
2880 if (sbi->s_jquota_fmt) {
2881 char *fmtname = "";
2882
2883 switch (sbi->s_jquota_fmt) {
2884 case QFMT_VFS_OLD:
2885 fmtname = "vfsold";
2886 break;
2887 case QFMT_VFS_V0:
2888 fmtname = "vfsv0";
2889 break;
2890 case QFMT_VFS_V1:
2891 fmtname = "vfsv1";
2892 break;
2893 }
2894 seq_printf(seq, ",jqfmt=%s", fmtname);
2895 }
2896
33458eab
TT		 2897 rcu_read_lock();
2898 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2899 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2900 if (usr_qf_name)
2901 seq_show_option(seq, "usrjquota", usr_qf_name);
2902 if (grp_qf_name)
2903 seq_show_option(seq, "grpjquota", grp_qf_name);
2904 rcu_read_unlock();
2adf6da8
TT		 2905#endif
2906}
2907
5a916be1
TT		 2908static const char *token2str(int token)
2909{
97d8a670 2910 const struct fs_parameter_spec *spec;
5a916be1 2911
97d8a670
LC		 2912 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2913 if (spec->opt == token && !spec->type)
5a916be1 2914 break;
97d8a670 2915 return spec->name;
5a916be1
TT		 2916}
2917
2adf6da8
TT		 2918/*
2919 * Show an option if
2920 * - it's set to a non-default value OR
2921 * - if the per-sb default is different from the global default
2922 */
66acdcf4
TT		 2923static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2924 int nodefs)
2adf6da8 2925{
2adf6da8
TT		 2926 struct ext4_sb_info *sbi = EXT4_SB(sb);
2927 struct ext4_super_block *es = sbi->s_es;
68afa7e0 2928 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
5a916be1 2929 const struct mount_opts *m;
66acdcf4 2930 char sep = nodefs ? '\n' : ',';
2adf6da8 2931
66acdcf4
TT		 2932#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2933#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2934
2935 if (sbi->s_sb_block != 1)
5a916be1
TT		 2936 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2937
2938 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2939 int want_set = m->flags & MOPT_SET;
2940 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
ba2e524d 2941 m->flags & MOPT_SKIP)
5a916be1 2942 continue;
68afa7e0 2943 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 2944 continue; /* skip if same as the default */
2945 if ((want_set &&
2946 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2947 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2948 continue; /* select Opt_noFoo vs Opt_Foo */
2949 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2950 }
5a916be1 2951
08cefc7a 2952 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2953 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2954 SEQ_OPTS_PRINT("resuid=%u",
2955 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2956 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2957 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2958 SEQ_OPTS_PRINT("resgid=%u",
2959 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2960 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2961 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2962 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2963 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2964 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2965 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2966 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2967 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2968 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2969 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2970 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2971 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2972 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
357fdad0 2973 if (sb->s_flags & SB_I_VERSION)
5a916be1 2974 SEQ_OPTS_PUTS("i_version");
66acdcf4 2975 if (nodefs || sbi->s_stripe)
5a916be1 2976 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0
TN		 2977 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2978 (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 2979 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2980 SEQ_OPTS_PUTS("data=journal");
2981 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2982 SEQ_OPTS_PUTS("data=ordered");
2983 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2984 SEQ_OPTS_PUTS("data=writeback");
2985 }
66acdcf4
TT		 2986 if (nodefs ||
2987 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2988 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2989 sbi->s_inode_readahead_blks);
2adf6da8 2990
ceec0376 2991 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 2992 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2993 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2994 if (nodefs || sbi->s_max_dir_size_kb)
2995 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2996 if (test_opt(sb, DATA_ERR_ABORT))
2997 SEQ_OPTS_PUTS("data_err=abort");
ed318a6c
EB		 2998
2999 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2adf6da8 3000
4f74d15f
EB		 3001 if (sb->s_flags & SB_INLINECRYPT)
3002 SEQ_OPTS_PUTS("inlinecrypt");
3003
9cb20f94
IW		 3004 if (test_opt(sb, DAX_ALWAYS)) {
3005 if (IS_EXT2_SB(sb))
3006 SEQ_OPTS_PUTS("dax");
3007 else
3008 SEQ_OPTS_PUTS("dax=always");
3009 } else if (test_opt2(sb, DAX_NEVER)) {
3010 SEQ_OPTS_PUTS("dax=never");
3011 } else if (test_opt2(sb, DAX_INODE)) {
3012 SEQ_OPTS_PUTS("dax=inode");
3013 }
3fa5d23e
OM		 3014
3015 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3016 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3017 SEQ_OPTS_PUTS("mb_optimize_scan=0");
3018 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3019 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3020 SEQ_OPTS_PUTS("mb_optimize_scan=1");
3021 }
3022
2adf6da8 3023 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 3024 return 0;
3025}
3026
66acdcf4
TT		 3027static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3028{
3029 return _ext4_show_options(seq, root->d_sb, 0);
3030}
3031
ebd173be 3032int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 3033{
3034 struct super_block *sb = seq->private;
3035 int rc;
3036
bc98a42c 3037 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 3038 rc = _ext4_show_options(seq, sb, 1);
3039 seq_puts(seq, "\n");
3040 return rc;
3041}
3042
617ba13b 3043static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 3044 int read_only)
3045{
617ba13b 3046 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 3047 int err = 0;
ac27a0ec 3048
617ba13b 3049 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 3050 ext4_msg(sb, KERN_ERR, "revision level too high, "
3051 "forcing read-only mode");
c89128a0 3052 err = -EROFS;
5adaccac 3053 goto done;
ac27a0ec
DK		 3054 }
3055 if (read_only)
281b5995 3056 goto done;
617ba13b 3057 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 3058 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3059 "running e2fsck is recommended");
c8b459f4 3060 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 3061 ext4_msg(sb, KERN_WARNING,
3062 "warning: mounting fs with errors, "
3063 "running e2fsck is recommended");
ed3ce80a 3064 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 3065 le16_to_cpu(es->s_mnt_count) >=
3066 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 3067 ext4_msg(sb, KERN_WARNING,
3068 "warning: maximal mount count reached, "
3069 "running e2fsck is recommended");
ac27a0ec 3070 else if (le32_to_cpu(es->s_checkinterval) &&
6a0678a7
AB		 3071 (ext4_get_tstamp(es, s_lastcheck) +
3072 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
b31e1552
ES		 3073 ext4_msg(sb, KERN_WARNING,
3074 "warning: checktime reached, "
3075 "running e2fsck is recommended");
0b8e58a1 3076 if (!sbi->s_journal)
0390131b 3077 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 3078 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 3079 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 3080 le16_add_cpu(&es->s_mnt_count, 1);
6a0678a7 3081 ext4_update_tstamp(es, s_mtime);
02f310fc 3082 if (sbi->s_journal) {
e2b911c5 3083 ext4_set_feature_journal_needs_recovery(sb);
02f310fc
JK		 3084 if (ext4_has_feature_orphan_file(sb))
3085 ext4_set_feature_orphan_present(sb);
3086 }
ac27a0ec 3087
4392fbc4 3088 err = ext4_commit_super(sb);
281b5995 3089done:
ac27a0ec 3090 if (test_opt(sb, DEBUG))
a9df9a49 3091 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 3092 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 3093 sb->s_blocksize,
3094 sbi->s_groups_count,
617ba13b
MC		 3095 EXT4_BLOCKS_PER_GROUP(sb),
3096 EXT4_INODES_PER_GROUP(sb),
a2595b8a 3097 sbi->s_mount_opt, sbi->s_mount_opt2);
c89128a0 3098 return err;
ac27a0ec
DK		 3099}
3100
117fff10
TT		 3101int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3102{
3103 struct ext4_sb_info *sbi = EXT4_SB(sb);
7c990728 3104 struct flex_groups **old_groups, **new_groups;
37b0b6b8 3105 int size, i, j;
117fff10
TT		 3106
3107 if (!sbi->s_log_groups_per_flex)
3108 return 0;
3109
3110 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3111 if (size <= sbi->s_flex_groups_allocated)
3112 return 0;
3113
7c990728
SJS		 3114 new_groups = kvzalloc(roundup_pow_of_two(size *
3115 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
117fff10 3116 if (!new_groups) {
7c990728
SJS		 3117 ext4_msg(sb, KERN_ERR,
3118 "not enough memory for %d flex group pointers", size);
117fff10
TT		 3119 return -ENOMEM;
3120 }
7c990728
SJS		 3121 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3122 new_groups[i] = kvzalloc(roundup_pow_of_two(
3123 sizeof(struct flex_groups)),
3124 GFP_KERNEL);
3125 if (!new_groups[i]) {
37b0b6b8
DC		 3126 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3127 kvfree(new_groups[j]);
7c990728
SJS		 3128 kvfree(new_groups);
3129 ext4_msg(sb, KERN_ERR,
3130 "not enough memory for %d flex groups", size);
3131 return -ENOMEM;
3132 }
117fff10 3133 }
7c990728
SJS		 3134 rcu_read_lock();
3135 old_groups = rcu_dereference(sbi->s_flex_groups);
3136 if (old_groups)
3137 memcpy(new_groups, old_groups,
3138 (sbi->s_flex_groups_allocated *
3139 sizeof(struct flex_groups *)));
3140 rcu_read_unlock();
3141 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3142 sbi->s_flex_groups_allocated = size;
3143 if (old_groups)
3144 ext4_kvfree_array_rcu(old_groups);
117fff10
TT		 3145 return 0;
3146}
3147
772cb7c8
JS		 3148static int ext4_fill_flex_info(struct super_block *sb)
3149{
3150 struct ext4_sb_info *sbi = EXT4_SB(sb);
3151 struct ext4_group_desc *gdp = NULL;
7c990728 3152 struct flex_groups *fg;
772cb7c8 3153 ext4_group_t flex_group;
117fff10 3154 int i, err;
772cb7c8 3155
503358ae 3156 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 3157 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 3158 sbi->s_log_groups_per_flex = 0;
3159 return 1;
3160 }
3161
117fff10
TT		 3162 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3163 if (err)
9933fc0a 3164 goto failed;
772cb7c8 3165
772cb7c8 3166 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 3167 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 3168
3169 flex_group = ext4_flex_group(sbi, i);
7c990728
SJS		 3170 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3171 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
90ba983f 3172 atomic64_add(ext4_free_group_clusters(sb, gdp),
7c990728
SJS		 3173 &fg->free_clusters);
3174 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
772cb7c8
JS		 3175 }
3176
3177 return 1;
3178failed:
3179 return 0;
3180}
3181
e2b911c5 3182static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 3183 struct ext4_group_desc *gdp)
717d50e4 3184{
b47820ed 3185 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 3186 __u16 crc = 0;
feb0ab32 3187 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 3188 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 3189
9aa5d32b 3190 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 3191 /* Use new metadata_csum algorithm */
feb0ab32 3192 __u32 csum32;
b47820ed 3193 __u16 dummy_csum = 0;
feb0ab32 3194
feb0ab32
DW		 3195 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3196 sizeof(le_group));
b47820ed
DJ		 3197 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3198 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3199 sizeof(dummy_csum));
3200 offset += sizeof(dummy_csum);
3201 if (offset < sbi->s_desc_size)
3202 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3203 sbi->s_desc_size - offset);
feb0ab32
DW		 3204
3205 crc = csum32 & 0xFFFF;
3206 goto out;
717d50e4
AD		 3207 }
3208
feb0ab32 3209 /* old crc16 code */
e2b911c5 3210 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 3211 return 0;
3212
feb0ab32
DW		 3213 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3214 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3215 crc = crc16(crc, (__u8 *)gdp, offset);
3216 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3217 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 3218 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 3219 offset < le16_to_cpu(sbi->s_es->s_desc_size))
3220 crc = crc16(crc, (__u8 *)gdp + offset,
3221 le16_to_cpu(sbi->s_es->s_desc_size) -
3222 offset);
3223
3224out:
717d50e4
AD		 3225 return cpu_to_le16(crc);
3226}
3227
feb0ab32 3228int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 3229 struct ext4_group_desc *gdp)
3230{
feb0ab32 3231 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 3232 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 3233 return 0;
3234
3235 return 1;
3236}
3237
feb0ab32
DW		 3238void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3239 struct ext4_group_desc *gdp)
3240{
3241 if (!ext4_has_group_desc_csum(sb))
3242 return;
e2b911c5 3243 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 3244}
3245
ac27a0ec 3246/* Called at mount-time, super-block is locked */
bfff6873 3247static int ext4_check_descriptors(struct super_block *sb,
829fa70d 3248 ext4_fsblk_t sb_block,
bfff6873 3249 ext4_group_t *first_not_zeroed)
ac27a0ec 3250{
617ba13b
MC		 3251 struct ext4_sb_info *sbi = EXT4_SB(sb);
3252 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3253 ext4_fsblk_t last_block;
44de022c 3254 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
bd81d8ee
LV		 3255 ext4_fsblk_t block_bitmap;
3256 ext4_fsblk_t inode_bitmap;
3257 ext4_fsblk_t inode_table;
ce421581 3258 int flexbg_flag = 0;
bfff6873 3259 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 3260
e2b911c5 3261 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 3262 flexbg_flag = 1;
3263
af5bc92d 3264 ext4_debug("Checking group descriptors");
ac27a0ec 3265
197cd65a
AM		 3266 for (i = 0; i < sbi->s_groups_count; i++) {
3267 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3268
ce421581 3269 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 3270 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 3271 else
3272 last_block = first_block +
617ba13b 3273 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 3274
bfff6873
LC		 3275 if ((grp == sbi->s_groups_count) &&
3276 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3277 grp = i;
3278
8fadc143 3279 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 3280 if (block_bitmap == sb_block) {
3281 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3282 "Block bitmap for group %u overlaps "
3283 "superblock", i);
18db4b4e
TT		 3284 if (!sb_rdonly(sb))
3285 return 0;
829fa70d 3286 }
77260807
TT		 3287 if (block_bitmap >= sb_block + 1 &&
3288 block_bitmap <= last_bg_block) {
3289 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3290 "Block bitmap for group %u overlaps "
3291 "block group descriptors", i);
3292 if (!sb_rdonly(sb))
3293 return 0;
3294 }
2b2d6d01 3295 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 3296 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 3297 "Block bitmap for group %u not in group "
b31e1552 3298 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 3299 return 0;
3300 }
8fadc143 3301 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 3302 if (inode_bitmap == sb_block) {
3303 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3304 "Inode bitmap for group %u overlaps "
3305 "superblock", i);
18db4b4e
TT		 3306 if (!sb_rdonly(sb))
3307 return 0;
829fa70d 3308 }
77260807
TT		 3309 if (inode_bitmap >= sb_block + 1 &&
3310 inode_bitmap <= last_bg_block) {
3311 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3312 "Inode bitmap for group %u overlaps "
3313 "block group descriptors", i);
3314 if (!sb_rdonly(sb))
3315 return 0;
3316 }
2b2d6d01 3317 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 3318 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 3319 "Inode bitmap for group %u not in group "
b31e1552 3320 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 3321 return 0;
3322 }
8fadc143 3323 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 3324 if (inode_table == sb_block) {
3325 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3326 "Inode table for group %u overlaps "
3327 "superblock", i);
18db4b4e
TT		 3328 if (!sb_rdonly(sb))
3329 return 0;
829fa70d 3330 }
77260807
TT		 3331 if (inode_table >= sb_block + 1 &&
3332 inode_table <= last_bg_block) {
3333 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3334 "Inode table for group %u overlaps "
3335 "block group descriptors", i);
3336 if (!sb_rdonly(sb))
3337 return 0;
3338 }
bd81d8ee 3339 if (inode_table < first_block ||
2b2d6d01 3340 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 3341 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 3342 "Inode table for group %u not in group "
b31e1552 3343 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 3344 return 0;
3345 }
955ce5f5 3346 ext4_lock_group(sb, i);
feb0ab32 3347 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 3348 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3349 "Checksum for group %u failed (%u!=%u)",
e2b911c5 3350 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 3351 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 3352 if (!sb_rdonly(sb)) {
955ce5f5 3353 ext4_unlock_group(sb, i);
8a266467 3354 return 0;
7ee1ec4c 3355 }
717d50e4 3356 }
955ce5f5 3357 ext4_unlock_group(sb, i);
ce421581
JS		 3358 if (!flexbg_flag)
3359 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 3360 }
bfff6873
LC		 3361 if (NULL != first_not_zeroed)
3362 *first_not_zeroed = grp;
ac27a0ec
DK		 3363 return 1;
3364}
3365
cd2291a4
ES		 3366/*
3367 * Maximal extent format file size.
3368 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3369 * extent format containers, within a sector_t, and within i_blocks
3370 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3371 * so that won't be a limiting factor.
3372 *
f17722f9
LC		 3373 * However there is other limiting factor. We do store extents in the form
3374 * of starting block and length, hence the resulting length of the extent
3375 * covering maximum file size must fit into on-disk format containers as
3376 * well. Given that length is always by 1 unit bigger than max unit (because
3377 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3378 *
cd2291a4
ES		 3379 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3380 */
f287a1a5 3381static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 3382{
3383 loff_t res;
3384 loff_t upper_limit = MAX_LFS_FILESIZE;
3385
72deb455
CH		 3386 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3387
3388 if (!has_huge_files) {
cd2291a4
ES		 3389 upper_limit = (1LL << 32) - 1;
3390
3391 /* total blocks in file system block size */
3392 upper_limit >>= (blkbits - 9);
3393 upper_limit <<= blkbits;
3394 }
3395
f17722f9
LC		 3396 /*
3397 * 32-bit extent-start container, ee_block. We lower the maxbytes
3398 * by one fs block, so ee_len can cover the extent of maximum file
3399 * size
3400 */
3401 res = (1LL << 32) - 1;
cd2291a4 3402 res <<= blkbits;
cd2291a4
ES		 3403
3404 /* Sanity check against vm- & vfs- imposed limits */
3405 if (res > upper_limit)
3406 res = upper_limit;
3407
3408 return res;
3409}
ac27a0ec 3410
ac27a0ec 3411/*
cd2291a4 3412 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 3413 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3414 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 3415 */
f287a1a5 3416static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 3417{
5c93e8ec 3418 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
0fc1b451 3419 int meta_blocks;
5c93e8ec 3420 unsigned int ppb = 1 << (bits - 2);
75ca6ad4
RH		 3421
3422 /*
3423 * This is calculated to be the largest file size for a dense, block
0b8e58a1
AD		 3424 * mapped file such that the file's total number of 512-byte sectors,
3425 * including data and all indirect blocks, does not exceed (2^48 - 1).
3426 *
3427 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3428 * number of 512-byte sectors of the file.
