projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge tag 'ext4_for_linus-6.10-rc1' of git://git.kernel.org/pub/scm/linux/kernel...
[linux-2.6-block.git] / fs / ext4 / super.c
CommitLineData
f5166768 1// SPDX-License-Identifier: GPL-2.0
ac27a0ec 2/*
617ba13b 3 * linux/fs/ext4/super.c
ac27a0ec
DK		 4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/fs/minix/inode.c
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 *
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
18 */
19
20#include <linux/module.h>
21#include <linux/string.h>
22#include <linux/fs.h>
23#include <linux/time.h>
c5ca7c76 24#include <linux/vmalloc.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
66114cad 28#include <linux/backing-dev.h>
ac27a0ec 29#include <linux/parser.h>
ac27a0ec 30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ef510424 41#include <linux/dax.h>
7c0f6ba6 42#include <linux/uaccess.h>
ee73f9a5 43#include <linux/iversion.h>
c83ad55e 44#include <linux/unicode.h>
c6a564ff 45#include <linux/part_stat.h>
bfff6873
LC		 46#include <linux/kthread.h>
47#include <linux/freezer.h>
9a089b21 48#include <linux/fsnotify.h>
e5a185c2
LC		 49#include <linux/fs_context.h>
50#include <linux/fs_parser.h>
bfff6873 51
3dcf5451 52#include "ext4.h"
4a092d73 53#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 54#include "ext4_jbd2.h"
ac27a0ec
DK		 55#include "xattr.h"
56#include "acl.h"
3661d286 57#include "mballoc.h"
0c9ec4be 58#include "fsmap.h"
ac27a0ec 59
9bffad1e
TT		 60#define CREATE_TRACE_POINTS
61#include <trace/events/ext4.h>
62
0b75a840 63static struct ext4_lazy_init *ext4_li_info;
59ebc7fd 64static DEFINE_MUTEX(ext4_li_mtx);
e294a537 65static struct ratelimit_state ext4_mount_msg_ratelimit;
9f6200bb 66
617ba13b 67static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 68 unsigned long journal_devnum);
2adf6da8 69static int ext4_show_options(struct seq_file *seq, struct dentry *root);
2d01ddc8 70static void ext4_update_super(struct super_block *sb);
4392fbc4 71static int ext4_commit_super(struct super_block *sb);
11215630 72static int ext4_mark_recovery_complete(struct super_block *sb,
2b2d6d01 73 struct ext4_super_block *es);
11215630
JK		 74static int ext4_clear_journal_err(struct super_block *sb,
75 struct ext4_super_block *es);
617ba13b 76static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 77static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 78static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 79static int ext4_freeze(struct super_block *sb);
2035e776
TT		 80static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
bfff6873
LC		 82static void ext4_destroy_lazyinit_thread(void);
83static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 84static void ext4_clear_request_list(void);
c6cb7e77
EW		 85static struct inode *ext4_get_journal_inode(struct super_block *sb,
86 unsigned int journal_inum);
da812f61 87static int ext4_validate_options(struct fs_context *fc);
b6bd2435
LC		 88static int ext4_check_opt_consistency(struct fs_context *fc,
89 struct super_block *sb);
85456054 90static void ext4_apply_options(struct fs_context *fc, struct super_block *sb);
02f960f8 91static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param);
cebe85d5
LC		 92static int ext4_get_tree(struct fs_context *fc);
93static int ext4_reconfigure(struct fs_context *fc);
94static void ext4_fc_free(struct fs_context *fc);
95static int ext4_init_fs_context(struct fs_context *fc);
1489dffd 96static void ext4_kill_sb(struct super_block *sb);
cebe85d5 97static const struct fs_parameter_spec ext4_param_specs[];
ac27a0ec 98
e74031fd
JK		 99/*
100 * Lock ordering
101 *
e74031fd 102 * page fault path:
d4f5258e
JK		 103 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
104 * -> page lock -> i_data_sem (rw)
e74031fd
JK		 105 *
106 * buffered write path:
c1e8d7c6 107 * sb_start_write -> i_mutex -> mmap_lock
e74031fd
JK		 108 * sb_start_write -> i_mutex -> transaction start -> page lock ->
109 * i_data_sem (rw)
110 *
111 * truncate:
d4f5258e
JK		 112 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
113 * page lock
114 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
1d39834f 115 * i_data_sem (rw)
e74031fd
JK		 116 *
117 * direct IO:
c1e8d7c6 118 * sb_start_write -> i_mutex -> mmap_lock
1d39834f 119 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
e74031fd
JK		 120 *
121 * writepages:
122 * transaction start -> page lock(s) -> i_data_sem (rw)
123 */
124
02f960f8
LC		 125static const struct fs_context_operations ext4_context_ops = {
126 .parse_param = ext4_parse_param,
cebe85d5
LC		 127 .get_tree = ext4_get_tree,
128 .reconfigure = ext4_reconfigure,
129 .free = ext4_fc_free,
02f960f8
LC		 130};
131
132
c290ea01 133#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
2035e776 134static struct file_system_type ext2_fs_type = {
cebe85d5
LC		 135 .owner = THIS_MODULE,
136 .name = "ext2",
137 .init_fs_context = ext4_init_fs_context,
138 .parameters = ext4_param_specs,
1489dffd 139 .kill_sb = ext4_kill_sb,
cebe85d5 140 .fs_flags = FS_REQUIRES_DEV,
2035e776 141};
7f78e035 142MODULE_ALIAS_FS("ext2");
fa7614dd 143MODULE_ALIAS("ext2");
4b41828b 144#define IS_EXT2_SB(sb) ((sb)->s_type == &ext2_fs_type)
2035e776
TT		 145#else
146#define IS_EXT2_SB(sb) (0)
147#endif
148
149
ba69f9ab 150static struct file_system_type ext3_fs_type = {
cebe85d5
LC		 151 .owner = THIS_MODULE,
152 .name = "ext3",
153 .init_fs_context = ext4_init_fs_context,
154 .parameters = ext4_param_specs,
1489dffd 155 .kill_sb = ext4_kill_sb,
cebe85d5 156 .fs_flags = FS_REQUIRES_DEV,
ba69f9ab 157};
7f78e035 158MODULE_ALIAS_FS("ext3");
fa7614dd 159MODULE_ALIAS("ext3");
4b41828b 160#define IS_EXT3_SB(sb) ((sb)->s_type == &ext3_fs_type)
bd81d8ee 161
fa491b14 162
67c0f556 163static inline void __ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags,
fa491b14 164 bh_end_io_t *end_io)
165{
166 /*
167 * buffer's verified bit is no longer valid after reading from
168 * disk again due to write out error, clear it to make sure we
169 * recheck the buffer contents.
170 */
171 clear_buffer_verified(bh);
172
173 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
174 get_bh(bh);
1420c4a5 175 submit_bh(REQ_OP_READ | op_flags, bh);
fa491b14 176}
177
67c0f556 178void ext4_read_bh_nowait(struct buffer_head *bh, blk_opf_t op_flags,
fa491b14 179 bh_end_io_t *end_io)
180{
181 BUG_ON(!buffer_locked(bh));
182
183 if (ext4_buffer_uptodate(bh)) {
184 unlock_buffer(bh);
185 return;
186 }
187 __ext4_read_bh(bh, op_flags, end_io);
188}
189
67c0f556 190int ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags, bh_end_io_t *end_io)
fa491b14 191{
192 BUG_ON(!buffer_locked(bh));
193
194 if (ext4_buffer_uptodate(bh)) {
195 unlock_buffer(bh);
196 return 0;
197 }
198
199 __ext4_read_bh(bh, op_flags, end_io);
200
201 wait_on_buffer(bh);
202 if (buffer_uptodate(bh))
203 return 0;
204 return -EIO;
205}
206
67c0f556 207int ext4_read_bh_lock(struct buffer_head *bh, blk_opf_t op_flags, bool wait)
fa491b14 208{
0b73284c
ZY		 209 lock_buffer(bh);
210 if (!wait) {
fa491b14 211 ext4_read_bh_nowait(bh, op_flags, NULL);
212 return 0;
213 }
0b73284c 214 return ext4_read_bh(bh, op_flags, NULL);
fa491b14 215}
216
fb265c9c 217/*
8394a6ab 218 * This works like __bread_gfp() except it uses ERR_PTR for error
fb265c9c
TT		 219 * returns. Currently with sb_bread it's impossible to distinguish
220 * between ENOMEM and EIO situations (since both result in a NULL
221 * return.
