]> git.ipfire.org Git - thirdparty/kernel/stable.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / stable.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: fix race between writepages and remount
[thirdparty/kernel/stable.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
5354b2af
TT		 1051 if (!grp || !gdp)
1052 return;
9af0b3d1
WS		 1053 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1054 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1055 &grp->bb_state);
1056 if (!ret)
1057 percpu_counter_sub(&sbi->s_freeclusters_counter,
1058 grp->bb_free);
db79e6d1
WS		 1059 }
1060
9af0b3d1
WS		 1061 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1062 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1063 &grp->bb_state);
1064 if (!ret && gdp) {
db79e6d1
WS		 1065 int count;
1066
1067 count = ext4_free_inodes_count(sb, gdp);
1068 percpu_counter_sub(&sbi->s_freeinodes_counter,
1069 count);
1070 }
db79e6d1
WS		 1071 }
1072}
1073
617ba13b 1074void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 1075{
617ba13b 1076 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 1077
617ba13b 1078 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 1079 return;
1080
12062ddd 1081 ext4_warning(sb,
ac27a0ec
DK		 1082 "updating to rev %d because of new feature flag, "
1083 "running e2fsck is recommended",
617ba13b 1084 EXT4_DYNAMIC_REV);
ac27a0ec 1085
617ba13b
MC		 1086 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1087 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1088 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 1089 /* leave es->s_feature_*compat flags alone */
1090 /* es->s_uuid will be set by e2fsck if empty */
1091
1092 /*
1093 * The rest of the superblock fields should be zero, and if not it
1094 * means they are likely already in use, so leave them alone. We
1095 * can leave it up to e2fsck to clean up any inconsistencies there.
1096 */
1097}
1098
dd2e31af
CH		 1099static void ext4_bdev_mark_dead(struct block_device *bdev)
1100{
1101 ext4_force_shutdown(bdev->bd_holder, EXT4_GOING_FLAGS_NOLOGFLUSH);
1102}
1103
1104static const struct blk_holder_ops ext4_holder_ops = {
1105 .mark_dead = ext4_bdev_mark_dead,
1106};
1107
ac27a0ec
DK		 1108/*
1109 * Open the external journal device
1110 */
b31e1552 1111static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 1112{
1113 struct block_device *bdev;
ac27a0ec 1114
05bdb996 1115 bdev = blkdev_get_by_dev(dev, BLK_OPEN_READ | BLK_OPEN_WRITE, sb,
dd2e31af 1116 &ext4_holder_ops);
ac27a0ec
DK		 1117 if (IS_ERR(bdev))
1118 goto fail;
1119 return bdev;
1120
1121fail:
ea3edd4d
CH		 1122 ext4_msg(sb, KERN_ERR,
1123 "failed to open journal device unknown-block(%u,%u) %ld",
1124 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
ac27a0ec
DK		 1125 return NULL;
1126}
1127
1128/*
1129 * Release the journal device
1130 */
4385bab1 1131static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 1132{
1133 struct block_device *bdev;
ee7ed3aa 1134 bdev = sbi->s_journal_bdev;
ac27a0ec 1135 if (bdev) {
26fb5290
ZC		 1136 /*
1137 * Invalidate the journal device's buffers. We don't want them
1138 * floating about in memory - the physical journal device may
1139 * hotswapped, and it breaks the `ro-after' testing code.
1140 */
1141 invalidate_bdev(bdev);
a42fb5a7 1142 blkdev_put(bdev, sbi->s_sb);
ee7ed3aa 1143 sbi->s_journal_bdev = NULL;
ac27a0ec 1144 }
ac27a0ec
DK		 1145}
1146
1147static inline struct inode *orphan_list_entry(struct list_head *l)
1148{
617ba13b 1149 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 1150}
1151
617ba13b 1152static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 1153{
1154 struct list_head *l;
1155
b31e1552
ES		 1156 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1157 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 1158
1159 printk(KERN_ERR "sb_info orphan list:\n");
1160 list_for_each(l, &sbi->s_orphan) {
1161 struct inode *inode = orphan_list_entry(l);
1162 printk(KERN_ERR " "
1163 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1164 inode->i_sb->s_id, inode->i_ino, inode,
1165 inode->i_mode, inode->i_nlink,
1166 NEXT_ORPHAN(inode));
1167 }
1168}
1169
957153fc
JK		 1170#ifdef CONFIG_QUOTA
1171static int ext4_quota_off(struct super_block *sb, int type);
1172
f3c1c42e 1173static inline void ext4_quotas_off(struct super_block *sb, int type)
957153fc 1174{
f3c1c42e 1175 BUG_ON(type > EXT4_MAXQUOTAS);
957153fc 1176
964edf66 1177 /* Use our quota_off function to clear inode flags etc. */
f3c1c42e 1178 for (type--; type >= 0; type--)
964edf66 1179 ext4_quota_off(sb, type);
957153fc 1180}
33458eab
TT		 1181
1182/*
1183 * This is a helper function which is used in the mount/remount
1184 * codepaths (which holds s_umount) to fetch the quota file name.
1185 */
1186static inline char *get_qf_name(struct super_block *sb,
1187 struct ext4_sb_info *sbi,
1188 int type)
1189{
1190 return rcu_dereference_protected(sbi->s_qf_names[type],
1191 lockdep_is_held(&sb->s_umount));
1192}
957153fc 1193#else
f3c1c42e 1194static inline void ext4_quotas_off(struct super_block *sb, int type)
957153fc
JK		 1195{
1196}
1197#endif
1198
1f79467c
JY		 1199static int ext4_percpu_param_init(struct ext4_sb_info *sbi)
1200{
1201 ext4_fsblk_t block;
1202 int err;
1203
1204 block = ext4_count_free_clusters(sbi->s_sb);
1205 ext4_free_blocks_count_set(sbi->s_es, EXT4_C2B(sbi, block));
1206 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
1207 GFP_KERNEL);
1208 if (!err) {
1209 unsigned long freei = ext4_count_free_inodes(sbi->s_sb);
1210 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
1211 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
1212 GFP_KERNEL);
1213 }
1214 if (!err)
1215 err = percpu_counter_init(&sbi->s_dirs_counter,
1216 ext4_count_dirs(sbi->s_sb), GFP_KERNEL);
1217 if (!err)
1218 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
1219 GFP_KERNEL);
1220 if (!err)
1221 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
1222 GFP_KERNEL);
1223 if (!err)
1224 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
1225
1226 if (err)
1227 ext4_msg(sbi->s_sb, KERN_ERR, "insufficient memory");
1228
1229 return err;
1230}
1231
1232static void ext4_percpu_param_destroy(struct ext4_sb_info *sbi)
1233{
1234 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1235 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1236 percpu_counter_destroy(&sbi->s_dirs_counter);
1237 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1238 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1239 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1240}
1241
6ef68498
JY		 1242static void ext4_group_desc_free(struct ext4_sb_info *sbi)
1243{
1244 struct buffer_head **group_desc;
1245 int i;
1246
1247 rcu_read_lock();
1248 group_desc = rcu_dereference(sbi->s_group_desc);
1249 for (i = 0; i < sbi->s_gdb_count; i++)
1250 brelse(group_desc[i]);
1251 kvfree(group_desc);
1252 rcu_read_unlock();
1253}
1254
dcbf8758
JY		 1255static void ext4_flex_groups_free(struct ext4_sb_info *sbi)
1256{
1257 struct flex_groups **flex_groups;
1258 int i;
1259
1260 rcu_read_lock();
1261 flex_groups = rcu_dereference(sbi->s_flex_groups);
1262 if (flex_groups) {
1263 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1264 kvfree(flex_groups[i]);
1265 kvfree(flex_groups);
1266 }
1267 rcu_read_unlock();
1268}
1269
2b2d6d01 1270static void ext4_put_super(struct super_block *sb)
ac27a0ec 1271{
617ba13b
MC		 1272 struct ext4_sb_info *sbi = EXT4_SB(sb);
1273 struct ext4_super_block *es = sbi->s_es;
97abd7d4 1274 int aborted = 0;
856dd6c5 1275 int err;
ac27a0ec 1276
5e47868f
RH		 1277 /*
1278 * Unregister sysfs before destroying jbd2 journal.
1279 * Since we could still access attr_journal_task attribute via sysfs
1280 * path which could have sbi->s_journal->j_task as NULL
b98535d0
YB		 1281 * Unregister sysfs before flush sbi->s_error_work.
1282 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1283 * read metadata verify failed then will queue error work.
1284 * flush_stashed_error_work will call start_this_handle may trigger
1285 * BUG_ON.
5e47868f
RH		 1286 */
1287 ext4_unregister_sysfs(sb);
1288
4808cb5b 1289 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
bb0fbc78
LC		 1290 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1291 &sb->s_uuid);
4808cb5b 1292
b98535d0 1293 ext4_unregister_li_request(sb);
f3c1c42e 1294 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
b98535d0
YB		 1295
1296 flush_work(&sbi->s_error_work);
1297 destroy_workqueue(sbi->rsv_conversion_wq);
1298 ext4_release_orphan_info(sb);
1299
0390131b 1300 if (sbi->s_journal) {
97abd7d4 1301 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 1302 err = jbd2_journal_destroy(sbi->s_journal);
1303 sbi->s_journal = NULL;
878520ac 1304 if ((err < 0) && !aborted) {
54d3adbc 1305 ext4_abort(sb, -err, "Couldn't clean up the journal");
878520ac 1306 }
0390131b 1307 }
d4edac31 1308
d3922a77 1309 ext4_es_unregister_shrinker(sbi);
292a089d 1310 timer_shutdown_sync(&sbi->s_err_report);
d4edac31
JB		 1311 ext4_release_system_zone(sb);
1312 ext4_mb_release(sb);
1313 ext4_ext_release(sb);
d4edac31 1314
bc98a42c 1315 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 1316 ext4_clear_feature_journal_needs_recovery(sb);
02f310fc 1317 ext4_clear_feature_orphan_present(sb);
ac27a0ec 1318 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 1319 }
bc98a42c 1320 if (!sb_rdonly(sb))
4392fbc4 1321 ext4_commit_super(sb);
a8e25a83 1322
6ef68498 1323 ext4_group_desc_free(sbi);
dcbf8758 1324 ext4_flex_groups_free(sbi);
1f79467c 1325 ext4_percpu_param_destroy(sbi);
ac27a0ec 1326#ifdef CONFIG_QUOTA
856dd6c5 1327 for (int i = 0; i < EXT4_MAXQUOTAS; i++)
33458eab 1328 kfree(get_qf_name(sb, sbi, i));
ac27a0ec
DK		 1329#endif
1330
1331 /* Debugging code just in case the in-memory inode orphan list
1332 * isn't empty. The on-disk one can be non-empty if we've
1333 * detected an error and taken the fs readonly, but the
1334 * in-memory list had better be clean by this point. */
1335 if (!list_empty(&sbi->s_orphan))
1336 dump_orphan_list(sb, sbi);
837c23fb 1337 ASSERT(list_empty(&sbi->s_orphan));
ac27a0ec 1338
89d96a6f 1339 sync_blockdev(sb->s_bdev);
f98393a6 1340 invalidate_bdev(sb->s_bdev);
93e92cfc 1341 if (sbi->s_journal_bdev) {
ee7ed3aa 1342 sync_blockdev(sbi->s_journal_bdev);
617ba13b 1343 ext4_blkdev_remove(sbi);
ac27a0ec 1344 }
50c15df6
CX		 1345
1346 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1347 sbi->s_ea_inode_cache = NULL;
1348
1349 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1350 sbi->s_ea_block_cache = NULL;
1351
618f0031
PS		 1352 ext4_stop_mmpd(sbi);
1353
9060dd2c 1354 brelse(sbi->s_sbh);
ac27a0ec 1355 sb->s_fs_info = NULL;
3197ebdb
TT		 1356 /*
1357 * Now that we are completely done shutting down the
1358 * superblock, we need to actually destroy the kobject.
1359 */
3197ebdb
TT		 1360 kobject_put(&sbi->s_kobj);
1361 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 1362 if (sbi->s_chksum_driver)
1363 crypto_free_shash(sbi->s_chksum_driver);
705895b6 1364 kfree(sbi->s_blockgroup_lock);
8012b866 1365 fs_put_dax(sbi->s_daxdev, NULL);
ac4acb1f 1366 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5298d4bf 1367#if IS_ENABLED(CONFIG_UNICODE)
f8f4acb6 1368 utf8_unload(sb->s_encoding);
c83ad55e 1369#endif
ac27a0ec 1370 kfree(sbi);
ac27a0ec
DK		 1371}
1372
e18b890b 1373static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 1374
1375/*
1376 * Called inside transaction, so use GFP_NOFS
1377 */
617ba13b 1378static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 1379{
617ba13b 1380 struct ext4_inode_info *ei;
ac27a0ec 1381
fd60b288 1382 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 1383 if (!ei)
1384 return NULL;
0b8e58a1 1385
ee73f9a5 1386 inode_set_iversion(&ei->vfs_inode, 1);
7ea71af9 1387 ei->i_flags = 0;
202ee5df 1388 spin_lock_init(&ei->i_raw_lock);
38727786 1389 ei->i_prealloc_node = RB_ROOT;
27bc446e 1390 atomic_set(&ei->i_prealloc_active, 0);
38727786 1391 rwlock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 1392 ext4_es_init_tree(&ei->i_es_tree);
1393 rwlock_init(&ei->i_es_lock);
edaa53ca 1394 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 1395 ei->i_es_all_nr = 0;
edaa53ca 1396 ei->i_es_shk_nr = 0;
dd475925 1397 ei->i_es_shrink_lblk = 0;
d2a17637 1398 ei->i_reserved_data_blocks = 0;
d2a17637 1399 spin_lock_init(&(ei->i_block_reservation_lock));
1dc0aa46 1400 ext4_init_pending_tree(&ei->i_pending_tree);
a9e7f447
DM		 1401#ifdef CONFIG_QUOTA
1402 ei->i_reserved_quota = 0;
96c7e0d9 1403 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 1404#endif
8aefcd55 1405 ei->jinode = NULL;
2e8fa54e 1406 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 1407 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 1408 ei->i_sync_tid = 0;
1409 ei->i_datasync_tid = 0;
e27f41e1 1410 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 1411 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
aa75f4d3
HS		 1412 ext4_fc_init_inode(&ei->vfs_inode);
1413 mutex_init(&ei->i_fc_lock);
ac27a0ec
DK		 1414 return &ei->vfs_inode;
1415}
1416
7ff9c073
TT		 1417static int ext4_drop_inode(struct inode *inode)
1418{
1419 int drop = generic_drop_inode(inode);
1420
29b3692e
EB		 1421 if (!drop)
1422 drop = fscrypt_drop_inode(inode);
1423
7ff9c073
TT		 1424 trace_ext4_drop_inode(inode, drop);
1425 return drop;
1426}
1427
94053139 1428static void ext4_free_in_core_inode(struct inode *inode)
fa0d7e3d 1429{
2c58d548 1430 fscrypt_free_inode(inode);
aa75f4d3
HS		 1431 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1432 pr_warn("%s: inode %ld still in fc list",
1433 __func__, inode->i_ino);
1434 }
fa0d7e3d
NP		 1435 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1436}
1437
617ba13b 1438static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1439{
9f7dd93d 1440 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1441 ext4_msg(inode->i_sb, KERN_ERR,
1442 "Inode %lu (%p): orphan list check failed!",
1443 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1444 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1445 EXT4_I(inode), sizeof(struct ext4_inode_info),
1446 true);
1447 dump_stack();
1448 }
6fed8395
JX		 1449
1450 if (EXT4_I(inode)->i_reserved_data_blocks)
1451 ext4_msg(inode->i_sb, KERN_ERR,
1452 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1453 inode->i_ino, EXT4_I(inode),
1454 EXT4_I(inode)->i_reserved_data_blocks);
ac27a0ec
DK		 1455}
1456
f5db130d
CH		 1457static void ext4_shutdown(struct super_block *sb)
1458{
1459 ext4_force_shutdown(sb, EXT4_GOING_FLAGS_NOLOGFLUSH);
1460}
1461
51cc5068 1462static void init_once(void *foo)
ac27a0ec 1463{
c30365b9 1464 struct ext4_inode_info *ei = foo;
ac27a0ec 1465
a35afb83 1466 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1467 init_rwsem(&ei->xattr_sem);
0e855ac8 1468 init_rwsem(&ei->i_data_sem);
a35afb83 1469 inode_init_once(&ei->vfs_inode);
aa75f4d3 1470 ext4_fc_init_inode(&ei->vfs_inode);
ac27a0ec
DK		 1471}
1472
e67bc2b3 1473static int __init init_inodecache(void)
ac27a0ec 1474{
f8dd7c70
DW		 1475 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1476 sizeof(struct ext4_inode_info), 0,
1477 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1478 SLAB_ACCOUNT),
1479 offsetof(struct ext4_inode_info, i_data),
1480 sizeof_field(struct ext4_inode_info, i_data),
1481 init_once);
617ba13b 1482 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1483 return -ENOMEM;
1484 return 0;
1485}
1486
1487static void destroy_inodecache(void)
1488{
8c0a8537
KS		 1489 /*
1490 * Make sure all delayed rcu free inodes are flushed before we
1491 * destroy cache.
1492 */
1493 rcu_barrier();
617ba13b 1494 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1495}
1496
0930fcc1 1497void ext4_clear_inode(struct inode *inode)
ac27a0ec 1498{
aa75f4d3 1499 ext4_fc_del(inode);
0930fcc1 1500 invalidate_inode_buffers(inode);
dbd5768f 1501 clear_inode(inode);
27bc446e 1502 ext4_discard_preallocations(inode, 0);
51865fda 1503 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
f4c2d372 1504 dquot_drop(inode);
8aefcd55
TT		 1505 if (EXT4_I(inode)->jinode) {
1506 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1507 EXT4_I(inode)->jinode);
1508 jbd2_free_inode(EXT4_I(inode)->jinode);
1509 EXT4_I(inode)->jinode = NULL;
1510 }
3d204e24 1511 fscrypt_put_encryption_info(inode);
c93d8f88 1512 fsverity_cleanup_inode(inode);
ac27a0ec
DK		 1513}
1514
1b961ac0 1515static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1516 u64 ino, u32 generation)
ac27a0ec 1517{
ac27a0ec 1518 struct inode *inode;
ac27a0ec 1519
8a363970 1520 /*
ac27a0ec
DK		 1521 * Currently we don't know the generation for parent directory, so
1522 * a generation of 0 means "accept any"
1523 */
8a363970 1524 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1d1fe1ee
DH		 1525 if (IS_ERR(inode))
1526 return ERR_CAST(inode);
1527 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1528 iput(inode);
1529 return ERR_PTR(-ESTALE);
1530 }
1b961ac0
CH		 1531
1532 return inode;
1533}
1534
1535static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1536 int fh_len, int fh_type)
1b961ac0
CH		 1537{
1538 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1539 ext4_nfs_get_inode);
1540}
1541
1542static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1543 int fh_len, int fh_type)
1b961ac0
CH		 1544{
1545 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1546 ext4_nfs_get_inode);
ac27a0ec
DK		 1547}
1548
fde87268
TT		 1549static int ext4_nfs_commit_metadata(struct inode *inode)
1550{
1551 struct writeback_control wbc = {
1552 .sync_mode = WB_SYNC_ALL
1553 };
1554
1555 trace_ext4_nfs_commit_metadata(inode);
1556 return ext4_write_inode(inode, &wbc);
1557}
1558
ac27a0ec 1559#ifdef CONFIG_QUOTA
d6006186 1560static const char * const quotatypes[] = INITQFNAMES;
689c958c 1561#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1562
617ba13b
MC		 1563static int ext4_write_dquot(struct dquot *dquot);
1564static int ext4_acquire_dquot(struct dquot *dquot);
1565static int ext4_release_dquot(struct dquot *dquot);
1566static int ext4_mark_dquot_dirty(struct dquot *dquot);
1567static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1568static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1569 const struct path *path);
617ba13b 1570static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1571 size_t len, loff_t off);
617ba13b 1572static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1573 const char *data, size_t len, loff_t off);
7c319d32
AK		 1574static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1575 unsigned int flags);
ac27a0ec 1576
96c7e0d9
JK		 1577static struct dquot **ext4_get_dquots(struct inode *inode)
1578{
1579 return EXT4_I(inode)->i_dquot;
1580}
1581
61e225dc 1582static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1583 .get_reserved_space = ext4_get_reserved_space,
1584 .write_dquot = ext4_write_dquot,
1585 .acquire_dquot = ext4_acquire_dquot,
1586 .release_dquot = ext4_release_dquot,
1587 .mark_dirty = ext4_mark_dquot_dirty,
1588 .write_info = ext4_write_info,
1589 .alloc_dquot = dquot_alloc,
1590 .destroy_dquot = dquot_destroy,
1591 .get_projid = ext4_get_projid,
1592 .get_inode_usage = ext4_get_inode_usage,
ebc11f7b 1593 .get_next_id = dquot_get_next_id,
ac27a0ec
DK		 1594};
1595
0d54b217 1596static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1597 .quota_on = ext4_quota_on,
ca0e05e4 1598 .quota_off = ext4_quota_off,
287a8095 1599 .quota_sync = dquot_quota_sync,
0a240339 1600 .get_state = dquot_get_state,
287a8095
CH		 1601 .set_info = dquot_set_dqinfo,
1602 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1603 .set_dqblk = dquot_set_dqblk,
1604 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1605};
1606#endif
1607
ee9b6d61 1608static const struct super_operations ext4_sops = {
617ba13b 1609 .alloc_inode = ext4_alloc_inode,
94053139 1610 .free_inode = ext4_free_in_core_inode,
617ba13b 1611 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1612 .write_inode = ext4_write_inode,
1613 .dirty_inode = ext4_dirty_inode,
7ff9c073 1614 .drop_inode = ext4_drop_inode,
0930fcc1 1615 .evict_inode = ext4_evict_inode,
617ba13b 1616 .put_super = ext4_put_super,
617ba13b 1617 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1618 .freeze_fs = ext4_freeze,
1619 .unfreeze_fs = ext4_unfreeze,
617ba13b 1620 .statfs = ext4_statfs,
617ba13b 1621 .show_options = ext4_show_options,
f5db130d 1622 .shutdown = ext4_shutdown,
ac27a0ec 1623#ifdef CONFIG_QUOTA
617ba13b
MC		 1624 .quota_read = ext4_quota_read,
1625 .quota_write = ext4_quota_write,
96c7e0d9 1626 .get_dquots = ext4_get_dquots,
ac27a0ec
DK		 1627#endif
1628};
1629
39655164 1630static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1631 .fh_to_dentry = ext4_fh_to_dentry,
1632 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1633 .get_parent = ext4_get_parent,
fde87268 1634 .commit_metadata = ext4_nfs_commit_metadata,
ac27a0ec
DK		 1635};
1636
1637enum {
1638 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
ba2e524d 1639 Opt_resgid, Opt_resuid, Opt_sb,
72578c33 1640 Opt_nouid32, Opt_debug, Opt_removed,
2d544ec9 1641 Opt_user_xattr, Opt_acl,
72578c33 1642 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1643 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1644 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1645 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1646 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
4f74d15f 1647 Opt_inlinecrypt,
ba2e524d 1648 Opt_usrjquota, Opt_grpjquota, Opt_quota,
ee4a3fcd 1649 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1ff20307 1650 Opt_usrquota, Opt_grpquota, Opt_prjquota,
9cb20f94 1651 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
327eaf73 1652 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
4437992b 1653 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1449032b 1654 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1655 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1656 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1657 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1658 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
21175ca4 1659 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
e5a185c2 1660 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
8016e29f 1661#ifdef CONFIG_EXT4_DEBUG
99c880de 1662 Opt_fc_debug_max_replay, Opt_fc_debug_force
8016e29f 1663#endif
ac27a0ec
DK		 1664};
1665
e5a185c2 1666static const struct constant_table ext4_param_errors[] = {
ba2e524d
LC		 1667 {"continue", EXT4_MOUNT_ERRORS_CONT},
1668 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1669 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
e5a185c2
LC		 1670 {}
1671};
1672
1673static const struct constant_table ext4_param_data[] = {
ba2e524d
LC		 1674 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1675 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1676 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
e5a185c2
LC		 1677 {}
1678};
1679
1680static const struct constant_table ext4_param_data_err[] = {
1681 {"abort", Opt_data_err_abort},
1682 {"ignore", Opt_data_err_ignore},
1683 {}
1684};
1685
1686static const struct constant_table ext4_param_jqfmt[] = {
ba2e524d
LC		 1687 {"vfsold", QFMT_VFS_OLD},
1688 {"vfsv0", QFMT_VFS_V0},
1689 {"vfsv1", QFMT_VFS_V1},
e5a185c2
LC		 1690 {}
1691};
1692
1693static const struct constant_table ext4_param_dax[] = {
1694 {"always", Opt_dax_always},
1695 {"inode", Opt_dax_inode},
1696 {"never", Opt_dax_never},
1697 {}
1698};
1699
1700/* String parameter that allows empty argument */
1701#define fsparam_string_empty(NAME, OPT) \
1702 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1703
1704/*
1705 * Mount option specification
1706 * We don't use fsparam_flag_no because of the way we set the
1707 * options and the way we show them in _ext4_show_options(). To
1708 * keep the changes to a minimum, let's keep the negative options
1709 * separate for now.
