projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'ext4_for_linus_stable' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux-2.6-block.git] / fs / ext4 / super.c
CommitLineData
f5166768 1// SPDX-License-Identifier: GPL-2.0
ac27a0ec 2/*
617ba13b 3 * linux/fs/ext4/super.c
ac27a0ec
DK		 4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/fs/minix/inode.c
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 *
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
18 */
19
20#include <linux/module.h>
21#include <linux/string.h>
22#include <linux/fs.h>
23#include <linux/time.h>
c5ca7c76 24#include <linux/vmalloc.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
66114cad 28#include <linux/backing-dev.h>
ac27a0ec 29#include <linux/parser.h>
ac27a0ec 30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ef510424 41#include <linux/dax.h>
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{
0b73284c
ZY		 208 lock_buffer(bh);
209 if (!wait) {
fa491b14 210 ext4_read_bh_nowait(bh, op_flags, NULL);
211 return 0;
212 }
0b73284c 213 return ext4_read_bh(bh, op_flags, NULL);
fa491b14 214}
215
fb265c9c 216/*
8394a6ab 217 * This works like __bread_gfp() except it uses ERR_PTR for error
fb265c9c
TT		 218 * returns. Currently with sb_bread it's impossible to distinguish
219 * between ENOMEM and EIO situations (since both result in a NULL
220 * return.
221 */
8394a6ab 222static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
67c0f556
BVA		 223 sector_t block,
224 blk_opf_t op_flags, gfp_t gfp)
fb265c9c 225{
2d069c08 226 struct buffer_head *bh;
227 int ret;
fb265c9c 228
8394a6ab 229 bh = sb_getblk_gfp(sb, block, gfp);
fb265c9c
TT		 230 if (bh == NULL)
231 return ERR_PTR(-ENOMEM);
cf2834a5 232 if (ext4_buffer_uptodate(bh))
fb265c9c 233 return bh;
2d069c08 234
235 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
236 if (ret) {
237 put_bh(bh);
238 return ERR_PTR(ret);
239 }
240 return bh;
fb265c9c
TT		 241}
242
8394a6ab 243struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
67c0f556 244 blk_opf_t op_flags)
8394a6ab 245{
246 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
247}
248
249struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
250 sector_t block)
251{
252 return __ext4_sb_bread_gfp(sb, block, 0, 0);
253}
254
5df1d412 255void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
256{
257 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
258
259 if (likely(bh)) {
0b73284c
ZY		 260 if (trylock_buffer(bh))
261 ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL);
5df1d412 262 brelse(bh);
263 }
fb265c9c
TT		 264}
265
d25425f8
DW		 266static int ext4_verify_csum_type(struct super_block *sb,
267 struct ext4_super_block *es)
268{
e2b911c5 269 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 270 return 1;
271
272 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
273}
274
bbc605cd
LC		 275__le32 ext4_superblock_csum(struct super_block *sb,
276 struct ext4_super_block *es)
a9c47317
DW		 277{
278 struct ext4_sb_info *sbi = EXT4_SB(sb);
279 int offset = offsetof(struct ext4_super_block, s_checksum);
280 __u32 csum;
281
282 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
283
284 return cpu_to_le32(csum);
285}
286
c197855e
SH		 287static int ext4_superblock_csum_verify(struct super_block *sb,
288 struct ext4_super_block *es)
a9c47317 289{
9aa5d32b 290 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 291 return 1;
292
293 return es->s_checksum == ext4_superblock_csum(sb, es);
294}
295
06db49e6 296void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 297{
06db49e6
TT		 298 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
299
9aa5d32b 300 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 301 return;
302
303 es->s_checksum = ext4_superblock_csum(sb, es);
304}
305
8fadc143
AR		 306ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
307 struct ext4_group_desc *bg)
bd81d8ee 308{
3a14589c 309 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 310 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 311 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 312}
313
8fadc143
AR		 314ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
315 struct ext4_group_desc *bg)
bd81d8ee 316{
5272f837 317 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 318 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 319 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 320}
321
8fadc143
AR		 322ext4_fsblk_t ext4_inode_table(struct super_block *sb,
323 struct ext4_group_desc *bg)
bd81d8ee 324{
5272f837 325 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 326 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 327 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 328}
329
021b65bb
TT		 330__u32 ext4_free_group_clusters(struct super_block *sb,
331 struct ext4_group_desc *bg)
560671a0
AK		 332{
333 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
334 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 335 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 336}
337
338__u32 ext4_free_inodes_count(struct super_block *sb,
339 struct ext4_group_desc *bg)
340{
341 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
342 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 343 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 344}
345
346__u32 ext4_used_dirs_count(struct super_block *sb,
347 struct ext4_group_desc *bg)
348{
349 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
350 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 351 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 352}
353
354__u32 ext4_itable_unused_count(struct super_block *sb,
355 struct ext4_group_desc *bg)
356{
357 return le16_to_cpu(bg->bg_itable_unused_lo) |
358 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 359 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 360}
361
8fadc143
AR		 362void ext4_block_bitmap_set(struct super_block *sb,
363 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 364{
3a14589c 365 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 366 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
367 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 368}
369
8fadc143
AR		 370void ext4_inode_bitmap_set(struct super_block *sb,
371 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 372{
5272f837 373 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 374 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
375 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 376}
377
8fadc143
AR		 378void ext4_inode_table_set(struct super_block *sb,
379 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 380{
5272f837 381 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 382 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
383 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 384}
385
021b65bb
TT		 386void ext4_free_group_clusters_set(struct super_block *sb,
387 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 388{
389 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
390 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
391 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
392}
393
394void ext4_free_inodes_set(struct super_block *sb,
395 struct ext4_group_desc *bg, __u32 count)
396{
397 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
398 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
399 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
400}
401
402void ext4_used_dirs_set(struct super_block *sb,
403 struct ext4_group_desc *bg, __u32 count)
404{
405 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
406 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
407 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
408}
409
410void ext4_itable_unused_set(struct super_block *sb,
411 struct ext4_group_desc *bg, __u32 count)
412{
413 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
414 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
415 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
416}
417
c92dc856 418static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
6a0678a7 419{
6a0678a7
AB		 420 now = clamp_val(now, 0, (1ull << 40) - 1);
421
422 *lo = cpu_to_le32(lower_32_bits(now));
423 *hi = upper_32_bits(now);
424}
425
426static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
427{
428 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
429}
430#define ext4_update_tstamp(es, tstamp) \
c92dc856
JK		 431 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
432 ktime_get_real_seconds())
6a0678a7
AB		 433#define ext4_get_tstamp(es, tstamp) \
434 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
d3d1faf6 435
bdfe0cbd
TT		 436/*
437 * The del_gendisk() function uninitializes the disk-specific data
438 * structures, including the bdi structure, without telling anyone
439 * else. Once this happens, any attempt to call mark_buffer_dirty()
440 * (for example, by ext4_commit_super), will cause a kernel OOPS.
441 * This is a kludge to prevent these oops until we can put in a proper
442 * hook in del_gendisk() to inform the VFS and file system layers.
443 */
444static int block_device_ejected(struct super_block *sb)
445{
446 struct inode *bd_inode = sb->s_bdev->bd_inode;
447 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
448
449 return bdi->dev == NULL;
450}
451
18aadd47
BJ		 452static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
453{
454 struct super_block *sb = journal->j_private;
455 struct ext4_sb_info *sbi = EXT4_SB(sb);
456 int error = is_journal_aborted(journal);
5d3ee208 457 struct ext4_journal_cb_entry *jce;
18aadd47 458
5d3ee208 459 BUG_ON(txn->t_state == T_FINISHED);
a0154344
DJ		 460
461 ext4_process_freed_data(sb, txn->t_tid);
462
18aadd47 463 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 464 while (!list_empty(&txn->t_private_list)) {
465 jce = list_entry(txn->t_private_list.next,
466 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 467 list_del_init(&jce->jce_list);
468 spin_unlock(&sbi->s_md_lock);
469 jce->jce_func(sb, jce, error);
470 spin_lock(&sbi->s_md_lock);
471 }
472 spin_unlock(&sbi->s_md_lock);
473}
1c13d5c0 474
afb585a9
MFO		 475/*
476 * This writepage callback for write_cache_pages()
477 * takes care of a few cases after page cleaning.
478 *
479 * write_cache_pages() already checks for dirty pages
480 * and calls clear_page_dirty_for_io(), which we want,
481 * to write protect the pages.
482 *
483 * However, we may have to redirty a page (see below.)
484 */
d585bdbe 485static int ext4_journalled_writepage_callback(struct folio *folio,
afb585a9
MFO		 486 struct writeback_control *wbc,
487 void *data)
488{
489 transaction_t *transaction = (transaction_t *) data;
490 struct buffer_head *bh, *head;
491 struct journal_head *jh;
492
d585bdbe 493 bh = head = folio_buffers(folio);
afb585a9
MFO		 494 do {
495 /*
496 * We have to redirty a page in these cases:
497 * 1) If buffer is dirty, it means the page was dirty because it
498 * contains a buffer that needs checkpointing. So the dirty bit
499 * needs to be preserved so that checkpointing writes the buffer
500 * properly.
501 * 2) If buffer is not part of the committing transaction
502 * (we may have just accidentally come across this buffer because
503 * inode range tracking is not exact) or if the currently running
504 * transaction already contains this buffer as well, dirty bit
505 * needs to be preserved so that the buffer gets writeprotected
506 * properly on running transaction's commit.
507 */
508 jh = bh2jh(bh);
509 if (buffer_dirty(bh) ||
510 (jh && (jh->b_transaction != transaction ||
511 jh->b_next_transaction))) {
d585bdbe 512 folio_redirty_for_writepage(wbc, folio);
afb585a9
MFO		 513 goto out;
514 }
515 } while ((bh = bh->b_this_page) != head);
516
517out:
518 return AOP_WRITEPAGE_ACTIVATE;
519}
520
521static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
522{
523 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
524 struct writeback_control wbc = {
525 .sync_mode = WB_SYNC_ALL,
526 .nr_to_write = LONG_MAX,
527 .range_start = jinode->i_dirty_start,
528 .range_end = jinode->i_dirty_end,
529 };
530
531 return write_cache_pages(mapping, &wbc,
532 ext4_journalled_writepage_callback,
533 jinode->i_transaction);
534}
535
536static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
537{
538 int ret;
539
540 if (ext4_should_journal_data(jinode->i_vfs_inode))
541 ret = ext4_journalled_submit_inode_data_buffers(jinode);
542 else
59205c8d 543 ret = ext4_normal_submit_inode_data_buffers(jinode);
afb585a9
MFO		 544 return ret;
545}
546
547static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
548{
549 int ret = 0;
550
551 if (!ext4_should_journal_data(jinode->i_vfs_inode))
552 ret = jbd2_journal_finish_inode_data_buffers(jinode);
553
554 return ret;
555}
556
1dc1097f
JK		 557static bool system_going_down(void)
558{
559 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
560 || system_state == SYSTEM_RESTART;
561}
562
02a7780e
JK		 563struct ext4_err_translation {
564 int code;
565 int errno;
566};
567
568#define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
569
570static struct ext4_err_translation err_translation[] = {
571 EXT4_ERR_TRANSLATE(EIO),
572 EXT4_ERR_TRANSLATE(ENOMEM),
573 EXT4_ERR_TRANSLATE(EFSBADCRC),
574 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
575 EXT4_ERR_TRANSLATE(ENOSPC),
576 EXT4_ERR_TRANSLATE(ENOKEY),
577 EXT4_ERR_TRANSLATE(EROFS),
578 EXT4_ERR_TRANSLATE(EFBIG),
579 EXT4_ERR_TRANSLATE(EEXIST),
580 EXT4_ERR_TRANSLATE(ERANGE),
581 EXT4_ERR_TRANSLATE(EOVERFLOW),
582 EXT4_ERR_TRANSLATE(EBUSY),
583 EXT4_ERR_TRANSLATE(ENOTDIR),
584 EXT4_ERR_TRANSLATE(ENOTEMPTY),
585 EXT4_ERR_TRANSLATE(ESHUTDOWN),
586 EXT4_ERR_TRANSLATE(EFAULT),
587};
588
589static int ext4_errno_to_code(int errno)
590{
591 int i;
592
593 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
594 if (err_translation[i].errno == errno)
595 return err_translation[i].code;
596 return EXT4_ERR_UNKNOWN;
597}
598
2d01ddc8
JK		 599static void save_error_info(struct super_block *sb, int error,
600 __u32 ino, __u64 block,
601 const char *func, unsigned int line)
40676623 602{
c92dc856 603 struct ext4_sb_info *sbi = EXT4_SB(sb);
40676623 604
02a7780e
JK		 605 /* We default to EFSCORRUPTED error... */
606 if (error == 0)
607 error = EFSCORRUPTED;
c92dc856
JK		 608
609 spin_lock(&sbi->s_error_lock);
610 sbi->s_add_error_count++;
611 sbi->s_last_error_code = error;
612 sbi->s_last_error_line = line;
613 sbi->s_last_error_ino = ino;
614 sbi->s_last_error_block = block;
615 sbi->s_last_error_func = func;
616 sbi->s_last_error_time = ktime_get_real_seconds();
617 if (!sbi->s_first_error_time) {
618 sbi->s_first_error_code = error;
619 sbi->s_first_error_line = line;
620 sbi->s_first_error_ino = ino;
621 sbi->s_first_error_block = block;
622 sbi->s_first_error_func = func;
623 sbi->s_first_error_time = sbi->s_last_error_time;
624 }
625 spin_unlock(&sbi->s_error_lock);
40676623
JK		 626}
627
ac27a0ec
DK		 628/* Deal with the reporting of failure conditions on a filesystem such as
629 * inconsistencies detected or read IO failures.
630 *
631 * On ext2, we can store the error state of the filesystem in the
617ba13b 632 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 633 * write ordering constraints on the superblock which prevent us from
634 * writing it out straight away; and given that the journal is about to
635 * be aborted, we can't rely on the current, or future, transactions to
636 * write out the superblock safely.
637 *
dab291af 638 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 639 * the journal instead. On recovery, the journal will complain about
ac27a0ec 640 * that error until we've noted it down and cleared it.
014c9caa
JK		 641 *
642 * If force_ro is set, we unconditionally force the filesystem into an
643 * ABORT|READONLY state, unless the error response on the fs has been set to
644 * panic in which case we take the easy way out and panic immediately. This is
645 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
646 * at a critical moment in log management.
ac27a0ec 647 */
e789ca0c
JK		 648static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
649 __u32 ino, __u64 block,
650 const char *func, unsigned int line)
ac27a0ec 651{
b08070ec 652 journal_t *journal = EXT4_SB(sb)->s_journal;
2d01ddc8 653 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
b08070ec 654
e789ca0c 655 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
327eaf73
TT		 656 if (test_opt(sb, WARN_ON_ERROR))
657 WARN_ON_ONCE(1);
658
2d01ddc8
JK		 659 if (!continue_fs && !sb_rdonly(sb)) {
660 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
661 if (journal)
662 jbd2_journal_abort(journal, -EIO);
663 }
664
665 if (!bdev_read_only(sb->s_bdev)) {
e789ca0c 666 save_error_info(sb, error, ino, block, func, line);
2d01ddc8
JK		 667 /*
668 * In case the fs should keep running, we need to writeout
669 * superblock through the journal. Due to lock ordering
670 * constraints, it may not be safe to do it right here so we
671 * defer superblock flushing to a workqueue.
672 */
bb9464e0 673 if (continue_fs && journal)
2d01ddc8
JK		 674 schedule_work(&EXT4_SB(sb)->s_error_work);
675 else
676 ext4_commit_super(sb);
677 }
e789ca0c 678
1dc1097f
JK		 679 /*
680 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
681 * could panic during 'reboot -f' as the underlying device got already
682 * disabled.
683 */
014c9caa 684 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
617ba13b 685 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 686 sb->s_id);
4327ba52 687 }
ac2f7ca5
YB		 688
689 if (sb_rdonly(sb) || continue_fs)
690 return;
691
014c9caa
JK		 692 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
693 /*
694 * Make sure updated value of ->s_mount_flags will be visible before
695 * ->s_flags update
696 */
697 smp_wmb();
698 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 699}
700
c92dc856
JK		 701static void flush_stashed_error_work(struct work_struct *work)
702{
703 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
704 s_error_work);
2d01ddc8
JK		 705 journal_t *journal = sbi->s_journal;
706 handle_t *handle;
c92dc856 707
2d01ddc8
JK		 708 /*
709 * If the journal is still running, we have to write out superblock
710 * through the journal to avoid collisions of other journalled sb
711 * updates.
712 *
713 * We use directly jbd2 functions here to avoid recursing back into
714 * ext4 error handling code during handling of previous errors.
715 */
716 if (!sb_rdonly(sbi->s_sb) && journal) {
558d6450 717 struct buffer_head *sbh = sbi->s_sbh;
2d01ddc8
JK		 718 handle = jbd2_journal_start(journal, 1);
719 if (IS_ERR(handle))
720 goto write_directly;
558d6450 721 if (jbd2_journal_get_write_access(handle, sbh)) {
2d01ddc8
JK		 722 jbd2_journal_stop(handle);
723 goto write_directly;
724 }
725 ext4_update_super(sbi->s_sb);
558d6450
YB		 726 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
727 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
728 "superblock detected");
729 clear_buffer_write_io_error(sbh);
730 set_buffer_uptodate(sbh);
731 }
732
733 if (jbd2_journal_dirty_metadata(handle, sbh)) {
2d01ddc8
JK		 734 jbd2_journal_stop(handle);
735 goto write_directly;
736 }
737 jbd2_journal_stop(handle);
d578b994 738 ext4_notify_error_sysfs(sbi);
2d01ddc8
JK		 739 return;
740 }
741write_directly:
742 /*
743 * Write through journal failed. Write sb directly to get error info
744 * out and hope for the best.
745 */
4392fbc4 746 ext4_commit_super(sbi->s_sb);
d578b994 747 ext4_notify_error_sysfs(sbi);
ac27a0ec
DK		 748}
749
efbed4dc
TT		 750#define ext4_error_ratelimit(sb) \
751 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
752 "EXT4-fs error")
753
12062ddd 754void __ext4_error(struct super_block *sb, const char *function,
014c9caa 755 unsigned int line, bool force_ro, int error, __u64 block,
54d3adbc 756 const char *fmt, ...)
ac27a0ec 757{
0ff2ea7d 758 struct va_format vaf;
ac27a0ec
DK		 759 va_list args;
760
0db1ff22
TT		 761 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
762 return;
763
ccf0f32a 764 trace_ext4_error(sb, function, line);
efbed4dc
TT		 765 if (ext4_error_ratelimit(sb)) {
766 va_start(args, fmt);
767 vaf.fmt = fmt;
768 vaf.va = &args;
769 printk(KERN_CRIT
770 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
771 sb->s_id, function, line, current->comm, &vaf);
772 va_end(args);
773 }
9a089b21
GKB		 774 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
775
e789ca0c 776 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
ac27a0ec
DK		 777}
778
e7c96e8e 779void __ext4_error_inode(struct inode *inode, const char *function,
54d3adbc 780 unsigned int line, ext4_fsblk_t block, int error,
e7c96e8e 781 const char *fmt, ...)
273df556
FM		 782{
783 va_list args;
f7c21177 784 struct va_format vaf;
273df556 785
0db1ff22
TT		 786 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
787 return;
788
ccf0f32a 789 trace_ext4_error(inode->i_sb, function, line);
efbed4dc
TT		 790 if (ext4_error_ratelimit(inode->i_sb)) {
791 va_start(args, fmt);
792 vaf.fmt = fmt;
793 vaf.va = &args;
794 if (block)
795 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
796 "inode #%lu: block %llu: comm %s: %pV\n",
797 inode->i_sb->s_id, function, line, inode->i_ino,
798 block, current->comm, &vaf);
799 else
800 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
801 "inode #%lu: comm %s: %pV\n",
802 inode->i_sb->s_id, function, line, inode->i_ino,
803 current->comm, &vaf);
804 va_end(args);
805 }
9a089b21
GKB		 806 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
807
e789ca0c
JK		 808 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
809 function, line);
273df556
FM		 810}
811
e7c96e8e
JP		 812void __ext4_error_file(struct file *file, const char *function,
813 unsigned int line, ext4_fsblk_t block,
814 const char *fmt, ...)
273df556
FM		 815{
816 va_list args;
f7c21177 817 struct va_format vaf;
496ad9aa 818 struct inode *inode = file_inode(file);
273df556
FM		 819 char pathname[80], *path;
820
0db1ff22
TT		 821 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
822 return;
823
ccf0f32a 824 trace_ext4_error(inode->i_sb, function, line);
efbed4dc 825 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 826 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 827 if (IS_ERR(path))
828 path = "(unknown)";
829 va_start(args, fmt);
830 vaf.fmt = fmt;
831 vaf.va = &args;
832 if (block)
833 printk(KERN_CRIT
834 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
835 "block %llu: comm %s: path %s: %pV\n",
836 inode->i_sb->s_id, function, line, inode->i_ino,
837 block, current->comm, path, &vaf);
838 else
839 printk(KERN_CRIT
840 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
841 "comm %s: path %s: %pV\n",
842 inode->i_sb->s_id, function, line, inode->i_ino,
843 current->comm, path, &vaf);
844 va_end(args);
845 }
9a089b21
GKB		 846 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
847
e789ca0c
JK		 848 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
849 function, line);
273df556
FM		 850}
851
722887dd
TT		 852const char *ext4_decode_error(struct super_block *sb, int errno,
853 char nbuf[16])
ac27a0ec
DK		 854{
855 char *errstr = NULL;
856
857 switch (errno) {
6a797d27
DW		 858 case -EFSCORRUPTED:
859 errstr = "Corrupt filesystem";
860 break;
861 case -EFSBADCRC:
862 errstr = "Filesystem failed CRC";
863 break;
ac27a0ec
DK		 864 case -EIO:
865 errstr = "IO failure";
866 break;
867 case -ENOMEM:
868 errstr = "Out of memory";
869 break;
870 case -EROFS:
78f1ddbb
TT		 871 if (!sb || (EXT4_SB(sb)->s_journal &&
872 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 873 errstr = "Journal has aborted";
874 else
875 errstr = "Readonly filesystem";
876 break;
877 default:
878 /* If the caller passed in an extra buffer for unknown
879 * errors, textualise them now. Else we just return
880 * NULL. */
881 if (nbuf) {
882 /* Check for truncated error codes... */
883 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
884 errstr = nbuf;
885 }
886 break;
887 }
888
889 return errstr;
890}
891
617ba13b 892/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 893 * automatically and invokes the appropriate error response. */
894
c398eda0
TT		 895void __ext4_std_error(struct super_block *sb, const char *function,
896 unsigned int line, int errno)
ac27a0ec
DK		 897{
898 char nbuf[16];
899 const char *errstr;
900
0db1ff22
TT		 901 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
902 return;
903
ac27a0ec
DK		 904 /* Special case: if the error is EROFS, and we're not already
905 * inside a transaction, then there's really no point in logging
906 * an error. */
bc98a42c 907 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
ac27a0ec
DK		 908 return;
909
efbed4dc
TT		 910 if (ext4_error_ratelimit(sb)) {
911 errstr = ext4_decode_error(sb, errno, nbuf);
912 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
913 sb->s_id, function, line, errstr);
914 }
9a089b21 915 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
ac27a0ec 916
e789ca0c 917 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
ac27a0ec
DK		 918}
919
e7c96e8e
JP		 920void __ext4_msg(struct super_block *sb,
921 const char *prefix, const char *fmt, ...)
b31e1552 922{
0ff2ea7d 923 struct va_format vaf;
b31e1552
ES		 924 va_list args;
925
da812f61
LC		 926 if (sb) {
927 atomic_inc(&EXT4_SB(sb)->s_msg_count);
928 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
929 "EXT4-fs"))
930 return;
931 }
efbed4dc 932
b31e1552 933 va_start(args, fmt);
0ff2ea7d
JP		 934 vaf.fmt = fmt;
935 vaf.va = &args;
da812f61
LC		 936 if (sb)
937 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
938 else
939 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
b31e1552
ES		 940 va_end(args);
941}
942
1cf006ed
DM		 943static int ext4_warning_ratelimit(struct super_block *sb)
944{
945 atomic_inc(&EXT4_SB(sb)->s_warning_count);
946 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
947 "EXT4-fs warning");
948}
b03a2f7e 949
12062ddd 950void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 951 unsigned int line, const char *fmt, ...)
ac27a0ec 952{
0ff2ea7d 953 struct va_format vaf;
ac27a0ec
DK		 954 va_list args;
955
b03a2f7e 956 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 957 return;
958
ac27a0ec 959 va_start(args, fmt);
0ff2ea7d
JP		 960 vaf.fmt = fmt;
961 vaf.va = &args;
962 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
963 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 964 va_end(args);
965}
966
b03a2f7e
AD		 967void __ext4_warning_inode(const struct inode *inode, const char *function,
968 unsigned int line, const char *fmt, ...)
969{
970 struct va_format vaf;
971 va_list args;
972
973 if (!ext4_warning_ratelimit(inode->i_sb))
974 return;
975
976 va_start(args, fmt);
977 vaf.fmt = fmt;
978 vaf.va = &args;
979 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
980 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
981 function, line, inode->i_ino, current->comm, &vaf);
982 va_end(args);
983}
984
e29136f8
TT		 985void __ext4_grp_locked_error(const char *function, unsigned int line,
986 struct super_block *sb, ext4_group_t grp,
987 unsigned long ino, ext4_fsblk_t block,
988 const char *fmt, ...)
5d1b1b3f
AK		 989__releases(bitlock)
990__acquires(bitlock)
991{
0ff2ea7d 992 struct va_format vaf;
5d1b1b3f 993 va_list args;
5d1b1b3f 994
0db1ff22
TT		 995 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
996 return;
997
ccf0f32a 998 trace_ext4_error(sb, function, line);
efbed4dc
TT		 999 if (ext4_error_ratelimit(sb)) {
1000 va_start(args, fmt);
1001 vaf.fmt = fmt;
1002 vaf.va = &args;
1003 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1004 sb->s_id, function, line, grp);
1005 if (ino)
1006 printk(KERN_CONT "inode %lu: ", ino);
1007 if (block)
1008 printk(KERN_CONT "block %llu:",
1009 (unsigned long long) block);
1010 printk(KERN_CONT "%pV\n", &vaf);
1011 va_end(args);
1012 }
5d1b1b3f
AK		 1013
1014 if (test_opt(sb, ERRORS_CONT)) {
c92dc856
JK		 1015 if (test_opt(sb, WARN_ON_ERROR))
1016 WARN_ON_ONCE(1);
e789ca0c 1017 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2d01ddc8
JK		 1018 if (!bdev_read_only(sb->s_bdev)) {
1019 save_error_info(sb, EFSCORRUPTED, ino, block, function,
1020 line);
e789ca0c 1021 schedule_work(&EXT4_SB(sb)->s_error_work);
2d01ddc8 1022 }
5d1b1b3f
AK		 1023 return;
1024 }
1025 ext4_unlock_group(sb, grp);
e789ca0c 1026 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
5d1b1b3f
AK		 1027 /*
1028 * We only get here in the ERRORS_RO case; relocking the group
1029 * may be dangerous, but nothing bad will happen since the
1030 * filesystem will have already been marked read/only and the
1031 * journal has been aborted. We return 1 as a hint to callers
1032 * who might what to use the return value from
25985edc 1033 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 1034 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1035 * aggressively from the ext4 function in question, with a
1036 * more appropriate error code.
1037 */
1038 ext4_lock_group(sb, grp);
1039 return;
1040}
1041
db79e6d1
WS		 1042void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1043 ext4_group_t group,
1044 unsigned int flags)
1045{
1046 struct ext4_sb_info *sbi = EXT4_SB(sb);
1047 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1048 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
9af0b3d1
WS		 1049 int ret;
1050
1051 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1052 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1053 &grp->bb_state);
1054 if (!ret)
1055 percpu_counter_sub(&sbi->s_freeclusters_counter,
1056 grp->bb_free);
db79e6d1
WS		 1057 }
1058
9af0b3d1
WS		 1059 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1060 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1061 &grp->bb_state);
1062 if (!ret && gdp) {
db79e6d1
WS		 1063 int count;
1064
1065 count = ext4_free_inodes_count(sb, gdp);
1066 percpu_counter_sub(&sbi->s_freeinodes_counter,
1067 count);
1068 }
db79e6d1
WS		 1069 }
1070}
1071
617ba13b 1072void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 1073{
617ba13b 1074 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 1075
617ba13b 1076 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 1077 return;
1078
12062ddd 1079 ext4_warning(sb,
ac27a0ec
DK		 1080 "updating to rev %d because of new feature flag, "
1081 "running e2fsck is recommended",
617ba13b 1082 EXT4_DYNAMIC_REV);
ac27a0ec 1083
617ba13b
MC		 1084 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1085 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1086 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 1087 /* leave es->s_feature_*compat flags alone */
1088 /* es->s_uuid will be set by e2fsck if empty */
1089
1090 /*
1091 * The rest of the superblock fields should be zero, and if not it
1092 * means they are likely already in use, so leave them alone. We
1093 * can leave it up to e2fsck to clean up any inconsistencies there.
1094 */
1095}
1096
1097/*
1098 * Open the external journal device
1099 */
b31e1552 1100static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 1101{
1102 struct block_device *bdev;
ac27a0ec 1103
d4d77629 1104 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 1105 if (IS_ERR(bdev))
1106 goto fail;
1107 return bdev;
1108
1109fail:
ea3edd4d
CH		 1110 ext4_msg(sb, KERN_ERR,
1111 "failed to open journal device unknown-block(%u,%u) %ld",
1112 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
ac27a0ec
DK		 1113 return NULL;
1114}
1115
1116/*
1117 * Release the journal device
1118 */
4385bab1 1119static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 1120{
4385bab1 1121 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 1122}
1123
4385bab1 1124static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 1125{
1126 struct block_device *bdev;
ee7ed3aa 1127 bdev = sbi->s_journal_bdev;
ac27a0ec 1128 if (bdev) {
4385bab1 1129 ext4_blkdev_put(bdev);
ee7ed3aa 1130 sbi->s_journal_bdev = NULL;
ac27a0ec 1131 }
ac27a0ec
DK		 1132}
1133
1134static inline struct inode *orphan_list_entry(struct list_head *l)
1135{
617ba13b 1136 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 1137}
1138
617ba13b 1139static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 1140{
1141 struct list_head *l;
1142
b31e1552
ES		 1143 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1144 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 1145
1146 printk(KERN_ERR "sb_info orphan list:\n");
1147 list_for_each(l, &sbi->s_orphan) {
1148 struct inode *inode = orphan_list_entry(l);
1149 printk(KERN_ERR " "
1150 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1151 inode->i_sb->s_id, inode->i_ino, inode,
1152 inode->i_mode, inode->i_nlink,
1153 NEXT_ORPHAN(inode));
1154 }
1155}
1156
957153fc
JK		 1157#ifdef CONFIG_QUOTA
1158static int ext4_quota_off(struct super_block *sb, int type);
1159
1160static inline void ext4_quota_off_umount(struct super_block *sb)
1161{
1162 int type;
1163
964edf66
JK		 1164 /* Use our quota_off function to clear inode flags etc. */
1165 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1166 ext4_quota_off(sb, type);
957153fc 1167}
33458eab
TT		 1168
1169/*
1170 * This is a helper function which is used in the mount/remount
1171 * codepaths (which holds s_umount) to fetch the quota file name.