222 */
8394a6ab 223static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
67c0f556
BVA		 224 sector_t block,
225 blk_opf_t op_flags, gfp_t gfp)
fb265c9c 226{
2d069c08 227 struct buffer_head *bh;
228 int ret;
fb265c9c 229
8394a6ab 230 bh = sb_getblk_gfp(sb, block, gfp);
fb265c9c
TT		 231 if (bh == NULL)
232 return ERR_PTR(-ENOMEM);
cf2834a5 233 if (ext4_buffer_uptodate(bh))
fb265c9c 234 return bh;
2d069c08 235
236 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
237 if (ret) {
238 put_bh(bh);
239 return ERR_PTR(ret);
240 }
241 return bh;
fb265c9c
TT		 242}
243
8394a6ab 244struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
67c0f556 245 blk_opf_t op_flags)
8394a6ab 246{
8a83ac54
MWO		 247 gfp_t gfp = mapping_gfp_constraint(sb->s_bdev->bd_inode->i_mapping,
248 ~__GFP_FS) | __GFP_MOVABLE;
249
250 return __ext4_sb_bread_gfp(sb, block, op_flags, gfp);
8394a6ab 251}
252
253struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
254 sector_t block)
255{
8a83ac54
MWO		 256 gfp_t gfp = mapping_gfp_constraint(sb->s_bdev->bd_inode->i_mapping,
257 ~__GFP_FS);
258
259 return __ext4_sb_bread_gfp(sb, block, 0, gfp);
8394a6ab 260}
261
5df1d412 262void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
263{
e509ad4d 264 struct buffer_head *bh = bdev_getblk(sb->s_bdev, block,
1cbf0a58 265 sb->s_blocksize, GFP_NOWAIT | __GFP_NOWARN);
5df1d412 266
267 if (likely(bh)) {
0b73284c
ZY		 268 if (trylock_buffer(bh))
269 ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL);
5df1d412 270 brelse(bh);
271 }
fb265c9c
TT		 272}
273
d25425f8
DW		 274static int ext4_verify_csum_type(struct super_block *sb,
275 struct ext4_super_block *es)
276{
e2b911c5 277 if (!ext4_has_feature_metadata_csum(sb))
d25425f8
DW		 278 return 1;
279
280 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
281}
282
bbc605cd
LC		 283__le32 ext4_superblock_csum(struct super_block *sb,
284 struct ext4_super_block *es)
a9c47317
DW		 285{
286 struct ext4_sb_info *sbi = EXT4_SB(sb);
287 int offset = offsetof(struct ext4_super_block, s_checksum);
288 __u32 csum;
289
290 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
291
292 return cpu_to_le32(csum);
293}
294
c197855e
SH		 295static int ext4_superblock_csum_verify(struct super_block *sb,
296 struct ext4_super_block *es)
a9c47317 297{
9aa5d32b 298 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 299 return 1;
300
301 return es->s_checksum == ext4_superblock_csum(sb, es);
302}
303
06db49e6 304void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 305{
06db49e6
TT		 306 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
307
9aa5d32b 308 if (!ext4_has_metadata_csum(sb))
a9c47317
DW		 309 return;
310
311 es->s_checksum = ext4_superblock_csum(sb, es);
312}
313
8fadc143
AR		 314ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
315 struct ext4_group_desc *bg)
bd81d8ee 316{
3a14589c 317 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 318 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 319 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 320}
321
8fadc143
AR		 322ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
323 struct ext4_group_desc *bg)
bd81d8ee 324{
5272f837 325 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 326 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 327 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 328}
329
8fadc143
AR		 330ext4_fsblk_t ext4_inode_table(struct super_block *sb,
331 struct ext4_group_desc *bg)
bd81d8ee 332{
5272f837 333 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 334 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 335 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 336}
337
021b65bb
TT		 338__u32 ext4_free_group_clusters(struct super_block *sb,
339 struct ext4_group_desc *bg)
560671a0
AK		 340{
341 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
342 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 343 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 344}
345
346__u32 ext4_free_inodes_count(struct super_block *sb,
347 struct ext4_group_desc *bg)
348{
349 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
350 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 351 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 352}
353
354__u32 ext4_used_dirs_count(struct super_block *sb,
355 struct ext4_group_desc *bg)
356{
357 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
358 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 359 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 360}
361
362__u32 ext4_itable_unused_count(struct super_block *sb,
363 struct ext4_group_desc *bg)
364{
365 return le16_to_cpu(bg->bg_itable_unused_lo) |
366 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 367 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 368}
369
8fadc143
AR		 370void ext4_block_bitmap_set(struct super_block *sb,
371 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 372{
3a14589c 373 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 374 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
375 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 376}
377
8fadc143
AR		 378void ext4_inode_bitmap_set(struct super_block *sb,
379 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 380{
5272f837 381 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 382 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
383 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 384}
385
8fadc143
AR		 386void ext4_inode_table_set(struct super_block *sb,
387 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 388{
5272f837 389 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 390 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
391 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 392}
393
021b65bb
TT		 394void ext4_free_group_clusters_set(struct super_block *sb,
395 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 396{
397 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
398 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
399 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
400}
401
402void ext4_free_inodes_set(struct super_block *sb,
403 struct ext4_group_desc *bg, __u32 count)
404{
405 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
406 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
407 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
408}
409
410void ext4_used_dirs_set(struct super_block *sb,
411 struct ext4_group_desc *bg, __u32 count)
412{
413 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
414 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
415 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
416}
417
418void ext4_itable_unused_set(struct super_block *sb,
419 struct ext4_group_desc *bg, __u32 count)
420{
421 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
422 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
423 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
424}
425
c92dc856 426static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
6a0678a7 427{
6a0678a7
AB		 428 now = clamp_val(now, 0, (1ull << 40) - 1);
429
430 *lo = cpu_to_le32(lower_32_bits(now));
431 *hi = upper_32_bits(now);
432}
433
434static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
435{
436 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
437}
438#define ext4_update_tstamp(es, tstamp) \
c92dc856
JK		 439 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
440 ktime_get_real_seconds())
6a0678a7
AB		 441#define ext4_get_tstamp(es, tstamp) \
442 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
d3d1faf6 443
ff0722de
VK		 444#define EXT4_SB_REFRESH_INTERVAL_SEC (3600) /* seconds (1 hour) */
445#define EXT4_SB_REFRESH_INTERVAL_KB (16384) /* kilobytes (16MB) */
446
447/*
448 * The ext4_maybe_update_superblock() function checks and updates the
449 * superblock if needed.
450 *
451 * This function is designed to update the on-disk superblock only under
452 * certain conditions to prevent excessive disk writes and unnecessary
453 * waking of the disk from sleep. The superblock will be updated if:
454 * 1. More than an hour has passed since the last superblock update, and
455 * 2. More than 16MB have been written since the last superblock update.
456 *
457 * @sb: The superblock
458 */
459static void ext4_maybe_update_superblock(struct super_block *sb)
460{
461 struct ext4_sb_info *sbi = EXT4_SB(sb);
462 struct ext4_super_block *es = sbi->s_es;
463 journal_t *journal = sbi->s_journal;
464 time64_t now;
465 __u64 last_update;
466 __u64 lifetime_write_kbytes;
467 __u64 diff_size;
468
469 if (sb_rdonly(sb) || !(sb->s_flags & SB_ACTIVE) ||
470 !journal || (journal->j_flags & JBD2_UNMOUNT))
471 return;
472
473 now = ktime_get_real_seconds();
474 last_update = ext4_get_tstamp(es, s_wtime);
475
476 if (likely(now - last_update < EXT4_SB_REFRESH_INTERVAL_SEC))
477 return;
478
479 lifetime_write_kbytes = sbi->s_kbytes_written +
480 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
481 sbi->s_sectors_written_start) >> 1);
482
483 /* Get the number of kilobytes not written to disk to account
484 * for statistics and compare with a multiple of 16 MB. This
485 * is used to determine when the next superblock commit should
486 * occur (i.e. not more often than once per 16MB if there was
487 * less written in an hour).
488 */
489 diff_size = lifetime_write_kbytes - le64_to_cpu(es->s_kbytes_written);
490
491 if (diff_size > EXT4_SB_REFRESH_INTERVAL_KB)
bb15cea2 492 schedule_work(&EXT4_SB(sb)->s_sb_upd_work);
ff0722de
VK		 493}
494
bdfe0cbd
TT		 495/*
496 * The del_gendisk() function uninitializes the disk-specific data
497 * structures, including the bdi structure, without telling anyone
498 * else. Once this happens, any attempt to call mark_buffer_dirty()
499 * (for example, by ext4_commit_super), will cause a kernel OOPS.
500 * This is a kludge to prevent these oops until we can put in a proper
501 * hook in del_gendisk() to inform the VFS and file system layers.
502 */
503static int block_device_ejected(struct super_block *sb)
504{
505 struct inode *bd_inode = sb->s_bdev->bd_inode;
506 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
507
508 return bdi->dev == NULL;
509}
510
18aadd47
BJ		 511static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
512{
513 struct super_block *sb = journal->j_private;
514 struct ext4_sb_info *sbi = EXT4_SB(sb);
515 int error = is_journal_aborted(journal);
5d3ee208 516 struct ext4_journal_cb_entry *jce;
18aadd47 517
5d3ee208 518 BUG_ON(txn->t_state == T_FINISHED);
a0154344
DJ		 519
520 ext4_process_freed_data(sb, txn->t_tid);
ff0722de 521 ext4_maybe_update_superblock(sb);
a0154344 522
18aadd47 523 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 524 while (!list_empty(&txn->t_private_list)) {
525 jce = list_entry(txn->t_private_list.next,
526 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 527 list_del_init(&jce->jce_list);
528 spin_unlock(&sbi->s_md_lock);
529 jce->jce_func(sb, jce, error);
530 spin_lock(&sbi->s_md_lock);
531 }
532 spin_unlock(&sbi->s_md_lock);
533}
1c13d5c0 534
afb585a9
MFO		 535/*
536 * This writepage callback for write_cache_pages()
537 * takes care of a few cases after page cleaning.
538 *
539 * write_cache_pages() already checks for dirty pages
540 * and calls clear_page_dirty_for_io(), which we want,
541 * to write protect the pages.
542 *
543 * However, we may have to redirty a page (see below.)
544 */
d585bdbe 545static int ext4_journalled_writepage_callback(struct folio *folio,
afb585a9
MFO		 546 struct writeback_control *wbc,
547 void *data)
548{
549 transaction_t *transaction = (transaction_t *) data;
550 struct buffer_head *bh, *head;
551 struct journal_head *jh;
552
d585bdbe 553 bh = head = folio_buffers(folio);
afb585a9
MFO		 554 do {
555 /*
556 * We have to redirty a page in these cases:
557 * 1) If buffer is dirty, it means the page was dirty because it
558 * contains a buffer that needs checkpointing. So the dirty bit
559 * needs to be preserved so that checkpointing writes the buffer
560 * properly.
561 * 2) If buffer is not part of the committing transaction
562 * (we may have just accidentally come across this buffer because
563 * inode range tracking is not exact) or if the currently running
564 * transaction already contains this buffer as well, dirty bit
565 * needs to be preserved so that the buffer gets writeprotected
566 * properly on running transaction's commit.
567 */
568 jh = bh2jh(bh);
569 if (buffer_dirty(bh) ||
570 (jh && (jh->b_transaction != transaction ||
571 jh->b_next_transaction))) {
d585bdbe 572 folio_redirty_for_writepage(wbc, folio);
afb585a9
MFO		 573 goto out;
574 }
575 } while ((bh = bh->b_this_page) != head);
576
577out:
578 return AOP_WRITEPAGE_ACTIVATE;
579}
580
581static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
582{
583 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
584 struct writeback_control wbc = {
585 .sync_mode = WB_SYNC_ALL,
586 .nr_to_write = LONG_MAX,
587 .range_start = jinode->i_dirty_start,
588 .range_end = jinode->i_dirty_end,
589 };
590
591 return write_cache_pages(mapping, &wbc,
592 ext4_journalled_writepage_callback,
593 jinode->i_transaction);
594}
595
596static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
597{
598 int ret;
599
600 if (ext4_should_journal_data(jinode->i_vfs_inode))
601 ret = ext4_journalled_submit_inode_data_buffers(jinode);
602 else
59205c8d 603 ret = ext4_normal_submit_inode_data_buffers(jinode);
afb585a9
MFO		 604 return ret;
605}
606
607static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
608{
609 int ret = 0;
610
611 if (!ext4_should_journal_data(jinode->i_vfs_inode))
612 ret = jbd2_journal_finish_inode_data_buffers(jinode);
613
614 return ret;
615}
616
1dc1097f
JK		 617static bool system_going_down(void)
618{
619 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
620 || system_state == SYSTEM_RESTART;
621}
622
02a7780e
JK		 623struct ext4_err_translation {
624 int code;
625 int errno;
626};
627
628#define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
629
630static struct ext4_err_translation err_translation[] = {
631 EXT4_ERR_TRANSLATE(EIO),
632 EXT4_ERR_TRANSLATE(ENOMEM),
633 EXT4_ERR_TRANSLATE(EFSBADCRC),
634 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
635 EXT4_ERR_TRANSLATE(ENOSPC),
636 EXT4_ERR_TRANSLATE(ENOKEY),
637 EXT4_ERR_TRANSLATE(EROFS),
638 EXT4_ERR_TRANSLATE(EFBIG),
639 EXT4_ERR_TRANSLATE(EEXIST),
640 EXT4_ERR_TRANSLATE(ERANGE),
641 EXT4_ERR_TRANSLATE(EOVERFLOW),
642 EXT4_ERR_TRANSLATE(EBUSY),
643 EXT4_ERR_TRANSLATE(ENOTDIR),
644 EXT4_ERR_TRANSLATE(ENOTEMPTY),
645 EXT4_ERR_TRANSLATE(ESHUTDOWN),
646 EXT4_ERR_TRANSLATE(EFAULT),
647};
648
649static int ext4_errno_to_code(int errno)
650{
651 int i;
652
653 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
654 if (err_translation[i].errno == errno)
655 return err_translation[i].code;
656 return EXT4_ERR_UNKNOWN;
657}
658
2d01ddc8
JK		 659static void save_error_info(struct super_block *sb, int error,
660 __u32 ino, __u64 block,
661 const char *func, unsigned int line)
40676623 662{
c92dc856 663 struct ext4_sb_info *sbi = EXT4_SB(sb);
40676623 664
02a7780e
JK		 665 /* We default to EFSCORRUPTED error... */
666 if (error == 0)
667 error = EFSCORRUPTED;
c92dc856
JK		 668
669 spin_lock(&sbi->s_error_lock);
670 sbi->s_add_error_count++;
671 sbi->s_last_error_code = error;
672 sbi->s_last_error_line = line;
673 sbi->s_last_error_ino = ino;
674 sbi->s_last_error_block = block;
675 sbi->s_last_error_func = func;
676 sbi->s_last_error_time = ktime_get_real_seconds();
677 if (!sbi->s_first_error_time) {
678 sbi->s_first_error_code = error;
679 sbi->s_first_error_line = line;
680 sbi->s_first_error_ino = ino;
681 sbi->s_first_error_block = block;
682 sbi->s_first_error_func = func;
683 sbi->s_first_error_time = sbi->s_last_error_time;
684 }
685 spin_unlock(&sbi->s_error_lock);
40676623
JK		 686}
687
ac27a0ec
DK		 688/* Deal with the reporting of failure conditions on a filesystem such as
689 * inconsistencies detected or read IO failures.