0fc1b451 3429 */
72deb455 3430 if (!has_huge_files) {
0fc1b451 3431 /*
72deb455
CH		 3432 * !has_huge_files or implies that the inode i_block field
3433 * represents total file blocks in 2^32 512-byte sectors ==
3434 * size of vfs inode i_blocks * 8
0fc1b451
AK		 3435 */
3436 upper_limit = (1LL << 32) - 1;
3437
3438 /* total blocks in file system block size */
3439 upper_limit >>= (bits - 9);
3440
3441 } else {
8180a562
AK		 3442 /*
3443 * We use 48 bit ext4_inode i_blocks
3444 * With EXT4_HUGE_FILE_FL set the i_blocks
3445 * represent total number of blocks in
3446 * file system block size
3447 */
0fc1b451
AK		 3448 upper_limit = (1LL << 48) - 1;
3449
0fc1b451
AK		 3450 }
3451
5c93e8ec
ZY		 3452 /* Compute how many blocks we can address by block tree */
3453 res += ppb;
3454 res += ppb * ppb;
3455 res += ((loff_t)ppb) * ppb * ppb;
3456 /* Compute how many metadata blocks are needed */
3457 meta_blocks = 1;
3458 meta_blocks += 1 + ppb;
3459 meta_blocks += 1 + ppb + ppb * ppb;
3460 /* Does block tree limit file size? */
3461 if (res + meta_blocks <= upper_limit)
3462 goto check_lfs;
3463
3464 res = upper_limit;
3465 /* How many metadata blocks are needed for addressing upper_limit? */
3466 upper_limit -= EXT4_NDIR_BLOCKS;
0fc1b451
AK		 3467 /* indirect blocks */
3468 meta_blocks = 1;
5c93e8ec 3469 upper_limit -= ppb;
0fc1b451 3470 /* double indirect blocks */
5c93e8ec
ZY		 3471 if (upper_limit < ppb * ppb) {
3472 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3473 res -= meta_blocks;
3474 goto check_lfs;
3475 }
3476 meta_blocks += 1 + ppb;
3477 upper_limit -= ppb * ppb;
3478 /* tripple indirect blocks for the rest */
3479 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3480 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3481 res -= meta_blocks;
3482check_lfs:
ac27a0ec 3483 res <<= bits;
0fc1b451
AK		 3484 if (res > MAX_LFS_FILESIZE)
3485 res = MAX_LFS_FILESIZE;
3486
5c93e8ec 3487 return res;
ac27a0ec
DK		 3488}
3489
617ba13b 3490static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 3491 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 3492{
617ba13b 3493 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 3494 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 3495 int has_super = 0;
3496
3497 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3498
e2b911c5 3499 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 3500 return logical_sb_block + nr + 1;
ac27a0ec 3501 bg = sbi->s_desc_per_block * nr;
617ba13b 3502 if (ext4_bg_has_super(sb, bg))
ac27a0ec 3503 has_super = 1;
0b8e58a1 3504
bd63f6b0
DW		 3505 /*
3506 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3507 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3508 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3509 * compensate.
3510 */
3511 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 3512 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 3513 has_super++;
3514
617ba13b 3515 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 3516}
3517
c9de560d
AT		 3518/**
3519 * ext4_get_stripe_size: Get the stripe size.
3520 * @sbi: In memory super block info
3521 *
3522 * If we have specified it via mount option, then
3523 * use the mount option value. If the value specified at mount time is
3524 * greater than the blocks per group use the super block value.
3525 * If the super block value is greater than blocks per group return 0.
3526 * Allocator needs it be less than blocks per group.
3527 *
3528 */
3529static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3530{
3531 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3532 unsigned long stripe_width =
3533 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 3534 int ret;
c9de560d
AT		 3535
3536 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 3537 ret = sbi->s_stripe;
5469d7c3 3538 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 3539 ret = stripe_width;
5469d7c3 3540 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 3541 ret = stride;
3542 else
3543 ret = 0;
c9de560d 3544
3eb08658
DE		 3545 /*
3546 * If the stripe width is 1, this makes no sense and
3547 * we set it to 0 to turn off stripe handling code.
3548 */
3549 if (ret <= 1)
3550 ret = 0;
c9de560d 3551
3eb08658 3552 return ret;
c9de560d 3553}
ac27a0ec 3554
a13fb1a4
ES		 3555/*
3556 * Check whether this filesystem can be mounted based on
3557 * the features present and the RDONLY/RDWR mount requested.
3558 * Returns 1 if this filesystem can be mounted as requested,
3559 * 0 if it cannot be.
3560 */
25c6d98f 3561int ext4_feature_set_ok(struct super_block *sb, int readonly)
a13fb1a4 3562{
e2b911c5 3563 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 3564 ext4_msg(sb, KERN_ERR,
3565 "Couldn't mount because of "
3566 "unsupported optional features (%x)",
3567 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3568 ~EXT4_FEATURE_INCOMPAT_SUPP));
3569 return 0;
3570 }
3571
5298d4bf 3572#if !IS_ENABLED(CONFIG_UNICODE)
c83ad55e
GKB		 3573 if (ext4_has_feature_casefold(sb)) {
3574 ext4_msg(sb, KERN_ERR,
3575 "Filesystem with casefold feature cannot be "
3576 "mounted without CONFIG_UNICODE");
3577 return 0;
3578 }
3579#endif
3580
a13fb1a4
ES		 3581 if (readonly)
3582 return 1;
3583
e2b911c5 3584 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 3585 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 3586 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 3587 return 1;
3588 }
3589
a13fb1a4 3590 /* Check that feature set is OK for a read-write mount */
e2b911c5 3591 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 3592 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3593 "unsupported optional features (%x)",
3594 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3595 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3596 return 0;
3597 }
e2b911c5 3598 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 3599 ext4_msg(sb, KERN_ERR,
3600 "Can't support bigalloc feature without "
3601 "extents feature\n");
3602 return 0;
3603 }
7c319d32 3604
9db176bc 3605#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
d65d87a0
TT		 3606 if (!readonly && (ext4_has_feature_quota(sb) ||
3607 ext4_has_feature_project(sb))) {
7c319d32 3608 ext4_msg(sb, KERN_ERR,
d65d87a0 3609 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
689c958c
LX		 3610 return 0;
3611 }
7c319d32 3612#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 3613 return 1;
3614}
3615
66e61a9e
TT		 3616/*
3617 * This function is called once a day if we have errors logged
3618 * on the file system
3619 */
235699a8 3620static void print_daily_error_info(struct timer_list *t)
66e61a9e 3621{
235699a8
KC		 3622 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3623 struct super_block *sb = sbi->s_sb;
3624 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 3625
3626 if (es->s_error_count)
ae0f78de
TT		 3627 /* fsck newer than v1.41.13 is needed to clean this condition. */
3628 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 3629 le32_to_cpu(es->s_error_count));
3630 if (es->s_first_error_time) {
6a0678a7
AB		 3631 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3632 sb->s_id,
3633 ext4_get_tstamp(es, s_first_error_time),
66e61a9e
TT		 3634 (int) sizeof(es->s_first_error_func),
3635 es->s_first_error_func,
3636 le32_to_cpu(es->s_first_error_line));
3637 if (es->s_first_error_ino)
651e1c3b 3638 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3639 le32_to_cpu(es->s_first_error_ino));
3640 if (es->s_first_error_block)
651e1c3b 3641 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3642 le64_to_cpu(es->s_first_error_block));
651e1c3b 3643 printk(KERN_CONT "\n");
66e61a9e
TT		 3644 }
3645 if (es->s_last_error_time) {
6a0678a7
AB		 3646 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3647 sb->s_id,
3648 ext4_get_tstamp(es, s_last_error_time),
66e61a9e
TT		 3649 (int) sizeof(es->s_last_error_func),
3650 es->s_last_error_func,
3651 le32_to_cpu(es->s_last_error_line));
3652 if (es->s_last_error_ino)
651e1c3b 3653 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3654 le32_to_cpu(es->s_last_error_ino));
3655 if (es->s_last_error_block)
651e1c3b 3656 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3657 le64_to_cpu(es->s_last_error_block));
651e1c3b 3658 printk(KERN_CONT "\n");
66e61a9e
TT		 3659 }
3660 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3661}
3662
bfff6873
LC		 3663/* Find next suitable group and run ext4_init_inode_table */
3664static int ext4_run_li_request(struct ext4_li_request *elr)
3665{
3666 struct ext4_group_desc *gdp = NULL;
3d392b26
TT		 3667 struct super_block *sb = elr->lr_super;
3668 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3669 ext4_group_t group = elr->lr_next_group;
3d392b26 3670 unsigned int prefetch_ios = 0;
bfff6873 3671 int ret = 0;
39fec688 3672 u64 start_time;
bfff6873 3673
3d392b26
TT		 3674 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3675 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3676 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3677 if (prefetch_ios)
3678 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3679 prefetch_ios);
3680 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3681 prefetch_ios);
3682 if (group >= elr->lr_next_group) {
3683 ret = 1;
3684 if (elr->lr_first_not_zeroed != ngroups &&
3685 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3686 elr->lr_next_group = elr->lr_first_not_zeroed;
3687 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3688 ret = 0;
3689 }
3690 }
3691 return ret;
3692 }
bfff6873 3693
3d392b26 3694 for (; group < ngroups; group++) {
bfff6873
LC		 3695 gdp = ext4_get_group_desc(sb, group, NULL);
3696 if (!gdp) {
3697 ret = 1;
3698 break;
3699 }
3700
3701 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3702 break;
3703 }
3704
7f511862 3705 if (group >= ngroups)
bfff6873
LC		 3706 ret = 1;
3707
3708 if (!ret) {
39fec688 3709 start_time = ktime_get_real_ns();
bfff6873
LC		 3710 ret = ext4_init_inode_table(sb, group,
3711 elr->lr_timeout ? 0 : 1);
3d392b26 3712 trace_ext4_lazy_itable_init(sb, group);
bfff6873 3713 if (elr->lr_timeout == 0) {
39fec688
SX		 3714 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3715 EXT4_SB(elr->lr_super)->s_li_wait_mult);
bfff6873
LC		 3716 }
3717 elr->lr_next_sched = jiffies + elr->lr_timeout;
3718 elr->lr_next_group = group + 1;
3719 }
bfff6873
LC		 3720 return ret;
3721}
3722
3723/*
3724 * Remove lr_request from the list_request and free the
4ed5c033 3725 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 3726 */
3727static void ext4_remove_li_request(struct ext4_li_request *elr)
3728{
bfff6873
LC		 3729 if (!elr)
3730 return;
3731
bfff6873 3732 list_del(&elr->lr_request);
3d392b26 3733 EXT4_SB(elr->lr_super)->s_li_request = NULL;
bfff6873
LC		 3734 kfree(elr);
3735}
3736
3737static void ext4_unregister_li_request(struct super_block *sb)
3738{
1bb933fb
LC		 3739 mutex_lock(&ext4_li_mtx);
3740 if (!ext4_li_info) {
3741 mutex_unlock(&ext4_li_mtx);
bfff6873 3742 return;
1bb933fb 3743 }
bfff6873
LC		 3744
3745 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 3746 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 3747 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 3748 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 3749}
3750
8f1f7453
ES		 3751static struct task_struct *ext4_lazyinit_task;
3752
bfff6873
LC		 3753/*
3754 * This is the function where ext4lazyinit thread lives. It walks
3755 * through the request list searching for next scheduled filesystem.
3756 * When such a fs is found, run the lazy initialization request
3757 * (ext4_rn_li_request) and keep track of the time spend in this
3758 * function. Based on that time we compute next schedule time of
3759 * the request. When walking through the list is complete, compute
3760 * next waking time and put itself into sleep.
3761 */
3762static int ext4_lazyinit_thread(void *arg)
3763{
c30365b9 3764 struct ext4_lazy_init *eli = arg;
bfff6873
LC		 3765 struct list_head *pos, *n;
3766 struct ext4_li_request *elr;
4ed5c033 3767 unsigned long next_wakeup, cur;
bfff6873
LC		 3768
3769 BUG_ON(NULL == eli);
3770
bfff6873
LC		 3771cont_thread:
3772 while (true) {
3773 next_wakeup = MAX_JIFFY_OFFSET;
3774
3775 mutex_lock(&eli->li_list_mtx);
3776 if (list_empty(&eli->li_request_list)) {
3777 mutex_unlock(&eli->li_list_mtx);
3778 goto exit_thread;
3779 }
bfff6873 3780 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3781 int err = 0;
3782 int progress = 0;
bfff6873
LC		 3783 elr = list_entry(pos, struct ext4_li_request,
3784 lr_request);
3785
e22834f0
DM		 3786 if (time_before(jiffies, elr->lr_next_sched)) {
3787 if (time_before(elr->lr_next_sched, next_wakeup))
3788 next_wakeup = elr->lr_next_sched;
3789 continue;
3790 }
3791 if (down_read_trylock(&elr->lr_super->s_umount)) {
3792 if (sb_start_write_trylock(elr->lr_super)) {
3793 progress = 1;
3794 /*
3795 * We hold sb->s_umount, sb can not
3796 * be removed from the list, it is
3797 * now safe to drop li_list_mtx
3798 */
3799 mutex_unlock(&eli->li_list_mtx);
3800 err = ext4_run_li_request(elr);
3801 sb_end_write(elr->lr_super);
3802 mutex_lock(&eli->li_list_mtx);
3803 n = pos->next;
b2c78cd0 3804 }
e22834f0
DM		 3805 up_read((&elr->lr_super->s_umount));
3806 }
3807 /* error, remove the lazy_init job */
3808 if (err) {
3809 ext4_remove_li_request(elr);
3810 continue;
3811 }
3812 if (!progress) {
3813 elr->lr_next_sched = jiffies +
3814 (prandom_u32()
3815 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873 3816 }
bfff6873
LC		 3817 if (time_before(elr->lr_next_sched, next_wakeup))
3818 next_wakeup = elr->lr_next_sched;
3819 }
3820 mutex_unlock(&eli->li_list_mtx);
3821
a0acae0e 3822 try_to_freeze();
bfff6873 3823
4ed5c033
LC		 3824 cur = jiffies;
3825 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3826 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3827 cond_resched();
3828 continue;
3829 }
3830
4ed5c033
LC		 3831 schedule_timeout_interruptible(next_wakeup - cur);
3832
8f1f7453
ES		 3833 if (kthread_should_stop()) {
3834 ext4_clear_request_list();
3835 goto exit_thread;
3836 }
bfff6873
LC		 3837 }
3838
3839exit_thread:
3840 /*
3841 * It looks like the request list is empty, but we need
3842 * to check it under the li_list_mtx lock, to prevent any
3843 * additions into it, and of course we should lock ext4_li_mtx
3844 * to atomically free the list and ext4_li_info, because at
3845 * this point another ext4 filesystem could be registering
3846 * new one.
3847 */
3848 mutex_lock(&ext4_li_mtx);
3849 mutex_lock(&eli->li_list_mtx);
3850 if (!list_empty(&eli->li_request_list)) {
3851 mutex_unlock(&eli->li_list_mtx);
3852 mutex_unlock(&ext4_li_mtx);
3853 goto cont_thread;
3854 }
3855 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3856 kfree(ext4_li_info);
3857 ext4_li_info = NULL;
3858 mutex_unlock(&ext4_li_mtx);
3859
3860 return 0;
3861}
3862
3863static void ext4_clear_request_list(void)
3864{
3865 struct list_head *pos, *n;
3866 struct ext4_li_request *elr;
3867
3868 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3869 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3870 elr = list_entry(pos, struct ext4_li_request,
3871 lr_request);
3872 ext4_remove_li_request(elr);
3873 }
3874 mutex_unlock(&ext4_li_info->li_list_mtx);
3875}
3876
3877static int ext4_run_lazyinit_thread(void)
3878{
8f1f7453
ES		 3879 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3880 ext4_li_info, "ext4lazyinit");
3881 if (IS_ERR(ext4_lazyinit_task)) {
3882 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3883 ext4_clear_request_list();
bfff6873
LC		 3884 kfree(ext4_li_info);
3885 ext4_li_info = NULL;
92b97816 3886 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3887 "initialization thread\n",
3888 err);
3889 return err;
3890 }
3891 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3892 return 0;
3893}
3894
3895/*
3896 * Check whether it make sense to run itable init. thread or not.
3897 * If there is at least one uninitialized inode table, return
3898 * corresponding group number, else the loop goes through all
3899 * groups and return total number of groups.
3900 */
3901static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3902{
3903 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3904 struct ext4_group_desc *gdp = NULL;
3905
8844618d
TT		 3906 if (!ext4_has_group_desc_csum(sb))
3907 return ngroups;
3908
bfff6873
LC		 3909 for (group = 0; group < ngroups; group++) {
3910 gdp = ext4_get_group_desc(sb, group, NULL);
3911 if (!gdp)
3912 continue;
3913
50122847 3914 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
bfff6873
LC		 3915 break;
3916 }
3917
3918 return group;
3919}
3920
3921static int ext4_li_info_new(void)
3922{
3923 struct ext4_lazy_init *eli = NULL;
3924
3925 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3926 if (!eli)
3927 return -ENOMEM;
3928
bfff6873
LC		 3929 INIT_LIST_HEAD(&eli->li_request_list);
3930 mutex_init(&eli->li_list_mtx);
3931
bfff6873
LC		 3932 eli->li_state |= EXT4_LAZYINIT_QUIT;
3933
3934 ext4_li_info = eli;
3935
3936 return 0;
3937}
3938
3939static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3940 ext4_group_t start)
3941{
bfff6873 3942 struct ext4_li_request *elr;
bfff6873
LC		 3943
3944 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3945 if (!elr)
3946 return NULL;
3947
3948 elr->lr_super = sb;
3d392b26 3949 elr->lr_first_not_zeroed = start;
21175ca4 3950 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3d392b26
TT		 3951 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3952 elr->lr_next_group = start;
21175ca4
HS		 3953 } else {
3954 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3d392b26 3955 }
bfff6873
LC		 3956
3957 /*
3958 * Randomize first schedule time of the request to
3959 * spread the inode table initialization requests
3960 * better.
3961 */
dd1f723b
TT		 3962 elr->lr_next_sched = jiffies + (prandom_u32() %
3963 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3964 return elr;
3965}
3966
7f511862
TT		 3967int ext4_register_li_request(struct super_block *sb,
3968 ext4_group_t first_not_zeroed)
bfff6873
LC		 3969{
3970 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3971 struct ext4_li_request *elr = NULL;
49598e04 3972 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 3973 int ret = 0;
bfff6873 3974
7f511862 3975 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3976 if (sbi->s_li_request != NULL) {
3977 /*
3978 * Reset timeout so it can be computed again, because
3979 * s_li_wait_mult might have changed.
3980 */
3981 sbi->s_li_request->lr_timeout = 0;
7f511862 3982 goto out;
51ce6511 3983 }
bfff6873 3984
21175ca4 3985 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3d392b26
TT		 3986 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3987 !test_opt(sb, INIT_INODE_TABLE)))
7f511862 3988 goto out;
bfff6873
LC		 3989
3990 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3991 if (!elr) {
3992 ret = -ENOMEM;
3993 goto out;
3994 }
bfff6873
LC		 3995
3996 if (NULL == ext4_li_info) {
3997 ret = ext4_li_info_new();
3998 if (ret)
3999 goto out;
4000 }
4001
4002 mutex_lock(&ext4_li_info->li_list_mtx);
4003 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
4004 mutex_unlock(&ext4_li_info->li_list_mtx);
4005
4006 sbi->s_li_request = elr;
46e4690b
TM		 4007 /*
4008 * set elr to NULL here since it has been inserted to
4009 * the request_list and the removal and free of it is
4010 * handled by ext4_clear_request_list from now on.