1172 */
1173static inline char *get_qf_name(struct super_block *sb,
1174 struct ext4_sb_info *sbi,
1175 int type)
1176{
1177 return rcu_dereference_protected(sbi->s_qf_names[type],
1178 lockdep_is_held(&sb->s_umount));
1179}
957153fc
JK		 1180#else
1181static inline void ext4_quota_off_umount(struct super_block *sb)
1182{
1183}
1184#endif
1185
2b2d6d01 1186static void ext4_put_super(struct super_block *sb)
ac27a0ec 1187{
617ba13b
MC		 1188 struct ext4_sb_info *sbi = EXT4_SB(sb);
1189 struct ext4_super_block *es = sbi->s_es;
1d0c3924 1190 struct buffer_head **group_desc;
7c990728 1191 struct flex_groups **flex_groups;
97abd7d4 1192 int aborted = 0;
ef2cabf7 1193 int i, err;
ac27a0ec 1194
5e47868f
RH		 1195 /*
1196 * Unregister sysfs before destroying jbd2 journal.
1197 * Since we could still access attr_journal_task attribute via sysfs
1198 * path which could have sbi->s_journal->j_task as NULL
b98535d0
YB		 1199 * Unregister sysfs before flush sbi->s_error_work.
1200 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1201 * read metadata verify failed then will queue error work.
1202 * flush_stashed_error_work will call start_this_handle may trigger
1203 * BUG_ON.
5e47868f
RH		 1204 */
1205 ext4_unregister_sysfs(sb);
1206
4808cb5b 1207 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
bb0fbc78
LC		 1208 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1209 &sb->s_uuid);
4808cb5b 1210
b98535d0
YB		 1211 ext4_unregister_li_request(sb);
1212 ext4_quota_off_umount(sb);
1213
1214 flush_work(&sbi->s_error_work);
1215 destroy_workqueue(sbi->rsv_conversion_wq);
1216 ext4_release_orphan_info(sb);
1217
0390131b 1218 if (sbi->s_journal) {
97abd7d4 1219 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 1220 err = jbd2_journal_destroy(sbi->s_journal);
1221 sbi->s_journal = NULL;
878520ac 1222 if ((err < 0) && !aborted) {
54d3adbc 1223 ext4_abort(sb, -err, "Couldn't clean up the journal");
878520ac 1224 }
0390131b 1225 }
d4edac31 1226
d3922a77 1227 ext4_es_unregister_shrinker(sbi);
292a089d 1228 timer_shutdown_sync(&sbi->s_err_report);
d4edac31
JB		 1229 ext4_release_system_zone(sb);
1230 ext4_mb_release(sb);
1231 ext4_ext_release(sb);
d4edac31 1232
bc98a42c 1233 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 1234 ext4_clear_feature_journal_needs_recovery(sb);
02f310fc 1235 ext4_clear_feature_orphan_present(sb);
ac27a0ec 1236 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 1237 }
bc98a42c 1238 if (!sb_rdonly(sb))
4392fbc4 1239 ext4_commit_super(sb);
a8e25a83 1240
1d0c3924
TT		 1241 rcu_read_lock();
1242 group_desc = rcu_dereference(sbi->s_group_desc);
ac27a0ec 1243 for (i = 0; i < sbi->s_gdb_count; i++)
1d0c3924
TT		 1244 brelse(group_desc[i]);
1245 kvfree(group_desc);
7c990728
SJS		 1246 flex_groups = rcu_dereference(sbi->s_flex_groups);
1247 if (flex_groups) {
1248 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1249 kvfree(flex_groups[i]);
1250 kvfree(flex_groups);
1251 }
1d0c3924 1252 rcu_read_unlock();
57042651 1253 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 1254 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1255 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 1256 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
efc61345 1257 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
bbd55937 1258 percpu_free_rwsem(&sbi->s_writepages_rwsem);
ac27a0ec 1259#ifdef CONFIG_QUOTA
a2d4a646 1260 for (i = 0; i < EXT4_MAXQUOTAS; i++)
33458eab 1261 kfree(get_qf_name(sb, sbi, i));
ac27a0ec
DK		 1262#endif
1263
1264 /* Debugging code just in case the in-memory inode orphan list
1265 * isn't empty. The on-disk one can be non-empty if we've
1266 * detected an error and taken the fs readonly, but the
1267 * in-memory list had better be clean by this point. */
1268 if (!list_empty(&sbi->s_orphan))
1269 dump_orphan_list(sb, sbi);
837c23fb 1270 ASSERT(list_empty(&sbi->s_orphan));
ac27a0ec 1271
89d96a6f 1272 sync_blockdev(sb->s_bdev);
f98393a6 1273 invalidate_bdev(sb->s_bdev);
ee7ed3aa 1274 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
ac27a0ec
DK		 1275 /*
1276 * Invalidate the journal device's buffers. We don't want them
1277 * floating about in memory - the physical journal device may
1278 * hotswapped, and it breaks the `ro-after' testing code.
1279 */
ee7ed3aa
CX		 1280 sync_blockdev(sbi->s_journal_bdev);
1281 invalidate_bdev(sbi->s_journal_bdev);
617ba13b 1282 ext4_blkdev_remove(sbi);
ac27a0ec 1283 }
50c15df6
CX		 1284
1285 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1286 sbi->s_ea_inode_cache = NULL;
1287
1288 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1289 sbi->s_ea_block_cache = NULL;
1290
618f0031
PS		 1291 ext4_stop_mmpd(sbi);
1292
9060dd2c 1293 brelse(sbi->s_sbh);
ac27a0ec 1294 sb->s_fs_info = NULL;
3197ebdb
TT		 1295 /*
1296 * Now that we are completely done shutting down the
1297 * superblock, we need to actually destroy the kobject.
1298 */
3197ebdb
TT		 1299 kobject_put(&sbi->s_kobj);
1300 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 1301 if (sbi->s_chksum_driver)
1302 crypto_free_shash(sbi->s_chksum_driver);
705895b6 1303 kfree(sbi->s_blockgroup_lock);
8012b866 1304 fs_put_dax(sbi->s_daxdev, NULL);
ac4acb1f 1305 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5298d4bf 1306#if IS_ENABLED(CONFIG_UNICODE)
f8f4acb6 1307 utf8_unload(sb->s_encoding);
c83ad55e 1308#endif
ac27a0ec 1309 kfree(sbi);
ac27a0ec
DK		 1310}
1311
e18b890b 1312static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 1313
1314/*
1315 * Called inside transaction, so use GFP_NOFS
1316 */
617ba13b 1317static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 1318{
617ba13b 1319 struct ext4_inode_info *ei;
ac27a0ec 1320
fd60b288 1321 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 1322 if (!ei)
1323 return NULL;
0b8e58a1 1324
ee73f9a5 1325 inode_set_iversion(&ei->vfs_inode, 1);
7ea71af9 1326 ei->i_flags = 0;
202ee5df 1327 spin_lock_init(&ei->i_raw_lock);
c9de560d 1328 INIT_LIST_HEAD(&ei->i_prealloc_list);
27bc446e 1329 atomic_set(&ei->i_prealloc_active, 0);
c9de560d 1330 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 1331 ext4_es_init_tree(&ei->i_es_tree);
1332 rwlock_init(&ei->i_es_lock);
edaa53ca 1333 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 1334 ei->i_es_all_nr = 0;
edaa53ca 1335 ei->i_es_shk_nr = 0;
dd475925 1336 ei->i_es_shrink_lblk = 0;
d2a17637 1337 ei->i_reserved_data_blocks = 0;
d2a17637 1338 spin_lock_init(&(ei->i_block_reservation_lock));
1dc0aa46 1339 ext4_init_pending_tree(&ei->i_pending_tree);
a9e7f447
DM		 1340#ifdef CONFIG_QUOTA
1341 ei->i_reserved_quota = 0;
96c7e0d9 1342 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 1343#endif
8aefcd55 1344 ei->jinode = NULL;
2e8fa54e 1345 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 1346 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 1347 ei->i_sync_tid = 0;
1348 ei->i_datasync_tid = 0;
e27f41e1 1349 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 1350 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
aa75f4d3
HS		 1351 ext4_fc_init_inode(&ei->vfs_inode);
1352 mutex_init(&ei->i_fc_lock);
ac27a0ec
DK		 1353 return &ei->vfs_inode;
1354}
1355
7ff9c073
TT		 1356static int ext4_drop_inode(struct inode *inode)
1357{
1358 int drop = generic_drop_inode(inode);
1359
29b3692e
EB		 1360 if (!drop)
1361 drop = fscrypt_drop_inode(inode);
1362
7ff9c073
TT		 1363 trace_ext4_drop_inode(inode, drop);
1364 return drop;
1365}
1366
94053139 1367static void ext4_free_in_core_inode(struct inode *inode)
fa0d7e3d 1368{
2c58d548 1369 fscrypt_free_inode(inode);
aa75f4d3
HS		 1370 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1371 pr_warn("%s: inode %ld still in fc list",
1372 __func__, inode->i_ino);
1373 }
fa0d7e3d
NP		 1374 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1375}
1376
617ba13b 1377static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1378{
9f7dd93d 1379 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1380 ext4_msg(inode->i_sb, KERN_ERR,
1381 "Inode %lu (%p): orphan list check failed!",
1382 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1383 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1384 EXT4_I(inode), sizeof(struct ext4_inode_info),
1385 true);
1386 dump_stack();
1387 }
6fed8395
JX		 1388
1389 if (EXT4_I(inode)->i_reserved_data_blocks)
1390 ext4_msg(inode->i_sb, KERN_ERR,
1391 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1392 inode->i_ino, EXT4_I(inode),
1393 EXT4_I(inode)->i_reserved_data_blocks);
ac27a0ec
DK		 1394}
1395
51cc5068 1396static void init_once(void *foo)
ac27a0ec 1397{
c30365b9 1398 struct ext4_inode_info *ei = foo;
ac27a0ec 1399
a35afb83 1400 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1401 init_rwsem(&ei->xattr_sem);
0e855ac8 1402 init_rwsem(&ei->i_data_sem);
a35afb83 1403 inode_init_once(&ei->vfs_inode);
aa75f4d3 1404 ext4_fc_init_inode(&ei->vfs_inode);
ac27a0ec
DK		 1405}
1406
e67bc2b3 1407static int __init init_inodecache(void)
ac27a0ec 1408{
f8dd7c70
DW		 1409 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1410 sizeof(struct ext4_inode_info), 0,
1411 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1412 SLAB_ACCOUNT),
1413 offsetof(struct ext4_inode_info, i_data),
1414 sizeof_field(struct ext4_inode_info, i_data),
1415 init_once);
617ba13b 1416 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1417 return -ENOMEM;
1418 return 0;
1419}
1420
1421static void destroy_inodecache(void)
1422{
8c0a8537
KS		 1423 /*
1424 * Make sure all delayed rcu free inodes are flushed before we
1425 * destroy cache.
1426 */
1427 rcu_barrier();
617ba13b 1428 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1429}
1430
0930fcc1 1431void ext4_clear_inode(struct inode *inode)
ac27a0ec 1432{
aa75f4d3 1433 ext4_fc_del(inode);
0930fcc1 1434 invalidate_inode_buffers(inode);
dbd5768f 1435 clear_inode(inode);
27bc446e 1436 ext4_discard_preallocations(inode, 0);
51865fda 1437 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
f4c2d372 1438 dquot_drop(inode);
8aefcd55
TT		 1439 if (EXT4_I(inode)->jinode) {
1440 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1441 EXT4_I(inode)->jinode);
1442 jbd2_free_inode(EXT4_I(inode)->jinode);
1443 EXT4_I(inode)->jinode = NULL;
1444 }
3d204e24 1445 fscrypt_put_encryption_info(inode);
c93d8f88 1446 fsverity_cleanup_inode(inode);
ac27a0ec
DK		 1447}
1448
1b961ac0 1449static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1450 u64 ino, u32 generation)
ac27a0ec 1451{
ac27a0ec 1452 struct inode *inode;
ac27a0ec 1453
8a363970 1454 /*
ac27a0ec
DK		 1455 * Currently we don't know the generation for parent directory, so
1456 * a generation of 0 means "accept any"
1457 */
8a363970 1458 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1d1fe1ee
DH		 1459 if (IS_ERR(inode))
1460 return ERR_CAST(inode);
1461 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1462 iput(inode);
1463 return ERR_PTR(-ESTALE);
1464 }
1b961ac0
CH		 1465
1466 return inode;
1467}
1468
1469static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1470 int fh_len, int fh_type)
1b961ac0
CH		 1471{
1472 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1473 ext4_nfs_get_inode);
1474}
1475
1476static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1477 int fh_len, int fh_type)
1b961ac0
CH		 1478{
1479 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1480 ext4_nfs_get_inode);
ac27a0ec
DK		 1481}
1482
fde87268
TT		 1483static int ext4_nfs_commit_metadata(struct inode *inode)
1484{
1485 struct writeback_control wbc = {
1486 .sync_mode = WB_SYNC_ALL
1487 };
1488
1489 trace_ext4_nfs_commit_metadata(inode);
1490 return ext4_write_inode(inode, &wbc);
1491}
1492
ac27a0ec 1493#ifdef CONFIG_QUOTA
d6006186 1494static const char * const quotatypes[] = INITQFNAMES;
689c958c 1495#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1496
617ba13b
MC		 1497static int ext4_write_dquot(struct dquot *dquot);
1498static int ext4_acquire_dquot(struct dquot *dquot);
1499static int ext4_release_dquot(struct dquot *dquot);
1500static int ext4_mark_dquot_dirty(struct dquot *dquot);
1501static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1502static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1503 const struct path *path);
617ba13b 1504static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1505 size_t len, loff_t off);
617ba13b 1506static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1507 const char *data, size_t len, loff_t off);
7c319d32
AK		 1508static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1509 unsigned int flags);
ac27a0ec 1510
96c7e0d9
JK		 1511static struct dquot **ext4_get_dquots(struct inode *inode)
1512{
1513 return EXT4_I(inode)->i_dquot;
1514}
1515
61e225dc 1516static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1517 .get_reserved_space = ext4_get_reserved_space,
1518 .write_dquot = ext4_write_dquot,
1519 .acquire_dquot = ext4_acquire_dquot,
1520 .release_dquot = ext4_release_dquot,
1521 .mark_dirty = ext4_mark_dquot_dirty,
1522 .write_info = ext4_write_info,
1523 .alloc_dquot = dquot_alloc,
1524 .destroy_dquot = dquot_destroy,
1525 .get_projid = ext4_get_projid,
1526 .get_inode_usage = ext4_get_inode_usage,
ebc11f7b 1527 .get_next_id = dquot_get_next_id,
ac27a0ec
DK		 1528};
1529
0d54b217 1530static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1531 .quota_on = ext4_quota_on,
ca0e05e4 1532 .quota_off = ext4_quota_off,
287a8095 1533 .quota_sync = dquot_quota_sync,
0a240339 1534 .get_state = dquot_get_state,
287a8095
CH		 1535 .set_info = dquot_set_dqinfo,
1536 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1537 .set_dqblk = dquot_set_dqblk,
1538 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1539};
1540#endif
1541
ee9b6d61 1542static const struct super_operations ext4_sops = {
617ba13b 1543 .alloc_inode = ext4_alloc_inode,
94053139 1544 .free_inode = ext4_free_in_core_inode,
617ba13b 1545 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1546 .write_inode = ext4_write_inode,
1547 .dirty_inode = ext4_dirty_inode,
7ff9c073 1548 .drop_inode = ext4_drop_inode,
0930fcc1 1549 .evict_inode = ext4_evict_inode,
617ba13b 1550 .put_super = ext4_put_super,
617ba13b 1551 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1552 .freeze_fs = ext4_freeze,
1553 .unfreeze_fs = ext4_unfreeze,
617ba13b 1554 .statfs = ext4_statfs,
617ba13b 1555 .show_options = ext4_show_options,
ac27a0ec 1556#ifdef CONFIG_QUOTA
617ba13b
MC		 1557 .quota_read = ext4_quota_read,
1558 .quota_write = ext4_quota_write,
96c7e0d9 1559 .get_dquots = ext4_get_dquots,
ac27a0ec
DK		 1560#endif
1561};
1562
39655164 1563static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1564 .fh_to_dentry = ext4_fh_to_dentry,
1565 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1566 .get_parent = ext4_get_parent,
fde87268 1567 .commit_metadata = ext4_nfs_commit_metadata,
ac27a0ec
DK		 1568};
1569
1570enum {
1571 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
ba2e524d 1572 Opt_resgid, Opt_resuid, Opt_sb,
72578c33 1573 Opt_nouid32, Opt_debug, Opt_removed,
2d544ec9 1574 Opt_user_xattr, Opt_acl,
72578c33 1575 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1576 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1577 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1578 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1579 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
4f74d15f 1580 Opt_inlinecrypt,
ba2e524d 1581 Opt_usrjquota, Opt_grpjquota, Opt_quota,
ee4a3fcd 1582 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1ff20307 1583 Opt_usrquota, Opt_grpquota, Opt_prjquota,
9cb20f94 1584 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
327eaf73 1585 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
4437992b 1586 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1449032b 1587 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1588 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1589 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1590 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1591 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
21175ca4 1592 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
e5a185c2 1593 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
8016e29f 1594#ifdef CONFIG_EXT4_DEBUG
99c880de 1595 Opt_fc_debug_max_replay, Opt_fc_debug_force
8016e29f 1596#endif
ac27a0ec
DK		 1597};
1598
e5a185c2 1599static const struct constant_table ext4_param_errors[] = {
ba2e524d
LC		 1600 {"continue", EXT4_MOUNT_ERRORS_CONT},
1601 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1602 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
e5a185c2
LC		 1603 {}
1604};
1605
1606static const struct constant_table ext4_param_data[] = {
ba2e524d
LC		 1607 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1608 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1609 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
e5a185c2
LC		 1610 {}
1611};
1612
1613static const struct constant_table ext4_param_data_err[] = {
1614 {"abort", Opt_data_err_abort},
1615 {"ignore", Opt_data_err_ignore},
1616 {}
1617};
1618
1619static const struct constant_table ext4_param_jqfmt[] = {
ba2e524d
LC		 1620 {"vfsold", QFMT_VFS_OLD},
1621 {"vfsv0", QFMT_VFS_V0},
1622 {"vfsv1", QFMT_VFS_V1},
e5a185c2
LC		 1623 {}
1624};
1625
1626static const struct constant_table ext4_param_dax[] = {
1627 {"always", Opt_dax_always},
1628 {"inode", Opt_dax_inode},
1629 {"never", Opt_dax_never},
1630 {}
1631};
1632
1633/* String parameter that allows empty argument */
1634#define fsparam_string_empty(NAME, OPT) \
1635 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1636
1637/*
1638 * Mount option specification
1639 * We don't use fsparam_flag_no because of the way we set the
1640 * options and the way we show them in _ext4_show_options(). To
1641 * keep the changes to a minimum, let's keep the negative options
1642 * separate for now.
1643 */
1644static const struct fs_parameter_spec ext4_param_specs[] = {
1645 fsparam_flag ("bsddf", Opt_bsd_df),
1646 fsparam_flag ("minixdf", Opt_minix_df),
1647 fsparam_flag ("grpid", Opt_grpid),
1648 fsparam_flag ("bsdgroups", Opt_grpid),
1649 fsparam_flag ("nogrpid", Opt_nogrpid),
1650 fsparam_flag ("sysvgroups", Opt_nogrpid),
1651 fsparam_u32 ("resgid", Opt_resgid),
1652 fsparam_u32 ("resuid", Opt_resuid),
1653 fsparam_u32 ("sb", Opt_sb),
1654 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1655 fsparam_flag ("nouid32", Opt_nouid32),
1656 fsparam_flag ("debug", Opt_debug),
1657 fsparam_flag ("oldalloc", Opt_removed),
1658 fsparam_flag ("orlov", Opt_removed),
1659 fsparam_flag ("user_xattr", Opt_user_xattr),
e5a185c2 1660 fsparam_flag ("acl", Opt_acl),
e5a185c2
LC		 1661 fsparam_flag ("norecovery", Opt_noload),
1662 fsparam_flag ("noload", Opt_noload),
1663 fsparam_flag ("bh", Opt_removed),
1664 fsparam_flag ("nobh", Opt_removed),
1665 fsparam_u32 ("commit", Opt_commit),
1666 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1667 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1668 fsparam_u32 ("journal_dev", Opt_journal_dev),
1669 fsparam_bdev ("journal_path", Opt_journal_path),
1670 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1671 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1672 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1673 fsparam_flag ("abort", Opt_abort),
1674 fsparam_enum ("data", Opt_data, ext4_param_data),
1675 fsparam_enum ("data_err", Opt_data_err,
1676 ext4_param_data_err),
1677 fsparam_string_empty
1678 ("usrjquota", Opt_usrjquota),
1679 fsparam_string_empty
1680 ("grpjquota", Opt_grpjquota),
1681 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1682 fsparam_flag ("grpquota", Opt_grpquota),
1683 fsparam_flag ("quota", Opt_quota),
1684 fsparam_flag ("noquota", Opt_noquota),
1685 fsparam_flag ("usrquota", Opt_usrquota),
1686 fsparam_flag ("prjquota", Opt_prjquota),
1687 fsparam_flag ("barrier", Opt_barrier),
1688 fsparam_u32 ("barrier", Opt_barrier),
1689 fsparam_flag ("nobarrier", Opt_nobarrier),
1ff20307 1690 fsparam_flag ("i_version", Opt_removed),
e5a185c2
LC		 1691 fsparam_flag ("dax", Opt_dax),
1692 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1693 fsparam_u32 ("stripe", Opt_stripe),
1694 fsparam_flag ("delalloc", Opt_delalloc),
1695 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1696 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1697 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
e5a185c2
LC		 1698 fsparam_u32 ("debug_want_extra_isize",
1699 Opt_debug_want_extra_isize),
1700 fsparam_flag ("mblk_io_submit", Opt_removed),
1701 fsparam_flag ("nomblk_io_submit", Opt_removed),
1702 fsparam_flag ("block_validity", Opt_block_validity),
1703 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1704 fsparam_u32 ("inode_readahead_blks",
1705 Opt_inode_readahead_blks),
1706 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1707 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1708 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1709 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1710 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1711 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1712 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1713 fsparam_flag ("discard", Opt_discard),
1714 fsparam_flag ("nodiscard", Opt_nodiscard),
1715 fsparam_u32 ("init_itable", Opt_init_itable),
1716 fsparam_flag ("init_itable", Opt_init_itable),
1717 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1718#ifdef CONFIG_EXT4_DEBUG
1719 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1720 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1721#endif
1722 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1723 fsparam_flag ("test_dummy_encryption",
1724 Opt_test_dummy_encryption),
1725 fsparam_string ("test_dummy_encryption",
1726 Opt_test_dummy_encryption),
1727 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1728 fsparam_flag ("nombcache", Opt_nombcache),
1729 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1730 fsparam_flag ("prefetch_block_bitmaps",
1731 Opt_removed),
1732 fsparam_flag ("no_prefetch_block_bitmaps",
1733 Opt_no_prefetch_block_bitmaps),
1734 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1735 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1736 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1737 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1738 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1739 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1740 {}
1741};
1742
b3881f74 1743#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
196e402a 1744
26092bf5
TT		 1745#define MOPT_SET 0x0001
1746#define MOPT_CLEAR 0x0002
1747#define MOPT_NOSUPPORT 0x0004
1748#define MOPT_EXPLICIT 0x0008
ac27a0ec 1749#ifdef CONFIG_QUOTA
26092bf5 1750#define MOPT_Q 0
ba2e524d 1751#define MOPT_QFMT 0x0010
26092bf5
TT		 1752#else
1753#define MOPT_Q MOPT_NOSUPPORT
1754#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1755#endif
ba2e524d
LC		 1756#define MOPT_NO_EXT2 0x0020
1757#define MOPT_NO_EXT3 0x0040
8dc0aa8c 1758#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ba2e524d
LC		 1759#define MOPT_SKIP 0x0080
1760#define MOPT_2 0x0100
26092bf5
TT		 1761
1762static const struct mount_opts {
1763 int token;
1764 int mount_opt;
1765 int flags;
1766} ext4_mount_opts[] = {
1767 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1768 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1769 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1770 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1771 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1772 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1773 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1774 MOPT_EXT4_ONLY | MOPT_SET},
1775 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1776 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1777 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1778 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1779 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1780 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1781 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1782 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1783 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1784 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
cb8435dc 1785 {Opt_commit, 0, MOPT_NO_EXT2},
c6d3d56d
DW		 1786 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1787 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1788 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1789 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1790 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1791 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1792 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1793 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
ba2e524d 1794 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
26092bf5
TT		 1795 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1796 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1797 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1798 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1799 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
ba2e524d
LC		 1800 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1801 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1802 {Opt_journal_path, 0, MOPT_NO_EXT2},
1803 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1804 {Opt_data, 0, MOPT_NO_EXT2},
26092bf5 1805 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
03010a33 1806#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5 1807 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
ac27a0ec 1808#else
26092bf5 1809 {Opt_acl, 0, MOPT_NOSUPPORT},
ac27a0ec 1810#endif
26092bf5
TT		 1811 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1812 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1813 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1814 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1815 MOPT_SET | MOPT_Q},
1816 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1817 MOPT_SET | MOPT_Q},
49da9392
JK		 1818 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1819 MOPT_SET | MOPT_Q},
26092bf5 1820 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1821 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1822 MOPT_CLEAR | MOPT_Q},
ba2e524d
LC		 1823 {Opt_usrjquota, 0, MOPT_Q},
1824 {Opt_grpjquota, 0, MOPT_Q},
1825 {Opt_jqfmt, 0, MOPT_QFMT},
cdb7ee4c 1826 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
21175ca4 1827 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
3d392b26 1828 MOPT_SET},
99c880de 1829#ifdef CONFIG_EXT4_DEBUG
0f0672ff
HS		 1830 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1831 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
8016e29f 1832#endif
26092bf5
TT		 1833 {Opt_err, 0, 0}
1834};
1835
5298d4bf 1836#if IS_ENABLED(CONFIG_UNICODE)
c83ad55e
GKB		 1837static const struct ext4_sb_encodings {
1838 __u16 magic;
1839 char *name;
49bd03cc 1840 unsigned int version;
c83ad55e 1841} ext4_sb_encoding_map[] = {
49bd03cc 1842 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
c83ad55e
GKB		 1843};
1844
aa8bf298
CH		 1845static const struct ext4_sb_encodings *
1846ext4_sb_read_encoding(const struct ext4_super_block *es)
c83ad55e
GKB		 1847{
1848 __u16 magic = le16_to_cpu(es->s_encoding);
1849 int i;
1850
1851 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1852 if (magic == ext4_sb_encoding_map[i].magic)
aa8bf298 1853 return &ext4_sb_encoding_map[i];
c83ad55e 1854
aa8bf298 1855 return NULL;
c83ad55e
GKB		 1856}
1857#endif
1858
6e47a3cc
LC		 1859#define EXT4_SPEC_JQUOTA (1 << 0)
1860#define EXT4_SPEC_JQFMT (1 << 1)
1861#define EXT4_SPEC_DATAJ (1 << 2)
1862#define EXT4_SPEC_SB_BLOCK (1 << 3)
1863#define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1864#define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
6e47a3cc
LC		 1865#define EXT4_SPEC_s_want_extra_isize (1 << 7)
1866#define EXT4_SPEC_s_max_batch_time (1 << 8)
1867#define EXT4_SPEC_s_min_batch_time (1 << 9)
1868#define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1869#define EXT4_SPEC_s_li_wait_mult (1 << 11)
1870#define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1871#define EXT4_SPEC_s_stripe (1 << 13)
1872#define EXT4_SPEC_s_resuid (1 << 14)
1873#define EXT4_SPEC_s_resgid (1 << 15)
1874#define EXT4_SPEC_s_commit_interval (1 << 16)
1875#define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
7edfd85b 1876#define EXT4_SPEC_s_sb_block (1 << 18)
27b38686 1877#define EXT4_SPEC_mb_optimize_scan (1 << 19)
6e47a3cc 1878
461c3af0 1879struct ext4_fs_context {
e6e268cb 1880 char *s_qf_names[EXT4_MAXQUOTAS];
85456054 1881 struct fscrypt_dummy_policy dummy_enc_policy;
e6e268cb 1882 int s_jquota_fmt; /* Format of quota to use */
6e47a3cc
LC		 1883#ifdef CONFIG_EXT4_DEBUG
1884 int s_fc_debug_max_replay;
1885#endif
e6e268cb 1886 unsigned short qname_spec;
6e47a3cc
LC		 1887 unsigned long vals_s_flags; /* Bits to set in s_flags */
1888 unsigned long mask_s_flags; /* Bits changed in s_flags */
e6e268cb 1889 unsigned long journal_devnum;
6e47a3cc
LC		 1890 unsigned long s_commit_interval;
1891 unsigned long s_stripe;
1892 unsigned int s_inode_readahead_blks;
1893 unsigned int s_want_extra_isize;
1894 unsigned int s_li_wait_mult;
1895 unsigned int s_max_dir_size_kb;
e6e268cb 1896 unsigned int journal_ioprio;
6e47a3cc
LC		 1897 unsigned int vals_s_mount_opt;
1898 unsigned int mask_s_mount_opt;
1899 unsigned int vals_s_mount_opt2;
1900 unsigned int mask_s_mount_opt2;
e3952fcc
LC		 1901 unsigned long vals_s_mount_flags;
1902 unsigned long mask_s_mount_flags;
b6bd2435 1903 unsigned int opt_flags; /* MOPT flags */
6e47a3cc
LC		 1904 unsigned int spec;
1905 u32 s_max_batch_time;
1906 u32 s_min_batch_time;
1907 kuid_t s_resuid;
1908 kgid_t s_resgid;
7edfd85b 1909 ext4_fsblk_t s_sb_block;
b237e304
HS		 1910};
1911
cebe85d5
LC		 1912static void ext4_fc_free(struct fs_context *fc)
1913{
1914 struct ext4_fs_context *ctx = fc->fs_private;
1915 int i;
1916
1917 if (!ctx)
1918 return;
1919
1920 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1921 kfree(ctx->s_qf_names[i]);
1922
85456054 1923 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
cebe85d5
LC		 1924 kfree(ctx);
1925}
1926
1927int ext4_init_fs_context(struct fs_context *fc)
1928{
da9e4802 1929 struct ext4_fs_context *ctx;
cebe85d5
LC		 1930
1931 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
1932 if (!ctx)
1933 return -ENOMEM;
1934
1935 fc->fs_private = ctx;
1936 fc->ops = &ext4_context_ops;
1937
1938 return 0;
1939}
1940
e6e268cb
LC		 1941#ifdef CONFIG_QUOTA
1942/*
1943 * Note the name of the specified quota file.
1944 */
1945static int note_qf_name(struct fs_context *fc, int qtype,
1946 struct fs_parameter *param)
1947{
1948 struct ext4_fs_context *ctx = fc->fs_private;
1949 char *qname;
1950
1951 if (param->size < 1) {
1952 ext4_msg(NULL, KERN_ERR, "Missing quota name");
1953 return -EINVAL;
1954 }
1955 if (strchr(param->string, '/')) {
1956 ext4_msg(NULL, KERN_ERR,
1957 "quotafile must be on filesystem root");
1958 return -EINVAL;
1959 }
1960 if (ctx->s_qf_names[qtype]) {
1961 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
1962 ext4_msg(NULL, KERN_ERR,
1963 "%s quota file already specified",
1964 QTYPE2NAME(qtype));
1965 return -EINVAL;
1966 }
1967 return 0;
1968 }
1969
1970 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
1971 if (!qname) {
1972 ext4_msg(NULL, KERN_ERR,
1973 "Not enough memory for storing quotafile name");
1974 return -ENOMEM;
1975 }
1976 ctx->s_qf_names[qtype] = qname;
1977 ctx->qname_spec |= 1 << qtype;
6e47a3cc 1978 ctx->spec |= EXT4_SPEC_JQUOTA;
e6e268cb
LC		 1979 return 0;
1980}
1981
1982/*
1983 * Clear the name of the specified quota file.