690 *
691 * On ext2, we can store the error state of the filesystem in the
617ba13b 692 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 693 * write ordering constraints on the superblock which prevent us from
694 * writing it out straight away; and given that the journal is about to
695 * be aborted, we can't rely on the current, or future, transactions to
696 * write out the superblock safely.
697 *
dab291af 698 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 699 * the journal instead. On recovery, the journal will complain about
ac27a0ec 700 * that error until we've noted it down and cleared it.
014c9caa
JK		 701 *
702 * If force_ro is set, we unconditionally force the filesystem into an
703 * ABORT|READONLY state, unless the error response on the fs has been set to
704 * panic in which case we take the easy way out and panic immediately. This is
705 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
706 * at a critical moment in log management.
ac27a0ec 707 */
e789ca0c
JK		 708static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
709 __u32 ino, __u64 block,
710 const char *func, unsigned int line)
ac27a0ec 711{
b08070ec 712 journal_t *journal = EXT4_SB(sb)->s_journal;
2d01ddc8 713 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
b08070ec 714
e789ca0c 715 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
327eaf73
TT		 716 if (test_opt(sb, WARN_ON_ERROR))
717 WARN_ON_ONCE(1);
718
2d01ddc8 719 if (!continue_fs && !sb_rdonly(sb)) {
95257987 720 set_bit(EXT4_FLAGS_SHUTDOWN, &EXT4_SB(sb)->s_ext4_flags);
2d01ddc8
JK		 721 if (journal)
722 jbd2_journal_abort(journal, -EIO);
723 }
724
725 if (!bdev_read_only(sb->s_bdev)) {
e789ca0c 726 save_error_info(sb, error, ino, block, func, line);
2d01ddc8
JK		 727 /*
728 * In case the fs should keep running, we need to writeout
729 * superblock through the journal. Due to lock ordering
730 * constraints, it may not be safe to do it right here so we
731 * defer superblock flushing to a workqueue.
732 */
bb9464e0 733 if (continue_fs && journal)
bb15cea2 734 schedule_work(&EXT4_SB(sb)->s_sb_upd_work);
2d01ddc8
JK		 735 else
736 ext4_commit_super(sb);
737 }
e789ca0c 738
1dc1097f
JK		 739 /*
740 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
741 * could panic during 'reboot -f' as the underlying device got already
742 * disabled.
743 */
014c9caa 744 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
617ba13b 745 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec 746 sb->s_id);
4327ba52 747 }
ac2f7ca5
YB		 748
749 if (sb_rdonly(sb) || continue_fs)
750 return;
751
014c9caa
JK		 752 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
753 /*
754 * Make sure updated value of ->s_mount_flags will be visible before
755 * ->s_flags update
756 */
757 smp_wmb();
758 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 759}
760
bb15cea2 761static void update_super_work(struct work_struct *work)
c92dc856
JK		 762{
763 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
bb15cea2 764 s_sb_upd_work);
2d01ddc8
JK		 765 journal_t *journal = sbi->s_journal;
766 handle_t *handle;
c92dc856 767
2d01ddc8
JK		 768 /*
769 * If the journal is still running, we have to write out superblock
770 * through the journal to avoid collisions of other journalled sb
771 * updates.
772 *
773 * We use directly jbd2 functions here to avoid recursing back into
774 * ext4 error handling code during handling of previous errors.
775 */
776 if (!sb_rdonly(sbi->s_sb) && journal) {
558d6450 777 struct buffer_head *sbh = sbi->s_sbh;
ee6a12d0
TT		 778 bool call_notify_err = false;
779
2d01ddc8
JK		 780 handle = jbd2_journal_start(journal, 1);
781 if (IS_ERR(handle))
782 goto write_directly;
558d6450 783 if (jbd2_journal_get_write_access(handle, sbh)) {
2d01ddc8
JK		 784 jbd2_journal_stop(handle);
785 goto write_directly;
786 }
ff0722de
VK		 787
788 if (sbi->s_add_error_count > 0)
789 call_notify_err = true;
790
2d01ddc8 791 ext4_update_super(sbi->s_sb);
558d6450
YB		 792 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
793 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
794 "superblock detected");
795 clear_buffer_write_io_error(sbh);
796 set_buffer_uptodate(sbh);
797 }
798
799 if (jbd2_journal_dirty_metadata(handle, sbh)) {
2d01ddc8
JK		 800 jbd2_journal_stop(handle);
801 goto write_directly;
802 }
803 jbd2_journal_stop(handle);
ff0722de
VK		 804
805 if (call_notify_err)
806 ext4_notify_error_sysfs(sbi);
807
2d01ddc8
JK		 808 return;
809 }
810write_directly:
811 /*
812 * Write through journal failed. Write sb directly to get error info
813 * out and hope for the best.
814 */
4392fbc4 815 ext4_commit_super(sbi->s_sb);
d578b994 816 ext4_notify_error_sysfs(sbi);
ac27a0ec
DK		 817}
818
efbed4dc
TT		 819#define ext4_error_ratelimit(sb) \
820 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
821 "EXT4-fs error")
822
12062ddd 823void __ext4_error(struct super_block *sb, const char *function,
014c9caa 824 unsigned int line, bool force_ro, int error, __u64 block,
54d3adbc 825 const char *fmt, ...)
ac27a0ec 826{
0ff2ea7d 827 struct va_format vaf;
ac27a0ec
DK		 828 va_list args;
829
eb8ab444 830 if (unlikely(ext4_forced_shutdown(sb)))
0db1ff22
TT		 831 return;
832
ccf0f32a 833 trace_ext4_error(sb, function, line);
efbed4dc
TT		 834 if (ext4_error_ratelimit(sb)) {
835 va_start(args, fmt);
836 vaf.fmt = fmt;
837 vaf.va = &args;
838 printk(KERN_CRIT
839 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
840 sb->s_id, function, line, current->comm, &vaf);
841 va_end(args);
842 }
9a089b21
GKB		 843 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
844
e789ca0c 845 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
ac27a0ec
DK		 846}
847
e7c96e8e 848void __ext4_error_inode(struct inode *inode, const char *function,
54d3adbc 849 unsigned int line, ext4_fsblk_t block, int error,
e7c96e8e 850 const char *fmt, ...)
273df556
FM		 851{
852 va_list args;
f7c21177 853 struct va_format vaf;
273df556 854
eb8ab444 855 if (unlikely(ext4_forced_shutdown(inode->i_sb)))
0db1ff22
TT		 856 return;
857
ccf0f32a 858 trace_ext4_error(inode->i_sb, function, line);
efbed4dc
TT		 859 if (ext4_error_ratelimit(inode->i_sb)) {
860 va_start(args, fmt);
861 vaf.fmt = fmt;
862 vaf.va = &args;
863 if (block)
864 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
865 "inode #%lu: block %llu: comm %s: %pV\n",
866 inode->i_sb->s_id, function, line, inode->i_ino,
867 block, current->comm, &vaf);
868 else
869 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
870 "inode #%lu: comm %s: %pV\n",
871 inode->i_sb->s_id, function, line, inode->i_ino,
872 current->comm, &vaf);
873 va_end(args);
874 }
9a089b21
GKB		 875 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
876
e789ca0c
JK		 877 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
878 function, line);
273df556
FM		 879}
880
e7c96e8e
JP		 881void __ext4_error_file(struct file *file, const char *function,
882 unsigned int line, ext4_fsblk_t block,
883 const char *fmt, ...)
273df556
FM		 884{
885 va_list args;
f7c21177 886 struct va_format vaf;
496ad9aa 887 struct inode *inode = file_inode(file);
273df556
FM		 888 char pathname[80], *path;
889
eb8ab444 890 if (unlikely(ext4_forced_shutdown(inode->i_sb)))
0db1ff22
TT		 891 return;
892
ccf0f32a 893 trace_ext4_error(inode->i_sb, function, line);
efbed4dc 894 if (ext4_error_ratelimit(inode->i_sb)) {
9bf39ab2 895 path = file_path(file, pathname, sizeof(pathname));
efbed4dc
TT		 896 if (IS_ERR(path))
897 path = "(unknown)";
898 va_start(args, fmt);
899 vaf.fmt = fmt;
900 vaf.va = &args;
901 if (block)
902 printk(KERN_CRIT
903 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
904 "block %llu: comm %s: path %s: %pV\n",
905 inode->i_sb->s_id, function, line, inode->i_ino,
906 block, current->comm, path, &vaf);
907 else
908 printk(KERN_CRIT
909 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
910 "comm %s: path %s: %pV\n",
911 inode->i_sb->s_id, function, line, inode->i_ino,
912 current->comm, path, &vaf);
913 va_end(args);
914 }
9a089b21
GKB		 915 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
916
e789ca0c
JK		 917 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
918 function, line);
273df556
FM		 919}
920
722887dd
TT		 921const char *ext4_decode_error(struct super_block *sb, int errno,
922 char nbuf[16])
ac27a0ec
DK		 923{
924 char *errstr = NULL;
925
926 switch (errno) {
6a797d27
DW		 927 case -EFSCORRUPTED:
928 errstr = "Corrupt filesystem";
929 break;
930 case -EFSBADCRC:
931 errstr = "Filesystem failed CRC";
932 break;
ac27a0ec
DK		 933 case -EIO:
934 errstr = "IO failure";
935 break;
936 case -ENOMEM:
937 errstr = "Out of memory";
938 break;
939 case -EROFS:
78f1ddbb
TT		 940 if (!sb || (EXT4_SB(sb)->s_journal &&
941 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 942 errstr = "Journal has aborted";
943 else
944 errstr = "Readonly filesystem";
945 break;
946 default:
947 /* If the caller passed in an extra buffer for unknown
948 * errors, textualise them now. Else we just return
949 * NULL. */
950 if (nbuf) {
951 /* Check for truncated error codes... */
952 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
953 errstr = nbuf;
954 }
955 break;
956 }
957
958 return errstr;
959}
960
617ba13b 961/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 962 * automatically and invokes the appropriate error response. */
963
c398eda0
TT		 964void __ext4_std_error(struct super_block *sb, const char *function,
965 unsigned int line, int errno)
ac27a0ec
DK		 966{
967 char nbuf[16];
968 const char *errstr;
969
eb8ab444 970 if (unlikely(ext4_forced_shutdown(sb)))
0db1ff22
TT		 971 return;
972
ac27a0ec
DK		 973 /* Special case: if the error is EROFS, and we're not already
974 * inside a transaction, then there's really no point in logging
975 * an error. */
bc98a42c 976 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
ac27a0ec
DK		 977 return;
978
efbed4dc
TT		 979 if (ext4_error_ratelimit(sb)) {
980 errstr = ext4_decode_error(sb, errno, nbuf);
981 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
982 sb->s_id, function, line, errstr);
983 }
9a089b21 984 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
ac27a0ec 985
e789ca0c 986 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
ac27a0ec
DK		 987}
988
e7c96e8e
JP		 989void __ext4_msg(struct super_block *sb,
990 const char *prefix, const char *fmt, ...)