1710 */
1711static const struct fs_parameter_spec ext4_param_specs[] = {
1712 fsparam_flag ("bsddf", Opt_bsd_df),
1713 fsparam_flag ("minixdf", Opt_minix_df),
1714 fsparam_flag ("grpid", Opt_grpid),
1715 fsparam_flag ("bsdgroups", Opt_grpid),
1716 fsparam_flag ("nogrpid", Opt_nogrpid),
1717 fsparam_flag ("sysvgroups", Opt_nogrpid),
1718 fsparam_u32 ("resgid", Opt_resgid),
1719 fsparam_u32 ("resuid", Opt_resuid),
1720 fsparam_u32 ("sb", Opt_sb),
1721 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1722 fsparam_flag ("nouid32", Opt_nouid32),
1723 fsparam_flag ("debug", Opt_debug),
1724 fsparam_flag ("oldalloc", Opt_removed),
1725 fsparam_flag ("orlov", Opt_removed),
1726 fsparam_flag ("user_xattr", Opt_user_xattr),
e5a185c2 1727 fsparam_flag ("acl", Opt_acl),
e5a185c2
LC		 1728 fsparam_flag ("norecovery", Opt_noload),
1729 fsparam_flag ("noload", Opt_noload),
1730 fsparam_flag ("bh", Opt_removed),
1731 fsparam_flag ("nobh", Opt_removed),
1732 fsparam_u32 ("commit", Opt_commit),
1733 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1734 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1735 fsparam_u32 ("journal_dev", Opt_journal_dev),
1736 fsparam_bdev ("journal_path", Opt_journal_path),
1737 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1738 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1739 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1740 fsparam_flag ("abort", Opt_abort),
1741 fsparam_enum ("data", Opt_data, ext4_param_data),
1742 fsparam_enum ("data_err", Opt_data_err,
1743 ext4_param_data_err),
1744 fsparam_string_empty
1745 ("usrjquota", Opt_usrjquota),
1746 fsparam_string_empty
1747 ("grpjquota", Opt_grpjquota),
1748 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1749 fsparam_flag ("grpquota", Opt_grpquota),
1750 fsparam_flag ("quota", Opt_quota),
1751 fsparam_flag ("noquota", Opt_noquota),
1752 fsparam_flag ("usrquota", Opt_usrquota),
1753 fsparam_flag ("prjquota", Opt_prjquota),
1754 fsparam_flag ("barrier", Opt_barrier),
1755 fsparam_u32 ("barrier", Opt_barrier),
1756 fsparam_flag ("nobarrier", Opt_nobarrier),
1ff20307 1757 fsparam_flag ("i_version", Opt_removed),
e5a185c2
LC		 1758 fsparam_flag ("dax", Opt_dax),
1759 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1760 fsparam_u32 ("stripe", Opt_stripe),
1761 fsparam_flag ("delalloc", Opt_delalloc),
1762 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1763 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1764 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
e5a185c2
LC		 1765 fsparam_u32 ("debug_want_extra_isize",
1766 Opt_debug_want_extra_isize),
1767 fsparam_flag ("mblk_io_submit", Opt_removed),
1768 fsparam_flag ("nomblk_io_submit", Opt_removed),
1769 fsparam_flag ("block_validity", Opt_block_validity),
1770 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1771 fsparam_u32 ("inode_readahead_blks",
1772 Opt_inode_readahead_blks),
1773 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1774 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1775 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1776 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1777 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1778 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1779 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1780 fsparam_flag ("discard", Opt_discard),
1781 fsparam_flag ("nodiscard", Opt_nodiscard),
1782 fsparam_u32 ("init_itable", Opt_init_itable),
1783 fsparam_flag ("init_itable", Opt_init_itable),
1784 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1785#ifdef CONFIG_EXT4_DEBUG
1786 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1787 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1788#endif
1789 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1790 fsparam_flag ("test_dummy_encryption",
1791 Opt_test_dummy_encryption),
1792 fsparam_string ("test_dummy_encryption",
1793 Opt_test_dummy_encryption),
1794 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1795 fsparam_flag ("nombcache", Opt_nombcache),
1796 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1797 fsparam_flag ("prefetch_block_bitmaps",
1798 Opt_removed),
1799 fsparam_flag ("no_prefetch_block_bitmaps",
1800 Opt_no_prefetch_block_bitmaps),
1801 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1802 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1803 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1804 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1805 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1806 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1807 {}
1808};
1809
b3881f74 1810#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
196e402a 1811
26092bf5
TT		 1812#define MOPT_SET 0x0001
1813#define MOPT_CLEAR 0x0002
1814#define MOPT_NOSUPPORT 0x0004
1815#define MOPT_EXPLICIT 0x0008
ac27a0ec 1816#ifdef CONFIG_QUOTA
26092bf5 1817#define MOPT_Q 0
ba2e524d 1818#define MOPT_QFMT 0x0010
26092bf5
TT		 1819#else
1820#define MOPT_Q MOPT_NOSUPPORT
1821#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1822#endif
ba2e524d
LC		 1823#define MOPT_NO_EXT2 0x0020
1824#define MOPT_NO_EXT3 0x0040
8dc0aa8c 1825#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ba2e524d
LC		 1826#define MOPT_SKIP 0x0080
1827#define MOPT_2 0x0100
26092bf5
TT		 1828
1829static const struct mount_opts {
1830 int token;
1831 int mount_opt;
1832 int flags;
1833} ext4_mount_opts[] = {
1834 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1835 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1836 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1837 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1838 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1839 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1840 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1841 MOPT_EXT4_ONLY | MOPT_SET},
1842 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1843 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1844 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1845 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1846 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1847 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1848 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1849 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1850 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1851 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
cb8435dc 1852 {Opt_commit, 0, MOPT_NO_EXT2},
c6d3d56d
DW		 1853 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1854 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1855 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1856 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1857 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1858 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1859 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1860 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
ba2e524d 1861 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
26092bf5
TT		 1862 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1863 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1864 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1865 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1866 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
ba2e524d
LC		 1867 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1868 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1869 {Opt_journal_path, 0, MOPT_NO_EXT2},
1870 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1871 {Opt_data, 0, MOPT_NO_EXT2},
26092bf5 1872 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
03010a33 1873#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5 1874 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
ac27a0ec 1875#else
26092bf5 1876 {Opt_acl, 0, MOPT_NOSUPPORT},
ac27a0ec 1877#endif
26092bf5
TT		 1878 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1879 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1880 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1881 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1882 MOPT_SET | MOPT_Q},
1883 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1884 MOPT_SET | MOPT_Q},
49da9392
JK		 1885 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1886 MOPT_SET | MOPT_Q},
26092bf5 1887 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1888 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1889 MOPT_CLEAR | MOPT_Q},
ba2e524d
LC		 1890 {Opt_usrjquota, 0, MOPT_Q},
1891 {Opt_grpjquota, 0, MOPT_Q},
1892 {Opt_jqfmt, 0, MOPT_QFMT},
cdb7ee4c 1893 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
21175ca4 1894 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
3d392b26 1895 MOPT_SET},
99c880de 1896#ifdef CONFIG_EXT4_DEBUG
0f0672ff
HS		 1897 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1898 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
8016e29f 1899#endif
26092bf5
TT		 1900 {Opt_err, 0, 0}
1901};
1902
5298d4bf 1903#if IS_ENABLED(CONFIG_UNICODE)
c83ad55e
GKB		 1904static const struct ext4_sb_encodings {
1905 __u16 magic;
1906 char *name;
49bd03cc 1907 unsigned int version;
c83ad55e 1908} ext4_sb_encoding_map[] = {
49bd03cc 1909 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
c83ad55e
GKB		 1910};
1911
aa8bf298
CH		 1912static const struct ext4_sb_encodings *
1913ext4_sb_read_encoding(const struct ext4_super_block *es)
c83ad55e
GKB		 1914{
1915 __u16 magic = le16_to_cpu(es->s_encoding);
1916 int i;
1917
1918 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1919 if (magic == ext4_sb_encoding_map[i].magic)
aa8bf298 1920 return &ext4_sb_encoding_map[i];
c83ad55e 1921
aa8bf298 1922 return NULL;
c83ad55e
GKB		 1923}
1924#endif
1925
6e47a3cc
LC		 1926#define EXT4_SPEC_JQUOTA (1 << 0)
1927#define EXT4_SPEC_JQFMT (1 << 1)
1928#define EXT4_SPEC_DATAJ (1 << 2)
1929#define EXT4_SPEC_SB_BLOCK (1 << 3)
1930#define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1931#define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
6e47a3cc
LC		 1932#define EXT4_SPEC_s_want_extra_isize (1 << 7)
1933#define EXT4_SPEC_s_max_batch_time (1 << 8)
1934#define EXT4_SPEC_s_min_batch_time (1 << 9)
1935#define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1936#define EXT4_SPEC_s_li_wait_mult (1 << 11)
1937#define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1938#define EXT4_SPEC_s_stripe (1 << 13)
1939#define EXT4_SPEC_s_resuid (1 << 14)
1940#define EXT4_SPEC_s_resgid (1 << 15)
1941#define EXT4_SPEC_s_commit_interval (1 << 16)
1942#define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
7edfd85b 1943#define EXT4_SPEC_s_sb_block (1 << 18)
27b38686 1944#define EXT4_SPEC_mb_optimize_scan (1 << 19)
6e47a3cc 1945
461c3af0 1946struct ext4_fs_context {
e6e268cb 1947 char *s_qf_names[EXT4_MAXQUOTAS];
85456054 1948 struct fscrypt_dummy_policy dummy_enc_policy;
e6e268cb 1949 int s_jquota_fmt; /* Format of quota to use */
6e47a3cc
LC		 1950#ifdef CONFIG_EXT4_DEBUG
1951 int s_fc_debug_max_replay;
1952#endif
e6e268cb 1953 unsigned short qname_spec;
6e47a3cc
LC		 1954 unsigned long vals_s_flags; /* Bits to set in s_flags */
1955 unsigned long mask_s_flags; /* Bits changed in s_flags */
e6e268cb 1956 unsigned long journal_devnum;
6e47a3cc
LC		 1957 unsigned long s_commit_interval;
1958 unsigned long s_stripe;
1959 unsigned int s_inode_readahead_blks;
1960 unsigned int s_want_extra_isize;
1961 unsigned int s_li_wait_mult;
1962 unsigned int s_max_dir_size_kb;
e6e268cb 1963 unsigned int journal_ioprio;
6e47a3cc
LC		 1964 unsigned int vals_s_mount_opt;
1965 unsigned int mask_s_mount_opt;
1966 unsigned int vals_s_mount_opt2;
1967 unsigned int mask_s_mount_opt2;
e3952fcc
LC		 1968 unsigned long vals_s_mount_flags;
1969 unsigned long mask_s_mount_flags;
b6bd2435 1970 unsigned int opt_flags; /* MOPT flags */
6e47a3cc
LC		 1971 unsigned int spec;
1972 u32 s_max_batch_time;
1973 u32 s_min_batch_time;
1974 kuid_t s_resuid;
1975 kgid_t s_resgid;
7edfd85b 1976 ext4_fsblk_t s_sb_block;
b237e304
HS		 1977};
1978
cebe85d5
LC		 1979static void ext4_fc_free(struct fs_context *fc)
1980{
1981 struct ext4_fs_context *ctx = fc->fs_private;
1982 int i;
1983
1984 if (!ctx)
1985 return;
1986
1987 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1988 kfree(ctx->s_qf_names[i]);
1989
85456054 1990 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
cebe85d5
LC		 1991 kfree(ctx);
1992}
1993
1994int ext4_init_fs_context(struct fs_context *fc)
1995{
da9e4802 1996 struct ext4_fs_context *ctx;
cebe85d5
LC		 1997
1998 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
1999 if (!ctx)
2000 return -ENOMEM;
2001
2002 fc->fs_private = ctx;
2003 fc->ops = &ext4_context_ops;
2004
2005 return 0;
2006}
2007
e6e268cb
LC		 2008#ifdef CONFIG_QUOTA
2009/*
2010 * Note the name of the specified quota file.
2011 */
2012static int note_qf_name(struct fs_context *fc, int qtype,
2013 struct fs_parameter *param)
2014{
2015 struct ext4_fs_context *ctx = fc->fs_private;
2016 char *qname;
2017
2018 if (param->size < 1) {
2019 ext4_msg(NULL, KERN_ERR, "Missing quota name");
2020 return -EINVAL;
2021 }
2022 if (strchr(param->string, '/')) {
2023 ext4_msg(NULL, KERN_ERR,
2024 "quotafile must be on filesystem root");
2025 return -EINVAL;
2026 }
2027 if (ctx->s_qf_names[qtype]) {
2028 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
2029 ext4_msg(NULL, KERN_ERR,
2030 "%s quota file already specified",
2031 QTYPE2NAME(qtype));
2032 return -EINVAL;
2033 }
2034 return 0;
2035 }
2036
2037 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
2038 if (!qname) {
2039 ext4_msg(NULL, KERN_ERR,
2040 "Not enough memory for storing quotafile name");
2041 return -ENOMEM;
2042 }
2043 ctx->s_qf_names[qtype] = qname;
2044 ctx->qname_spec |= 1 << qtype;
6e47a3cc 2045 ctx->spec |= EXT4_SPEC_JQUOTA;
e6e268cb
LC		 2046 return 0;
2047}
2048
2049/*
2050 * Clear the name of the specified quota file.
2051 */
2052static int unnote_qf_name(struct fs_context *fc, int qtype)
2053{
2054 struct ext4_fs_context *ctx = fc->fs_private;
2055
2056 if (ctx->s_qf_names[qtype])
2057 kfree(ctx->s_qf_names[qtype]);
2058
2059 ctx->s_qf_names[qtype] = NULL;
2060 ctx->qname_spec |= 1 << qtype;
6e47a3cc 2061 ctx->spec |= EXT4_SPEC_JQUOTA;
e6e268cb
LC		 2062 return 0;
2063}
2064#endif
2065
85456054
EB		 2066static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2067 struct ext4_fs_context *ctx)
2068{
2069 int err;
2070
2071 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2072 ext4_msg(NULL, KERN_WARNING,
2073 "test_dummy_encryption option not supported");
2074 return -EINVAL;
2075 }
2076 err = fscrypt_parse_test_dummy_encryption(param,
2077 &ctx->dummy_enc_policy);
2078 if (err == -EINVAL) {
2079 ext4_msg(NULL, KERN_WARNING,
2080 "Value of option \"%s\" is unrecognized", param->key);
2081 } else if (err == -EEXIST) {
2082 ext4_msg(NULL, KERN_WARNING,
2083 "Conflicting test_dummy_encryption options");
2084 return -EINVAL;
2085 }
2086 return err;
2087}
2088
6e47a3cc 2089#define EXT4_SET_CTX(name) \
4c246728
LC		 2090static inline void ctx_set_##name(struct ext4_fs_context *ctx, \
2091 unsigned long flag) \
6e47a3cc
LC		 2092{ \
2093 ctx->mask_s_##name |= flag; \
2094 ctx->vals_s_##name |= flag; \
e3952fcc
LC		 2095}
2096
2097#define EXT4_CLEAR_CTX(name) \
4c246728
LC		 2098static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \
2099 unsigned long flag) \
6e47a3cc
LC		 2100{ \
2101 ctx->mask_s_##name |= flag; \
2102 ctx->vals_s_##name &= ~flag; \
e3952fcc
LC		 2103}
2104
2105#define EXT4_TEST_CTX(name) \
4c246728
LC		 2106static inline unsigned long \
2107ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
6e47a3cc 2108{ \
4c246728 2109 return (ctx->vals_s_##name & flag); \
e3952fcc 2110}
6e47a3cc 2111
e3952fcc 2112EXT4_SET_CTX(flags); /* set only */
6e47a3cc 2113EXT4_SET_CTX(mount_opt);
e3952fcc
LC		 2114EXT4_CLEAR_CTX(mount_opt);
2115EXT4_TEST_CTX(mount_opt);
6e47a3cc 2116EXT4_SET_CTX(mount_opt2);
e3952fcc
LC		 2117EXT4_CLEAR_CTX(mount_opt2);
2118EXT4_TEST_CTX(mount_opt2);
2119
2120static inline void ctx_set_mount_flag(struct ext4_fs_context *ctx, int bit)
2121{
2122 set_bit(bit, &ctx->mask_s_mount_flags);
2123 set_bit(bit, &ctx->vals_s_mount_flags);
2124}
6e47a3cc 2125
02f960f8 2126static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
26092bf5 2127{
461c3af0 2128 struct ext4_fs_context *ctx = fc->fs_private;
461c3af0 2129 struct fs_parse_result result;
26092bf5 2130 const struct mount_opts *m;
461c3af0 2131 int is_remount;
08cefc7a
EB		 2132 kuid_t uid;
2133 kgid_t gid;
461c3af0
LC		 2134 int token;
2135
2136 token = fs_parse(fc, ext4_param_specs, param, &result);
2137 if (token < 0)
2138 return token;
2139 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
26092bf5 2140
ba2e524d
LC		 2141 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2142 if (token == m->token)
2143 break;
2144
2145 ctx->opt_flags |= m->flags;
2146
2147 if (m->flags & MOPT_EXPLICIT) {
2148 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2149 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2150 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2151 ctx_set_mount_opt2(ctx,
2152 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2153 } else
2154 return -EINVAL;
2155 }
2156
2157 if (m->flags & MOPT_NOSUPPORT) {
2158 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2159 param->key);
2160 return 0;
2161 }
2162
2163 switch (token) {
57f73c2c 2164#ifdef CONFIG_QUOTA
ba2e524d 2165 case Opt_usrjquota:
461c3af0 2166 if (!*param->string)
e6e268cb 2167 return unnote_qf_name(fc, USRQUOTA);
461c3af0 2168 else
e6e268cb 2169 return note_qf_name(fc, USRQUOTA, param);
ba2e524d 2170 case Opt_grpjquota:
461c3af0 2171 if (!*param->string)
e6e268cb 2172 return unnote_qf_name(fc, GRPQUOTA);
461c3af0 2173 else
e6e268cb 2174 return note_qf_name(fc, GRPQUOTA, param);
57f73c2c 2175#endif
26092bf5 2176 case Opt_sb:
7edfd85b
LC		 2177 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2178 ext4_msg(NULL, KERN_WARNING,
2179 "Ignoring %s option on remount", param->key);
2180 } else {
2181 ctx->s_sb_block = result.uint_32;
2182 ctx->spec |= EXT4_SPEC_s_sb_block;
2183 }
02f960f8 2184 return 0;
26092bf5 2185 case Opt_removed:
da812f61 2186 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
461c3af0 2187 param->key);
02f960f8 2188 return 0;
26092bf5 2189 case Opt_abort:
e3952fcc 2190 ctx_set_mount_flag(ctx, EXT4_MF_FS_ABORTED);
02f960f8 2191 return 0;
4f74d15f
EB		 2192 case Opt_inlinecrypt:
2193#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
6e47a3cc 2194 ctx_set_flags(ctx, SB_INLINECRYPT);
4f74d15f 2195#else
da812f61 2196 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
4f74d15f 2197#endif
02f960f8 2198 return 0;
461c3af0 2199 case Opt_errors:
6e47a3cc 2200 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
ba2e524d
LC		 2201 ctx_set_mount_opt(ctx, result.uint_32);
2202 return 0;
2203#ifdef CONFIG_QUOTA
2204 case Opt_jqfmt:
2205 ctx->s_jquota_fmt = result.uint_32;
2206 ctx->spec |= EXT4_SPEC_JQFMT;
2207 return 0;
2208#endif
2209 case Opt_data:
2210 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2211 ctx_set_mount_opt(ctx, result.uint_32);
2212 ctx->spec |= EXT4_SPEC_DATAJ;
2213 return 0;
2214 case Opt_commit:
461c3af0 2215 if (result.uint_32 == 0)
934b0de1 2216 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE;
461c3af0 2217 else if (result.uint_32 > INT_MAX / HZ) {
da812f61 2218 ext4_msg(NULL, KERN_ERR,
9ba55543 2219 "Invalid commit interval %d, "
2220 "must be smaller than %d",
461c3af0 2221 result.uint_32, INT_MAX / HZ);
da812f61 2222 return -EINVAL;
9ba55543 2223 }
6e47a3cc
LC		 2224 ctx->s_commit_interval = HZ * result.uint_32;
2225 ctx->spec |= EXT4_SPEC_s_commit_interval;
ba2e524d
LC		 2226 return 0;
2227 case Opt_debug_want_extra_isize:
6e47a3cc 2228 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
da812f61 2229 ext4_msg(NULL, KERN_ERR,
461c3af0 2230 "Invalid want_extra_isize %d", result.uint_32);
da812f61 2231 return -EINVAL;
9803387c 2232 }
6e47a3cc
LC		 2233 ctx->s_want_extra_isize = result.uint_32;
2234 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
ba2e524d
LC		 2235 return 0;
2236 case Opt_max_batch_time:
6e47a3cc
LC		 2237 ctx->s_max_batch_time = result.uint_32;
2238 ctx->spec |= EXT4_SPEC_s_max_batch_time;
ba2e524d
LC		 2239 return 0;
2240 case Opt_min_batch_time:
6e47a3cc
LC		 2241 ctx->s_min_batch_time = result.uint_32;
2242 ctx->spec |= EXT4_SPEC_s_min_batch_time;
ba2e524d
LC		 2243 return 0;
2244 case Opt_inode_readahead_blks:
461c3af0
LC		 2245 if (result.uint_32 &&
2246 (result.uint_32 > (1 << 30) ||
2247 !is_power_of_2(result.uint_32))) {
da812f61 2248 ext4_msg(NULL, KERN_ERR,
e33e60ea
JK		 2249 "EXT4-fs: inode_readahead_blks must be "
2250 "0 or a power of 2 smaller than 2^31");
da812f61 2251 return -EINVAL;
5f3633e3 2252 }
6e47a3cc
LC		 2253 ctx->s_inode_readahead_blks = result.uint_32;
2254 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
ba2e524d
LC		 2255 return 0;
2256 case Opt_init_itable:
6e47a3cc
LC		 2257 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2258 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
461c3af0 2259 if (param->type == fs_value_is_string)
6e47a3cc
LC		 2260 ctx->s_li_wait_mult = result.uint_32;
2261 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
ba2e524d
LC		 2262 return 0;
2263 case Opt_max_dir_size_kb:
6e47a3cc
LC		 2264 ctx->s_max_dir_size_kb = result.uint_32;
2265 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
ba2e524d 2266 return 0;
8016e29f 2267#ifdef CONFIG_EXT4_DEBUG
ba2e524d 2268 case Opt_fc_debug_max_replay:
6e47a3cc
LC		 2269 ctx->s_fc_debug_max_replay = result.uint_32;
2270 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
ba2e524d 2271 return 0;
8016e29f 2272#endif
ba2e524d 2273 case Opt_stripe:
6e47a3cc
LC		 2274 ctx->s_stripe = result.uint_32;
2275 ctx->spec |= EXT4_SPEC_s_stripe;
ba2e524d
LC		 2276 return 0;
2277 case Opt_resuid:
461c3af0 2278 uid = make_kuid(current_user_ns(), result.uint_32);
5f3633e3 2279 if (!uid_valid(uid)) {
da812f61 2280 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
461c3af0 2281 result.uint_32);
da812f61 2282 return -EINVAL;
26092bf5 2283 }
6e47a3cc
LC		 2284 ctx->s_resuid = uid;
2285 ctx->spec |= EXT4_SPEC_s_resuid;
ba2e524d
LC		 2286 return 0;
2287 case Opt_resgid:
461c3af0 2288 gid = make_kgid(current_user_ns(), result.uint_32);
5f3633e3 2289 if (!gid_valid(gid)) {
da812f61 2290 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
461c3af0 2291 result.uint_32);
da812f61 2292 return -EINVAL;
5f3633e3 2293 }
6e47a3cc
LC		 2294 ctx->s_resgid = gid;
2295 ctx->spec |= EXT4_SPEC_s_resgid;
ba2e524d
LC		 2296 return 0;
2297 case Opt_journal_dev:
5f3633e3 2298 if (is_remount) {
da812f61 2299 ext4_msg(NULL, KERN_ERR,
5f3633e3 2300 "Cannot specify journal on remount");
da812f61 2301 return -EINVAL;
5f3633e3 2302 }
461c3af0 2303 ctx->journal_devnum = result.uint_32;
6e47a3cc 2304 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
ba2e524d
LC		 2305 return 0;
2306 case Opt_journal_path:
2307 {
ad4eec61
ES		 2308 struct inode *journal_inode;
2309 struct path path;
2310 int error;
2311
2312 if (is_remount) {
da812f61 2313 ext4_msg(NULL, KERN_ERR,
ad4eec61 2314 "Cannot specify journal on remount");
da812f61 2315 return -EINVAL;
ad4eec61 2316 }
ad4eec61 2317
e3ea75ee 2318 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path);
ad4eec61 2319 if (error) {
da812f61 2320 ext4_msg(NULL, KERN_ERR, "error: could not find "
461c3af0 2321 "journal device path");
da812f61 2322 return -EINVAL;
ad4eec61
ES		 2323 }
2324
2b0143b5 2325 journal_inode = d_inode(path.dentry);
461c3af0 2326 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
6e47a3cc 2327 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
ad4eec61 2328 path_put(&path);
ba2e524d
LC		 2329 return 0;
2330 }
2331 case Opt_journal_ioprio:
461c3af0 2332 if (result.uint_32 > 7) {
da812f61 2333 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
5f3633e3 2334 " (must be 0-7)");
da812f61 2335 return -EINVAL;
5f3633e3 2336 }
461c3af0
LC		 2337 ctx->journal_ioprio =
2338 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
6e47a3cc 2339 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
ba2e524d
LC		 2340 return 0;
2341 case Opt_test_dummy_encryption:
85456054 2342 return ext4_parse_test_dummy_encryption(param, ctx);
ba2e524d
LC		 2343 case Opt_dax:
2344 case Opt_dax_type:
ef83b6e8 2345#ifdef CONFIG_FS_DAX
ba2e524d
LC		 2346 {
2347 int type = (token == Opt_dax) ?
2348 Opt_dax : result.uint_32;
2349
2350 switch (type) {
9cb20f94
IW		 2351 case Opt_dax:
2352 case Opt_dax_always:
ba2e524d 2353 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
6e47a3cc 2354 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
9cb20f94
IW		 2355 break;
2356 case Opt_dax_never:
ba2e524d 2357 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
6e47a3cc 2358 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
9cb20f94
IW		 2359 break;
2360 case Opt_dax_inode:
6e47a3cc
LC		 2361 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2362 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
9cb20f94 2363 /* Strictly for printing options */
ba2e524d 2364 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
9cb20f94
IW		 2365 break;
2366 }
ba2e524d
LC		 2367 return 0;
2368 }
ef83b6e8 2369#else
da812f61 2370 ext4_msg(NULL, KERN_INFO, "dax option not supported");
da812f61 2371 return -EINVAL;
5f3633e3 2372#endif
ba2e524d
LC		 2373 case Opt_data_err:
2374 if (result.uint_32 == Opt_data_err_abort)
2375 ctx_set_mount_opt(ctx, m->mount_opt);
2376 else if (result.uint_32 == Opt_data_err_ignore)
2377 ctx_clear_mount_opt(ctx, m->mount_opt);
2378 return 0;
2379 case Opt_mb_optimize_scan:
27b38686
OM		 2380 if (result.int_32 == 1) {
2381 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2382 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2383 } else if (result.int_32 == 0) {
2384 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2385 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2386 } else {
da812f61 2387 ext4_msg(NULL, KERN_WARNING,
196e402a 2388 "mb_optimize_scan should be set to 0 or 1.");
da812f61 2389 return -EINVAL;
196e402a 2390 }
ba2e524d
LC		 2391 return 0;
2392 }
2393
2394 /*
2395 * At this point we should only be getting options requiring MOPT_SET,
2396 * or MOPT_CLEAR. Anything else is a bug
2397 */
2398 if (m->token == Opt_err) {
2399 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2400 param->key);
2401 WARN_ON(1);
2402 return -EINVAL;
2403 }
2404
2405 else {
461c3af0
LC		 2406 unsigned int set = 0;
2407
2408 if ((param->type == fs_value_is_flag) ||
2409 result.uint_32 > 0)
2410 set = 1;
2411
5f3633e3 2412 if (m->flags & MOPT_CLEAR)
461c3af0 2413 set = !set;
5f3633e3 2414 else if (unlikely(!(m->flags & MOPT_SET))) {
da812f61 2415 ext4_msg(NULL, KERN_WARNING,
461c3af0
LC		 2416 "buggy handling of option %s",
2417 param->key);
5f3633e3 2418 WARN_ON(1);
da812f61 2419 return -EINVAL;
5f3633e3 2420 }
995a3ed6 2421 if (m->flags & MOPT_2) {
461c3af0 2422 if (set != 0)
6e47a3cc 2423 ctx_set_mount_opt2(ctx, m->mount_opt);
995a3ed6 2424 else
6e47a3cc 2425 ctx_clear_mount_opt2(ctx, m->mount_opt);
995a3ed6 2426 } else {
461c3af0 2427 if (set != 0)
6e47a3cc 2428 ctx_set_mount_opt(ctx, m->mount_opt);
995a3ed6 2429 else
6e47a3cc 2430 ctx_clear_mount_opt(ctx, m->mount_opt);
995a3ed6 2431 }
26092bf5 2432 }
ba2e524d 2433
02f960f8 2434 return 0;
26092bf5
TT		 2435}
2436
7edfd85b 2437static int parse_options(struct fs_context *fc, char *options)
26092bf5 2438{
461c3af0 2439 struct fs_parameter param;
461c3af0
LC		 2440 int ret;
2441 char *key;
26092bf5
TT		 2442
2443 if (!options)
7edfd85b 2444 return 0;
461c3af0
LC		 2445
2446 while ((key = strsep(&options, ",")) != NULL) {
2447 if (*key) {
2448 size_t v_len = 0;
2449 char *value = strchr(key, '=');
2450
2451 param.type = fs_value_is_flag;
2452 param.string = NULL;
2453
2454 if (value) {
2455 if (value == key)
2456 continue;
2457
2458 *value++ = 0;
2459 v_len = strlen(value);
2460 param.string = kmemdup_nul(value, v_len,
2461 GFP_KERNEL);
2462 if (!param.string)
7edfd85b 2463 return -ENOMEM;
461c3af0
LC		 2464 param.type = fs_value_is_string;
2465 }
2466
2467 param.key = key;
2468 param.size = v_len;
2469
02f960f8 2470 ret = ext4_parse_param(fc, &param);
461c3af0
LC		 2471 if (param.string)
2472 kfree(param.string);
2473 if (ret < 0)
7edfd85b 2474 return ret;
461c3af0 2475 }
ac27a0ec 2476 }
461c3af0 2477
7edfd85b 2478 ret = ext4_validate_options(fc);
da812f61 2479 if (ret < 0)
7edfd85b 2480 return ret;
da812f61 2481
7edfd85b
LC		 2482 return 0;
2483}
2484
2485static int parse_apply_sb_mount_options(struct super_block *sb,
2486 struct ext4_fs_context *m_ctx)
2487{
2488 struct ext4_sb_info *sbi = EXT4_SB(sb);
2489 char *s_mount_opts = NULL;
2490 struct ext4_fs_context *s_ctx = NULL;
2491 struct fs_context *fc = NULL;
2492 int ret = -ENOMEM;
2493
2494 if (!sbi->s_es->s_mount_opts[0])
e6e268cb
LC		 2495 return 0;
2496
7edfd85b
LC		 2497 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2498 sizeof(sbi->s_es->s_mount_opts),
2499 GFP_KERNEL);
2500 if (!s_mount_opts)
2501 return ret;
2502
2503 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2504 if (!fc)
2505 goto out_free;
2506
2507 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2508 if (!s_ctx)
2509 goto out_free;
2510
2511 fc->fs_private = s_ctx;
2512 fc->s_fs_info = sbi;
2513
2514 ret = parse_options(fc, s_mount_opts);
6e47a3cc 2515 if (ret < 0)
7edfd85b 2516 goto parse_failed;
e6e268cb 2517
7edfd85b
LC		 2518 ret = ext4_check_opt_consistency(fc, sb);
2519 if (ret < 0) {
2520parse_failed:
2521 ext4_msg(sb, KERN_WARNING,
2522 "failed to parse options in superblock: %s",
2523 s_mount_opts);
2524 ret = 0;
2525 goto out_free;
2526 }
2527
2528 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2529 m_ctx->journal_devnum = s_ctx->journal_devnum;
2530 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2531 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2532
85456054
EB		 2533 ext4_apply_options(fc, sb);
2534 ret = 0;
7edfd85b
LC		 2535
2536out_free:
c069db76
EB		 2537 if (fc) {
2538 ext4_fc_free(fc);
2539 kfree(fc);
2540 }
7edfd85b
LC		 2541 kfree(s_mount_opts);
2542 return ret;
4c94bff9
LC		 2543}
2544
e6e268cb
LC		 2545static void ext4_apply_quota_options(struct fs_context *fc,
2546 struct super_block *sb)
2547{
2548#ifdef CONFIG_QUOTA
6e47a3cc 2549 bool quota_feature = ext4_has_feature_quota(sb);
e6e268cb
LC		 2550 struct ext4_fs_context *ctx = fc->fs_private;
2551 struct ext4_sb_info *sbi = EXT4_SB(sb);
2552 char *qname;
2553 int i;
2554
6e47a3cc
LC		 2555 if (quota_feature)
2556 return;
2557
2558 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2559 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2560 if (!(ctx->qname_spec & (1 << i)))
2561 continue;
2562
2563 qname = ctx->s_qf_names[i]; /* May be NULL */
4c1bd5a9
LC		 2564 if (qname)
2565 set_opt(sb, QUOTA);
6e47a3cc 2566 ctx->s_qf_names[i] = NULL;
13b215a9
LC		 2567 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2568 lockdep_is_held(&sb->s_umount));
2569 if (qname)
10e4f310 2570 kfree_rcu_mightsleep(qname);
6e47a3cc 2571 }
e6e268cb 2572 }
6e47a3cc
LC		 2573
2574 if (ctx->spec & EXT4_SPEC_JQFMT)
2575 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
e6e268cb
LC		 2576#endif
2577}
2578
2579/*
2580 * Check quota settings consistency.