4011 */
4012 elr = NULL;
bfff6873
LC		 4013
4014 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4015 ret = ext4_run_lazyinit_thread();
4016 if (ret)
4017 goto out;
4018 }
bfff6873 4019out:
beed5ecb
NK		 4020 mutex_unlock(&ext4_li_mtx);
4021 if (ret)
bfff6873 4022 kfree(elr);
bfff6873
LC		 4023 return ret;
4024}
4025
4026/*
4027 * We do not need to lock anything since this is called on
4028 * module unload.
4029 */
4030static void ext4_destroy_lazyinit_thread(void)
4031{
4032 /*
4033 * If thread exited earlier
4034 * there's nothing to be done.
4035 */
8f1f7453 4036 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 4037 return;
4038
8f1f7453 4039 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 4040}
4041
25ed6e8a
DW		 4042static int set_journal_csum_feature_set(struct super_block *sb)
4043{
4044 int ret = 1;
4045 int compat, incompat;
4046 struct ext4_sb_info *sbi = EXT4_SB(sb);
4047
9aa5d32b 4048 if (ext4_has_metadata_csum(sb)) {
db9ee220 4049 /* journal checksum v3 */
25ed6e8a 4050 compat = 0;
db9ee220 4051 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 4052 } else {
4053 /* journal checksum v1 */
4054 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4055 incompat = 0;
4056 }
4057
feb8c6d3
DW		 4058 jbd2_journal_clear_features(sbi->s_journal,
4059 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4060 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4061 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 4062 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4063 ret = jbd2_journal_set_features(sbi->s_journal,
4064 compat, 0,
4065 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4066 incompat);
4067 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4068 ret = jbd2_journal_set_features(sbi->s_journal,
4069 compat, 0,
4070 incompat);
4071 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4072 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4073 } else {
feb8c6d3
DW		 4074 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4075 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 4076 }
4077
4078 return ret;
4079}
4080
952fc18e
TT		 4081/*
4082 * Note: calculating the overhead so we can be compatible with
4083 * historical BSD practice is quite difficult in the face of
4084 * clusters/bigalloc. This is because multiple metadata blocks from
4085 * different block group can end up in the same allocation cluster.
4086 * Calculating the exact overhead in the face of clustered allocation
4087 * requires either O(all block bitmaps) in memory or O(number of block
4088 * groups**2) in time. We will still calculate the superblock for
4089 * older file systems --- and if we come across with a bigalloc file
4090 * system with zero in s_overhead_clusters the estimate will be close to
4091 * correct especially for very large cluster sizes --- but for newer
4092 * file systems, it's better to calculate this figure once at mkfs
4093 * time, and store it in the superblock. If the superblock value is
4094 * present (even for non-bigalloc file systems), we will use it.
4095 */
4096static int count_overhead(struct super_block *sb, ext4_group_t grp,
4097 char *buf)
4098{
4099 struct ext4_sb_info *sbi = EXT4_SB(sb);
4100 struct ext4_group_desc *gdp;
4101 ext4_fsblk_t first_block, last_block, b;
4102 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4103 int s, j, count = 0;
10b01ee9 4104 int has_super = ext4_bg_has_super(sb, grp);
952fc18e 4105
e2b911c5 4106 if (!ext4_has_feature_bigalloc(sb))
10b01ee9
TT		 4107 return (has_super + ext4_bg_num_gdb(sb, grp) +
4108 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
0548bbb8
TT		 4109 sbi->s_itb_per_group + 2);
4110
952fc18e
TT		 4111 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4112 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4113 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4114 for (i = 0; i < ngroups; i++) {
4115 gdp = ext4_get_group_desc(sb, i, NULL);
4116 b = ext4_block_bitmap(sb, gdp);
4117 if (b >= first_block && b <= last_block) {
4118 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4119 count++;
4120 }
4121 b = ext4_inode_bitmap(sb, gdp);
4122 if (b >= first_block && b <= last_block) {
4123 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4124 count++;
4125 }
4126 b = ext4_inode_table(sb, gdp);
4127 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4128 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4129 int c = EXT4_B2C(sbi, b - first_block);
4130 ext4_set_bit(c, buf);
4131 count++;
4132 }
4133 if (i != grp)
4134 continue;
4135 s = 0;
4136 if (ext4_bg_has_super(sb, grp)) {
4137 ext4_set_bit(s++, buf);
4138 count++;
4139 }
c48ae41b
TT		 4140 j = ext4_bg_num_gdb(sb, grp);
4141 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4142 ext4_error(sb, "Invalid number of block group "
4143 "descriptor blocks: %d", j);
4144 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 4145 }
c48ae41b
TT		 4146 count += j;
4147 for (; j > 0; j--)
4148 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 4149 }
4150 if (!count)
4151 return 0;
4152 return EXT4_CLUSTERS_PER_GROUP(sb) -
4153 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4154}
4155
4156/*
4157 * Compute the overhead and stash it in sbi->s_overhead
4158 */
4159int ext4_calculate_overhead(struct super_block *sb)
4160{
4161 struct ext4_sb_info *sbi = EXT4_SB(sb);
4162 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 4163 struct inode *j_inode;
4164 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 4165 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4166 ext4_fsblk_t overhead = 0;
4fdb5543 4167 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 4168
952fc18e
TT		 4169 if (!buf)
4170 return -ENOMEM;
4171
4172 /*
4173 * Compute the overhead (FS structures). This is constant
4174 * for a given filesystem unless the number of block groups
4175 * changes so we cache the previous value until it does.
4176 */
4177
4178 /*
4179 * All of the blocks before first_data_block are overhead
4180 */
4181 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4182
4183 /*
4184 * Add the overhead found in each block group
4185 */
4186 for (i = 0; i < ngroups; i++) {
4187 int blks;
4188
4189 blks = count_overhead(sb, i, buf);
4190 overhead += blks;
4191 if (blks)
4192 memset(buf, 0, PAGE_SIZE);
4193 cond_resched();
4194 }
3c816ded
EW		 4195
4196 /*
4197 * Add the internal journal blocks whether the journal has been
4198 * loaded or not
4199 */
ee7ed3aa 4200 if (sbi->s_journal && !sbi->s_journal_bdev)
ede7dc7f 4201 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
f1eec3b0
RH		 4202 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4203 /* j_inum for internal journal is non-zero */
3c816ded
EW		 4204 j_inode = ext4_get_journal_inode(sb, j_inum);
4205 if (j_inode) {
4206 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4207 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4208 iput(j_inode);
4209 } else {
4210 ext4_msg(sb, KERN_ERR, "can't get journal size");
4211 }
4212 }
952fc18e
TT		 4213 sbi->s_overhead = overhead;
4214 smp_wmb();
4215 free_page((unsigned long) buf);
4216 return 0;
4217}
4218
b5799018 4219static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 4220{
4221 ext4_fsblk_t resv_clusters;
b5799018 4222 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 4223
30fac0f7
JK		 4224 /*
4225 * There's no need to reserve anything when we aren't using extents.
4226 * The space estimates are exact, there are no unwritten extents,
4227 * hole punching doesn't need new metadata... This is needed especially
4228 * to keep ext2/3 backward compatibility.
4229 */
e2b911c5 4230 if (!ext4_has_feature_extents(sb))
b5799018 4231 return;
27dd4385
LC		 4232 /*
4233 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4234 * This should cover the situations where we can not afford to run
4235 * out of space like for example punch hole, or converting
556615dc 4236 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 4237 * allocation would require 1, or 2 blocks, higher numbers are
4238 * very rare.
4239 */
b5799018
TT		 4240 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4241 sbi->s_cluster_bits);
27dd4385
LC		 4242
4243 do_div(resv_clusters, 50);
4244 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4245
b5799018 4246 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 4247}
4248
ca9b404f
RA		 4249static const char *ext4_quota_mode(struct super_block *sb)
4250{
4251#ifdef CONFIG_QUOTA
4252 if (!ext4_quota_capable(sb))
4253 return "none";
4254
4255 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4256 return "journalled";
4257 else
4258 return "writeback";
4259#else
4260 return "disabled";
4261#endif
4262}
4263
188c299e
JK		 4264static void ext4_setup_csum_trigger(struct super_block *sb,
4265 enum ext4_journal_trigger_type type,
4266 void (*trigger)(
4267 struct jbd2_buffer_trigger_type *type,
4268 struct buffer_head *bh,
4269 void *mapped_data,
4270 size_t size))
4271{
4272 struct ext4_sb_info *sbi = EXT4_SB(sb);
4273
4274 sbi->s_journal_triggers[type].sb = sb;
4275 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4276}
4277
7edfd85b
LC		 4278static void ext4_free_sbi(struct ext4_sb_info *sbi)
4279{
4280 if (!sbi)
4281 return;
4282
4283 kfree(sbi->s_blockgroup_lock);
8012b866 4284 fs_put_dax(sbi->s_daxdev, NULL);
7edfd85b
LC		 4285 kfree(sbi);
4286}
4287
4288static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4289{
4290 struct ext4_sb_info *sbi;
4291
4292 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4293 if (!sbi)
4294 return NULL;
4295
8012b866
SR		 4296 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4297 NULL, NULL);
7edfd85b
LC		 4298
4299 sbi->s_blockgroup_lock =
4300 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4301
4302 if (!sbi->s_blockgroup_lock)
4303 goto err_out;
4304
4305 sb->s_fs_info = sbi;
4306 sbi->s_sb = sb;
4307 return sbi;
4308err_out:
8012b866 4309 fs_put_dax(sbi->s_daxdev, NULL);
7edfd85b
LC		 4310 kfree(sbi);
4311 return NULL;
4312}
4313
960e0ab6 4314static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
ac27a0ec 4315{
1d0c3924 4316 struct buffer_head *bh, **group_desc;
617ba13b 4317 struct ext4_super_block *es = NULL;
7edfd85b 4318 struct ext4_sb_info *sbi = EXT4_SB(sb);
7c990728 4319 struct flex_groups **flex_groups;
617ba13b 4320 ext4_fsblk_t block;
70bbb3e0 4321 ext4_fsblk_t logical_sb_block;
ac27a0ec 4322 unsigned long offset = 0;
ac27a0ec
DK		 4323 unsigned long def_mount_opts;
4324 struct inode *root;
dcc7dae3 4325 int ret = -ENOMEM;
281b5995 4326 int blocksize, clustersize;
4ec11028
TT		 4327 unsigned int db_count;
4328 unsigned int i;
ef5fd681 4329 int needs_recovery, has_huge_files;
bd81d8ee 4330 __u64 blocks_count;
07aa2ea1 4331 int err = 0;
bfff6873 4332 ext4_group_t first_not_zeroed;
7edfd85b 4333 struct ext4_fs_context *ctx = fc->fs_private;
960e0ab6 4334 int silent = fc->sb_flags & SB_SILENT;
b237e304
HS		 4335
4336 /* Set defaults for the variables that will be set during parsing */
e4e58e5d
OM		 4337 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
4338 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
705895b6 4339
240799cd 4340 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
8446fe92
CH		 4341 sbi->s_sectors_written_start =
4342 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
ac27a0ec 4343
07aa2ea1 4344 /* -EINVAL is default */
dcc7dae3 4345 ret = -EINVAL;
617ba13b 4346 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 4347 if (!blocksize) {
b31e1552 4348 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 4349 goto out_fail;
4350 }
4351
4352 /*
617ba13b 4353 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 4354 * block sizes. We need to calculate the offset from buffer start.
4355 */
617ba13b 4356 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
7edfd85b 4357 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
70bbb3e0 4358 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 4359 } else {
7edfd85b 4360 logical_sb_block = sbi->s_sb_block;
ac27a0ec
DK		 4361 }
4362
8394a6ab 4363 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4364 if (IS_ERR(bh)) {
b31e1552 4365 ext4_msg(sb, KERN_ERR, "unable to read superblock");
8394a6ab 4366 ret = PTR_ERR(bh);
ac27a0ec
DK		 4367 goto out_fail;
4368 }
4369 /*
4370 * Note: s_es must be initialized as soon as possible because
617ba13b 4371 * some ext4 macro-instructions depend on its value
ac27a0ec 4372 */
2716b802 4373 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 4374 sbi->s_es = es;
4375 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 4376 if (sb->s_magic != EXT4_SUPER_MAGIC)
4377 goto cantfind_ext4;
afc32f7e 4378 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 4379
feb0ab32 4380 /* Warn if metadata_csum and gdt_csum are both set. */
e2b911c5
DW		 4381 if (ext4_has_feature_metadata_csum(sb) &&
4382 ext4_has_feature_gdt_csum(sb))
363307e6 4383 ext4_warning(sb, "metadata_csum and uninit_bg are "
feb0ab32
DW		 4384 "redundant flags; please run fsck.");
4385
d25425f8
DW		 4386 /* Check for a known checksum algorithm */
4387 if (!ext4_verify_csum_type(sb, es)) {
4388 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4389 "unknown checksum algorithm.");
4390 silent = 1;
4391 goto cantfind_ext4;
4392 }
02f310fc
JK		 4393 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4394 ext4_orphan_file_block_trigger);
d25425f8 4395
0441984a 4396 /* Load the checksum driver */
a45403b5
TT		 4397 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4398 if (IS_ERR(sbi->s_chksum_driver)) {
4399 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4400 ret = PTR_ERR(sbi->s_chksum_driver);
4401 sbi->s_chksum_driver = NULL;
4402 goto failed_mount;
0441984a
DW		 4403 }
4404
a9c47317
DW		 4405 /* Check superblock checksum */
4406 if (!ext4_superblock_csum_verify(sb, es)) {
4407 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4408 "invalid superblock checksum. Run e2fsck?");
4409 silent = 1;
6a797d27 4410 ret = -EFSBADCRC;
a9c47317
DW		 4411 goto cantfind_ext4;
4412 }
4413
4414 /* Precompute checksum seed for all metadata */
e2b911c5 4415 if (ext4_has_feature_csum_seed(sb))
8c81bd8f 4416 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
dec214d0 4417 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
a9c47317
DW		 4418 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4419 sizeof(es->s_uuid));
4420
ac27a0ec
DK		 4421 /* Set defaults before we parse the mount options */
4422 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 4423 set_opt(sb, INIT_INODE_TABLE);
617ba13b 4424 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 4425 set_opt(sb, DEBUG);
87f26807 4426 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 4427 set_opt(sb, GRPID);
617ba13b 4428 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 4429 set_opt(sb, NO_UID32);
ea663336 4430 /* xattr user namespace & acls are now defaulted on */
ea663336 4431 set_opt(sb, XATTR_USER);
03010a33 4432#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 4433 set_opt(sb, POSIX_ACL);
2e7842b8 4434#endif
995a3ed6
HS		 4435 if (ext4_has_feature_fast_commit(sb))
4436 set_opt2(sb, JOURNAL_FAST_COMMIT);
98c1a759
DW		 4437 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4438 if (ext4_has_metadata_csum(sb))
4439 set_opt(sb, JOURNAL_CHECKSUM);
4440
617ba13b 4441 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 4442 set_opt(sb, JOURNAL_DATA);
617ba13b 4443 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 4444 set_opt(sb, ORDERED_DATA);
617ba13b 4445 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 4446 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 4447
4448 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 4449 set_opt(sb, ERRORS_PANIC);
bb4f397a 4450 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 4451 set_opt(sb, ERRORS_CONT);
bb4f397a 4452 else
fd8c37ec 4453 set_opt(sb, ERRORS_RO);
45f1a9c3
DW		 4454 /* block_validity enabled by default; disable with noblock_validity */
4455 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 4456 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 4457 set_opt(sb, DISCARD);
ac27a0ec 4458
08cefc7a
EB		 4459 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4460 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 4461 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4462 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4463 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 4464
8b67f04a 4465 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 4466 set_opt(sb, BARRIER);
ac27a0ec 4467
dd919b98
AK		 4468 /*
4469 * enable delayed allocation by default
4470 * Use -o nodelalloc to turn it off
4471 */
bc0b75f7 4472 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 4473 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 4474 set_opt(sb, DELALLOC);
dd919b98 4475
51ce6511
LC		 4476 /*
4477 * set default s_li_wait_mult for lazyinit, for the case there is
4478 * no mount option specified.
4479 */
4480 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4481
c9200760
TT		 4482 if (le32_to_cpu(es->s_log_block_size) >
4483 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4f97a681 4484 ext4_msg(sb, KERN_ERR,
c9200760
TT		 4485 "Invalid log block size: %u",
4486 le32_to_cpu(es->s_log_block_size));
4f97a681
TT		 4487 goto failed_mount;
4488 }
c9200760
TT		 4489 if (le32_to_cpu(es->s_log_cluster_size) >
4490 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4f97a681 4491 ext4_msg(sb, KERN_ERR,
c9200760
TT		 4492 "Invalid log cluster size: %u",
4493 le32_to_cpu(es->s_log_cluster_size));
4f97a681
TT		 4494 goto failed_mount;
4495 }
4496
c9200760
TT		 4497 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4498
4499 if (blocksize == PAGE_SIZE)
4500 set_opt(sb, DIOREAD_NOLOCK);
4f97a681 4501
9803387c
TT		 4502 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4503 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4504 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4505 } else {
4506 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4507 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4508 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4509 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4510 sbi->s_first_ino);
4511 goto failed_mount;
4512 }
4513 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4514 (!is_power_of_2(sbi->s_inode_size)) ||
4515 (sbi->s_inode_size > blocksize)) {
4516 ext4_msg(sb, KERN_ERR,
4517 "unsupported inode size: %d",
4518 sbi->s_inode_size);
4f97a681 4519 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
9803387c
TT		 4520 goto failed_mount;
4521 }
4522 /*
4523 * i_atime_extra is the last extra field available for
4524 * [acm]times in struct ext4_inode. Checking for that
4525 * field should suffice to ensure we have extra space
4526 * for all three.