b31e1552 991{
0ff2ea7d 992 struct va_format vaf;
b31e1552
ES		 993 va_list args;
994
da812f61
LC		 995 if (sb) {
996 atomic_inc(&EXT4_SB(sb)->s_msg_count);
997 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
998 "EXT4-fs"))
999 return;
1000 }
efbed4dc 1001
b31e1552 1002 va_start(args, fmt);
0ff2ea7d
JP		 1003 vaf.fmt = fmt;
1004 vaf.va = &args;
da812f61
LC		 1005 if (sb)
1006 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
1007 else
1008 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
b31e1552
ES		 1009 va_end(args);
1010}
1011
1cf006ed
DM		 1012static int ext4_warning_ratelimit(struct super_block *sb)
1013{
1014 atomic_inc(&EXT4_SB(sb)->s_warning_count);
1015 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
1016 "EXT4-fs warning");
1017}
b03a2f7e 1018
12062ddd 1019void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 1020 unsigned int line, const char *fmt, ...)
ac27a0ec 1021{
0ff2ea7d 1022 struct va_format vaf;
ac27a0ec
DK		 1023 va_list args;
1024
b03a2f7e 1025 if (!ext4_warning_ratelimit(sb))
efbed4dc
TT		 1026 return;
1027
ac27a0ec 1028 va_start(args, fmt);
0ff2ea7d
JP		 1029 vaf.fmt = fmt;
1030 vaf.va = &args;
1031 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
1032 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 1033 va_end(args);
1034}
1035
b03a2f7e
AD		 1036void __ext4_warning_inode(const struct inode *inode, const char *function,
1037 unsigned int line, const char *fmt, ...)
1038{
1039 struct va_format vaf;
1040 va_list args;
1041
1042 if (!ext4_warning_ratelimit(inode->i_sb))
1043 return;
1044
1045 va_start(args, fmt);
1046 vaf.fmt = fmt;
1047 vaf.va = &args;
1048 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
1049 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
1050 function, line, inode->i_ino, current->comm, &vaf);
1051 va_end(args);
1052}
1053
e29136f8
TT		 1054void __ext4_grp_locked_error(const char *function, unsigned int line,
1055 struct super_block *sb, ext4_group_t grp,
1056 unsigned long ino, ext4_fsblk_t block,
1057 const char *fmt, ...)
5d1b1b3f
AK		 1058__releases(bitlock)
1059__acquires(bitlock)
1060{
0ff2ea7d 1061 struct va_format vaf;
5d1b1b3f 1062 va_list args;
5d1b1b3f 1063
eb8ab444 1064 if (unlikely(ext4_forced_shutdown(sb)))
0db1ff22
TT		 1065 return;
1066
ccf0f32a 1067 trace_ext4_error(sb, function, line);
efbed4dc
TT		 1068 if (ext4_error_ratelimit(sb)) {
1069 va_start(args, fmt);
1070 vaf.fmt = fmt;
1071 vaf.va = &args;
1072 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1073 sb->s_id, function, line, grp);
1074 if (ino)
1075 printk(KERN_CONT "inode %lu: ", ino);
1076 if (block)
1077 printk(KERN_CONT "block %llu:",
1078 (unsigned long long) block);
1079 printk(KERN_CONT "%pV\n", &vaf);
1080 va_end(args);
1081 }
5d1b1b3f
AK		 1082
1083 if (test_opt(sb, ERRORS_CONT)) {
c92dc856
JK		 1084 if (test_opt(sb, WARN_ON_ERROR))
1085 WARN_ON_ONCE(1);
e789ca0c 1086 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2d01ddc8
JK		 1087 if (!bdev_read_only(sb->s_bdev)) {
1088 save_error_info(sb, EFSCORRUPTED, ino, block, function,
1089 line);
bb15cea2 1090 schedule_work(&EXT4_SB(sb)->s_sb_upd_work);
2d01ddc8 1091 }
5d1b1b3f
AK		 1092 return;
1093 }
1094 ext4_unlock_group(sb, grp);
e789ca0c 1095 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
5d1b1b3f
AK		 1096 /*
1097 * We only get here in the ERRORS_RO case; relocking the group
1098 * may be dangerous, but nothing bad will happen since the
1099 * filesystem will have already been marked read/only and the
1100 * journal has been aborted. We return 1 as a hint to callers
1101 * who might what to use the return value from
25985edc 1102 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 1103 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1104 * aggressively from the ext4 function in question, with a
1105 * more appropriate error code.
1106 */
1107 ext4_lock_group(sb, grp);
1108 return;
1109}
1110
db79e6d1
WS		 1111void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1112 ext4_group_t group,
1113 unsigned int flags)
1114{
1115 struct ext4_sb_info *sbi = EXT4_SB(sb);
1116 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1117 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
9af0b3d1
WS		 1118 int ret;
1119
5354b2af
TT		 1120 if (!grp || !gdp)
1121 return;
9af0b3d1
WS		 1122 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1123 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1124 &grp->bb_state);
1125 if (!ret)
1126 percpu_counter_sub(&sbi->s_freeclusters_counter,
1127 grp->bb_free);
db79e6d1
WS		 1128 }
1129
9af0b3d1
WS		 1130 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1131 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1132 &grp->bb_state);
1133 if (!ret && gdp) {
db79e6d1
WS		 1134 int count;
1135
1136 count = ext4_free_inodes_count(sb, gdp);
1137 percpu_counter_sub(&sbi->s_freeinodes_counter,
1138 count);
1139 }
db79e6d1
WS		 1140 }
1141}
1142
617ba13b 1143void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 1144{
617ba13b 1145 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 1146
617ba13b 1147 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 1148 return;
1149
12062ddd 1150 ext4_warning(sb,
ac27a0ec
DK		 1151 "updating to rev %d because of new feature flag, "
1152 "running e2fsck is recommended",
617ba13b 1153 EXT4_DYNAMIC_REV);
ac27a0ec 1154
617ba13b
MC		 1155 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1156 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1157 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 1158 /* leave es->s_feature_*compat flags alone */
1159 /* es->s_uuid will be set by e2fsck if empty */
1160
1161 /*
1162 * The rest of the superblock fields should be zero, and if not it
1163 * means they are likely already in use, so leave them alone. We
1164 * can leave it up to e2fsck to clean up any inconsistencies there.
1165 */
1166}
1167
ac27a0ec
DK		 1168static inline struct inode *orphan_list_entry(struct list_head *l)
1169{
617ba13b 1170 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 1171}
1172
617ba13b 1173static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 1174{
1175 struct list_head *l;
1176
b31e1552
ES		 1177 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1178 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 1179
1180 printk(KERN_ERR "sb_info orphan list:\n");
1181 list_for_each(l, &sbi->s_orphan) {
1182 struct inode *inode = orphan_list_entry(l);
1183 printk(KERN_ERR " "
1184 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1185 inode->i_sb->s_id, inode->i_ino, inode,
1186 inode->i_mode, inode->i_nlink,
1187 NEXT_ORPHAN(inode));
1188 }
1189}
1190
957153fc
JK		 1191#ifdef CONFIG_QUOTA
1192static int ext4_quota_off(struct super_block *sb, int type);
1193
f3c1c42e 1194static inline void ext4_quotas_off(struct super_block *sb, int type)
957153fc 1195{
f3c1c42e 1196 BUG_ON(type > EXT4_MAXQUOTAS);
957153fc 1197
964edf66 1198 /* Use our quota_off function to clear inode flags etc. */
f3c1c42e 1199 for (type--; type >= 0; type--)
964edf66 1200 ext4_quota_off(sb, type);
957153fc 1201}
33458eab
TT		 1202
1203/*
1204 * This is a helper function which is used in the mount/remount
1205 * codepaths (which holds s_umount) to fetch the quota file name.
1206 */
1207static inline char *get_qf_name(struct super_block *sb,
1208 struct ext4_sb_info *sbi,
1209 int type)
1210{
1211 return rcu_dereference_protected(sbi->s_qf_names[type],
1212 lockdep_is_held(&sb->s_umount));
1213}
957153fc 1214#else
f3c1c42e 1215static inline void ext4_quotas_off(struct super_block *sb, int type)
957153fc
JK		 1216{
1217}
1218#endif
1219
1f79467c
JY		 1220static int ext4_percpu_param_init(struct ext4_sb_info *sbi)
1221{
1222 ext4_fsblk_t block;
1223 int err;
1224
1225 block = ext4_count_free_clusters(sbi->s_sb);
1226 ext4_free_blocks_count_set(sbi->s_es, EXT4_C2B(sbi, block));
1227 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
1228 GFP_KERNEL);
1229 if (!err) {
1230 unsigned long freei = ext4_count_free_inodes(sbi->s_sb);
1231 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
1232 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
1233 GFP_KERNEL);
1234 }
1235 if (!err)
1236 err = percpu_counter_init(&sbi->s_dirs_counter,
1237 ext4_count_dirs(sbi->s_sb), GFP_KERNEL);
1238 if (!err)
1239 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
1240 GFP_KERNEL);
1241 if (!err)
1242 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
1243 GFP_KERNEL);
1244 if (!err)
1245 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
1246
1247 if (err)
1248 ext4_msg(sbi->s_sb, KERN_ERR, "insufficient memory");
1249
1250 return err;
1251}
1252
1253static void ext4_percpu_param_destroy(struct ext4_sb_info *sbi)
1254{
1255 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1256 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1257 percpu_counter_destroy(&sbi->s_dirs_counter);
1258 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1259 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1260 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1261}
1262
6ef68498
JY		 1263static void ext4_group_desc_free(struct ext4_sb_info *sbi)
1264{
1265 struct buffer_head **group_desc;
1266 int i;
1267
1268 rcu_read_lock();
1269 group_desc = rcu_dereference(sbi->s_group_desc);
1270 for (i = 0; i < sbi->s_gdb_count; i++)
1271 brelse(group_desc[i]);
1272 kvfree(group_desc);
1273 rcu_read_unlock();
1274}
1275
dcbf8758
JY		 1276static void ext4_flex_groups_free(struct ext4_sb_info *sbi)
1277{
1278 struct flex_groups **flex_groups;
1279 int i;
1280
1281 rcu_read_lock();
1282 flex_groups = rcu_dereference(sbi->s_flex_groups);
1283 if (flex_groups) {
1284 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1285 kvfree(flex_groups[i]);
1286 kvfree(flex_groups);
1287 }
1288 rcu_read_unlock();
1289}
1290
2b2d6d01 1291static void ext4_put_super(struct super_block *sb)
ac27a0ec 1292{
617ba13b
MC		 1293 struct ext4_sb_info *sbi = EXT4_SB(sb);
1294 struct ext4_super_block *es = sbi->s_es;
97abd7d4 1295 int aborted = 0;
856dd6c5 1296 int err;
ac27a0ec 1297
5e47868f
RH		 1298 /*
1299 * Unregister sysfs before destroying jbd2 journal.