1984 */
1985static int unnote_qf_name(struct fs_context *fc, int qtype)
1986{
1987 struct ext4_fs_context *ctx = fc->fs_private;
1988
1989 if (ctx->s_qf_names[qtype])
1990 kfree(ctx->s_qf_names[qtype]);
1991
1992 ctx->s_qf_names[qtype] = NULL;
1993 ctx->qname_spec |= 1 << qtype;
6e47a3cc 1994 ctx->spec |= EXT4_SPEC_JQUOTA;
e6e268cb
LC		 1995 return 0;
1996}
1997#endif
1998
85456054
EB		 1999static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2000 struct ext4_fs_context *ctx)
2001{
2002 int err;
2003
2004 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2005 ext4_msg(NULL, KERN_WARNING,
2006 "test_dummy_encryption option not supported");
2007 return -EINVAL;
2008 }
2009 err = fscrypt_parse_test_dummy_encryption(param,
2010 &ctx->dummy_enc_policy);
2011 if (err == -EINVAL) {
2012 ext4_msg(NULL, KERN_WARNING,
2013 "Value of option \"%s\" is unrecognized", param->key);
2014 } else if (err == -EEXIST) {
2015 ext4_msg(NULL, KERN_WARNING,
2016 "Conflicting test_dummy_encryption options");
2017 return -EINVAL;
2018 }
2019 return err;
2020}
2021
6e47a3cc 2022#define EXT4_SET_CTX(name) \
4c246728
LC		 2023static inline void ctx_set_##name(struct ext4_fs_context *ctx, \
2024 unsigned long flag) \
6e47a3cc
LC		 2025{ \
2026 ctx->mask_s_##name |= flag; \
2027 ctx->vals_s_##name |= flag; \
e3952fcc
LC		 2028}
2029
2030#define EXT4_CLEAR_CTX(name) \
4c246728
LC		 2031static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \
2032 unsigned long flag) \
6e47a3cc
LC		 2033{ \
2034 ctx->mask_s_##name |= flag; \
2035 ctx->vals_s_##name &= ~flag; \
e3952fcc
LC		 2036}
2037
2038#define EXT4_TEST_CTX(name) \
4c246728
LC		 2039static inline unsigned long \
2040ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
6e47a3cc 2041{ \
4c246728 2042 return (ctx->vals_s_##name & flag); \
e3952fcc 2043}
6e47a3cc 2044
e3952fcc 2045EXT4_SET_CTX(flags); /* set only */
6e47a3cc 2046EXT4_SET_CTX(mount_opt);
e3952fcc
LC		 2047EXT4_CLEAR_CTX(mount_opt);
2048EXT4_TEST_CTX(mount_opt);
6e47a3cc 2049EXT4_SET_CTX(mount_opt2);
e3952fcc
LC		 2050EXT4_CLEAR_CTX(mount_opt2);
2051EXT4_TEST_CTX(mount_opt2);
2052
2053static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
2054{
2055 set_bit(bit, &ctx->mask_s_mount_flags);
2056 set_bit(bit, &ctx->vals_s_mount_flags);
2057}
6e47a3cc 2058
02f960f8 2059static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
26092bf5 2060{
461c3af0 2061 struct ext4_fs_context *ctx = fc->fs_private;
461c3af0 2062 struct fs_parse_result result;
26092bf5 2063 const struct mount_opts *m;
461c3af0 2064 int is_remount;
08cefc7a
EB		 2065 kuid_t uid;
2066 kgid_t gid;
461c3af0
LC		 2067 int token;
2068
2069 token = fs_parse(fc, ext4_param_specs, param, &result);
2070 if (token < 0)
2071 return token;
2072 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
26092bf5 2073
ba2e524d
LC		 2074 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2075 if (token == m->token)
2076 break;
2077
2078 ctx->opt_flags |= m->flags;
2079
2080 if (m->flags & MOPT_EXPLICIT) {
2081 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2082 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2083 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2084 ctx_set_mount_opt2(ctx,
2085 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2086 } else
2087 return -EINVAL;
2088 }
2089
2090 if (m->flags & MOPT_NOSUPPORT) {
2091 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2092 param->key);
2093 return 0;
2094 }
2095
2096 switch (token) {
57f73c2c 2097#ifdef CONFIG_QUOTA
ba2e524d 2098 case Opt_usrjquota:
461c3af0 2099 if (!*param->string)
e6e268cb 2100 return unnote_qf_name(fc, USRQUOTA);
461c3af0 2101 else
e6e268cb 2102 return note_qf_name(fc, USRQUOTA, param);
ba2e524d 2103 case Opt_grpjquota:
461c3af0 2104 if (!*param->string)
e6e268cb 2105 return unnote_qf_name(fc, GRPQUOTA);
461c3af0 2106 else
e6e268cb 2107 return note_qf_name(fc, GRPQUOTA, param);
57f73c2c 2108#endif
26092bf5 2109 case Opt_sb:
7edfd85b
LC		 2110 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2111 ext4_msg(NULL, KERN_WARNING,
2112 "Ignoring %s option on remount", param->key);
2113 } else {
2114 ctx->s_sb_block = result.uint_32;
2115 ctx->spec |= EXT4_SPEC_s_sb_block;
2116 }
02f960f8 2117 return 0;
26092bf5 2118 case Opt_removed:
da812f61 2119 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
461c3af0 2120 param->key);
02f960f8 2121 return 0;
26092bf5 2122 case Opt_abort:
e3952fcc 2123 ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
02f960f8 2124 return 0;
4f74d15f
EB		 2125 case Opt_inlinecrypt:
2126#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
6e47a3cc 2127 ctx_set_flags(ctx, SB_INLINECRYPT);
4f74d15f 2128#else
da812f61 2129 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
4f74d15f 2130#endif
02f960f8 2131 return 0;
461c3af0 2132 case Opt_errors:
6e47a3cc 2133 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
ba2e524d
LC		 2134 ctx_set_mount_opt(ctx, result.uint_32);
2135 return 0;
2136#ifdef CONFIG_QUOTA
2137 case Opt_jqfmt:
2138 ctx->s_jquota_fmt = result.uint_32;
2139 ctx->spec |= EXT4_SPEC_JQFMT;
2140 return 0;
2141#endif
2142 case Opt_data:
2143 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2144 ctx_set_mount_opt(ctx, result.uint_32);
2145 ctx->spec |= EXT4_SPEC_DATAJ;
2146 return 0;
2147 case Opt_commit:
461c3af0 2148 if (result.uint_32 == 0)
934b0de1 2149 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE;
461c3af0 2150 else if (result.uint_32 > INT_MAX / HZ) {
da812f61 2151 ext4_msg(NULL, KERN_ERR,
9ba55543 2152 "Invalid commit interval %d, "
2153 "must be smaller than %d",
461c3af0 2154 result.uint_32, INT_MAX / HZ);
da812f61 2155 return -EINVAL;
9ba55543 2156 }
6e47a3cc
LC		 2157 ctx->s_commit_interval = HZ * result.uint_32;
2158 ctx->spec |= EXT4_SPEC_s_commit_interval;
ba2e524d
LC		 2159 return 0;
2160 case Opt_debug_want_extra_isize:
6e47a3cc 2161 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
da812f61 2162 ext4_msg(NULL, KERN_ERR,
461c3af0 2163 "Invalid want_extra_isize %d", result.uint_32);
da812f61 2164 return -EINVAL;
9803387c 2165 }
6e47a3cc
LC		 2166 ctx->s_want_extra_isize = result.uint_32;
2167 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
ba2e524d
LC		 2168 return 0;
2169 case Opt_max_batch_time:
6e47a3cc
LC		 2170 ctx->s_max_batch_time = result.uint_32;
2171 ctx->spec |= EXT4_SPEC_s_max_batch_time;
ba2e524d
LC		 2172 return 0;
2173 case Opt_min_batch_time:
6e47a3cc
LC		 2174 ctx->s_min_batch_time = result.uint_32;
2175 ctx->spec |= EXT4_SPEC_s_min_batch_time;
ba2e524d
LC		 2176 return 0;
2177 case Opt_inode_readahead_blks:
461c3af0
LC		 2178 if (result.uint_32 &&
2179 (result.uint_32 > (1 << 30) ||
2180 !is_power_of_2(result.uint_32))) {
da812f61 2181 ext4_msg(NULL, KERN_ERR,
e33e60ea
JK		 2182 "EXT4-fs: inode_readahead_blks must be "
2183 "0 or a power of 2 smaller than 2^31");
da812f61 2184 return -EINVAL;
5f3633e3 2185 }
6e47a3cc
LC		 2186 ctx->s_inode_readahead_blks = result.uint_32;
2187 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
ba2e524d
LC		 2188 return 0;
2189 case Opt_init_itable:
6e47a3cc
LC		 2190 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2191 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
461c3af0 2192 if (param->type == fs_value_is_string)
6e47a3cc
LC		 2193 ctx->s_li_wait_mult = result.uint_32;
2194 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
ba2e524d
LC		 2195 return 0;
2196 case Opt_max_dir_size_kb:
6e47a3cc
LC		 2197 ctx->s_max_dir_size_kb = result.uint_32;
2198 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
ba2e524d 2199 return 0;
8016e29f 2200#ifdef CONFIG_EXT4_DEBUG
ba2e524d 2201 case Opt_fc_debug_max_replay:
6e47a3cc
LC		 2202 ctx->s_fc_debug_max_replay = result.uint_32;
2203 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
ba2e524d 2204 return 0;
8016e29f 2205#endif
ba2e524d 2206 case Opt_stripe:
6e47a3cc
LC		 2207 ctx->s_stripe = result.uint_32;
2208 ctx->spec |= EXT4_SPEC_s_stripe;
ba2e524d
LC		 2209 return 0;
2210 case Opt_resuid:
461c3af0 2211 uid = make_kuid(current_user_ns(), result.uint_32);
5f3633e3 2212 if (!uid_valid(uid)) {
da812f61 2213 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
461c3af0 2214 result.uint_32);
da812f61 2215 return -EINVAL;
26092bf5 2216 }
6e47a3cc
LC		 2217 ctx->s_resuid = uid;
2218 ctx->spec |= EXT4_SPEC_s_resuid;
ba2e524d
LC		 2219 return 0;
2220 case Opt_resgid:
461c3af0 2221 gid = make_kgid(current_user_ns(), result.uint_32);
5f3633e3 2222 if (!gid_valid(gid)) {
da812f61 2223 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
461c3af0 2224 result.uint_32);
da812f61 2225 return -EINVAL;
5f3633e3 2226 }
6e47a3cc
LC		 2227 ctx->s_resgid = gid;
2228 ctx->spec |= EXT4_SPEC_s_resgid;
ba2e524d
LC		 2229 return 0;
2230 case Opt_journal_dev:
5f3633e3 2231 if (is_remount) {
da812f61 2232 ext4_msg(NULL, KERN_ERR,
5f3633e3 2233 "Cannot specify journal on remount");
da812f61 2234 return -EINVAL;
5f3633e3 2235 }
461c3af0 2236 ctx->journal_devnum = result.uint_32;
6e47a3cc 2237 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
ba2e524d
LC		 2238 return 0;
2239 case Opt_journal_path:
2240 {
ad4eec61
ES		 2241 struct inode *journal_inode;
2242 struct path path;
2243 int error;
2244
2245 if (is_remount) {
da812f61 2246 ext4_msg(NULL, KERN_ERR,
ad4eec61 2247 "Cannot specify journal on remount");
da812f61 2248 return -EINVAL;
ad4eec61 2249 }
ad4eec61 2250
e3ea75ee 2251 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path);
ad4eec61 2252 if (error) {
da812f61 2253 ext4_msg(NULL, KERN_ERR, "error: could not find "
461c3af0 2254 "journal device path");
da812f61 2255 return -EINVAL;
ad4eec61
ES		 2256 }
2257
2b0143b5 2258 journal_inode = d_inode(path.dentry);
461c3af0 2259 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
6e47a3cc 2260 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
ad4eec61 2261 path_put(&path);
ba2e524d
LC		 2262 return 0;
2263 }
2264 case Opt_journal_ioprio:
461c3af0 2265 if (result.uint_32 > 7) {
da812f61 2266 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
5f3633e3 2267 " (must be 0-7)");
da812f61 2268 return -EINVAL;
5f3633e3 2269 }
461c3af0
LC		 2270 ctx->journal_ioprio =
2271 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
6e47a3cc 2272 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
ba2e524d
LC		 2273 return 0;
2274 case Opt_test_dummy_encryption:
85456054 2275 return ext4_parse_test_dummy_encryption(param, ctx);
ba2e524d
LC		 2276 case Opt_dax:
2277 case Opt_dax_type:
ef83b6e8 2278#ifdef CONFIG_FS_DAX
ba2e524d
LC		 2279 {
2280 int type = (token == Opt_dax) ?
2281 Opt_dax : result.uint_32;
2282
2283 switch (type) {
9cb20f94
IW		 2284 case Opt_dax:
2285 case Opt_dax_always:
ba2e524d 2286 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
6e47a3cc 2287 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
9cb20f94
IW		 2288 break;
2289 case Opt_dax_never:
ba2e524d 2290 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
6e47a3cc 2291 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
9cb20f94
IW		 2292 break;
2293 case Opt_dax_inode:
6e47a3cc
LC		 2294 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2295 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
9cb20f94 2296 /* Strictly for printing options */
ba2e524d 2297 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
9cb20f94
IW		 2298 break;
2299 }
ba2e524d
LC		 2300 return 0;
2301 }
ef83b6e8 2302#else
da812f61 2303 ext4_msg(NULL, KERN_INFO, "dax option not supported");
da812f61 2304 return -EINVAL;
5f3633e3 2305#endif
ba2e524d
LC		 2306 case Opt_data_err:
2307 if (result.uint_32 == Opt_data_err_abort)
2308 ctx_set_mount_opt(ctx, m->mount_opt);
2309 else if (result.uint_32 == Opt_data_err_ignore)
2310 ctx_clear_mount_opt(ctx, m->mount_opt);
2311 return 0;
2312 case Opt_mb_optimize_scan:
27b38686
OM		 2313 if (result.int_32 == 1) {
2314 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2315 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2316 } else if (result.int_32 == 0) {
2317 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2318 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2319 } else {
da812f61 2320 ext4_msg(NULL, KERN_WARNING,
196e402a 2321 "mb_optimize_scan should be set to 0 or 1.");
da812f61 2322 return -EINVAL;
196e402a 2323 }
ba2e524d
LC		 2324 return 0;
2325 }
2326
2327 /*
2328 * At this point we should only be getting options requiring MOPT_SET,
2329 * or MOPT_CLEAR. Anything else is a bug
2330 */
2331 if (m->token == Opt_err) {
2332 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2333 param->key);
2334 WARN_ON(1);
2335 return -EINVAL;
2336 }
2337
2338 else {
461c3af0
LC		 2339 unsigned int set = 0;
2340
2341 if ((param->type == fs_value_is_flag) ||
2342 result.uint_32 > 0)
2343 set = 1;
2344
5f3633e3 2345 if (m->flags & MOPT_CLEAR)
461c3af0 2346 set = !set;
5f3633e3 2347 else if (unlikely(!(m->flags & MOPT_SET))) {
da812f61 2348 ext4_msg(NULL, KERN_WARNING,
461c3af0
LC		 2349 "buggy handling of option %s",
2350 param->key);
5f3633e3 2351 WARN_ON(1);
da812f61 2352 return -EINVAL;
5f3633e3 2353 }
995a3ed6 2354 if (m->flags & MOPT_2) {
461c3af0 2355 if (set != 0)
6e47a3cc 2356 ctx_set_mount_opt2(ctx, m->mount_opt);
995a3ed6 2357 else
6e47a3cc 2358 ctx_clear_mount_opt2(ctx, m->mount_opt);
995a3ed6 2359 } else {
461c3af0 2360 if (set != 0)
6e47a3cc 2361 ctx_set_mount_opt(ctx, m->mount_opt);
995a3ed6 2362 else
6e47a3cc 2363 ctx_clear_mount_opt(ctx, m->mount_opt);
995a3ed6 2364 }
26092bf5 2365 }
ba2e524d 2366
02f960f8 2367 return 0;
26092bf5
TT		 2368}
2369
7edfd85b 2370static int parse_options(struct fs_context *fc, char *options)
26092bf5 2371{
461c3af0 2372 struct fs_parameter param;
461c3af0
LC		 2373 int ret;
2374 char *key;
26092bf5
TT		 2375
2376 if (!options)
7edfd85b 2377 return 0;
461c3af0
LC		 2378
2379 while ((key = strsep(&options, ",")) != NULL) {
2380 if (*key) {
2381 size_t v_len = 0;
2382 char *value = strchr(key, '=');
2383
2384 param.type = fs_value_is_flag;
2385 param.string = NULL;
2386
2387 if (value) {
2388 if (value == key)
2389 continue;
2390
2391 *value++ = 0;
2392 v_len = strlen(value);
2393 param.string = kmemdup_nul(value, v_len,
2394 GFP_KERNEL);
2395 if (!param.string)
7edfd85b 2396 return -ENOMEM;
461c3af0
LC		 2397 param.type = fs_value_is_string;
2398 }
2399
2400 param.key = key;
2401 param.size = v_len;
2402
02f960f8 2403 ret = ext4_parse_param(fc, &param);
461c3af0
LC		 2404 if (param.string)
2405 kfree(param.string);
2406 if (ret < 0)
7edfd85b 2407 return ret;
461c3af0 2408 }
ac27a0ec 2409 }
461c3af0 2410
7edfd85b 2411 ret = ext4_validate_options(fc);
da812f61 2412 if (ret < 0)
7edfd85b 2413 return ret;
da812f61 2414
7edfd85b
LC		 2415 return 0;
2416}
2417
2418static int parse_apply_sb_mount_options(struct super_block *sb,
2419 struct ext4_fs_context *m_ctx)
2420{
2421 struct ext4_sb_info *sbi = EXT4_SB(sb);
2422 char *s_mount_opts = NULL;
2423 struct ext4_fs_context *s_ctx = NULL;
2424 struct fs_context *fc = NULL;
2425 int ret = -ENOMEM;
2426
2427 if (!sbi->s_es->s_mount_opts[0])
e6e268cb
LC		 2428 return 0;
2429
7edfd85b
LC		 2430 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2431 sizeof(sbi->s_es->s_mount_opts),
2432 GFP_KERNEL);
2433 if (!s_mount_opts)
2434 return ret;
2435
2436 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2437 if (!fc)
2438 goto out_free;
2439
2440 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2441 if (!s_ctx)
2442 goto out_free;
2443
2444 fc->fs_private = s_ctx;
2445 fc->s_fs_info = sbi;
2446
2447 ret = parse_options(fc, s_mount_opts);
6e47a3cc 2448 if (ret < 0)
7edfd85b 2449 goto parse_failed;
e6e268cb 2450
7edfd85b
LC		 2451 ret = ext4_check_opt_consistency(fc, sb);
2452 if (ret < 0) {
2453parse_failed:
2454 ext4_msg(sb, KERN_WARNING,
2455 "failed to parse options in superblock: %s",
2456 s_mount_opts);
2457 ret = 0;
2458 goto out_free;
2459 }
2460
2461 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2462 m_ctx->journal_devnum = s_ctx->journal_devnum;
2463 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2464 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2465
85456054
EB		 2466 ext4_apply_options(fc, sb);
2467 ret = 0;
7edfd85b
LC		 2468
2469out_free:
c069db76
EB		 2470 if (fc) {
2471 ext4_fc_free(fc);
2472 kfree(fc);
2473 }
7edfd85b
LC		 2474 kfree(s_mount_opts);
2475 return ret;
4c94bff9
LC		 2476}
2477
e6e268cb
LC		 2478static void ext4_apply_quota_options(struct fs_context *fc,
2479 struct super_block *sb)
2480{
2481#ifdef CONFIG_QUOTA
6e47a3cc 2482 bool quota_feature = ext4_has_feature_quota(sb);
e6e268cb
LC		 2483 struct ext4_fs_context *ctx = fc->fs_private;
2484 struct ext4_sb_info *sbi = EXT4_SB(sb);
2485 char *qname;
2486 int i;
2487
6e47a3cc
LC		 2488 if (quota_feature)
2489 return;
2490
2491 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2492 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2493 if (!(ctx->qname_spec & (1 << i)))
2494 continue;
2495
2496 qname = ctx->s_qf_names[i]; /* May be NULL */
4c1bd5a9
LC		 2497 if (qname)
2498 set_opt(sb, QUOTA);
6e47a3cc 2499 ctx->s_qf_names[i] = NULL;
13b215a9
LC		 2500 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2501 lockdep_is_held(&sb->s_umount));
2502 if (qname)
2503 kfree_rcu(qname);
6e47a3cc 2504 }
e6e268cb 2505 }
6e47a3cc
LC		 2506
2507 if (ctx->spec & EXT4_SPEC_JQFMT)
2508 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
e6e268cb
LC		 2509#endif
2510}
2511
2512/*
2513 * Check quota settings consistency.
2514 */
2515static int ext4_check_quota_consistency(struct fs_context *fc,
2516 struct super_block *sb)
2517{
2518#ifdef CONFIG_QUOTA
2519 struct ext4_fs_context *ctx = fc->fs_private;
2520 struct ext4_sb_info *sbi = EXT4_SB(sb);
2521 bool quota_feature = ext4_has_feature_quota(sb);
2522 bool quota_loaded = sb_any_quota_loaded(sb);
6e47a3cc
LC		 2523 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2524 int quota_flags, i;
2525
2526 /*
2527 * We do the test below only for project quotas. 'usrquota' and
2528 * 'grpquota' mount options are allowed even without quota feature
2529 * to support legacy quotas in quota files.
2530 */
2531 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2532 !ext4_has_feature_project(sb)) {
2533 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2534 "Cannot enable project quota enforcement.");
2535 return -EINVAL;
2536 }
e6e268cb 2537
6e47a3cc
LC		 2538 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2539 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2540 if (quota_loaded &&
2541 ctx->mask_s_mount_opt & quota_flags &&
2542 !ctx_test_mount_opt(ctx, quota_flags))
2543 goto err_quota_change;
2544
2545 if (ctx->spec & EXT4_SPEC_JQUOTA) {
e6e268cb
LC		 2546
2547 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2548 if (!(ctx->qname_spec & (1 << i)))
2549 continue;
2550
6e47a3cc
LC		 2551 if (quota_loaded &&
2552 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
e6e268cb
LC		 2553 goto err_jquota_change;
2554
2555 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
13b215a9 2556 strcmp(get_qf_name(sb, sbi, i),
e6e268cb
LC		 2557 ctx->s_qf_names[i]) != 0)
2558 goto err_jquota_specified;
2559 }
6e47a3cc
LC		 2560
2561 if (quota_feature) {
2562 ext4_msg(NULL, KERN_INFO,
2563 "Journaled quota options ignored when "
2564 "QUOTA feature is enabled");
2565 return 0;
2566 }
e6e268cb
LC		 2567 }
2568
6e47a3cc 2569 if (ctx->spec & EXT4_SPEC_JQFMT) {
e6e268cb 2570 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
6e47a3cc 2571 goto err_jquota_change;
e6e268cb
LC		 2572 if (quota_feature) {
2573 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2574 "ignored when QUOTA feature is enabled");
2575 return 0;
2576 }
2577 }
6e47a3cc
LC		 2578
2579 /* Make sure we don't mix old and new quota format */
2580 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2581 ctx->s_qf_names[USRQUOTA]);
2582 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2583 ctx->s_qf_names[GRPQUOTA]);
2584
2585 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2586 test_opt(sb, USRQUOTA));
2587
2588 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2589 test_opt(sb, GRPQUOTA));
2590
2591 if (usr_qf_name) {
2592 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2593 usrquota = false;
2594 }
2595 if (grp_qf_name) {
2596 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2597 grpquota = false;
2598 }
2599
2600 if (usr_qf_name || grp_qf_name) {
2601 if (usrquota || grpquota) {
2602 ext4_msg(NULL, KERN_ERR, "old and new quota "
2603 "format mixing");
2604 return -EINVAL;
2605 }
2606
2607 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2608 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2609 "not specified");
2610 return -EINVAL;
2611 }
2612 }
2613
e6e268cb
LC		 2614 return 0;
2615
2616err_quota_change:
2617 ext4_msg(NULL, KERN_ERR,
2618 "Cannot change quota options when quota turned on");
2619 return -EINVAL;
2620err_jquota_change:
2621 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2622 "options when quota turned on");
2623 return -EINVAL;
2624err_jquota_specified:
2625 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2626 QTYPE2NAME(i));
2627 return -EINVAL;
e6e268cb
LC		 2628#else
2629 return 0;
2630#endif
2631}
2632
5f41fdae
EB		 2633static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2634 struct super_block *sb)
2635{
5f41fdae
EB		 2636 const struct ext4_fs_context *ctx = fc->fs_private;
2637 const struct ext4_sb_info *sbi = EXT4_SB(sb);
2638
85456054 2639 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
5f41fdae
EB		 2640 return 0;
2641
2642 if (!ext4_has_feature_encrypt(sb)) {
2643 ext4_msg(NULL, KERN_WARNING,
2644 "test_dummy_encryption requires encrypt feature");
2645 return -EINVAL;
2646 }
2647 /*
2648 * This mount option is just for testing, and it's not worthwhile to
2649 * implement the extra complexity (e.g. RCU protection) that would be
2650 * needed to allow it to be set or changed during remount. We do allow
2651 * it to be specified during remount, but only if there is no change.