2581 */
2582static int ext4_check_quota_consistency(struct fs_context *fc,
2583 struct super_block *sb)
2584{
2585#ifdef CONFIG_QUOTA
2586 struct ext4_fs_context *ctx = fc->fs_private;
2587 struct ext4_sb_info *sbi = EXT4_SB(sb);
2588 bool quota_feature = ext4_has_feature_quota(sb);
2589 bool quota_loaded = sb_any_quota_loaded(sb);
6e47a3cc
LC		 2590 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2591 int quota_flags, i;
2592
2593 /*
2594 * We do the test below only for project quotas. 'usrquota' and
2595 * 'grpquota' mount options are allowed even without quota feature
2596 * to support legacy quotas in quota files.
2597 */
2598 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2599 !ext4_has_feature_project(sb)) {
2600 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2601 "Cannot enable project quota enforcement.");
2602 return -EINVAL;
2603 }
e6e268cb 2604
6e47a3cc
LC		 2605 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2606 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2607 if (quota_loaded &&
2608 ctx->mask_s_mount_opt & quota_flags &&
2609 !ctx_test_mount_opt(ctx, quota_flags))
2610 goto err_quota_change;
2611
2612 if (ctx->spec & EXT4_SPEC_JQUOTA) {
e6e268cb
LC		 2613
2614 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2615 if (!(ctx->qname_spec & (1 << i)))
2616 continue;
2617
6e47a3cc
LC		 2618 if (quota_loaded &&
2619 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
e6e268cb
LC		 2620 goto err_jquota_change;
2621
2622 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
13b215a9 2623 strcmp(get_qf_name(sb, sbi, i),
e6e268cb
LC		 2624 ctx->s_qf_names[i]) != 0)
2625 goto err_jquota_specified;
2626 }
6e47a3cc
LC		 2627
2628 if (quota_feature) {
2629 ext4_msg(NULL, KERN_INFO,
2630 "Journaled quota options ignored when "
2631 "QUOTA feature is enabled");
2632 return 0;
2633 }
e6e268cb
LC		 2634 }
2635
6e47a3cc 2636 if (ctx->spec & EXT4_SPEC_JQFMT) {
e6e268cb 2637 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
6e47a3cc 2638 goto err_jquota_change;
e6e268cb
LC		 2639 if (quota_feature) {
2640 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2641 "ignored when QUOTA feature is enabled");
2642 return 0;
2643 }
2644 }
6e47a3cc
LC		 2645
2646 /* Make sure we don't mix old and new quota format */
2647 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2648 ctx->s_qf_names[USRQUOTA]);
2649 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2650 ctx->s_qf_names[GRPQUOTA]);
2651
2652 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2653 test_opt(sb, USRQUOTA));
2654
2655 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2656 test_opt(sb, GRPQUOTA));
2657
2658 if (usr_qf_name) {
2659 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2660 usrquota = false;
2661 }
2662 if (grp_qf_name) {
2663 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2664 grpquota = false;
2665 }
2666
2667 if (usr_qf_name || grp_qf_name) {
2668 if (usrquota || grpquota) {
2669 ext4_msg(NULL, KERN_ERR, "old and new quota "
2670 "format mixing");
2671 return -EINVAL;
2672 }
2673
2674 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2675 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2676 "not specified");
2677 return -EINVAL;
2678 }
2679 }
2680
e6e268cb
LC		 2681 return 0;
2682
2683err_quota_change:
2684 ext4_msg(NULL, KERN_ERR,
2685 "Cannot change quota options when quota turned on");
2686 return -EINVAL;
2687err_jquota_change:
2688 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2689 "options when quota turned on");
2690 return -EINVAL;
2691err_jquota_specified:
2692 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2693 QTYPE2NAME(i));
2694 return -EINVAL;
e6e268cb
LC		 2695#else
2696 return 0;
2697#endif
2698}
2699
5f41fdae
EB		 2700static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2701 struct super_block *sb)
2702{
5f41fdae
EB		 2703 const struct ext4_fs_context *ctx = fc->fs_private;
2704 const struct ext4_sb_info *sbi = EXT4_SB(sb);
2705
85456054 2706 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
5f41fdae
EB		 2707 return 0;
2708
2709 if (!ext4_has_feature_encrypt(sb)) {
2710 ext4_msg(NULL, KERN_WARNING,
2711 "test_dummy_encryption requires encrypt feature");
2712 return -EINVAL;
2713 }
2714 /*
2715 * This mount option is just for testing, and it's not worthwhile to
2716 * implement the extra complexity (e.g. RCU protection) that would be
2717 * needed to allow it to be set or changed during remount. We do allow
2718 * it to be specified during remount, but only if there is no change.
2719 */
85456054
EB		 2720 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2721 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2722 &ctx->dummy_enc_policy))
2723 return 0;
5f41fdae 2724 ext4_msg(NULL, KERN_WARNING,
85456054 2725 "Can't set or change test_dummy_encryption on remount");
5f41fdae
EB		 2726 return -EINVAL;
2727 }
85456054
EB		 2728 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2729 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2730 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2731 &ctx->dummy_enc_policy))
2732 return 0;
2733 ext4_msg(NULL, KERN_WARNING,
2734 "Conflicting test_dummy_encryption options");
2735 return -EINVAL;
2736 }
7959eb19 2737 return 0;
85456054
EB		 2738}
2739
2740static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2741 struct super_block *sb)
2742{
2743 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2744 /* if already set, it was already verified to be the same */
2745 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2746 return;
2747 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2748 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2749 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
5f41fdae
EB		 2750}
2751
b6bd2435
LC		 2752static int ext4_check_opt_consistency(struct fs_context *fc,
2753 struct super_block *sb)
2754{
2755 struct ext4_fs_context *ctx = fc->fs_private;
6e47a3cc
LC		 2756 struct ext4_sb_info *sbi = fc->s_fs_info;
2757 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
5f41fdae 2758 int err;
b6bd2435
LC		 2759
2760 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2761 ext4_msg(NULL, KERN_ERR,
2762 "Mount option(s) incompatible with ext2");
2763 return -EINVAL;
2764 }
2765 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2766 ext4_msg(NULL, KERN_ERR,
2767 "Mount option(s) incompatible with ext3");
2768 return -EINVAL;
2769 }
2770
6e47a3cc
LC		 2771 if (ctx->s_want_extra_isize >
2772 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2773 ext4_msg(NULL, KERN_ERR,
2774 "Invalid want_extra_isize %d",
2775 ctx->s_want_extra_isize);
2776 return -EINVAL;
2777 }
2778
2779 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DIOREAD_NOLOCK)) {
2780 int blocksize =
2781 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2782 if (blocksize < PAGE_SIZE)
2783 ext4_msg(NULL, KERN_WARNING, "Warning: mounting with an "
2784 "experimental mount option 'dioread_nolock' "
2785 "for blocksize < PAGE_SIZE");
2786 }
2787
5f41fdae
EB		 2788 err = ext4_check_test_dummy_encryption(fc, sb);
2789 if (err)
2790 return err;
6e47a3cc
LC		 2791
2792 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2793 if (!sbi->s_journal) {
2794 ext4_msg(NULL, KERN_WARNING,
2795 "Remounting file system with no journal "
2796 "so ignoring journalled data option");
2797 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
4c246728
LC		 2798 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2799 test_opt(sb, DATA_FLAGS)) {
6e47a3cc
LC		 2800 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2801 "on remount");
2802 return -EINVAL;
2803 }
2804 }
2805
2806 if (is_remount) {
2807 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2808 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2809 ext4_msg(NULL, KERN_ERR, "can't mount with "
2810 "both data=journal and dax");
2811 return -EINVAL;
2812 }
2813
2814 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2815 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2816 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2817fail_dax_change_remount:
2818 ext4_msg(NULL, KERN_ERR, "can't change "
2819 "dax mount option while remounting");
2820 return -EINVAL;
2821 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2822 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2823 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2824 goto fail_dax_change_remount;
2825 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2826 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2827 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2828 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2829 goto fail_dax_change_remount;
2830 }
2831 }
2832
b6bd2435
LC		 2833 return ext4_check_quota_consistency(fc, sb);
2834}
2835
85456054 2836static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
4c94bff9 2837{
6e47a3cc 2838 struct ext4_fs_context *ctx = fc->fs_private;
da812f61 2839 struct ext4_sb_info *sbi = fc->s_fs_info;
6e47a3cc
LC		 2840
2841 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2842 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2843 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2844 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2845 sbi->s_mount_flags &= ~ctx->mask_s_mount_flags;
2846 sbi->s_mount_flags |= ctx->vals_s_mount_flags;
2847 sb->s_flags &= ~ctx->mask_s_flags;
2848 sb->s_flags |= ctx->vals_s_flags;
2849
2850#define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2851 APPLY(s_commit_interval);
2852 APPLY(s_stripe);
2853 APPLY(s_max_batch_time);
2854 APPLY(s_min_batch_time);
2855 APPLY(s_want_extra_isize);
2856 APPLY(s_inode_readahead_blks);
2857 APPLY(s_max_dir_size_kb);
2858 APPLY(s_li_wait_mult);
2859 APPLY(s_resgid);
2860 APPLY(s_resuid);
2861
2862#ifdef CONFIG_EXT4_DEBUG
2863 APPLY(s_fc_debug_max_replay);
2864#endif
2865
2866 ext4_apply_quota_options(fc, sb);
85456054 2867 ext4_apply_test_dummy_encryption(ctx, sb);
6e47a3cc
LC		 2868}
2869
2870
2871static int ext4_validate_options(struct fs_context *fc)
2872{
ac27a0ec 2873#ifdef CONFIG_QUOTA
6e47a3cc 2874 struct ext4_fs_context *ctx = fc->fs_private;
4c94bff9 2875 char *usr_qf_name, *grp_qf_name;
6e47a3cc
LC		 2876
2877 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2878 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2879
33458eab 2880 if (usr_qf_name || grp_qf_name) {
6e47a3cc
LC		 2881 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2882 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
ac27a0ec 2883
6e47a3cc
LC		 2884 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2885 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
ac27a0ec 2886
6e47a3cc
LC		 2887 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2888 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
da812f61 2889 ext4_msg(NULL, KERN_ERR, "old and new quota "
6e47a3cc 2890 "format mixing");
da812f61 2891 return -EINVAL;
ac27a0ec 2892 }
ac27a0ec
DK		 2893 }
2894#endif
6e47a3cc 2895 return 1;
ac27a0ec
DK		 2896}
2897
2adf6da8
TT		 2898static inline void ext4_show_quota_options(struct seq_file *seq,
2899 struct super_block *sb)
2900{
2901#if defined(CONFIG_QUOTA)
2902 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 2903 char *usr_qf_name, *grp_qf_name;
2adf6da8
TT		 2904
2905 if (sbi->s_jquota_fmt) {
2906 char *fmtname = "";
2907
2908 switch (sbi->s_jquota_fmt) {
2909 case QFMT_VFS_OLD:
2910 fmtname = "vfsold";
2911 break;
2912 case QFMT_VFS_V0:
2913 fmtname = "vfsv0";
2914 break;
2915 case QFMT_VFS_V1:
2916 fmtname = "vfsv1";
2917 break;
2918 }
2919 seq_printf(seq, ",jqfmt=%s", fmtname);
2920 }
2921
33458eab
TT		 2922 rcu_read_lock();
2923 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2924 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2925 if (usr_qf_name)
2926 seq_show_option(seq, "usrjquota", usr_qf_name);
2927 if (grp_qf_name)
2928 seq_show_option(seq, "grpjquota", grp_qf_name);
2929 rcu_read_unlock();
2adf6da8
TT		 2930#endif
2931}
2932
5a916be1
TT		 2933static const char *token2str(int token)
2934{
97d8a670 2935 const struct fs_parameter_spec *spec;
5a916be1 2936
97d8a670
LC		 2937 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2938 if (spec->opt == token && !spec->type)
5a916be1 2939 break;
97d8a670 2940 return spec->name;
5a916be1
TT		 2941}
2942
2adf6da8
TT		 2943/*
2944 * Show an option if
2945 * - it's set to a non-default value OR
2946 * - if the per-sb default is different from the global default
2947 */
66acdcf4
TT		 2948static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2949 int nodefs)
2adf6da8 2950{
2adf6da8
TT		 2951 struct ext4_sb_info *sbi = EXT4_SB(sb);
2952 struct ext4_super_block *es = sbi->s_es;
e3645d72 2953 int def_errors;
5a916be1 2954 const struct mount_opts *m;
66acdcf4 2955 char sep = nodefs ? '\n' : ',';
2adf6da8 2956
66acdcf4
TT		 2957#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2958#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2959
2960 if (sbi->s_sb_block != 1)
5a916be1
TT		 2961 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2962
2963 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2964 int want_set = m->flags & MOPT_SET;
e3645d72
ZY		 2965 int opt_2 = m->flags & MOPT_2;
2966 unsigned int mount_opt, def_mount_opt;
2967
5a916be1 2968 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
ba2e524d 2969 m->flags & MOPT_SKIP)
5a916be1 2970 continue;
e3645d72
ZY		 2971
2972 if (opt_2) {
2973 mount_opt = sbi->s_mount_opt2;
2974 def_mount_opt = sbi->s_def_mount_opt2;
2975 } else {
2976 mount_opt = sbi->s_mount_opt;
2977 def_mount_opt = sbi->s_def_mount_opt;
2978 }
2979 /* skip if same as the default */
2980 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt)))
2981 continue;
2982 /* select Opt_noFoo vs Opt_Foo */
5a916be1 2983 if ((want_set &&
e3645d72
ZY		 2984 (mount_opt & m->mount_opt) != m->mount_opt) ||
2985 (!want_set && (mount_opt & m->mount_opt)))
2986 continue;
5a916be1 2987 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2988 }
5a916be1 2989
08cefc7a 2990 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2991 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2992 SEQ_OPTS_PRINT("resuid=%u",
2993 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2994 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2995 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2996 SEQ_OPTS_PRINT("resgid=%u",
2997 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2998 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2999 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
3000 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 3001 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 3002 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 3003 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 3004 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 3005 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 3006 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 3007 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 3008 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 3009 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 3010 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
66acdcf4 3011 if (nodefs || sbi->s_stripe)
5a916be1 3012 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0 3013 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
e3645d72 3014 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5a916be1
TT		 3015 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3016 SEQ_OPTS_PUTS("data=journal");
3017 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3018 SEQ_OPTS_PUTS("data=ordered");
3019 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
3020 SEQ_OPTS_PUTS("data=writeback");
3021 }
66acdcf4
TT		 3022 if (nodefs ||
3023 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 3024 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
3025 sbi->s_inode_readahead_blks);
2adf6da8 3026
ceec0376 3027 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 3028 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 3029 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 3030 if (nodefs || sbi->s_max_dir_size_kb)
3031 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 3032 if (test_opt(sb, DATA_ERR_ABORT))
3033 SEQ_OPTS_PUTS("data_err=abort");
ed318a6c
EB		 3034
3035 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2adf6da8 3036
4f74d15f
EB		 3037 if (sb->s_flags & SB_INLINECRYPT)
3038 SEQ_OPTS_PUTS("inlinecrypt");
3039
9cb20f94
IW		 3040 if (test_opt(sb, DAX_ALWAYS)) {
3041 if (IS_EXT2_SB(sb))
3042 SEQ_OPTS_PUTS("dax");
3043 else
3044 SEQ_OPTS_PUTS("dax=always");
3045 } else if (test_opt2(sb, DAX_NEVER)) {
3046 SEQ_OPTS_PUTS("dax=never");
3047 } else if (test_opt2(sb, DAX_INODE)) {
3048 SEQ_OPTS_PUTS("dax=inode");
3049 }
3fa5d23e
OM		 3050
3051 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3052 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3053 SEQ_OPTS_PUTS("mb_optimize_scan=0");
3054 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3055 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3056 SEQ_OPTS_PUTS("mb_optimize_scan=1");
3057 }
3058
2adf6da8 3059 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 3060 return 0;
3061}
3062
66acdcf4
TT		 3063static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3064{
3065 return _ext4_show_options(seq, root->d_sb, 0);
3066}
3067
ebd173be 3068int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 3069{
3070 struct super_block *sb = seq->private;
3071 int rc;
3072
bc98a42c 3073 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 3074 rc = _ext4_show_options(seq, sb, 1);
3075 seq_puts(seq, "\n");
3076 return rc;
3077}
3078
617ba13b 3079static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 3080 int read_only)
3081{
617ba13b 3082 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 3083 int err = 0;
ac27a0ec 3084
617ba13b 3085 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 3086 ext4_msg(sb, KERN_ERR, "revision level too high, "
3087 "forcing read-only mode");
c89128a0 3088 err = -EROFS;
5adaccac 3089 goto done;
ac27a0ec
DK		 3090 }
3091 if (read_only)
281b5995 3092 goto done;
617ba13b 3093 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 3094 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3095 "running e2fsck is recommended");
c8b459f4 3096 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 3097 ext4_msg(sb, KERN_WARNING,
3098 "warning: mounting fs with errors, "
3099 "running e2fsck is recommended");
ed3ce80a 3100 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 3101 le16_to_cpu(es->s_mnt_count) >=
3102 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 3103 ext4_msg(sb, KERN_WARNING,
3104 "warning: maximal mount count reached, "
3105 "running e2fsck is recommended");
ac27a0ec 3106 else if (le32_to_cpu(es->s_checkinterval) &&
6a0678a7
AB		 3107 (ext4_get_tstamp(es, s_lastcheck) +
3108 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
b31e1552
ES		 3109 ext4_msg(sb, KERN_WARNING,
3110 "warning: checktime reached, "
3111 "running e2fsck is recommended");
0b8e58a1 3112 if (!sbi->s_journal)
0390131b 3113 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 3114 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 3115 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 3116 le16_add_cpu(&es->s_mnt_count, 1);
6a0678a7 3117 ext4_update_tstamp(es, s_mtime);
02f310fc 3118 if (sbi->s_journal) {
e2b911c5 3119 ext4_set_feature_journal_needs_recovery(sb);
02f310fc
JK		 3120 if (ext4_has_feature_orphan_file(sb))
3121 ext4_set_feature_orphan_present(sb);
3122 }
ac27a0ec 3123
4392fbc4 3124 err = ext4_commit_super(sb);
281b5995 3125done:
ac27a0ec 3126 if (test_opt(sb, DEBUG))
a9df9a49 3127 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 3128 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 3129 sb->s_blocksize,
3130 sbi->s_groups_count,
617ba13b
MC		 3131 EXT4_BLOCKS_PER_GROUP(sb),
3132 EXT4_INODES_PER_GROUP(sb),
a2595b8a 3133 sbi->s_mount_opt, sbi->s_mount_opt2);
c89128a0 3134 return err;
ac27a0ec
DK		 3135}
3136
117fff10
TT		 3137int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3138{
3139 struct ext4_sb_info *sbi = EXT4_SB(sb);
7c990728 3140 struct flex_groups **old_groups, **new_groups;
37b0b6b8 3141 int size, i, j;
117fff10
TT		 3142
3143 if (!sbi->s_log_groups_per_flex)
3144 return 0;
3145
3146 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3147 if (size <= sbi->s_flex_groups_allocated)
3148 return 0;
3149
7c990728
SJS		 3150 new_groups = kvzalloc(roundup_pow_of_two(size *
3151 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
117fff10 3152 if (!new_groups) {
7c990728
SJS		 3153 ext4_msg(sb, KERN_ERR,
3154 "not enough memory for %d flex group pointers", size);
117fff10
TT		 3155 return -ENOMEM;
3156 }
7c990728
SJS		 3157 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3158 new_groups[i] = kvzalloc(roundup_pow_of_two(
3159 sizeof(struct flex_groups)),
3160 GFP_KERNEL);
3161 if (!new_groups[i]) {
37b0b6b8
DC		 3162 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3163 kvfree(new_groups[j]);
7c990728
SJS		 3164 kvfree(new_groups);
3165 ext4_msg(sb, KERN_ERR,
3166 "not enough memory for %d flex groups", size);
3167 return -ENOMEM;
3168 }
117fff10 3169 }
7c990728
SJS		 3170 rcu_read_lock();
3171 old_groups = rcu_dereference(sbi->s_flex_groups);
3172 if (old_groups)
3173 memcpy(new_groups, old_groups,
3174 (sbi->s_flex_groups_allocated *
3175 sizeof(struct flex_groups *)));
3176 rcu_read_unlock();
3177 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3178 sbi->s_flex_groups_allocated = size;
3179 if (old_groups)
3180 ext4_kvfree_array_rcu(old_groups);
117fff10
TT		 3181 return 0;
3182}
3183
772cb7c8
JS		 3184static int ext4_fill_flex_info(struct super_block *sb)
3185{
3186 struct ext4_sb_info *sbi = EXT4_SB(sb);
3187 struct ext4_group_desc *gdp = NULL;
7c990728 3188 struct flex_groups *fg;
772cb7c8 3189 ext4_group_t flex_group;
117fff10 3190 int i, err;
772cb7c8 3191
503358ae 3192 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 3193 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 3194 sbi->s_log_groups_per_flex = 0;
3195 return 1;
3196 }
3197
117fff10
TT		 3198 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3199 if (err)
9933fc0a 3200 goto failed;
772cb7c8 3201
772cb7c8 3202 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 3203 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 3204
3205 flex_group = ext4_flex_group(sbi, i);
7c990728
SJS		 3206 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3207 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
90ba983f 3208 atomic64_add(ext4_free_group_clusters(sb, gdp),
7c990728
SJS		 3209 &fg->free_clusters);
3210 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
772cb7c8
JS		 3211 }
3212
3213 return 1;
3214failed:
3215 return 0;
3216}
3217
e2b911c5 3218static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 3219 struct ext4_group_desc *gdp)
717d50e4 3220{
b47820ed 3221 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 3222 __u16 crc = 0;
feb0ab32 3223 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 3224 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 3225
9aa5d32b 3226 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 3227 /* Use new metadata_csum algorithm */
feb0ab32 3228 __u32 csum32;
b47820ed 3229 __u16 dummy_csum = 0;
feb0ab32 3230
feb0ab32
DW		 3231 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3232 sizeof(le_group));
b47820ed
DJ		 3233 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3234 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3235 sizeof(dummy_csum));
3236 offset += sizeof(dummy_csum);
3237 if (offset < sbi->s_desc_size)
3238 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3239 sbi->s_desc_size - offset);
feb0ab32
DW		 3240
3241 crc = csum32 & 0xFFFF;
3242 goto out;
717d50e4
AD		 3243 }
3244
feb0ab32 3245 /* old crc16 code */
e2b911c5 3246 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 3247 return 0;
3248
feb0ab32
DW		 3249 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3250 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3251 crc = crc16(crc, (__u8 *)gdp, offset);
3252 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3253 /* for checksum of struct ext4_group_desc do the rest...*/
4f043518 3254 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size)
feb0ab32 3255 crc = crc16(crc, (__u8 *)gdp + offset,
4f043518 3256 sbi->s_desc_size - offset);
feb0ab32
DW		 3257
3258out:
717d50e4
AD		 3259 return cpu_to_le16(crc);
3260}
3261
feb0ab32 3262int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 3263 struct ext4_group_desc *gdp)
3264{
feb0ab32 3265 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 3266 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 3267 return 0;
3268
3269 return 1;
3270}
3271
feb0ab32
DW		 3272void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3273 struct ext4_group_desc *gdp)
3274{
3275 if (!ext4_has_group_desc_csum(sb))
3276 return;
e2b911c5 3277 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 3278}
3279
ac27a0ec 3280/* Called at mount-time, super-block is locked */
bfff6873 3281static int ext4_check_descriptors(struct super_block *sb,
829fa70d 3282 ext4_fsblk_t sb_block,
bfff6873 3283 ext4_group_t *first_not_zeroed)
ac27a0ec 3284{
617ba13b
MC		 3285 struct ext4_sb_info *sbi = EXT4_SB(sb);
3286 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3287 ext4_fsblk_t last_block;
44de022c 3288 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
bd81d8ee
LV		 3289 ext4_fsblk_t block_bitmap;
3290 ext4_fsblk_t inode_bitmap;
3291 ext4_fsblk_t inode_table;
ce421581 3292 int flexbg_flag = 0;
bfff6873 3293 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 3294
e2b911c5 3295 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 3296 flexbg_flag = 1;
3297
af5bc92d 3298 ext4_debug("Checking group descriptors");
ac27a0ec 3299
197cd65a
AM		 3300 for (i = 0; i < sbi->s_groups_count; i++) {
3301 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3302
ce421581 3303 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 3304 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 3305 else
3306 last_block = first_block +
617ba13b 3307 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 3308
bfff6873
LC		 3309 if ((grp == sbi->s_groups_count) &&
3310 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3311 grp = i;
3312
8fadc143 3313 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 3314 if (block_bitmap == sb_block) {
3315 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3316 "Block bitmap for group %u overlaps "
3317 "superblock", i);
18db4b4e
TT		 3318 if (!sb_rdonly(sb))
3319 return 0;
829fa70d 3320 }
77260807
TT		 3321 if (block_bitmap >= sb_block + 1 &&
3322 block_bitmap <= last_bg_block) {
3323 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3324 "Block bitmap for group %u overlaps "
3325 "block group descriptors", i);
3326 if (!sb_rdonly(sb))
3327 return 0;
3328 }
2b2d6d01 3329 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 3330 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 3331 "Block bitmap for group %u not in group "
b31e1552 3332 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 3333 return 0;
3334 }
8fadc143 3335 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 3336 if (inode_bitmap == sb_block) {
3337 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3338 "Inode bitmap for group %u overlaps "
3339 "superblock", i);
18db4b4e
TT		 3340 if (!sb_rdonly(sb))
3341 return 0;
829fa70d 3342 }
77260807
TT		 3343 if (inode_bitmap >= sb_block + 1 &&
3344 inode_bitmap <= last_bg_block) {
3345 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3346 "Inode bitmap for group %u overlaps "
3347 "block group descriptors", i);
3348 if (!sb_rdonly(sb))
3349 return 0;
3350 }
2b2d6d01 3351 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 3352 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 3353 "Inode bitmap for group %u not in group "
b31e1552 3354 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 3355 return 0;
3356 }
8fadc143 3357 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 3358 if (inode_table == sb_block) {
3359 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3360 "Inode table for group %u overlaps "
3361 "superblock", i);
18db4b4e
TT		 3362 if (!sb_rdonly(sb))
3363 return 0;
829fa70d 3364 }
77260807
TT		 3365 if (inode_table >= sb_block + 1 &&
3366 inode_table <= last_bg_block) {
3367 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3368 "Inode table for group %u overlaps "
3369 "block group descriptors", i);
3370 if (!sb_rdonly(sb))
3371 return 0;
3372 }
bd81d8ee 3373 if (inode_table < first_block ||
2b2d6d01 3374 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 3375 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 3376 "Inode table for group %u not in group "
b31e1552 3377 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 3378 return 0;
3379 }
955ce5f5 3380 ext4_lock_group(sb, i);
feb0ab32 3381 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 3382 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3383 "Checksum for group %u failed (%u!=%u)",
e2b911c5 3384 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 3385 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 3386 if (!sb_rdonly(sb)) {
955ce5f5 3387 ext4_unlock_group(sb, i);
8a266467 3388 return 0;
7ee1ec4c 3389 }
717d50e4 3390 }
955ce5f5 3391 ext4_unlock_group(sb, i);
ce421581
JS		 3392 if (!flexbg_flag)
3393 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 3394 }
bfff6873
LC		 3395 if (NULL != first_not_zeroed)
3396 *first_not_zeroed = grp;
ac27a0ec
DK		 3397 return 1;
3398}
3399
cd2291a4
ES		 3400/*
3401 * Maximal extent format file size.