1300 * Since we could still access attr_journal_task attribute via sysfs
1301 * path which could have sbi->s_journal->j_task as NULL
bb15cea2 1302 * Unregister sysfs before flush sbi->s_sb_upd_work.
b98535d0
YB		 1303 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1304 * read metadata verify failed then will queue error work.
bb15cea2 1305 * update_super_work will call start_this_handle may trigger
b98535d0 1306 * BUG_ON.
5e47868f
RH		 1307 */
1308 ext4_unregister_sysfs(sb);
1309
4808cb5b 1310 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
bb0fbc78
LC		 1311 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1312 &sb->s_uuid);
4808cb5b 1313
b98535d0 1314 ext4_unregister_li_request(sb);
f3c1c42e 1315 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
b98535d0 1316
bb15cea2 1317 flush_work(&sbi->s_sb_upd_work);
b98535d0
YB		 1318 destroy_workqueue(sbi->rsv_conversion_wq);
1319 ext4_release_orphan_info(sb);
1320
0390131b 1321 if (sbi->s_journal) {
97abd7d4 1322 aborted = is_journal_aborted(sbi->s_journal);
0390131b
FM		 1323 err = jbd2_journal_destroy(sbi->s_journal);
1324 sbi->s_journal = NULL;
878520ac 1325 if ((err < 0) && !aborted) {
54d3adbc 1326 ext4_abort(sb, -err, "Couldn't clean up the journal");
878520ac 1327 }
0390131b 1328 }
d4edac31 1329
d3922a77 1330 ext4_es_unregister_shrinker(sbi);
292a089d 1331 timer_shutdown_sync(&sbi->s_err_report);
d4edac31
JB		 1332 ext4_release_system_zone(sb);
1333 ext4_mb_release(sb);
1334 ext4_ext_release(sb);
d4edac31 1335
bc98a42c 1336 if (!sb_rdonly(sb) && !aborted) {
e2b911c5 1337 ext4_clear_feature_journal_needs_recovery(sb);
02f310fc 1338 ext4_clear_feature_orphan_present(sb);
ac27a0ec 1339 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 1340 }
bc98a42c 1341 if (!sb_rdonly(sb))
4392fbc4 1342 ext4_commit_super(sb);
a8e25a83 1343
6ef68498 1344 ext4_group_desc_free(sbi);
dcbf8758 1345 ext4_flex_groups_free(sbi);
1f79467c 1346 ext4_percpu_param_destroy(sbi);
ac27a0ec 1347#ifdef CONFIG_QUOTA
856dd6c5 1348 for (int i = 0; i < EXT4_MAXQUOTAS; i++)
33458eab 1349 kfree(get_qf_name(sb, sbi, i));
ac27a0ec
DK		 1350#endif
1351
1352 /* Debugging code just in case the in-memory inode orphan list
1353 * isn't empty. The on-disk one can be non-empty if we've
1354 * detected an error and taken the fs readonly, but the
1355 * in-memory list had better be clean by this point. */
1356 if (!list_empty(&sbi->s_orphan))
1357 dump_orphan_list(sb, sbi);
837c23fb 1358 ASSERT(list_empty(&sbi->s_orphan));
ac27a0ec 1359
89d96a6f 1360 sync_blockdev(sb->s_bdev);
f98393a6 1361 invalidate_bdev(sb->s_bdev);
61ead714 1362 if (sbi->s_journal_bdev_file) {
1489dffd
CH		 1363 /*
1364 * Invalidate the journal device's buffers. We don't want them
1365 * floating about in memory - the physical journal device may
1366 * hotswapped, and it breaks the `ro-after' testing code.
1367 */
61ead714
CB		 1368 sync_blockdev(file_bdev(sbi->s_journal_bdev_file));
1369 invalidate_bdev(file_bdev(sbi->s_journal_bdev_file));
ac27a0ec 1370 }
50c15df6
CX		 1371
1372 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1373 sbi->s_ea_inode_cache = NULL;
1374
1375 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1376 sbi->s_ea_block_cache = NULL;
1377
618f0031
PS		 1378 ext4_stop_mmpd(sbi);
1379
9060dd2c 1380 brelse(sbi->s_sbh);
ac27a0ec 1381 sb->s_fs_info = NULL;
3197ebdb
TT		 1382 /*
1383 * Now that we are completely done shutting down the
1384 * superblock, we need to actually destroy the kobject.
1385 */
3197ebdb
TT		 1386 kobject_put(&sbi->s_kobj);
1387 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 1388 if (sbi->s_chksum_driver)
1389 crypto_free_shash(sbi->s_chksum_driver);
705895b6 1390 kfree(sbi->s_blockgroup_lock);
8012b866 1391 fs_put_dax(sbi->s_daxdev, NULL);
ac4acb1f 1392 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5298d4bf 1393#if IS_ENABLED(CONFIG_UNICODE)
f8f4acb6 1394 utf8_unload(sb->s_encoding);
c83ad55e 1395#endif
ac27a0ec 1396 kfree(sbi);
ac27a0ec
DK		 1397}
1398
e18b890b 1399static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 1400
1401/*
1402 * Called inside transaction, so use GFP_NOFS
1403 */
617ba13b 1404static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 1405{
617ba13b 1406 struct ext4_inode_info *ei;
ac27a0ec 1407
fd60b288 1408 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 1409 if (!ei)
1410 return NULL;
0b8e58a1 1411
ee73f9a5 1412 inode_set_iversion(&ei->vfs_inode, 1);
7ea71af9 1413 ei->i_flags = 0;
202ee5df 1414 spin_lock_init(&ei->i_raw_lock);
38727786 1415 ei->i_prealloc_node = RB_ROOT;
27bc446e 1416 atomic_set(&ei->i_prealloc_active, 0);
38727786 1417 rwlock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 1418 ext4_es_init_tree(&ei->i_es_tree);
1419 rwlock_init(&ei->i_es_lock);
edaa53ca 1420 INIT_LIST_HEAD(&ei->i_es_list);
eb68d0e2 1421 ei->i_es_all_nr = 0;
edaa53ca 1422 ei->i_es_shk_nr = 0;
dd475925 1423 ei->i_es_shrink_lblk = 0;
d2a17637 1424 ei->i_reserved_data_blocks = 0;
d2a17637 1425 spin_lock_init(&(ei->i_block_reservation_lock));
1dc0aa46 1426 ext4_init_pending_tree(&ei->i_pending_tree);
a9e7f447
DM		 1427#ifdef CONFIG_QUOTA
1428 ei->i_reserved_quota = 0;
96c7e0d9 1429 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
a9e7f447 1430#endif
8aefcd55 1431 ei->jinode = NULL;
2e8fa54e 1432 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 1433 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 1434 ei->i_sync_tid = 0;
1435 ei->i_datasync_tid = 0;
e27f41e1 1436 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 1437 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
aa75f4d3
HS		 1438 ext4_fc_init_inode(&ei->vfs_inode);
1439 mutex_init(&ei->i_fc_lock);
ac27a0ec
DK		 1440 return &ei->vfs_inode;
1441}
1442
7ff9c073
TT		 1443static int ext4_drop_inode(struct inode *inode)
1444{
1445 int drop = generic_drop_inode(inode);
1446
29b3692e
EB		 1447 if (!drop)
1448 drop = fscrypt_drop_inode(inode);
1449
7ff9c073
TT		 1450 trace_ext4_drop_inode(inode, drop);
1451 return drop;
1452}
1453
94053139 1454static void ext4_free_in_core_inode(struct inode *inode)
fa0d7e3d 1455{
2c58d548 1456 fscrypt_free_inode(inode);
aa75f4d3
HS		 1457 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1458 pr_warn("%s: inode %ld still in fc list",
1459 __func__, inode->i_ino);
1460 }
fa0d7e3d
NP		 1461 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1462}
1463
617ba13b 1464static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 1465{
9f7dd93d 1466 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 1467 ext4_msg(inode->i_sb, KERN_ERR,
1468 "Inode %lu (%p): orphan list check failed!",
1469 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 1470 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1471 EXT4_I(inode), sizeof(struct ext4_inode_info),
1472 true);
1473 dump_stack();
1474 }
6fed8395
JX		 1475
1476 if (EXT4_I(inode)->i_reserved_data_blocks)
1477 ext4_msg(inode->i_sb, KERN_ERR,
1478 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1479 inode->i_ino, EXT4_I(inode),
1480 EXT4_I(inode)->i_reserved_data_blocks);
ac27a0ec
DK		 1481}
1482
f5db130d
CH		 1483static void ext4_shutdown(struct super_block *sb)
1484{
1485 ext4_force_shutdown(sb, EXT4_GOING_FLAGS_NOLOGFLUSH);
1486}
1487
51cc5068 1488static void init_once(void *foo)
ac27a0ec 1489{
c30365b9 1490 struct ext4_inode_info *ei = foo;
ac27a0ec 1491
a35afb83 1492 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1493 init_rwsem(&ei->xattr_sem);
0e855ac8 1494 init_rwsem(&ei->i_data_sem);
a35afb83 1495 inode_init_once(&ei->vfs_inode);
aa75f4d3 1496 ext4_fc_init_inode(&ei->vfs_inode);
ac27a0ec
DK		 1497}
1498
e67bc2b3 1499static int __init init_inodecache(void)
ac27a0ec 1500{
f8dd7c70
DW		 1501 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1502 sizeof(struct ext4_inode_info), 0,
f88c3fb8 1503 SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
f8dd7c70
DW		 1504 offsetof(struct ext4_inode_info, i_data),
1505 sizeof_field(struct ext4_inode_info, i_data),
1506 init_once);
617ba13b 1507 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1508 return -ENOMEM;
1509 return 0;
1510}
1511
1512static void destroy_inodecache(void)
1513{
8c0a8537
KS		 1514 /*
1515 * Make sure all delayed rcu free inodes are flushed before we
1516 * destroy cache.