2652 */
85456054
EB		 2653 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2654 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2655 &ctx->dummy_enc_policy))
2656 return 0;
5f41fdae 2657 ext4_msg(NULL, KERN_WARNING,
85456054 2658 "Can't set or change test_dummy_encryption on remount");
5f41fdae
EB		 2659 return -EINVAL;
2660 }
85456054
EB		 2661 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2662 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2663 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2664 &ctx->dummy_enc_policy))
2665 return 0;
2666 ext4_msg(NULL, KERN_WARNING,
2667 "Conflicting test_dummy_encryption options");
2668 return -EINVAL;
2669 }
7959eb19 2670 return 0;
85456054
EB		 2671}
2672
2673static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2674 struct super_block *sb)
2675{
2676 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2677 /* if already set, it was already verified to be the same */
2678 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2679 return;
2680 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2681 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2682 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
5f41fdae
EB		 2683}
2684
b6bd2435
LC		 2685static int ext4_check_opt_consistency(struct fs_context *fc,
2686 struct super_block *sb)
2687{
2688 struct ext4_fs_context *ctx = fc->fs_private;
6e47a3cc
LC		 2689 struct ext4_sb_info *sbi = fc->s_fs_info;
2690 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
5f41fdae 2691 int err;
b6bd2435
LC		 2692
2693 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2694 ext4_msg(NULL, KERN_ERR,
2695 "Mount option(s) incompatible with ext2");
2696 return -EINVAL;
2697 }
2698 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2699 ext4_msg(NULL, KERN_ERR,
2700 "Mount option(s) incompatible with ext3");
2701 return -EINVAL;
2702 }
2703
6e47a3cc
LC		 2704 if (ctx->s_want_extra_isize >
2705 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2706 ext4_msg(NULL, KERN_ERR,
2707 "Invalid want_extra_isize %d",
2708 ctx->s_want_extra_isize);
2709 return -EINVAL;
2710 }
2711
2712 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) {
2713 int blocksize =
2714 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2715 if (blocksize < PAGE_SIZE)
2716 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an "
2717 "experimental mount option 'dioread_nolock' "
2718 "for blocksize < PAGE_SIZE");
2719 }
2720
5f41fdae
EB		 2721 err = ext4_check_test_dummy_encryption(fc, sb);
2722 if (err)
2723 return err;
6e47a3cc
LC		 2724
2725 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2726 if (!sbi->s_journal) {
2727 ext4_msg(NULL, KERN_WARNING,
2728 "Remounting file system with no journal "
2729 "so ignoring journalled data option");
2730 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
4c246728
LC		 2731 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2732 test_opt(sb, DATA_FLAGS)) {
6e47a3cc
LC		 2733 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2734 "on remount");
2735 return -EINVAL;
2736 }
2737 }
2738
2739 if (is_remount) {
2740 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2741 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2742 ext4_msg(NULL, KERN_ERR, "can't mount with "
2743 "both data=journal and dax");
2744 return -EINVAL;
2745 }
2746
2747 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2748 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2749 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2750fail_dax_change_remount:
2751 ext4_msg(NULL, KERN_ERR, "can't change "
2752 "dax mount option while remounting");
2753 return -EINVAL;
2754 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2755 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2756 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2757 goto fail_dax_change_remount;
2758 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2759 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2760 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2761 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2762 goto fail_dax_change_remount;
2763 }
2764 }
2765
b6bd2435
LC		 2766 return ext4_check_quota_consistency(fc, sb);
2767}
2768
85456054 2769static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
4c94bff9 2770{
6e47a3cc 2771 struct ext4_fs_context *ctx = fc->fs_private;
da812f61 2772 struct ext4_sb_info *sbi = fc->s_fs_info;
6e47a3cc
LC		 2773
2774 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2775 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2776 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2777 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2778 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags;
2779 sbi->s_mount_flags |= ctx->vals_s_mount_flags;
2780 sb->s_flags &= ~ctx->mask_s_flags;
2781 sb->s_flags |= ctx->vals_s_flags;
2782
2783#define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2784 APPLY(s_commit_interval);
2785 APPLY(s_stripe);
2786 APPLY(s_max_batch_time);
2787 APPLY(s_min_batch_time);
2788 APPLY(s_want_extra_isize);
2789 APPLY(s_inode_readahead_blks);
2790 APPLY(s_max_dir_size_kb);
2791 APPLY(s_li_wait_mult);
2792 APPLY(s_resgid);
2793 APPLY(s_resuid);
2794
2795#ifdef CONFIG_EXT4_DEBUG
2796 APPLY(s_fc_debug_max_replay);
2797#endif
2798
2799 ext4_apply_quota_options(fc, sb);
85456054 2800 ext4_apply_test_dummy_encryption(ctx, sb);
6e47a3cc
LC		 2801}
2802
2803
2804static int ext4_validate_options(struct fs_context *fc)
2805{
ac27a0ec 2806#ifdef CONFIG_QUOTA
6e47a3cc 2807 struct ext4_fs_context *ctx = fc->fs_private;
4c94bff9 2808 char *usr_qf_name, *grp_qf_name;
6e47a3cc
LC		 2809
2810 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2811 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2812
33458eab 2813 if (usr_qf_name || grp_qf_name) {
6e47a3cc
LC		 2814 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2815 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
ac27a0ec 2816
6e47a3cc
LC		 2817 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2818 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
ac27a0ec 2819
6e47a3cc
LC		 2820 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2821 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
da812f61 2822 ext4_msg(NULL, KERN_ERR, "old and new quota "
6e47a3cc 2823 "format mixing");
da812f61 2824 return -EINVAL;
ac27a0ec 2825 }
ac27a0ec
DK		 2826 }
2827#endif
6e47a3cc 2828 return 1;
ac27a0ec
DK		 2829}
2830
2adf6da8
TT		 2831static inline void ext4_show_quota_options(struct seq_file *seq,
2832 struct super_block *sb)
2833{
2834#if defined(CONFIG_QUOTA)
2835 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 2836 char *usr_qf_name, *grp_qf_name;
2adf6da8
TT		 2837
2838 if (sbi->s_jquota_fmt) {
2839 char *fmtname = "";
2840
2841 switch (sbi->s_jquota_fmt) {
2842 case QFMT_VFS_OLD:
2843 fmtname = "vfsold";
2844 break;
2845 case QFMT_VFS_V0:
2846 fmtname = "vfsv0";
2847 break;
2848 case QFMT_VFS_V1:
2849 fmtname = "vfsv1";
2850 break;
2851 }
2852 seq_printf(seq, ",jqfmt=%s", fmtname);
2853 }
2854
33458eab
TT		 2855 rcu_read_lock();
2856 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2857 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2858 if (usr_qf_name)
2859 seq_show_option(seq, "usrjquota", usr_qf_name);
2860 if (grp_qf_name)
2861 seq_show_option(seq, "grpjquota", grp_qf_name);
2862 rcu_read_unlock();
2adf6da8
TT		 2863#endif
2864}
2865
5a916be1
TT		 2866static const char *token2str(int token)
2867{
97d8a670 2868 const struct fs_parameter_spec *spec;
5a916be1 2869
97d8a670
LC		 2870 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2871 if (spec->opt == token && !spec->type)
5a916be1 2872 break;
97d8a670 2873 return spec->name;
5a916be1
TT		 2874}
2875
2adf6da8
TT		 2876/*
2877 * Show an option if
2878 * - it's set to a non-default value OR
2879 * - if the per-sb default is different from the global default
2880 */
66acdcf4
TT		 2881static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2882 int nodefs)
2adf6da8 2883{
2adf6da8
TT		 2884 struct ext4_sb_info *sbi = EXT4_SB(sb);
2885 struct ext4_super_block *es = sbi->s_es;
e3645d72 2886 int def_errors;
5a916be1 2887 const struct mount_opts *m;
66acdcf4 2888 char sep = nodefs ? '\n' : ',';
2adf6da8 2889
66acdcf4
TT		 2890#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2891#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2892
2893 if (sbi->s_sb_block != 1)
5a916be1
TT		 2894 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2895
2896 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2897 int want_set = m->flags & MOPT_SET;
e3645d72
ZY		 2898 int opt_2 = m->flags & MOPT_2;
2899 unsigned int mount_opt, def_mount_opt;
2900
5a916be1 2901 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
ba2e524d 2902 m->flags & MOPT_SKIP)
5a916be1 2903 continue;
e3645d72
ZY		 2904
2905 if (opt_2) {
2906 mount_opt = sbi->s_mount_opt2;
2907 def_mount_opt = sbi->s_def_mount_opt2;
2908 } else {
2909 mount_opt = sbi->s_mount_opt;
2910 def_mount_opt = sbi->s_def_mount_opt;
2911 }
2912 /* skip if same as the default */
2913 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt)))
2914 continue;
2915 /* select Opt_noFoo vs Opt_Foo */
5a916be1 2916 if ((want_set &&
e3645d72
ZY		 2917 (mount_opt & m->mount_opt) != m->mount_opt) ||
2918 (!want_set && (mount_opt & m->mount_opt)))
2919 continue;
5a916be1 2920 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2921 }
5a916be1 2922
08cefc7a 2923 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2924 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2925 SEQ_OPTS_PRINT("resuid=%u",
2926 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2927 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2928 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2929 SEQ_OPTS_PRINT("resgid=%u",
2930 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2931 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2932 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2933 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 2934 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 2935 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 2936 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 2937 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 2938 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 2939 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 2940 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 2941 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 2942 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 2943 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
66acdcf4 2944 if (nodefs || sbi->s_stripe)
5a916be1 2945 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0 2946 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
e3645d72 2947 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5a916be1
TT		 2948 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2949 SEQ_OPTS_PUTS("data=journal");
2950 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2951 SEQ_OPTS_PUTS("data=ordered");
2952 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2953 SEQ_OPTS_PUTS("data=writeback");
2954 }
66acdcf4
TT		 2955 if (nodefs ||
2956 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 2957 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2958 sbi->s_inode_readahead_blks);
2adf6da8 2959
ceec0376 2960 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 2961 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 2962 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 2963 if (nodefs || sbi->s_max_dir_size_kb)
2964 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 2965 if (test_opt(sb, DATA_ERR_ABORT))
2966 SEQ_OPTS_PUTS("data_err=abort");
ed318a6c
EB		 2967
2968 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2adf6da8 2969
4f74d15f
EB		 2970 if (sb->s_flags & SB_INLINECRYPT)
2971 SEQ_OPTS_PUTS("inlinecrypt");
2972
9cb20f94
IW		 2973 if (test_opt(sb, DAX_ALWAYS)) {
2974 if (IS_EXT2_SB(sb))
2975 SEQ_OPTS_PUTS("dax");
2976 else
2977 SEQ_OPTS_PUTS("dax=always");
2978 } else if (test_opt2(sb, DAX_NEVER)) {
2979 SEQ_OPTS_PUTS("dax=never");
2980 } else if (test_opt2(sb, DAX_INODE)) {
2981 SEQ_OPTS_PUTS("dax=inode");
2982 }
3fa5d23e
OM		 2983
2984 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
2985 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
2986 SEQ_OPTS_PUTS("mb_optimize_scan=0");
2987 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
2988 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
2989 SEQ_OPTS_PUTS("mb_optimize_scan=1");
2990 }
2991
2adf6da8 2992 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 2993 return 0;
2994}
2995
66acdcf4
TT		 2996static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2997{
2998 return _ext4_show_options(seq, root->d_sb, 0);
2999}
3000
ebd173be 3001int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 3002{
3003 struct super_block *sb = seq->private;
3004 int rc;
3005
bc98a42c 3006 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 3007 rc = _ext4_show_options(seq, sb, 1);
3008 seq_puts(seq, "\n");
3009 return rc;
3010}
3011
617ba13b 3012static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 3013 int read_only)
3014{
617ba13b 3015 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 3016 int err = 0;
ac27a0ec 3017
617ba13b 3018 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 3019 ext4_msg(sb, KERN_ERR, "revision level too high, "
3020 "forcing read-only mode");
c89128a0 3021 err = -EROFS;
5adaccac 3022 goto done;
ac27a0ec
DK		 3023 }
3024 if (read_only)
281b5995 3025 goto done;
617ba13b 3026 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 3027 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3028 "running e2fsck is recommended");
c8b459f4 3029 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 3030 ext4_msg(sb, KERN_WARNING,
3031 "warning: mounting fs with errors, "
3032 "running e2fsck is recommended");
ed3ce80a 3033 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 3034 le16_to_cpu(es->s_mnt_count) >=
3035 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 3036 ext4_msg(sb, KERN_WARNING,
3037 "warning: maximal mount count reached, "
3038 "running e2fsck is recommended");
ac27a0ec 3039 else if (le32_to_cpu(es->s_checkinterval) &&
6a0678a7
AB		 3040 (ext4_get_tstamp(es, s_lastcheck) +
3041 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
b31e1552
ES		 3042 ext4_msg(sb, KERN_WARNING,
3043 "warning: checktime reached, "
3044 "running e2fsck is recommended");
0b8e58a1 3045 if (!sbi->s_journal)
0390131b 3046 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 3047 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 3048 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 3049 le16_add_cpu(&es->s_mnt_count, 1);
6a0678a7 3050 ext4_update_tstamp(es, s_mtime);
02f310fc 3051 if (sbi->s_journal) {
e2b911c5 3052 ext4_set_feature_journal_needs_recovery(sb);
02f310fc
JK		 3053 if (ext4_has_feature_orphan_file(sb))
3054 ext4_set_feature_orphan_present(sb);
3055 }
ac27a0ec 3056
4392fbc4 3057 err = ext4_commit_super(sb);
281b5995 3058done:
ac27a0ec 3059 if (test_opt(sb, DEBUG))
a9df9a49 3060 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 3061 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 3062 sb->s_blocksize,
3063 sbi->s_groups_count,
617ba13b
MC		 3064 EXT4_BLOCKS_PER_GROUP(sb),
3065 EXT4_INODES_PER_GROUP(sb),
a2595b8a 3066 sbi->s_mount_opt, sbi->s_mount_opt2);
c89128a0 3067 return err;
ac27a0ec
DK		 3068}
3069
117fff10
TT		 3070int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3071{
3072 struct ext4_sb_info *sbi = EXT4_SB(sb);
7c990728 3073 struct flex_groups **old_groups, **new_groups;
37b0b6b8 3074 int size, i, j;
117fff10
TT		 3075
3076 if (!sbi->s_log_groups_per_flex)
3077 return 0;
3078
3079 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3080 if (size <= sbi->s_flex_groups_allocated)
3081 return 0;
3082
7c990728
SJS		 3083 new_groups = kvzalloc(roundup_pow_of_two(size *
3084 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
117fff10 3085 if (!new_groups) {
7c990728
SJS		 3086 ext4_msg(sb, KERN_ERR,
3087 "not enough memory for %d flex group pointers", size);
117fff10
TT		 3088 return -ENOMEM;
3089 }
7c990728
SJS		 3090 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3091 new_groups[i] = kvzalloc(roundup_pow_of_two(
3092 sizeof(struct flex_groups)),
3093 GFP_KERNEL);
3094 if (!new_groups[i]) {
37b0b6b8
DC		 3095 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3096 kvfree(new_groups[j]);
7c990728
SJS		 3097 kvfree(new_groups);
3098 ext4_msg(sb, KERN_ERR,
3099 "not enough memory for %d flex groups", size);
3100 return -ENOMEM;
3101 }
117fff10 3102 }
7c990728
SJS		 3103 rcu_read_lock();
3104 old_groups = rcu_dereference(sbi->s_flex_groups);
3105 if (old_groups)
3106 memcpy(new_groups, old_groups,
3107 (sbi->s_flex_groups_allocated *
3108 sizeof(struct flex_groups *)));
3109 rcu_read_unlock();
3110 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3111 sbi->s_flex_groups_allocated = size;
3112 if (old_groups)
3113 ext4_kvfree_array_rcu(old_groups);
117fff10
TT		 3114 return 0;
3115}
3116
772cb7c8
JS		 3117static int ext4_fill_flex_info(struct super_block *sb)
3118{
3119 struct ext4_sb_info *sbi = EXT4_SB(sb);
3120 struct ext4_group_desc *gdp = NULL;
7c990728 3121 struct flex_groups *fg;
772cb7c8 3122 ext4_group_t flex_group;
117fff10 3123 int i, err;
772cb7c8 3124
503358ae 3125 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 3126 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 3127 sbi->s_log_groups_per_flex = 0;
3128 return 1;
3129 }
3130
117fff10
TT		 3131 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3132 if (err)
9933fc0a 3133 goto failed;
772cb7c8 3134
772cb7c8 3135 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 3136 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 3137
3138 flex_group = ext4_flex_group(sbi, i);
7c990728
SJS		 3139 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3140 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
90ba983f 3141 atomic64_add(ext4_free_group_clusters(sb, gdp),
7c990728
SJS		 3142 &fg->free_clusters);
3143 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
772cb7c8
JS		 3144 }
3145
3146 return 1;
3147failed:
3148 return 0;
3149}
3150
e2b911c5 3151static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 3152 struct ext4_group_desc *gdp)
717d50e4 3153{
b47820ed 3154 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 3155 __u16 crc = 0;
feb0ab32 3156 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 3157 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 3158
9aa5d32b 3159 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 3160 /* Use new metadata_csum algorithm */
feb0ab32 3161 __u32 csum32;
b47820ed 3162 __u16 dummy_csum = 0;
feb0ab32 3163
feb0ab32
DW		 3164 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3165 sizeof(le_group));
b47820ed
DJ		 3166 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3167 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3168 sizeof(dummy_csum));
3169 offset += sizeof(dummy_csum);
3170 if (offset < sbi->s_desc_size)
3171 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3172 sbi->s_desc_size - offset);
feb0ab32
DW		 3173
3174 crc = csum32 & 0xFFFF;
3175 goto out;
717d50e4
AD		 3176 }
3177
feb0ab32 3178 /* old crc16 code */
e2b911c5 3179 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 3180 return 0;
3181
feb0ab32
DW		 3182 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3183 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3184 crc = crc16(crc, (__u8 *)gdp, offset);
3185 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3186 /* for checksum of struct ext4_group_desc do the rest...*/
e2b911c5 3187 if (ext4_has_feature_64bit(sb) &&
feb0ab32
DW		 3188 offset < le16_to_cpu(sbi->s_es->s_desc_size))
3189 crc = crc16(crc, (__u8 *)gdp + offset,
3190 le16_to_cpu(sbi->s_es->s_desc_size) -
3191 offset);
3192
3193out:
717d50e4
AD		 3194 return cpu_to_le16(crc);
3195}
3196
feb0ab32 3197int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 3198 struct ext4_group_desc *gdp)
3199{
feb0ab32 3200 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 3201 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 3202 return 0;
3203
3204 return 1;
3205}
3206
feb0ab32
DW		 3207void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3208 struct ext4_group_desc *gdp)
3209{
3210 if (!ext4_has_group_desc_csum(sb))
3211 return;
e2b911c5 3212 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 3213}
3214
ac27a0ec 3215/* Called at mount-time, super-block is locked */
bfff6873 3216static int ext4_check_descriptors(struct super_block *sb,
829fa70d 3217 ext4_fsblk_t sb_block,
bfff6873 3218 ext4_group_t *first_not_zeroed)
ac27a0ec 3219{
617ba13b
MC		 3220 struct ext4_sb_info *sbi = EXT4_SB(sb);
3221 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3222 ext4_fsblk_t last_block;
44de022c 3223 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
bd81d8ee
LV		 3224 ext4_fsblk_t block_bitmap;
3225 ext4_fsblk_t inode_bitmap;
3226 ext4_fsblk_t inode_table;
ce421581 3227 int flexbg_flag = 0;
bfff6873 3228 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 3229
e2b911c5 3230 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 3231 flexbg_flag = 1;
3232
af5bc92d 3233 ext4_debug("Checking group descriptors");
ac27a0ec 3234
197cd65a
AM		 3235 for (i = 0; i < sbi->s_groups_count; i++) {
3236 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3237
ce421581 3238 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 3239 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 3240 else
3241 last_block = first_block +
617ba13b 3242 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 3243
bfff6873
LC		 3244 if ((grp == sbi->s_groups_count) &&
3245 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3246 grp = i;
3247
8fadc143 3248 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 3249 if (block_bitmap == sb_block) {
3250 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3251 "Block bitmap for group %u overlaps "
3252 "superblock", i);
18db4b4e
TT		 3253 if (!sb_rdonly(sb))
3254 return 0;
829fa70d 3255 }
77260807
TT		 3256 if (block_bitmap >= sb_block + 1 &&
3257 block_bitmap <= last_bg_block) {
3258 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3259 "Block bitmap for group %u overlaps "
3260 "block group descriptors", i);
3261 if (!sb_rdonly(sb))
3262 return 0;
3263 }
2b2d6d01 3264 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 3265 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 3266 "Block bitmap for group %u not in group "
b31e1552 3267 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 3268 return 0;
3269 }
8fadc143 3270 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 3271 if (inode_bitmap == sb_block) {
3272 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3273 "Inode bitmap for group %u overlaps "
3274 "superblock", i);
18db4b4e
TT		 3275 if (!sb_rdonly(sb))
3276 return 0;
829fa70d 3277 }
77260807
TT		 3278 if (inode_bitmap >= sb_block + 1 &&
3279 inode_bitmap <= last_bg_block) {
3280 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3281 "Inode bitmap for group %u overlaps "
3282 "block group descriptors", i);
3283 if (!sb_rdonly(sb))
3284 return 0;
3285 }
2b2d6d01 3286 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 3287 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 3288 "Inode bitmap for group %u not in group "
b31e1552 3289 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 3290 return 0;
3291 }
8fadc143 3292 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 3293 if (inode_table == sb_block) {
3294 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3295 "Inode table for group %u overlaps "
3296 "superblock", i);
18db4b4e
TT		 3297 if (!sb_rdonly(sb))
3298 return 0;
829fa70d 3299 }
77260807
TT		 3300 if (inode_table >= sb_block + 1 &&
3301 inode_table <= last_bg_block) {
3302 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3303 "Inode table for group %u overlaps "
3304 "block group descriptors", i);
3305 if (!sb_rdonly(sb))
3306 return 0;
3307 }
bd81d8ee 3308 if (inode_table < first_block ||
2b2d6d01 3309 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 3310 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 3311 "Inode table for group %u not in group "
b31e1552 3312 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 3313 return 0;
3314 }
955ce5f5 3315 ext4_lock_group(sb, i);
feb0ab32 3316 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 3317 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3318 "Checksum for group %u failed (%u!=%u)",
e2b911c5 3319 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 3320 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 3321 if (!sb_rdonly(sb)) {
955ce5f5 3322 ext4_unlock_group(sb, i);
8a266467 3323 return 0;
7ee1ec4c 3324 }
717d50e4 3325 }
955ce5f5 3326 ext4_unlock_group(sb, i);
ce421581
JS		 3327 if (!flexbg_flag)
3328 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 3329 }
bfff6873
LC		 3330 if (NULL != first_not_zeroed)
3331 *first_not_zeroed = grp;
ac27a0ec
DK		 3332 return 1;
3333}
3334
cd2291a4
ES		 3335/*
3336 * Maximal extent format file size.
3337 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3338 * extent format containers, within a sector_t, and within i_blocks
3339 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3340 * so that won't be a limiting factor.
3341 *
f17722f9
LC		 3342 * However there is other limiting factor. We do store extents in the form
3343 * of starting block and length, hence the resulting length of the extent
3344 * covering maximum file size must fit into on-disk format containers as
3345 * well. Given that length is always by 1 unit bigger than max unit (because
3346 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3347 *
cd2291a4
ES		 3348 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3349 */
f287a1a5 3350static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 3351{
3352 loff_t res;
3353 loff_t upper_limit = MAX_LFS_FILESIZE;
3354
72deb455
CH		 3355 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3356
3357 if (!has_huge_files) {
cd2291a4
ES		 3358 upper_limit = (1LL << 32) - 1;
3359
3360 /* total blocks in file system block size */
3361 upper_limit >>= (blkbits - 9);
3362 upper_limit <<= blkbits;
3363 }
3364
f17722f9
LC		 3365 /*
3366 * 32-bit extent-start container, ee_block. We lower the maxbytes
3367 * by one fs block, so ee_len can cover the extent of maximum file
3368 * size
3369 */
3370 res = (1LL << 32) - 1;
cd2291a4 3371 res <<= blkbits;
cd2291a4
ES		 3372
3373 /* Sanity check against vm- & vfs- imposed limits */
3374 if (res > upper_limit)
3375 res = upper_limit;
3376
3377 return res;
3378}
ac27a0ec 3379
ac27a0ec 3380/*
cd2291a4 3381 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 3382 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3383 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 3384 */
f287a1a5 3385static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 3386{
5c93e8ec 3387 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
0fc1b451 3388 int meta_blocks;
5c93e8ec 3389 unsigned int ppb = 1 << (bits - 2);
75ca6ad4
RH		 3390
3391 /*
3392 * This is calculated to be the largest file size for a dense, block
0b8e58a1
AD		 3393 * mapped file such that the file's total number of 512-byte sectors,
3394 * including data and all indirect blocks, does not exceed (2^48 - 1).
3395 *
3396 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3397 * number of 512-byte sectors of the file.
0fc1b451 3398 */
72deb455 3399 if (!has_huge_files) {
0fc1b451 3400 /*
72deb455
CH		 3401 * !has_huge_files or implies that the inode i_block field
3402 * represents total file blocks in 2^32 512-byte sectors ==
3403 * size of vfs inode i_blocks * 8
0fc1b451
AK		 3404 */
3405 upper_limit = (1LL << 32) - 1;
3406
3407 /* total blocks in file system block size */
3408 upper_limit >>= (bits - 9);
3409
3410 } else {
8180a562
AK		 3411 /*
3412 * We use 48 bit ext4_inode i_blocks
3413 * With EXT4_HUGE_FILE_FL set the i_blocks
3414 * represent total number of blocks in
3415 * file system block size
3416 */
0fc1b451
AK		 3417 upper_limit = (1LL << 48) - 1;
3418
0fc1b451
AK		 3419 }
3420
5c93e8ec
ZY		 3421 /* Compute how many blocks we can address by block tree */
3422 res += ppb;
3423 res += ppb * ppb;
3424 res += ((loff_t)ppb) * ppb * ppb;
3425 /* Compute how many metadata blocks are needed */
3426 meta_blocks = 1;
3427 meta_blocks += 1 + ppb;
3428 meta_blocks += 1 + ppb + ppb * ppb;
3429 /* Does block tree limit file size? */
3430 if (res + meta_blocks <= upper_limit)
3431 goto check_lfs;
3432
3433 res = upper_limit;
3434 /* How many metadata blocks are needed for addressing upper_limit? */
3435 upper_limit -= EXT4_NDIR_BLOCKS;
0fc1b451
AK		 3436 /* indirect blocks */
3437 meta_blocks = 1;
5c93e8ec 3438 upper_limit -= ppb;
0fc1b451 3439 /* double indirect blocks */
5c93e8ec
ZY		 3440 if (upper_limit < ppb * ppb) {
3441 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3442 res -= meta_blocks;
3443 goto check_lfs;
3444 }
3445 meta_blocks += 1 + ppb;
3446 upper_limit -= ppb * ppb;
3447 /* tripple indirect blocks for the rest */
3448 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3449 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3450 res -= meta_blocks;
3451check_lfs:
ac27a0ec 3452 res <<= bits;
0fc1b451
AK		 3453 if (res > MAX_LFS_FILESIZE)
3454 res = MAX_LFS_FILESIZE;
3455
5c93e8ec 3456 return res;
ac27a0ec
DK		 3457}
3458
617ba13b 3459static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 3460 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 3461{
617ba13b 3462 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 3463 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 3464 int has_super = 0;
3465
3466 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3467
e2b911c5 3468 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 3469 return logical_sb_block + nr + 1;
ac27a0ec 3470 bg = sbi->s_desc_per_block * nr;
617ba13b 3471 if (ext4_bg_has_super(sb, bg))
ac27a0ec 3472 has_super = 1;
0b8e58a1 3473
bd63f6b0
DW		 3474 /*
3475 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3476 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3477 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3478 * compensate.
3479 */
3480 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 3481 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 3482 has_super++;
3483
617ba13b 3484 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 3485}
3486
c9de560d
AT		 3487/**
3488 * ext4_get_stripe_size: Get the stripe size.
3489 * @sbi: In memory super block info
3490 *
3491 * If we have specified it via mount option, then
3492 * use the mount option value. If the value specified at mount time is
3493 * greater than the blocks per group use the super block value.
3494 * If the super block value is greater than blocks per group return 0.
3495 * Allocator needs it be less than blocks per group.
3496 *
3497 */
3498static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3499{
3500 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3501 unsigned long stripe_width =
3502 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 3503 int ret;
c9de560d
AT		 3504
3505 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 3506 ret = sbi->s_stripe;
5469d7c3 3507 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 3508 ret = stripe_width;
5469d7c3 3509 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 3510 ret = stride;
3511 else
3512 ret = 0;
c9de560d 3513
3eb08658
DE		 3514 /*
3515 * If the stripe width is 1, this makes no sense and
3516 * we set it to 0 to turn off stripe handling code.
3517 */
3518 if (ret <= 1)
3519 ret = 0;
c9de560d 3520
3eb08658 3521 return ret;
c9de560d 3522}
ac27a0ec 3523
a13fb1a4
ES		 3524/*
3525 * Check whether this filesystem can be mounted based on
3526 * the features present and the RDONLY/RDWR mount requested.
3527 * Returns 1 if this filesystem can be mounted as requested,
3528 * 0 if it cannot be.
3529 */
25c6d98f 3530int ext4_feature_set_ok(struct super_block *sb, int readonly)
a13fb1a4 3531{
e2b911c5 3532 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 3533 ext4_msg(sb, KERN_ERR,
3534 "Couldn't mount because of "
3535 "unsupported optional features (%x)",
3536 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3537 ~EXT4_FEATURE_INCOMPAT_SUPP));
3538 return 0;
3539 }
3540
5298d4bf 3541#if !IS_ENABLED(CONFIG_UNICODE)
c83ad55e
GKB		 3542 if (ext4_has_feature_casefold(sb)) {
3543 ext4_msg(sb, KERN_ERR,
3544 "Filesystem with casefold feature cannot be "
3545 "mounted without CONFIG_UNICODE");
3546 return 0;
3547 }
3548#endif
3549
a13fb1a4
ES		 3550 if (readonly)
3551 return 1;
3552
e2b911c5 3553 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 3554 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 3555 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 3556 return 1;
3557 }
3558
a13fb1a4 3559 /* Check that feature set is OK for a read-write mount */
e2b911c5 3560 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 3561 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3562 "unsupported optional features (%x)",
3563 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3564 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3565 return 0;
3566 }
e2b911c5 3567 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 3568 ext4_msg(sb, KERN_ERR,
3569 "Can't support bigalloc feature without "
3570 "extents feature\n");
3571 return 0;
3572 }
7c319d32 3573
9db176bc 3574#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
d65d87a0
TT		 3575 if (!readonly && (ext4_has_feature_quota(sb) ||
3576 ext4_has_feature_project(sb))) {
7c319d32 3577 ext4_msg(sb, KERN_ERR,
d65d87a0 3578 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
689c958c
LX		 3579 return 0;
3580 }
7c319d32 3581#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 3582 return 1;
3583}
3584
66e61a9e
TT		 3585/*
3586 * This function is called once a day if we have errors logged
3587 * on the file system
3588 */
235699a8 3589static void print_daily_error_info(struct timer_list *t)
66e61a9e 3590{
235699a8
KC		 3591 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3592 struct super_block *sb = sbi->s_sb;
3593 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 3594
3595 if (es->s_error_count)
ae0f78de
TT		 3596 /* fsck newer than v1.41.13 is needed to clean this condition. */
3597 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 3598 le32_to_cpu(es->s_error_count));
3599 if (es->s_first_error_time) {
6a0678a7
AB		 3600 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3601 sb->s_id,
3602 ext4_get_tstamp(es, s_first_error_time),
66e61a9e
TT		 3603 (int) sizeof(es->s_first_error_func),
3604 es->s_first_error_func,
3605 le32_to_cpu(es->s_first_error_line));
3606 if (es->s_first_error_ino)
651e1c3b 3607 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3608 le32_to_cpu(es->s_first_error_ino));
3609 if (es->s_first_error_block)
651e1c3b 3610 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3611 le64_to_cpu(es->s_first_error_block));
651e1c3b 3612 printk(KERN_CONT "\n");
66e61a9e
TT		 3613 }
3614 if (es->s_last_error_time) {
6a0678a7
AB		 3615 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3616 sb->s_id,
3617 ext4_get_tstamp(es, s_last_error_time),
66e61a9e
TT		 3618 (int) sizeof(es->s_last_error_func),
3619 es->s_last_error_func,
3620 le32_to_cpu(es->s_last_error_line));
3621 if (es->s_last_error_ino)
651e1c3b 3622 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3623 le32_to_cpu(es->s_last_error_ino));
3624 if (es->s_last_error_block)
651e1c3b 3625 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3626 le64_to_cpu(es->s_last_error_block));
651e1c3b 3627 printk(KERN_CONT "\n");
66e61a9e
TT		 3628 }
3629 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3630}
3631
bfff6873
LC		 3632/* Find next suitable group and run ext4_init_inode_table */
3633static int ext4_run_li_request(struct ext4_li_request *elr)
3634{
3635 struct ext4_group_desc *gdp = NULL;
3d392b26
TT		 3636 struct super_block *sb = elr->lr_super;
3637 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3638 ext4_group_t group = elr->lr_next_group;
3d392b26 3639 unsigned int prefetch_ios = 0;
bfff6873 3640 int ret = 0;
39fec688 3641 u64 start_time;
bfff6873 3642
3d392b26
TT		 3643 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3644 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3645 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3646 if (prefetch_ios)
3647 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3648 prefetch_ios);
3649 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3650 prefetch_ios);
3651 if (group >= elr->lr_next_group) {
3652 ret = 1;
3653 if (elr->lr_first_not_zeroed != ngroups &&
3654 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3655 elr->lr_next_group = elr->lr_first_not_zeroed;
3656 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3657 ret = 0;
3658 }
3659 }
3660 return ret;
3661 }
bfff6873 3662
3d392b26 3663 for (; group < ngroups; group++) {
bfff6873
LC		 3664 gdp = ext4_get_group_desc(sb, group, NULL);
3665 if (!gdp) {
3666 ret = 1;
3667 break;
3668 }
3669
3670 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3671 break;
3672 }
3673
7f511862 3674 if (group >= ngroups)
bfff6873
LC		 3675 ret = 1;
3676
3677 if (!ret) {
39fec688 3678 start_time = ktime_get_real_ns();
bfff6873
LC		 3679 ret = ext4_init_inode_table(sb, group,
3680 elr->lr_timeout ? 0 : 1);
3d392b26 3681 trace_ext4_lazy_itable_init(sb, group);
bfff6873 3682 if (elr->lr_timeout == 0) {
39fec688
SX		 3683 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3684 EXT4_SB(elr->lr_super)->s_li_wait_mult);
bfff6873
LC		 3685 }
3686 elr->lr_next_sched = jiffies + elr->lr_timeout;
3687 elr->lr_next_group = group + 1;
3688 }
bfff6873
LC		 3689 return ret;
3690}
3691
3692/*
3693 * Remove lr_request from the list_request and free the
4ed5c033 3694 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 3695 */
3696static void ext4_remove_li_request(struct ext4_li_request *elr)
3697{
bfff6873
LC		 3698 if (!elr)
3699 return;
3700
bfff6873 3701 list_del(&elr->lr_request);
3d392b26 3702 EXT4_SB(elr->lr_super)->s_li_request = NULL;
bfff6873
LC		 3703 kfree(elr);
3704}
3705
3706static void ext4_unregister_li_request(struct super_block *sb)
3707{
1bb933fb
LC		 3708 mutex_lock(&ext4_li_mtx);
3709 if (!ext4_li_info) {
3710 mutex_unlock(&ext4_li_mtx);
bfff6873 3711 return;
1bb933fb 3712 }
bfff6873
LC		 3713
3714 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 3715 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 3716 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 3717 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 3718}
3719
8f1f7453
ES		 3720static struct task_struct *ext4_lazyinit_task;
3721
bfff6873
LC		 3722/*
3723 * This is the function where ext4lazyinit thread lives. It walks
3724 * through the request list searching for next scheduled filesystem.
3725 * When such a fs is found, run the lazy initialization request
3726 * (ext4_rn_li_request) and keep track of the time spend in this
3727 * function. Based on that time we compute next schedule time of
3728 * the request. When walking through the list is complete, compute
3729 * next waking time and put itself into sleep.