3402 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3403 * extent format containers, within a sector_t, and within i_blocks
3404 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3405 * so that won't be a limiting factor.
3406 *
f17722f9
LC		 3407 * However there is other limiting factor. We do store extents in the form
3408 * of starting block and length, hence the resulting length of the extent
3409 * covering maximum file size must fit into on-disk format containers as
3410 * well. Given that length is always by 1 unit bigger than max unit (because
3411 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3412 *
cd2291a4
ES		 3413 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3414 */
f287a1a5 3415static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 3416{
3417 loff_t res;
3418 loff_t upper_limit = MAX_LFS_FILESIZE;
3419
72deb455
CH		 3420 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3421
3422 if (!has_huge_files) {
cd2291a4
ES		 3423 upper_limit = (1LL << 32) - 1;
3424
3425 /* total blocks in file system block size */
3426 upper_limit >>= (blkbits - 9);
3427 upper_limit <<= blkbits;
3428 }
3429
f17722f9
LC		 3430 /*
3431 * 32-bit extent-start container, ee_block. We lower the maxbytes
3432 * by one fs block, so ee_len can cover the extent of maximum file
3433 * size
3434 */
3435 res = (1LL << 32) - 1;
cd2291a4 3436 res <<= blkbits;
cd2291a4
ES		 3437
3438 /* Sanity check against vm- & vfs- imposed limits */
3439 if (res > upper_limit)
3440 res = upper_limit;
3441
3442 return res;
3443}
ac27a0ec 3444
ac27a0ec 3445/*
cd2291a4 3446 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 3447 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3448 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 3449 */
f287a1a5 3450static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 3451{
5c93e8ec 3452 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
0fc1b451 3453 int meta_blocks;
5c93e8ec 3454 unsigned int ppb = 1 << (bits - 2);
75ca6ad4
RH		 3455
3456 /*
3457 * This is calculated to be the largest file size for a dense, block
0b8e58a1
AD		 3458 * mapped file such that the file's total number of 512-byte sectors,
3459 * including data and all indirect blocks, does not exceed (2^48 - 1).
3460 *
3461 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3462 * number of 512-byte sectors of the file.
0fc1b451 3463 */
72deb455 3464 if (!has_huge_files) {
0fc1b451 3465 /*
72deb455
CH		 3466 * !has_huge_files or implies that the inode i_block field
3467 * represents total file blocks in 2^32 512-byte sectors ==
3468 * size of vfs inode i_blocks * 8
0fc1b451
AK		 3469 */
3470 upper_limit = (1LL << 32) - 1;
3471
3472 /* total blocks in file system block size */
3473 upper_limit >>= (bits - 9);
3474
3475 } else {
8180a562
AK		 3476 /*
3477 * We use 48 bit ext4_inode i_blocks
3478 * With EXT4_HUGE_FILE_FL set the i_blocks
3479 * represent total number of blocks in
3480 * file system block size
3481 */
0fc1b451
AK		 3482 upper_limit = (1LL << 48) - 1;
3483
0fc1b451
AK		 3484 }
3485
5c93e8ec
ZY		 3486 /* Compute how many blocks we can address by block tree */
3487 res += ppb;
3488 res += ppb * ppb;
3489 res += ((loff_t)ppb) * ppb * ppb;
3490 /* Compute how many metadata blocks are needed */
3491 meta_blocks = 1;
3492 meta_blocks += 1 + ppb;
3493 meta_blocks += 1 + ppb + ppb * ppb;
3494 /* Does block tree limit file size? */
3495 if (res + meta_blocks <= upper_limit)
3496 goto check_lfs;
3497
3498 res = upper_limit;
3499 /* How many metadata blocks are needed for addressing upper_limit? */
3500 upper_limit -= EXT4_NDIR_BLOCKS;
0fc1b451
AK		 3501 /* indirect blocks */
3502 meta_blocks = 1;
5c93e8ec 3503 upper_limit -= ppb;
0fc1b451 3504 /* double indirect blocks */
5c93e8ec
ZY		 3505 if (upper_limit < ppb * ppb) {
3506 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3507 res -= meta_blocks;
3508 goto check_lfs;
3509 }
3510 meta_blocks += 1 + ppb;
3511 upper_limit -= ppb * ppb;
3512 /* tripple indirect blocks for the rest */
3513 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3514 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3515 res -= meta_blocks;
3516check_lfs:
ac27a0ec 3517 res <<= bits;
0fc1b451
AK		 3518 if (res > MAX_LFS_FILESIZE)
3519 res = MAX_LFS_FILESIZE;
3520
5c93e8ec 3521 return res;
ac27a0ec
DK		 3522}
3523
617ba13b 3524static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 3525 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 3526{
617ba13b 3527 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 3528 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 3529 int has_super = 0;
3530
3531 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3532
e2b911c5 3533 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 3534 return logical_sb_block + nr + 1;
ac27a0ec 3535 bg = sbi->s_desc_per_block * nr;
617ba13b 3536 if (ext4_bg_has_super(sb, bg))
ac27a0ec 3537 has_super = 1;
0b8e58a1 3538
bd63f6b0
DW		 3539 /*
3540 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3541 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3542 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3543 * compensate.
3544 */
3545 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 3546 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 3547 has_super++;
3548
617ba13b 3549 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 3550}
3551
c9de560d
AT		 3552/**
3553 * ext4_get_stripe_size: Get the stripe size.
3554 * @sbi: In memory super block info
3555 *
3556 * If we have specified it via mount option, then
3557 * use the mount option value. If the value specified at mount time is
3558 * greater than the blocks per group use the super block value.
3559 * If the super block value is greater than blocks per group return 0.
3560 * Allocator needs it be less than blocks per group.
3561 *
3562 */
3563static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3564{
3565 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3566 unsigned long stripe_width =
3567 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 3568 int ret;
c9de560d
AT		 3569
3570 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 3571 ret = sbi->s_stripe;
5469d7c3 3572 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 3573 ret = stripe_width;
5469d7c3 3574 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 3575 ret = stride;
3576 else
3577 ret = 0;
c9de560d 3578
3eb08658
DE		 3579 /*
3580 * If the stripe width is 1, this makes no sense and
3581 * we set it to 0 to turn off stripe handling code.
3582 */
3583 if (ret <= 1)
3584 ret = 0;
c9de560d 3585
3eb08658 3586 return ret;
c9de560d 3587}
ac27a0ec 3588
a13fb1a4
ES		 3589/*
3590 * Check whether this filesystem can be mounted based on
3591 * the features present and the RDONLY/RDWR mount requested.
3592 * Returns 1 if this filesystem can be mounted as requested,
3593 * 0 if it cannot be.
3594 */
25c6d98f 3595int ext4_feature_set_ok(struct super_block *sb, int readonly)
a13fb1a4 3596{
e2b911c5 3597 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 3598 ext4_msg(sb, KERN_ERR,
3599 "Couldn't mount because of "
3600 "unsupported optional features (%x)",
3601 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3602 ~EXT4_FEATURE_INCOMPAT_SUPP));
3603 return 0;
3604 }
3605
5298d4bf 3606#if !IS_ENABLED(CONFIG_UNICODE)
c83ad55e
GKB		 3607 if (ext4_has_feature_casefold(sb)) {
3608 ext4_msg(sb, KERN_ERR,
3609 "Filesystem with casefold feature cannot be "
3610 "mounted without CONFIG_UNICODE");
3611 return 0;
3612 }
3613#endif
3614
a13fb1a4
ES		 3615 if (readonly)
3616 return 1;
3617
e2b911c5 3618 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 3619 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 3620 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 3621 return 1;
3622 }
3623
a13fb1a4 3624 /* Check that feature set is OK for a read-write mount */
e2b911c5 3625 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 3626 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3627 "unsupported optional features (%x)",
3628 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3629 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3630 return 0;
3631 }
e2b911c5 3632 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 3633 ext4_msg(sb, KERN_ERR,
3634 "Can't support bigalloc feature without "
3635 "extents feature\n");
3636 return 0;
3637 }
7c319d32 3638
9db176bc 3639#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
d65d87a0
TT		 3640 if (!readonly && (ext4_has_feature_quota(sb) ||
3641 ext4_has_feature_project(sb))) {
7c319d32 3642 ext4_msg(sb, KERN_ERR,
d65d87a0 3643 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
689c958c
LX		 3644 return 0;
3645 }
7c319d32 3646#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 3647 return 1;
3648}
3649
66e61a9e
TT		 3650/*
3651 * This function is called once a day if we have errors logged
3652 * on the file system
3653 */
235699a8 3654static void print_daily_error_info(struct timer_list *t)
66e61a9e 3655{
235699a8
KC		 3656 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3657 struct super_block *sb = sbi->s_sb;
3658 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 3659
3660 if (es->s_error_count)
ae0f78de
TT		 3661 /* fsck newer than v1.41.13 is needed to clean this condition. */
3662 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 3663 le32_to_cpu(es->s_error_count));
3664 if (es->s_first_error_time) {
6a0678a7
AB		 3665 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3666 sb->s_id,
3667 ext4_get_tstamp(es, s_first_error_time),
66e61a9e
TT		 3668 (int) sizeof(es->s_first_error_func),
3669 es->s_first_error_func,
3670 le32_to_cpu(es->s_first_error_line));
3671 if (es->s_first_error_ino)
651e1c3b 3672 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3673 le32_to_cpu(es->s_first_error_ino));
3674 if (es->s_first_error_block)
651e1c3b 3675 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3676 le64_to_cpu(es->s_first_error_block));
651e1c3b 3677 printk(KERN_CONT "\n");
66e61a9e
TT		 3678 }
3679 if (es->s_last_error_time) {
6a0678a7
AB		 3680 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3681 sb->s_id,
3682 ext4_get_tstamp(es, s_last_error_time),
66e61a9e
TT		 3683 (int) sizeof(es->s_last_error_func),
3684 es->s_last_error_func,
3685 le32_to_cpu(es->s_last_error_line));
3686 if (es->s_last_error_ino)
651e1c3b 3687 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3688 le32_to_cpu(es->s_last_error_ino));
3689 if (es->s_last_error_block)
651e1c3b 3690 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3691 le64_to_cpu(es->s_last_error_block));
651e1c3b 3692 printk(KERN_CONT "\n");
66e61a9e
TT		 3693 }
3694 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3695}
3696
bfff6873
LC		 3697/* Find next suitable group and run ext4_init_inode_table */
3698static int ext4_run_li_request(struct ext4_li_request *elr)
3699{
3700 struct ext4_group_desc *gdp = NULL;
3d392b26
TT		 3701 struct super_block *sb = elr->lr_super;
3702 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3703 ext4_group_t group = elr->lr_next_group;
3d392b26 3704 unsigned int prefetch_ios = 0;
bfff6873 3705 int ret = 0;
4f3d1e45 3706 int nr = EXT4_SB(sb)->s_mb_prefetch;
39fec688 3707 u64 start_time;
bfff6873 3708
3d392b26 3709 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
4f3d1e45
OM		 3710 elr->lr_next_group = ext4_mb_prefetch(sb, group, nr, &prefetch_ios);
3711 ext4_mb_prefetch_fini(sb, elr->lr_next_group, nr);
3712 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, nr);
3d392b26
TT		 3713 if (group >= elr->lr_next_group) {
3714 ret = 1;
3715 if (elr->lr_first_not_zeroed != ngroups &&
3716 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3717 elr->lr_next_group = elr->lr_first_not_zeroed;
3718 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3719 ret = 0;
3720 }
3721 }
3722 return ret;
3723 }
bfff6873 3724
3d392b26 3725 for (; group < ngroups; group++) {
bfff6873
LC		 3726 gdp = ext4_get_group_desc(sb, group, NULL);
3727 if (!gdp) {
3728 ret = 1;
3729 break;
3730 }
3731
3732 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3733 break;
3734 }
3735
7f511862 3736 if (group >= ngroups)
bfff6873
LC		 3737 ret = 1;
3738
3739 if (!ret) {
39fec688 3740 start_time = ktime_get_real_ns();
bfff6873
LC		 3741 ret = ext4_init_inode_table(sb, group,
3742 elr->lr_timeout ? 0 : 1);
3d392b26 3743 trace_ext4_lazy_itable_init(sb, group);
bfff6873 3744 if (elr->lr_timeout == 0) {
39fec688
SX		 3745 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3746 EXT4_SB(elr->lr_super)->s_li_wait_mult);
bfff6873
LC		 3747 }
3748 elr->lr_next_sched = jiffies + elr->lr_timeout;
3749 elr->lr_next_group = group + 1;
3750 }
bfff6873
LC		 3751 return ret;
3752}
3753
3754/*
3755 * Remove lr_request from the list_request and free the
4ed5c033 3756 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 3757 */
3758static void ext4_remove_li_request(struct ext4_li_request *elr)
3759{
bfff6873
LC		 3760 if (!elr)
3761 return;
3762
bfff6873 3763 list_del(&elr->lr_request);
3d392b26 3764 EXT4_SB(elr->lr_super)->s_li_request = NULL;
bfff6873
LC		 3765 kfree(elr);
3766}
3767
3768static void ext4_unregister_li_request(struct super_block *sb)
3769{
1bb933fb
LC		 3770 mutex_lock(&ext4_li_mtx);
3771 if (!ext4_li_info) {
3772 mutex_unlock(&ext4_li_mtx);
bfff6873 3773 return;
1bb933fb 3774 }
bfff6873
LC		 3775
3776 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 3777 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 3778 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 3779 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 3780}
3781
8f1f7453
ES		 3782static struct task_struct *ext4_lazyinit_task;
3783
bfff6873
LC		 3784/*
3785 * This is the function where ext4lazyinit thread lives. It walks
3786 * through the request list searching for next scheduled filesystem.
3787 * When such a fs is found, run the lazy initialization request
3788 * (ext4_rn_li_request) and keep track of the time spend in this
3789 * function. Based on that time we compute next schedule time of
3790 * the request. When walking through the list is complete, compute
3791 * next waking time and put itself into sleep.
3792 */
3793static int ext4_lazyinit_thread(void *arg)
3794{
c30365b9 3795 struct ext4_lazy_init *eli = arg;
bfff6873
LC		 3796 struct list_head *pos, *n;
3797 struct ext4_li_request *elr;
4ed5c033 3798 unsigned long next_wakeup, cur;
bfff6873
LC		 3799
3800 BUG_ON(NULL == eli);
3b575495 3801 set_freezable();
bfff6873 3802
bfff6873
LC		 3803cont_thread:
3804 while (true) {
3805 next_wakeup = MAX_JIFFY_OFFSET;
3806
3807 mutex_lock(&eli->li_list_mtx);
3808 if (list_empty(&eli->li_request_list)) {
3809 mutex_unlock(&eli->li_list_mtx);
3810 goto exit_thread;
3811 }
bfff6873 3812 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3813 int err = 0;
3814 int progress = 0;
bfff6873
LC		 3815 elr = list_entry(pos, struct ext4_li_request,
3816 lr_request);
3817
e22834f0
DM		 3818 if (time_before(jiffies, elr->lr_next_sched)) {
3819 if (time_before(elr->lr_next_sched, next_wakeup))
3820 next_wakeup = elr->lr_next_sched;
3821 continue;
3822 }
3823 if (down_read_trylock(&elr->lr_super->s_umount)) {
3824 if (sb_start_write_trylock(elr->lr_super)) {
3825 progress = 1;
3826 /*
3827 * We hold sb->s_umount, sb can not
3828 * be removed from the list, it is
3829 * now safe to drop li_list_mtx
3830 */
3831 mutex_unlock(&eli->li_list_mtx);
3832 err = ext4_run_li_request(elr);
3833 sb_end_write(elr->lr_super);
3834 mutex_lock(&eli->li_list_mtx);
3835 n = pos->next;
b2c78cd0 3836 }
e22834f0
DM		 3837 up_read((&elr->lr_super->s_umount));
3838 }
3839 /* error, remove the lazy_init job */
3840 if (err) {
3841 ext4_remove_li_request(elr);
3842 continue;
3843 }
3844 if (!progress) {
3845 elr->lr_next_sched = jiffies +
8032bf12 3846 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
bfff6873 3847 }
bfff6873
LC		 3848 if (time_before(elr->lr_next_sched, next_wakeup))
3849 next_wakeup = elr->lr_next_sched;
3850 }
3851 mutex_unlock(&eli->li_list_mtx);
3852
a0acae0e 3853 try_to_freeze();
bfff6873 3854
4ed5c033
LC		 3855 cur = jiffies;
3856 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3857 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3858 cond_resched();
3859 continue;
3860 }
3861
4ed5c033
LC		 3862 schedule_timeout_interruptible(next_wakeup - cur);
3863
8f1f7453
ES		 3864 if (kthread_should_stop()) {
3865 ext4_clear_request_list();
3866 goto exit_thread;
3867 }
bfff6873
LC		 3868 }
3869
3870exit_thread:
3871 /*
3872 * It looks like the request list is empty, but we need
3873 * to check it under the li_list_mtx lock, to prevent any
3874 * additions into it, and of course we should lock ext4_li_mtx
3875 * to atomically free the list and ext4_li_info, because at
3876 * this point another ext4 filesystem could be registering
3877 * new one.
3878 */
3879 mutex_lock(&ext4_li_mtx);
3880 mutex_lock(&eli->li_list_mtx);
3881 if (!list_empty(&eli->li_request_list)) {
3882 mutex_unlock(&eli->li_list_mtx);
3883 mutex_unlock(&ext4_li_mtx);
3884 goto cont_thread;
3885 }
3886 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3887 kfree(ext4_li_info);
3888 ext4_li_info = NULL;
3889 mutex_unlock(&ext4_li_mtx);
3890
3891 return 0;
3892}
3893
3894static void ext4_clear_request_list(void)
3895{
3896 struct list_head *pos, *n;
3897 struct ext4_li_request *elr;
3898
3899 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3900 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3901 elr = list_entry(pos, struct ext4_li_request,
3902 lr_request);
3903 ext4_remove_li_request(elr);
3904 }
3905 mutex_unlock(&ext4_li_info->li_list_mtx);
3906}
3907
3908static int ext4_run_lazyinit_thread(void)
3909{
8f1f7453
ES		 3910 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3911 ext4_li_info, "ext4lazyinit");
3912 if (IS_ERR(ext4_lazyinit_task)) {
3913 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3914 ext4_clear_request_list();
bfff6873
LC		 3915 kfree(ext4_li_info);
3916 ext4_li_info = NULL;
92b97816 3917 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3918 "initialization thread\n",
3919 err);
3920 return err;
3921 }
3922 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3923 return 0;
3924}
3925
3926/*
3927 * Check whether it make sense to run itable init. thread or not.
3928 * If there is at least one uninitialized inode table, return
3929 * corresponding group number, else the loop goes through all
3930 * groups and return total number of groups.
3931 */
3932static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3933{
3934 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3935 struct ext4_group_desc *gdp = NULL;
3936
8844618d
TT		 3937 if (!ext4_has_group_desc_csum(sb))
3938 return ngroups;
3939
bfff6873
LC		 3940 for (group = 0; group < ngroups; group++) {
3941 gdp = ext4_get_group_desc(sb, group, NULL);
3942 if (!gdp)
3943 continue;
3944
50122847 3945 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
bfff6873
LC		 3946 break;
3947 }
3948
3949 return group;
3950}
3951
3952static int ext4_li_info_new(void)
3953{
3954 struct ext4_lazy_init *eli = NULL;
3955
3956 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3957 if (!eli)
3958 return -ENOMEM;
3959
bfff6873
LC		 3960 INIT_LIST_HEAD(&eli->li_request_list);
3961 mutex_init(&eli->li_list_mtx);
3962
bfff6873
LC		 3963 eli->li_state |= EXT4_LAZYINIT_QUIT;
3964
3965 ext4_li_info = eli;
3966
3967 return 0;
3968}
3969
3970static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3971 ext4_group_t start)
3972{
bfff6873 3973 struct ext4_li_request *elr;
bfff6873
LC		 3974
3975 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3976 if (!elr)
3977 return NULL;
3978
3979 elr->lr_super = sb;
3d392b26 3980 elr->lr_first_not_zeroed = start;
21175ca4 3981 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3d392b26
TT		 3982 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3983 elr->lr_next_group = start;
21175ca4
HS		 3984 } else {
3985 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3d392b26 3986 }
bfff6873
LC		 3987
3988 /*
3989 * Randomize first schedule time of the request to
3990 * spread the inode table initialization requests
3991 * better.
3992 */
8032bf12 3993 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
bfff6873
LC		 3994 return elr;
3995}
3996
7f511862
TT		 3997int ext4_register_li_request(struct super_block *sb,
3998 ext4_group_t first_not_zeroed)
bfff6873
LC		 3999{
4000 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 4001 struct ext4_li_request *elr = NULL;
49598e04 4002 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 4003 int ret = 0;
bfff6873 4004
7f511862 4005 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 4006 if (sbi->s_li_request != NULL) {
4007 /*
4008 * Reset timeout so it can be computed again, because
4009 * s_li_wait_mult might have changed.
4010 */
4011 sbi->s_li_request->lr_timeout = 0;
7f511862 4012 goto out;
51ce6511 4013 }
bfff6873 4014
426d15ad
JT		 4015 if (sb_rdonly(sb) ||
4016 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
4017 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
7f511862 4018 goto out;
bfff6873
LC		 4019
4020 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 4021 if (!elr) {
4022 ret = -ENOMEM;
4023 goto out;
4024 }
bfff6873
LC		 4025
4026 if (NULL == ext4_li_info) {
4027 ret = ext4_li_info_new();
4028 if (ret)
4029 goto out;
4030 }
4031
4032 mutex_lock(&ext4_li_info->li_list_mtx);
4033 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
4034 mutex_unlock(&ext4_li_info->li_list_mtx);
4035
4036 sbi->s_li_request = elr;
46e4690b
TM		 4037 /*
4038 * set elr to NULL here since it has been inserted to
4039 * the request_list and the removal and free of it is
4040 * handled by ext4_clear_request_list from now on.
4041 */
4042 elr = NULL;
bfff6873
LC		 4043
4044 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4045 ret = ext4_run_lazyinit_thread();
4046 if (ret)
4047 goto out;
4048 }
bfff6873 4049out:
beed5ecb
NK		 4050 mutex_unlock(&ext4_li_mtx);
4051 if (ret)
bfff6873 4052 kfree(elr);
bfff6873
LC		 4053 return ret;
4054}
4055
4056/*
4057 * We do not need to lock anything since this is called on
4058 * module unload.
4059 */
4060static void ext4_destroy_lazyinit_thread(void)
4061{
4062 /*
4063 * If thread exited earlier
4064 * there's nothing to be done.
4065 */
8f1f7453 4066 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 4067 return;
4068
8f1f7453 4069 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 4070}
4071
25ed6e8a
DW		 4072static int set_journal_csum_feature_set(struct super_block *sb)
4073{
4074 int ret = 1;
4075 int compat, incompat;
4076 struct ext4_sb_info *sbi = EXT4_SB(sb);
4077
9aa5d32b 4078 if (ext4_has_metadata_csum(sb)) {
db9ee220 4079 /* journal checksum v3 */
25ed6e8a 4080 compat = 0;
db9ee220 4081 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 4082 } else {
4083 /* journal checksum v1 */
4084 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4085 incompat = 0;
4086 }
4087
feb8c6d3
DW		 4088 jbd2_journal_clear_features(sbi->s_journal,
4089 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4090 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4091 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 4092 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4093 ret = jbd2_journal_set_features(sbi->s_journal,
4094 compat, 0,
4095 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4096 incompat);
4097 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4098 ret = jbd2_journal_set_features(sbi->s_journal,
4099 compat, 0,
4100 incompat);
4101 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4102 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4103 } else {
feb8c6d3
DW		 4104 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4105 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 4106 }
4107
4108 return ret;
4109}
4110
952fc18e
TT		 4111/*
4112 * Note: calculating the overhead so we can be compatible with
4113 * historical BSD practice is quite difficult in the face of
4114 * clusters/bigalloc. This is because multiple metadata blocks from
4115 * different block group can end up in the same allocation cluster.
4116 * Calculating the exact overhead in the face of clustered allocation
4117 * requires either O(all block bitmaps) in memory or O(number of block
4118 * groups**2) in time. We will still calculate the superblock for
4119 * older file systems --- and if we come across with a bigalloc file
4120 * system with zero in s_overhead_clusters the estimate will be close to
4121 * correct especially for very large cluster sizes --- but for newer
4122 * file systems, it's better to calculate this figure once at mkfs
4123 * time, and store it in the superblock. If the superblock value is
4124 * present (even for non-bigalloc file systems), we will use it.
4125 */
4126static int count_overhead(struct super_block *sb, ext4_group_t grp,
4127 char *buf)
4128{
4129 struct ext4_sb_info *sbi = EXT4_SB(sb);
4130 struct ext4_group_desc *gdp;
4131 ext4_fsblk_t first_block, last_block, b;
4132 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4133 int s, j, count = 0;
10b01ee9 4134 int has_super = ext4_bg_has_super(sb, grp);
952fc18e 4135
e2b911c5 4136 if (!ext4_has_feature_bigalloc(sb))
10b01ee9
TT		 4137 return (has_super + ext4_bg_num_gdb(sb, grp) +
4138 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
0548bbb8
TT		 4139 sbi->s_itb_per_group + 2);
4140
952fc18e
TT		 4141 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4142 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4143 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4144 for (i = 0; i < ngroups; i++) {
4145 gdp = ext4_get_group_desc(sb, i, NULL);
4146 b = ext4_block_bitmap(sb, gdp);
4147 if (b >= first_block && b <= last_block) {
4148 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4149 count++;
4150 }
4151 b = ext4_inode_bitmap(sb, gdp);
4152 if (b >= first_block && b <= last_block) {
4153 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4154 count++;
4155 }
4156 b = ext4_inode_table(sb, gdp);
4157 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4158 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4159 int c = EXT4_B2C(sbi, b - first_block);
4160 ext4_set_bit(c, buf);
4161 count++;
4162 }
4163 if (i != grp)
4164 continue;
4165 s = 0;
4166 if (ext4_bg_has_super(sb, grp)) {
4167 ext4_set_bit(s++, buf);
4168 count++;
4169 }
c48ae41b
TT		 4170 j = ext4_bg_num_gdb(sb, grp);
4171 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4172 ext4_error(sb, "Invalid number of block group "
4173 "descriptor blocks: %d", j);
4174 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 4175 }
c48ae41b
TT		 4176 count += j;
4177 for (; j > 0; j--)
4178 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 4179 }
4180 if (!count)
4181 return 0;
4182 return EXT4_CLUSTERS_PER_GROUP(sb) -
4183 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4184}
4185
4186/*
4187 * Compute the overhead and stash it in sbi->s_overhead
4188 */
4189int ext4_calculate_overhead(struct super_block *sb)
4190{
4191 struct ext4_sb_info *sbi = EXT4_SB(sb);
4192 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 4193 struct inode *j_inode;
4194 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 4195 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4196 ext4_fsblk_t overhead = 0;
4fdb5543 4197 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 4198
952fc18e
TT		 4199 if (!buf)
4200 return -ENOMEM;
4201
4202 /*
4203 * Compute the overhead (FS structures). This is constant
4204 * for a given filesystem unless the number of block groups
4205 * changes so we cache the previous value until it does.