1517 */
1518 rcu_barrier();
617ba13b 1519 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1520}
1521
0930fcc1 1522void ext4_clear_inode(struct inode *inode)
ac27a0ec 1523{
aa75f4d3 1524 ext4_fc_del(inode);
0930fcc1 1525 invalidate_inode_buffers(inode);
dbd5768f 1526 clear_inode(inode);
2ffd2a6a 1527 ext4_discard_preallocations(inode);
51865fda 1528 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
f4c2d372 1529 dquot_drop(inode);
8aefcd55
TT		 1530 if (EXT4_I(inode)->jinode) {
1531 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1532 EXT4_I(inode)->jinode);
1533 jbd2_free_inode(EXT4_I(inode)->jinode);
1534 EXT4_I(inode)->jinode = NULL;
1535 }
3d204e24 1536 fscrypt_put_encryption_info(inode);
c93d8f88 1537 fsverity_cleanup_inode(inode);
ac27a0ec
DK		 1538}
1539
1b961ac0 1540static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1541 u64 ino, u32 generation)
ac27a0ec 1542{
ac27a0ec 1543 struct inode *inode;
ac27a0ec 1544
8a363970 1545 /*
ac27a0ec
DK		 1546 * Currently we don't know the generation for parent directory, so
1547 * a generation of 0 means "accept any"
1548 */
8a363970 1549 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1d1fe1ee
DH		 1550 if (IS_ERR(inode))
1551 return ERR_CAST(inode);
1552 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1553 iput(inode);
1554 return ERR_PTR(-ESTALE);
1555 }
1b961ac0
CH		 1556
1557 return inode;
1558}
1559
1560static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1561 int fh_len, int fh_type)
1b961ac0
CH		 1562{
1563 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1564 ext4_nfs_get_inode);
1565}
1566
1567static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1568 int fh_len, int fh_type)
1b961ac0
CH		 1569{
1570 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1571 ext4_nfs_get_inode);
ac27a0ec
DK		 1572}
1573
fde87268
TT		 1574static int ext4_nfs_commit_metadata(struct inode *inode)
1575{
1576 struct writeback_control wbc = {
1577 .sync_mode = WB_SYNC_ALL
1578 };
1579
1580 trace_ext4_nfs_commit_metadata(inode);
1581 return ext4_write_inode(inode, &wbc);
1582}
1583
ac27a0ec 1584#ifdef CONFIG_QUOTA
d6006186 1585static const char * const quotatypes[] = INITQFNAMES;
689c958c 1586#define QTYPE2NAME(t) (quotatypes[t])
ac27a0ec 1587
617ba13b
MC		 1588static int ext4_write_dquot(struct dquot *dquot);
1589static int ext4_acquire_dquot(struct dquot *dquot);
1590static int ext4_release_dquot(struct dquot *dquot);
1591static int ext4_mark_dquot_dirty(struct dquot *dquot);
1592static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1593static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 1594 const struct path *path);
617ba13b 1595static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1596 size_t len, loff_t off);
617ba13b 1597static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1598 const char *data, size_t len, loff_t off);
7c319d32
AK		 1599static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1600 unsigned int flags);
ac27a0ec 1601
ccb49011 1602static struct dquot __rcu **ext4_get_dquots(struct inode *inode)
96c7e0d9
JK		 1603{
1604 return EXT4_I(inode)->i_dquot;
1605}
1606
61e225dc 1607static const struct dquot_operations ext4_quota_operations = {
7a9ca53a
TE		 1608 .get_reserved_space = ext4_get_reserved_space,
1609 .write_dquot = ext4_write_dquot,
1610 .acquire_dquot = ext4_acquire_dquot,
1611 .release_dquot = ext4_release_dquot,
1612 .mark_dirty = ext4_mark_dquot_dirty,
1613 .write_info = ext4_write_info,
1614 .alloc_dquot = dquot_alloc,
1615 .destroy_dquot = dquot_destroy,
1616 .get_projid = ext4_get_projid,
1617 .get_inode_usage = ext4_get_inode_usage,
ebc11f7b 1618 .get_next_id = dquot_get_next_id,
ac27a0ec
DK		 1619};
1620
0d54b217 1621static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1622 .quota_on = ext4_quota_on,
ca0e05e4 1623 .quota_off = ext4_quota_off,
287a8095 1624 .quota_sync = dquot_quota_sync,
0a240339 1625 .get_state = dquot_get_state,
287a8095
CH		 1626 .set_info = dquot_set_dqinfo,
1627 .get_dqblk = dquot_get_dqblk,
6332b9b5
ES		 1628 .set_dqblk = dquot_set_dqblk,
1629 .get_nextdqblk = dquot_get_next_dqblk,
ac27a0ec
DK		 1630};
1631#endif
1632
ee9b6d61 1633static const struct super_operations ext4_sops = {
617ba13b 1634 .alloc_inode = ext4_alloc_inode,
94053139 1635 .free_inode = ext4_free_in_core_inode,
617ba13b 1636 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1637 .write_inode = ext4_write_inode,
1638 .dirty_inode = ext4_dirty_inode,
7ff9c073 1639 .drop_inode = ext4_drop_inode,
0930fcc1 1640 .evict_inode = ext4_evict_inode,
617ba13b 1641 .put_super = ext4_put_super,
617ba13b 1642 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1643 .freeze_fs = ext4_freeze,
1644 .unfreeze_fs = ext4_unfreeze,
617ba13b 1645 .statfs = ext4_statfs,
617ba13b 1646 .show_options = ext4_show_options,
f5db130d 1647 .shutdown = ext4_shutdown,
ac27a0ec 1648#ifdef CONFIG_QUOTA
617ba13b
MC		 1649 .quota_read = ext4_quota_read,
1650 .quota_write = ext4_quota_write,
96c7e0d9 1651 .get_dquots = ext4_get_dquots,
ac27a0ec
DK		 1652#endif
1653};
1654
39655164 1655static const struct export_operations ext4_export_ops = {
e21fc203 1656 .encode_fh = generic_encode_ino32_fh,
1b961ac0
CH		 1657 .fh_to_dentry = ext4_fh_to_dentry,
1658 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1659 .get_parent = ext4_get_parent,
fde87268 1660 .commit_metadata = ext4_nfs_commit_metadata,
ac27a0ec
DK		 1661};
1662
1663enum {
1664 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
ba2e524d 1665 Opt_resgid, Opt_resuid, Opt_sb,
72578c33 1666 Opt_nouid32, Opt_debug, Opt_removed,
2d544ec9 1667 Opt_user_xattr, Opt_acl,
72578c33 1668 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1669 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1670 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1671 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
6ddb2447 1672 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
4f74d15f 1673 Opt_inlinecrypt,
ba2e524d 1674 Opt_usrjquota, Opt_grpjquota, Opt_quota,
ee4a3fcd 1675 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1ff20307 1676 Opt_usrquota, Opt_grpquota, Opt_prjquota,
9cb20f94 1677 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
327eaf73 1678 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
4437992b 1679 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1449032b 1680 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1681 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1682 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1683 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
cdb7ee4c 1684 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
21175ca4 1685 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
e5a185c2 1686 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
8016e29f 1687#ifdef CONFIG_EXT4_DEBUG
99c880de 1688 Opt_fc_debug_max_replay, Opt_fc_debug_force
8016e29f 1689#endif
ac27a0ec
DK		 1690};
1691
e5a185c2 1692static const struct constant_table ext4_param_errors[] = {
ba2e524d
LC		 1693 {"continue", EXT4_MOUNT_ERRORS_CONT},
1694 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1695 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
e5a185c2
LC		 1696 {}
1697};
1698
1699static const struct constant_table ext4_param_data[] = {
ba2e524d
LC		 1700 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1701 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1702 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
e5a185c2
LC		 1703 {}
1704};
1705
1706static const struct constant_table ext4_param_data_err[] = {
1707 {"abort", Opt_data_err_abort},
1708 {"ignore", Opt_data_err_ignore},
1709 {}
1710};
1711
1712static const struct constant_table ext4_param_jqfmt[] = {
ba2e524d
LC		 1713 {"vfsold", QFMT_VFS_OLD},
1714 {"vfsv0", QFMT_VFS_V0},
1715 {"vfsv1", QFMT_VFS_V1},
e5a185c2
LC		 1716 {}
1717};
1718
1719static const struct constant_table ext4_param_dax[] = {
1720 {"always", Opt_dax_always},
1721 {"inode", Opt_dax_inode},
1722 {"never", Opt_dax_never},
1723 {}
1724};
1725
e5a185c2
LC		 1726/*
1727 * Mount option specification
1728 * We don't use fsparam_flag_no because of the way we set the
1729 * options and the way we show them in _ext4_show_options(). To
1730 * keep the changes to a minimum, let's keep the negative options
1731 * separate for now.
1732 */
1733static const struct fs_parameter_spec ext4_param_specs[] = {
1734 fsparam_flag ("bsddf", Opt_bsd_df),
1735 fsparam_flag ("minixdf", Opt_minix_df),
1736 fsparam_flag ("grpid", Opt_grpid),
1737 fsparam_flag ("bsdgroups", Opt_grpid),
1738 fsparam_flag ("nogrpid", Opt_nogrpid),
1739 fsparam_flag ("sysvgroups", Opt_nogrpid),
1740 fsparam_u32 ("resgid", Opt_resgid),
1741 fsparam_u32 ("resuid", Opt_resuid),
1742 fsparam_u32 ("sb", Opt_sb),
1743 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1744 fsparam_flag ("nouid32", Opt_nouid32),
1745 fsparam_flag ("debug", Opt_debug),
1746 fsparam_flag ("oldalloc", Opt_removed),
1747 fsparam_flag ("orlov", Opt_removed),
1748 fsparam_flag ("user_xattr", Opt_user_xattr),
e5a185c2 1749 fsparam_flag ("acl", Opt_acl),
e5a185c2
LC		 1750 fsparam_flag ("norecovery", Opt_noload),
1751 fsparam_flag ("noload", Opt_noload),
1752 fsparam_flag ("bh", Opt_removed),
1753 fsparam_flag ("nobh", Opt_removed),
1754 fsparam_u32 ("commit", Opt_commit),
1755 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1756 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1757 fsparam_u32 ("journal_dev", Opt_journal_dev),
1758 fsparam_bdev ("journal_path", Opt_journal_path),
1759 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1760 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1761 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1762 fsparam_flag ("abort", Opt_abort),
1763 fsparam_enum ("data", Opt_data, ext4_param_data),
1764 fsparam_enum ("data_err", Opt_data_err,
1765 ext4_param_data_err),
1766 fsparam_string_empty
1767 ("usrjquota", Opt_usrjquota),
1768 fsparam_string_empty
1769 ("grpjquota", Opt_grpjquota),
1770 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1771 fsparam_flag ("grpquota", Opt_grpquota),
1772 fsparam_flag ("quota", Opt_quota),
1773 fsparam_flag ("noquota", Opt_noquota),
1774 fsparam_flag ("usrquota", Opt_usrquota),
1775 fsparam_flag ("prjquota", Opt_prjquota),
1776 fsparam_flag ("barrier", Opt_barrier),
1777 fsparam_u32 ("barrier", Opt_barrier),
1778 fsparam_flag ("nobarrier", Opt_nobarrier),
1ff20307 1779 fsparam_flag ("i_version", Opt_removed),
e5a185c2
LC		 1780 fsparam_flag ("dax", Opt_dax),
1781 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1782 fsparam_u32 ("stripe", Opt_stripe),
1783 fsparam_flag ("delalloc", Opt_delalloc),
1784 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1785 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1786 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
e5a185c2
LC		 1787 fsparam_u32 ("debug_want_extra_isize",
1788 Opt_debug_want_extra_isize),
1789 fsparam_flag ("mblk_io_submit", Opt_removed),
1790 fsparam_flag ("nomblk_io_submit", Opt_removed),
1791 fsparam_flag ("block_validity", Opt_block_validity),