3730 */
3731static int ext4_lazyinit_thread(void *arg)
3732{
c30365b9 3733 struct ext4_lazy_init *eli = arg;
bfff6873
LC		 3734 struct list_head *pos, *n;
3735 struct ext4_li_request *elr;
4ed5c033 3736 unsigned long next_wakeup, cur;
bfff6873
LC		 3737
3738 BUG_ON(NULL == eli);
3b575495 3739 set_freezable();
bfff6873 3740
bfff6873
LC		 3741cont_thread:
3742 while (true) {
3743 next_wakeup = MAX_JIFFY_OFFSET;
3744
3745 mutex_lock(&eli->li_list_mtx);
3746 if (list_empty(&eli->li_request_list)) {
3747 mutex_unlock(&eli->li_list_mtx);
3748 goto exit_thread;
3749 }
bfff6873 3750 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3751 int err = 0;
3752 int progress = 0;
bfff6873
LC		 3753 elr = list_entry(pos, struct ext4_li_request,
3754 lr_request);
3755
e22834f0
DM		 3756 if (time_before(jiffies, elr->lr_next_sched)) {
3757 if (time_before(elr->lr_next_sched, next_wakeup))
3758 next_wakeup = elr->lr_next_sched;
3759 continue;
3760 }
3761 if (down_read_trylock(&elr->lr_super->s_umount)) {
3762 if (sb_start_write_trylock(elr->lr_super)) {
3763 progress = 1;
3764 /*
3765 * We hold sb->s_umount, sb can not
3766 * be removed from the list, it is
3767 * now safe to drop li_list_mtx
3768 */
3769 mutex_unlock(&eli->li_list_mtx);
3770 err = ext4_run_li_request(elr);
3771 sb_end_write(elr->lr_super);
3772 mutex_lock(&eli->li_list_mtx);
3773 n = pos->next;
b2c78cd0 3774 }
e22834f0
DM		 3775 up_read((&elr->lr_super->s_umount));
3776 }
3777 /* error, remove the lazy_init job */
3778 if (err) {
3779 ext4_remove_li_request(elr);
3780 continue;
3781 }
3782 if (!progress) {
3783 elr->lr_next_sched = jiffies +
8032bf12 3784 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
bfff6873 3785 }
bfff6873
LC		 3786 if (time_before(elr->lr_next_sched, next_wakeup))
3787 next_wakeup = elr->lr_next_sched;
3788 }
3789 mutex_unlock(&eli->li_list_mtx);
3790
a0acae0e 3791 try_to_freeze();
bfff6873 3792
4ed5c033
LC		 3793 cur = jiffies;
3794 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3795 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3796 cond_resched();
3797 continue;
3798 }
3799
4ed5c033
LC		 3800 schedule_timeout_interruptible(next_wakeup - cur);
3801
8f1f7453
ES		 3802 if (kthread_should_stop()) {
3803 ext4_clear_request_list();
3804 goto exit_thread;
3805 }
bfff6873
LC		 3806 }
3807
3808exit_thread:
3809 /*
3810 * It looks like the request list is empty, but we need
3811 * to check it under the li_list_mtx lock, to prevent any
3812 * additions into it, and of course we should lock ext4_li_mtx
3813 * to atomically free the list and ext4_li_info, because at
3814 * this point another ext4 filesystem could be registering
3815 * new one.
3816 */
3817 mutex_lock(&ext4_li_mtx);
3818 mutex_lock(&eli->li_list_mtx);
3819 if (!list_empty(&eli->li_request_list)) {
3820 mutex_unlock(&eli->li_list_mtx);
3821 mutex_unlock(&ext4_li_mtx);
3822 goto cont_thread;
3823 }
3824 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3825 kfree(ext4_li_info);
3826 ext4_li_info = NULL;
3827 mutex_unlock(&ext4_li_mtx);
3828
3829 return 0;
3830}
3831
3832static void ext4_clear_request_list(void)
3833{
3834 struct list_head *pos, *n;
3835 struct ext4_li_request *elr;
3836
3837 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3838 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3839 elr = list_entry(pos, struct ext4_li_request,
3840 lr_request);
3841 ext4_remove_li_request(elr);
3842 }
3843 mutex_unlock(&ext4_li_info->li_list_mtx);
3844}
3845
3846static int ext4_run_lazyinit_thread(void)
3847{
8f1f7453
ES		 3848 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3849 ext4_li_info, "ext4lazyinit");
3850 if (IS_ERR(ext4_lazyinit_task)) {
3851 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3852 ext4_clear_request_list();
bfff6873
LC		 3853 kfree(ext4_li_info);
3854 ext4_li_info = NULL;
92b97816 3855 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3856 "initialization thread\n",
3857 err);
3858 return err;
3859 }
3860 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3861 return 0;
3862}
3863
3864/*
3865 * Check whether it make sense to run itable init. thread or not.
3866 * If there is at least one uninitialized inode table, return
3867 * corresponding group number, else the loop goes through all
3868 * groups and return total number of groups.
3869 */
3870static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3871{
3872 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3873 struct ext4_group_desc *gdp = NULL;
3874
8844618d
TT		 3875 if (!ext4_has_group_desc_csum(sb))
3876 return ngroups;
3877
bfff6873
LC		 3878 for (group = 0; group < ngroups; group++) {
3879 gdp = ext4_get_group_desc(sb, group, NULL);
3880 if (!gdp)
3881 continue;
3882
50122847 3883 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
bfff6873
LC		 3884 break;
3885 }
3886
3887 return group;
3888}
3889
3890static int ext4_li_info_new(void)
3891{
3892 struct ext4_lazy_init *eli = NULL;
3893
3894 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3895 if (!eli)
3896 return -ENOMEM;
3897
bfff6873
LC		 3898 INIT_LIST_HEAD(&eli->li_request_list);
3899 mutex_init(&eli->li_list_mtx);
3900
bfff6873
LC		 3901 eli->li_state |= EXT4_LAZYINIT_QUIT;
3902
3903 ext4_li_info = eli;
3904
3905 return 0;
3906}
3907
3908static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3909 ext4_group_t start)
3910{
bfff6873 3911 struct ext4_li_request *elr;
bfff6873
LC		 3912
3913 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3914 if (!elr)
3915 return NULL;
3916
3917 elr->lr_super = sb;
3d392b26 3918 elr->lr_first_not_zeroed = start;
21175ca4 3919 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3d392b26
TT		 3920 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3921 elr->lr_next_group = start;
21175ca4
HS		 3922 } else {
3923 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3d392b26 3924 }
bfff6873
LC		 3925
3926 /*
3927 * Randomize first schedule time of the request to
3928 * spread the inode table initialization requests
3929 * better.
3930 */
8032bf12 3931 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
bfff6873
LC		 3932 return elr;
3933}
3934
7f511862
TT		 3935int ext4_register_li_request(struct super_block *sb,
3936 ext4_group_t first_not_zeroed)
bfff6873
LC		 3937{
3938 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3939 struct ext4_li_request *elr = NULL;
49598e04 3940 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 3941 int ret = 0;
bfff6873 3942
7f511862 3943 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3944 if (sbi->s_li_request != NULL) {
3945 /*
3946 * Reset timeout so it can be computed again, because
3947 * s_li_wait_mult might have changed.
3948 */
3949 sbi->s_li_request->lr_timeout = 0;
7f511862 3950 goto out;
51ce6511 3951 }
bfff6873 3952
426d15ad
JT		 3953 if (sb_rdonly(sb) ||
3954 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3955 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
7f511862 3956 goto out;
bfff6873
LC		 3957
3958 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3959 if (!elr) {
3960 ret = -ENOMEM;
3961 goto out;
3962 }
bfff6873
LC		 3963
3964 if (NULL == ext4_li_info) {
3965 ret = ext4_li_info_new();
3966 if (ret)
3967 goto out;
3968 }
3969
3970 mutex_lock(&ext4_li_info->li_list_mtx);
3971 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3972 mutex_unlock(&ext4_li_info->li_list_mtx);
3973
3974 sbi->s_li_request = elr;
46e4690b
TM		 3975 /*
3976 * set elr to NULL here since it has been inserted to
3977 * the request_list and the removal and free of it is
3978 * handled by ext4_clear_request_list from now on.
3979 */
3980 elr = NULL;
bfff6873
LC		 3981
3982 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3983 ret = ext4_run_lazyinit_thread();
3984 if (ret)
3985 goto out;
3986 }
bfff6873 3987out:
beed5ecb
NK		 3988 mutex_unlock(&ext4_li_mtx);
3989 if (ret)
bfff6873 3990 kfree(elr);
bfff6873
LC		 3991 return ret;
3992}
3993
3994/*
3995 * We do not need to lock anything since this is called on
3996 * module unload.
3997 */
3998static void ext4_destroy_lazyinit_thread(void)
3999{
4000 /*
4001 * If thread exited earlier
4002 * there's nothing to be done.
4003 */
8f1f7453 4004 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 4005 return;
4006
8f1f7453 4007 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 4008}
4009
25ed6e8a
DW		 4010static int set_journal_csum_feature_set(struct super_block *sb)
4011{
4012 int ret = 1;
4013 int compat, incompat;
4014 struct ext4_sb_info *sbi = EXT4_SB(sb);
4015
9aa5d32b 4016 if (ext4_has_metadata_csum(sb)) {
db9ee220 4017 /* journal checksum v3 */
25ed6e8a 4018 compat = 0;
db9ee220 4019 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 4020 } else {
4021 /* journal checksum v1 */
4022 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4023 incompat = 0;
4024 }
4025
feb8c6d3
DW		 4026 jbd2_journal_clear_features(sbi->s_journal,
4027 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4028 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4029 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 4030 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4031 ret = jbd2_journal_set_features(sbi->s_journal,
4032 compat, 0,
4033 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4034 incompat);
4035 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4036 ret = jbd2_journal_set_features(sbi->s_journal,
4037 compat, 0,
4038 incompat);
4039 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4040 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4041 } else {
feb8c6d3
DW		 4042 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4043 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 4044 }
4045
4046 return ret;
4047}
4048
952fc18e
TT		 4049/*
4050 * Note: calculating the overhead so we can be compatible with
4051 * historical BSD practice is quite difficult in the face of
4052 * clusters/bigalloc. This is because multiple metadata blocks from
4053 * different block group can end up in the same allocation cluster.
4054 * Calculating the exact overhead in the face of clustered allocation
4055 * requires either O(all block bitmaps) in memory or O(number of block
4056 * groups**2) in time. We will still calculate the superblock for
4057 * older file systems --- and if we come across with a bigalloc file
4058 * system with zero in s_overhead_clusters the estimate will be close to
4059 * correct especially for very large cluster sizes --- but for newer
4060 * file systems, it's better to calculate this figure once at mkfs
4061 * time, and store it in the superblock. If the superblock value is
4062 * present (even for non-bigalloc file systems), we will use it.
4063 */
4064static int count_overhead(struct super_block *sb, ext4_group_t grp,
4065 char *buf)
4066{
4067 struct ext4_sb_info *sbi = EXT4_SB(sb);
4068 struct ext4_group_desc *gdp;
4069 ext4_fsblk_t first_block, last_block, b;
4070 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4071 int s, j, count = 0;
10b01ee9 4072 int has_super = ext4_bg_has_super(sb, grp);
952fc18e 4073
e2b911c5 4074 if (!ext4_has_feature_bigalloc(sb))
10b01ee9
TT		 4075 return (has_super + ext4_bg_num_gdb(sb, grp) +
4076 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
0548bbb8
TT		 4077 sbi->s_itb_per_group + 2);
4078
952fc18e
TT		 4079 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4080 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4081 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4082 for (i = 0; i < ngroups; i++) {
4083 gdp = ext4_get_group_desc(sb, i, NULL);
4084 b = ext4_block_bitmap(sb, gdp);
4085 if (b >= first_block && b <= last_block) {
4086 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4087 count++;
4088 }
4089 b = ext4_inode_bitmap(sb, gdp);
4090 if (b >= first_block && b <= last_block) {
4091 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4092 count++;
4093 }
4094 b = ext4_inode_table(sb, gdp);
4095 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4096 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4097 int c = EXT4_B2C(sbi, b - first_block);
4098 ext4_set_bit(c, buf);
4099 count++;
4100 }
4101 if (i != grp)
4102 continue;
4103 s = 0;
4104 if (ext4_bg_has_super(sb, grp)) {
4105 ext4_set_bit(s++, buf);
4106 count++;
4107 }
c48ae41b
TT		 4108 j = ext4_bg_num_gdb(sb, grp);
4109 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4110 ext4_error(sb, "Invalid number of block group "
4111 "descriptor blocks: %d", j);
4112 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 4113 }
c48ae41b
TT		 4114 count += j;
4115 for (; j > 0; j--)
4116 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 4117 }
4118 if (!count)
4119 return 0;
4120 return EXT4_CLUSTERS_PER_GROUP(sb) -
4121 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4122}
4123
4124/*
4125 * Compute the overhead and stash it in sbi->s_overhead
4126 */
4127int ext4_calculate_overhead(struct super_block *sb)
4128{
4129 struct ext4_sb_info *sbi = EXT4_SB(sb);
4130 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 4131 struct inode *j_inode;
4132 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 4133 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4134 ext4_fsblk_t overhead = 0;
4fdb5543 4135 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 4136
952fc18e
TT		 4137 if (!buf)
4138 return -ENOMEM;
4139
4140 /*
4141 * Compute the overhead (FS structures). This is constant
4142 * for a given filesystem unless the number of block groups
4143 * changes so we cache the previous value until it does.
4144 */
4145
4146 /*
4147 * All of the blocks before first_data_block are overhead
4148 */
4149 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4150
4151 /*
4152 * Add the overhead found in each block group
4153 */
4154 for (i = 0; i < ngroups; i++) {
4155 int blks;
4156
4157 blks = count_overhead(sb, i, buf);
4158 overhead += blks;
4159 if (blks)
4160 memset(buf, 0, PAGE_SIZE);
4161 cond_resched();
4162 }
3c816ded
EW		 4163
4164 /*
4165 * Add the internal journal blocks whether the journal has been
4166 * loaded or not
4167 */
ee7ed3aa 4168 if (sbi->s_journal && !sbi->s_journal_bdev)
ede7dc7f 4169 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
f1eec3b0
RH		 4170 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4171 /* j_inum for internal journal is non-zero */
3c816ded
EW		 4172 j_inode = ext4_get_journal_inode(sb, j_inum);
4173 if (j_inode) {
4174 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4175 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4176 iput(j_inode);
4177 } else {
4178 ext4_msg(sb, KERN_ERR, "can't get journal size");
4179 }
4180 }
952fc18e
TT		 4181 sbi->s_overhead = overhead;
4182 smp_wmb();
4183 free_page((unsigned long) buf);
4184 return 0;
4185}
4186
b5799018 4187static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 4188{
4189 ext4_fsblk_t resv_clusters;
b5799018 4190 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 4191
30fac0f7
JK		 4192 /*
4193 * There's no need to reserve anything when we aren't using extents.
4194 * The space estimates are exact, there are no unwritten extents,
4195 * hole punching doesn't need new metadata... This is needed especially
4196 * to keep ext2/3 backward compatibility.
4197 */
e2b911c5 4198 if (!ext4_has_feature_extents(sb))
b5799018 4199 return;
27dd4385
LC		 4200 /*
4201 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4202 * This should cover the situations where we can not afford to run
4203 * out of space like for example punch hole, or converting
556615dc 4204 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 4205 * allocation would require 1, or 2 blocks, higher numbers are
4206 * very rare.
4207 */
b5799018
TT		 4208 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4209 sbi->s_cluster_bits);
27dd4385
LC		 4210
4211 do_div(resv_clusters, 50);
4212 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4213
b5799018 4214 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 4215}
4216
ca9b404f
RA		 4217static const char *ext4_quota_mode(struct super_block *sb)
4218{
4219#ifdef CONFIG_QUOTA
4220 if (!ext4_quota_capable(sb))
4221 return "none";
4222
4223 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4224 return "journalled";
4225 else
4226 return "writeback";
4227#else
4228 return "disabled";
4229#endif
4230}
4231
188c299e
JK		 4232static void ext4_setup_csum_trigger(struct super_block *sb,
4233 enum ext4_journal_trigger_type type,
4234 void (*trigger)(
4235 struct jbd2_buffer_trigger_type *type,
4236 struct buffer_head *bh,
4237 void *mapped_data,
4238 size_t size))
4239{
4240 struct ext4_sb_info *sbi = EXT4_SB(sb);
4241
4242 sbi->s_journal_triggers[type].sb = sb;
4243 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4244}
4245
7edfd85b
LC		 4246static void ext4_free_sbi(struct ext4_sb_info *sbi)
4247{
4248 if (!sbi)
4249 return;
4250
4251 kfree(sbi->s_blockgroup_lock);
8012b866 4252 fs_put_dax(sbi->s_daxdev, NULL);
7edfd85b
LC		 4253 kfree(sbi);
4254}
4255
4256static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4257{
4258 struct ext4_sb_info *sbi;
4259
4260 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4261 if (!sbi)
4262 return NULL;
4263
8012b866
SR		 4264 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4265 NULL, NULL);
7edfd85b
LC		 4266
4267 sbi->s_blockgroup_lock =
4268 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4269
4270 if (!sbi->s_blockgroup_lock)
4271 goto err_out;
4272
4273 sb->s_fs_info = sbi;
4274 sbi->s_sb = sb;
4275 return sbi;
4276err_out:
8012b866 4277 fs_put_dax(sbi->s_daxdev, NULL);
7edfd85b
LC		 4278 kfree(sbi);
4279 return NULL;
4280}
4281
5f6d662d
JY		 4282static void ext4_set_def_opts(struct super_block *sb,
4283 struct ext4_super_block *es)
4284{
4285 unsigned long def_mount_opts;
4286
4287 /* Set defaults before we parse the mount options */
4288 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4289 set_opt(sb, INIT_INODE_TABLE);
4290 if (def_mount_opts & EXT4_DEFM_DEBUG)
4291 set_opt(sb, DEBUG);
4292 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4293 set_opt(sb, GRPID);
4294 if (def_mount_opts & EXT4_DEFM_UID16)
4295 set_opt(sb, NO_UID32);
4296 /* xattr user namespace & acls are now defaulted on */
4297 set_opt(sb, XATTR_USER);
4298#ifdef CONFIG_EXT4_FS_POSIX_ACL
4299 set_opt(sb, POSIX_ACL);
4300#endif
4301 if (ext4_has_feature_fast_commit(sb))
4302 set_opt2(sb, JOURNAL_FAST_COMMIT);
4303 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4304 if (ext4_has_metadata_csum(sb))
4305 set_opt(sb, JOURNAL_CHECKSUM);
4306
4307 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4308 set_opt(sb, JOURNAL_DATA);
4309 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4310 set_opt(sb, ORDERED_DATA);
4311 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4312 set_opt(sb, WRITEBACK_DATA);
4313
4314 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4315 set_opt(sb, ERRORS_PANIC);
4316 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4317 set_opt(sb, ERRORS_CONT);
4318 else
4319 set_opt(sb, ERRORS_RO);
4320 /* block_validity enabled by default; disable with noblock_validity */
4321 set_opt(sb, BLOCK_VALIDITY);
4322 if (def_mount_opts & EXT4_DEFM_DISCARD)
4323 set_opt(sb, DISCARD);
4324
4325 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4326 set_opt(sb, BARRIER);
4327
4328 /*
4329 * enable delayed allocation by default
4330 * Use -o nodelalloc to turn it off
4331 */
4332 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4333 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4334 set_opt(sb, DELALLOC);
3df11e27
JY		 4335
4336 if (sb->s_blocksize == PAGE_SIZE)
4337 set_opt(sb, DIOREAD_NOLOCK);
5f6d662d
JY		 4338}
4339
c8267c51 4340static int ext4_handle_clustersize(struct super_block *sb)
4a8557b0
JY		 4341{
4342 struct ext4_sb_info *sbi = EXT4_SB(sb);
4343 struct ext4_super_block *es = sbi->s_es;
4344 int clustersize;
4345
4346 /* Handle clustersize */
4347 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4348 if (ext4_has_feature_bigalloc(sb)) {
c8267c51 4349 if (clustersize < sb->s_blocksize) {
4a8557b0
JY		 4350 ext4_msg(sb, KERN_ERR,
4351 "cluster size (%d) smaller than "
c8267c51 4352 "block size (%lu)", clustersize, sb->s_blocksize);
4a8557b0
JY		 4353 return -EINVAL;
4354 }
4355 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4356 le32_to_cpu(es->s_log_block_size);
4357 sbi->s_clusters_per_group =
4358 le32_to_cpu(es->s_clusters_per_group);
c8267c51 4359 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4a8557b0
JY		 4360 ext4_msg(sb, KERN_ERR,
4361 "#clusters per group too big: %lu",
4362 sbi->s_clusters_per_group);
4363 return -EINVAL;
4364 }
4365 if (sbi->s_blocks_per_group !=
c8267c51 4366 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4a8557b0
JY		 4367 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4368 "clusters per group (%lu) inconsistent",
4369 sbi->s_blocks_per_group,
4370 sbi->s_clusters_per_group);
4371 return -EINVAL;
4372 }
4373 } else {
c8267c51 4374 if (clustersize != sb->s_blocksize) {
4a8557b0
JY		 4375 ext4_msg(sb, KERN_ERR,
4376 "fragment/cluster size (%d) != "
c8267c51 4377 "block size (%lu)", clustersize, sb->s_blocksize);
4a8557b0
JY		 4378 return -EINVAL;
4379 }
c8267c51 4380 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4a8557b0
JY		 4381 ext4_msg(sb, KERN_ERR,
4382 "#blocks per group too big: %lu",
4383 sbi->s_blocks_per_group);
4384 return -EINVAL;
4385 }
4386 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4387 sbi->s_cluster_bits = 0;
4388 }
c8267c51 4389 sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4a8557b0
JY		 4390
4391 /* Do we have standard group size of clustersize * 8 blocks ? */
4392 if (sbi->s_blocks_per_group == clustersize << 3)
4393 set_opt2(sb, STD_GROUP_SIZE);
4394
4395 return 0;
4396}
4397
f7314a67
JY		 4398static void ext4_fast_commit_init(struct super_block *sb)
4399{
4400 struct ext4_sb_info *sbi = EXT4_SB(sb);
4401
4402 /* Initialize fast commit stuff */
4403 atomic_set(&sbi->s_fc_subtid, 0);
4404 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4405 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4406 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4407 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4408 sbi->s_fc_bytes = 0;
4409 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4410 sbi->s_fc_ineligible_tid = 0;
4411 spin_lock_init(&sbi->s_fc_lock);
4412 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4413 sbi->s_fc_replay_state.fc_regions = NULL;
4414 sbi->s_fc_replay_state.fc_regions_size = 0;
4415 sbi->s_fc_replay_state.fc_regions_used = 0;
4416 sbi->s_fc_replay_state.fc_regions_valid = 0;
4417 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4418 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4419 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4420}
4421
0e495f7c 4422static int ext4_inode_info_init(struct super_block *sb,
c8267c51 4423 struct ext4_super_block *es)
0e495f7c
JY		 4424{
4425 struct ext4_sb_info *sbi = EXT4_SB(sb);
4426
4427 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4428 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4429 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4430 } else {
4431 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4432 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4433 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4434 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4435 sbi->s_first_ino);
4436 return -EINVAL;
4437 }
4438 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4439 (!is_power_of_2(sbi->s_inode_size)) ||
c8267c51 4440 (sbi->s_inode_size > sb->s_blocksize)) {
0e495f7c
JY		 4441 ext4_msg(sb, KERN_ERR,
4442 "unsupported inode size: %d",
4443 sbi->s_inode_size);
c8267c51 4444 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
0e495f7c
JY		 4445 return -EINVAL;
4446 }
4447 /*
4448 * i_atime_extra is the last extra field available for
4449 * [acm]times in struct ext4_inode. Checking for that
4450 * field should suffice to ensure we have extra space
4451 * for all three.
4452 */
4453 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4454 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4455 sb->s_time_gran = 1;
4456 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4457 } else {
4458 sb->s_time_gran = NSEC_PER_SEC;
4459 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4460 }
4461 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4462 }
4463
4464 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4465 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4466 EXT4_GOOD_OLD_INODE_SIZE;
4467 if (ext4_has_feature_extra_isize(sb)) {
4468 unsigned v, max = (sbi->s_inode_size -
4469 EXT4_GOOD_OLD_INODE_SIZE);
4470
4471 v = le16_to_cpu(es->s_want_extra_isize);
4472 if (v > max) {
4473 ext4_msg(sb, KERN_ERR,
4474 "bad s_want_extra_isize: %d", v);
4475 return -EINVAL;
4476 }
4477 if (sbi->s_want_extra_isize < v)
4478 sbi->s_want_extra_isize = v;
4479
4480 v = le16_to_cpu(es->s_min_extra_isize);
4481 if (v > max) {
4482 ext4_msg(sb, KERN_ERR,
4483 "bad s_min_extra_isize: %d", v);
4484 return -EINVAL;
4485 }
4486 if (sbi->s_want_extra_isize < v)
4487 sbi->s_want_extra_isize = v;
4488 }
4489 }
4490
4491 return 0;
4492}
4493
39c135b0
JY		 4494#if IS_ENABLED(CONFIG_UNICODE)
4495static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4496{
4497 const struct ext4_sb_encodings *encoding_info;
4498 struct unicode_map *encoding;
4499 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4500
4501 if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4502 return 0;
4503
4504 encoding_info = ext4_sb_read_encoding(es);
4505 if (!encoding_info) {
4506 ext4_msg(sb, KERN_ERR,
4507 "Encoding requested by superblock is unknown");
4508 return -EINVAL;
4509 }
4510
4511 encoding = utf8_load(encoding_info->version);
4512 if (IS_ERR(encoding)) {
4513 ext4_msg(sb, KERN_ERR,
4514 "can't mount with superblock charset: %s-%u.%u.%u "
4515 "not supported by the kernel. flags: 0x%x.",
4516 encoding_info->name,
4517 unicode_major(encoding_info->version),
4518 unicode_minor(encoding_info->version),
4519 unicode_rev(encoding_info->version),
4520 encoding_flags);
4521 return -EINVAL;
4522 }
4523 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4524 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4525 unicode_major(encoding_info->version),
4526 unicode_minor(encoding_info->version),
4527 unicode_rev(encoding_info->version),
4528 encoding_flags);
4529
4530 sb->s_encoding = encoding;
4531 sb->s_encoding_flags = encoding_flags;
4532
4533 return 0;
4534}
4535#else
4536static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4537{
4538 return 0;
4539}
4540#endif
4541
b26458d1
JY		 4542static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4543{
4544 struct ext4_sb_info *sbi = EXT4_SB(sb);
4545
4546 /* Warn if metadata_csum and gdt_csum are both set. */
4547 if (ext4_has_feature_metadata_csum(sb) &&
4548 ext4_has_feature_gdt_csum(sb))
4549 ext4_warning(sb, "metadata_csum and uninit_bg are "
4550 "redundant flags; please run fsck.");
4551
4552 /* Check for a known checksum algorithm */
4553 if (!ext4_verify_csum_type(sb, es)) {
4554 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4555 "unknown checksum algorithm.");
4556 return -EINVAL;
4557 }
4558 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4559 ext4_orphan_file_block_trigger);
4560
4561 /* Load the checksum driver */
4562 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4563 if (IS_ERR(sbi->s_chksum_driver)) {
4564 int ret = PTR_ERR(sbi->s_chksum_driver);
4565 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4566 sbi->s_chksum_driver = NULL;
4567 return ret;
4568 }
4569
4570 /* Check superblock checksum */
4571 if (!ext4_superblock_csum_verify(sb, es)) {
4572 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4573 "invalid superblock checksum. Run e2fsck?");
4574 return -EFSBADCRC;
4575 }
4576
4577 /* Precompute checksum seed for all metadata */
4578 if (ext4_has_feature_csum_seed(sb))
4579 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4580 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4581 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4582 sizeof(es->s_uuid));
4583 return 0;
4584}
4585
d7f3542b
JY		 4586static int ext4_check_feature_compatibility(struct super_block *sb,
4587 struct ext4_super_block *es,
4588 int silent)
4589{
4590 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4591 (ext4_has_compat_features(sb) ||
4592 ext4_has_ro_compat_features(sb) ||
4593 ext4_has_incompat_features(sb)))
4594 ext4_msg(sb, KERN_WARNING,
4595 "feature flags set on rev 0 fs, "
4596 "running e2fsck is recommended");
4597
4598 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4599 set_opt2(sb, HURD_COMPAT);
4600 if (ext4_has_feature_64bit(sb)) {
4601 ext4_msg(sb, KERN_ERR,
4602 "The Hurd can't support 64-bit file systems");
4603 return -EINVAL;
4604 }
4605
4606 /*
4607 * ea_inode feature uses l_i_version field which is not
4608 * available in HURD_COMPAT mode.
4609 */
4610 if (ext4_has_feature_ea_inode(sb)) {
4611 ext4_msg(sb, KERN_ERR,
4612 "ea_inode feature is not supported for Hurd");
4613 return -EINVAL;
4614 }
4615 }
4616
4617 if (IS_EXT2_SB(sb)) {
4618 if (ext2_feature_set_ok(sb))
4619 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4620 "using the ext4 subsystem");
4621 else {
4622 /*
4623 * If we're probing be silent, if this looks like
4624 * it's actually an ext[34] filesystem.
4625 */
4626 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4627 return -EINVAL;
4628 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4629 "to feature incompatibilities");
4630 return -EINVAL;
4631 }
4632 }
4633
4634 if (IS_EXT3_SB(sb)) {
4635 if (ext3_feature_set_ok(sb))
4636 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4637 "using the ext4 subsystem");
4638 else {
4639 /*
4640 * If we're probing be silent, if this looks like
4641 * it's actually an ext4 filesystem.
4642 */
4643 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4644 return -EINVAL;
4645 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4646 "to feature incompatibilities");
4647 return -EINVAL;
4648 }
4649 }
4650
4651 /*
4652 * Check feature flags regardless of the revision level, since we
4653 * previously didn't change the revision level when setting the flags,
4654 * so there is a chance incompat flags are set on a rev 0 filesystem.
4655 */
4656 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4657 return -EINVAL;
4658
4659 return 0;
4660}
4661
bc62dbf9
JY		 4662static int ext4_geometry_check(struct super_block *sb,
4663 struct ext4_super_block *es)
4664{
4665 struct ext4_sb_info *sbi = EXT4_SB(sb);
4666 __u64 blocks_count;
4667
4668 /* check blocks count against device size */
4669 blocks_count = sb_bdev_nr_blocks(sb);
4670 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4671 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4672 "exceeds size of device (%llu blocks)",
4673 ext4_blocks_count(es), blocks_count);
4674 return -EINVAL;
4675 }
4676
4677 /*
4678 * It makes no sense for the first data block to be beyond the end
4679 * of the filesystem.