4206 */
4207
4208 /*
4209 * All of the blocks before first_data_block are overhead
4210 */
4211 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4212
4213 /*
4214 * Add the overhead found in each block group
4215 */
4216 for (i = 0; i < ngroups; i++) {
4217 int blks;
4218
4219 blks = count_overhead(sb, i, buf);
4220 overhead += blks;
4221 if (blks)
4222 memset(buf, 0, PAGE_SIZE);
4223 cond_resched();
4224 }
3c816ded
EW		 4225
4226 /*
4227 * Add the internal journal blocks whether the journal has been
4228 * loaded or not
4229 */
ee7ed3aa 4230 if (sbi->s_journal && !sbi->s_journal_bdev)
ede7dc7f 4231 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
f1eec3b0
RH		 4232 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4233 /* j_inum for internal journal is non-zero */
3c816ded
EW		 4234 j_inode = ext4_get_journal_inode(sb, j_inum);
4235 if (j_inode) {
4236 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4237 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4238 iput(j_inode);
4239 } else {
4240 ext4_msg(sb, KERN_ERR, "can't get journal size");
4241 }
4242 }
952fc18e
TT		 4243 sbi->s_overhead = overhead;
4244 smp_wmb();
4245 free_page((unsigned long) buf);
4246 return 0;
4247}
4248
b5799018 4249static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 4250{
4251 ext4_fsblk_t resv_clusters;
b5799018 4252 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 4253
30fac0f7
JK		 4254 /*
4255 * There's no need to reserve anything when we aren't using extents.
4256 * The space estimates are exact, there are no unwritten extents,
4257 * hole punching doesn't need new metadata... This is needed especially
4258 * to keep ext2/3 backward compatibility.
4259 */
e2b911c5 4260 if (!ext4_has_feature_extents(sb))
b5799018 4261 return;
27dd4385
LC		 4262 /*
4263 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4264 * This should cover the situations where we can not afford to run
4265 * out of space like for example punch hole, or converting
556615dc 4266 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 4267 * allocation would require 1, or 2 blocks, higher numbers are
4268 * very rare.
4269 */
b5799018
TT		 4270 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4271 sbi->s_cluster_bits);
27dd4385
LC		 4272
4273 do_div(resv_clusters, 50);
4274 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4275
b5799018 4276 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 4277}
4278
ca9b404f
RA		 4279static const char *ext4_quota_mode(struct super_block *sb)
4280{
4281#ifdef CONFIG_QUOTA
4282 if (!ext4_quota_capable(sb))
4283 return "none";
4284
4285 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4286 return "journalled";
4287 else
4288 return "writeback";
4289#else
4290 return "disabled";
4291#endif
4292}
4293
188c299e
JK		 4294static void ext4_setup_csum_trigger(struct super_block *sb,
4295 enum ext4_journal_trigger_type type,
4296 void (*trigger)(
4297 struct jbd2_buffer_trigger_type *type,
4298 struct buffer_head *bh,
4299 void *mapped_data,
4300 size_t size))
4301{
4302 struct ext4_sb_info *sbi = EXT4_SB(sb);
4303
4304 sbi->s_journal_triggers[type].sb = sb;
4305 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4306}
4307
7edfd85b
LC		 4308static void ext4_free_sbi(struct ext4_sb_info *sbi)
4309{
4310 if (!sbi)
4311 return;
4312
4313 kfree(sbi->s_blockgroup_lock);
8012b866 4314 fs_put_dax(sbi->s_daxdev, NULL);
7edfd85b
LC		 4315 kfree(sbi);
4316}
4317
4318static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4319{
4320 struct ext4_sb_info *sbi;
4321
4322 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4323 if (!sbi)
4324 return NULL;
4325
8012b866
SR		 4326 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4327 NULL, NULL);
7edfd85b
LC		 4328
4329 sbi->s_blockgroup_lock =
4330 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4331
4332 if (!sbi->s_blockgroup_lock)
4333 goto err_out;
4334
4335 sb->s_fs_info = sbi;
4336 sbi->s_sb = sb;
4337 return sbi;
4338err_out:
8012b866 4339 fs_put_dax(sbi->s_daxdev, NULL);
7edfd85b
LC		 4340 kfree(sbi);
4341 return NULL;
4342}
4343
5f6d662d
JY		 4344static void ext4_set_def_opts(struct super_block *sb,
4345 struct ext4_super_block *es)
4346{
4347 unsigned long def_mount_opts;
4348
4349 /* Set defaults before we parse the mount options */
4350 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4351 set_opt(sb, INIT_INODE_TABLE);
4352 if (def_mount_opts & EXT4_DEFM_DEBUG)
4353 set_opt(sb, DEBUG);
4354 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4355 set_opt(sb, GRPID);
4356 if (def_mount_opts & EXT4_DEFM_UID16)
4357 set_opt(sb, NO_UID32);
4358 /* xattr user namespace & acls are now defaulted on */
4359 set_opt(sb, XATTR_USER);
4360#ifdef CONFIG_EXT4_FS_POSIX_ACL
4361 set_opt(sb, POSIX_ACL);
4362#endif
4363 if (ext4_has_feature_fast_commit(sb))
4364 set_opt2(sb, JOURNAL_FAST_COMMIT);
4365 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4366 if (ext4_has_metadata_csum(sb))
4367 set_opt(sb, JOURNAL_CHECKSUM);
4368
4369 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4370 set_opt(sb, JOURNAL_DATA);
4371 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4372 set_opt(sb, ORDERED_DATA);
4373 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4374 set_opt(sb, WRITEBACK_DATA);
4375
4376 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4377 set_opt(sb, ERRORS_PANIC);
4378 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4379 set_opt(sb, ERRORS_CONT);
4380 else
4381 set_opt(sb, ERRORS_RO);
4382 /* block_validity enabled by default; disable with noblock_validity */
4383 set_opt(sb, BLOCK_VALIDITY);
4384 if (def_mount_opts & EXT4_DEFM_DISCARD)
4385 set_opt(sb, DISCARD);
4386
4387 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4388 set_opt(sb, BARRIER);
4389
4390 /*
4391 * enable delayed allocation by default
4392 * Use -o nodelalloc to turn it off
4393 */
4394 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4395 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4396 set_opt(sb, DELALLOC);
3df11e27
JY		 4397
4398 if (sb->s_blocksize == PAGE_SIZE)
4399 set_opt(sb, DIOREAD_NOLOCK);
5f6d662d
JY		 4400}
4401
c8267c51 4402static int ext4_handle_clustersize(struct super_block *sb)
4a8557b0
JY		 4403{
4404 struct ext4_sb_info *sbi = EXT4_SB(sb);
4405 struct ext4_super_block *es = sbi->s_es;
4406 int clustersize;
4407
4408 /* Handle clustersize */
4409 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4410 if (ext4_has_feature_bigalloc(sb)) {
c8267c51 4411 if (clustersize < sb->s_blocksize) {
4a8557b0
JY		 4412 ext4_msg(sb, KERN_ERR,
4413 "cluster size (%d) smaller than "
c8267c51 4414 "block size (%lu)", clustersize, sb->s_blocksize);
4a8557b0
JY		 4415 return -EINVAL;
4416 }
4417 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4418 le32_to_cpu(es->s_log_block_size);
4419 sbi->s_clusters_per_group =
4420 le32_to_cpu(es->s_clusters_per_group);
c8267c51 4421 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4a8557b0
JY		 4422 ext4_msg(sb, KERN_ERR,
4423 "#clusters per group too big: %lu",
4424 sbi->s_clusters_per_group);
4425 return -EINVAL;
4426 }
4427 if (sbi->s_blocks_per_group !=
c8267c51 4428 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4a8557b0
JY		 4429 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4430 "clusters per group (%lu) inconsistent",
4431 sbi->s_blocks_per_group,
4432 sbi->s_clusters_per_group);
4433 return -EINVAL;
4434 }
4435 } else {
c8267c51 4436 if (clustersize != sb->s_blocksize) {
4a8557b0
JY		 4437 ext4_msg(sb, KERN_ERR,
4438 "fragment/cluster size (%d) != "
c8267c51 4439 "block size (%lu)", clustersize, sb->s_blocksize);
4a8557b0
JY		 4440 return -EINVAL;
4441 }
c8267c51 4442 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4a8557b0
JY		 4443 ext4_msg(sb, KERN_ERR,
4444 "#blocks per group too big: %lu",
4445 sbi->s_blocks_per_group);
4446 return -EINVAL;
4447 }
4448 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4449 sbi->s_cluster_bits = 0;
4450 }
c8267c51 4451 sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4a8557b0
JY		 4452
4453 /* Do we have standard group size of clustersize * 8 blocks ? */
4454 if (sbi->s_blocks_per_group == clustersize << 3)
4455 set_opt2(sb, STD_GROUP_SIZE);
4456
4457 return 0;
4458}
4459
f7314a67
JY		 4460static void ext4_fast_commit_init(struct super_block *sb)
4461{
4462 struct ext4_sb_info *sbi = EXT4_SB(sb);
4463
4464 /* Initialize fast commit stuff */
4465 atomic_set(&sbi->s_fc_subtid, 0);
4466 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4467 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4468 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4469 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4470 sbi->s_fc_bytes = 0;
4471 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4472 sbi->s_fc_ineligible_tid = 0;
4473 spin_lock_init(&sbi->s_fc_lock);
4474 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4475 sbi->s_fc_replay_state.fc_regions = NULL;
4476 sbi->s_fc_replay_state.fc_regions_size = 0;
4477 sbi->s_fc_replay_state.fc_regions_used = 0;
4478 sbi->s_fc_replay_state.fc_regions_valid = 0;
4479 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4480 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4481 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4482}
4483
0e495f7c 4484static int ext4_inode_info_init(struct super_block *sb,
c8267c51 4485 struct ext4_super_block *es)
0e495f7c
JY		 4486{
4487 struct ext4_sb_info *sbi = EXT4_SB(sb);
4488
4489 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4490 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4491 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4492 } else {
4493 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4494 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4495 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4496 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4497 sbi->s_first_ino);
4498 return -EINVAL;
4499 }
4500 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4501 (!is_power_of_2(sbi->s_inode_size)) ||
c8267c51 4502 (sbi->s_inode_size > sb->s_blocksize)) {
0e495f7c
JY		 4503 ext4_msg(sb, KERN_ERR,
4504 "unsupported inode size: %d",
4505 sbi->s_inode_size);
c8267c51 4506 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
0e495f7c
JY		 4507 return -EINVAL;
4508 }
4509 /*
4510 * i_atime_extra is the last extra field available for
4511 * [acm]times in struct ext4_inode. Checking for that
4512 * field should suffice to ensure we have extra space
4513 * for all three.
4514 */
4515 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4516 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4517 sb->s_time_gran = 1;
4518 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4519 } else {
4520 sb->s_time_gran = NSEC_PER_SEC;
4521 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4522 }
4523 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4524 }
4525
4526 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4527 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4528 EXT4_GOOD_OLD_INODE_SIZE;
4529 if (ext4_has_feature_extra_isize(sb)) {
4530 unsigned v, max = (sbi->s_inode_size -
4531 EXT4_GOOD_OLD_INODE_SIZE);
4532
4533 v = le16_to_cpu(es->s_want_extra_isize);
4534 if (v > max) {
4535 ext4_msg(sb, KERN_ERR,
4536 "bad s_want_extra_isize: %d", v);
4537 return -EINVAL;
4538 }
4539 if (sbi->s_want_extra_isize < v)
4540 sbi->s_want_extra_isize = v;
4541
4542 v = le16_to_cpu(es->s_min_extra_isize);
4543 if (v > max) {
4544 ext4_msg(sb, KERN_ERR,
4545 "bad s_min_extra_isize: %d", v);
4546 return -EINVAL;
4547 }
4548 if (sbi->s_want_extra_isize < v)
4549 sbi->s_want_extra_isize = v;
4550 }
4551 }
4552
4553 return 0;
4554}
4555
39c135b0
JY		 4556#if IS_ENABLED(CONFIG_UNICODE)
4557static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4558{
4559 const struct ext4_sb_encodings *encoding_info;
4560 struct unicode_map *encoding;
4561 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4562
4563 if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4564 return 0;
4565
4566 encoding_info = ext4_sb_read_encoding(es);
4567 if (!encoding_info) {
4568 ext4_msg(sb, KERN_ERR,
4569 "Encoding requested by superblock is unknown");
4570 return -EINVAL;
4571 }
4572
4573 encoding = utf8_load(encoding_info->version);
4574 if (IS_ERR(encoding)) {
4575 ext4_msg(sb, KERN_ERR,
4576 "can't mount with superblock charset: %s-%u.%u.%u "
4577 "not supported by the kernel. flags: 0x%x.",
4578 encoding_info->name,
4579 unicode_major(encoding_info->version),
4580 unicode_minor(encoding_info->version),
4581 unicode_rev(encoding_info->version),
4582 encoding_flags);
4583 return -EINVAL;
4584 }
4585 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4586 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4587 unicode_major(encoding_info->version),
4588 unicode_minor(encoding_info->version),
4589 unicode_rev(encoding_info->version),
4590 encoding_flags);
4591
4592 sb->s_encoding = encoding;
4593 sb->s_encoding_flags = encoding_flags;
4594
4595 return 0;
4596}
4597#else
4598static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4599{
4600 return 0;
4601}
4602#endif
4603
b26458d1
JY		 4604static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4605{
4606 struct ext4_sb_info *sbi = EXT4_SB(sb);
4607
4608 /* Warn if metadata_csum and gdt_csum are both set. */
4609 if (ext4_has_feature_metadata_csum(sb) &&
4610 ext4_has_feature_gdt_csum(sb))
4611 ext4_warning(sb, "metadata_csum and uninit_bg are "
4612 "redundant flags; please run fsck.");
4613
4614 /* Check for a known checksum algorithm */
4615 if (!ext4_verify_csum_type(sb, es)) {
4616 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4617 "unknown checksum algorithm.");
4618 return -EINVAL;
4619 }
4620 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4621 ext4_orphan_file_block_trigger);
4622
4623 /* Load the checksum driver */
4624 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4625 if (IS_ERR(sbi->s_chksum_driver)) {
4626 int ret = PTR_ERR(sbi->s_chksum_driver);
4627 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4628 sbi->s_chksum_driver = NULL;
4629 return ret;
4630 }
4631
4632 /* Check superblock checksum */
4633 if (!ext4_superblock_csum_verify(sb, es)) {
4634 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4635 "invalid superblock checksum. Run e2fsck?");
4636 return -EFSBADCRC;
4637 }
4638
4639 /* Precompute checksum seed for all metadata */
4640 if (ext4_has_feature_csum_seed(sb))
4641 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4642 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4643 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4644 sizeof(es->s_uuid));
4645 return 0;
4646}
4647
d7f3542b
JY		 4648static int ext4_check_feature_compatibility(struct super_block *sb,
4649 struct ext4_super_block *es,
4650 int silent)
4651{
54902099
JY		 4652 struct ext4_sb_info *sbi = EXT4_SB(sb);
4653
d7f3542b
JY		 4654 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4655 (ext4_has_compat_features(sb) ||
4656 ext4_has_ro_compat_features(sb) ||
4657 ext4_has_incompat_features(sb)))
4658 ext4_msg(sb, KERN_WARNING,
4659 "feature flags set on rev 0 fs, "
4660 "running e2fsck is recommended");
4661
4662 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4663 set_opt2(sb, HURD_COMPAT);
4664 if (ext4_has_feature_64bit(sb)) {
4665 ext4_msg(sb, KERN_ERR,
4666 "The Hurd can't support 64-bit file systems");
4667 return -EINVAL;
4668 }
4669
4670 /*
4671 * ea_inode feature uses l_i_version field which is not
4672 * available in HURD_COMPAT mode.
4673 */
4674 if (ext4_has_feature_ea_inode(sb)) {
4675 ext4_msg(sb, KERN_ERR,
4676 "ea_inode feature is not supported for Hurd");
4677 return -EINVAL;
4678 }
4679 }
4680
4681 if (IS_EXT2_SB(sb)) {
4682 if (ext2_feature_set_ok(sb))
4683 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4684 "using the ext4 subsystem");
4685 else {
4686 /*
4687 * If we're probing be silent, if this looks like
4688 * it's actually an ext[34] filesystem.
4689 */
4690 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4691 return -EINVAL;
4692 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4693 "to feature incompatibilities");
4694 return -EINVAL;
4695 }
4696 }
4697
4698 if (IS_EXT3_SB(sb)) {
4699 if (ext3_feature_set_ok(sb))
4700 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4701 "using the ext4 subsystem");
4702 else {
4703 /*
4704 * If we're probing be silent, if this looks like
4705 * it's actually an ext4 filesystem.
4706 */
4707 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4708 return -EINVAL;
4709 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4710 "to feature incompatibilities");
4711 return -EINVAL;
4712 }
4713 }
4714
4715 /*
4716 * Check feature flags regardless of the revision level, since we
4717 * previously didn't change the revision level when setting the flags,
4718 * so there is a chance incompat flags are set on a rev 0 filesystem.
4719 */
4720 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4721 return -EINVAL;
4722
54902099
JY		 4723 if (sbi->s_daxdev) {
4724 if (sb->s_blocksize == PAGE_SIZE)
4725 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4726 else
4727 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
4728 }
4729
4730 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4731 if (ext4_has_feature_inline_data(sb)) {
4732 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4733 " that may contain inline data");
4734 return -EINVAL;
4735 }
4736 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4737 ext4_msg(sb, KERN_ERR,
4738 "DAX unsupported by block device.");
4739 return -EINVAL;
4740 }
4741 }
4742
4743 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4744 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4745 es->s_encryption_level);
4746 return -EINVAL;
4747 }
4748
d7f3542b
JY		 4749 return 0;
4750}
4751
68e62439 4752static int ext4_check_geometry(struct super_block *sb,
bc62dbf9
JY		 4753 struct ext4_super_block *es)
4754{
4755 struct ext4_sb_info *sbi = EXT4_SB(sb);
4756 __u64 blocks_count;
269e9226
JY		 4757 int err;
4758
4759 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
4760 ext4_msg(sb, KERN_ERR,
4761 "Number of reserved GDT blocks insanely large: %d",
4762 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4763 return -EINVAL;
4764 }
4765 /*
4766 * Test whether we have more sectors than will fit in sector_t,
4767 * and whether the max offset is addressable by the page cache.
4768 */
4769 err = generic_check_addressable(sb->s_blocksize_bits,
4770 ext4_blocks_count(es));
4771 if (err) {
4772 ext4_msg(sb, KERN_ERR, "filesystem"
4773 " too large to mount safely on this system");
4774 return err;
4775 }
bc62dbf9
JY		 4776
4777 /* check blocks count against device size */
4778 blocks_count = sb_bdev_nr_blocks(sb);
4779 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4780 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4781 "exceeds size of device (%llu blocks)",
4782 ext4_blocks_count(es), blocks_count);
4783 return -EINVAL;
4784 }
4785
4786 /*
4787 * It makes no sense for the first data block to be beyond the end
4788 * of the filesystem.
4789 */
4790 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4791 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4792 "block %u is beyond end of filesystem (%llu)",
4793 le32_to_cpu(es->s_first_data_block),
4794 ext4_blocks_count(es));
4795 return -EINVAL;
4796 }
4797 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4798 (sbi->s_cluster_ratio == 1)) {
4799 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4800 "block is 0 with a 1k block and cluster size");
4801 return -EINVAL;
4802 }
4803
4804 blocks_count = (ext4_blocks_count(es) -
4805 le32_to_cpu(es->s_first_data_block) +
4806 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4807 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4808 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4809 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4810 "(block count %llu, first data block %u, "
4811 "blocks per group %lu)", blocks_count,
4812 ext4_blocks_count(es),
4813 le32_to_cpu(es->s_first_data_block),
4814 EXT4_BLOCKS_PER_GROUP(sb));
4815 return -EINVAL;
4816 }
4817 sbi->s_groups_count = blocks_count;
4818 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4819 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4820 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4821 le32_to_cpu(es->s_inodes_count)) {
4822 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4823 le32_to_cpu(es->s_inodes_count),
4824 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4825 return -EINVAL;
4826 }
4827
4828 return 0;
4829}
4830
a4e6a511
JY		 4831static int ext4_group_desc_init(struct super_block *sb,
4832 struct ext4_super_block *es,
4833 ext4_fsblk_t logical_sb_block,
4834 ext4_group_t *first_not_zeroed)
4835{
4836 struct ext4_sb_info *sbi = EXT4_SB(sb);
4837 unsigned int db_count;
4838 ext4_fsblk_t block;
a4e6a511
JY		 4839 int i;
4840
4841 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4842 EXT4_DESC_PER_BLOCK(sb);
4843 if (ext4_has_feature_meta_bg(sb)) {
4844 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4845 ext4_msg(sb, KERN_WARNING,
4846 "first meta block group too large: %u "
4847 "(group descriptor block count %u)",
4848 le32_to_cpu(es->s_first_meta_bg), db_count);
4849 return -EINVAL;
4850 }
4851 }
4852 rcu_assign_pointer(sbi->s_group_desc,
4853 kvmalloc_array(db_count,
4854 sizeof(struct buffer_head *),
4855 GFP_KERNEL));
4856 if (sbi->s_group_desc == NULL) {
4857 ext4_msg(sb, KERN_ERR, "not enough memory");
4858 return -ENOMEM;
4859 }
4860
4861 bgl_lock_init(sbi->s_blockgroup_lock);
4862
4863 /* Pre-read the descriptors into the buffer cache */
4864 for (i = 0; i < db_count; i++) {
4865 block = descriptor_loc(sb, logical_sb_block, i);
4866 ext4_sb_breadahead_unmovable(sb, block);
4867 }
4868
4869 for (i = 0; i < db_count; i++) {
4870 struct buffer_head *bh;
4871
4872 block = descriptor_loc(sb, logical_sb_block, i);
4873 bh = ext4_sb_bread_unmovable(sb, block);
4874 if (IS_ERR(bh)) {
4875 ext4_msg(sb, KERN_ERR,
4876 "can't read group descriptor %d", i);
4877 sbi->s_gdb_count = i;
172e344e 4878 return PTR_ERR(bh);
a4e6a511
JY		 4879 }
4880 rcu_read_lock();
4881 rcu_dereference(sbi->s_group_desc)[i] = bh;
4882 rcu_read_unlock();
4883 }
4884 sbi->s_gdb_count = db_count;
4885 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) {
4886 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
172e344e 4887 return -EFSCORRUPTED;
a4e6a511 4888 }
172e344e 4889
a4e6a511 4890 return 0;
a4e6a511
JY		 4891}
4892
9c1dd22d
JY		 4893static int ext4_load_and_init_journal(struct super_block *sb,
4894 struct ext4_super_block *es,
4895 struct ext4_fs_context *ctx)
4896{
4897 struct ext4_sb_info *sbi = EXT4_SB(sb);
4898 int err;
4899
4900 err = ext4_load_journal(sb, es, ctx->journal_devnum);
4901 if (err)
4902 return err;
4903
4904 if (ext4_has_feature_64bit(sb) &&
4905 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4906 JBD2_FEATURE_INCOMPAT_64BIT)) {
4907 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4908 goto out;
4909 }
4910
4911 if (!set_journal_csum_feature_set(sb)) {
4912 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4913 "feature set");
4914 goto out;
4915 }
4916
4917 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4918 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4919 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4920 ext4_msg(sb, KERN_ERR,
4921 "Failed to set fast commit journal feature");
4922 goto out;
4923 }
4924
4925 /* We have now updated the journal if required, so we can
4926 * validate the data journaling mode. */
4927 switch (test_opt(sb, DATA_FLAGS)) {
4928 case 0:
4929 /* No mode set, assume a default based on the journal
4930 * capabilities: ORDERED_DATA if the journal can
4931 * cope, else JOURNAL_DATA
4932 */
4933 if (jbd2_journal_check_available_features
4934 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4935 set_opt(sb, ORDERED_DATA);
4936 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4937 } else {
4938 set_opt(sb, JOURNAL_DATA);
4939 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4940 }
4941 break;
4942
4943 case EXT4_MOUNT_ORDERED_DATA:
4944 case EXT4_MOUNT_WRITEBACK_DATA:
4945 if (!jbd2_journal_check_available_features
4946 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4947 ext4_msg(sb, KERN_ERR, "Journal does not support "
4948 "requested data journaling mode");
4949 goto out;
4950 }
4951 break;
4952 default:
4953 break;
4954 }
4955
4956 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4957 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4958 ext4_msg(sb, KERN_ERR, "can't mount with "
4959 "journal_async_commit in data=ordered mode");
4960 goto out;
4961 }
4962
4963 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
4964
4965 sbi->s_journal->j_submit_inode_data_buffers =
4966 ext4_journal_submit_inode_data_buffers;
4967 sbi->s_journal->j_finish_inode_data_buffers =
4968 ext4_journal_finish_inode_data_buffers;
4969
4970 return 0;
4971
4972out:
4973 /* flush s_error_work before journal destroy. */
4974 flush_work(&sbi->s_error_work);
4975 jbd2_journal_destroy(sbi->s_journal);
4976 sbi->s_journal = NULL;
9f2a1d9f 4977 return -EINVAL;
9c1dd22d
JY		 4978}
4979
68e62439 4980static int ext4_check_journal_data_mode(struct super_block *sb)
a5991e53
JY		 4981{
4982 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4983 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4984 "data=journal disables delayed allocation, "
4985 "dioread_nolock, O_DIRECT and fast_commit support!\n");
4986 /* can't mount with both data=journal and dioread_nolock. */
4987 clear_opt(sb, DIOREAD_NOLOCK);
4988 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4989 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4990 ext4_msg(sb, KERN_ERR, "can't mount with "
4991 "both data=journal and delalloc");
4992 return -EINVAL;
4993 }
4994 if (test_opt(sb, DAX_ALWAYS)) {
4995 ext4_msg(sb, KERN_ERR, "can't mount with "
4996 "both data=journal and dax");
4997 return -EINVAL;
4998 }
4999 if (ext4_has_feature_encrypt(sb)) {
5000 ext4_msg(sb, KERN_WARNING,
5001 "encrypted files will use data=ordered "
5002 "instead of data journaling mode");
5003 }
5004 if (test_opt(sb, DELALLOC))
5005 clear_opt(sb, DELALLOC);
5006 } else {
5007 sb->s_iflags |= SB_I_CGROUPWB;
5008 }
5009
5010 return 0;
5011}
5012
a7a79c29
JY		 5013static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
5014 int silent)
ac27a0ec 5015{
7edfd85b 5016 struct ext4_sb_info *sbi = EXT4_SB(sb);
a7a79c29 5017 struct ext4_super_block *es;
70bbb3e0 5018 ext4_fsblk_t logical_sb_block;
ac27a0ec 5019 unsigned long offset = 0;
a7a79c29
JY		 5020 struct buffer_head *bh;
5021 int ret = -EINVAL;
4a8557b0 5022 int blocksize;
705895b6 5023
617ba13b 5024 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 5025 if (!blocksize) {
b31e1552 5026 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
a7a79c29 5027 return -EINVAL;
ac27a0ec
DK		 5028 }
5029
5030 /*
617ba13b 5031 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 5032 * block sizes. We need to calculate the offset from buffer start.
5033 */
617ba13b 5034 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
7edfd85b 5035 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
70bbb3e0 5036 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 5037 } else {
7edfd85b 5038 logical_sb_block = sbi->s_sb_block;
ac27a0ec
DK		 5039 }
5040
8394a6ab 5041 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5042 if (IS_ERR(bh)) {
b31e1552 5043 ext4_msg(sb, KERN_ERR, "unable to read superblock");
a7a79c29 5044 return PTR_ERR(bh);
ac27a0ec
DK		 5045 }
5046 /*
5047 * Note: s_es must be initialized as soon as possible because
617ba13b 5048 * some ext4 macro-instructions depend on its value
ac27a0ec 5049 */
2716b802 5050 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 5051 sbi->s_es = es;
5052 sb->s_magic = le16_to_cpu(es->s_magic);
a5fc5119
JY		 5053 if (sb->s_magic != EXT4_SUPER_MAGIC) {
5054 if (!silent)
5055 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
a7a79c29 5056 goto out;
a5fc5119
JY		 5057 }
5058
a7a79c29
JY		 5059 if (le32_to_cpu(es->s_log_block_size) >
5060 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5061 ext4_msg(sb, KERN_ERR,
5062 "Invalid log block size: %u",
5063 le32_to_cpu(es->s_log_block_size));
5064 goto out;
5065 }
5066 if (le32_to_cpu(es->s_log_cluster_size) >
5067 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5068 ext4_msg(sb, KERN_ERR,
5069 "Invalid log cluster size: %u",
5070 le32_to_cpu(es->s_log_cluster_size));
5071 goto out;
5072 }
5073
5074 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
5075
5076 /*
5077 * If the default block size is not the same as the real block size,
5078 * we need to reload it.