1792 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1793 fsparam_u32 ("inode_readahead_blks",
1794 Opt_inode_readahead_blks),
1795 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1796 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1797 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1798 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1799 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1800 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1801 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1802 fsparam_flag ("discard", Opt_discard),
1803 fsparam_flag ("nodiscard", Opt_nodiscard),
1804 fsparam_u32 ("init_itable", Opt_init_itable),
1805 fsparam_flag ("init_itable", Opt_init_itable),
1806 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1807#ifdef CONFIG_EXT4_DEBUG
1808 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1809 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1810#endif
1811 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1812 fsparam_flag ("test_dummy_encryption",
1813 Opt_test_dummy_encryption),
1814 fsparam_string ("test_dummy_encryption",
1815 Opt_test_dummy_encryption),
1816 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1817 fsparam_flag ("nombcache", Opt_nombcache),
1818 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1819 fsparam_flag ("prefetch_block_bitmaps",
1820 Opt_removed),
1821 fsparam_flag ("no_prefetch_block_bitmaps",
1822 Opt_no_prefetch_block_bitmaps),
1823 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1824 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1825 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1826 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1827 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1828 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1829 {}
1830};
1831
b3881f74 1832#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
196e402a 1833
26092bf5
TT		 1834#define MOPT_SET 0x0001
1835#define MOPT_CLEAR 0x0002
1836#define MOPT_NOSUPPORT 0x0004
1837#define MOPT_EXPLICIT 0x0008
ac27a0ec 1838#ifdef CONFIG_QUOTA
26092bf5 1839#define MOPT_Q 0
ba2e524d 1840#define MOPT_QFMT 0x0010
26092bf5
TT		 1841#else
1842#define MOPT_Q MOPT_NOSUPPORT
1843#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1844#endif
ba2e524d
LC		 1845#define MOPT_NO_EXT2 0x0020
1846#define MOPT_NO_EXT3 0x0040
8dc0aa8c 1847#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ba2e524d
LC		 1848#define MOPT_SKIP 0x0080
1849#define MOPT_2 0x0100
26092bf5
TT		 1850
1851static const struct mount_opts {
1852 int token;
1853 int mount_opt;
1854 int flags;
1855} ext4_mount_opts[] = {
1856 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1857 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1858 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1859 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1860 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1861 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1862 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1863 MOPT_EXT4_ONLY | MOPT_SET},
1864 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1865 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1866 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1867 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1868 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1869 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1870 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1871 MOPT_EXT4_ONLY | MOPT_CLEAR},
327eaf73
TT		 1872 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1873 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
cb8435dc 1874 {Opt_commit, 0, MOPT_NO_EXT2},
c6d3d56d
DW		 1875 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1876 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c 1877 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1e381f60 1878 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
26092bf5 1879 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c 1880 EXT4_MOUNT_JOURNAL_CHECKSUM),
1e381f60 1881 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
8dc0aa8c 1882 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
ba2e524d 1883 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
26092bf5
TT		 1884 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1885 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1886 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1887 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1888 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
ba2e524d
LC		 1889 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1890 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1891 {Opt_journal_path, 0, MOPT_NO_EXT2},
1892 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1893 {Opt_data, 0, MOPT_NO_EXT2},
26092bf5 1894 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
03010a33 1895#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5 1896 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
ac27a0ec 1897#else
26092bf5 1898 {Opt_acl, 0, MOPT_NOSUPPORT},
ac27a0ec 1899#endif
26092bf5
TT		 1900 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1901 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1902 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1903 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1904 MOPT_SET | MOPT_Q},
1905 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1906 MOPT_SET | MOPT_Q},
49da9392
JK		 1907 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1908 MOPT_SET | MOPT_Q},
26092bf5 1909 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
49da9392
JK		 1910 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1911 MOPT_CLEAR | MOPT_Q},
ba2e524d
LC		 1912 {Opt_usrjquota, 0, MOPT_Q},
1913 {Opt_grpjquota, 0, MOPT_Q},
1914 {Opt_jqfmt, 0, MOPT_QFMT},
cdb7ee4c 1915 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
21175ca4 1916 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
3d392b26 1917 MOPT_SET},
99c880de 1918#ifdef CONFIG_EXT4_DEBUG
0f0672ff
HS		 1919 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1920 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
8016e29f 1921#endif
22b8d707 1922 {Opt_abort, EXT4_MOUNT2_ABORT, MOPT_SET | MOPT_2},
26092bf5
TT		 1923 {Opt_err, 0, 0}
1924};
1925
5298d4bf 1926#if IS_ENABLED(CONFIG_UNICODE)
c83ad55e
GKB		 1927static const struct ext4_sb_encodings {
1928 __u16 magic;
1929 char *name;
49bd03cc 1930 unsigned int version;
c83ad55e 1931} ext4_sb_encoding_map[] = {
49bd03cc 1932 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
c83ad55e
GKB		 1933};
1934
aa8bf298
CH		 1935static const struct ext4_sb_encodings *
1936ext4_sb_read_encoding(const struct ext4_super_block *es)
c83ad55e
GKB		 1937{
1938 __u16 magic = le16_to_cpu(es->s_encoding);
1939 int i;
1940
1941 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1942 if (magic == ext4_sb_encoding_map[i].magic)
aa8bf298 1943 return &ext4_sb_encoding_map[i];
c83ad55e 1944
aa8bf298 1945 return NULL;
c83ad55e
GKB		 1946}
1947#endif
1948
6e47a3cc
LC		 1949#define EXT4_SPEC_JQUOTA (1 << 0)
1950#define EXT4_SPEC_JQFMT (1 << 1)
1951#define EXT4_SPEC_DATAJ (1 << 2)
1952#define EXT4_SPEC_SB_BLOCK (1 << 3)
1953#define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1954#define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
6e47a3cc
LC		 1955#define EXT4_SPEC_s_want_extra_isize (1 << 7)
1956#define EXT4_SPEC_s_max_batch_time (1 << 8)
1957#define EXT4_SPEC_s_min_batch_time (1 << 9)
1958#define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1959#define EXT4_SPEC_s_li_wait_mult (1 << 11)
1960#define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1961#define EXT4_SPEC_s_stripe (1 << 13)
1962#define EXT4_SPEC_s_resuid (1 << 14)
1963#define EXT4_SPEC_s_resgid (1 << 15)
1964#define EXT4_SPEC_s_commit_interval (1 << 16)
1965#define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
7edfd85b 1966#define EXT4_SPEC_s_sb_block (1 << 18)
27b38686 1967#define EXT4_SPEC_mb_optimize_scan (1 << 19)
6e47a3cc 1968
461c3af0 1969struct ext4_fs_context {
e6e268cb 1970 char *s_qf_names[EXT4_MAXQUOTAS];
85456054 1971 struct fscrypt_dummy_policy dummy_enc_policy;
e6e268cb 1972 int s_jquota_fmt; /* Format of quota to use */
6e47a3cc
LC		 1973#ifdef CONFIG_EXT4_DEBUG
1974 int s_fc_debug_max_replay;
1975#endif
e6e268cb 1976 unsigned short qname_spec;
6e47a3cc
LC		 1977 unsigned long vals_s_flags; /* Bits to set in s_flags */
1978 unsigned long mask_s_flags; /* Bits changed in s_flags */
e6e268cb 1979 unsigned long journal_devnum;
6e47a3cc
LC		 1980 unsigned long s_commit_interval;
1981 unsigned long s_stripe;
1982 unsigned int s_inode_readahead_blks;
1983 unsigned int s_want_extra_isize;
1984 unsigned int s_li_wait_mult;
1985 unsigned int s_max_dir_size_kb;
e6e268cb 1986 unsigned int journal_ioprio;
6e47a3cc
LC		 1987 unsigned int vals_s_mount_opt;
1988 unsigned int mask_s_mount_opt;
1989 unsigned int vals_s_mount_opt2;
1990 unsigned int mask_s_mount_opt2;
b6bd2435 1991 unsigned int opt_flags; /* MOPT flags */
6e47a3cc
LC		 1992 unsigned int spec;
1993 u32 s_max_batch_time;
1994 u32 s_min_batch_time;
1995 kuid_t s_resuid;
1996 kgid_t s_resgid;
7edfd85b 1997 ext4_fsblk_t s_sb_block;
b237e304
HS		 1998};
1999
cebe85d5
LC		 2000static void ext4_fc_free(struct fs_context *fc)
2001{
2002 struct ext4_fs_context *ctx = fc->fs_private;
2003 int i;
2004
2005 if (!ctx)
2006 return;
2007
2008 for (i = 0; i < EXT4_MAXQUOTAS; i++)
2009 kfree(ctx->s_qf_names[i]);
2010
85456054 2011 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
cebe85d5
LC		 2012 kfree(ctx);
2013}
2014
2015int ext4_init_fs_context(struct fs_context *fc)
2016{
da9e4802 2017 struct ext4_fs_context *ctx;
cebe85d5
LC		 2018
2019 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2020 if (!ctx)
2021 return -ENOMEM;
2022
2023 fc->fs_private = ctx;
2024 fc->ops = &ext4_context_ops;
2025
2026 return 0;
2027}
2028
e6e268cb
LC		 2029#ifdef CONFIG_QUOTA
2030/*
2031 * Note the name of the specified quota file.
2032 */
2033static int note_qf_name(struct fs_context *fc, int qtype,
2034 struct fs_parameter *param)
2035{
2036 struct ext4_fs_context *ctx = fc->fs_private;
2037 char *qname;
2038
2039 if (param->size < 1) {
2040 ext4_msg(NULL, KERN_ERR, "Missing quota name");
2041 return -EINVAL;
2042 }
2043 if (strchr(param->string, '/')) {
2044 ext4_msg(NULL, KERN_ERR,
2045 "quotafile must be on filesystem root");
2046 return -EINVAL;
2047 }
2048 if (ctx->s_qf_names[qtype]) {
2049 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
2050 ext4_msg(NULL, KERN_ERR,
2051 "%s quota file already specified",
2052 QTYPE2NAME(qtype));
2053 return -EINVAL;
2054 }
2055 return 0;
2056 }
2057
2058 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
2059 if (!qname) {
2060 ext4_msg(NULL, KERN_ERR,
2061 "Not enough memory for storing quotafile name");
2062 return -ENOMEM;
2063 }
2064 ctx->s_qf_names[qtype] = qname;
2065 ctx->qname_spec |= 1 << qtype;
6e47a3cc 2066 ctx->spec |= EXT4_SPEC_JQUOTA;
e6e268cb
LC		 2067 return 0;
2068}
2069
2070/*
2071 * Clear the name of the specified quota file.