4527 */
4528 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4529 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4530 sb->s_time_gran = 1;
4531 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4532 } else {
4533 sb->s_time_gran = NSEC_PER_SEC;
4534 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4535 }
4536 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4537 }
4538 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4539 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4540 EXT4_GOOD_OLD_INODE_SIZE;
4541 if (ext4_has_feature_extra_isize(sb)) {
4542 unsigned v, max = (sbi->s_inode_size -
4543 EXT4_GOOD_OLD_INODE_SIZE);
4544
4545 v = le16_to_cpu(es->s_want_extra_isize);
4546 if (v > max) {
4547 ext4_msg(sb, KERN_ERR,
4548 "bad s_want_extra_isize: %d", v);
4549 goto failed_mount;
4550 }
4551 if (sbi->s_want_extra_isize < v)
4552 sbi->s_want_extra_isize = v;
4553
4554 v = le16_to_cpu(es->s_min_extra_isize);
4555 if (v > max) {
4556 ext4_msg(sb, KERN_ERR,
4557 "bad s_min_extra_isize: %d", v);
4558 goto failed_mount;
4559 }
4560 if (sbi->s_want_extra_isize < v)
4561 sbi->s_want_extra_isize = v;
4562 }
4563 }
4564
7edfd85b
LC		 4565 err = parse_apply_sb_mount_options(sb, ctx);
4566 if (err < 0)
4567 goto failed_mount;
4568
5a916be1 4569 sbi->s_def_mount_opt = sbi->s_mount_opt;
7edfd85b
LC		 4570
4571 err = ext4_check_opt_consistency(fc, sb);
4572 if (err < 0)
4573 goto failed_mount;
4574
85456054 4575 ext4_apply_options(fc, sb);
ac27a0ec 4576
5298d4bf 4577#if IS_ENABLED(CONFIG_UNICODE)
f8f4acb6 4578 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
c83ad55e
GKB		 4579 const struct ext4_sb_encodings *encoding_info;
4580 struct unicode_map *encoding;
aa8bf298 4581 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
c83ad55e 4582
aa8bf298
CH		 4583 encoding_info = ext4_sb_read_encoding(es);
4584 if (!encoding_info) {
c83ad55e
GKB		 4585 ext4_msg(sb, KERN_ERR,
4586 "Encoding requested by superblock is unknown");
4587 goto failed_mount;
4588 }
4589
4590 encoding = utf8_load(encoding_info->version);
4591 if (IS_ERR(encoding)) {
4592 ext4_msg(sb, KERN_ERR,
49bd03cc 4593 "can't mount with superblock charset: %s-%u.%u.%u "
c83ad55e 4594 "not supported by the kernel. flags: 0x%x.",
49bd03cc
CH		 4595 encoding_info->name,
4596 unicode_major(encoding_info->version),
4597 unicode_minor(encoding_info->version),
4598 unicode_rev(encoding_info->version),
c83ad55e
GKB		 4599 encoding_flags);
4600 goto failed_mount;
4601 }
4602 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
49bd03cc
CH		 4603 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4604 unicode_major(encoding_info->version),
4605 unicode_minor(encoding_info->version),
4606 unicode_rev(encoding_info->version),
4607 encoding_flags);
c83ad55e 4608
f8f4acb6
DR		 4609 sb->s_encoding = encoding;
4610 sb->s_encoding_flags = encoding_flags;
c83ad55e
GKB		 4611 }
4612#endif
4613
56889787 4614 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
556e0319 4615 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
781c036b 4616 /* can't mount with both data=journal and dioread_nolock. */
244adf64 4617 clear_opt(sb, DIOREAD_NOLOCK);
556e0319 4618 clear_opt2(sb, JOURNAL_FAST_COMMIT);
56889787
TT		 4619 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4620 ext4_msg(sb, KERN_ERR, "can't mount with "
4621 "both data=journal and delalloc");
4622 goto failed_mount;
4623 }
fc626fe3 4624 if (test_opt(sb, DAX_ALWAYS)) {
923ae0ff
RZ		 4625 ext4_msg(sb, KERN_ERR, "can't mount with "
4626 "both data=journal and dax");
4627 goto failed_mount;
4628 }
73b92a2a
SK		 4629 if (ext4_has_feature_encrypt(sb)) {
4630 ext4_msg(sb, KERN_WARNING,
4631 "encrypted files will use data=ordered "
4632 "instead of data journaling mode");
4633 }
56889787
TT		 4634 if (test_opt(sb, DELALLOC))
4635 clear_opt(sb, DELALLOC);
001e4a87
TH		 4636 } else {
4637 sb->s_iflags |= SB_I_CGROUPWB;
56889787
TT		 4638 }
4639
1751e8a6
LT		 4640 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4641 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 4642
617ba13b 4643 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
e2b911c5
DW		 4644 (ext4_has_compat_features(sb) ||
4645 ext4_has_ro_compat_features(sb) ||
4646 ext4_has_incompat_features(sb)))
b31e1552
ES		 4647 ext4_msg(sb, KERN_WARNING,
4648 "feature flags set on rev 0 fs, "
4649 "running e2fsck is recommended");
469108ff 4650
ed3654eb
TT		 4651 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4652 set_opt2(sb, HURD_COMPAT);
e2b911c5 4653 if (ext4_has_feature_64bit(sb)) {
ed3654eb
TT		 4654 ext4_msg(sb, KERN_ERR,
4655 "The Hurd can't support 64-bit file systems");
4656 goto failed_mount;
4657 }
dec214d0
TE		 4658
4659 /*
4660 * ea_inode feature uses l_i_version field which is not
4661 * available in HURD_COMPAT mode.
4662 */
4663 if (ext4_has_feature_ea_inode(sb)) {
4664 ext4_msg(sb, KERN_ERR,
4665 "ea_inode feature is not supported for Hurd");
4666 goto failed_mount;
4667 }
ed3654eb
TT		 4668 }
4669
2035e776
TT		 4670 if (IS_EXT2_SB(sb)) {
4671 if (ext2_feature_set_ok(sb))
4672 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4673 "using the ext4 subsystem");
4674 else {
0d9366d6
ES		 4675 /*
4676 * If we're probing be silent, if this looks like
4677 * it's actually an ext[34] filesystem.
4678 */
4679 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4680 goto failed_mount;
2035e776
TT		 4681 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4682 "to feature incompatibilities");
4683 goto failed_mount;
4684 }
4685 }
4686
4687 if (IS_EXT3_SB(sb)) {
4688 if (ext3_feature_set_ok(sb))
4689 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4690 "using the ext4 subsystem");
4691 else {
0d9366d6
ES		 4692 /*
4693 * If we're probing be silent, if this looks like
4694 * it's actually an ext4 filesystem.
4695 */
4696 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4697 goto failed_mount;
2035e776
TT		 4698 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4699 "to feature incompatibilities");
4700 goto failed_mount;
4701 }
4702 }
4703
ac27a0ec
DK		 4704 /*
4705 * Check feature flags regardless of the revision level, since we
4706 * previously didn't change the revision level when setting the flags,
4707 * so there is a chance incompat flags are set on a rev 0 filesystem.
4708 */
bc98a42c 4709 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
ac27a0ec 4710 goto failed_mount;
a13fb1a4 4711
5b9554dc
TT		 4712 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4713 ext4_msg(sb, KERN_ERR,
4714 "Number of reserved GDT blocks insanely large: %d",
4715 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4716 goto failed_mount;
4717 }
4718
89b93a7b 4719 if (sbi->s_daxdev) {
7b0800d0
CH		 4720 if (blocksize == PAGE_SIZE)
4721 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4722 else
4723 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
4724 }
a8ab6d38 4725
fc626fe3 4726 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
559db4c6
RZ		 4727 if (ext4_has_feature_inline_data(sb)) {
4728 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4729 " that may contain inline data");
361d24d4 4730 goto failed_mount;
559db4c6 4731 }
a8ab6d38 4732 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
24f3478d 4733 ext4_msg(sb, KERN_ERR,
361d24d4
ES		 4734 "DAX unsupported by block device.");
4735 goto failed_mount;
24f3478d 4736 }
923ae0ff
RZ		 4737 }
4738
e2b911c5 4739 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 4740 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4741 es->s_encryption_level);
4742 goto failed_mount;
4743 }
4744
ac27a0ec 4745 if (sb->s_blocksize != blocksize) {
afd09b61
AM		 4746 /*
4747 * bh must be released before kill_bdev(), otherwise
4748 * it won't be freed and its page also. kill_bdev()
4749 * is called by sb_set_blocksize().
4750 */
4751 brelse(bh);
ce40733c
AK		 4752 /* Validate the filesystem blocksize */
4753 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 4754 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 4755 blocksize);
afd09b61 4756 bh = NULL;
ac27a0ec
DK		 4757 goto failed_mount;
4758 }
4759
7edfd85b 4760 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
70bbb3e0 4761 offset = do_div(logical_sb_block, blocksize);
8394a6ab 4762 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4763 if (IS_ERR(bh)) {
b31e1552
ES		 4764 ext4_msg(sb, KERN_ERR,
4765 "Can't read superblock on 2nd try");
8394a6ab 4766 ret = PTR_ERR(bh);
4767 bh = NULL;
ac27a0ec
DK		 4768 goto failed_mount;
4769 }
2716b802 4770 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 4771 sbi->s_es = es;
617ba13b 4772 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 4773 ext4_msg(sb, KERN_ERR,
4774 "Magic mismatch, very weird!");
ac27a0ec
DK		 4775 goto failed_mount;
4776 }
4777 }
4778
e2b911c5 4779 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 4780 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4781 has_huge_files);
4782 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 4783
0d1ee42f 4784 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 4785 if (ext4_has_feature_64bit(sb)) {
8fadc143 4786 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 4787 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 4788 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 4789 ext4_msg(sb, KERN_ERR,
4790 "unsupported descriptor size %lu",
0d1ee42f
AR		 4791 sbi->s_desc_size);
4792 goto failed_mount;
4793 }
4794 } else
4795 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 4796
ac27a0ec 4797 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 4798 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 4799
617ba13b 4800 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 4801 if (sbi->s_inodes_per_block == 0)
617ba13b 4802 goto cantfind_ext4;
cd6bb35b
TT		 4803 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4804 sbi->s_inodes_per_group > blocksize * 8) {
4805 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
b9c538da 4806 sbi->s_inodes_per_group);
cd6bb35b
TT		 4807 goto failed_mount;
4808 }
ac27a0ec
DK		 4809 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4810 sbi->s_inodes_per_block;
0d1ee42f 4811 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec 4812 sbi->s_sbh = bh;
c878bea3 4813 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
e57aa839
FW		 4814 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4815 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 4816
2b2d6d01 4817 for (i = 0; i < 4; i++)
ac27a0ec
DK		 4818 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4819 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 4820 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 4821 i = le32_to_cpu(es->s_flags);
4822 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4823 sbi->s_hash_unsigned = 3;
4824 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 4825#ifdef __CHAR_UNSIGNED__
bc98a42c 4826 if (!sb_rdonly(sb))
23301410
TT		 4827 es->s_flags |=
4828 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4829 sbi->s_hash_unsigned = 3;
f99b2589 4830#else
bc98a42c 4831 if (!sb_rdonly(sb))
23301410
TT		 4832 es->s_flags |=
4833 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 4834#endif
23301410 4835 }
f99b2589 4836 }
ac27a0ec 4837
281b5995
TT		 4838 /* Handle clustersize */
4839 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
ef5fd681 4840 if (ext4_has_feature_bigalloc(sb)) {
281b5995
TT		 4841 if (clustersize < blocksize) {
4842 ext4_msg(sb, KERN_ERR,
4843 "cluster size (%d) smaller than "
4844 "block size (%d)", clustersize, blocksize);
4845 goto failed_mount;
4846 }
4847 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4848 le32_to_cpu(es->s_log_block_size);
4849 sbi->s_clusters_per_group =
4850 le32_to_cpu(es->s_clusters_per_group);
4851 if (sbi->s_clusters_per_group > blocksize * 8) {
4852 ext4_msg(sb, KERN_ERR,
4853 "#clusters per group too big: %lu",
4854 sbi->s_clusters_per_group);
4855 goto failed_mount;
4856 }
4857 if (sbi->s_blocks_per_group !=
4858 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4859 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4860 "clusters per group (%lu) inconsistent",
4861 sbi->s_blocks_per_group,
4862 sbi->s_clusters_per_group);
4863 goto failed_mount;
4864 }
4865 } else {
4866 if (clustersize != blocksize) {
bfe0a5f4
TT		 4867 ext4_msg(sb, KERN_ERR,
4868 "fragment/cluster size (%d) != "
4869 "block size (%d)", clustersize, blocksize);
4870 goto failed_mount;
281b5995
TT		 4871 }
4872 if (sbi->s_blocks_per_group > blocksize * 8) {
4873 ext4_msg(sb, KERN_ERR,
4874 "#blocks per group too big: %lu",
4875 sbi->s_blocks_per_group);
4876 goto failed_mount;
4877 }
4878 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4879 sbi->s_cluster_bits = 0;
ac27a0ec 4880 }
281b5995
TT		 4881 sbi->s_cluster_ratio = clustersize / blocksize;
4882
960fd856
TT		 4883 /* Do we have standard group size of clustersize * 8 blocks ? */
4884 if (sbi->s_blocks_per_group == clustersize << 3)
4885 set_opt2(sb, STD_GROUP_SIZE);
4886
bf43d84b
ES		 4887 /*
4888 * Test whether we have more sectors than will fit in sector_t,
4889 * and whether the max offset is addressable by the page cache.
4890 */
5a9ae68a 4891 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 4892 ext4_blocks_count(es));
5a9ae68a 4893 if (err) {
b31e1552 4894 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 4895 " too large to mount safely on this system");
ac27a0ec
DK		 4896 goto failed_mount;
4897 }
4898
617ba13b
MC		 4899 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4900 goto cantfind_ext4;
e7c95593 4901
0f2ddca6 4902 /* check blocks count against device size */
5513b241 4903 blocks_count = sb_bdev_nr_blocks(sb);
0f2ddca6 4904 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 4905 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4906 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 4907 ext4_blocks_count(es), blocks_count);
4908 goto failed_mount;
4909 }
4910
0b8e58a1
AD		 4911 /*
4912 * It makes no sense for the first data block to be beyond the end
4913 * of the filesystem.
4914 */
4915 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 4916 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 4917 "block %u is beyond end of filesystem (%llu)",
4918 le32_to_cpu(es->s_first_data_block),
4919 ext4_blocks_count(es));
e7c95593
ES		 4920 goto failed_mount;
4921 }
bfe0a5f4
TT		 4922 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4923 (sbi->s_cluster_ratio == 1)) {
4924 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4925 "block is 0 with a 1k block and cluster size");
4926 goto failed_mount;
4927 }
4928
bd81d8ee
LV		 4929 blocks_count = (ext4_blocks_count(es) -
4930 le32_to_cpu(es->s_first_data_block) +
4931 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4932 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 4933 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
df41460a 4934 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4ec11028 4935 "(block count %llu, first data block %u, "
df41460a 4936 "blocks per group %lu)", blocks_count,
4ec11028
TT		 4937 ext4_blocks_count(es),
4938 le32_to_cpu(es->s_first_data_block),
4939 EXT4_BLOCKS_PER_GROUP(sb));
4940 goto failed_mount;
4941 }
bd81d8ee 4942 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 4943 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4944 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
9e463084
TT		 4945 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4946 le32_to_cpu(es->s_inodes_count)) {
4947 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4948 le32_to_cpu(es->s_inodes_count),
4949 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4950 ret = -EINVAL;
4951 goto failed_mount;
4952 }
617ba13b
MC		 4953 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4954 EXT4_DESC_PER_BLOCK(sb);
3a4b77cd 4955 if (ext4_has_feature_meta_bg(sb)) {
2ba3e6e8 4956 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3a4b77cd
EG		 4957 ext4_msg(sb, KERN_WARNING,
4958 "first meta block group too large: %u "
4959 "(group descriptor block count %u)",
4960 le32_to_cpu(es->s_first_meta_bg), db_count);
4961 goto failed_mount;
4962 }
4963 }
1d0c3924
TT		 4964 rcu_assign_pointer(sbi->s_group_desc,
4965 kvmalloc_array(db_count,
4966 sizeof(struct buffer_head *),
4967 GFP_KERNEL));
ac27a0ec 4968 if (sbi->s_group_desc == NULL) {
b31e1552 4969 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 4970 ret = -ENOMEM;
ac27a0ec
DK		 4971 goto failed_mount;
4972 }
4973
705895b6 4974 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec 4975
85c8f176
AP		 4976 /* Pre-read the descriptors into the buffer cache */
4977 for (i = 0; i < db_count; i++) {
4978 block = descriptor_loc(sb, logical_sb_block, i);
5df1d412 4979 ext4_sb_breadahead_unmovable(sb, block);
85c8f176
AP		 4980 }
4981
ac27a0ec 4982 for (i = 0; i < db_count; i++) {
1d0c3924
TT		 4983 struct buffer_head *bh;
4984
70bbb3e0 4985 block = descriptor_loc(sb, logical_sb_block, i);
8394a6ab 4986 bh = ext4_sb_bread_unmovable(sb, block);
4987 if (IS_ERR(bh)) {
b31e1552
ES		 4988 ext4_msg(sb, KERN_ERR,
4989 "can't read group descriptor %d", i);
ac27a0ec 4990 db_count = i;
8394a6ab 4991 ret = PTR_ERR(bh);
ac27a0ec
DK		 4992 goto failed_mount2;
4993 }
1d0c3924
TT		 4994 rcu_read_lock();
4995 rcu_dereference(sbi->s_group_desc)[i] = bh;
4996 rcu_read_unlock();
ac27a0ec 4997 }
44de022c 4998 sbi->s_gdb_count = db_count;
829fa70d 4999 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
b31e1552 5000 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
6a797d27 5001 ret = -EFSCORRUPTED;
f9ae9cf5 5002 goto failed_mount2;
ac27a0ec 5003 }
772cb7c8 5004
235699a8 5005 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
c92dc856
JK		 5006 spin_lock_init(&sbi->s_error_lock);
5007 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
04496411 5008
a75ae78f 5009 /* Register extent status tree shrinker */
eb68d0e2 5010 if (ext4_es_register_shrinker(sbi))
ce7e010a 5011 goto failed_mount3;
ce7e010a 5012
c9de560d 5013 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 5014 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 5015
f9ae9cf5
TT		 5016 /*
5017 * set up enough so that it can read an inode
5018 */
f6e63f90 5019 sb->s_op = &ext4_sops;
617ba13b
MC		 5020 sb->s_export_op = &ext4_export_ops;
5021 sb->s_xattr = ext4_xattr_handlers;
643fa961 5022#ifdef CONFIG_FS_ENCRYPTION
a7550b30 5023 sb->s_cop = &ext4_cryptops;
ffcc4182 5024#endif
c93d8f88
EB		 5025#ifdef CONFIG_FS_VERITY
5026 sb->s_vop = &ext4_verityops;
5027#endif
ac27a0ec 5028#ifdef CONFIG_QUOTA
617ba13b 5029 sb->dq_op = &ext4_quota_operations;
e2b911c5 5030 if (ext4_has_feature_quota(sb))
1fa5efe3 5031 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 5032 else
5033 sb->s_qcop = &ext4_qctl_operations;
689c958c 5034 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 5035#endif
85787090 5036 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 5037
ac27a0ec 5038 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 5039 mutex_init(&sbi->s_orphan_lock);
ac27a0ec 5040
aa75f4d3
HS		 5041 /* Initialize fast commit stuff */
5042 atomic_set(&sbi->s_fc_subtid, 0);
aa75f4d3
HS		 5043 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
5044 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
5045 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
5046 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
5047 sbi->s_fc_bytes = 0;
9b5f6c9b 5048 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
e85c81ba 5049 sbi->s_fc_ineligible_tid = 0;
aa75f4d3
HS		 5050 spin_lock_init(&sbi->s_fc_lock);
5051 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
8016e29f
HS		 5052 sbi->s_fc_replay_state.fc_regions = NULL;
5053 sbi->s_fc_replay_state.fc_regions_size = 0;
5054 sbi->s_fc_replay_state.fc_regions_used = 0;
5055 sbi->s_fc_replay_state.fc_regions_valid = 0;
5056 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
5057 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
5058 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
aa75f4d3 5059
ac27a0ec
DK		 5060 sb->s_root = NULL;
5061
5062 needs_recovery = (es->s_last_orphan != 0 ||
02f310fc 5063 ext4_has_feature_orphan_present(sb) ||
e2b911c5 5064 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 5065
bc98a42c 5066 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 5067 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 5068 goto failed_mount3a;
c5e06d10 5069
ac27a0ec
DK		 5070 /*
5071 * The first inode we look at is the journal inode. Don't try
5072 * root first: it may be modified in the journal!