5079 */
5080 if (sb->s_blocksize == blocksize) {
5081 *lsb = logical_sb_block;
5082 sbi->s_sbh = bh;
5083 return 0;
5084 }
5085
5086 /*
5087 * bh must be released before kill_bdev(), otherwise
5088 * it won't be freed and its page also. kill_bdev()
5089 * is called by sb_set_blocksize().
5090 */
5091 brelse(bh);
5092 /* Validate the filesystem blocksize */
5093 if (!sb_set_blocksize(sb, blocksize)) {
5094 ext4_msg(sb, KERN_ERR, "bad block size %d",
5095 blocksize);
5096 bh = NULL;
5097 goto out;
5098 }
5099
5100 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5101 offset = do_div(logical_sb_block, blocksize);
5102 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5103 if (IS_ERR(bh)) {
5104 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5105 ret = PTR_ERR(bh);
5106 bh = NULL;
5107 goto out;
5108 }
5109 es = (struct ext4_super_block *)(bh->b_data + offset);
5110 sbi->s_es = es;
5111 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5112 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5113 goto out;
5114 }
5115 *lsb = logical_sb_block;
5116 sbi->s_sbh = bh;
5117 return 0;
5118out:
5119 brelse(bh);
5120 return ret;
5121}
5122
db9345d9
JY		 5123static void ext4_hash_info_init(struct super_block *sb)
5124{
5125 struct ext4_sb_info *sbi = EXT4_SB(sb);
5126 struct ext4_super_block *es = sbi->s_es;
5127 unsigned int i;
5128
5129 for (i = 0; i < 4; i++)
5130 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5131
5132 sbi->s_def_hash_version = es->s_def_hash_version;
5133 if (ext4_has_feature_dir_index(sb)) {
5134 i = le32_to_cpu(es->s_flags);
5135 if (i & EXT2_FLAGS_UNSIGNED_HASH)
5136 sbi->s_hash_unsigned = 3;
5137 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
5138#ifdef __CHAR_UNSIGNED__
5139 if (!sb_rdonly(sb))
5140 es->s_flags |=
5141 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5142 sbi->s_hash_unsigned = 3;
5143#else
5144 if (!sb_rdonly(sb))
5145 es->s_flags |=
5146 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
5147#endif
5148 }
5149 }
5150}
5151
107d2be9
JY		 5152static int ext4_block_group_meta_init(struct super_block *sb, int silent)
5153{
5154 struct ext4_sb_info *sbi = EXT4_SB(sb);
5155 struct ext4_super_block *es = sbi->s_es;
5156 int has_huge_files;
5157
5158 has_huge_files = ext4_has_feature_huge_file(sb);
5159 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
5160 has_huge_files);
5161 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
5162
5163 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
5164 if (ext4_has_feature_64bit(sb)) {
5165 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
5166 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
5167 !is_power_of_2(sbi->s_desc_size)) {
5168 ext4_msg(sb, KERN_ERR,
5169 "unsupported descriptor size %lu",
5170 sbi->s_desc_size);
5171 return -EINVAL;
5172 }
5173 } else
5174 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
5175
5176 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
5177 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
5178
5179 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
5180 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5181 if (!silent)
5182 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5183 return -EINVAL;
5184 }
5185 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
5186 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
5187 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
5188 sbi->s_inodes_per_group);
5189 return -EINVAL;
5190 }
5191 sbi->s_itb_per_group = sbi->s_inodes_per_group /
5192 sbi->s_inodes_per_block;
5193 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
5194 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
5195 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5196 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
5197
5198 return 0;
5199}
5200
a7a79c29
JY		 5201static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5202{
5203 struct ext4_super_block *es = NULL;
5204 struct ext4_sb_info *sbi = EXT4_SB(sb);
a7a79c29
JY		 5205 ext4_fsblk_t logical_sb_block;
5206 struct inode *root;
107d2be9 5207 int needs_recovery;
d4fab7b2 5208 int err;
a7a79c29
JY		 5209 ext4_group_t first_not_zeroed;
5210 struct ext4_fs_context *ctx = fc->fs_private;
5211 int silent = fc->sb_flags & SB_SILENT;
5212
5213 /* Set defaults for the variables that will be set during parsing */
5214 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5215 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5216
5217 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5218 sbi->s_sectors_written_start =
5219 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5220
a7a79c29
JY		 5221 err = ext4_load_super(sb, &logical_sb_block, silent);
5222 if (err)
5223 goto out_fail;
5224
5225 es = sbi->s_es;
afc32f7e 5226 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 5227
b26458d1
JY		 5228 err = ext4_init_metadata_csum(sb, es);
5229 if (err)
a5fc5119 5230 goto failed_mount;
a9c47317 5231
5f6d662d 5232 ext4_set_def_opts(sb, es);
ac27a0ec 5233
08cefc7a
EB		 5234 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
5235 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 5236 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5237 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5238 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 5239
51ce6511
LC		 5240 /*
5241 * set default s_li_wait_mult for lazyinit, for the case there is
5242 * no mount option specified.
5243 */
5244 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5245
d4fab7b2
TT		 5246 err = ext4_inode_info_init(sb, es);
5247 if (err)
0e495f7c 5248 goto failed_mount;
9803387c 5249
7edfd85b
LC		 5250 err = parse_apply_sb_mount_options(sb, ctx);
5251 if (err < 0)
5252 goto failed_mount;
5253
5a916be1 5254 sbi->s_def_mount_opt = sbi->s_mount_opt;
e3645d72 5255 sbi->s_def_mount_opt2 = sbi->s_mount_opt2;
7edfd85b
LC		 5256
5257 err = ext4_check_opt_consistency(fc, sb);
5258 if (err < 0)
5259 goto failed_mount;
5260
85456054 5261 ext4_apply_options(fc, sb);
ac27a0ec 5262
d4fab7b2
TT		 5263 err = ext4_encoding_init(sb, es);
5264 if (err)
39c135b0 5265 goto failed_mount;
c83ad55e 5266
d4fab7b2
TT		 5267 err = ext4_check_journal_data_mode(sb);
5268 if (err)
a5991e53 5269 goto failed_mount;
56889787 5270
1751e8a6
LT		 5271 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5272 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 5273
1ff20307
JL		 5274 /* i_version is always enabled now */
5275 sb->s_flags |= SB_I_VERSION;
5276
d4fab7b2
TT		 5277 err = ext4_check_feature_compatibility(sb, es, silent);
5278 if (err)
ac27a0ec 5279 goto failed_mount;
a13fb1a4 5280
d4fab7b2
TT		 5281 err = ext4_block_group_meta_init(sb, silent);
5282 if (err)
5b9554dc 5283 goto failed_mount;
5b9554dc 5284
db9345d9 5285 ext4_hash_info_init(sb);
ac27a0ec 5286
d4fab7b2
TT		 5287 err = ext4_handle_clustersize(sb);
5288 if (err)
4a8557b0 5289 goto failed_mount;
960fd856 5290
d4fab7b2
TT		 5291 err = ext4_check_geometry(sb, es);
5292 if (err)
0f2ddca6 5293 goto failed_mount;
0f2ddca6 5294
235699a8 5295 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
c92dc856
JK		 5296 spin_lock_init(&sbi->s_error_lock);
5297 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
04496411 5298
172e344e
YB		 5299 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed);
5300 if (err)
5301 goto failed_mount3;
5302
d4fab7b2
TT		 5303 err = ext4_es_register_shrinker(sbi);
5304 if (err)
ce7e010a 5305 goto failed_mount3;
ce7e010a 5306
c9de560d 5307 sbi->s_stripe = ext4_get_stripe_size(sbi);
c3defd99
KS		 5308 /*
5309 * It's hard to get stripe aligned blocks if stripe is not aligned with
5310 * cluster, just disable stripe and alert user to simpfy code and avoid
5311 * stripe aligned allocation which will rarely successes.
5312 */
5313 if (sbi->s_stripe > 0 && sbi->s_cluster_ratio > 1 &&
5314 sbi->s_stripe % sbi->s_cluster_ratio != 0) {
5315 ext4_msg(sb, KERN_WARNING,
5316 "stripe (%lu) is not aligned with cluster size (%u), "
5317 "stripe is disabled",
5318 sbi->s_stripe, sbi->s_cluster_ratio);
5319 sbi->s_stripe = 0;
5320 }
67a5da56 5321 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 5322
f9ae9cf5
TT		 5323 /*
5324 * set up enough so that it can read an inode
5325 */
f6e63f90 5326 sb->s_op = &ext4_sops;
617ba13b
MC		 5327 sb->s_export_op = &ext4_export_ops;
5328 sb->s_xattr = ext4_xattr_handlers;
643fa961 5329#ifdef CONFIG_FS_ENCRYPTION
a7550b30 5330 sb->s_cop = &ext4_cryptops;
ffcc4182 5331#endif
c93d8f88
EB		 5332#ifdef CONFIG_FS_VERITY
5333 sb->s_vop = &ext4_verityops;
5334#endif
ac27a0ec 5335#ifdef CONFIG_QUOTA
617ba13b 5336 sb->dq_op = &ext4_quota_operations;
e2b911c5 5337 if (ext4_has_feature_quota(sb))
1fa5efe3 5338 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 5339 else
5340 sb->s_qcop = &ext4_qctl_operations;
689c958c 5341 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 5342#endif
85787090 5343 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
f2fa2ffc 5344
ac27a0ec 5345 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 5346 mutex_init(&sbi->s_orphan_lock);
ac27a0ec 5347
f7314a67 5348 ext4_fast_commit_init(sb);
aa75f4d3 5349
ac27a0ec
DK		 5350 sb->s_root = NULL;
5351
5352 needs_recovery = (es->s_last_orphan != 0 ||
02f310fc 5353 ext4_has_feature_orphan_present(sb) ||
e2b911c5 5354 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 5355
3b50d501
TT		 5356 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) {
5357 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block));
5358 if (err)
50460fe8 5359 goto failed_mount3a;
3b50d501 5360 }
c5e06d10 5361
d4fab7b2 5362 err = -EINVAL;
ac27a0ec
DK		 5363 /*
5364 * The first inode we look at is the journal inode. Don't try
5365 * root first: it may be modified in the journal!
5366 */
e2b911c5 5367 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
9c1dd22d 5368 err = ext4_load_and_init_journal(sb, es, ctx);
4753d8a2 5369 if (err)
50460fe8 5370 goto failed_mount3a;
bc98a42c 5371 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 5372 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 5373 ext4_msg(sb, KERN_ERR, "required journal recovery "
5374 "suppressed and not mounted read-only");
43bd6f1b 5375 goto failed_mount3a;
ac27a0ec 5376 } else {
1e381f60 5377 /* Nojournal mode, all journal mount options are illegal */
89481b5f 5378 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1e381f60 5379 ext4_msg(sb, KERN_ERR, "can't mount with "
89481b5f 5380 "journal_async_commit, fs mounted w/o journal");
43bd6f1b 5381 goto failed_mount3a;
1e381f60 5382 }
89481b5f
BL		 5383
5384 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
1e381f60 5385 ext4_msg(sb, KERN_ERR, "can't mount with "
89481b5f 5386 "journal_checksum, fs mounted w/o journal");
43bd6f1b 5387 goto failed_mount3a;
1e381f60
DM		 5388 }
5389 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5390 ext4_msg(sb, KERN_ERR, "can't mount with "
5391 "commit=%lu, fs mounted w/o journal",
5392 sbi->s_commit_interval / HZ);
43bd6f1b 5393 goto failed_mount3a;
1e381f60
DM		 5394 }
5395 if (EXT4_MOUNT_DATA_FLAGS &
5396 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5397 ext4_msg(sb, KERN_ERR, "can't mount with "
5398 "data=, fs mounted w/o journal");
43bd6f1b 5399 goto failed_mount3a;
1e381f60 5400 }
50b29d8f 5401 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
1e381f60 5402 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 5403 clear_opt(sb, DATA_FLAGS);
995a3ed6 5404 clear_opt2(sb, JOURNAL_FAST_COMMIT);
0390131b
FM		 5405 sbi->s_journal = NULL;
5406 needs_recovery = 0;
ac27a0ec
DK		 5407 }
5408
cdb7ee4c
TE		 5409 if (!test_opt(sb, NO_MBCACHE)) {
5410 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5411 if (!sbi->s_ea_block_cache) {
dec214d0 5412 ext4_msg(sb, KERN_ERR,
cdb7ee4c 5413 "Failed to create ea_block_cache");
d4fab7b2 5414 err = -EINVAL;
dec214d0
TE		 5415 goto failed_mount_wq;
5416 }
cdb7ee4c
TE		 5417
5418 if (ext4_has_feature_ea_inode(sb)) {
5419 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5420 if (!sbi->s_ea_inode_cache) {
5421 ext4_msg(sb, KERN_ERR,
5422 "Failed to create ea_inode_cache");
d4fab7b2 5423 err = -EINVAL;
cdb7ee4c
TE		 5424 goto failed_mount_wq;
5425 }
5426 }
9c191f70
M		 5427 }
5428
952fc18e
TT		 5429 /*
5430 * Get the # of file system overhead blocks from the
5431 * superblock if present.
5432 */
85d825db
TT		 5433 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5434 /* ignore the precalculated value if it is ridiculous */
5435 if (sbi->s_overhead > ext4_blocks_count(es))
5436 sbi->s_overhead = 0;
5437 /*
5438 * If the bigalloc feature is not enabled recalculating the
5439 * overhead doesn't take long, so we might as well just redo
5440 * it to make sure we are using the correct value.
5441 */
5442 if (!ext4_has_feature_bigalloc(sb))
5443 sbi->s_overhead = 0;
5444 if (sbi->s_overhead == 0) {
07aa2ea1
LC		 5445 err = ext4_calculate_overhead(sb);
5446 if (err)
952fc18e
TT		 5447 goto failed_mount_wq;
5448 }
5449
fd89d5f2
TH		 5450 /*
5451 * The maximum number of concurrent works can be high and
5452 * concurrency isn't really necessary. Limit it to 1.
5453 */
2e8fa54e
JK		 5454 EXT4_SB(sb)->rsv_conversion_wq =
5455 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5456 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5457 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
d4fab7b2 5458 err = -ENOMEM;
2e8fa54e
JK		 5459 goto failed_mount4;
5460 }
5461
ac27a0ec 5462 /*
dab291af 5463 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 5464 * so we can safely mount the rest of the filesystem now.
5465 */
5466
8a363970 5467 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
1d1fe1ee 5468 if (IS_ERR(root)) {
b31e1552 5469 ext4_msg(sb, KERN_ERR, "get root inode failed");
d4fab7b2 5470 err = PTR_ERR(root);
32a9bb57 5471 root = NULL;
ac27a0ec
DK		 5472 goto failed_mount4;
5473 }
5474 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 5475 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 5476 iput(root);
d4fab7b2 5477 err = -EFSCORRUPTED;
ac27a0ec
DK		 5478 goto failed_mount4;
5479 }
b886ee3e 5480
48fde701 5481 sb->s_root = d_make_root(root);
1d1fe1ee 5482 if (!sb->s_root) {
b31e1552 5483 ext4_msg(sb, KERN_ERR, "get root dentry failed");
d4fab7b2 5484 err = -ENOMEM;
1d1fe1ee
DH		 5485 goto failed_mount4;
5486 }
ac27a0ec 5487
d4fab7b2
TT		 5488 err = ext4_setup_super(sb, es, sb_rdonly(sb));
5489 if (err == -EROFS) {
1751e8a6 5490 sb->s_flags |= SB_RDONLY;
d4fab7b2 5491 } else if (err)
c89128a0 5492 goto failed_mount4a;
ef7f3835 5493
b5799018 5494 ext4_set_resv_clusters(sb);
27dd4385 5495
0f5bde1d
JK		 5496 if (test_opt(sb, BLOCK_VALIDITY)) {
5497 err = ext4_setup_system_zone(sb);
5498 if (err) {
5499 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5500 "zone (%d)", err);
5501 goto failed_mount4a;
5502 }
f9ae9cf5 5503 }
8016e29f 5504 ext4_fc_replay_cleanup(sb);
f9ae9cf5
TT		 5505
5506 ext4_ext_init(sb);
196e402a
HS		 5507
5508 /*
5509 * Enable optimize_scan if number of groups is > threshold. This can be
5510 * turned off by passing "mb_optimize_scan=0". This can also be
5511 * turned on forcefully by passing "mb_optimize_scan=1".
5512 */
27b38686
OM		 5513 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5514 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5515 set_opt2(sb, MB_OPTIMIZE_SCAN);
5516 else
5517 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5518 }
196e402a 5519
f9ae9cf5
TT		 5520 err = ext4_mb_init(sb);
5521 if (err) {
5522 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5523 err);
dcf2d804 5524 goto failed_mount5;
c2774d84
AK		 5525 }
5526
027f14f5
TT		 5527 /*
5528 * We can only set up the journal commit callback once
5529 * mballoc is initialized
5530 */
5531 if (sbi->s_journal)
5532 sbi->s_journal->j_commit_callback =
5533 ext4_journal_commit_callback;
5534
d5e72c4e
TT		 5535 err = ext4_percpu_param_init(sbi);
5536 if (err)
d5e03cbb 5537 goto failed_mount6;
d5e03cbb 5538
e2b911c5 5539 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 5540 if (!ext4_fill_flex_info(sb)) {
5541 ext4_msg(sb, KERN_ERR,
5542 "unable to initialize "
5543 "flex_bg meta info!");
d4fab7b2 5544 err = -ENOMEM;
d5e03cbb
TT		 5545 goto failed_mount6;
5546 }
5547
bfff6873
LC		 5548 err = ext4_register_li_request(sb, first_not_zeroed);
5549 if (err)
dcf2d804 5550 goto failed_mount6;
bfff6873 5551
b5799018 5552 err = ext4_register_sysfs(sb);
dcf2d804
TM		 5553 if (err)
5554 goto failed_mount7;
3197ebdb 5555
02f310fc
JK		 5556 err = ext4_init_orphan_info(sb);
5557 if (err)
5558 goto failed_mount8;
9b2ff357
JK		 5559#ifdef CONFIG_QUOTA
5560 /* Enable quota usage during mount. */
bc98a42c 5561 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 5562 err = ext4_enable_quotas(sb);
5563 if (err)
02f310fc 5564 goto failed_mount9;
9b2ff357
JK		 5565 }
5566#endif /* CONFIG_QUOTA */
5567
bc71726c 5568 /*
5569 * Save the original bdev mapping's wb_err value which could be
5570 * used to detect the metadata async write error.
5571 */
5572 spin_lock_init(&sbi->s_bdev_wb_lock);
9704a322
ZX		 5573 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5574 &sbi->s_bdev_wb_err);
bc71726c 5575 sb->s_bdev->bd_super = sb;
617ba13b
MC		 5576 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5577 ext4_orphan_cleanup(sb, es);
5578 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
9b6641dd
YB		 5579 /*
5580 * Update the checksum after updating free space/inode counters and
5581 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5582 * checksum in the buffer cache until it is written out and
5583 * e2fsprogs programs trying to open a file system immediately
5584 * after it is mounted can fail.
5585 */
5586 ext4_superblock_csum_set(sb);
0390131b 5587 if (needs_recovery) {
b31e1552 5588 ext4_msg(sb, KERN_INFO, "recovery complete");
11215630
JK		 5589 err = ext4_mark_recovery_complete(sb, es);
5590 if (err)
d13f9963 5591 goto failed_mount10;
0390131b 5592 }
0390131b 5593
70200574
CH		 5594 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5595 ext4_msg(sb, KERN_WARNING,
5596 "mounting with \"discard\" option, but the device does not support discard");
79add3a3 5597
66e61a9e
TT		 5598 if (es->s_error_count)
5599 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 5600
efbed4dc
TT		 5601 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5602 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5603 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5604 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
1cf006ed
DM		 5605 atomic_set(&sbi->s_warning_count, 0);
5606 atomic_set(&sbi->s_msg_count, 0);
efbed4dc 5607
ac27a0ec
DK		 5608 return 0;
5609
d13f9963 5610failed_mount10:
f3c1c42e 5611 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
d13f9963 5612failed_mount9: __maybe_unused
02f310fc 5613 ext4_release_orphan_info(sb);
72ba7450 5614failed_mount8:
ebd173be 5615 ext4_unregister_sysfs(sb);
cb8d53d2 5616 kobject_put(&sbi->s_kobj);
dcf2d804
TM		 5617failed_mount7:
5618 ext4_unregister_li_request(sb);
5619failed_mount6:
f9ae9cf5 5620 ext4_mb_release(sb);
dcbf8758 5621 ext4_flex_groups_free(sbi);
1f79467c 5622 ext4_percpu_param_destroy(sbi);
00764937 5623failed_mount5:
f9ae9cf5
TT		 5624 ext4_ext_release(sb);
5625 ext4_release_system_zone(sb);
5626failed_mount4a:
94bf608a 5627 dput(sb->s_root);
32a9bb57 5628 sb->s_root = NULL;
94bf608a 5629failed_mount4:
b31e1552 5630 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 5631 if (EXT4_SB(sb)->rsv_conversion_wq)
5632 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 5633failed_mount_wq:
50c15df6
CX		 5634 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5635 sbi->s_ea_inode_cache = NULL;
5636
5637 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5638 sbi->s_ea_block_cache = NULL;
5639
0390131b 5640 if (sbi->s_journal) {
bb9464e0 5641 /* flush s_error_work before journal destroy. */
5642 flush_work(&sbi->s_error_work);
0390131b
FM		 5643 jbd2_journal_destroy(sbi->s_journal);
5644 sbi->s_journal = NULL;
5645 }
50460fe8 5646failed_mount3a:
d3922a77 5647 ext4_es_unregister_shrinker(sbi);
eb68d0e2 5648failed_mount3:
bb9464e0 5649 /* flush s_error_work before sbi destroy */
c92dc856 5650 flush_work(&sbi->s_error_work);
2a4ae3bc 5651 del_timer_sync(&sbi->s_err_report);
618f0031 5652 ext4_stop_mmpd(sbi);
a4e6a511 5653 ext4_group_desc_free(sbi);
ac27a0ec 5654failed_mount:
0441984a
DW		 5655 if (sbi->s_chksum_driver)
5656 crypto_free_shash(sbi->s_chksum_driver);
c83ad55e 5657
5298d4bf 5658#if IS_ENABLED(CONFIG_UNICODE)
f8f4acb6 5659 utf8_unload(sb->s_encoding);
c83ad55e
GKB		 5660#endif
5661
ac27a0ec 5662#ifdef CONFIG_QUOTA
856dd6c5 5663 for (unsigned int i = 0; i < EXT4_MAXQUOTAS; i++)
0ba33fac 5664 kfree(get_qf_name(sb, sbi, i));
ac27a0ec 5665#endif
ac4acb1f 5666 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
afd09b61 5667 /* ext4_blkdev_remove() calls kill_bdev(), release bh before it. */
a7a79c29 5668 brelse(sbi->s_sbh);
afd09b61 5669 ext4_blkdev_remove(sbi);
ac27a0ec 5670out_fail:
26fb5290 5671 invalidate_bdev(sb->s_bdev);
ac27a0ec 5672 sb->s_fs_info = NULL;
d4fab7b2 5673 return err;
ac27a0ec
DK		 5674}
5675
cebe85d5 5676static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
7edfd85b 5677{
cebe85d5 5678 struct ext4_fs_context *ctx = fc->fs_private;
7edfd85b 5679 struct ext4_sb_info *sbi;
7edfd85b 5680 const char *descr;
cebe85d5 5681 int ret;
7edfd85b
LC		 5682
5683 sbi = ext4_alloc_sbi(sb);
cebe85d5 5684 if (!sbi)
7c268d4c 5685 return -ENOMEM;
7edfd85b 5686
cebe85d5
LC		 5687 fc->s_fs_info = sbi;
5688
5689 /* Cleanup superblock name */
5690 strreplace(sb->s_id, '/', '!');
7edfd85b
LC		 5691
5692 sbi->s_sb_block = 1; /* Default super block location */
cebe85d5
LC		 5693 if (ctx->spec & EXT4_SPEC_s_sb_block)
5694 sbi->s_sb_block = ctx->s_sb_block;
7edfd85b 5695
960e0ab6 5696 ret = __ext4_fill_super(fc, sb);
7edfd85b
LC		 5697 if (ret < 0)
5698 goto free_sbi;
5699
cebe85d5 5700 if (sbi->s_journal) {
7edfd85b
LC		 5701 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5702 descr = " journalled data mode";
5703 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5704 descr = " ordered data mode";
5705 else
5706 descr = " writeback data mode";
5707 } else
5708 descr = "out journal";
5709
5710 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
6dcc98fb
TT		 5711 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU %s with%s. "
5712 "Quota mode: %s.", &sb->s_uuid,
5713 sb_rdonly(sb) ? "ro" : "r/w", descr,
bb0fbc78 5714 ext4_quota_mode(sb));
cebe85d5 5715
eb705421 5716 /* Update the s_overhead_clusters if necessary */
827891a3 5717 ext4_update_overhead(sb, false);
7edfd85b 5718 return 0;
cebe85d5 5719
7edfd85b
LC		 5720free_sbi:
5721 ext4_free_sbi(sbi);
cebe85d5 5722 fc->s_fs_info = NULL;
7edfd85b
LC		 5723 return ret;
5724}
5725
cebe85d5
LC		 5726static int ext4_get_tree(struct fs_context *fc)
5727{
5728 return get_tree_bdev(fc, ext4_fill_super);
5729}
5730
ac27a0ec
DK		 5731/*
5732 * Setup any per-fs journal parameters now. We'll do this both on
5733 * initial mount, once the journal has been initialised but before we've
5734 * done any recovery; and again on any subsequent remount.
5735 */
617ba13b 5736static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 5737{
617ba13b 5738 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5739
30773840
TT		 5740 journal->j_commit_interval = sbi->s_commit_interval;
5741 journal->j_min_batch_time = sbi->s_min_batch_time;
5742 journal->j_max_batch_time = sbi->s_max_batch_time;
6866d7b3 5743 ext4_fc_init(sb, journal);
ac27a0ec 5744
a931da6a 5745 write_lock(&journal->j_state_lock);
ac27a0ec 5746 if (test_opt(sb, BARRIER))
dab291af 5747 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 5748 else
dab291af 5749 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 5750 if (test_opt(sb, DATA_ERR_ABORT))
5751 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5752 else
5753 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
72945058
ZY		 5754 /*
5755 * Always enable journal cycle record option, letting the journal
5756 * records log transactions continuously between each mount.
5757 */
5758 journal->j_flags |= JBD2_CYCLE_RECORD;
a931da6a 5759 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 5760}
5761
c6cb7e77
EW		 5762static struct inode *ext4_get_journal_inode(struct super_block *sb,
5763 unsigned int journal_inum)
ac27a0ec
DK		 5764{
5765 struct inode *journal_inode;
ac27a0ec 5766
c6cb7e77
EW		 5767 /*
5768 * Test for the existence of a valid inode on disk. Bad things
5769 * happen if we iget() an unused inode, as the subsequent iput()
5770 * will try to delete it.