4680 */
4681 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4682 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4683 "block %u is beyond end of filesystem (%llu)",
4684 le32_to_cpu(es->s_first_data_block),
4685 ext4_blocks_count(es));
4686 return -EINVAL;
4687 }
4688 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4689 (sbi->s_cluster_ratio == 1)) {
4690 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4691 "block is 0 with a 1k block and cluster size");
4692 return -EINVAL;
4693 }
4694
4695 blocks_count = (ext4_blocks_count(es) -
4696 le32_to_cpu(es->s_first_data_block) +
4697 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4698 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4699 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4700 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4701 "(block count %llu, first data block %u, "
4702 "blocks per group %lu)", blocks_count,
4703 ext4_blocks_count(es),
4704 le32_to_cpu(es->s_first_data_block),
4705 EXT4_BLOCKS_PER_GROUP(sb));
4706 return -EINVAL;
4707 }
4708 sbi->s_groups_count = blocks_count;
4709 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4710 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4711 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4712 le32_to_cpu(es->s_inodes_count)) {
4713 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4714 le32_to_cpu(es->s_inodes_count),
4715 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4716 return -EINVAL;
4717 }
4718
4719 return 0;
4720}
4721
a4e6a511
JY		 4722static void ext4_group_desc_free(struct ext4_sb_info *sbi)
4723{
4724 struct buffer_head **group_desc;
4725 int i;
4726
4727 rcu_read_lock();
4728 group_desc = rcu_dereference(sbi->s_group_desc);
4729 for (i = 0; i < sbi->s_gdb_count; i++)
4730 brelse(group_desc[i]);
4731 kvfree(group_desc);
4732 rcu_read_unlock();
4733}
4734
4735static int ext4_group_desc_init(struct super_block *sb,
4736 struct ext4_super_block *es,
4737 ext4_fsblk_t logical_sb_block,
4738 ext4_group_t *first_not_zeroed)
4739{
4740 struct ext4_sb_info *sbi = EXT4_SB(sb);
4741 unsigned int db_count;
4742 ext4_fsblk_t block;
a4e6a511
JY		 4743 int i;
4744
4745 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4746 EXT4_DESC_PER_BLOCK(sb);
4747 if (ext4_has_feature_meta_bg(sb)) {
4748 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4749 ext4_msg(sb, KERN_WARNING,
4750 "first meta block group too large: %u "
4751 "(group descriptor block count %u)",
4752 le32_to_cpu(es->s_first_meta_bg), db_count);
4753 return -EINVAL;
4754 }
4755 }
4756 rcu_assign_pointer(sbi->s_group_desc,
4757 kvmalloc_array(db_count,
4758 sizeof(struct buffer_head *),
4759 GFP_KERNEL));
4760 if (sbi->s_group_desc == NULL) {
4761 ext4_msg(sb, KERN_ERR, "not enough memory");
4762 return -ENOMEM;
4763 }
4764
4765 bgl_lock_init(sbi->s_blockgroup_lock);
4766
4767 /* Pre-read the descriptors into the buffer cache */
4768 for (i = 0; i < db_count; i++) {
4769 block = descriptor_loc(sb, logical_sb_block, i);
4770 ext4_sb_breadahead_unmovable(sb, block);
4771 }
4772
4773 for (i = 0; i < db_count; i++) {
4774 struct buffer_head *bh;
4775
4776 block = descriptor_loc(sb, logical_sb_block, i);
4777 bh = ext4_sb_bread_unmovable(sb, block);
4778 if (IS_ERR(bh)) {
4779 ext4_msg(sb, KERN_ERR,
4780 "can't read group descriptor %d", i);
4781 sbi->s_gdb_count = i;
172e344e 4782 return PTR_ERR(bh);
a4e6a511
JY		 4783 }
4784 rcu_read_lock();
4785 rcu_dereference(sbi->s_group_desc)[i] = bh;
4786 rcu_read_unlock();
4787 }
4788 sbi->s_gdb_count = db_count;
4789 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) {
4790 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
172e344e 4791 return -EFSCORRUPTED;
a4e6a511 4792 }
172e344e 4793
a4e6a511 4794 return 0;
a4e6a511
JY		 4795}
4796
9c1dd22d
JY		 4797static int ext4_load_and_init_journal(struct super_block *sb,
4798 struct ext4_super_block *es,
4799 struct ext4_fs_context *ctx)
4800{
4801 struct ext4_sb_info *sbi = EXT4_SB(sb);
4802 int err;
4803
4804 err = ext4_load_journal(sb, es, ctx->journal_devnum);
4805 if (err)
4806 return err;
4807
4808 if (ext4_has_feature_64bit(sb) &&
4809 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4810 JBD2_FEATURE_INCOMPAT_64BIT)) {
4811 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4812 goto out;
4813 }
4814
4815 if (!set_journal_csum_feature_set(sb)) {
4816 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4817 "feature set");
4818 goto out;
4819 }
4820
4821 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4822 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4823 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4824 ext4_msg(sb, KERN_ERR,
4825 "Failed to set fast commit journal feature");
4826 goto out;
4827 }
4828
4829 /* We have now updated the journal if required, so we can
4830 * validate the data journaling mode. */
4831 switch (test_opt(sb, DATA_FLAGS)) {
4832 case 0:
4833 /* No mode set, assume a default based on the journal
4834 * capabilities: ORDERED_DATA if the journal can
4835 * cope, else JOURNAL_DATA
4836 */
4837 if (jbd2_journal_check_available_features
4838 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4839 set_opt(sb, ORDERED_DATA);
4840 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4841 } else {
4842 set_opt(sb, JOURNAL_DATA);
4843 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4844 }
4845 break;
4846
4847 case EXT4_MOUNT_ORDERED_DATA:
4848 case EXT4_MOUNT_WRITEBACK_DATA:
4849 if (!jbd2_journal_check_available_features
4850 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4851 ext4_msg(sb, KERN_ERR, "Journal does not support "
4852 "requested data journaling mode");
4853 goto out;
4854 }
4855 break;
4856 default:
4857 break;
4858 }
4859
4860 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4861 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4862 ext4_msg(sb, KERN_ERR, "can't mount with "
4863 "journal_async_commit in data=ordered mode");
4864 goto out;
4865 }
4866
4867 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
4868
4869 sbi->s_journal->j_submit_inode_data_buffers =
4870 ext4_journal_submit_inode_data_buffers;
4871 sbi->s_journal->j_finish_inode_data_buffers =
4872 ext4_journal_finish_inode_data_buffers;
4873
4874 return 0;
4875
4876out:
4877 /* flush s_error_work before journal destroy. */
4878 flush_work(&sbi->s_error_work);
4879 jbd2_journal_destroy(sbi->s_journal);
4880 sbi->s_journal = NULL;
9f2a1d9f 4881 return -EINVAL;
9c1dd22d
JY		 4882}
4883
a5991e53
JY		 4884static int ext4_journal_data_mode_check(struct super_block *sb)
4885{
4886 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4887 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4888 "data=journal disables delayed allocation, "
4889 "dioread_nolock, O_DIRECT and fast_commit support!\n");
4890 /* can't mount with both data=journal and dioread_nolock. */
4891 clear_opt(sb, DIOREAD_NOLOCK);
4892 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4893 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4894 ext4_msg(sb, KERN_ERR, "can't mount with "
4895 "both data=journal and delalloc");
4896 return -EINVAL;
4897 }
4898 if (test_opt(sb, DAX_ALWAYS)) {
4899 ext4_msg(sb, KERN_ERR, "can't mount with "
4900 "both data=journal and dax");
4901 return -EINVAL;
4902 }
4903 if (ext4_has_feature_encrypt(sb)) {
4904 ext4_msg(sb, KERN_WARNING,
4905 "encrypted files will use data=ordered "
4906 "instead of data journaling mode");
4907 }
4908 if (test_opt(sb, DELALLOC))
4909 clear_opt(sb, DELALLOC);
4910 } else {
4911 sb->s_iflags |= SB_I_CGROUPWB;
4912 }
4913
4914 return 0;
4915}
4916
a7a79c29
JY		 4917static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
4918 int silent)
ac27a0ec 4919{
7edfd85b 4920 struct ext4_sb_info *sbi = EXT4_SB(sb);
a7a79c29 4921 struct ext4_super_block *es;
70bbb3e0 4922 ext4_fsblk_t logical_sb_block;
ac27a0ec 4923 unsigned long offset = 0;
a7a79c29
JY		 4924 struct buffer_head *bh;
4925 int ret = -EINVAL;
4a8557b0 4926 int blocksize;
705895b6 4927
617ba13b 4928 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 4929 if (!blocksize) {
b31e1552 4930 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
a7a79c29 4931 return -EINVAL;
ac27a0ec
DK		 4932 }
4933
4934 /*
617ba13b 4935 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 4936 * block sizes. We need to calculate the offset from buffer start.
4937 */
617ba13b 4938 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
7edfd85b 4939 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
70bbb3e0 4940 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 4941 } else {
7edfd85b 4942 logical_sb_block = sbi->s_sb_block;
ac27a0ec
DK		 4943 }
4944
8394a6ab 4945 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4946 if (IS_ERR(bh)) {
b31e1552 4947 ext4_msg(sb, KERN_ERR, "unable to read superblock");
a7a79c29 4948 return PTR_ERR(bh);
ac27a0ec
DK		 4949 }
4950 /*
4951 * Note: s_es must be initialized as soon as possible because
617ba13b 4952 * some ext4 macro-instructions depend on its value
ac27a0ec 4953 */
2716b802 4954 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 4955 sbi->s_es = es;
4956 sb->s_magic = le16_to_cpu(es->s_magic);
a5fc5119
JY		 4957 if (sb->s_magic != EXT4_SUPER_MAGIC) {
4958 if (!silent)
4959 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
a7a79c29 4960 goto out;
a5fc5119
JY		 4961 }
4962
a7a79c29
JY		 4963 if (le32_to_cpu(es->s_log_block_size) >
4964 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4965 ext4_msg(sb, KERN_ERR,
4966 "Invalid log block size: %u",
4967 le32_to_cpu(es->s_log_block_size));
4968 goto out;
4969 }
4970 if (le32_to_cpu(es->s_log_cluster_size) >
4971 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4972 ext4_msg(sb, KERN_ERR,
4973 "Invalid log cluster size: %u",
4974 le32_to_cpu(es->s_log_cluster_size));
4975 goto out;
4976 }
4977
4978 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4979
4980 /*
4981 * If the default block size is not the same as the real block size,
4982 * we need to reload it.
4983 */
4984 if (sb->s_blocksize == blocksize) {
4985 *lsb = logical_sb_block;
4986 sbi->s_sbh = bh;
4987 return 0;
4988 }
4989
4990 /*
4991 * bh must be released before kill_bdev(), otherwise
4992 * it won't be freed and its page also. kill_bdev()
4993 * is called by sb_set_blocksize().
4994 */
4995 brelse(bh);
4996 /* Validate the filesystem blocksize */
4997 if (!sb_set_blocksize(sb, blocksize)) {
4998 ext4_msg(sb, KERN_ERR, "bad block size %d",
4999 blocksize);
5000 bh = NULL;
5001 goto out;
5002 }
5003
5004 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5005 offset = do_div(logical_sb_block, blocksize);
5006 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5007 if (IS_ERR(bh)) {
5008 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5009 ret = PTR_ERR(bh);
5010 bh = NULL;
5011 goto out;
5012 }
5013 es = (struct ext4_super_block *)(bh->b_data + offset);
5014 sbi->s_es = es;
5015 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5016 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5017 goto out;
5018 }
5019 *lsb = logical_sb_block;
5020 sbi->s_sbh = bh;
5021 return 0;
5022out:
5023 brelse(bh);
5024 return ret;
5025}
5026
5027static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5028{
5029 struct ext4_super_block *es = NULL;
5030 struct ext4_sb_info *sbi = EXT4_SB(sb);
5031 struct flex_groups **flex_groups;
5032 ext4_fsblk_t block;
5033 ext4_fsblk_t logical_sb_block;
5034 struct inode *root;
5035 int ret = -ENOMEM;
a7a79c29
JY		 5036 unsigned int i;
5037 int needs_recovery, has_huge_files;
5038 int err = 0;
5039 ext4_group_t first_not_zeroed;
5040 struct ext4_fs_context *ctx = fc->fs_private;
5041 int silent = fc->sb_flags & SB_SILENT;
5042
5043 /* Set defaults for the variables that will be set during parsing */
5044 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5045 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5046
5047 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5048 sbi->s_sectors_written_start =
5049 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5050
5051 /* -EINVAL is default */
5052 ret = -EINVAL;
5053 err = ext4_load_super(sb, &logical_sb_block, silent);
5054 if (err)
5055 goto out_fail;
5056
5057 es = sbi->s_es;
afc32f7e 5058 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 5059
b26458d1
JY		 5060 err = ext4_init_metadata_csum(sb, es);
5061 if (err)
a5fc5119 5062 goto failed_mount;
a9c47317 5063
5f6d662d 5064 ext4_set_def_opts(sb, es);
ac27a0ec 5065
08cefc7a
EB		 5066 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
5067 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 5068 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5069 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5070 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 5071
51ce6511
LC		 5072 /*
5073 * set default s_li_wait_mult for lazyinit, for the case there is
5074 * no mount option specified.
5075 */
5076 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5077
c8267c51 5078 if (ext4_inode_info_init(sb, es))
0e495f7c 5079 goto failed_mount;
9803387c 5080
7edfd85b
LC		 5081 err = parse_apply_sb_mount_options(sb, ctx);
5082 if (err < 0)
5083 goto failed_mount;
5084
5a916be1 5085 sbi->s_def_mount_opt = sbi->s_mount_opt;
e3645d72 5086 sbi->s_def_mount_opt2 = sbi->s_mount_opt2;
7edfd85b
LC		 5087
5088 err = ext4_check_opt_consistency(fc, sb);
5089 if (err < 0)
5090 goto failed_mount;
5091
85456054 5092 ext4_apply_options(fc, sb);
ac27a0ec 5093
39c135b0
JY		 5094 if (ext4_encoding_init(sb, es))
5095 goto failed_mount;
c83ad55e 5096
a5991e53
JY		 5097 if (ext4_journal_data_mode_check(sb))
5098 goto failed_mount;
56889787 5099
1751e8a6
LT		 5100 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5101 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 5102
1ff20307
JL		 5103 /* i_version is always enabled now */
5104 sb->s_flags |= SB_I_VERSION;
5105
d7f3542b 5106 if (ext4_check_feature_compatibility(sb, es, silent))
ac27a0ec 5107 goto failed_mount;
a13fb1a4 5108
c8267c51 5109 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
5b9554dc
TT		 5110 ext4_msg(sb, KERN_ERR,
5111 "Number of reserved GDT blocks insanely large: %d",
5112 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
5113 goto failed_mount;
5114 }
5115
89b93a7b 5116 if (sbi->s_daxdev) {
c8267c51 5117 if (sb->s_blocksize == PAGE_SIZE)
7b0800d0
CH		 5118 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
5119 else
5120 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
5121 }
a8ab6d38 5122
fc626fe3 5123 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
559db4c6
RZ		 5124 if (ext4_has_feature_inline_data(sb)) {
5125 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
5126 " that may contain inline data");
361d24d4 5127 goto failed_mount;
559db4c6 5128 }
a8ab6d38 5129 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
24f3478d 5130 ext4_msg(sb, KERN_ERR,
361d24d4
ES		 5131 "DAX unsupported by block device.");
5132 goto failed_mount;
24f3478d 5133 }
923ae0ff
RZ		 5134 }
5135
e2b911c5 5136 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
6ddb2447
TT		 5137 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
5138 es->s_encryption_level);
5139 goto failed_mount;
5140 }
5141
e2b911c5 5142 has_huge_files = ext4_has_feature_huge_file(sb);
f287a1a5
TT		 5143 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
5144 has_huge_files);
5145 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 5146
0d1ee42f 5147 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
e2b911c5 5148 if (ext4_has_feature_64bit(sb)) {
8fadc143 5149 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 5150 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 5151 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 5152 ext4_msg(sb, KERN_ERR,
5153 "unsupported descriptor size %lu",
0d1ee42f
AR		 5154 sbi->s_desc_size);
5155 goto failed_mount;
5156 }
5157 } else
5158 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 5159
ac27a0ec 5160 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 5161 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
0b8e58a1 5162
c8267c51 5163 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
a5fc5119
JY		 5164 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5165 if (!silent)
5166 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5167 goto failed_mount;
5168 }
cd6bb35b 5169 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
c8267c51 5170 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
cd6bb35b 5171 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
b9c538da 5172 sbi->s_inodes_per_group);
cd6bb35b
TT		 5173 goto failed_mount;
5174 }
ac27a0ec
DK		 5175 sbi->s_itb_per_group = sbi->s_inodes_per_group /
5176 sbi->s_inodes_per_block;
c8267c51 5177 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
c878bea3 5178 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
e57aa839
FW		 5179 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5180 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 5181
2b2d6d01 5182 for (i = 0; i < 4; i++)
ac27a0ec
DK		 5183 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5184 sbi->s_def_hash_version = es->s_def_hash_version;
e2b911c5 5185 if (ext4_has_feature_dir_index(sb)) {
23301410
TT		 5186 i = le32_to_cpu(es->s_flags);
5187 if (i & EXT2_FLAGS_UNSIGNED_HASH)
5188 sbi->s_hash_unsigned = 3;
5189 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 5190#ifdef __CHAR_UNSIGNED__
bc98a42c 5191 if (!sb_rdonly(sb))
23301410
TT		 5192 es->s_flags |=
5193 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5194 sbi->s_hash_unsigned = 3;
f99b2589 5195#else
bc98a42c 5196 if (!sb_rdonly(sb))
23301410
TT		 5197 es->s_flags |=
5198 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 5199#endif
23301410 5200 }
f99b2589 5201 }
ac27a0ec 5202
c8267c51 5203 if (ext4_handle_clustersize(sb))
4a8557b0 5204 goto failed_mount;
960fd856 5205
bf43d84b
ES		 5206 /*
5207 * Test whether we have more sectors than will fit in sector_t,
5208 * and whether the max offset is addressable by the page cache.
5209 */
5a9ae68a 5210 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 5211 ext4_blocks_count(es));
5a9ae68a 5212 if (err) {
b31e1552 5213 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 5214 " too large to mount safely on this system");
ac27a0ec
DK		 5215 goto failed_mount;
5216 }
5217
bc62dbf9 5218 if (ext4_geometry_check(sb, es))
0f2ddca6 5219 goto failed_mount;
0f2ddca6 5220
235699a8 5221 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
c92dc856
JK		 5222 spin_lock_init(&sbi->s_error_lock);
5223 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
04496411 5224
172e344e
YB		 5225 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed);
5226 if (err)
5227 goto failed_mount3;
5228
a75ae78f 5229 /* Register extent status tree shrinker */
eb68d0e2 5230 if (ext4_es_register_shrinker(sbi))
ce7e010a 5231 goto failed_mount3;
ce7e010a 5232
c9de560d 5233 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 5234 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 5235
f9ae9cf5
TT		 5236 /*
5237 * set up enough so that it can read an inode
5238 */
f6e63f90 5239 sb->s_op = &ext4_sops;
617ba13b
MC		 5240 sb->s_export_op = &ext4_export_ops;
5241 sb->s_xattr = ext4_xattr_handlers;
643fa961 5242#ifdef CONFIG_FS_ENCRYPTION
a7550b30 5243 sb->s_cop = &ext4_cryptops;
ffcc4182 5244#endif
c93d8f88
EB		 5245#ifdef CONFIG_FS_VERITY
5246 sb->s_vop = &ext4_verityops;
5247#endif
ac27a0ec 5248#ifdef CONFIG_QUOTA
617ba13b 5249 sb->dq_op = &ext4_quota_operations;
e2b911c5 5250 if (ext4_has_feature_quota(sb))
1fa5efe3 5251 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 5252 else
5253 sb->s_qcop = &ext4_qctl_operations;
689c958c 5254 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 5255#endif
85787090 5256 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 5257
ac27a0ec 5258 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 5259 mutex_init(&sbi->s_orphan_lock);
ac27a0ec 5260
f7314a67 5261 ext4_fast_commit_init(sb);
aa75f4d3 5262
ac27a0ec
DK		 5263 sb->s_root = NULL;
5264
5265 needs_recovery = (es->s_last_orphan != 0 ||
02f310fc 5266 ext4_has_feature_orphan_present(sb) ||
e2b911c5 5267 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 5268
bc98a42c 5269 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
c5e06d10 5270 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
50460fe8 5271 goto failed_mount3a;
c5e06d10 5272
ac27a0ec
DK		 5273 /*
5274 * The first inode we look at is the journal inode. Don't try
5275 * root first: it may be modified in the journal!
5276 */
e2b911c5 5277 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
9c1dd22d 5278 err = ext4_load_and_init_journal(sb, es, ctx);
4753d8a2 5279 if (err)
50460fe8 5280 goto failed_mount3a;
bc98a42c 5281 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 5282 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 5283 ext4_msg(sb, KERN_ERR, "required journal recovery "
5284 "suppressed and not mounted read-only");
43bd6f1b 5285 goto failed_mount3a;
ac27a0ec 5286 } else {
1e381f60 5287 /* Nojournal mode, all journal mount options are illegal */
89481b5f 5288 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1e381f60 5289 ext4_msg(sb, KERN_ERR, "can't mount with "
89481b5f 5290 "journal_async_commit, fs mounted w/o journal");
43bd6f1b 5291 goto failed_mount3a;
1e381f60 5292 }
89481b5f
BL		 5293
5294 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
1e381f60 5295 ext4_msg(sb, KERN_ERR, "can't mount with "
89481b5f 5296 "journal_checksum, fs mounted w/o journal");
43bd6f1b 5297 goto failed_mount3a;
1e381f60
DM		 5298 }
5299 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5300 ext4_msg(sb, KERN_ERR, "can't mount with "
5301 "commit=%lu, fs mounted w/o journal",
5302 sbi->s_commit_interval / HZ);
43bd6f1b 5303 goto failed_mount3a;
1e381f60
DM		 5304 }
5305 if (EXT4_MOUNT_DATA_FLAGS &
5306 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5307 ext4_msg(sb, KERN_ERR, "can't mount with "
5308 "data=, fs mounted w/o journal");
43bd6f1b 5309 goto failed_mount3a;
1e381f60 5310 }
50b29d8f 5311 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
1e381f60 5312 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 5313 clear_opt(sb, DATA_FLAGS);
995a3ed6 5314 clear_opt2(sb, JOURNAL_FAST_COMMIT);
0390131b
FM		 5315 sbi->s_journal = NULL;
5316 needs_recovery = 0;
ac27a0ec
DK		 5317 }
5318
cdb7ee4c
TE		 5319 if (!test_opt(sb, NO_MBCACHE)) {
5320 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5321 if (!sbi->s_ea_block_cache) {
dec214d0 5322 ext4_msg(sb, KERN_ERR,
cdb7ee4c 5323 "Failed to create ea_block_cache");
dec214d0
TE		 5324 goto failed_mount_wq;
5325 }
cdb7ee4c
TE		 5326
5327 if (ext4_has_feature_ea_inode(sb)) {
5328 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5329 if (!sbi->s_ea_inode_cache) {
5330 ext4_msg(sb, KERN_ERR,
5331 "Failed to create ea_inode_cache");
5332 goto failed_mount_wq;
5333 }
5334 }
9c191f70
M		 5335 }
5336
952fc18e
TT		 5337 /*
5338 * Get the # of file system overhead blocks from the
5339 * superblock if present.
5340 */
85d825db
TT		 5341 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5342 /* ignore the precalculated value if it is ridiculous */
5343 if (sbi->s_overhead > ext4_blocks_count(es))
5344 sbi->s_overhead = 0;
5345 /*
5346 * If the bigalloc feature is not enabled recalculating the
5347 * overhead doesn't take long, so we might as well just redo
5348 * it to make sure we are using the correct value.
5349 */
5350 if (!ext4_has_feature_bigalloc(sb))
5351 sbi->s_overhead = 0;
5352 if (sbi->s_overhead == 0) {
07aa2ea1
LC		 5353 err = ext4_calculate_overhead(sb);
5354 if (err)
952fc18e
TT		 5355 goto failed_mount_wq;
5356 }
5357
fd89d5f2
TH		 5358 /*
5359 * The maximum number of concurrent works can be high and
5360 * concurrency isn't really necessary. Limit it to 1.
5361 */
2e8fa54e
JK		 5362 EXT4_SB(sb)->rsv_conversion_wq =
5363 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5364 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5365 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 5366 ret = -ENOMEM;
2e8fa54e
JK		 5367 goto failed_mount4;
5368 }
5369
ac27a0ec 5370 /*
dab291af 5371 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 5372 * so we can safely mount the rest of the filesystem now.
5373 */
5374
8a363970 5375 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
1d1fe1ee 5376 if (IS_ERR(root)) {
b31e1552 5377 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 5378 ret = PTR_ERR(root);
32a9bb57 5379 root = NULL;
ac27a0ec
DK		 5380 goto failed_mount4;
5381 }
5382 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 5383 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 5384 iput(root);
ac27a0ec
DK		 5385 goto failed_mount4;
5386 }
b886ee3e 5387
48fde701 5388 sb->s_root = d_make_root(root);
1d1fe1ee 5389 if (!sb->s_root) {
b31e1552 5390 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 5391 ret = -ENOMEM;
5392 goto failed_mount4;
5393 }
ac27a0ec 5394
c89128a0
JK		 5395 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
5396 if (ret == -EROFS) {
1751e8a6 5397 sb->s_flags |= SB_RDONLY;
c89128a0
JK		 5398 ret = 0;
5399 } else if (ret)
5400 goto failed_mount4a;
ef7f3835 5401
b5799018 5402 ext4_set_resv_clusters(sb);
27dd4385 5403
0f5bde1d
JK		 5404 if (test_opt(sb, BLOCK_VALIDITY)) {
5405 err = ext4_setup_system_zone(sb);
5406 if (err) {
5407 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5408 "zone (%d)", err);
5409 goto failed_mount4a;
5410 }
f9ae9cf5 5411 }
8016e29f 5412 ext4_fc_replay_cleanup(sb);
f9ae9cf5
TT		 5413
5414 ext4_ext_init(sb);
196e402a
HS		 5415
5416 /*
5417 * Enable optimize_scan if number of groups is > threshold. This can be
5418 * turned off by passing "mb_optimize_scan=0". This can also be
5419 * turned on forcefully by passing "mb_optimize_scan=1".
5420 */
27b38686
OM		 5421 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5422 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5423 set_opt2(sb, MB_OPTIMIZE_SCAN);
5424 else
5425 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5426 }
196e402a 5427
f9ae9cf5
TT		 5428 err = ext4_mb_init(sb);
5429 if (err) {
5430 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5431 err);
dcf2d804 5432 goto failed_mount5;
c2774d84
AK		 5433 }
5434
027f14f5
TT		 5435 /*
5436 * We can only set up the journal commit callback once
5437 * mballoc is initialized
5438 */
5439 if (sbi->s_journal)
5440 sbi->s_journal->j_commit_callback =
5441 ext4_journal_commit_callback;
5442
d5e03cbb 5443 block = ext4_count_free_clusters(sb);
666245d9 5444 ext4_free_blocks_count_set(sbi->s_es,
d5e03cbb 5445 EXT4_C2B(sbi, block));
908c7f19
TH		 5446 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
5447 GFP_KERNEL);
d5e03cbb
TT		 5448 if (!err) {
5449 unsigned long freei = ext4_count_free_inodes(sb);
5450 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
908c7f19
TH		 5451 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
5452 GFP_KERNEL);
d5e03cbb
TT		 5453 }
5454 if (!err)
5455 err = percpu_counter_init(&sbi->s_dirs_counter,
908c7f19 5456 ext4_count_dirs(sb), GFP_KERNEL);
d5e03cbb 5457 if (!err)
908c7f19
TH		 5458 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
5459 GFP_KERNEL);
efc61345
EW		 5460 if (!err)
5461 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
5462 GFP_KERNEL);
c8585c6f 5463 if (!err)
bbd55937 5464 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
c8585c6f 5465
d5e03cbb
TT		 5466 if (err) {
5467 ext4_msg(sb, KERN_ERR, "insufficient memory");
5468 goto failed_mount6;
5469 }
5470
e2b911c5 5471 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 5472 if (!ext4_fill_flex_info(sb)) {
5473 ext4_msg(sb, KERN_ERR,
5474 "unable to initialize "
5475 "flex_bg meta info!");
8f6840c4 5476 ret = -ENOMEM;
d5e03cbb
TT		 5477 goto failed_mount6;
5478 }
5479
bfff6873
LC		 5480 err = ext4_register_li_request(sb, first_not_zeroed);
5481 if (err)
dcf2d804 5482 goto failed_mount6;
bfff6873 5483
b5799018 5484 err = ext4_register_sysfs(sb);
dcf2d804
TM		 5485 if (err)
5486 goto failed_mount7;
3197ebdb 5487
02f310fc
JK		 5488 err = ext4_init_orphan_info(sb);
5489 if (err)
5490 goto failed_mount8;
9b2ff357
JK		 5491#ifdef CONFIG_QUOTA
5492 /* Enable quota usage during mount. */
bc98a42c 5493 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 5494 err = ext4_enable_quotas(sb);
5495 if (err)
02f310fc 5496 goto failed_mount9;
9b2ff357
JK		 5497 }
5498#endif /* CONFIG_QUOTA */
5499
bc71726c 5500 /*
5501 * Save the original bdev mapping's wb_err value which could be
5502 * used to detect the metadata async write error.
5503 */
5504 spin_lock_init(&sbi->s_bdev_wb_lock);
9704a322
ZX		 5505 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5506 &sbi->s_bdev_wb_err);
bc71726c 5507 sb->s_bdev->bd_super = sb;
617ba13b
MC		 5508 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5509 ext4_orphan_cleanup(sb, es);
5510 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
9b6641dd
YB		 5511 /*
5512 * Update the checksum after updating free space/inode counters and
5513 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5514 * checksum in the buffer cache until it is written out and
5515 * e2fsprogs programs trying to open a file system immediately
5516 * after it is mounted can fail.