5073 */
e2b911c5 5074 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
7edfd85b 5075 err = ext4_load_journal(sb, es, ctx->journal_devnum);
4753d8a2 5076 if (err)
50460fe8 5077 goto failed_mount3a;
bc98a42c 5078 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 5079 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 5080 ext4_msg(sb, KERN_ERR, "required journal recovery "
5081 "suppressed and not mounted read-only");
744692dc 5082 goto failed_mount_wq;
ac27a0ec 5083 } else {
1e381f60
DM		 5084 /* Nojournal mode, all journal mount options are illegal */
5085 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
5086 ext4_msg(sb, KERN_ERR, "can't mount with "
5087 "journal_checksum, fs mounted w/o journal");
5088 goto failed_mount_wq;
5089 }
5090 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5091 ext4_msg(sb, KERN_ERR, "can't mount with "
5092 "journal_async_commit, fs mounted w/o journal");
5093 goto failed_mount_wq;
5094 }
5095 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5096 ext4_msg(sb, KERN_ERR, "can't mount with "
5097 "commit=%lu, fs mounted w/o journal",
5098 sbi->s_commit_interval / HZ);
5099 goto failed_mount_wq;
5100 }
5101 if (EXT4_MOUNT_DATA_FLAGS &
5102 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5103 ext4_msg(sb, KERN_ERR, "can't mount with "
5104 "data=, fs mounted w/o journal");
5105 goto failed_mount_wq;
5106 }
50b29d8f 5107 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
1e381f60 5108 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 5109 clear_opt(sb, DATA_FLAGS);
995a3ed6 5110 clear_opt2(sb, JOURNAL_FAST_COMMIT);
0390131b
FM		 5111 sbi->s_journal = NULL;
5112 needs_recovery = 0;
5113 goto no_journal;
ac27a0ec
DK		 5114 }
5115
e2b911c5 5116 if (ext4_has_feature_64bit(sb) &&
eb40a09c
JS		 5117 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
5118 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 5119 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 5120 goto failed_mount_wq;
eb40a09c
JS		 5121 }
5122
25ed6e8a
DW		 5123 if (!set_journal_csum_feature_set(sb)) {
5124 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
5125 "feature set");
5126 goto failed_mount_wq;
d4da6c9c 5127 }
818d276c 5128
a1e5e465
HS		 5129 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
5130 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
5131 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
5132 ext4_msg(sb, KERN_ERR,
5133 "Failed to set fast commit journal feature");
5134 goto failed_mount_wq;
5135 }
5136
ac27a0ec
DK		 5137 /* We have now updated the journal if required, so we can
5138 * validate the data journaling mode. */
5139 switch (test_opt(sb, DATA_FLAGS)) {
5140 case 0:
5141 /* No mode set, assume a default based on the journal
63f57933
AM		 5142 * capabilities: ORDERED_DATA if the journal can
5143 * cope, else JOURNAL_DATA
5144 */
dab291af 5145 if (jbd2_journal_check_available_features
27f394a7 5146 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
fd8c37ec 5147 set_opt(sb, ORDERED_DATA);
27f394a7
TN		 5148 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
5149 } else {
fd8c37ec 5150 set_opt(sb, JOURNAL_DATA);
27f394a7
TN		 5151 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
5152 }
ac27a0ec
DK		 5153 break;
5154
617ba13b
MC		 5155 case EXT4_MOUNT_ORDERED_DATA:
5156 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 5157 if (!jbd2_journal_check_available_features
5158 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 5159 ext4_msg(sb, KERN_ERR, "Journal does not support "
5160 "requested data journaling mode");
744692dc 5161 goto failed_mount_wq;
ac27a0ec 5162 }
5a150bde 5163 break;
ac27a0ec
DK		 5164 default:
5165 break;
5166 }
ab04df78
JK		 5167
5168 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
5169 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5170 ext4_msg(sb, KERN_ERR, "can't mount with "
5171 "journal_async_commit in data=ordered mode");
5172 goto failed_mount_wq;
5173 }
5174
7edfd85b 5175 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
ac27a0ec 5176
342af94e 5177 sbi->s_journal->j_submit_inode_data_buffers =
afb585a9 5178 ext4_journal_submit_inode_data_buffers;
342af94e 5179 sbi->s_journal->j_finish_inode_data_buffers =
afb585a9 5180 ext4_journal_finish_inode_data_buffers;
18aadd47 5181
ce7e010a 5182no_journal:
cdb7ee4c
TE		 5183 if (!test_opt(sb, NO_MBCACHE)) {
5184 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5185 if (!sbi->s_ea_block_cache) {
dec214d0 5186 ext4_msg(sb, KERN_ERR,
cdb7ee4c 5187 "Failed to create ea_block_cache");
dec214d0
TE		 5188 goto failed_mount_wq;
5189 }
cdb7ee4c
TE		 5190
5191 if (ext4_has_feature_ea_inode(sb)) {
5192 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5193 if (!sbi->s_ea_inode_cache) {
5194 ext4_msg(sb, KERN_ERR,
5195 "Failed to create ea_inode_cache");
5196 goto failed_mount_wq;
5197 }
5198 }
9c191f70
M		 5199 }
5200
c93d8f88
EB		 5201 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
5202 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
5203 goto failed_mount_wq;
5204 }
5205
952fc18e
TT		 5206 /*
5207 * Get the # of file system overhead blocks from the
5208 * superblock if present.
5209 */
85d825db
TT		 5210 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5211 /* ignore the precalculated value if it is ridiculous */
5212 if (sbi->s_overhead > ext4_blocks_count(es))
5213 sbi->s_overhead = 0;
5214 /*
5215 * If the bigalloc feature is not enabled recalculating the
5216 * overhead doesn't take long, so we might as well just redo
5217 * it to make sure we are using the correct value.
5218 */
5219 if (!ext4_has_feature_bigalloc(sb))
5220 sbi->s_overhead = 0;
5221 if (sbi->s_overhead == 0) {
07aa2ea1
LC		 5222 err = ext4_calculate_overhead(sb);
5223 if (err)
952fc18e
TT		 5224 goto failed_mount_wq;
5225 }
5226
fd89d5f2
TH		 5227 /*
5228 * The maximum number of concurrent works can be high and
5229 * concurrency isn't really necessary. Limit it to 1.
5230 */
2e8fa54e
JK		 5231 EXT4_SB(sb)->rsv_conversion_wq =
5232 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5233 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5234 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 5235 ret = -ENOMEM;
2e8fa54e
JK		 5236 goto failed_mount4;
5237 }
5238
ac27a0ec 5239 /*
dab291af 5240 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 5241 * so we can safely mount the rest of the filesystem now.
5242 */
5243
8a363970 5244 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
1d1fe1ee 5245 if (IS_ERR(root)) {
b31e1552 5246 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 5247 ret = PTR_ERR(root);
32a9bb57 5248 root = NULL;
ac27a0ec
DK		 5249 goto failed_mount4;
5250 }
5251 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 5252 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 5253 iput(root);
ac27a0ec
DK		 5254 goto failed_mount4;
5255 }
b886ee3e 5256
48fde701 5257 sb->s_root = d_make_root(root);
1d1fe1ee 5258 if (!sb->s_root) {
b31e1552 5259 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 5260 ret = -ENOMEM;
5261 goto failed_mount4;
5262 }
ac27a0ec 5263
c89128a0
JK		 5264 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
5265 if (ret == -EROFS) {
1751e8a6 5266 sb->s_flags |= SB_RDONLY;
c89128a0
JK		 5267 ret = 0;
5268 } else if (ret)
5269 goto failed_mount4a;
ef7f3835 5270
b5799018 5271 ext4_set_resv_clusters(sb);
27dd4385 5272
0f5bde1d
JK		 5273 if (test_opt(sb, BLOCK_VALIDITY)) {
5274 err = ext4_setup_system_zone(sb);
5275 if (err) {
5276 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5277 "zone (%d)", err);
5278 goto failed_mount4a;
5279 }
f9ae9cf5 5280 }
8016e29f 5281 ext4_fc_replay_cleanup(sb);
f9ae9cf5
TT		 5282
5283 ext4_ext_init(sb);
196e402a
HS		 5284
5285 /*
5286 * Enable optimize_scan if number of groups is > threshold. This can be
5287 * turned off by passing "mb_optimize_scan=0". This can also be
5288 * turned on forcefully by passing "mb_optimize_scan=1".
5289 */
27b38686
OM		 5290 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5291 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5292 set_opt2(sb, MB_OPTIMIZE_SCAN);
5293 else
5294 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5295 }
196e402a 5296
f9ae9cf5
TT		 5297 err = ext4_mb_init(sb);
5298 if (err) {
5299 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5300 err);
dcf2d804 5301 goto failed_mount5;
c2774d84
AK		 5302 }
5303
027f14f5
TT		 5304 /*
5305 * We can only set up the journal commit callback once
5306 * mballoc is initialized
5307 */
5308 if (sbi->s_journal)
5309 sbi->s_journal->j_commit_callback =
5310 ext4_journal_commit_callback;
5311
d5e03cbb 5312 block = ext4_count_free_clusters(sb);
666245d9 5313 ext4_free_blocks_count_set(sbi->s_es,
d5e03cbb 5314 EXT4_C2B(sbi, block));
908c7f19
TH		 5315 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5316 GFP_KERNEL);
d5e03cbb
TT		 5317 if (!err) {
5318 unsigned long freei = ext4_count_free_inodes(sb);
5319 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 5320 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5321 GFP_KERNEL);
d5e03cbb
TT		 5322 }
5323 if (!err)
5324 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 5325 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 5326 if (!err)
908c7f19
TH		 5327 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5328 GFP_KERNEL);
efc61345
EW		 5329 if (!err)
5330 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
5331 GFP_KERNEL);
c8585c6f 5332 if (!err)
bbd55937 5333 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
c8585c6f 5334
d5e03cbb
TT		 5335 if (err) {
5336 ext4_msg(sb, KERN_ERR, "insufficient memory");
5337 goto failed_mount6;
5338 }
5339
e2b911c5 5340 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 5341 if (!ext4_fill_flex_info(sb)) {
5342 ext4_msg(sb, KERN_ERR,
5343 "unable to initialize "
5344 "flex_bg meta info!");
8f6840c4 5345 ret = -ENOMEM;
d5e03cbb
TT		 5346 goto failed_mount6;
5347 }
5348
bfff6873
LC		 5349 err = ext4_register_li_request(sb, first_not_zeroed);
5350 if (err)
dcf2d804 5351 goto failed_mount6;
bfff6873 5352
b5799018 5353 err = ext4_register_sysfs(sb);
dcf2d804
TM		 5354 if (err)
5355 goto failed_mount7;
3197ebdb 5356
02f310fc
JK		 5357 err = ext4_init_orphan_info(sb);
5358 if (err)
5359 goto failed_mount8;
9b2ff357
JK		 5360#ifdef CONFIG_QUOTA
5361 /* Enable quota usage during mount. */
bc98a42c 5362 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 5363 err = ext4_enable_quotas(sb);
5364 if (err)
02f310fc 5365 goto failed_mount9;
9b2ff357
JK		 5366 }
5367#endif /* CONFIG_QUOTA */
5368
bc71726c 5369 /*
5370 * Save the original bdev mapping's wb_err value which could be
5371 * used to detect the metadata async write error.
5372 */
5373 spin_lock_init(&sbi->s_bdev_wb_lock);
9704a322
ZX		 5374 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5375 &sbi->s_bdev_wb_err);
bc71726c 5376 sb->s_bdev->bd_super = sb;
617ba13b
MC		 5377 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5378 ext4_orphan_cleanup(sb, es);
5379 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
9b6641dd
YB		 5380 /*
5381 * Update the checksum after updating free space/inode counters and
5382 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5383 * checksum in the buffer cache until it is written out and
5384 * e2fsprogs programs trying to open a file system immediately
5385 * after it is mounted can fail.
5386 */
5387 ext4_superblock_csum_set(sb);
0390131b 5388 if (needs_recovery) {
b31e1552 5389 ext4_msg(sb, KERN_INFO, "recovery complete");
11215630
JK		 5390 err = ext4_mark_recovery_complete(sb, es);
5391 if (err)
02f310fc 5392 goto failed_mount9;
0390131b 5393 }
0390131b 5394
70200574
CH		 5395 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5396 ext4_msg(sb, KERN_WARNING,
5397 "mounting with \"discard\" option, but the device does not support discard");
79add3a3 5398
66e61a9e
TT		 5399 if (es->s_error_count)
5400 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 5401
efbed4dc
TT		 5402 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5403 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5404 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5405 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
1cf006ed
DM		 5406 atomic_set(&sbi->s_warning_count, 0);
5407 atomic_set(&sbi->s_msg_count, 0);
efbed4dc 5408
ac27a0ec
DK		 5409 return 0;
5410
617ba13b 5411cantfind_ext4:
ac27a0ec 5412 if (!silent)
b31e1552 5413 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 5414 goto failed_mount;
5415
02f310fc
JK		 5416failed_mount9:
5417 ext4_release_orphan_info(sb);
72ba7450 5418failed_mount8:
ebd173be 5419 ext4_unregister_sysfs(sb);
cb8d53d2 5420 kobject_put(&sbi->s_kobj);
dcf2d804
TM		 5421failed_mount7:
5422 ext4_unregister_li_request(sb);
5423failed_mount6:
f9ae9cf5 5424 ext4_mb_release(sb);
7c990728
SJS		 5425 rcu_read_lock();
5426 flex_groups = rcu_dereference(sbi->s_flex_groups);
5427 if (flex_groups) {
5428 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5429 kvfree(flex_groups[i]);
5430 kvfree(flex_groups);
5431 }
5432 rcu_read_unlock();
d5e03cbb
TT		 5433 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5434 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5435 percpu_counter_destroy(&sbi->s_dirs_counter);
5436 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
efc61345 5437 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
bbd55937 5438 percpu_free_rwsem(&sbi->s_writepages_rwsem);
00764937 5439failed_mount5:
f9ae9cf5
TT		 5440 ext4_ext_release(sb);
5441 ext4_release_system_zone(sb);
5442failed_mount4a:
94bf608a 5443 dput(sb->s_root);
32a9bb57 5444 sb->s_root = NULL;
94bf608a 5445failed_mount4:
b31e1552 5446 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 5447 if (EXT4_SB(sb)->rsv_conversion_wq)
5448 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 5449failed_mount_wq:
50c15df6
CX		 5450 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5451 sbi->s_ea_inode_cache = NULL;
5452
5453 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5454 sbi->s_ea_block_cache = NULL;
5455
0390131b 5456 if (sbi->s_journal) {
bb9464e0 5457 /* flush s_error_work before journal destroy. */
5458 flush_work(&sbi->s_error_work);
0390131b
FM		 5459 jbd2_journal_destroy(sbi->s_journal);
5460 sbi->s_journal = NULL;
5461 }
50460fe8 5462failed_mount3a:
d3922a77 5463 ext4_es_unregister_shrinker(sbi);
eb68d0e2 5464failed_mount3:
bb9464e0 5465 /* flush s_error_work before sbi destroy */
c92dc856 5466 flush_work(&sbi->s_error_work);
2a4ae3bc 5467 del_timer_sync(&sbi->s_err_report);
618f0031 5468 ext4_stop_mmpd(sbi);
ac27a0ec 5469failed_mount2:
1d0c3924
TT		 5470 rcu_read_lock();
5471 group_desc = rcu_dereference(sbi->s_group_desc);
ac27a0ec 5472 for (i = 0; i < db_count; i++)
1d0c3924
TT		 5473 brelse(group_desc[i]);
5474 kvfree(group_desc);
5475 rcu_read_unlock();
ac27a0ec 5476failed_mount:
0441984a
DW		 5477 if (sbi->s_chksum_driver)
5478 crypto_free_shash(sbi->s_chksum_driver);
c83ad55e 5479
5298d4bf 5480#if IS_ENABLED(CONFIG_UNICODE)
f8f4acb6 5481 utf8_unload(sb->s_encoding);
c83ad55e
GKB		 5482#endif
5483
ac27a0ec 5484#ifdef CONFIG_QUOTA
a2d4a646 5485 for (i = 0; i < EXT4_MAXQUOTAS; i++)
0ba33fac 5486 kfree(get_qf_name(sb, sbi, i));
ac27a0ec 5487#endif
ac4acb1f 5488 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
afd09b61 5489 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
ac27a0ec 5490 brelse(bh);
afd09b61 5491 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 5492out_fail:
5493 sb->s_fs_info = NULL;
07aa2ea1 5494 return err ? err : ret;
ac27a0ec
DK		 5495}
5496
cebe85d5 5497static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
7edfd85b 5498{
cebe85d5 5499 struct ext4_fs_context *ctx = fc->fs_private;
7edfd85b 5500 struct ext4_sb_info *sbi;
7edfd85b 5501 const char *descr;
cebe85d5 5502 int ret;
7edfd85b
LC		 5503
5504 sbi = ext4_alloc_sbi(sb);
cebe85d5 5505 if (!sbi)
7c268d4c 5506 return -ENOMEM;
7edfd85b 5507
cebe85d5
LC		 5508 fc->s_fs_info = sbi;
5509
5510 /* Cleanup superblock name */
5511 strreplace(sb->s_id, '/', '!');
7edfd85b
LC		 5512
5513 sbi->s_sb_block = 1; /* Default super block location */
cebe85d5
LC		 5514 if (ctx->spec & EXT4_SPEC_s_sb_block)
5515 sbi->s_sb_block = ctx->s_sb_block;
7edfd85b 5516
960e0ab6 5517 ret = __ext4_fill_super(fc, sb);
7edfd85b
LC		 5518 if (ret < 0)
5519 goto free_sbi;
5520
cebe85d5 5521 if (sbi->s_journal) {
7edfd85b
LC		 5522 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5523 descr = " journalled data mode";
5524 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5525 descr = " ordered data mode";
5526 else
5527 descr = " writeback data mode";
5528 } else
5529 descr = "out journal";
5530
5531 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5532 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
cebe85d5
LC		 5533 "Quota mode: %s.", descr, ext4_quota_mode(sb));
5534
eb705421 5535 /* Update the s_overhead_clusters if necessary */
827891a3 5536 ext4_update_overhead(sb, false);
7edfd85b 5537 return 0;
cebe85d5 5538
7edfd85b
LC		 5539free_sbi:
5540 ext4_free_sbi(sbi);
cebe85d5 5541 fc->s_fs_info = NULL;
7edfd85b
LC		 5542 return ret;
5543}
5544
cebe85d5
LC		 5545static int ext4_get_tree(struct fs_context *fc)
5546{
5547 return get_tree_bdev(fc, ext4_fill_super);
5548}
5549
ac27a0ec
DK		 5550/*
5551 * Setup any per-fs journal parameters now. We'll do this both on
5552 * initial mount, once the journal has been initialised but before we've
5553 * done any recovery; and again on any subsequent remount.