5771 */
8a363970 5772 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
1d1fe1ee 5773 if (IS_ERR(journal_inode)) {
b31e1552 5774 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 5775 return NULL;
5776 }
5777 if (!journal_inode->i_nlink) {
5778 make_bad_inode(journal_inode);
5779 iput(journal_inode);
b31e1552 5780 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 5781 return NULL;
5782 }
5783
4978c659 5784 ext4_debug("Journal inode found at %p: %lld bytes\n",
ac27a0ec 5785 journal_inode, journal_inode->i_size);
105c78e1 5786 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
b31e1552 5787 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 5788 iput(journal_inode);
5789 return NULL;
5790 }
c6cb7e77
EW		 5791 return journal_inode;
5792}
5793
62913ae9
TT		 5794static int ext4_journal_bmap(journal_t *journal, sector_t *block)
5795{
5796 struct ext4_map_blocks map;
5797 int ret;
5798
5799 if (journal->j_inode == NULL)
5800 return 0;
5801
5802 map.m_lblk = *block;
5803 map.m_len = 1;
5804 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0);
5805 if (ret <= 0) {
5806 ext4_msg(journal->j_inode->i_sb, KERN_CRIT,
5807 "journal bmap failed: block %llu ret %d\n",
5808 *block, ret);
5809 jbd2_journal_abort(journal, ret ? ret : -EIO);
5810 return ret;
5811 }
5812 *block = map.m_pblk;
5813 return 0;
5814}
5815
c6cb7e77
EW		 5816static journal_t *ext4_get_journal(struct super_block *sb,
5817 unsigned int journal_inum)
5818{
5819 struct inode *journal_inode;
5820 journal_t *journal;
5821
11215630
JK		 5822 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5823 return NULL;
c6cb7e77
EW		 5824
5825 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5826 if (!journal_inode)
5827 return NULL;
ac27a0ec 5828
dab291af 5829 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 5830 if (!journal) {
b31e1552 5831 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 5832 iput(journal_inode);
5833 return NULL;
5834 }
5835 journal->j_private = sb;
62913ae9 5836 journal->j_bmap = ext4_journal_bmap;
617ba13b 5837 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 5838 return journal;
5839}
5840
617ba13b 5841static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 5842 dev_t j_dev)
5843{
2b2d6d01 5844 struct buffer_head *bh;
ac27a0ec 5845 journal_t *journal;
617ba13b
MC		 5846 ext4_fsblk_t start;
5847 ext4_fsblk_t len;
ac27a0ec 5848 int hblock, blocksize;
617ba13b 5849 ext4_fsblk_t sb_block;
ac27a0ec 5850 unsigned long offset;
2b2d6d01 5851 struct ext4_super_block *es;
ac27a0ec
DK		 5852 struct block_device *bdev;
5853
11215630
JK		 5854 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5855 return NULL;
0390131b 5856
b31e1552 5857 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 5858 if (bdev == NULL)
5859 return NULL;
5860
ac27a0ec 5861 blocksize = sb->s_blocksize;
e1defc4f 5862 hblock = bdev_logical_block_size(bdev);
ac27a0ec 5863 if (blocksize < hblock) {
b31e1552
ES		 5864 ext4_msg(sb, KERN_ERR,
5865 "blocksize too small for journal device");
ac27a0ec
DK		 5866 goto out_bdev;
5867 }
5868
617ba13b
MC		 5869 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5870 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 5871 set_blocksize(bdev, blocksize);
5872 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 5873 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5874 "external journal");
ac27a0ec
DK		 5875 goto out_bdev;
5876 }
5877
2716b802 5878 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 5879 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 5880 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 5881 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 5882 ext4_msg(sb, KERN_ERR, "external journal has "
5883 "bad superblock");
ac27a0ec
DK		 5884 brelse(bh);
5885 goto out_bdev;
5886 }
5887
df4763be
DW		 5888 if ((le32_to_cpu(es->s_feature_ro_compat) &
5889 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5890 es->s_checksum != ext4_superblock_csum(sb, es)) {
5891 ext4_msg(sb, KERN_ERR, "external journal has "
5892 "corrupt superblock");
5893 brelse(bh);
5894 goto out_bdev;
5895 }
5896
617ba13b 5897 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 5898 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 5899 brelse(bh);
5900 goto out_bdev;
5901 }
5902
bd81d8ee 5903 len = ext4_blocks_count(es);
ac27a0ec
DK		 5904 start = sb_block + 1;
5905 brelse(bh); /* we're done with the superblock */
5906
dab291af 5907 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 5908 start, len, blocksize);
5909 if (!journal) {
b31e1552 5910 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 5911 goto out_bdev;
5912 }
5913 journal->j_private = sb;
2d069c08 5914 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
b31e1552 5915 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 5916 goto out_journal;
5917 }
5918 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 5919 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5920 "user (unsupported) - %d",
ac27a0ec
DK		 5921 be32_to_cpu(journal->j_superblock->s_nr_users));
5922 goto out_journal;
5923 }
ee7ed3aa 5924 EXT4_SB(sb)->s_journal_bdev = bdev;
617ba13b 5925 ext4_init_journal_params(sb, journal);
ac27a0ec 5926 return journal;
0b8e58a1 5927
ac27a0ec 5928out_journal:
dab291af 5929 jbd2_journal_destroy(journal);
ac27a0ec 5930out_bdev:
2736e8ee 5931 blkdev_put(bdev, sb);
ac27a0ec
DK		 5932 return NULL;
5933}
5934
617ba13b
MC		 5935static int ext4_load_journal(struct super_block *sb,
5936 struct ext4_super_block *es,
ac27a0ec
DK		 5937 unsigned long journal_devnum)
5938{
5939 journal_t *journal;
5940 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5941 dev_t journal_dev;
5942 int err = 0;
5943 int really_read_only;
273108fa 5944 int journal_dev_ro;
ac27a0ec 5945
11215630
JK		 5946 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5947 return -EFSCORRUPTED;
0390131b 5948
ac27a0ec
DK		 5949 if (journal_devnum &&
5950 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 5951 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5952 "numbers have changed");
ac27a0ec
DK		 5953 journal_dev = new_decode_dev(journal_devnum);
5954 } else
5955 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5956
273108fa
LC		 5957 if (journal_inum && journal_dev) {
5958 ext4_msg(sb, KERN_ERR,
5959 "filesystem has both journal inode and journal device!");
5960 return -EINVAL;
5961 }
5962
5963 if (journal_inum) {
5964 journal = ext4_get_journal(sb, journal_inum);
5965 if (!journal)
5966 return -EINVAL;
5967 } else {
5968 journal = ext4_get_dev_journal(sb, journal_dev);
5969 if (!journal)
5970 return -EINVAL;
5971 }
5972
5973 journal_dev_ro = bdev_read_only(journal->j_dev);
5974 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5975
5976 if (journal_dev_ro && !sb_rdonly(sb)) {
5977 ext4_msg(sb, KERN_ERR,
5978 "journal device read-only, try mounting with '-o ro'");
5979 err = -EROFS;
5980 goto err_out;
5981 }
ac27a0ec
DK		 5982
5983 /*
5984 * Are we loading a blank journal or performing recovery after a
5985 * crash? For recovery, we need to check in advance whether we
5986 * can get read-write access to the device.
5987 */
e2b911c5 5988 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 5989 if (sb_rdonly(sb)) {
b31e1552
ES		 5990 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5991 "required on readonly filesystem");
ac27a0ec 5992 if (really_read_only) {
b31e1552 5993 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 5994 "unavailable, cannot proceed "
5995 "(try mounting with noload)");
273108fa
LC		 5996 err = -EROFS;
5997 goto err_out;
ac27a0ec 5998 }
b31e1552
ES		 5999 ext4_msg(sb, KERN_INFO, "write access will "
6000 "be enabled during recovery");
ac27a0ec
DK		 6001 }
6002 }
6003
90576c0b 6004 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 6005 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 6006
e2b911c5 6007 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 6008 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 6009 if (!err) {
6010 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
2ef6c32a
TT		 6011 __le16 orig_state;
6012 bool changed = false;
eee00237 6013
1c13d5c0
TT		 6014 if (save)
6015 memcpy(save, ((char *) es) +
6016 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 6017 err = jbd2_journal_load(journal);
2ef6c32a
TT		 6018 if (save && memcmp(((char *) es) + EXT4_S_ERR_START,
6019 save, EXT4_S_ERR_LEN)) {
1c13d5c0
TT		 6020 memcpy(((char *) es) + EXT4_S_ERR_START,
6021 save, EXT4_S_ERR_LEN);
2ef6c32a
TT		 6022 changed = true;
6023 }
1c13d5c0 6024 kfree(save);
2ef6c32a 6025 orig_state = es->s_state;
eee00237
YB		 6026 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state &
6027 EXT4_ERROR_FS);
2ef6c32a
TT		 6028 if (orig_state != es->s_state)
6029 changed = true;
eee00237 6030 /* Write out restored error information to the superblock */
2ef6c32a 6031 if (changed && !really_read_only) {
eee00237
YB		 6032 int err2;
6033 err2 = ext4_commit_super(sb);
6034 err = err ? : err2;
6035 }
1c13d5c0 6036 }
ac27a0ec
DK		 6037
6038 if (err) {
b31e1552 6039 ext4_msg(sb, KERN_ERR, "error loading journal");
273108fa 6040 goto err_out;
ac27a0ec
DK		 6041 }
6042
617ba13b 6043 EXT4_SB(sb)->s_journal = journal;
11215630
JK		 6044 err = ext4_clear_journal_err(sb, es);
6045 if (err) {
6046 EXT4_SB(sb)->s_journal = NULL;
6047 jbd2_journal_destroy(journal);
6048 return err;
6049 }
ac27a0ec 6050
c41303ce 6051 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 6052 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
6053 es->s_journal_dev = cpu_to_le32(journal_devnum);
3039d8b8
BL		 6054 ext4_commit_super(sb);
6055 }
6056 if (!really_read_only && journal_inum &&
6057 journal_inum != le32_to_cpu(es->s_journal_inum)) {
6058 es->s_journal_inum = cpu_to_le32(journal_inum);
4392fbc4 6059 ext4_commit_super(sb);
ac27a0ec
DK		 6060 }
6061
6062 return 0;
273108fa
LC		 6063
6064err_out:
6065 jbd2_journal_destroy(journal);
6066 return err;
ac27a0ec
DK		 6067}
6068
2d01ddc8
JK		 6069/* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
6070static void ext4_update_super(struct super_block *sb)
ac27a0ec 6071{
c92dc856 6072 struct ext4_sb_info *sbi = EXT4_SB(sb);
e92ad03f
JK		 6073 struct ext4_super_block *es = sbi->s_es;
6074 struct buffer_head *sbh = sbi->s_sbh;
a17712c8 6075
05c2c00f 6076 lock_buffer(sbh);
71290b36
TT		 6077 /*
6078 * If the file system is mounted read-only, don't update the
6079 * superblock write time. This avoids updating the superblock
6080 * write time when we are mounting the root file system
6081 * read/only but we need to replay the journal; at that point,
6082 * for people who are east of GMT and who make their clock
6083 * tick in localtime for Windows bug-for-bug compatibility,
6084 * the clock is set in the future, and this will cause e2fsck
6085 * to complain and force a full file system check.
6086 */
1751e8a6 6087 if (!(sb->s_flags & SB_RDONLY))
6a0678a7 6088 ext4_update_tstamp(es, s_wtime);
8446fe92 6089 es->s_kbytes_written =
0bc9bc1d 6090 cpu_to_le64(sbi->s_kbytes_written +
8446fe92 6091 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
0bc9bc1d 6092 sbi->s_sectors_written_start) >> 1));
e92ad03f 6093 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
d5e03cbb 6094 ext4_free_blocks_count_set(es,
e92ad03f
JK		 6095 EXT4_C2B(sbi, percpu_counter_sum_positive(
6096 &sbi->s_freeclusters_counter)));
6097 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
d5e03cbb
TT		 6098 es->s_free_inodes_count =
6099 cpu_to_le32(percpu_counter_sum_positive(
e92ad03f 6100 &sbi->s_freeinodes_counter));
c92dc856
JK		 6101 /* Copy error information to the on-disk superblock */
6102 spin_lock(&sbi->s_error_lock);
6103 if (sbi->s_add_error_count > 0) {
6104 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6105 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6106 __ext4_update_tstamp(&es->s_first_error_time,
6107 &es->s_first_error_time_hi,
6108 sbi->s_first_error_time);
6109 strncpy(es->s_first_error_func, sbi->s_first_error_func,
6110 sizeof(es->s_first_error_func));
6111 es->s_first_error_line =
6112 cpu_to_le32(sbi->s_first_error_line);
6113 es->s_first_error_ino =
6114 cpu_to_le32(sbi->s_first_error_ino);
6115 es->s_first_error_block =
6116 cpu_to_le64(sbi->s_first_error_block);
6117 es->s_first_error_errcode =
6118 ext4_errno_to_code(sbi->s_first_error_code);
6119 }
6120 __ext4_update_tstamp(&es->s_last_error_time,
6121 &es->s_last_error_time_hi,
6122 sbi->s_last_error_time);
6123 strncpy(es->s_last_error_func, sbi->s_last_error_func,
6124 sizeof(es->s_last_error_func));
6125 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6126 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6127 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6128 es->s_last_error_errcode =
6129 ext4_errno_to_code(sbi->s_last_error_code);
6130 /*
6131 * Start the daily error reporting function if it hasn't been
6132 * started already
6133 */
6134 if (!es->s_error_count)
6135 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6136 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6137 sbi->s_add_error_count = 0;
6138 }
6139 spin_unlock(&sbi->s_error_lock);
6140
06db49e6 6141 ext4_superblock_csum_set(sb);
2d01ddc8
JK		 6142 unlock_buffer(sbh);
6143}
6144
6145static int ext4_commit_super(struct super_block *sb)
6146{
6147 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2d01ddc8 6148
f88f1466
FC		 6149 if (!sbh)
6150 return -EINVAL;
6151 if (block_device_ejected(sb))
6152 return -ENODEV;
2d01ddc8
JK		 6153
6154 ext4_update_super(sb);
6155
15baa7dc
ZY		 6156 lock_buffer(sbh);
6157 /* Buffer got discarded which means block device got invalidated */
6158 if (!buffer_mapped(sbh)) {
6159 unlock_buffer(sbh);
6160 return -EIO;
6161 }
6162
e8680786 6163 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4743f839
PKS		 6164 /*
6165 * Oh, dear. A previous attempt to write the
6166 * superblock failed. This could happen because the
6167 * USB device was yanked out. Or it could happen to
6168 * be a transient write error and maybe the block will
6169 * be remapped. Nothing we can do but to retry the
6170 * write and hope for the best.
6171 */
6172 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6173 "superblock detected");
6174 clear_buffer_write_io_error(sbh);
6175 set_buffer_uptodate(sbh);
6176 }
15baa7dc
ZY		 6177 get_bh(sbh);
6178 /* Clear potential dirty bit if it was journalled update */
6179 clear_buffer_dirty(sbh);
6180 sbh->b_end_io = end_buffer_write_sync;
1420c4a5
BVA		 6181 submit_bh(REQ_OP_WRITE | REQ_SYNC |
6182 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
15baa7dc 6183 wait_on_buffer(sbh);
4392fbc4
JK		 6184 if (buffer_write_io_error(sbh)) {
6185 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6186 "superblock");
6187 clear_buffer_write_io_error(sbh);
6188 set_buffer_uptodate(sbh);
15baa7dc 6189 return -EIO;
914258bf 6190 }
15baa7dc 6191 return 0;
ac27a0ec
DK		 6192}
6193
ac27a0ec
DK		 6194/*
6195 * Have we just finished recovery? If so, and if we are mounting (or
6196 * remounting) the filesystem readonly, then we will end up with a
6197 * consistent fs on disk. Record that fact.
6198 */
11215630
JK		 6199static int ext4_mark_recovery_complete(struct super_block *sb,
6200 struct ext4_super_block *es)
ac27a0ec 6201{
11215630 6202 int err;
617ba13b 6203 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 6204
e2b911c5 6205 if (!ext4_has_feature_journal(sb)) {
11215630
JK		 6206 if (journal != NULL) {
6207 ext4_error(sb, "Journal got removed while the fs was "
6208 "mounted!");
6209 return -EFSCORRUPTED;
6210 }
6211 return 0;
0390131b 6212 }
dab291af 6213 jbd2_journal_lock_updates(journal);
01d5d965 6214 err = jbd2_journal_flush(journal, 0);
11215630 6215 if (err < 0)
7ffe1ea8
HK		 6216 goto out;
6217
02f310fc
JK		 6218 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6219 ext4_has_feature_orphan_present(sb))) {
6220 if (!ext4_orphan_file_empty(sb)) {
6221 ext4_error(sb, "Orphan file not empty on read-only fs.");
6222 err = -EFSCORRUPTED;
6223 goto out;
6224 }
e2b911c5 6225 ext4_clear_feature_journal_needs_recovery(sb);
02f310fc 6226 ext4_clear_feature_orphan_present(sb);
4392fbc4 6227 ext4_commit_super(sb);
ac27a0ec 6228 }
7ffe1ea8 6229out:
dab291af 6230 jbd2_journal_unlock_updates(journal);
11215630 6231 return err;
ac27a0ec
DK		 6232}
6233
6234/*
6235 * If we are mounting (or read-write remounting) a filesystem whose journal
6236 * has recorded an error from a previous lifetime, move that error to the
6237 * main filesystem now.
6238 */
11215630 6239static int ext4_clear_journal_err(struct super_block *sb,
2b2d6d01 6240 struct ext4_super_block *es)
ac27a0ec
DK		 6241{
6242 journal_t *journal;
6243 int j_errno;
6244 const char *errstr;
6245
11215630
JK		 6246 if (!ext4_has_feature_journal(sb)) {
6247 ext4_error(sb, "Journal got removed while the fs was mounted!");
6248 return -EFSCORRUPTED;
6249 }
0390131b 6250
617ba13b 6251 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 6252
6253 /*
6254 * Now check for any error status which may have been recorded in the
617ba13b 6255 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 6256 */
6257
dab291af 6258 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 6259 if (j_errno) {
6260 char nbuf[16];
6261
617ba13b 6262 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 6263 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 6264 "from previous mount: %s", errstr);
ac27a0ec 6265
617ba13b
MC		 6266 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6267 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
f57886ca
YB		 6268 j_errno = ext4_commit_super(sb);
6269 if (j_errno)
6270 return j_errno;
6271 ext4_warning(sb, "Marked fs in need of filesystem check.");
ac27a0ec 6272
dab291af 6273 jbd2_journal_clear_err(journal);
d796c52e 6274 jbd2_journal_update_sb_errno(journal);
ac27a0ec 6275 }
11215630 6276 return 0;
ac27a0ec
DK		 6277}
6278
6279/*
6280 * Force the running and committing transactions to commit,
6281 * and wait on the commit.
6282 */
617ba13b 6283int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 6284{
6285 journal_t *journal;
ac27a0ec 6286
bc98a42c 6287 if (sb_rdonly(sb))
ac27a0ec
DK		 6288 return 0;
6289
617ba13b 6290 journal = EXT4_SB(sb)->s_journal;
b1deefc9 6291 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 6292}
6293
617ba13b 6294static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 6295{
14ce0cb4 6296 int ret = 0;
9eddacf9 6297 tid_t target;
06a407f1 6298 bool needs_barrier = false;
8d5d02e6 6299 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 6300
49598e04 6301 if (unlikely(ext4_forced_shutdown(sbi)))
0db1ff22
TT		 6302 return 0;
6303
9bffad1e 6304 trace_ext4_sync_fs(sb, wait);
2e8fa54e 6305 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 6306 /*
6307 * Writeback quota in non-journalled quota case - journalled quota has
6308 * no dirty dquots
6309 */
6310 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 6311 /*
6312 * Data writeback is possible w/o journal transaction, so barrier must
6313 * being sent at the end of the function. But we can skip it if
6314 * transaction_commit will do it for us.
6315 */
bda32530
TT		 6316 if (sbi->s_journal) {
6317 target = jbd2_get_latest_transaction(sbi->s_journal);
6318 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6319 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6320 needs_barrier = true;
6321
6322 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6323 if (wait)
6324 ret = jbd2_log_wait_commit(sbi->s_journal,
6325 target);
6326 }
6327 } else if (wait && test_opt(sb, BARRIER))
06a407f1 6328 needs_barrier = true;
06a407f1
DM		 6329 if (needs_barrier) {
6330 int err;
c6bf3f0e 6331 err = blkdev_issue_flush(sb->s_bdev);
06a407f1
DM		 6332 if (!ret)
6333 ret = err;
0390131b 6334 }
06a407f1
DM		 6335
6336 return ret;
6337}
6338
ac27a0ec
DK		 6339/*
6340 * LVM calls this function before a (read-only) snapshot is created. This
6341 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 6342 *
6343 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 6344 * state independently. It relies on upper layer to stop all data & metadata
6345 * modifications.
ac27a0ec 6346 */
c4be0c1d 6347static int ext4_freeze(struct super_block *sb)
ac27a0ec 6348{
c4be0c1d
TS		 6349 int error = 0;
6350 journal_t *journal;
ac27a0ec 6351
bc98a42c 6352 if (sb_rdonly(sb))
9ca92389 6353 return 0;
ac27a0ec 6354
9ca92389 6355 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 6356
bb044576
TT		 6357 if (journal) {
6358 /* Now we set up the journal barrier. */
6359 jbd2_journal_lock_updates(journal);
ac27a0ec 6360
bb044576
TT		 6361 /*
6362 * Don't clear the needs_recovery flag if we failed to
6363 * flush the journal.
6364 */
01d5d965 6365 error = jbd2_journal_flush(journal, 0);
bb044576
TT		 6366 if (error < 0)
6367 goto out;
c642dc9e
ES		 6368
6369 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 6370 ext4_clear_feature_journal_needs_recovery(sb);
02f310fc
JK		 6371 if (ext4_orphan_file_empty(sb))
6372 ext4_clear_feature_orphan_present(sb);
bb044576 6373 }
9ca92389 6374
4392fbc4 6375 error = ext4_commit_super(sb);
6b0310fb 6376out:
bb044576
TT		 6377 if (journal)
6378 /* we rely on upper layer to stop further updates */
6379 jbd2_journal_unlock_updates(journal);
6b0310fb 6380 return error;
ac27a0ec
DK		 6381}
6382
6383/*
6384 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6385 * flag here, even though the filesystem is not technically dirty yet.
6386 */
c4be0c1d 6387static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 6388{
bc98a42c 6389 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
9ca92389
TT		 6390 return 0;
6391
c642dc9e
ES		 6392 if (EXT4_SB(sb)->s_journal) {
6393 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 6394 ext4_set_feature_journal_needs_recovery(sb);
02f310fc
JK		 6395 if (ext4_has_feature_orphan_file(sb))
6396 ext4_set_feature_orphan_present(sb);
c642dc9e
ES		 6397 }
6398
4392fbc4 6399 ext4_commit_super(sb);
c4be0c1d 6400 return 0;
ac27a0ec
DK		 6401}
6402
673c6100
TT		 6403/*
6404 * Structure to save mount options for ext4_remount's benefit
6405 */
6406struct ext4_mount_options {
6407 unsigned long s_mount_opt;
a2595b8a 6408 unsigned long s_mount_opt2;
08cefc7a
EB		 6409 kuid_t s_resuid;
6410 kgid_t s_resgid;
673c6100
TT		 6411 unsigned long s_commit_interval;
6412 u32 s_min_batch_time, s_max_batch_time;
6413#ifdef CONFIG_QUOTA
6414 int s_jquota_fmt;
a2d4a646 6415 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 6416#endif
6417};
6418
960e0ab6 6419static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
ac27a0ec 6420{
7edfd85b 6421 struct ext4_fs_context *ctx = fc->fs_private;
2b2d6d01 6422 struct ext4_super_block *es;
617ba13b 6423 struct ext4_sb_info *sbi = EXT4_SB(sb);
960e0ab6 6424 unsigned long old_sb_flags;
617ba13b 6425 struct ext4_mount_options old_opts;
8a266467 6426 ext4_group_t g;
c5e06d10 6427 int err = 0;
a937cf1c 6428 int alloc_ctx;
ac27a0ec 6429#ifdef CONFIG_QUOTA
3bbef91b 6430 int enable_quota = 0;
03dafb5f 6431 int i, j;
33458eab 6432 char *to_free[EXT4_MAXQUOTAS];
ac27a0ec
DK		 6433#endif
6434
21ac738e 6435
ac27a0ec
DK		 6436 /* Store the original options */
6437 old_sb_flags = sb->s_flags;
6438 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 6439 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 6440 old_opts.s_resuid = sbi->s_resuid;
6441 old_opts.s_resgid = sbi->s_resgid;
6442 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 6443 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6444 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 6445#ifdef CONFIG_QUOTA
6446 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 6447 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 6448 if (sbi->s_qf_names[i]) {
33458eab
TT		 6449 char *qf_name = get_qf_name(sb, sbi, i);
6450
6451 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
03dafb5f
CG		 6452 if (!old_opts.s_qf_names[i]) {
6453 for (j = 0; j < i; j++)
6454 kfree(old_opts.s_qf_names[j]);
6455 return -ENOMEM;
6456 }
6457 } else
6458 old_opts.s_qf_names[i] = NULL;
ac27a0ec 6459#endif
e4e58e5d
OM		 6460 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6461 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6462 ctx->journal_ioprio =
6463 sbi->s_journal->j_task->io_context->ioprio;
6464 else
6465 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6466
6467 }
ac27a0ec 6468
a937cf1c
BL		 6469 /*
6470 * Changing the DIOREAD_NOLOCK or DELALLOC mount options may cause
6471 * two calls to ext4_should_dioread_nolock() to return inconsistent
6472 * values, triggering WARN_ON in ext4_add_complete_io(). we grab
6473 * here s_writepages_rwsem to avoid race between writepages ops and
6474 * remount.
6475 */
6476 alloc_ctx = ext4_writepages_down_write(sb);
7edfd85b 6477 ext4_apply_options(fc, sb);
a937cf1c 6478 ext4_writepages_up_write(sb, alloc_ctx);
ac27a0ec 6479
6b992ff2 6480 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 6481 test_opt(sb, JOURNAL_CHECKSUM)) {
6482 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 6483 "during remount not supported; ignoring");
6484 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 6485 }
6486
6ae6514b
PS		 6487 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6488 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6489 ext4_msg(sb, KERN_ERR, "can't mount with "
6490 "both data=journal and delalloc");
6491 err = -EINVAL;
6492 goto restore_opts;
6493 }
6494 if (test_opt(sb, DIOREAD_NOLOCK)) {
6495 ext4_msg(sb, KERN_ERR, "can't mount with "
6496 "both data=journal and dioread_nolock");
6497 err = -EINVAL;
6498 goto restore_opts;
6499 }
ab04df78
JK		 6500 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6501 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6502 ext4_msg(sb, KERN_ERR, "can't mount with "
6503 "journal_async_commit in data=ordered mode");
6504 err = -EINVAL;
6505 goto restore_opts;
6506 }
923ae0ff
RZ		 6507 }
6508
cdb7ee4c
TE		 6509 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6510 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6511 err = -EINVAL;
6512 goto restore_opts;
6513 }
6514
9b5f6c9b 6515 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
124e7c61 6516 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
ac27a0ec 6517
1751e8a6
LT		 6518 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6519 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 6520
6521 es = sbi->s_es;
6522
b3881f74 6523 if (sbi->s_journal) {
0390131b 6524 ext4_init_journal_params(sb, sbi->s_journal);
7edfd85b 6525 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
b3881f74 6526 }
ac27a0ec 6527
c92dc856
JK		 6528 /* Flush outstanding errors before changing fs state */
6529 flush_work(&sbi->s_error_work);
6530
960e0ab6 6531 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
9b5f6c9b 6532 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
ac27a0ec
DK		 6533 err = -EROFS;
6534 goto restore_opts;
6535 }
6536
960e0ab6 6537 if (fc->sb_flags & SB_RDONLY) {
38c03b34
TT		 6538 err = sync_filesystem(sb);
6539 if (err < 0)
6540 goto restore_opts;
0f0dd62f
CH		 6541 err = dquot_suspend(sb, -1);
6542 if (err < 0)
c79d967d 6543 goto restore_opts;
c79d967d 6544
ac27a0ec
DK		 6545 /*
6546 * First of all, the unconditional stuff we have to do
6547 * to disable replay of the journal when we next remount
6548 */
1751e8a6 6549 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 6550
6551 /*
6552 * OK, test if we are remounting a valid rw partition
6553 * readonly, and if so set the rdonly flag and then
6554 * mark the partition as valid again.
6555 */
617ba13b
MC		 6556 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6557 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 6558 es->s_state = cpu_to_le16(sbi->s_mount_state);
6559
11215630
JK		 6560 if (sbi->s_journal) {
6561 /*
6562 * We let remount-ro finish even if marking fs
6563 * as clean failed...
6564 */
0390131b 6565 ext4_mark_recovery_complete(sb, es);
11215630 6566 }
ac27a0ec 6567 } else {
a13fb1a4 6568 /* Make sure we can mount this feature set readwrite */
e2b911c5 6569 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 6570 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 6571 err = -EROFS;
6572 goto restore_opts;
6573 }
8a266467
TT		 6574 /*
6575 * Make sure the group descriptor checksums
0b8e58a1 6576 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 6577 */
6578 for (g = 0; g < sbi->s_groups_count; g++) {
6579 struct ext4_group_desc *gdp =
6580 ext4_get_group_desc(sb, g, NULL);
6581
feb0ab32 6582 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 6583 ext4_msg(sb, KERN_ERR,
6584 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 6585 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 6586 le16_to_cpu(gdp->bg_checksum));
6a797d27 6587 err = -EFSBADCRC;
8a266467
TT		 6588 goto restore_opts;
6589 }
6590 }
6591
ead6596b
ES		 6592 /*
6593 * If we have an unprocessed orphan list hanging
6594 * around from a previously readonly bdev mount,
6595 * require a full umount/remount for now.