5517 */
5518 ext4_superblock_csum_set(sb);
0390131b 5519 if (needs_recovery) {
b31e1552 5520 ext4_msg(sb, KERN_INFO, "recovery complete");
11215630
JK		 5521 err = ext4_mark_recovery_complete(sb, es);
5522 if (err)
02f310fc 5523 goto failed_mount9;
0390131b 5524 }
0390131b 5525
70200574
CH		 5526 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5527 ext4_msg(sb, KERN_WARNING,
5528 "mounting with \"discard\" option, but the device does not support discard");
79add3a3 5529
66e61a9e
TT		 5530 if (es->s_error_count)
5531 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 5532
efbed4dc
TT		 5533 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5534 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5535 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5536 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
1cf006ed
DM		 5537 atomic_set(&sbi->s_warning_count, 0);
5538 atomic_set(&sbi->s_msg_count, 0);
efbed4dc 5539
ac27a0ec
DK		 5540 return 0;
5541
02f310fc
JK		 5542failed_mount9:
5543 ext4_release_orphan_info(sb);
72ba7450 5544failed_mount8:
ebd173be 5545 ext4_unregister_sysfs(sb);
cb8d53d2 5546 kobject_put(&sbi->s_kobj);
dcf2d804
TM		 5547failed_mount7:
5548 ext4_unregister_li_request(sb);
5549failed_mount6:
f9ae9cf5 5550 ext4_mb_release(sb);
7c990728
SJS		 5551 rcu_read_lock();
5552 flex_groups = rcu_dereference(sbi->s_flex_groups);
5553 if (flex_groups) {
5554 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5555 kvfree(flex_groups[i]);
5556 kvfree(flex_groups);
5557 }
5558 rcu_read_unlock();
d5e03cbb
TT		 5559 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5560 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5561 percpu_counter_destroy(&sbi->s_dirs_counter);
5562 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
efc61345 5563 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
bbd55937 5564 percpu_free_rwsem(&sbi->s_writepages_rwsem);
00764937 5565failed_mount5:
f9ae9cf5
TT		 5566 ext4_ext_release(sb);
5567 ext4_release_system_zone(sb);
5568failed_mount4a:
94bf608a 5569 dput(sb->s_root);
32a9bb57 5570 sb->s_root = NULL;
94bf608a 5571failed_mount4:
b31e1552 5572 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 5573 if (EXT4_SB(sb)->rsv_conversion_wq)
5574 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 5575failed_mount_wq:
50c15df6
CX		 5576 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5577 sbi->s_ea_inode_cache = NULL;
5578
5579 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5580 sbi->s_ea_block_cache = NULL;
5581
0390131b 5582 if (sbi->s_journal) {
bb9464e0 5583 /* flush s_error_work before journal destroy. */
5584 flush_work(&sbi->s_error_work);
0390131b
FM		 5585 jbd2_journal_destroy(sbi->s_journal);
5586 sbi->s_journal = NULL;
5587 }
50460fe8 5588failed_mount3a:
d3922a77 5589 ext4_es_unregister_shrinker(sbi);
eb68d0e2 5590failed_mount3:
bb9464e0 5591 /* flush s_error_work before sbi destroy */
c92dc856 5592 flush_work(&sbi->s_error_work);
2a4ae3bc 5593 del_timer_sync(&sbi->s_err_report);
618f0031 5594 ext4_stop_mmpd(sbi);
a4e6a511 5595 ext4_group_desc_free(sbi);
ac27a0ec 5596failed_mount:
0441984a
DW		 5597 if (sbi->s_chksum_driver)
5598 crypto_free_shash(sbi->s_chksum_driver);
c83ad55e 5599
5298d4bf 5600#if IS_ENABLED(CONFIG_UNICODE)
f8f4acb6 5601 utf8_unload(sb->s_encoding);
c83ad55e
GKB		 5602#endif
5603
ac27a0ec 5604#ifdef CONFIG_QUOTA
a2d4a646 5605 for (i = 0; i < EXT4_MAXQUOTAS; i++)
0ba33fac 5606 kfree(get_qf_name(sb, sbi, i));
ac27a0ec 5607#endif
ac4acb1f 5608 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
afd09b61 5609 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
a7a79c29 5610 brelse(sbi->s_sbh);
afd09b61 5611 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 5612out_fail:
5613 sb->s_fs_info = NULL;
07aa2ea1 5614 return err ? err : ret;
ac27a0ec
DK		 5615}
5616
cebe85d5 5617static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
7edfd85b 5618{
cebe85d5 5619 struct ext4_fs_context *ctx = fc->fs_private;
7edfd85b 5620 struct ext4_sb_info *sbi;
7edfd85b 5621 const char *descr;
cebe85d5 5622 int ret;
7edfd85b
LC		 5623
5624 sbi = ext4_alloc_sbi(sb);
cebe85d5 5625 if (!sbi)
7c268d4c 5626 return -ENOMEM;
7edfd85b 5627
cebe85d5
LC		 5628 fc->s_fs_info = sbi;
5629
5630 /* Cleanup superblock name */
5631 strreplace(sb->s_id, '/', '!');
7edfd85b
LC		 5632
5633 sbi->s_sb_block = 1; /* Default super block location */
cebe85d5
LC		 5634 if (ctx->spec & EXT4_SPEC_s_sb_block)
5635 sbi->s_sb_block = ctx->s_sb_block;
7edfd85b 5636
960e0ab6 5637 ret = __ext4_fill_super(fc, sb);
7edfd85b
LC		 5638 if (ret < 0)
5639 goto free_sbi;
5640
cebe85d5 5641 if (sbi->s_journal) {
7edfd85b
LC		 5642 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5643 descr = " journalled data mode";
5644 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5645 descr = " ordered data mode";
5646 else
5647 descr = " writeback data mode";
5648 } else
5649 descr = "out journal";
5650
5651 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
bb0fbc78
LC		 5652 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU with%s. "
5653 "Quota mode: %s.", &sb->s_uuid, descr,
5654 ext4_quota_mode(sb));
cebe85d5 5655
eb705421 5656 /* Update the s_overhead_clusters if necessary */
827891a3 5657 ext4_update_overhead(sb, false);
7edfd85b 5658 return 0;
cebe85d5 5659
7edfd85b
LC		 5660free_sbi:
5661 ext4_free_sbi(sbi);
cebe85d5 5662 fc->s_fs_info = NULL;
7edfd85b
LC		 5663 return ret;
5664}
5665
cebe85d5
LC		 5666static int ext4_get_tree(struct fs_context *fc)
5667{
5668 return get_tree_bdev(fc, ext4_fill_super);
5669}
5670
ac27a0ec
DK		 5671/*
5672 * Setup any per-fs journal parameters now. We'll do this both on
5673 * initial mount, once the journal has been initialised but before we've
5674 * done any recovery; and again on any subsequent remount.
5675 */
617ba13b 5676static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 5677{
617ba13b 5678 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5679
30773840
TT		 5680 journal->j_commit_interval = sbi->s_commit_interval;
5681 journal->j_min_batch_time = sbi->s_min_batch_time;
5682 journal->j_max_batch_time = sbi->s_max_batch_time;
6866d7b3 5683 ext4_fc_init(sb, journal);
ac27a0ec 5684
a931da6a 5685 write_lock(&journal->j_state_lock);
ac27a0ec 5686 if (test_opt(sb, BARRIER))
dab291af 5687 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 5688 else
dab291af 5689 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 5690 if (test_opt(sb, DATA_ERR_ABORT))
5691 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5692 else
5693 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 5694 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 5695}
5696
c6cb7e77
EW		 5697static struct inode *ext4_get_journal_inode(struct super_block *sb,
5698 unsigned int journal_inum)
ac27a0ec
DK		 5699{
5700 struct inode *journal_inode;
ac27a0ec 5701
c6cb7e77
EW		 5702 /*
5703 * Test for the existence of a valid inode on disk. Bad things
5704 * happen if we iget() an unused inode, as the subsequent iput()
5705 * will try to delete it.
5706 */
8a363970 5707 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
1d1fe1ee 5708 if (IS_ERR(journal_inode)) {
b31e1552 5709 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 5710 return NULL;
5711 }
5712 if (!journal_inode->i_nlink) {
5713 make_bad_inode(journal_inode);
5714 iput(journal_inode);
b31e1552 5715 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 5716 return NULL;
5717 }
5718
4978c659 5719 ext4_debug("Journal inode found at %p: %lld bytes\n",
ac27a0ec 5720 journal_inode, journal_inode->i_size);
105c78e1 5721 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
b31e1552 5722 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 5723 iput(journal_inode);
5724 return NULL;
5725 }
c6cb7e77
EW		 5726 return journal_inode;
5727}
5728
62913ae9
TT		 5729static int ext4_journal_bmap(journal_t *journal, sector_t *block)
5730{
5731 struct ext4_map_blocks map;
5732 int ret;
5733
5734 if (journal->j_inode == NULL)
5735 return 0;
5736
5737 map.m_lblk = *block;
5738 map.m_len = 1;
5739 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0);
5740 if (ret <= 0) {
5741 ext4_msg(journal->j_inode->i_sb, KERN_CRIT,
5742 "journal bmap failed: block %llu ret %d\n",
5743 *block, ret);
5744 jbd2_journal_abort(journal, ret ? ret : -EIO);
5745 return ret;
5746 }
5747 *block = map.m_pblk;
5748 return 0;
5749}
5750
c6cb7e77
EW		 5751static journal_t *ext4_get_journal(struct super_block *sb,
5752 unsigned int journal_inum)
5753{
5754 struct inode *journal_inode;
5755 journal_t *journal;
5756
11215630
JK		 5757 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5758 return NULL;
c6cb7e77
EW		 5759
5760 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5761 if (!journal_inode)
5762 return NULL;
ac27a0ec 5763
dab291af 5764 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 5765 if (!journal) {
b31e1552 5766 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 5767 iput(journal_inode);
5768 return NULL;
5769 }
5770 journal->j_private = sb;
62913ae9 5771 journal->j_bmap = ext4_journal_bmap;
617ba13b 5772 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 5773 return journal;
5774}
5775
617ba13b 5776static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 5777 dev_t j_dev)
5778{
2b2d6d01 5779 struct buffer_head *bh;
ac27a0ec 5780 journal_t *journal;
617ba13b
MC		 5781 ext4_fsblk_t start;
5782 ext4_fsblk_t len;
ac27a0ec 5783 int hblock, blocksize;
617ba13b 5784 ext4_fsblk_t sb_block;
ac27a0ec 5785 unsigned long offset;
2b2d6d01 5786 struct ext4_super_block *es;
ac27a0ec
DK		 5787 struct block_device *bdev;
5788
11215630
JK		 5789 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5790 return NULL;
0390131b 5791
b31e1552 5792 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 5793 if (bdev == NULL)
5794 return NULL;
5795
ac27a0ec 5796 blocksize = sb->s_blocksize;
e1defc4f 5797 hblock = bdev_logical_block_size(bdev);
ac27a0ec 5798 if (blocksize < hblock) {
b31e1552
ES		 5799 ext4_msg(sb, KERN_ERR,
5800 "blocksize too small for journal device");
ac27a0ec
DK		 5801 goto out_bdev;
5802 }
5803
617ba13b
MC		 5804 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5805 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 5806 set_blocksize(bdev, blocksize);
5807 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 5808 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5809 "external journal");
ac27a0ec
DK		 5810 goto out_bdev;
5811 }
5812
2716b802 5813 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 5814 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 5815 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 5816 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 5817 ext4_msg(sb, KERN_ERR, "external journal has "
5818 "bad superblock");
ac27a0ec
DK		 5819 brelse(bh);
5820 goto out_bdev;
5821 }
5822
df4763be
DW		 5823 if ((le32_to_cpu(es->s_feature_ro_compat) &
5824 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5825 es->s_checksum != ext4_superblock_csum(sb, es)) {
5826 ext4_msg(sb, KERN_ERR, "external journal has "
5827 "corrupt superblock");
5828 brelse(bh);
5829 goto out_bdev;
5830 }
5831
617ba13b 5832 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 5833 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 5834 brelse(bh);
5835 goto out_bdev;
5836 }
5837
bd81d8ee 5838 len = ext4_blocks_count(es);
ac27a0ec
DK		 5839 start = sb_block + 1;
5840 brelse(bh); /* we're done with the superblock */
5841
dab291af 5842 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 5843 start, len, blocksize);
5844 if (!journal) {
b31e1552 5845 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 5846 goto out_bdev;
5847 }
5848 journal->j_private = sb;
2d069c08 5849 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
b31e1552 5850 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 5851 goto out_journal;
5852 }
5853 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 5854 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5855 "user (unsupported) - %d",
ac27a0ec
DK		 5856 be32_to_cpu(journal->j_superblock->s_nr_users));
5857 goto out_journal;
5858 }
ee7ed3aa 5859 EXT4_SB(sb)->s_journal_bdev = bdev;
617ba13b 5860 ext4_init_journal_params(sb, journal);
ac27a0ec 5861 return journal;
0b8e58a1 5862
ac27a0ec 5863out_journal:
dab291af 5864 jbd2_journal_destroy(journal);
ac27a0ec 5865out_bdev:
617ba13b 5866 ext4_blkdev_put(bdev);
ac27a0ec
DK		 5867 return NULL;
5868}
5869
617ba13b
MC		 5870static int ext4_load_journal(struct super_block *sb,
5871 struct ext4_super_block *es,
ac27a0ec
DK		 5872 unsigned long journal_devnum)
5873{
5874 journal_t *journal;
5875 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5876 dev_t journal_dev;
5877 int err = 0;
5878 int really_read_only;
273108fa 5879 int journal_dev_ro;
ac27a0ec 5880
11215630
JK		 5881 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5882 return -EFSCORRUPTED;
0390131b 5883
ac27a0ec
DK		 5884 if (journal_devnum &&
5885 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 5886 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5887 "numbers have changed");
ac27a0ec
DK		 5888 journal_dev = new_decode_dev(journal_devnum);
5889 } else
5890 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5891
273108fa
LC		 5892 if (journal_inum && journal_dev) {
5893 ext4_msg(sb, KERN_ERR,
5894 "filesystem has both journal inode and journal device!");
5895 return -EINVAL;
5896 }
5897
5898 if (journal_inum) {
5899 journal = ext4_get_journal(sb, journal_inum);
5900 if (!journal)
5901 return -EINVAL;
5902 } else {
5903 journal = ext4_get_dev_journal(sb, journal_dev);
5904 if (!journal)
5905 return -EINVAL;
5906 }
5907
5908 journal_dev_ro = bdev_read_only(journal->j_dev);
5909 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5910
5911 if (journal_dev_ro && !sb_rdonly(sb)) {
5912 ext4_msg(sb, KERN_ERR,
5913 "journal device read-only, try mounting with '-o ro'");
5914 err = -EROFS;
5915 goto err_out;
5916 }
ac27a0ec
DK		 5917
5918 /*
5919 * Are we loading a blank journal or performing recovery after a
5920 * crash? For recovery, we need to check in advance whether we
5921 * can get read-write access to the device.
5922 */
e2b911c5 5923 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 5924 if (sb_rdonly(sb)) {
b31e1552
ES		 5925 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5926 "required on readonly filesystem");
ac27a0ec 5927 if (really_read_only) {
b31e1552 5928 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 5929 "unavailable, cannot proceed "
5930 "(try mounting with noload)");
273108fa
LC		 5931 err = -EROFS;
5932 goto err_out;
ac27a0ec 5933 }
b31e1552
ES		 5934 ext4_msg(sb, KERN_INFO, "write access will "
5935 "be enabled during recovery");
ac27a0ec
DK		 5936 }
5937 }
5938
90576c0b 5939 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 5940 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 5941
e2b911c5 5942 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 5943 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 5944 if (!err) {
5945 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
eee00237 5946
1c13d5c0
TT		 5947 if (save)
5948 memcpy(save, ((char *) es) +
5949 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 5950 err = jbd2_journal_load(journal);
1c13d5c0
TT		 5951 if (save)
5952 memcpy(((char *) es) + EXT4_S_ERR_START,
5953 save, EXT4_S_ERR_LEN);
5954 kfree(save);
eee00237
YB		 5955 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state &
5956 EXT4_ERROR_FS);
5957 /* Write out restored error information to the superblock */
5958 if (!bdev_read_only(sb->s_bdev)) {
5959 int err2;
5960 err2 = ext4_commit_super(sb);
5961 err = err ? : err2;
5962 }
1c13d5c0 5963 }
ac27a0ec
DK		 5964
5965 if (err) {
b31e1552 5966 ext4_msg(sb, KERN_ERR, "error loading journal");
273108fa 5967 goto err_out;
ac27a0ec
DK		 5968 }
5969
617ba13b 5970 EXT4_SB(sb)->s_journal = journal;
11215630
JK		 5971 err = ext4_clear_journal_err(sb, es);
5972 if (err) {
5973 EXT4_SB(sb)->s_journal = NULL;
5974 jbd2_journal_destroy(journal);
5975 return err;
5976 }
ac27a0ec 5977
c41303ce 5978 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 5979 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5980 es->s_journal_dev = cpu_to_le32(journal_devnum);
3039d8b8
BL		 5981 ext4_commit_super(sb);
5982 }
5983 if (!really_read_only && journal_inum &&
5984 journal_inum != le32_to_cpu(es->s_journal_inum)) {
5985 es->s_journal_inum = cpu_to_le32(journal_inum);
4392fbc4 5986 ext4_commit_super(sb);
ac27a0ec
DK		 5987 }
5988
5989 return 0;
273108fa
LC		 5990
5991err_out:
5992 jbd2_journal_destroy(journal);
5993 return err;
ac27a0ec
DK		 5994}
5995
2d01ddc8
JK		 5996/* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
5997static void ext4_update_super(struct super_block *sb)
ac27a0ec 5998{
c92dc856 5999 struct ext4_sb_info *sbi = EXT4_SB(sb);
e92ad03f
JK		 6000 struct ext4_super_block *es = sbi->s_es;
6001 struct buffer_head *sbh = sbi->s_sbh;
a17712c8 6002
05c2c00f 6003 lock_buffer(sbh);
71290b36
TT		 6004 /*
6005 * If the file system is mounted read-only, don't update the
6006 * superblock write time. This avoids updating the superblock
6007 * write time when we are mounting the root file system
6008 * read/only but we need to replay the journal; at that point,
6009 * for people who are east of GMT and who make their clock
6010 * tick in localtime for Windows bug-for-bug compatibility,
6011 * the clock is set in the future, and this will cause e2fsck
6012 * to complain and force a full file system check.
6013 */
1751e8a6 6014 if (!(sb->s_flags & SB_RDONLY))
6a0678a7 6015 ext4_update_tstamp(es, s_wtime);
8446fe92 6016 es->s_kbytes_written =
0bc9bc1d 6017 cpu_to_le64(sbi->s_kbytes_written +
8446fe92 6018 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
0bc9bc1d 6019 sbi->s_sectors_written_start) >> 1));
e92ad03f 6020 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
d5e03cbb 6021 ext4_free_blocks_count_set(es,
e92ad03f
JK		 6022 EXT4_C2B(sbi, percpu_counter_sum_positive(
6023 &sbi->s_freeclusters_counter)));
6024 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
d5e03cbb
TT		 6025 es->s_free_inodes_count =
6026 cpu_to_le32(percpu_counter_sum_positive(
e92ad03f 6027 &sbi->s_freeinodes_counter));
c92dc856
JK		 6028 /* Copy error information to the on-disk superblock */
6029 spin_lock(&sbi->s_error_lock);
6030 if (sbi->s_add_error_count > 0) {
6031 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6032 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6033 __ext4_update_tstamp(&es->s_first_error_time,
6034 &es->s_first_error_time_hi,
6035 sbi->s_first_error_time);
6036 strncpy(es->s_first_error_func, sbi->s_first_error_func,
6037 sizeof(es->s_first_error_func));
6038 es->s_first_error_line =
6039 cpu_to_le32(sbi->s_first_error_line);
6040 es->s_first_error_ino =
6041 cpu_to_le32(sbi->s_first_error_ino);
6042 es->s_first_error_block =
6043 cpu_to_le64(sbi->s_first_error_block);
6044 es->s_first_error_errcode =
6045 ext4_errno_to_code(sbi->s_first_error_code);
6046 }
6047 __ext4_update_tstamp(&es->s_last_error_time,
6048 &es->s_last_error_time_hi,
6049 sbi->s_last_error_time);
6050 strncpy(es->s_last_error_func, sbi->s_last_error_func,
6051 sizeof(es->s_last_error_func));
6052 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6053 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6054 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6055 es->s_last_error_errcode =
6056 ext4_errno_to_code(sbi->s_last_error_code);
6057 /*
6058 * Start the daily error reporting function if it hasn't been
6059 * started already
6060 */
6061 if (!es->s_error_count)
6062 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6063 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6064 sbi->s_add_error_count = 0;
6065 }
6066 spin_unlock(&sbi->s_error_lock);
6067
06db49e6 6068 ext4_superblock_csum_set(sb);
2d01ddc8
JK		 6069 unlock_buffer(sbh);
6070}
6071
6072static int ext4_commit_super(struct super_block *sb)
6073{
6074 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2d01ddc8 6075
f88f1466
FC		 6076 if (!sbh)
6077 return -EINVAL;
6078 if (block_device_ejected(sb))
6079 return -ENODEV;
2d01ddc8
JK		 6080
6081 ext4_update_super(sb);
6082
15baa7dc
ZY		 6083 lock_buffer(sbh);
6084 /* Buffer got discarded which means block device got invalidated */
6085 if (!buffer_mapped(sbh)) {
6086 unlock_buffer(sbh);
6087 return -EIO;
6088 }
6089
e8680786 6090 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4743f839
PKS		 6091 /*
6092 * Oh, dear. A previous attempt to write the
6093 * superblock failed. This could happen because the
6094 * USB device was yanked out. Or it could happen to
6095 * be a transient write error and maybe the block will
6096 * be remapped. Nothing we can do but to retry the
6097 * write and hope for the best.
6098 */
6099 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6100 "superblock detected");
6101 clear_buffer_write_io_error(sbh);
6102 set_buffer_uptodate(sbh);
6103 }
15baa7dc
ZY		 6104 get_bh(sbh);
6105 /* Clear potential dirty bit if it was journalled update */
6106 clear_buffer_dirty(sbh);
6107 sbh->b_end_io = end_buffer_write_sync;
1420c4a5
BVA		 6108 submit_bh(REQ_OP_WRITE | REQ_SYNC |
6109 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
15baa7dc 6110 wait_on_buffer(sbh);
4392fbc4
JK		 6111 if (buffer_write_io_error(sbh)) {
6112 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6113 "superblock");
6114 clear_buffer_write_io_error(sbh);
6115 set_buffer_uptodate(sbh);
15baa7dc 6116 return -EIO;
914258bf 6117 }
15baa7dc 6118 return 0;
ac27a0ec
DK		 6119}
6120
ac27a0ec
DK		 6121/*
6122 * Have we just finished recovery? If so, and if we are mounting (or
6123 * remounting) the filesystem readonly, then we will end up with a
6124 * consistent fs on disk. Record that fact.
6125 */
11215630
JK		 6126static int ext4_mark_recovery_complete(struct super_block *sb,
6127 struct ext4_super_block *es)
ac27a0ec 6128{
11215630 6129 int err;
617ba13b 6130 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 6131
e2b911c5 6132 if (!ext4_has_feature_journal(sb)) {
11215630
JK		 6133 if (journal != NULL) {
6134 ext4_error(sb, "Journal got removed while the fs was "
6135 "mounted!");
6136 return -EFSCORRUPTED;
6137 }
6138 return 0;
0390131b 6139 }
dab291af 6140 jbd2_journal_lock_updates(journal);
01d5d965 6141 err = jbd2_journal_flush(journal, 0);
11215630 6142 if (err < 0)
7ffe1ea8
HK		 6143 goto out;
6144
02f310fc
JK		 6145 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6146 ext4_has_feature_orphan_present(sb))) {
6147 if (!ext4_orphan_file_empty(sb)) {
6148 ext4_error(sb, "Orphan file not empty on read-only fs.");
6149 err = -EFSCORRUPTED;
6150 goto out;
6151 }
e2b911c5 6152 ext4_clear_feature_journal_needs_recovery(sb);
02f310fc 6153 ext4_clear_feature_orphan_present(sb);
4392fbc4 6154 ext4_commit_super(sb);
ac27a0ec 6155 }
7ffe1ea8 6156out:
dab291af 6157 jbd2_journal_unlock_updates(journal);
11215630 6158 return err;
ac27a0ec
DK		 6159}
6160
6161/*
6162 * If we are mounting (or read-write remounting) a filesystem whose journal
6163 * has recorded an error from a previous lifetime, move that error to the
6164 * main filesystem now.
6165 */
11215630 6166static int ext4_clear_journal_err(struct super_block *sb,
2b2d6d01 6167 struct ext4_super_block *es)
ac27a0ec
DK		 6168{
6169 journal_t *journal;
6170 int j_errno;
6171 const char *errstr;
6172
11215630
JK		 6173 if (!ext4_has_feature_journal(sb)) {
6174 ext4_error(sb, "Journal got removed while the fs was mounted!");
6175 return -EFSCORRUPTED;
6176 }
0390131b 6177
617ba13b 6178 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 6179
6180 /*
6181 * Now check for any error status which may have been recorded in the
617ba13b 6182 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 6183 */
6184
dab291af 6185 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 6186 if (j_errno) {
6187 char nbuf[16];
6188
617ba13b 6189 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 6190 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 6191 "from previous mount: %s", errstr);
ac27a0ec 6192
617ba13b
MC		 6193 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6194 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
f57886ca
YB		 6195 j_errno = ext4_commit_super(sb);
6196 if (j_errno)
6197 return j_errno;
6198 ext4_warning(sb, "Marked fs in need of filesystem check.");
ac27a0ec 6199
dab291af 6200 jbd2_journal_clear_err(journal);
d796c52e 6201 jbd2_journal_update_sb_errno(journal);
ac27a0ec 6202 }
11215630 6203 return 0;
ac27a0ec
DK		 6204}
6205
6206/*
6207 * Force the running and committing transactions to commit,
6208 * and wait on the commit.
6209 */
617ba13b 6210int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 6211{
6212 journal_t *journal;
ac27a0ec 6213
bc98a42c 6214 if (sb_rdonly(sb))
ac27a0ec
DK		 6215 return 0;
6216
617ba13b 6217 journal = EXT4_SB(sb)->s_journal;
b1deefc9 6218 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 6219}
6220
617ba13b 6221static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 6222{
14ce0cb4 6223 int ret = 0;
9eddacf9 6224 tid_t target;
06a407f1 6225 bool needs_barrier = false;
8d5d02e6 6226 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 6227
49598e04 6228 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 6229 return 0;
6230
9bffad1e 6231 trace_ext4_sync_fs(sb, wait);
2e8fa54e 6232 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 6233 /*
6234 * Writeback quota in non-journalled quota case - journalled quota has
6235 * no dirty dquots
6236 */
6237 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 6238 /*
6239 * Data writeback is possible w/o journal transaction, so barrier must
6240 * being sent at the end of the function. But we can skip it if
6241 * transaction_commit will do it for us.
6242 */
bda32530
TT		 6243 if (sbi->s_journal) {
6244 target = jbd2_get_latest_transaction(sbi->s_journal);
6245 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6246 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6247 needs_barrier = true;
6248
6249 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6250 if (wait)
6251 ret = jbd2_log_wait_commit(sbi->s_journal,
6252 target);
6253 }
6254 } else if (wait && test_opt(sb, BARRIER))
06a407f1 6255 needs_barrier = true;
06a407f1
DM		 6256 if (needs_barrier) {
6257 int err;
c6bf3f0e 6258 err = blkdev_issue_flush(sb->s_bdev);
06a407f1
DM		 6259 if (!ret)
6260 ret = err;
0390131b 6261 }
06a407f1
DM		 6262
6263 return ret;
6264}
6265
ac27a0ec
DK		 6266/*
6267 * LVM calls this function before a (read-only) snapshot is created. This
6268 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 6269 *
6270 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 6271 * state independently. It relies on upper layer to stop all data & metadata
6272 * modifications.
ac27a0ec 6273 */
c4be0c1d 6274static int ext4_freeze(struct super_block *sb)
ac27a0ec 6275{
c4be0c1d
TS		 6276 int error = 0;
6277 journal_t *journal;
ac27a0ec 6278
bc98a42c 6279 if (sb_rdonly(sb))
9ca92389 6280 return 0;
ac27a0ec 6281
9ca92389 6282 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 6283
bb044576
TT		 6284 if (journal) {
6285 /* Now we set up the journal barrier. */
6286 jbd2_journal_lock_updates(journal);
ac27a0ec 6287
bb044576
TT		 6288 /*
6289 * Don't clear the needs_recovery flag if we failed to
6290 * flush the journal.
6291 */
01d5d965 6292 error = jbd2_journal_flush(journal, 0);
bb044576
TT		 6293 if (error < 0)
6294 goto out;
c642dc9e
ES		 6295
6296 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 6297 ext4_clear_feature_journal_needs_recovery(sb);
02f310fc
JK		 6298 if (ext4_orphan_file_empty(sb))
6299 ext4_clear_feature_orphan_present(sb);
bb044576 6300 }
9ca92389 6301
4392fbc4 6302 error = ext4_commit_super(sb);
6b0310fb 6303out:
bb044576
TT		 6304 if (journal)
6305 /* we rely on upper layer to stop further updates */
6306 jbd2_journal_unlock_updates(journal);
6b0310fb 6307 return error;
ac27a0ec
DK		 6308}
6309
6310/*
6311 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6312 * flag here, even though the filesystem is not technically dirty yet.
6313 */
c4be0c1d 6314static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 6315{
bc98a42c 6316 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 6317 return 0;
6318
c642dc9e
ES		 6319 if (EXT4_SB(sb)->s_journal) {
6320 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 6321 ext4_set_feature_journal_needs_recovery(sb);
02f310fc
JK		 6322 if (ext4_has_feature_orphan_file(sb))
6323 ext4_set_feature_orphan_present(sb);
c642dc9e
ES		 6324 }
6325
4392fbc4 6326 ext4_commit_super(sb);
c4be0c1d 6327 return 0;
ac27a0ec
DK		 6328}
6329
673c6100
TT		 6330/*
6331 * Structure to save mount options for ext4_remount's benefit
6332 */
6333struct ext4_mount_options {
6334 unsigned long s_mount_opt;
a2595b8a 6335 unsigned long s_mount_opt2;
08cefc7a
EB		 6336 kuid_t s_resuid;
6337 kgid_t s_resgid;
673c6100
TT		 6338 unsigned long s_commit_interval;
6339 u32 s_min_batch_time, s_max_batch_time;
6340#ifdef CONFIG_QUOTA
6341 int s_jquota_fmt;
a2d4a646 6342 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 6343#endif
6344};
6345
960e0ab6 6346static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
ac27a0ec 6347{
7edfd85b 6348 struct ext4_fs_context *ctx = fc->fs_private;
2b2d6d01 6349 struct ext4_super_block *es;
617ba13b 6350 struct ext4_sb_info *sbi = EXT4_SB(sb);
960e0ab6 6351 unsigned long old_sb_flags;
617ba13b 6352 struct ext4_mount_options old_opts;
8a266467 6353 ext4_group_t g;
c5e06d10 6354 int err = 0;
ac27a0ec 6355#ifdef CONFIG_QUOTA
3bbef91b 6356 int enable_quota = 0;
03dafb5f 6357 int i, j;
33458eab 6358 char *to_free[EXT4_MAXQUOTAS];
ac27a0ec
DK		 6359#endif
6360
21ac738e 6361
ac27a0ec
DK		 6362 /* Store the original options */
6363 old_sb_flags = sb->s_flags;
6364 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 6365 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 6366 old_opts.s_resuid = sbi->s_resuid;
6367 old_opts.s_resgid = sbi->s_resgid;
6368 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 6369 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6370 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 6371#ifdef CONFIG_QUOTA
6372 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 6373 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 6374 if (sbi->s_qf_names[i]) {
33458eab
TT		 6375 char *qf_name = get_qf_name(sb, sbi, i);
6376
6377 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
03dafb5f
CG		 6378 if (!old_opts.s_qf_names[i]) {
6379 for (j = 0; j < i; j++)
6380 kfree(old_opts.s_qf_names[j]);
6381 return -ENOMEM;
6382 }
6383 } else
6384 old_opts.s_qf_names[i] = NULL;
ac27a0ec 6385#endif
e4e58e5d
OM		 6386 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6387 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6388 ctx->journal_ioprio =
6389 sbi->s_journal->j_task->io_context->ioprio;
6390 else
6391 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6392
6393 }
ac27a0ec 6394
7edfd85b 6395 ext4_apply_options(fc, sb);
ac27a0ec 6396
6b992ff2 6397 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 6398 test_opt(sb, JOURNAL_CHECKSUM)) {
6399 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 6400 "during remount not supported; ignoring");
6401 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 6402 }
6403
6ae6514b
PS		 6404 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6405 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6406 ext4_msg(sb, KERN_ERR, "can't mount with "
6407 "both data=journal and delalloc");
6408 err = -EINVAL;
6409 goto restore_opts;
6410 }
6411 if (test_opt(sb, DIOREAD_NOLOCK)) {
6412 ext4_msg(sb, KERN_ERR, "can't mount with "
6413 "both data=journal and dioread_nolock");
6414 err = -EINVAL;
6415 goto restore_opts;
6416 }
ab04df78
JK		 6417 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6418 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6419 ext4_msg(sb, KERN_ERR, "can't mount with "
6420 "journal_async_commit in data=ordered mode");
6421 err = -EINVAL;
6422 goto restore_opts;
6423 }
923ae0ff
RZ		 6424 }
6425
cdb7ee4c
TE		 6426 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6427 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6428 err = -EINVAL;
6429 goto restore_opts;
6430 }
6431
9b5f6c9b 6432 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
124e7c61 6433 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
ac27a0ec 6434
1751e8a6
LT		 6435 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6436 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 6437
6438 es = sbi->s_es;
6439
b3881f74 6440 if (sbi->s_journal) {
0390131b 6441 ext4_init_journal_params(sb, sbi->s_journal);
7edfd85b 6442 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
b3881f74 6443 }
ac27a0ec 6444
c92dc856
JK		 6445 /* Flush outstanding errors before changing fs state */
6446 flush_work(&sbi->s_error_work);
6447
960e0ab6 6448 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
9b5f6c9b 6449 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
ac27a0ec
DK		 6450 err = -EROFS;
6451 goto restore_opts;
6452 }
6453
960e0ab6 6454 if (fc->sb_flags & SB_RDONLY) {
38c03b34
TT		 6455 err = sync_filesystem(sb);
6456 if (err < 0)
6457 goto restore_opts;
0f0dd62f
CH		 6458 err = dquot_suspend(sb, -1);
6459 if (err < 0)
c79d967d 6460 goto restore_opts;
c79d967d 6461
ac27a0ec
DK		 6462 /*
6463 * First of all, the unconditional stuff we have to do
6464 * to disable replay of the journal when we next remount
6465 */
1751e8a6 6466 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 6467
6468 /*
6469 * OK, test if we are remounting a valid rw partition
6470 * readonly, and if so set the rdonly flag and then
6471 * mark the partition as valid again.