5554 */
617ba13b 5555static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 5556{
617ba13b 5557 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5558
30773840
TT		 5559 journal->j_commit_interval = sbi->s_commit_interval;
5560 journal->j_min_batch_time = sbi->s_min_batch_time;
5561 journal->j_max_batch_time = sbi->s_max_batch_time;
6866d7b3 5562 ext4_fc_init(sb, journal);
ac27a0ec 5563
a931da6a 5564 write_lock(&journal->j_state_lock);
ac27a0ec 5565 if (test_opt(sb, BARRIER))
dab291af 5566 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 5567 else
dab291af 5568 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 5569 if (test_opt(sb, DATA_ERR_ABORT))
5570 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5571 else
5572 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 5573 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 5574}
5575
c6cb7e77
EW		 5576static struct inode *ext4_get_journal_inode(struct super_block *sb,
5577 unsigned int journal_inum)
ac27a0ec
DK		 5578{
5579 struct inode *journal_inode;
ac27a0ec 5580
c6cb7e77
EW		 5581 /*
5582 * Test for the existence of a valid inode on disk. Bad things
5583 * happen if we iget() an unused inode, as the subsequent iput()
5584 * will try to delete it.
5585 */
8a363970 5586 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
1d1fe1ee 5587 if (IS_ERR(journal_inode)) {
b31e1552 5588 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 5589 return NULL;
5590 }
5591 if (!journal_inode->i_nlink) {
5592 make_bad_inode(journal_inode);
5593 iput(journal_inode);
b31e1552 5594 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 5595 return NULL;
5596 }
5597
4978c659 5598 ext4_debug("Journal inode found at %p: %lld bytes\n",
ac27a0ec 5599 journal_inode, journal_inode->i_size);
1d1fe1ee 5600 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 5601 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 5602 iput(journal_inode);
5603 return NULL;
5604 }
c6cb7e77
EW		 5605 return journal_inode;
5606}
5607
5608static journal_t *ext4_get_journal(struct super_block *sb,
5609 unsigned int journal_inum)
5610{
5611 struct inode *journal_inode;
5612 journal_t *journal;
5613
11215630
JK		 5614 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5615 return NULL;
c6cb7e77
EW		 5616
5617 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5618 if (!journal_inode)
5619 return NULL;
ac27a0ec 5620
dab291af 5621 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 5622 if (!journal) {
b31e1552 5623 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 5624 iput(journal_inode);
5625 return NULL;
5626 }
5627 journal->j_private = sb;
617ba13b 5628 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 5629 return journal;
5630}
5631
617ba13b 5632static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 5633 dev_t j_dev)
5634{
2b2d6d01 5635 struct buffer_head *bh;
ac27a0ec 5636 journal_t *journal;
617ba13b
MC		 5637 ext4_fsblk_t start;
5638 ext4_fsblk_t len;
ac27a0ec 5639 int hblock, blocksize;
617ba13b 5640 ext4_fsblk_t sb_block;
ac27a0ec 5641 unsigned long offset;
2b2d6d01 5642 struct ext4_super_block *es;
ac27a0ec
DK		 5643 struct block_device *bdev;
5644
11215630
JK		 5645 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5646 return NULL;
0390131b 5647
b31e1552 5648 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 5649 if (bdev == NULL)
5650 return NULL;
5651
ac27a0ec 5652 blocksize = sb->s_blocksize;
e1defc4f 5653 hblock = bdev_logical_block_size(bdev);
ac27a0ec 5654 if (blocksize < hblock) {
b31e1552
ES		 5655 ext4_msg(sb, KERN_ERR,
5656 "blocksize too small for journal device");
ac27a0ec
DK		 5657 goto out_bdev;
5658 }
5659
617ba13b
MC		 5660 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5661 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 5662 set_blocksize(bdev, blocksize);
5663 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 5664 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5665 "external journal");
ac27a0ec
DK		 5666 goto out_bdev;
5667 }
5668
2716b802 5669 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 5670 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 5671 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 5672 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 5673 ext4_msg(sb, KERN_ERR, "external journal has "
5674 "bad superblock");
ac27a0ec
DK		 5675 brelse(bh);
5676 goto out_bdev;
5677 }
5678
df4763be
DW		 5679 if ((le32_to_cpu(es->s_feature_ro_compat) &
5680 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5681 es->s_checksum != ext4_superblock_csum(sb, es)) {
5682 ext4_msg(sb, KERN_ERR, "external journal has "
5683 "corrupt superblock");
5684 brelse(bh);
5685 goto out_bdev;
5686 }
5687
617ba13b 5688 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 5689 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 5690 brelse(bh);
5691 goto out_bdev;
5692 }
5693
bd81d8ee 5694 len = ext4_blocks_count(es);
ac27a0ec
DK		 5695 start = sb_block + 1;
5696 brelse(bh); /* we're done with the superblock */
5697
dab291af 5698 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 5699 start, len, blocksize);
5700 if (!journal) {
b31e1552 5701 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 5702 goto out_bdev;
5703 }
5704 journal->j_private = sb;
2d069c08 5705 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
b31e1552 5706 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 5707 goto out_journal;
5708 }
5709 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 5710 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5711 "user (unsupported) - %d",
ac27a0ec
DK		 5712 be32_to_cpu(journal->j_superblock->s_nr_users));
5713 goto out_journal;
5714 }
ee7ed3aa 5715 EXT4_SB(sb)->s_journal_bdev = bdev;
617ba13b 5716 ext4_init_journal_params(sb, journal);
ac27a0ec 5717 return journal;
0b8e58a1 5718
ac27a0ec 5719out_journal:
dab291af 5720 jbd2_journal_destroy(journal);
ac27a0ec 5721out_bdev:
617ba13b 5722 ext4_blkdev_put(bdev);
ac27a0ec
DK		 5723 return NULL;
5724}
5725
617ba13b
MC		 5726static int ext4_load_journal(struct super_block *sb,
5727 struct ext4_super_block *es,
ac27a0ec
DK		 5728 unsigned long journal_devnum)
5729{
5730 journal_t *journal;
5731 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5732 dev_t journal_dev;
5733 int err = 0;
5734 int really_read_only;
273108fa 5735 int journal_dev_ro;
ac27a0ec 5736
11215630
JK		 5737 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5738 return -EFSCORRUPTED;
0390131b 5739
ac27a0ec
DK		 5740 if (journal_devnum &&
5741 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 5742 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5743 "numbers have changed");
ac27a0ec
DK		 5744 journal_dev = new_decode_dev(journal_devnum);
5745 } else
5746 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5747
273108fa
LC		 5748 if (journal_inum && journal_dev) {
5749 ext4_msg(sb, KERN_ERR,
5750 "filesystem has both journal inode and journal device!");
5751 return -EINVAL;
5752 }
5753
5754 if (journal_inum) {
5755 journal = ext4_get_journal(sb, journal_inum);
5756 if (!journal)
5757 return -EINVAL;
5758 } else {
5759 journal = ext4_get_dev_journal(sb, journal_dev);
5760 if (!journal)
5761 return -EINVAL;
5762 }
5763
5764 journal_dev_ro = bdev_read_only(journal->j_dev);
5765 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5766
5767 if (journal_dev_ro && !sb_rdonly(sb)) {
5768 ext4_msg(sb, KERN_ERR,
5769 "journal device read-only, try mounting with '-o ro'");
5770 err = -EROFS;
5771 goto err_out;
5772 }
ac27a0ec
DK		 5773
5774 /*
5775 * Are we loading a blank journal or performing recovery after a
5776 * crash? For recovery, we need to check in advance whether we
5777 * can get read-write access to the device.
5778 */
e2b911c5 5779 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 5780 if (sb_rdonly(sb)) {
b31e1552
ES		 5781 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5782 "required on readonly filesystem");
ac27a0ec 5783 if (really_read_only) {
b31e1552 5784 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 5785 "unavailable, cannot proceed "
5786 "(try mounting with noload)");
273108fa
LC		 5787 err = -EROFS;
5788 goto err_out;
ac27a0ec 5789 }
b31e1552
ES		 5790 ext4_msg(sb, KERN_INFO, "write access will "
5791 "be enabled during recovery");
ac27a0ec
DK		 5792 }
5793 }
5794
90576c0b 5795 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 5796 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 5797
e2b911c5 5798 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 5799 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 5800 if (!err) {
5801 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5802 if (save)
5803 memcpy(save, ((char *) es) +
5804 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 5805 err = jbd2_journal_load(journal);
1c13d5c0
TT		 5806 if (save)
5807 memcpy(((char *) es) + EXT4_S_ERR_START,
5808 save, EXT4_S_ERR_LEN);
5809 kfree(save);
5810 }
ac27a0ec
DK		 5811
5812 if (err) {
b31e1552 5813 ext4_msg(sb, KERN_ERR, "error loading journal");
273108fa 5814 goto err_out;
ac27a0ec
DK		 5815 }
5816
617ba13b 5817 EXT4_SB(sb)->s_journal = journal;
11215630
JK		 5818 err = ext4_clear_journal_err(sb, es);
5819 if (err) {
5820 EXT4_SB(sb)->s_journal = NULL;
5821 jbd2_journal_destroy(journal);
5822 return err;
5823 }
ac27a0ec 5824
c41303ce 5825 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 5826 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5827 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 5828
5829 /* Make sure we flush the recovery flag to disk. */
4392fbc4 5830 ext4_commit_super(sb);
ac27a0ec
DK		 5831 }
5832
5833 return 0;
273108fa
LC		 5834
5835err_out:
5836 jbd2_journal_destroy(journal);
5837 return err;
ac27a0ec
DK		 5838}
5839
2d01ddc8
JK		 5840/* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5841static void ext4_update_super(struct super_block *sb)
ac27a0ec 5842{
c92dc856 5843 struct ext4_sb_info *sbi = EXT4_SB(sb);
e92ad03f
JK		 5844 struct ext4_super_block *es = sbi->s_es;
5845 struct buffer_head *sbh = sbi->s_sbh;
a17712c8 5846
05c2c00f 5847 lock_buffer(sbh);
71290b36
TT		 5848 /*
5849 * If the file system is mounted read-only, don't update the
5850 * superblock write time. This avoids updating the superblock
5851 * write time when we are mounting the root file system
5852 * read/only but we need to replay the journal; at that point,
5853 * for people who are east of GMT and who make their clock
5854 * tick in localtime for Windows bug-for-bug compatibility,
5855 * the clock is set in the future, and this will cause e2fsck
5856 * to complain and force a full file system check.
5857 */
1751e8a6 5858 if (!(sb->s_flags & SB_RDONLY))
6a0678a7 5859 ext4_update_tstamp(es, s_wtime);
8446fe92 5860 es->s_kbytes_written =
0bc9bc1d 5861 cpu_to_le64(sbi->s_kbytes_written +
8446fe92 5862 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
0bc9bc1d 5863 sbi->s_sectors_written_start) >> 1));
e92ad03f 5864 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
d5e03cbb 5865 ext4_free_blocks_count_set(es,
e92ad03f
JK		 5866 EXT4_C2B(sbi, percpu_counter_sum_positive(
5867 &sbi->s_freeclusters_counter)));
5868 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
d5e03cbb
TT		 5869 es->s_free_inodes_count =
5870 cpu_to_le32(percpu_counter_sum_positive(
e92ad03f 5871 &sbi->s_freeinodes_counter));
c92dc856
JK		 5872 /* Copy error information to the on-disk superblock */
5873 spin_lock(&sbi->s_error_lock);
5874 if (sbi->s_add_error_count > 0) {
5875 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5876 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
5877 __ext4_update_tstamp(&es->s_first_error_time,
5878 &es->s_first_error_time_hi,
5879 sbi->s_first_error_time);
5880 strncpy(es->s_first_error_func, sbi->s_first_error_func,
5881 sizeof(es->s_first_error_func));
5882 es->s_first_error_line =
5883 cpu_to_le32(sbi->s_first_error_line);
5884 es->s_first_error_ino =
5885 cpu_to_le32(sbi->s_first_error_ino);
5886 es->s_first_error_block =
5887 cpu_to_le64(sbi->s_first_error_block);
5888 es->s_first_error_errcode =
5889 ext4_errno_to_code(sbi->s_first_error_code);
5890 }
5891 __ext4_update_tstamp(&es->s_last_error_time,
5892 &es->s_last_error_time_hi,
5893 sbi->s_last_error_time);
5894 strncpy(es->s_last_error_func, sbi->s_last_error_func,
5895 sizeof(es->s_last_error_func));
5896 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
5897 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
5898 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
5899 es->s_last_error_errcode =
5900 ext4_errno_to_code(sbi->s_last_error_code);
5901 /*
5902 * Start the daily error reporting function if it hasn't been
5903 * started already
5904 */
5905 if (!es->s_error_count)
5906 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
5907 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
5908 sbi->s_add_error_count = 0;
5909 }
5910 spin_unlock(&sbi->s_error_lock);
5911
06db49e6 5912 ext4_superblock_csum_set(sb);
2d01ddc8
JK		 5913 unlock_buffer(sbh);
5914}
5915
5916static int ext4_commit_super(struct super_block *sb)
5917{
5918 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2d01ddc8 5919
f88f1466
FC		 5920 if (!sbh)
5921 return -EINVAL;
5922 if (block_device_ejected(sb))
5923 return -ENODEV;
2d01ddc8
JK		 5924
5925 ext4_update_super(sb);
5926
15baa7dc
ZY		 5927 lock_buffer(sbh);
5928 /* Buffer got discarded which means block device got invalidated */
5929 if (!buffer_mapped(sbh)) {
5930 unlock_buffer(sbh);
5931 return -EIO;
5932 }
5933
e8680786 5934 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4743f839
PKS		 5935 /*
5936 * Oh, dear. A previous attempt to write the
5937 * superblock failed. This could happen because the
5938 * USB device was yanked out. Or it could happen to
5939 * be a transient write error and maybe the block will
5940 * be remapped. Nothing we can do but to retry the
5941 * write and hope for the best.
5942 */
5943 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5944 "superblock detected");
5945 clear_buffer_write_io_error(sbh);
5946 set_buffer_uptodate(sbh);
5947 }
15baa7dc
ZY		 5948 get_bh(sbh);
5949 /* Clear potential dirty bit if it was journalled update */
5950 clear_buffer_dirty(sbh);
5951 sbh->b_end_io = end_buffer_write_sync;
1420c4a5
BVA		 5952 submit_bh(REQ_OP_WRITE | REQ_SYNC |
5953 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
15baa7dc 5954 wait_on_buffer(sbh);
4392fbc4
JK		 5955 if (buffer_write_io_error(sbh)) {
5956 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5957 "superblock");
5958 clear_buffer_write_io_error(sbh);
5959 set_buffer_uptodate(sbh);
15baa7dc 5960 return -EIO;
914258bf 5961 }
15baa7dc 5962 return 0;
ac27a0ec
DK		 5963}
5964
ac27a0ec
DK		 5965/*
5966 * Have we just finished recovery? If so, and if we are mounting (or
5967 * remounting) the filesystem readonly, then we will end up with a
5968 * consistent fs on disk. Record that fact.
5969 */
11215630
JK		 5970static int ext4_mark_recovery_complete(struct super_block *sb,
5971 struct ext4_super_block *es)
ac27a0ec 5972{
11215630 5973 int err;
617ba13b 5974 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 5975
e2b911c5 5976 if (!ext4_has_feature_journal(sb)) {
11215630
JK		 5977 if (journal != NULL) {
5978 ext4_error(sb, "Journal got removed while the fs was "
5979 "mounted!");
5980 return -EFSCORRUPTED;
5981 }
5982 return 0;
0390131b 5983 }
dab291af 5984 jbd2_journal_lock_updates(journal);
01d5d965 5985 err = jbd2_journal_flush(journal, 0);
11215630 5986 if (err < 0)
7ffe1ea8
HK		 5987 goto out;
5988
02f310fc
JK		 5989 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
5990 ext4_has_feature_orphan_present(sb))) {
5991 if (!ext4_orphan_file_empty(sb)) {
5992 ext4_error(sb, "Orphan file not empty on read-only fs.");
5993 err = -EFSCORRUPTED;
5994 goto out;
5995 }
e2b911c5 5996 ext4_clear_feature_journal_needs_recovery(sb);
02f310fc 5997 ext4_clear_feature_orphan_present(sb);
4392fbc4 5998 ext4_commit_super(sb);
ac27a0ec 5999 }
7ffe1ea8 6000out:
dab291af 6001 jbd2_journal_unlock_updates(journal);
11215630 6002 return err;
ac27a0ec
DK		 6003}
6004
6005/*
6006 * If we are mounting (or read-write remounting) a filesystem whose journal
6007 * has recorded an error from a previous lifetime, move that error to the
6008 * main filesystem now.
6009 */
11215630 6010static int ext4_clear_journal_err(struct super_block *sb,
2b2d6d01 6011 struct ext4_super_block *es)
ac27a0ec
DK		 6012{
6013 journal_t *journal;
6014 int j_errno;
6015 const char *errstr;
6016
11215630
JK		 6017 if (!ext4_has_feature_journal(sb)) {
6018 ext4_error(sb, "Journal got removed while the fs was mounted!");
6019 return -EFSCORRUPTED;
6020 }
0390131b 6021
617ba13b 6022 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 6023
6024 /*
6025 * Now check for any error status which may have been recorded in the
617ba13b 6026 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 6027 */
6028
dab291af 6029 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 6030 if (j_errno) {
6031 char nbuf[16];
6032
617ba13b 6033 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 6034 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 6035 "from previous mount: %s", errstr);
12062ddd 6036 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 6037
617ba13b
MC		 6038 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6039 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4392fbc4 6040 ext4_commit_super(sb);
ac27a0ec 6041
dab291af 6042 jbd2_journal_clear_err(journal);
d796c52e 6043 jbd2_journal_update_sb_errno(journal);
ac27a0ec 6044 }
11215630 6045 return 0;
ac27a0ec
DK		 6046}
6047
6048/*
6049 * Force the running and committing transactions to commit,
6050 * and wait on the commit.