6596 */
02f310fc 6597 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
b31e1552 6598 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 6599 "remount RDWR because of unprocessed "
6600 "orphan inode list. Please "
b31e1552 6601 "umount/remount instead");
ead6596b
ES		 6602 err = -EINVAL;
6603 goto restore_opts;
6604 }
6605
ac27a0ec
DK		 6606 /*
6607 * Mounting a RDONLY partition read-write, so reread
6608 * and store the current valid flag. (It may have
6609 * been changed by e2fsck since we originally mounted
6610 * the partition.)
6611 */
11215630
JK		 6612 if (sbi->s_journal) {
6613 err = ext4_clear_journal_err(sb, es);
6614 if (err)
6615 goto restore_opts;
6616 }
c878bea3
TT		 6617 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6618 ~EXT4_FC_REPLAY);
c89128a0
JK		 6619
6620 err = ext4_setup_super(sb, es, 0);
6621 if (err)
6622 goto restore_opts;
6623
1b292439 6624 sb->s_flags &= ~SB_RDONLY;
3b50d501
TT		 6625 if (ext4_has_feature_mmp(sb)) {
6626 err = ext4_multi_mount_protect(sb,
6627 le64_to_cpu(es->s_mmp_block));
6628 if (err)
c5e06d10 6629 goto restore_opts;
3b50d501 6630 }
3bbef91b 6631#ifdef CONFIG_QUOTA
c79d967d 6632 enable_quota = 1;
3bbef91b 6633#endif
ac27a0ec
DK		 6634 }
6635 }
bfff6873 6636
0f5bde1d
JK		 6637 /*
6638 * Handle creation of system zone data early because it can fail.
6639 * Releasing of existing data is done when we are sure remount will
6640 * succeed.
6641 */
dd0db94f 6642 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
0f5bde1d
JK		 6643 err = ext4_setup_system_zone(sb);
6644 if (err)
6645 goto restore_opts;
6646 }
d176b1f6 6647
c89128a0 6648 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
4392fbc4 6649 err = ext4_commit_super(sb);
c89128a0
JK		 6650 if (err)
6651 goto restore_opts;
6652 }
0390131b 6653
ac27a0ec 6654#ifdef CONFIG_QUOTA
7c319d32
AK		 6655 if (enable_quota) {
6656 if (sb_any_quota_suspended(sb))
6657 dquot_resume(sb, -1);
e2b911c5 6658 else if (ext4_has_feature_quota(sb)) {
7c319d32 6659 err = ext4_enable_quotas(sb);
07724f98 6660 if (err)
7c319d32 6661 goto restore_opts;
7c319d32
AK		 6662 }
6663 }
4c0b4818
TT		 6664 /* Release old quota file names */
6665 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6666 kfree(old_opts.s_qf_names[i]);
ac27a0ec 6667#endif
dd0db94f 6668 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
0f5bde1d 6669 ext4_release_system_zone(sb);
d4c402d9 6670
eb1f822c
TT		 6671 /*
6672 * Reinitialize lazy itable initialization thread based on
6673 * current settings
6674 */
6675 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6676 ext4_unregister_li_request(sb);
6677 else {
6678 ext4_group_t first_not_zeroed;
6679 first_not_zeroed = ext4_has_uninit_itable(sb);
6680 ext4_register_li_request(sb, first_not_zeroed);
6681 }
6682
61bb4a1c
TT		 6683 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6684 ext4_stop_mmpd(sbi);
6685
ac27a0ec 6686 return 0;
0b8e58a1 6687
ac27a0ec 6688restore_opts:
4c0b4818
TT		 6689 /*
6690 * If there was a failing r/w to ro transition, we may need to
6691 * re-enable quota
6692 */
6693 if ((sb->s_flags & SB_RDONLY) && !(old_sb_flags & SB_RDONLY) &&
6694 sb_any_quota_suspended(sb))
6695 dquot_resume(sb, -1);
a937cf1c
BL		 6696
6697 alloc_ctx = ext4_writepages_down_write(sb);
ac27a0ec
DK		 6698 sb->s_flags = old_sb_flags;
6699 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 6700 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 6701 sbi->s_resuid = old_opts.s_resuid;
6702 sbi->s_resgid = old_opts.s_resgid;
6703 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 6704 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6705 sbi->s_max_batch_time = old_opts.s_max_batch_time;
a937cf1c
BL		 6706 ext4_writepages_up_write(sb, alloc_ctx);
6707
dd0db94f 6708 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
0f5bde1d 6709 ext4_release_system_zone(sb);
ac27a0ec
DK		 6710#ifdef CONFIG_QUOTA
6711 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 6712 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
33458eab
TT		 6713 to_free[i] = get_qf_name(sb, sbi, i);
6714 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
ac27a0ec 6715 }
33458eab
TT		 6716 synchronize_rcu();
6717 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6718 kfree(to_free[i]);
ac27a0ec 6719#endif
61bb4a1c
TT		 6720 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6721 ext4_stop_mmpd(sbi);
ac27a0ec
DK		 6722 return err;
6723}
6724
cebe85d5 6725static int ext4_reconfigure(struct fs_context *fc)
7edfd85b 6726{
cebe85d5 6727 struct super_block *sb = fc->root->d_sb;
7edfd85b
LC		 6728 int ret;
6729
cebe85d5 6730 fc->s_fs_info = EXT4_SB(sb);
7edfd85b 6731
cebe85d5 6732 ret = ext4_check_opt_consistency(fc, sb);
7edfd85b 6733 if (ret < 0)
cebe85d5 6734 return ret;
7edfd85b 6735
960e0ab6 6736 ret = __ext4_remount(fc, sb);
7edfd85b 6737 if (ret < 0)
cebe85d5 6738 return ret;
7edfd85b 6739
6dcc98fb
TT		 6740 ext4_msg(sb, KERN_INFO, "re-mounted %pU %s. Quota mode: %s.",
6741 &sb->s_uuid, sb_rdonly(sb) ? "ro" : "r/w",
6742 ext4_quota_mode(sb));
7edfd85b 6743
7edfd85b 6744 return 0;
7edfd85b
LC		 6745}
6746
689c958c
LX		 6747#ifdef CONFIG_QUOTA
6748static int ext4_statfs_project(struct super_block *sb,
6749 kprojid_t projid, struct kstatfs *buf)
6750{
6751 struct kqid qid;
6752 struct dquot *dquot;
6753 u64 limit;
6754 u64 curblock;
6755
6756 qid = make_kqid_projid(projid);
6757 dquot = dqget(sb, qid);
6758 if (IS_ERR(dquot))
6759 return PTR_ERR(dquot);
7b9ca4c6 6760 spin_lock(&dquot->dq_dqb_lock);
689c958c 6761
a08fe66e
CX		 6762 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6763 dquot->dq_dqb.dqb_bhardlimit);
57c32ea4
CX		 6764 limit >>= sb->s_blocksize_bits;
6765
689c958c 6766 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 6767 curblock = (dquot->dq_dqb.dqb_curspace +
6768 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 6769 buf->f_blocks = limit;
6770 buf->f_bfree = buf->f_bavail =
6771 (buf->f_blocks > curblock) ?
6772 (buf->f_blocks - curblock) : 0;
6773 }
6774
a08fe66e
CX		 6775 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6776 dquot->dq_dqb.dqb_ihardlimit);
689c958c
LX		 6777 if (limit && buf->f_files > limit) {
6778 buf->f_files = limit;
6779 buf->f_ffree =
6780 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6781 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6782 }
6783
7b9ca4c6 6784 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 6785 dqput(dquot);
6786 return 0;
6787}
6788#endif
6789
2b2d6d01 6790static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 6791{
6792 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 6793 struct ext4_sb_info *sbi = EXT4_SB(sb);
6794 struct ext4_super_block *es = sbi->s_es;
27dd4385 6795 ext4_fsblk_t overhead = 0, resv_blocks;
d02a9391 6796 s64 bfree;
27dd4385 6797 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 6798
952fc18e
TT		 6799 if (!test_opt(sb, MINIX_DF))
6800 overhead = sbi->s_overhead;
ac27a0ec 6801
617ba13b 6802 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 6803 buf->f_bsize = sb->s_blocksize;
b72f78cb 6804 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 6805 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6806 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 6807 /* prevent underflow in case that few free space is available */
57042651 6808 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 6809 buf->f_bavail = buf->f_bfree -
6810 (ext4_r_blocks_count(es) + resv_blocks);
6811 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 6812 buf->f_bavail = 0;
6813 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 6814 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 6815 buf->f_namelen = EXT4_NAME_LEN;
9591c3a3 6816 buf->f_fsid = uuid_to_fsid(es->s_uuid);
0b8e58a1 6817
689c958c
LX		 6818#ifdef CONFIG_QUOTA
6819 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6820 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6821 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6822#endif
ac27a0ec
DK		 6823 return 0;
6824}
6825
ac27a0ec
DK		 6826
6827#ifdef CONFIG_QUOTA
6828
bc8230ee
JK		 6829/*
6830 * Helper functions so that transaction is started before we acquire dqio_sem
6831 * to keep correct lock ordering of transaction > dqio_sem
6832 */
ac27a0ec
DK		 6833static inline struct inode *dquot_to_inode(struct dquot *dquot)
6834{
4c376dca 6835 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 6836}
6837
617ba13b 6838static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 6839{
6840 int ret, err;
6841 handle_t *handle;
6842 struct inode *inode;
6843
6844 inode = dquot_to_inode(dquot);
9924a92a 6845 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 6846 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 6847 if (IS_ERR(handle))
6848 return PTR_ERR(handle);
6849 ret = dquot_commit(dquot);
617ba13b 6850 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6851 if (!ret)
6852 ret = err;
6853 return ret;
6854}
6855
617ba13b 6856static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 6857{
6858 int ret, err;
6859 handle_t *handle;
6860
9924a92a 6861 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 6862 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 6863 if (IS_ERR(handle))
6864 return PTR_ERR(handle);
6865 ret = dquot_acquire(dquot);
617ba13b 6866 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6867 if (!ret)
6868 ret = err;
6869 return ret;
6870}
6871
617ba13b 6872static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 6873{
6874 int ret, err;
6875 handle_t *handle;
6876
9924a92a 6877 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 6878 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 6879 if (IS_ERR(handle)) {
6880 /* Release dquot anyway to avoid endless cycle in dqput() */
6881 dquot_release(dquot);
ac27a0ec 6882 return PTR_ERR(handle);
9c3013e9 6883 }
ac27a0ec 6884 ret = dquot_release(dquot);
617ba13b 6885 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6886 if (!ret)
6887 ret = err;
6888 return ret;
6889}
6890
617ba13b 6891static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 6892{
262b4662 6893 struct super_block *sb = dquot->dq_sb;
262b4662 6894
f177ee08 6895 if (ext4_is_quota_journalled(sb)) {
ac27a0ec 6896 dquot_mark_dquot_dirty(dquot);
617ba13b 6897 return ext4_write_dquot(dquot);
ac27a0ec
DK		 6898 } else {
6899 return dquot_mark_dquot_dirty(dquot);
6900 }
6901}
6902
617ba13b 6903static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 6904{
6905 int ret, err;
6906 handle_t *handle;
6907
6908 /* Data block + inode block */
f9c1f248 6909 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 6910 if (IS_ERR(handle))
6911 return PTR_ERR(handle);
6912 ret = dquot_commit_info(sb, type);
617ba13b 6913 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6914 if (!ret)
6915 ret = err;
6916 return ret;
6917}
6918
daf647d2
TT		 6919static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6920{
6921 struct ext4_inode_info *ei = EXT4_I(inode);
6922
6923 /* The first argument of lockdep_set_subclass has to be
6924 * *exactly* the same as the argument to init_rwsem() --- in
6925 * this case, in init_once() --- or lockdep gets unhappy
6926 * because the name of the lock is set using the
6927 * stringification of the argument to init_rwsem().
6928 */
6929 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6930 lockdep_set_subclass(&ei->i_data_sem, subclass);
6931}
6932
ac27a0ec
DK		 6933/*
6934 * Standard function to be called on quota_on
6935 */
617ba13b 6936static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 6937 const struct path *path)
ac27a0ec
DK		 6938{
6939 int err;
ac27a0ec
DK		 6940
6941 if (!test_opt(sb, QUOTA))
6942 return -EINVAL;
0623543b 6943
ac27a0ec 6944 /* Quotafile not on the same filesystem? */
d8c9584e 6945 if (path->dentry->d_sb != sb)
ac27a0ec 6946 return -EXDEV;
e0770e91
JK		 6947
6948 /* Quota already enabled for this file? */
6949 if (IS_NOQUOTA(d_inode(path->dentry)))
6950 return -EBUSY;
6951
0623543b
JK		 6952 /* Journaling quota? */
6953 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 6954 /* Quotafile not in fs root? */
f00c9e44 6955 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 6956 ext4_msg(sb, KERN_WARNING,
6957 "Quota file not on filesystem root. "
6958 "Journaled quota will not work");
91389240
JK		 6959 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6960 } else {
6961 /*
6962 * Clear the flag just in case mount options changed since
6963 * last time.
6964 */
6965 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 6966 }
0623543b 6967
daf647d2
TT		 6968 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6969 err = dquot_quota_on(sb, type, format_id, path);
15fc69bb 6970 if (!err) {
957153fc
JK		 6971 struct inode *inode = d_inode(path->dentry);
6972 handle_t *handle;
6973
61a92987
JK		 6974 /*
6975 * Set inode flags to prevent userspace from messing with quota
6976 * files. If this fails, we return success anyway since quotas
6977 * are already enabled and this is not a hard failure.
6978 */
957153fc
JK		 6979 inode_lock(inode);
6980 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6981 if (IS_ERR(handle))
6982 goto unlock_inode;
6983 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6984 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6985 S_NOATIME | S_IMMUTABLE);
4209ae12 6986 err = ext4_mark_inode_dirty(handle, inode);
957153fc
JK		 6987 ext4_journal_stop(handle);
6988 unlock_inode:
6989 inode_unlock(inode);
15fc69bb
JK		 6990 if (err)
6991 dquot_quota_off(sb, type);
957153fc 6992 }
15fc69bb
JK		 6993 if (err)
6994 lockdep_set_quota_inode(path->dentry->d_inode,
6995 I_DATA_SEM_NORMAL);
daf647d2 6996 return err;
ac27a0ec
DK		 6997}
6998
07342ec2
BL		 6999static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
7000{
7001 switch (type) {
7002 case USRQUOTA:
7003 return qf_inum == EXT4_USR_QUOTA_INO;
7004 case GRPQUOTA:
7005 return qf_inum == EXT4_GRP_QUOTA_INO;
7006 case PRJQUOTA:
7007 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
7008 default:
7009 BUG();
7010 }
7011}
7012
7c319d32
AK		 7013static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
7014 unsigned int flags)
7015{
7016 int err;
7017 struct inode *qf_inode;
a2d4a646 7018 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 7019 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 7020 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7021 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 7022 };
7023
e2b911c5 7024 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 7025
7026 if (!qf_inums[type])
7027 return -EPERM;
7028
07342ec2
BL		 7029 if (!ext4_check_quota_inum(type, qf_inums[type])) {
7030 ext4_error(sb, "Bad quota inum: %lu, type: %d",
7031 qf_inums[type], type);
7032 return -EUCLEAN;
7033 }
7034
8a363970 7035 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7c319d32 7036 if (IS_ERR(qf_inode)) {
07342ec2
BL		 7037 ext4_error(sb, "Bad quota inode: %lu, type: %d",
7038 qf_inums[type], type);
7c319d32
AK		 7039 return PTR_ERR(qf_inode);
7040 }
7041
bcb13850
JK		 7042 /* Don't account quota for quota files to avoid recursion */
7043 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 7044 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7212b95e 7045 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
daf647d2
TT		 7046 if (err)
7047 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
61157b24 7048 iput(qf_inode);
7c319d32
AK		 7049
7050 return err;
7051}
7052
7053/* Enable usage tracking for all quota types. */
25c6d98f 7054int ext4_enable_quotas(struct super_block *sb)
7c319d32
AK		 7055{
7056 int type, err = 0;
a2d4a646 7057 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 7058 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 7059 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7060 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 7061 };
49da9392
JK		 7062 bool quota_mopt[EXT4_MAXQUOTAS] = {
7063 test_opt(sb, USRQUOTA),
7064 test_opt(sb, GRPQUOTA),
7065 test_opt(sb, PRJQUOTA),
7066 };
7c319d32 7067
91389240 7068 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 7069 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 7070 if (qf_inums[type]) {
7071 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 7072 DQUOT_USAGE_ENABLED |
7073 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 7074 if (err) {
7075 ext4_warning(sb,
72ba7450 7076 "Failed to enable quota tracking "
07342ec2
BL		 7077 "(type=%d, err=%d, ino=%lu). "
7078 "Please run e2fsck to fix.", type,
7079 err, qf_inums[type]);
7f144fd0 7080
f3c1c42e 7081 ext4_quotas_off(sb, type);
7c319d32
AK		 7082 return err;
7083 }
7084 }
7085 }
7086 return 0;
7087}
7088
ca0e05e4
DM		 7089static int ext4_quota_off(struct super_block *sb, int type)
7090{
21f97697
JK		 7091 struct inode *inode = sb_dqopt(sb)->files[type];
7092 handle_t *handle;
957153fc 7093 int err;
21f97697 7094
87009d86
DM		 7095 /* Force all delayed allocation blocks to be allocated.
7096 * Caller already holds s_umount sem */
7097 if (test_opt(sb, DELALLOC))
ca0e05e4 7098 sync_filesystem(sb);
ca0e05e4 7099
957153fc 7100 if (!inode || !igrab(inode))
0b268590
AG		 7101 goto out;
7102
957153fc 7103 err = dquot_quota_off(sb, type);
964edf66 7104 if (err || ext4_has_feature_quota(sb))
957153fc
JK		 7105 goto out_put;
7106
7107 inode_lock(inode);
61a92987
JK		 7108 /*
7109 * Update modification times of quota files when userspace can
7110 * start looking at them. If we fail, we return success anyway since
7111 * this is not a hard failure and quotas are already disabled.
7112 */
9924a92a 7113 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
4209ae12
HS		 7114 if (IS_ERR(handle)) {
7115 err = PTR_ERR(handle);
957153fc 7116 goto out_unlock;
4209ae12 7117 }
957153fc
JK		 7118 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7119 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
eeca7ea1 7120 inode->i_mtime = inode->i_ctime = current_time(inode);
4209ae12 7121 err = ext4_mark_inode_dirty(handle, inode);
21f97697 7122 ext4_journal_stop(handle);
957153fc
JK		 7123out_unlock:
7124 inode_unlock(inode);
7125out_put:
964edf66 7126 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 7127 iput(inode);
7128 return err;
21f97697 7129out:
ca0e05e4
DM		 7130 return dquot_quota_off(sb, type);
7131}
7132
ac27a0ec
DK		 7133/* Read data from quotafile - avoid pagecache and such because we cannot afford
7134 * acquiring the locks... As quota files are never truncated and quota code
25985edc 7135 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 7136 * we don't have to be afraid of races */
617ba13b 7137static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 7138 size_t len, loff_t off)
7139{
7140 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 7141 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 7142 int offset = off & (sb->s_blocksize - 1);
7143 int tocopy;
7144 size_t toread;
7145 struct buffer_head *bh;
7146 loff_t i_size = i_size_read(inode);
7147
7148 if (off > i_size)
7149 return 0;
7150 if (off+len > i_size)
7151 len = i_size-off;
7152 toread = len;
7153 while (toread > 0) {
66267814 7154 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1c215028
TT		 7155 bh = ext4_bread(NULL, inode, blk, 0);
7156 if (IS_ERR(bh))
7157 return PTR_ERR(bh);
ac27a0ec
DK		 7158 if (!bh) /* A hole? */
7159 memset(data, 0, tocopy);
7160 else
7161 memcpy(data, bh->b_data+offset, tocopy);
7162 brelse(bh);
7163 offset = 0;
7164 toread -= tocopy;
7165 data += tocopy;
7166 blk++;
7167 }
7168 return len;
7169}
7170
7171/* Write to quotafile (we know the transaction is already started and has
7172 * enough credits) */
617ba13b 7173static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 7174 const char *data, size_t len, loff_t off)
7175{
7176 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 7177 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4209ae12 7178 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
c5e298ae 7179 int retries = 0;
ac27a0ec
DK		 7180 struct buffer_head *bh;
7181 handle_t *handle = journal_current_handle();
7182
380a0091 7183 if (!handle) {
b31e1552
ES		 7184 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7185 " cancelled because transaction is not started",
9c3013e9
JK		 7186 (unsigned long long)off, (unsigned long long)len);
7187 return -EIO;
7188 }
67eeb568
DM		 7189 /*
7190 * Since we account only one data block in transaction credits,
7191 * then it is impossible to cross a block boundary.
7192 */
7193 if (sb->s_blocksize - offset < len) {
7194 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7195 " cancelled because not block aligned",
7196 (unsigned long long)off, (unsigned long long)len);
7197 return -EIO;
7198 }
7199
c5e298ae
TT		 7200 do {
7201 bh = ext4_bread(handle, inode, blk,
7202 EXT4_GET_BLOCKS_CREATE |
7203 EXT4_GET_BLOCKS_METADATA_NOFAIL);
45586c70 7204 } while (PTR_ERR(bh) == -ENOSPC &&
c5e298ae 7205 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 7206 if (IS_ERR(bh))
7207 return PTR_ERR(bh);
67eeb568
DM		 7208 if (!bh)
7209 goto out;
5d601255 7210 BUFFER_TRACE(bh, "get write access");
188c299e 7211 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
62d2b5f2
JK		 7212 if (err) {
7213 brelse(bh);
1c215028 7214 return err;
ac27a0ec 7215 }
67eeb568
DM		 7216 lock_buffer(bh);
7217 memcpy(bh->b_data+offset, data, len);
7218 flush_dcache_page(bh->b_page);
7219 unlock_buffer(bh);
62d2b5f2 7220 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 7221 brelse(bh);
ac27a0ec 7222out:
67eeb568
DM		 7223 if (inode->i_size < off + len) {
7224 i_size_write(inode, off + len);
617ba13b 7225 EXT4_I(inode)->i_disksize = inode->i_size;
4209ae12
HS		 7226 err2 = ext4_mark_inode_dirty(handle, inode);
7227 if (unlikely(err2 && !err))
7228 err = err2;
ac27a0ec 7229 }
4209ae12 7230 return err ? err : len;
ac27a0ec 7231}
ac27a0ec
DK		 7232#endif
7233
c290ea01 7234#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 7235static inline void register_as_ext2(void)
7236{
7237 int err = register_filesystem(&ext2_fs_type);
7238 if (err)
7239 printk(KERN_WARNING
7240 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7241}
7242
7243static inline void unregister_as_ext2(void)
7244{
7245 unregister_filesystem(&ext2_fs_type);
7246}
2035e776
TT		 7247
7248static inline int ext2_feature_set_ok(struct super_block *sb)
7249{
e2b911c5 7250 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 7251 return 0;
bc98a42c 7252 if (sb_rdonly(sb))
2035e776 7253 return 1;
e2b911c5 7254 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 7255 return 0;
7256 return 1;
7257}
24b58424
TT		 7258#else
7259static inline void register_as_ext2(void) { }
7260static inline void unregister_as_ext2(void) { }
2035e776 7261static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 7262#endif
7263
24b58424
TT		 7264static inline void register_as_ext3(void)
7265{
7266 int err = register_filesystem(&ext3_fs_type);
7267 if (err)
7268 printk(KERN_WARNING
7269 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7270}
7271
7272static inline void unregister_as_ext3(void)
7273{
7274 unregister_filesystem(&ext3_fs_type);
7275}
2035e776
TT		 7276
7277static inline int ext3_feature_set_ok(struct super_block *sb)
7278{
e2b911c5 7279 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 7280 return 0;
e2b911c5 7281 if (!ext4_has_feature_journal(sb))
2035e776 7282 return 0;
bc98a42c 7283 if (sb_rdonly(sb))
2035e776 7284 return 1;
e2b911c5 7285 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 7286 return 0;
7287 return 1;
7288}
24b58424 7289
03010a33 7290static struct file_system_type ext4_fs_type = {
cebe85d5
LC		 7291 .owner = THIS_MODULE,
7292 .name = "ext4",
7293 .init_fs_context = ext4_init_fs_context,
7294 .parameters = ext4_param_specs,
7295 .kill_sb = kill_block_super,
7296 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
03010a33 7297};
7f78e035 7298MODULE_ALIAS_FS("ext4");
03010a33 7299
e9e3bcec
ES		 7300/* Shared across all ext4 file systems */
7301wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 7302
5dabfc78 7303static int __init ext4_init_fs(void)
ac27a0ec 7304{
e9e3bcec 7305 int i, err;
c9de560d 7306
e294a537 7307 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8 7308 ext4_li_info = NULL;
07c0c5d8 7309
9a4c8019 7310 /* Build-time check for flags consistency */
12e9b892 7311 ext4_check_flag_values();
e9e3bcec 7312
e142d052 7313 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 7314 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 7315
51865fda 7316 err = ext4_init_es();
6fd058f7
TT		 7317 if (err)
7318 return err;
51865fda 7319
1dc0aa46 7320 err = ext4_init_pending();
22cfe4b4
EB		 7321 if (err)
7322 goto out7;
7323
7324 err = ext4_init_post_read_processing();
1dc0aa46
EW		 7325 if (err)
7326 goto out6;
7327
51865fda
ZL		 7328 err = ext4_init_pageio();
7329 if (err)
b5799018 7330 goto out5;
51865fda 7331
5dabfc78 7332 err = ext4_init_system_zone();
bd2d0210 7333 if (err)
b5799018 7334 goto out4;
857ac889 7335
b5799018 7336 err = ext4_init_sysfs();
dd68314c 7337 if (err)
b5799018 7338 goto out3;
857ac889 7339
5dabfc78 7340 err = ext4_init_mballoc();
c9de560d
AT		 7341 if (err)
7342 goto out2;
ac27a0ec
DK		 7343 err = init_inodecache();
7344 if (err)
7345 goto out1;
aa75f4d3
HS		 7346
7347 err = ext4_fc_init_dentry_cache();
7348 if (err)
7349 goto out05;
7350
24b58424 7351 register_as_ext3();
2035e776 7352 register_as_ext2();
03010a33 7353 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 7354 if (err)
7355 goto out;
bfff6873 7356
ac27a0ec
DK		 7357 return 0;
7358out:
24b58424
TT		 7359 unregister_as_ext2();
7360 unregister_as_ext3();
ab047d51 7361 ext4_fc_destroy_dentry_cache();
aa75f4d3 7362out05:
ac27a0ec
DK		 7363 destroy_inodecache();
7364out1:
5dabfc78 7365 ext4_exit_mballoc();
9c191f70 7366out2:
b5799018
TT		 7367 ext4_exit_sysfs();
7368out3:
5dabfc78 7369 ext4_exit_system_zone();
b5799018 7370out4:
5dabfc78 7371 ext4_exit_pageio();
b5799018 7372out5:
22cfe4b4 7373 ext4_exit_post_read_processing();
1dc0aa46 7374out6:
22cfe4b4
EB		 7375 ext4_exit_pending();
7376out7:
51865fda
ZL		 7377 ext4_exit_es();
7378
ac27a0ec
DK		 7379 return err;
7380}
7381
5dabfc78 7382static void __exit ext4_exit_fs(void)
ac27a0ec 7383{
bfff6873 7384 ext4_destroy_lazyinit_thread();
24b58424
TT		 7385 unregister_as_ext2();
7386 unregister_as_ext3();
03010a33 7387 unregister_filesystem(&ext4_fs_type);
ab047d51 7388 ext4_fc_destroy_dentry_cache();
ac27a0ec 7389 destroy_inodecache();
5dabfc78 7390 ext4_exit_mballoc();
b5799018 7391 ext4_exit_sysfs();
5dabfc78
TT		 7392 ext4_exit_system_zone();
7393 ext4_exit_pageio();
22cfe4b4 7394 ext4_exit_post_read_processing();
dd12ed14 7395 ext4_exit_es();
1dc0aa46 7396 ext4_exit_pending();
ac27a0ec
DK		 7397}
7398
7399MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 7400MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 7401MODULE_LICENSE("GPL");
7ef79ad5 7402MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 7403module_init(ext4_init_fs)
7404module_exit(ext4_exit_fs)
Mirror https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git