6472 */
617ba13b
MC		 6473 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6474 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 6475 es->s_state = cpu_to_le16(sbi->s_mount_state);
6476
11215630
JK		 6477 if (sbi->s_journal) {
6478 /*
6479 * We let remount-ro finish even if marking fs
6480 * as clean failed...
6481 */
0390131b 6482 ext4_mark_recovery_complete(sb, es);
11215630 6483 }
ac27a0ec 6484 } else {
a13fb1a4 6485 /* Make sure we can mount this feature set readwrite */
e2b911c5 6486 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 6487 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 6488 err = -EROFS;
6489 goto restore_opts;
6490 }
8a266467
TT		 6491 /*
6492 * Make sure the group descriptor checksums
0b8e58a1 6493 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 6494 */
6495 for (g = 0; g < sbi->s_groups_count; g++) {
6496 struct ext4_group_desc *gdp =
6497 ext4_get_group_desc(sb, g, NULL);
6498
feb0ab32 6499 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 6500 ext4_msg(sb, KERN_ERR,
6501 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 6502 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 6503 le16_to_cpu(gdp->bg_checksum));
6a797d27 6504 err = -EFSBADCRC;
8a266467
TT		 6505 goto restore_opts;
6506 }
6507 }
6508
ead6596b
ES		 6509 /*
6510 * If we have an unprocessed orphan list hanging
6511 * around from a previously readonly bdev mount,
6512 * require a full umount/remount for now.
6513 */
02f310fc 6514 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
b31e1552 6515 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 6516 "remount RDWR because of unprocessed "
6517 "orphan inode list. Please "
b31e1552 6518 "umount/remount instead");
ead6596b
ES		 6519 err = -EINVAL;
6520 goto restore_opts;
6521 }
6522
ac27a0ec
DK		 6523 /*
6524 * Mounting a RDONLY partition read-write, so reread
6525 * and store the current valid flag. (It may have
6526 * been changed by e2fsck since we originally mounted
6527 * the partition.)
6528 */
11215630
JK		 6529 if (sbi->s_journal) {
6530 err = ext4_clear_journal_err(sb, es);
6531 if (err)
6532 goto restore_opts;
6533 }
c878bea3
TT		 6534 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6535 ~EXT4_FC_REPLAY);
c89128a0
JK		 6536
6537 err = ext4_setup_super(sb, es, 0);
6538 if (err)
6539 goto restore_opts;
6540
6541 sb->s_flags &= ~SB_RDONLY;
e2b911c5 6542 if (ext4_has_feature_mmp(sb))
c5e06d10
JL		 6543 if (ext4_multi_mount_protect(sb,
6544 le64_to_cpu(es->s_mmp_block))) {
6545 err = -EROFS;
6546 goto restore_opts;
6547 }
3bbef91b 6548#ifdef CONFIG_QUOTA
c79d967d 6549 enable_quota = 1;
3bbef91b 6550#endif
ac27a0ec
DK		 6551 }
6552 }
bfff6873
LC		 6553
6554 /*
6555 * Reinitialize lazy itable initialization thread based on
6556 * current settings
6557 */
bc98a42c 6558 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
bfff6873
LC		 6559 ext4_unregister_li_request(sb);
6560 else {
6561 ext4_group_t first_not_zeroed;
6562 first_not_zeroed = ext4_has_uninit_itable(sb);
6563 ext4_register_li_request(sb, first_not_zeroed);
6564 }
6565
0f5bde1d
JK		 6566 /*
6567 * Handle creation of system zone data early because it can fail.
6568 * Releasing of existing data is done when we are sure remount will
6569 * succeed.
6570 */
dd0db94f 6571 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
0f5bde1d
JK		 6572 err = ext4_setup_system_zone(sb);
6573 if (err)
6574 goto restore_opts;
6575 }
d176b1f6 6576
c89128a0 6577 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
4392fbc4 6578 err = ext4_commit_super(sb);
c89128a0
JK		 6579 if (err)
6580 goto restore_opts;
6581 }
0390131b 6582
ac27a0ec
DK		 6583#ifdef CONFIG_QUOTA
6584 /* Release old quota file names */
a2d4a646 6585 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 6586 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 6587 if (enable_quota) {
6588 if (sb_any_quota_suspended(sb))
6589 dquot_resume(sb, -1);
e2b911c5 6590 else if (ext4_has_feature_quota(sb)) {
7c319d32 6591 err = ext4_enable_quotas(sb);
07724f98 6592 if (err)
7c319d32 6593 goto restore_opts;
7c319d32
AK		 6594 }
6595 }
ac27a0ec 6596#endif
dd0db94f 6597 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
0f5bde1d 6598 ext4_release_system_zone(sb);
d4c402d9 6599
61bb4a1c
TT		 6600 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6601 ext4_stop_mmpd(sbi);
6602
ac27a0ec 6603 return 0;
0b8e58a1 6604
ac27a0ec
DK		 6605restore_opts:
6606 sb->s_flags = old_sb_flags;
6607 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 6608 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 6609 sbi->s_resuid = old_opts.s_resuid;
6610 sbi->s_resgid = old_opts.s_resgid;
6611 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 6612 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6613 sbi->s_max_batch_time = old_opts.s_max_batch_time;
dd0db94f 6614 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
0f5bde1d 6615 ext4_release_system_zone(sb);
ac27a0ec
DK		 6616#ifdef CONFIG_QUOTA
6617 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 6618 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
33458eab
TT		 6619 to_free[i] = get_qf_name(sb, sbi, i);
6620 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
ac27a0ec 6621 }
33458eab
TT		 6622 synchronize_rcu();
6623 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6624 kfree(to_free[i]);
ac27a0ec 6625#endif
61bb4a1c
TT		 6626 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6627 ext4_stop_mmpd(sbi);
ac27a0ec
DK		 6628 return err;
6629}
6630
cebe85d5 6631static int ext4_reconfigure(struct fs_context *fc)
7edfd85b 6632{
cebe85d5 6633 struct super_block *sb = fc->root->d_sb;
7edfd85b
LC		 6634 int ret;
6635
cebe85d5 6636 fc->s_fs_info = EXT4_SB(sb);
7edfd85b 6637
cebe85d5 6638 ret = ext4_check_opt_consistency(fc, sb);
7edfd85b 6639 if (ret < 0)
cebe85d5 6640 return ret;
7edfd85b 6641
960e0ab6 6642 ret = __ext4_remount(fc, sb);
7edfd85b 6643 if (ret < 0)
cebe85d5 6644 return ret;
7edfd85b 6645
bb0fbc78
LC		 6646 ext4_msg(sb, KERN_INFO, "re-mounted %pU. Quota mode: %s.",
6647 &sb->s_uuid, ext4_quota_mode(sb));
7edfd85b 6648
7edfd85b 6649 return 0;
7edfd85b
LC		 6650}
6651
689c958c
LX		 6652#ifdef CONFIG_QUOTA
6653static int ext4_statfs_project(struct super_block *sb,
6654 kprojid_t projid, struct kstatfs *buf)
6655{
6656 struct kqid qid;
6657 struct dquot *dquot;
6658 u64 limit;
6659 u64 curblock;
6660
6661 qid = make_kqid_projid(projid);
6662 dquot = dqget(sb, qid);
6663 if (IS_ERR(dquot))
6664 return PTR_ERR(dquot);
7b9ca4c6 6665 spin_lock(&dquot->dq_dqb_lock);
689c958c 6666
a08fe66e
CX		 6667 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6668 dquot->dq_dqb.dqb_bhardlimit);
57c32ea4
CX		 6669 limit >>= sb->s_blocksize_bits;
6670
689c958c 6671 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 6672 curblock = (dquot->dq_dqb.dqb_curspace +
6673 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 6674 buf->f_blocks = limit;
6675 buf->f_bfree = buf->f_bavail =
6676 (buf->f_blocks > curblock) ?
6677 (buf->f_blocks - curblock) : 0;
6678 }
6679
a08fe66e
CX		 6680 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6681 dquot->dq_dqb.dqb_ihardlimit);
689c958c
LX		 6682 if (limit && buf->f_files > limit) {
6683 buf->f_files = limit;
6684 buf->f_ffree =
6685 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6686 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6687 }
6688
7b9ca4c6 6689 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 6690 dqput(dquot);
6691 return 0;
6692}
6693#endif
6694
2b2d6d01 6695static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 6696{
6697 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 6698 struct ext4_sb_info *sbi = EXT4_SB(sb);
6699 struct ext4_super_block *es = sbi->s_es;
27dd4385 6700 ext4_fsblk_t overhead = 0, resv_blocks;
d02a9391 6701 s64 bfree;
27dd4385 6702 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 6703
952fc18e
TT		 6704 if (!test_opt(sb, MINIX_DF))
6705 overhead = sbi->s_overhead;
ac27a0ec 6706
617ba13b 6707 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 6708 buf->f_bsize = sb->s_blocksize;
b72f78cb 6709 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 6710 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6711 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 6712 /* prevent underflow in case that few free space is available */
57042651 6713 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 6714 buf->f_bavail = buf->f_bfree -
6715 (ext4_r_blocks_count(es) + resv_blocks);
6716 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 6717 buf->f_bavail = 0;
6718 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 6719 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 6720 buf->f_namelen = EXT4_NAME_LEN;
9591c3a3 6721 buf->f_fsid = uuid_to_fsid(es->s_uuid);
0b8e58a1 6722
689c958c
LX		 6723#ifdef CONFIG_QUOTA
6724 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6725 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6726 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6727#endif
ac27a0ec
DK		 6728 return 0;
6729}
6730
ac27a0ec
DK		 6731
6732#ifdef CONFIG_QUOTA
6733
bc8230ee
JK		 6734/*
6735 * Helper functions so that transaction is started before we acquire dqio_sem
6736 * to keep correct lock ordering of transaction > dqio_sem
6737 */
ac27a0ec
DK		 6738static inline struct inode *dquot_to_inode(struct dquot *dquot)
6739{
4c376dca 6740 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 6741}
6742
617ba13b 6743static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 6744{
6745 int ret, err;
6746 handle_t *handle;
6747 struct inode *inode;
6748
6749 inode = dquot_to_inode(dquot);
9924a92a 6750 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 6751 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 6752 if (IS_ERR(handle))
6753 return PTR_ERR(handle);
6754 ret = dquot_commit(dquot);
617ba13b 6755 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6756 if (!ret)
6757 ret = err;
6758 return ret;
6759}
6760
617ba13b 6761static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 6762{
6763 int ret, err;
6764 handle_t *handle;
6765
9924a92a 6766 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 6767 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 6768 if (IS_ERR(handle))
6769 return PTR_ERR(handle);
6770 ret = dquot_acquire(dquot);
617ba13b 6771 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6772 if (!ret)
6773 ret = err;
6774 return ret;
6775}
6776
617ba13b 6777static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 6778{
6779 int ret, err;
6780 handle_t *handle;
6781
9924a92a 6782 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 6783 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 6784 if (IS_ERR(handle)) {
6785 /* Release dquot anyway to avoid endless cycle in dqput() */
6786 dquot_release(dquot);
ac27a0ec 6787 return PTR_ERR(handle);
9c3013e9 6788 }
ac27a0ec 6789 ret = dquot_release(dquot);
617ba13b 6790 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6791 if (!ret)
6792 ret = err;
6793 return ret;
6794}
6795
617ba13b 6796static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 6797{
262b4662 6798 struct super_block *sb = dquot->dq_sb;
262b4662 6799
f177ee08 6800 if (ext4_is_quota_journalled(sb)) {
ac27a0ec 6801 dquot_mark_dquot_dirty(dquot);
617ba13b 6802 return ext4_write_dquot(dquot);
ac27a0ec
DK		 6803 } else {
6804 return dquot_mark_dquot_dirty(dquot);
6805 }
6806}
6807
617ba13b 6808static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 6809{
6810 int ret, err;
6811 handle_t *handle;
6812
6813 /* Data block + inode block */
f9c1f248 6814 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 6815 if (IS_ERR(handle))
6816 return PTR_ERR(handle);
6817 ret = dquot_commit_info(sb, type);
617ba13b 6818 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6819 if (!ret)
6820 ret = err;
6821 return ret;
6822}
6823
daf647d2
TT		 6824static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6825{
6826 struct ext4_inode_info *ei = EXT4_I(inode);
6827
6828 /* The first argument of lockdep_set_subclass has to be
6829 * *exactly* the same as the argument to init_rwsem() --- in
6830 * this case, in init_once() --- or lockdep gets unhappy
6831 * because the name of the lock is set using the
6832 * stringification of the argument to init_rwsem().
6833 */
6834 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6835 lockdep_set_subclass(&ei->i_data_sem, subclass);
6836}
6837
ac27a0ec
DK		 6838/*
6839 * Standard function to be called on quota_on
6840 */
617ba13b 6841static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 6842 const struct path *path)
ac27a0ec
DK		 6843{
6844 int err;
ac27a0ec
DK		 6845
6846 if (!test_opt(sb, QUOTA))
6847 return -EINVAL;
0623543b 6848
ac27a0ec 6849 /* Quotafile not on the same filesystem? */
d8c9584e 6850 if (path->dentry->d_sb != sb)
ac27a0ec 6851 return -EXDEV;
e0770e91
JK		 6852
6853 /* Quota already enabled for this file? */
6854 if (IS_NOQUOTA(d_inode(path->dentry)))
6855 return -EBUSY;
6856
0623543b
JK		 6857 /* Journaling quota? */
6858 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 6859 /* Quotafile not in fs root? */
f00c9e44 6860 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 6861 ext4_msg(sb, KERN_WARNING,
6862 "Quota file not on filesystem root. "
6863 "Journaled quota will not work");
91389240
JK		 6864 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6865 } else {
6866 /*
6867 * Clear the flag just in case mount options changed since
6868 * last time.
6869 */
6870 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 6871 }
0623543b
JK		 6872
6873 /*
6874 * When we journal data on quota file, we have to flush journal to see
6875 * all updates to the file when we bypass pagecache...
6876 */
0390131b 6877 if (EXT4_SB(sb)->s_journal &&
2b0143b5 6878 ext4_should_journal_data(d_inode(path->dentry))) {
0623543b
JK		 6879 /*
6880 * We don't need to lock updates but journal_flush() could
6881 * otherwise be livelocked...
6882 */
6883 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
01d5d965 6884 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal, 0);
0623543b 6885 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 6886 if (err)
7ffe1ea8 6887 return err;
0623543b 6888 }
957153fc 6889
daf647d2
TT		 6890 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6891 err = dquot_quota_on(sb, type, format_id, path);
15fc69bb 6892 if (!err) {
957153fc
JK		 6893 struct inode *inode = d_inode(path->dentry);
6894 handle_t *handle;
6895
61a92987
JK		 6896 /*
6897 * Set inode flags to prevent userspace from messing with quota
6898 * files. If this fails, we return success anyway since quotas
6899 * are already enabled and this is not a hard failure.
6900 */
957153fc
JK		 6901 inode_lock(inode);
6902 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6903 if (IS_ERR(handle))
6904 goto unlock_inode;
6905 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6906 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6907 S_NOATIME | S_IMMUTABLE);
4209ae12 6908 err = ext4_mark_inode_dirty(handle, inode);
957153fc
JK		 6909 ext4_journal_stop(handle);
6910 unlock_inode:
6911 inode_unlock(inode);
15fc69bb
JK		 6912 if (err)
6913 dquot_quota_off(sb, type);
957153fc 6914 }
15fc69bb
JK		 6915 if (err)
6916 lockdep_set_quota_inode(path->dentry->d_inode,
6917 I_DATA_SEM_NORMAL);
daf647d2 6918 return err;
ac27a0ec
DK		 6919}
6920
07342ec2
BL		 6921static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
6922{
6923 switch (type) {
6924 case USRQUOTA:
6925 return qf_inum == EXT4_USR_QUOTA_INO;
6926 case GRPQUOTA:
6927 return qf_inum == EXT4_GRP_QUOTA_INO;
6928 case PRJQUOTA:
6929 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
6930 default:
6931 BUG();
6932 }
6933}
6934
7c319d32
AK		 6935static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6936 unsigned int flags)
6937{
6938 int err;
6939 struct inode *qf_inode;
a2d4a646 6940 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 6941 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 6942 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6943 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 6944 };
6945
e2b911c5 6946 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 6947
6948 if (!qf_inums[type])
6949 return -EPERM;
6950
07342ec2
BL		 6951 if (!ext4_check_quota_inum(type, qf_inums[type])) {
6952 ext4_error(sb, "Bad quota inum: %lu, type: %d",
6953 qf_inums[type], type);
6954 return -EUCLEAN;
6955 }
6956
8a363970 6957 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7c319d32 6958 if (IS_ERR(qf_inode)) {
07342ec2
BL		 6959 ext4_error(sb, "Bad quota inode: %lu, type: %d",
6960 qf_inums[type], type);
7c319d32
AK		 6961 return PTR_ERR(qf_inode);
6962 }
6963
bcb13850
JK		 6964 /* Don't account quota for quota files to avoid recursion */
6965 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 6966 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7212b95e 6967 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
daf647d2
TT		 6968 if (err)
6969 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
61157b24 6970 iput(qf_inode);
7c319d32
AK		 6971
6972 return err;
6973}
6974
6975/* Enable usage tracking for all quota types. */
25c6d98f 6976int ext4_enable_quotas(struct super_block *sb)
7c319d32
AK		 6977{
6978 int type, err = 0;
a2d4a646 6979 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 6980 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 6981 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6982 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 6983 };
49da9392
JK		 6984 bool quota_mopt[EXT4_MAXQUOTAS] = {
6985 test_opt(sb, USRQUOTA),
6986 test_opt(sb, GRPQUOTA),
6987 test_opt(sb, PRJQUOTA),
6988 };
7c319d32 6989
91389240 6990 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 6991 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 6992 if (qf_inums[type]) {
6993 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 6994 DQUOT_USAGE_ENABLED |
6995 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 6996 if (err) {
6997 ext4_warning(sb,
72ba7450 6998 "Failed to enable quota tracking "
07342ec2
BL		 6999 "(type=%d, err=%d, ino=%lu). "
7000 "Please run e2fsck to fix.", type,
7001 err, qf_inums[type]);
4013d47a
JK		 7002 for (type--; type >= 0; type--) {
7003 struct inode *inode;
7004
7005 inode = sb_dqopt(sb)->files[type];
7006 if (inode)
7007 inode = igrab(inode);
7f144fd0 7008 dquot_quota_off(sb, type);
4013d47a
JK		 7009 if (inode) {
7010 lockdep_set_quota_inode(inode,
7011 I_DATA_SEM_NORMAL);
7012 iput(inode);
7013 }
7014 }
7f144fd0 7015
7c319d32
AK		 7016 return err;
7017 }
7018 }
7019 }
7020 return 0;
7021}
7022
ca0e05e4
DM		 7023static int ext4_quota_off(struct super_block *sb, int type)
7024{
21f97697
JK		 7025 struct inode *inode = sb_dqopt(sb)->files[type];
7026 handle_t *handle;
957153fc 7027 int err;
21f97697 7028
87009d86
DM		 7029 /* Force all delayed allocation blocks to be allocated.
7030 * Caller already holds s_umount sem */
7031 if (test_opt(sb, DELALLOC))
ca0e05e4 7032 sync_filesystem(sb);
ca0e05e4 7033
957153fc 7034 if (!inode || !igrab(inode))
0b268590
AG		 7035 goto out;
7036
957153fc 7037 err = dquot_quota_off(sb, type);
964edf66 7038 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 7039 goto out_put;
7040
7041 inode_lock(inode);
61a92987
JK		 7042 /*
7043 * Update modification times of quota files when userspace can
7044 * start looking at them. If we fail, we return success anyway since
7045 * this is not a hard failure and quotas are already disabled.
7046 */
9924a92a 7047 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
4209ae12
HS		 7048 if (IS_ERR(handle)) {
7049 err = PTR_ERR(handle);
957153fc 7050 goto out_unlock;
4209ae12 7051 }
957153fc
JK		 7052 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7053 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 7054 inode->i_mtime = inode->i_ctime = current_time(inode);
4209ae12 7055 err = ext4_mark_inode_dirty(handle, inode);
21f97697 7056 ext4_journal_stop(handle);
957153fc
JK		 7057out_unlock:
7058 inode_unlock(inode);
7059out_put:
964edf66 7060 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 7061 iput(inode);
7062 return err;
21f97697 7063out:
ca0e05e4
DM		 7064 return dquot_quota_off(sb, type);
7065}
7066
ac27a0ec
DK		 7067/* Read data from quotafile - avoid pagecache and such because we cannot afford
7068 * acquiring the locks... As quota files are never truncated and quota code
25985edc 7069 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 7070 * we don't have to be afraid of races */
617ba13b 7071static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 7072 size_t len, loff_t off)
7073{
7074 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 7075 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 7076 int offset = off & (sb->s_blocksize - 1);
7077 int tocopy;
7078 size_t toread;
7079 struct buffer_head *bh;
7080 loff_t i_size = i_size_read(inode);
7081
7082 if (off > i_size)
7083 return 0;
7084 if (off+len > i_size)
7085 len = i_size-off;
7086 toread = len;
7087 while (toread > 0) {
66267814 7088 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1c215028
TT		 7089 bh = ext4_bread(NULL, inode, blk, 0);
7090 if (IS_ERR(bh))
7091 return PTR_ERR(bh);
ac27a0ec
DK		 7092 if (!bh) /* A hole? */
7093 memset(data, 0, tocopy);
7094 else
7095 memcpy(data, bh->b_data+offset, tocopy);
7096 brelse(bh);
7097 offset = 0;
7098 toread -= tocopy;
7099 data += tocopy;
7100 blk++;
7101 }
7102 return len;
7103}
7104
7105/* Write to quotafile (we know the transaction is already started and has
7106 * enough credits) */
617ba13b 7107static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 7108 const char *data, size_t len, loff_t off)
7109{
7110 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 7111 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4209ae12 7112 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
c5e298ae 7113 int retries = 0;
ac27a0ec
DK		 7114 struct buffer_head *bh;
7115 handle_t *handle = journal_current_handle();
7116
380a0091 7117 if (!handle) {
b31e1552
ES		 7118 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7119 " cancelled because transaction is not started",
9c3013e9
JK		 7120 (unsigned long long)off, (unsigned long long)len);
7121 return -EIO;
7122 }
67eeb568
DM		 7123 /*
7124 * Since we account only one data block in transaction credits,
7125 * then it is impossible to cross a block boundary.
7126 */
7127 if (sb->s_blocksize - offset < len) {
7128 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7129 " cancelled because not block aligned",
7130 (unsigned long long)off, (unsigned long long)len);
7131 return -EIO;
7132 }
7133
c5e298ae
TT		 7134 do {
7135 bh = ext4_bread(handle, inode, blk,
7136 EXT4_GET_BLOCKS_CREATE |
7137 EXT4_GET_BLOCKS_METADATA_NOFAIL);
45586c70 7138 } while (PTR_ERR(bh) == -ENOSPC &&
c5e298ae 7139 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 7140 if (IS_ERR(bh))
7141 return PTR_ERR(bh);
67eeb568
DM		 7142 if (!bh)
7143 goto out;
5d601255 7144 BUFFER_TRACE(bh, "get write access");
188c299e 7145 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
62d2b5f2
JK		 7146 if (err) {
7147 brelse(bh);
1c215028 7148 return err;
ac27a0ec 7149 }
67eeb568
DM		 7150 lock_buffer(bh);
7151 memcpy(bh->b_data+offset, data, len);
7152 flush_dcache_page(bh->b_page);
7153 unlock_buffer(bh);
62d2b5f2 7154 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 7155 brelse(bh);
ac27a0ec 7156out:
67eeb568
DM		 7157 if (inode->i_size < off + len) {
7158 i_size_write(inode, off + len);
617ba13b 7159 EXT4_I(inode)->i_disksize = inode->i_size;
4209ae12
HS		 7160 err2 = ext4_mark_inode_dirty(handle, inode);
7161 if (unlikely(err2 && !err))
7162 err = err2;
ac27a0ec 7163 }
4209ae12 7164 return err ? err : len;
ac27a0ec 7165}
ac27a0ec
DK		 7166#endif
7167
c290ea01 7168#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 7169static inline void register_as_ext2(void)
7170{
7171 int err = register_filesystem(&ext2_fs_type);
7172 if (err)
7173 printk(KERN_WARNING
7174 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7175}
7176
7177static inline void unregister_as_ext2(void)
7178{
7179 unregister_filesystem(&ext2_fs_type);
7180}
2035e776
TT		 7181
7182static inline int ext2_feature_set_ok(struct super_block *sb)
7183{
e2b911c5 7184 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 7185 return 0;
bc98a42c 7186 if (sb_rdonly(sb))
2035e776 7187 return 1;
e2b911c5 7188 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 7189 return 0;
7190 return 1;
7191}
24b58424
TT		 7192#else
7193static inline void register_as_ext2(void) { }
7194static inline void unregister_as_ext2(void) { }
2035e776 7195static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 7196#endif
7197
24b58424
TT		 7198static inline void register_as_ext3(void)
7199{
7200 int err = register_filesystem(&ext3_fs_type);
7201 if (err)
7202 printk(KERN_WARNING
7203 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7204}
7205
7206static inline void unregister_as_ext3(void)
7207{
7208 unregister_filesystem(&ext3_fs_type);
7209}
2035e776
TT		 7210
7211static inline int ext3_feature_set_ok(struct super_block *sb)
7212{
e2b911c5 7213 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 7214 return 0;
e2b911c5 7215 if (!ext4_has_feature_journal(sb))
2035e776 7216 return 0;
bc98a42c 7217 if (sb_rdonly(sb))
2035e776 7218 return 1;
e2b911c5 7219 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 7220 return 0;
7221 return 1;
7222}
24b58424 7223
03010a33 7224static struct file_system_type ext4_fs_type = {
cebe85d5
LC		 7225 .owner = THIS_MODULE,
7226 .name = "ext4",
7227 .init_fs_context = ext4_init_fs_context,
7228 .parameters = ext4_param_specs,
7229 .kill_sb = kill_block_super,
7230 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
03010a33 7231};
7f78e035 7232MODULE_ALIAS_FS("ext4");
03010a33 7233
e9e3bcec
ES		 7234/* Shared across all ext4 file systems */
7235wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 7236
5dabfc78 7237static int __init ext4_init_fs(void)
ac27a0ec 7238{
e9e3bcec 7239 int i, err;
c9de560d 7240
e294a537 7241 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8 7242 ext4_li_info = NULL;
07c0c5d8 7243
9a4c8019 7244 /* Build-time check for flags consistency */
12e9b892 7245 ext4_check_flag_values();
e9e3bcec 7246
e142d052 7247 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 7248 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 7249
51865fda 7250 err = ext4_init_es();
6fd058f7
TT		 7251 if (err)
7252 return err;
51865fda 7253
1dc0aa46 7254 err = ext4_init_pending();
22cfe4b4
EB		 7255 if (err)
7256 goto out7;
7257
7258 err = ext4_init_post_read_processing();
1dc0aa46
EW		 7259 if (err)
7260 goto out6;
7261
51865fda
ZL		 7262 err = ext4_init_pageio();
7263 if (err)
b5799018 7264 goto out5;
51865fda 7265
5dabfc78 7266 err = ext4_init_system_zone();
bd2d0210 7267 if (err)
b5799018 7268 goto out4;
857ac889 7269
b5799018 7270 err = ext4_init_sysfs();
dd68314c 7271 if (err)
b5799018 7272 goto out3;
857ac889 7273
5dabfc78 7274 err = ext4_init_mballoc();
c9de560d
AT		 7275 if (err)
7276 goto out2;
ac27a0ec
DK		 7277 err = init_inodecache();
7278 if (err)
7279 goto out1;
aa75f4d3
HS		 7280
7281 err = ext4_fc_init_dentry_cache();
7282 if (err)
7283 goto out05;
7284
24b58424 7285 register_as_ext3();
2035e776 7286 register_as_ext2();
03010a33 7287 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 7288 if (err)
7289 goto out;
bfff6873 7290
ac27a0ec
DK		 7291 return 0;
7292out:
24b58424
TT		 7293 unregister_as_ext2();
7294 unregister_as_ext3();
ab047d51 7295 ext4_fc_destroy_dentry_cache();
aa75f4d3 7296out05:
ac27a0ec
DK		 7297 destroy_inodecache();
7298out1:
5dabfc78 7299 ext4_exit_mballoc();
9c191f70 7300out2:
b5799018
TT		 7301 ext4_exit_sysfs();
7302out3:
5dabfc78 7303 ext4_exit_system_zone();
b5799018 7304out4:
5dabfc78 7305 ext4_exit_pageio();
b5799018 7306out5:
22cfe4b4 7307 ext4_exit_post_read_processing();
1dc0aa46 7308out6:
22cfe4b4
EB		 7309 ext4_exit_pending();
7310out7:
51865fda
ZL		 7311 ext4_exit_es();
7312
ac27a0ec
DK		 7313 return err;
7314}
7315
5dabfc78 7316static void __exit ext4_exit_fs(void)
ac27a0ec 7317{
bfff6873 7318 ext4_destroy_lazyinit_thread();
24b58424
TT		 7319 unregister_as_ext2();
7320 unregister_as_ext3();
03010a33 7321 unregister_filesystem(&ext4_fs_type);
ab047d51 7322 ext4_fc_destroy_dentry_cache();
ac27a0ec 7323 destroy_inodecache();
5dabfc78 7324 ext4_exit_mballoc();
b5799018 7325 ext4_exit_sysfs();
5dabfc78
TT		 7326 ext4_exit_system_zone();
7327 ext4_exit_pageio();
22cfe4b4 7328 ext4_exit_post_read_processing();
dd12ed14 7329 ext4_exit_es();
1dc0aa46 7330 ext4_exit_pending();
ac27a0ec
DK		 7331}
7332
7333MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 7334MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 7335MODULE_LICENSE("GPL");
7ef79ad5 7336MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 7337module_init(ext4_init_fs)
7338module_exit(ext4_exit_fs)
Linux 6.x block layer and io_uring tree(s)