6051 */
617ba13b 6052int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 6053{
6054 journal_t *journal;
ac27a0ec 6055
bc98a42c 6056 if (sb_rdonly(sb))
ac27a0ec
DK		 6057 return 0;
6058
617ba13b 6059 journal = EXT4_SB(sb)->s_journal;
b1deefc9 6060 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 6061}
6062
617ba13b 6063static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 6064{
14ce0cb4 6065 int ret = 0;
9eddacf9 6066 tid_t target;
06a407f1 6067 bool needs_barrier = false;
8d5d02e6 6068 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 6069
49598e04 6070 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 6071 return 0;
6072
9bffad1e 6073 trace_ext4_sync_fs(sb, wait);
2e8fa54e 6074 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 6075 /*
6076 * Writeback quota in non-journalled quota case - journalled quota has
6077 * no dirty dquots
6078 */
6079 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 6080 /*
6081 * Data writeback is possible w/o journal transaction, so barrier must
6082 * being sent at the end of the function. But we can skip it if
6083 * transaction_commit will do it for us.
6084 */
bda32530
TT		 6085 if (sbi->s_journal) {
6086 target = jbd2_get_latest_transaction(sbi->s_journal);
6087 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6088 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6089 needs_barrier = true;
6090
6091 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6092 if (wait)
6093 ret = jbd2_log_wait_commit(sbi->s_journal,
6094 target);
6095 }
6096 } else if (wait && test_opt(sb, BARRIER))
06a407f1 6097 needs_barrier = true;
06a407f1
DM		 6098 if (needs_barrier) {
6099 int err;
c6bf3f0e 6100 err = blkdev_issue_flush(sb->s_bdev);
06a407f1
DM		 6101 if (!ret)
6102 ret = err;
0390131b 6103 }
06a407f1
DM		 6104
6105 return ret;
6106}
6107
ac27a0ec
DK		 6108/*
6109 * LVM calls this function before a (read-only) snapshot is created. This
6110 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 6111 *
6112 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 6113 * state independently. It relies on upper layer to stop all data & metadata
6114 * modifications.
ac27a0ec 6115 */
c4be0c1d 6116static int ext4_freeze(struct super_block *sb)
ac27a0ec 6117{
c4be0c1d
TS		 6118 int error = 0;
6119 journal_t *journal;
ac27a0ec 6120
bc98a42c 6121 if (sb_rdonly(sb))
9ca92389 6122 return 0;
ac27a0ec 6123
9ca92389 6124 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 6125
bb044576
TT		 6126 if (journal) {
6127 /* Now we set up the journal barrier. */
6128 jbd2_journal_lock_updates(journal);
ac27a0ec 6129
bb044576
TT		 6130 /*
6131 * Don't clear the needs_recovery flag if we failed to
6132 * flush the journal.
6133 */
01d5d965 6134 error = jbd2_journal_flush(journal, 0);
bb044576
TT		 6135 if (error < 0)
6136 goto out;
c642dc9e
ES		 6137
6138 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 6139 ext4_clear_feature_journal_needs_recovery(sb);
02f310fc
JK		 6140 if (ext4_orphan_file_empty(sb))
6141 ext4_clear_feature_orphan_present(sb);
bb044576 6142 }
9ca92389 6143
4392fbc4 6144 error = ext4_commit_super(sb);
6b0310fb 6145out:
bb044576
TT		 6146 if (journal)
6147 /* we rely on upper layer to stop further updates */
6148 jbd2_journal_unlock_updates(journal);
6b0310fb 6149 return error;
ac27a0ec
DK		 6150}
6151
6152/*
6153 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6154 * flag here, even though the filesystem is not technically dirty yet.
6155 */
c4be0c1d 6156static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 6157{
bc98a42c 6158 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 6159 return 0;
6160
c642dc9e
ES		 6161 if (EXT4_SB(sb)->s_journal) {
6162 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 6163 ext4_set_feature_journal_needs_recovery(sb);
02f310fc
JK		 6164 if (ext4_has_feature_orphan_file(sb))
6165 ext4_set_feature_orphan_present(sb);
c642dc9e
ES		 6166 }
6167
4392fbc4 6168 ext4_commit_super(sb);
c4be0c1d 6169 return 0;
ac27a0ec
DK		 6170}
6171
673c6100
TT		 6172/*
6173 * Structure to save mount options for ext4_remount's benefit
6174 */
6175struct ext4_mount_options {
6176 unsigned long s_mount_opt;
a2595b8a 6177 unsigned long s_mount_opt2;
08cefc7a
EB		 6178 kuid_t s_resuid;
6179 kgid_t s_resgid;
673c6100
TT		 6180 unsigned long s_commit_interval;
6181 u32 s_min_batch_time, s_max_batch_time;
6182#ifdef CONFIG_QUOTA
6183 int s_jquota_fmt;
a2d4a646 6184 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 6185#endif
6186};
6187
960e0ab6 6188static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
ac27a0ec 6189{
7edfd85b 6190 struct ext4_fs_context *ctx = fc->fs_private;
2b2d6d01 6191 struct ext4_super_block *es;
617ba13b 6192 struct ext4_sb_info *sbi = EXT4_SB(sb);
960e0ab6 6193 unsigned long old_sb_flags;
617ba13b 6194 struct ext4_mount_options old_opts;
8a266467 6195 ext4_group_t g;
c5e06d10 6196 int err = 0;
ac27a0ec 6197#ifdef CONFIG_QUOTA
3bbef91b 6198 int enable_quota = 0;
03dafb5f 6199 int i, j;
33458eab 6200 char *to_free[EXT4_MAXQUOTAS];
ac27a0ec
DK		 6201#endif
6202
21ac738e 6203
ac27a0ec
DK		 6204 /* Store the original options */
6205 old_sb_flags = sb->s_flags;
6206 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 6207 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 6208 old_opts.s_resuid = sbi->s_resuid;
6209 old_opts.s_resgid = sbi->s_resgid;
6210 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 6211 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6212 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 6213#ifdef CONFIG_QUOTA
6214 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 6215 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 6216 if (sbi->s_qf_names[i]) {
33458eab
TT		 6217 char *qf_name = get_qf_name(sb, sbi, i);
6218
6219 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
03dafb5f
CG		 6220 if (!old_opts.s_qf_names[i]) {
6221 for (j = 0; j < i; j++)
6222 kfree(old_opts.s_qf_names[j]);
6223 return -ENOMEM;
6224 }
6225 } else
6226 old_opts.s_qf_names[i] = NULL;
ac27a0ec 6227#endif
e4e58e5d
OM		 6228 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6229 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6230 ctx->journal_ioprio =
6231 sbi->s_journal->j_task->io_context->ioprio;
6232 else
6233 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6234
6235 }
ac27a0ec 6236
7edfd85b 6237 ext4_apply_options(fc, sb);
ac27a0ec 6238
6b992ff2 6239 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 6240 test_opt(sb, JOURNAL_CHECKSUM)) {
6241 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 6242 "during remount not supported; ignoring");
6243 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 6244 }
6245
6ae6514b
PS		 6246 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6247 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6248 ext4_msg(sb, KERN_ERR, "can't mount with "
6249 "both data=journal and delalloc");
6250 err = -EINVAL;
6251 goto restore_opts;
6252 }
6253 if (test_opt(sb, DIOREAD_NOLOCK)) {
6254 ext4_msg(sb, KERN_ERR, "can't mount with "
6255 "both data=journal and dioread_nolock");
6256 err = -EINVAL;
6257 goto restore_opts;
6258 }
ab04df78
JK		 6259 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6260 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6261 ext4_msg(sb, KERN_ERR, "can't mount with "
6262 "journal_async_commit in data=ordered mode");
6263 err = -EINVAL;
6264 goto restore_opts;
6265 }
923ae0ff
RZ		 6266 }
6267
cdb7ee4c
TE		 6268 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6269 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6270 err = -EINVAL;
6271 goto restore_opts;
6272 }
6273
9b5f6c9b 6274 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
124e7c61 6275 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
ac27a0ec 6276
1751e8a6
LT		 6277 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6278 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 6279
6280 es = sbi->s_es;
6281
b3881f74 6282 if (sbi->s_journal) {
0390131b 6283 ext4_init_journal_params(sb, sbi->s_journal);
7edfd85b 6284 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
b3881f74 6285 }
ac27a0ec 6286
c92dc856
JK		 6287 /* Flush outstanding errors before changing fs state */
6288 flush_work(&sbi->s_error_work);
6289
960e0ab6 6290 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
9b5f6c9b 6291 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
ac27a0ec
DK		 6292 err = -EROFS;
6293 goto restore_opts;
6294 }
6295
960e0ab6 6296 if (fc->sb_flags & SB_RDONLY) {
38c03b34
TT		 6297 err = sync_filesystem(sb);
6298 if (err < 0)
6299 goto restore_opts;
0f0dd62f
CH		 6300 err = dquot_suspend(sb, -1);
6301 if (err < 0)
c79d967d 6302 goto restore_opts;
c79d967d 6303
ac27a0ec
DK		 6304 /*
6305 * First of all, the unconditional stuff we have to do
6306 * to disable replay of the journal when we next remount
6307 */
1751e8a6 6308 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 6309
6310 /*
6311 * OK, test if we are remounting a valid rw partition
6312 * readonly, and if so set the rdonly flag and then
6313 * mark the partition as valid again.
6314 */
617ba13b
MC		 6315 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6316 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 6317 es->s_state = cpu_to_le16(sbi->s_mount_state);
6318
11215630
JK		 6319 if (sbi->s_journal) {
6320 /*
6321 * We let remount-ro finish even if marking fs
6322 * as clean failed...
6323 */
0390131b 6324 ext4_mark_recovery_complete(sb, es);
11215630 6325 }
ac27a0ec 6326 } else {
a13fb1a4 6327 /* Make sure we can mount this feature set readwrite */
e2b911c5 6328 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 6329 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 6330 err = -EROFS;
6331 goto restore_opts;
6332 }
8a266467
TT		 6333 /*
6334 * Make sure the group descriptor checksums
0b8e58a1 6335 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 6336 */
6337 for (g = 0; g < sbi->s_groups_count; g++) {
6338 struct ext4_group_desc *gdp =
6339 ext4_get_group_desc(sb, g, NULL);
6340
feb0ab32 6341 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 6342 ext4_msg(sb, KERN_ERR,
6343 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 6344 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 6345 le16_to_cpu(gdp->bg_checksum));
6a797d27 6346 err = -EFSBADCRC;
8a266467
TT		 6347 goto restore_opts;
6348 }
6349 }
6350
ead6596b
ES		 6351 /*
6352 * If we have an unprocessed orphan list hanging
6353 * around from a previously readonly bdev mount,
6354 * require a full umount/remount for now.
6355 */
02f310fc 6356 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
b31e1552 6357 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 6358 "remount RDWR because of unprocessed "
6359 "orphan inode list. Please "
b31e1552 6360 "umount/remount instead");
ead6596b
ES		 6361 err = -EINVAL;
6362 goto restore_opts;
6363 }
6364
ac27a0ec
DK		 6365 /*
6366 * Mounting a RDONLY partition read-write, so reread
6367 * and store the current valid flag. (It may have
6368 * been changed by e2fsck since we originally mounted
6369 * the partition.)
6370 */
11215630
JK		 6371 if (sbi->s_journal) {
6372 err = ext4_clear_journal_err(sb, es);
6373 if (err)
6374 goto restore_opts;
6375 }
c878bea3
TT		 6376 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6377 ~EXT4_FC_REPLAY);
c89128a0
JK		 6378
6379 err = ext4_setup_super(sb, es, 0);
6380 if (err)
6381 goto restore_opts;
6382
6383 sb->s_flags &= ~SB_RDONLY;
e2b911c5 6384 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 6385 if (ext4_multi_mount_protect(sb,
6386 le64_to_cpu(es->s_mmp_block))) {
6387 err = -EROFS;
6388 goto restore_opts;
6389 }
3bbef91b 6390#ifdef CONFIG_QUOTA
c79d967d 6391 enable_quota = 1;
3bbef91b 6392#endif
ac27a0ec
DK		 6393 }
6394 }
bfff6873
LC		 6395
6396 /*
6397 * Reinitialize lazy itable initialization thread based on
6398 * current settings
6399 */
bc98a42c 6400 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 6401 ext4_unregister_li_request(sb);
6402 else {
6403 ext4_group_t first_not_zeroed;
6404 first_not_zeroed = ext4_has_uninit_itable(sb);
6405 ext4_register_li_request(sb, first_not_zeroed);
6406 }
6407
0f5bde1d
JK		 6408 /*
6409 * Handle creation of system zone data early because it can fail.
6410 * Releasing of existing data is done when we are sure remount will
6411 * succeed.
6412 */
dd0db94f 6413 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
0f5bde1d
JK		 6414 err = ext4_setup_system_zone(sb);
6415 if (err)
6416 goto restore_opts;
6417 }
d176b1f6 6418
c89128a0 6419 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
4392fbc4 6420 err = ext4_commit_super(sb);
c89128a0
JK		 6421 if (err)
6422 goto restore_opts;
6423 }
0390131b 6424
ac27a0ec
DK		 6425#ifdef CONFIG_QUOTA
6426 /* Release old quota file names */
a2d4a646 6427 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 6428 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 6429 if (enable_quota) {
6430 if (sb_any_quota_suspended(sb))
6431 dquot_resume(sb, -1);
e2b911c5 6432 else if (ext4_has_feature_quota(sb)) {
7c319d32 6433 err = ext4_enable_quotas(sb);
07724f98 6434 if (err)
7c319d32 6435 goto restore_opts;
7c319d32
AK		 6436 }
6437 }
ac27a0ec 6438#endif
dd0db94f 6439 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
0f5bde1d 6440 ext4_release_system_zone(sb);
d4c402d9 6441
61bb4a1c
TT		 6442 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6443 ext4_stop_mmpd(sbi);
6444
ac27a0ec 6445 return 0;
0b8e58a1 6446
ac27a0ec
DK		 6447restore_opts:
6448 sb->s_flags = old_sb_flags;
6449 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 6450 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 6451 sbi->s_resuid = old_opts.s_resuid;
6452 sbi->s_resgid = old_opts.s_resgid;
6453 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 6454 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6455 sbi->s_max_batch_time = old_opts.s_max_batch_time;
dd0db94f 6456 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
0f5bde1d 6457 ext4_release_system_zone(sb);
ac27a0ec
DK		 6458#ifdef CONFIG_QUOTA
6459 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 6460 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
33458eab
TT		 6461 to_free[i] = get_qf_name(sb, sbi, i);
6462 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
ac27a0ec 6463 }
33458eab
TT		 6464 synchronize_rcu();
6465 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6466 kfree(to_free[i]);
ac27a0ec 6467#endif
61bb4a1c
TT		 6468 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6469 ext4_stop_mmpd(sbi);
ac27a0ec
DK		 6470 return err;
6471}
6472
cebe85d5 6473static int ext4_reconfigure(struct fs_context *fc)
7edfd85b 6474{
cebe85d5 6475 struct super_block *sb = fc->root->d_sb;
7edfd85b
LC		 6476 int ret;
6477
cebe85d5 6478 fc->s_fs_info = EXT4_SB(sb);
7edfd85b 6479
cebe85d5 6480 ret = ext4_check_opt_consistency(fc, sb);
7edfd85b 6481 if (ret < 0)
cebe85d5 6482 return ret;
7edfd85b 6483
960e0ab6 6484 ret = __ext4_remount(fc, sb);
7edfd85b 6485 if (ret < 0)
cebe85d5 6486 return ret;
7edfd85b 6487
cebe85d5
LC		 6488 ext4_msg(sb, KERN_INFO, "re-mounted. Quota mode: %s.",
6489 ext4_quota_mode(sb));
7edfd85b 6490
7edfd85b 6491 return 0;
7edfd85b
LC		 6492}
6493
689c958c
LX		 6494#ifdef CONFIG_QUOTA
6495static int ext4_statfs_project(struct super_block *sb,
6496 kprojid_t projid, struct kstatfs *buf)
6497{
6498 struct kqid qid;
6499 struct dquot *dquot;
6500 u64 limit;
6501 u64 curblock;
6502
6503 qid = make_kqid_projid(projid);
6504 dquot = dqget(sb, qid);
6505 if (IS_ERR(dquot))
6506 return PTR_ERR(dquot);
7b9ca4c6 6507 spin_lock(&dquot->dq_dqb_lock);
689c958c 6508
a08fe66e
CX		 6509 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6510 dquot->dq_dqb.dqb_bhardlimit);
57c32ea4
CX		 6511 limit >>= sb->s_blocksize_bits;
6512
689c958c 6513 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 6514 curblock = (dquot->dq_dqb.dqb_curspace +
6515 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 6516 buf->f_blocks = limit;
6517 buf->f_bfree = buf->f_bavail =
6518 (buf->f_blocks > curblock) ?
6519 (buf->f_blocks - curblock) : 0;
6520 }
6521
a08fe66e
CX		 6522 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6523 dquot->dq_dqb.dqb_ihardlimit);
689c958c
LX		 6524 if (limit && buf->f_files > limit) {
6525 buf->f_files = limit;
6526 buf->f_ffree =
6527 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6528 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6529 }
6530
7b9ca4c6 6531 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 6532 dqput(dquot);
6533 return 0;
6534}
6535#endif
6536
2b2d6d01 6537static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 6538{
6539 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 6540 struct ext4_sb_info *sbi = EXT4_SB(sb);
6541 struct ext4_super_block *es = sbi->s_es;
27dd4385 6542 ext4_fsblk_t overhead = 0, resv_blocks;
d02a9391 6543 s64 bfree;
27dd4385 6544 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 6545
952fc18e
TT		 6546 if (!test_opt(sb, MINIX_DF))
6547 overhead = sbi->s_overhead;
ac27a0ec 6548
617ba13b 6549 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 6550 buf->f_bsize = sb->s_blocksize;
b72f78cb 6551 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 6552 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6553 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 6554 /* prevent underflow in case that few free space is available */
57042651 6555 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 6556 buf->f_bavail = buf->f_bfree -
6557 (ext4_r_blocks_count(es) + resv_blocks);
6558 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 6559 buf->f_bavail = 0;
6560 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 6561 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 6562 buf->f_namelen = EXT4_NAME_LEN;
9591c3a3 6563 buf->f_fsid = uuid_to_fsid(es->s_uuid);
0b8e58a1 6564
689c958c
LX		 6565#ifdef CONFIG_QUOTA
6566 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6567 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6568 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6569#endif
ac27a0ec
DK		 6570 return 0;
6571}
6572
ac27a0ec
DK		 6573
6574#ifdef CONFIG_QUOTA
6575
bc8230ee
JK		 6576/*
6577 * Helper functions so that transaction is started before we acquire dqio_sem
6578 * to keep correct lock ordering of transaction > dqio_sem
6579 */
ac27a0ec
DK		 6580static inline struct inode *dquot_to_inode(struct dquot *dquot)
6581{
4c376dca 6582 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 6583}
6584
617ba13b 6585static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 6586{
6587 int ret, err;
6588 handle_t *handle;
6589 struct inode *inode;
6590
6591 inode = dquot_to_inode(dquot);
9924a92a 6592 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 6593 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 6594 if (IS_ERR(handle))
6595 return PTR_ERR(handle);
6596 ret = dquot_commit(dquot);
617ba13b 6597 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6598 if (!ret)
6599 ret = err;
6600 return ret;
6601}
6602
617ba13b 6603static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 6604{
6605 int ret, err;
6606 handle_t *handle;
6607
9924a92a 6608 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 6609 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 6610 if (IS_ERR(handle))
6611 return PTR_ERR(handle);
6612 ret = dquot_acquire(dquot);
617ba13b 6613 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6614 if (!ret)
6615 ret = err;
6616 return ret;
6617}
6618
617ba13b 6619static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 6620{
6621 int ret, err;
6622 handle_t *handle;
6623
9924a92a 6624 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 6625 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 6626 if (IS_ERR(handle)) {
6627 /* Release dquot anyway to avoid endless cycle in dqput() */
6628 dquot_release(dquot);
ac27a0ec 6629 return PTR_ERR(handle);
9c3013e9 6630 }
ac27a0ec 6631 ret = dquot_release(dquot);
617ba13b 6632 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6633 if (!ret)
6634 ret = err;
6635 return ret;
6636}
6637
617ba13b 6638static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 6639{
262b4662 6640 struct super_block *sb = dquot->dq_sb;
262b4662 6641
f177ee08 6642 if (ext4_is_quota_journalled(sb)) {
ac27a0ec 6643 dquot_mark_dquot_dirty(dquot);
617ba13b 6644 return ext4_write_dquot(dquot);
ac27a0ec
DK		 6645 } else {
6646 return dquot_mark_dquot_dirty(dquot);
6647 }
6648}
6649
617ba13b 6650static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 6651{
6652 int ret, err;
6653 handle_t *handle;
6654
6655 /* Data block + inode block */
2b0143b5 6656 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 6657 if (IS_ERR(handle))
6658 return PTR_ERR(handle);
6659 ret = dquot_commit_info(sb, type);
617ba13b 6660 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6661 if (!ret)
6662 ret = err;
6663 return ret;
6664}
6665
daf647d2
TT		 6666static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6667{
6668 struct ext4_inode_info *ei = EXT4_I(inode);
6669
6670 /* The first argument of lockdep_set_subclass has to be
6671 * *exactly* the same as the argument to init_rwsem() --- in
6672 * this case, in init_once() --- or lockdep gets unhappy
6673 * because the name of the lock is set using the
6674 * stringification of the argument to init_rwsem().
6675 */
6676 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6677 lockdep_set_subclass(&ei->i_data_sem, subclass);
6678}
6679
ac27a0ec
DK		 6680/*
6681 * Standard function to be called on quota_on
6682 */
617ba13b 6683static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 6684 const struct path *path)
ac27a0ec
DK		 6685{
6686 int err;
ac27a0ec
DK		 6687
6688 if (!test_opt(sb, QUOTA))
6689 return -EINVAL;
0623543b 6690
ac27a0ec 6691 /* Quotafile not on the same filesystem? */
d8c9584e 6692 if (path->dentry->d_sb != sb)
ac27a0ec 6693 return -EXDEV;
e0770e91
JK		 6694
6695 /* Quota already enabled for this file? */
6696 if (IS_NOQUOTA(d_inode(path->dentry)))
6697 return -EBUSY;
6698
0623543b
JK		 6699 /* Journaling quota? */
6700 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 6701 /* Quotafile not in fs root? */
f00c9e44 6702 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 6703 ext4_msg(sb, KERN_WARNING,
6704 "Quota file not on filesystem root. "
6705 "Journaled quota will not work");
91389240
JK		 6706 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6707 } else {
6708 /*
6709 * Clear the flag just in case mount options changed since
6710 * last time.
6711 */
6712 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 6713 }
0623543b
JK		 6714
6715 /*
6716 * When we journal data on quota file, we have to flush journal to see
6717 * all updates to the file when we bypass pagecache...
6718 */
0390131b 6719 if (EXT4_SB(sb)->s_journal &&
2b0143b5 6720 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 6721 /*
6722 * We don't need to lock updates but journal_flush() could
6723 * otherwise be livelocked...
6724 */
6725 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
01d5d965 6726 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
0623543b 6727 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 6728 if (err)
7ffe1ea8 6729 return err;
0623543b 6730 }
957153fc 6731
daf647d2
TT		 6732 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6733 err = dquot_quota_on(sb, type, format_id, path);
15fc69bb 6734 if (!err) {
957153fc
JK		 6735 struct inode *inode = d_inode(path->dentry);
6736 handle_t *handle;
6737
61a92987
JK		 6738 /*
6739 * Set inode flags to prevent userspace from messing with quota
6740 * files. If this fails, we return success anyway since quotas
6741 * are already enabled and this is not a hard failure.
6742 */
957153fc
JK		 6743 inode_lock(inode);
6744 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6745 if (IS_ERR(handle))
6746 goto unlock_inode;
6747 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6748 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6749 S_NOATIME | S_IMMUTABLE);
4209ae12 6750 err = ext4_mark_inode_dirty(handle, inode);
957153fc
JK		 6751 ext4_journal_stop(handle);
6752 unlock_inode:
6753 inode_unlock(inode);
15fc69bb
JK		 6754 if (err)
6755 dquot_quota_off(sb, type);
957153fc 6756 }
15fc69bb
JK		 6757 if (err)
6758 lockdep_set_quota_inode(path->dentry->d_inode,
6759 I_DATA_SEM_NORMAL);
daf647d2 6760 return err;
ac27a0ec
DK		 6761}
6762
7c319d32
AK		 6763static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6764 unsigned int flags)
6765{
6766 int err;
6767 struct inode *qf_inode;
a2d4a646 6768 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 6769 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 6770 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6771 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 6772 };
6773
e2b911c5 6774 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 6775
6776 if (!qf_inums[type])
6777 return -EPERM;
6778
8a363970 6779 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7c319d32
AK		 6780 if (IS_ERR(qf_inode)) {
6781 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6782 return PTR_ERR(qf_inode);
6783 }
6784
bcb13850
JK		 6785 /* Don't account quota for quota files to avoid recursion */
6786 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 6787 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7212b95e 6788 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
daf647d2
TT		 6789 if (err)
6790 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
61157b24 6791 iput(qf_inode);
7c319d32
AK		 6792
6793 return err;
6794}
6795
6796/* Enable usage tracking for all quota types. */
25c6d98f 6797int ext4_enable_quotas(struct super_block *sb)
7c319d32
AK		 6798{
6799 int type, err = 0;
a2d4a646 6800 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 6801 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 6802 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6803 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 6804 };
49da9392
JK		 6805 bool quota_mopt[EXT4_MAXQUOTAS] = {
6806 test_opt(sb, USRQUOTA),
6807 test_opt(sb, GRPQUOTA),
6808 test_opt(sb, PRJQUOTA),
6809 };
7c319d32 6810
91389240 6811 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 6812 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 6813 if (qf_inums[type]) {
6814 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 6815 DQUOT_USAGE_ENABLED |
6816 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 6817 if (err) {
6818 ext4_warning(sb,
72ba7450
TT		 6819 "Failed to enable quota tracking "
6820 "(type=%d, err=%d). Please run "
6821 "e2fsck to fix.", type, err);
4013d47a
JK		 6822 for (type--; type >= 0; type--) {
6823 struct inode *inode;
6824
6825 inode = sb_dqopt(sb)->files[type];
6826 if (inode)
6827 inode = igrab(inode);
7f144fd0 6828 dquot_quota_off(sb, type);
4013d47a
JK		 6829 if (inode) {
6830 lockdep_set_quota_inode(inode,
6831 I_DATA_SEM_NORMAL);
6832 iput(inode);
6833 }
6834 }
7f144fd0 6835
7c319d32
AK		 6836 return err;
6837 }
6838 }
6839 }
6840 return 0;
6841}
6842
ca0e05e4
DM		 6843static int ext4_quota_off(struct super_block *sb, int type)
6844{
21f97697
JK		 6845 struct inode *inode = sb_dqopt(sb)->files[type];
6846 handle_t *handle;
957153fc 6847 int err;
21f97697 6848
87009d86
DM		 6849 /* Force all delayed allocation blocks to be allocated.
6850 * Caller already holds s_umount sem */
6851 if (test_opt(sb, DELALLOC))
ca0e05e4 6852 sync_filesystem(sb);
ca0e05e4 6853
957153fc 6854 if (!inode || !igrab(inode))
0b268590
AG		 6855 goto out;
6856
957153fc 6857 err = dquot_quota_off(sb, type);
964edf66 6858 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 6859 goto out_put;
6860
6861 inode_lock(inode);
61a92987
JK		 6862 /*
6863 * Update modification times of quota files when userspace can
6864 * start looking at them. If we fail, we return success anyway since
6865 * this is not a hard failure and quotas are already disabled.
6866 */
9924a92a 6867 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
4209ae12
HS		 6868 if (IS_ERR(handle)) {
6869 err = PTR_ERR(handle);
957153fc 6870 goto out_unlock;
4209ae12 6871 }
957153fc
JK		 6872 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6873 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 6874 inode->i_mtime = inode->i_ctime = current_time(inode);
4209ae12 6875 err = ext4_mark_inode_dirty(handle, inode);
21f97697 6876 ext4_journal_stop(handle);
957153fc
JK		 6877out_unlock:
6878 inode_unlock(inode);
6879out_put:
964edf66 6880 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 6881 iput(inode);
6882 return err;
21f97697 6883out:
ca0e05e4
DM		 6884 return dquot_quota_off(sb, type);
6885}
6886
ac27a0ec
DK		 6887/* Read data from quotafile - avoid pagecache and such because we cannot afford
6888 * acquiring the locks... As quota files are never truncated and quota code
25985edc 6889 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 6890 * we don't have to be afraid of races */
617ba13b 6891static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 6892 size_t len, loff_t off)
6893{
6894 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 6895 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 6896 int offset = off & (sb->s_blocksize - 1);
6897 int tocopy;
6898 size_t toread;
6899 struct buffer_head *bh;
6900 loff_t i_size = i_size_read(inode);
6901
6902 if (off > i_size)
6903 return 0;
6904 if (off+len > i_size)
6905 len = i_size-off;
6906 toread = len;
6907 while (toread > 0) {
6908 tocopy = sb->s_blocksize - offset < toread ?
6909 sb->s_blocksize - offset : toread;
1c215028
TT		 6910 bh = ext4_bread(NULL, inode, blk, 0);
6911 if (IS_ERR(bh))
6912 return PTR_ERR(bh);
ac27a0ec
DK		 6913 if (!bh) /* A hole? */
6914 memset(data, 0, tocopy);
6915 else
6916 memcpy(data, bh->b_data+offset, tocopy);
6917 brelse(bh);
6918 offset = 0;
6919 toread -= tocopy;
6920 data += tocopy;
6921 blk++;
6922 }
6923 return len;
6924}
6925
6926/* Write to quotafile (we know the transaction is already started and has
6927 * enough credits) */
617ba13b 6928static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 6929 const char *data, size_t len, loff_t off)
6930{
6931 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 6932 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4209ae12 6933 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
c5e298ae 6934 int retries = 0;
ac27a0ec
DK		 6935 struct buffer_head *bh;
6936 handle_t *handle = journal_current_handle();
6937
380a0091 6938 if (!handle) {
b31e1552
ES		 6939 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6940 " cancelled because transaction is not started",
9c3013e9
JK		 6941 (unsigned long long)off, (unsigned long long)len);
6942 return -EIO;
6943 }
67eeb568
DM		 6944 /*
6945 * Since we account only one data block in transaction credits,
6946 * then it is impossible to cross a block boundary.
6947 */
6948 if (sb->s_blocksize - offset < len) {
6949 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6950 " cancelled because not block aligned",
6951 (unsigned long long)off, (unsigned long long)len);
6952 return -EIO;
6953 }
6954
c5e298ae
TT		 6955 do {
6956 bh = ext4_bread(handle, inode, blk,
6957 EXT4_GET_BLOCKS_CREATE |
6958 EXT4_GET_BLOCKS_METADATA_NOFAIL);
45586c70 6959 } while (PTR_ERR(bh) == -ENOSPC &&
c5e298ae 6960 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 6961 if (IS_ERR(bh))
6962 return PTR_ERR(bh);
67eeb568
DM		 6963 if (!bh)
6964 goto out;
5d601255 6965 BUFFER_TRACE(bh, "get write access");
188c299e 6966 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
62d2b5f2
JK		 6967 if (err) {
6968 brelse(bh);
1c215028 6969 return err;
ac27a0ec 6970 }
67eeb568
DM		 6971 lock_buffer(bh);
6972 memcpy(bh->b_data+offset, data, len);
6973 flush_dcache_page(bh->b_page);
6974 unlock_buffer(bh);
62d2b5f2 6975 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 6976 brelse(bh);
ac27a0ec 6977out:
67eeb568
DM		 6978 if (inode->i_size < off + len) {
6979 i_size_write(inode, off + len);
617ba13b 6980 EXT4_I(inode)->i_disksize = inode->i_size;
4209ae12
HS		 6981 err2 = ext4_mark_inode_dirty(handle, inode);
6982 if (unlikely(err2 && !err))
6983 err = err2;
ac27a0ec 6984 }
4209ae12 6985 return err ? err : len;
ac27a0ec 6986}
ac27a0ec
DK		 6987#endif
6988
c290ea01 6989#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 6990static inline void register_as_ext2(void)
6991{
6992 int err = register_filesystem(&ext2_fs_type);
6993 if (err)
6994 printk(KERN_WARNING
6995 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6996}
6997
6998static inline void unregister_as_ext2(void)
6999{
7000 unregister_filesystem(&ext2_fs_type);
7001}
2035e776
TT		 7002
7003static inline int ext2_feature_set_ok(struct super_block *sb)
7004{
e2b911c5 7005 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 7006 return 0;
bc98a42c 7007 if (sb_rdonly(sb))
2035e776 7008 return 1;
e2b911c5 7009 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 7010 return 0;
7011 return 1;
7012}
24b58424
TT		 7013#else
7014static inline void register_as_ext2(void) { }
7015static inline void unregister_as_ext2(void) { }
2035e776 7016static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 7017#endif
7018
24b58424
TT		 7019static inline void register_as_ext3(void)
7020{
7021 int err = register_filesystem(&ext3_fs_type);
7022 if (err)
7023 printk(KERN_WARNING
7024 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7025}
7026
7027static inline void unregister_as_ext3(void)
7028{
7029 unregister_filesystem(&ext3_fs_type);
7030}
2035e776
TT		 7031
7032static inline int ext3_feature_set_ok(struct super_block *sb)
7033{
e2b911c5 7034 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 7035 return 0;
e2b911c5 7036 if (!ext4_has_feature_journal(sb))
2035e776 7037 return 0;
bc98a42c 7038 if (sb_rdonly(sb))
2035e776 7039 return 1;
e2b911c5 7040 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 7041 return 0;
7042 return 1;
7043}
24b58424 7044
03010a33 7045static struct file_system_type ext4_fs_type = {
cebe85d5
LC		 7046 .owner = THIS_MODULE,
7047 .name = "ext4",
7048 .init_fs_context = ext4_init_fs_context,
7049 .parameters = ext4_param_specs,
7050 .kill_sb = kill_block_super,
7051 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
03010a33 7052};
7f78e035 7053MODULE_ALIAS_FS("ext4");
03010a33 7054
e9e3bcec
ES		 7055/* Shared across all ext4 file systems */
7056wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 7057
5dabfc78 7058static int __init ext4_init_fs(void)
ac27a0ec 7059{
e9e3bcec 7060 int i, err;
c9de560d 7061
e294a537 7062 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8 7063 ext4_li_info = NULL;
07c0c5d8 7064
9a4c8019 7065 /* Build-time check for flags consistency */
12e9b892 7066 ext4_check_flag_values();
e9e3bcec 7067
e142d052 7068 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 7069 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 7070
51865fda 7071 err = ext4_init_es();
6fd058f7
TT		 7072 if (err)
7073 return err;
51865fda 7074
1dc0aa46 7075 err = ext4_init_pending();
22cfe4b4
EB		 7076 if (err)
7077 goto out7;
7078
7079 err = ext4_init_post_read_processing();
1dc0aa46
EW		 7080 if (err)
7081 goto out6;
7082
51865fda
ZL		 7083 err = ext4_init_pageio();
7084 if (err)
b5799018 7085 goto out5;
51865fda 7086
5dabfc78 7087 err = ext4_init_system_zone();
bd2d0210 7088 if (err)
b5799018 7089 goto out4;
857ac889 7090
b5799018 7091 err = ext4_init_sysfs();
dd68314c 7092 if (err)
b5799018 7093 goto out3;
857ac889 7094
5dabfc78 7095 err = ext4_init_mballoc();
c9de560d
AT		 7096 if (err)
7097 goto out2;
ac27a0ec
DK		 7098 err = init_inodecache();
7099 if (err)
7100 goto out1;
aa75f4d3
HS		 7101
7102 err = ext4_fc_init_dentry_cache();
7103 if (err)
7104 goto out05;
7105
24b58424 7106 register_as_ext3();
2035e776 7107 register_as_ext2();
03010a33 7108 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 7109 if (err)
7110 goto out;
bfff6873 7111
ac27a0ec
DK		 7112 return 0;
7113out:
24b58424
TT		 7114 unregister_as_ext2();
7115 unregister_as_ext3();
ab047d51 7116 ext4_fc_destroy_dentry_cache();
aa75f4d3 7117out05:
ac27a0ec
DK		 7118 destroy_inodecache();
7119out1:
5dabfc78 7120 ext4_exit_mballoc();
9c191f70 7121out2:
b5799018
TT		 7122 ext4_exit_sysfs();
7123out3:
5dabfc78 7124 ext4_exit_system_zone();
b5799018 7125out4:
5dabfc78 7126 ext4_exit_pageio();
b5799018 7127out5:
22cfe4b4 7128 ext4_exit_post_read_processing();
1dc0aa46 7129out6:
22cfe4b4
EB		 7130 ext4_exit_pending();
7131out7:
51865fda
ZL		 7132 ext4_exit_es();
7133
ac27a0ec
DK		 7134 return err;
7135}
7136
5dabfc78 7137static void __exit ext4_exit_fs(void)
ac27a0ec 7138{
bfff6873 7139 ext4_destroy_lazyinit_thread();
24b58424
TT		 7140 unregister_as_ext2();
7141 unregister_as_ext3();
03010a33 7142 unregister_filesystem(&ext4_fs_type);
ab047d51 7143 ext4_fc_destroy_dentry_cache();
ac27a0ec 7144 destroy_inodecache();
5dabfc78 7145 ext4_exit_mballoc();
b5799018 7146 ext4_exit_sysfs();
5dabfc78
TT		 7147 ext4_exit_system_zone();
7148 ext4_exit_pageio();
22cfe4b4 7149 ext4_exit_post_read_processing();
dd12ed14 7150 ext4_exit_es();
1dc0aa46 7151 ext4_exit_pending();
ac27a0ec
DK		 7152}
7153
7154MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 7155MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 7156MODULE_LICENSE("GPL");
7ef79ad5 7157MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 7158module_init(ext4_init_fs)
7159module_exit(ext4_exit_fs)
Michael's Linux tree.