2072 */
2073static int unnote_qf_name(struct fs_context *fc, int qtype)
2074{
2075 struct ext4_fs_context *ctx = fc->fs_private;
2076
ea7d09ad 2077 kfree(ctx->s_qf_names[qtype]);
e6e268cb
LC		 2078
2079 ctx->s_qf_names[qtype] = NULL;
2080 ctx->qname_spec |= 1 << qtype;
6e47a3cc 2081 ctx->spec |= EXT4_SPEC_JQUOTA;
e6e268cb
LC		 2082 return 0;
2083}
2084#endif
2085
85456054
EB		 2086static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2087 struct ext4_fs_context *ctx)
2088{
2089 int err;
2090
2091 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2092 ext4_msg(NULL, KERN_WARNING,
2093 "test_dummy_encryption option not supported");
2094 return -EINVAL;
2095 }
2096 err = fscrypt_parse_test_dummy_encryption(param,
2097 &ctx->dummy_enc_policy);
2098 if (err == -EINVAL) {
2099 ext4_msg(NULL, KERN_WARNING,
2100 "Value of option \"%s\" is unrecognized", param->key);
2101 } else if (err == -EEXIST) {
2102 ext4_msg(NULL, KERN_WARNING,
2103 "Conflicting test_dummy_encryption options");
2104 return -EINVAL;
2105 }
2106 return err;
2107}
2108
6e47a3cc 2109#define EXT4_SET_CTX(name) \
4c246728
LC		 2110static inline void ctx_set_##name(struct ext4_fs_context *ctx, \
2111 unsigned long flag) \
6e47a3cc
LC		 2112{ \
2113 ctx->mask_s_##name |= flag; \
2114 ctx->vals_s_##name |= flag; \
e3952fcc
LC		 2115}
2116
2117#define EXT4_CLEAR_CTX(name) \
4c246728
LC		 2118static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \
2119 unsigned long flag) \
6e47a3cc
LC		 2120{ \
2121 ctx->mask_s_##name |= flag; \
2122 ctx->vals_s_##name &= ~flag; \
e3952fcc
LC		 2123}
2124
2125#define EXT4_TEST_CTX(name) \
4c246728
LC		 2126static inline unsigned long \
2127ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
6e47a3cc 2128{ \
4c246728 2129 return (ctx->vals_s_##name & flag); \
e3952fcc 2130}
6e47a3cc 2131
e3952fcc 2132EXT4_SET_CTX(flags); /* set only */
6e47a3cc 2133EXT4_SET_CTX(mount_opt);
e3952fcc
LC		 2134EXT4_CLEAR_CTX(mount_opt);
2135EXT4_TEST_CTX(mount_opt);
6e47a3cc 2136EXT4_SET_CTX(mount_opt2);
e3952fcc
LC		 2137EXT4_CLEAR_CTX(mount_opt2);
2138EXT4_TEST_CTX(mount_opt2);
2139
02f960f8 2140static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
26092bf5 2141{
461c3af0 2142 struct ext4_fs_context *ctx = fc->fs_private;
461c3af0 2143 struct fs_parse_result result;
26092bf5 2144 const struct mount_opts *m;
461c3af0 2145 int is_remount;
08cefc7a
EB		 2146 kuid_t uid;
2147 kgid_t gid;
461c3af0
LC		 2148 int token;
2149
2150 token = fs_parse(fc, ext4_param_specs, param, &result);
2151 if (token < 0)
2152 return token;
2153 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
26092bf5 2154
ba2e524d
LC		 2155 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2156 if (token == m->token)
2157 break;
2158
2159 ctx->opt_flags |= m->flags;
2160
2161 if (m->flags & MOPT_EXPLICIT) {
2162 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2163 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2164 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2165 ctx_set_mount_opt2(ctx,
2166 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2167 } else
2168 return -EINVAL;
2169 }
2170
2171 if (m->flags & MOPT_NOSUPPORT) {
2172 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2173 param->key);
2174 return 0;
2175 }
2176
2177 switch (token) {
57f73c2c 2178#ifdef CONFIG_QUOTA
ba2e524d 2179 case Opt_usrjquota:
461c3af0 2180 if (!*param->string)
e6e268cb 2181 return unnote_qf_name(fc, USRQUOTA);
461c3af0 2182 else
e6e268cb 2183 return note_qf_name(fc, USRQUOTA, param);
ba2e524d 2184 case Opt_grpjquota:
461c3af0 2185 if (!*param->string)
e6e268cb 2186 return unnote_qf_name(fc, GRPQUOTA);
461c3af0 2187 else
e6e268cb 2188 return note_qf_name(fc, GRPQUOTA, param);
57f73c2c 2189#endif
26092bf5 2190 case Opt_sb:
7edfd85b
LC		 2191 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2192 ext4_msg(NULL, KERN_WARNING,
2193 "Ignoring %s option on remount", param->key);
2194 } else {
2195 ctx->s_sb_block = result.uint_32;
2196 ctx->spec |= EXT4_SPEC_s_sb_block;
2197 }
02f960f8 2198 return 0;
26092bf5 2199 case Opt_removed:
da812f61 2200 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
461c3af0 2201 param->key);
02f960f8 2202 return 0;
4f74d15f
EB		 2203 case Opt_inlinecrypt:
2204#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
6e47a3cc 2205 ctx_set_flags(ctx, SB_INLINECRYPT);
4f74d15f 2206#else
da812f61 2207 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
4f74d15f 2208#endif
02f960f8 2209 return 0;
461c3af0 2210 case Opt_errors:
6e47a3cc 2211 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
ba2e524d
LC		 2212 ctx_set_mount_opt(ctx, result.uint_32);
2213 return 0;
2214#ifdef CONFIG_QUOTA
2215 case Opt_jqfmt:
2216 ctx->s_jquota_fmt = result.uint_32;
2217 ctx->spec |= EXT4_SPEC_JQFMT;
2218 return 0;
2219#endif
2220 case Opt_data:
2221 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2222 ctx_set_mount_opt(ctx, result.uint_32);
2223 ctx->spec |= EXT4_SPEC_DATAJ;
2224 return 0;
2225 case Opt_commit:
461c3af0 2226 if (result.uint_32 == 0)
934b0de1 2227 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE;
461c3af0 2228 else if (result.uint_32 > INT_MAX / HZ) {
da812f61 2229 ext4_msg(NULL, KERN_ERR,
9ba55543 2230 "Invalid commit interval %d, "
2231 "must be smaller than %d",
461c3af0 2232 result.uint_32, INT_MAX / HZ);
da812f61 2233 return -EINVAL;
9ba55543 2234 }
6e47a3cc
LC		 2235 ctx->s_commit_interval = HZ * result.uint_32;
2236 ctx->spec |= EXT4_SPEC_s_commit_interval;
ba2e524d
LC		 2237 return 0;
2238 case Opt_debug_want_extra_isize:
6e47a3cc 2239 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
da812f61 2240 ext4_msg(NULL, KERN_ERR,
461c3af0 2241 "Invalid want_extra_isize %d", result.uint_32);
da812f61 2242 return -EINVAL;
9803387c 2243 }
6e47a3cc
LC		 2244 ctx->s_want_extra_isize = result.uint_32;
2245 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
ba2e524d
LC		 2246 return 0;
2247 case Opt_max_batch_time:
6e47a3cc
LC		 2248 ctx->s_max_batch_time = result.uint_32;
2249 ctx->spec |= EXT4_SPEC_s_max_batch_time;
ba2e524d
LC		 2250 return 0;
2251 case Opt_min_batch_time:
6e47a3cc
LC		 2252 ctx->s_min_batch_time = result.uint_32;
2253 ctx->spec |= EXT4_SPEC_s_min_batch_time;
ba2e524d
LC		 2254 return 0;
2255 case Opt_inode_readahead_blks:
461c3af0
LC		 2256 if (result.uint_32 &&
2257 (result.uint_32 > (1 << 30) ||
2258 !is_power_of_2(result.uint_32))) {
da812f61 2259 ext4_msg(NULL, KERN_ERR,
e33e60ea
JK		 2260 "EXT4-fs: inode_readahead_blks must be "
2261 "0 or a power of 2 smaller than 2^31");
da812f61 2262 return -EINVAL;
5f3633e3 2263 }
6e47a3cc
LC		 2264 ctx->s_inode_readahead_blks = result.uint_32;
2265 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
ba2e524d
LC		 2266 return 0;
2267 case Opt_init_itable:
6e47a3cc
LC		 2268 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2269 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
461c3af0 2270 if (param->type == fs_value_is_string)
6e47a3cc
LC		 2271 ctx->s_li_wait_mult = result.uint_32;
2272 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
ba2e524d
LC		 2273 return 0;
2274 case Opt_max_dir_size_kb:
6e47a3cc
LC		 2275 ctx->s_max_dir_size_kb = result.uint_32;
2276 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
ba2e524d 2277 return 0;
8016e29f 2278#ifdef CONFIG_EXT4_DEBUG
ba2e524d 2279 case Opt_fc_debug_max_replay:
6e47a3cc
LC		 2280 ctx->s_fc_debug_max_replay = result.uint_32;
2281 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
ba2e524d 2282 return 0;
8016e29f 2283#endif
ba2e524d 2284 case Opt_stripe:
6e47a3cc
LC		 2285 ctx->s_stripe = result.uint_32;
2286 ctx->spec |= EXT4_SPEC_s_stripe;
ba2e524d
LC		 2287 return 0;
2288 case Opt_resuid:
461c3af0 2289 uid = make_kuid(current_user_ns(), result.uint_32);
5f3633e3 2290 if (!uid_valid(uid)) {
da812f61 2291 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
461c3af0 2292 result.uint_32);
da812f61 2293 return -EINVAL;
26092bf5 2294 }
6e47a3cc
LC		 2295 ctx->s_resuid = uid;
2296 ctx->spec |= EXT4_SPEC_s_resuid;
ba2e524d
LC		 2297 return 0;
2298 case Opt_resgid:
461c3af0 2299 gid = make_kgid(current_user_ns(), result.uint_32);
5f3633e3 2300 if (!gid_valid(gid)) {
da812f61 2301 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
461c3af0 2302 result.uint_32);
da812f61 2303 return -EINVAL;
5f3633e3 2304 }
6e47a3cc
LC		 2305 ctx->s_resgid = gid;
2306 ctx->spec |= EXT4_SPEC_s_resgid;
ba2e524d
LC		 2307 return 0;
2308 case Opt_journal_dev:
5f3633e3 2309 if (is_remount) {
da812f61 2310 ext4_msg(NULL, KERN_ERR,
5f3633e3 2311 "Cannot specify journal on remount");
da812f61 2312 return -EINVAL;
5f3633e3 2313 }
461c3af0 2314 ctx->journal_devnum = result.uint_32;
6e47a3cc 2315 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
ba2e524d
LC		 2316 return 0;
2317 case Opt_journal_path:
2318 {
ad4eec61
ES		 2319 struct inode *journal_inode;
2320 struct path path;
2321 int error;
2322
2323 if (is_remount) {
da812f61 2324 ext4_msg(NULL, KERN_ERR,
ad4eec61 2325 "Cannot specify journal on remount");
da812f61 2326 return -EINVAL;
ad4eec61 2327 }
ad4eec61 2328
e3ea75ee 2329 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path);
ad4eec61 2330 if (error) {
da812f61 2331 ext4_msg(NULL, KERN_ERR, "error: could not find "
461c3af0 2332 "journal device path");
da812f61 2333 return -EINVAL;
ad4eec61
ES		 2334 }
2335
2b0143b5 2336 journal_inode = d_inode(path.dentry);
461c3af0 2337 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
6e47a3cc 2338 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
ad4eec61 2339 path_put(&path);
ba2e524d
LC		 2340 return 0;
2341 }
2342 case Opt_journal_ioprio:
461c3af0 2343 if (result.uint_32 > 7) {
da812f61 2344 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
5f3633e3 2345 " (must be 0-7)");
da812f61 2346 return -EINVAL;
5f3633e3 2347 }
461c3af0
LC		 2348 ctx->journal_ioprio =
2349 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
6e47a3cc 2350 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
ba2e524d
LC		 2351 return 0;
2352 case Opt_test_dummy_encryption:
85456054 2353 return ext4_parse_test_dummy_encryption(param, ctx);
ba2e524d
LC		 2354 case Opt_dax:
2355 case Opt_dax_type:
ef83b6e8 2356#ifdef CONFIG_FS_DAX
ba2e524d
LC		 2357 {
2358 int type = (token == Opt_dax) ?
2359 Opt_dax : result.uint_32;
2360
2361 switch (type) {
9cb20f94
IW		 2362 case Opt_dax:
2363 case Opt_dax_always:
ba2e524d 2364 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
6e47a3cc 2365 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
9cb20f94
IW		 2366 break;
2367 case Opt_dax_never:
ba2e524d 2368 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
6e47a3cc 2369 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
9cb20f94
IW		 2370 break;
2371 case Opt_dax_inode:
6e47a3cc
LC		 2372 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2373 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
9cb20f94 2374 /* Strictly for printing options */
ba2e524d 2375 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
9cb20f94
IW		 2376 break;
2377 }
ba2e524d
LC		 2378 return 0;
2379 }
ef83b6e8 2380#else
da812f61 2381 ext4_msg(NULL, KERN_INFO, "dax option not supported");
da812f61 2382 return -EINVAL;
5f3633e3 2383#endif
ba2e524d
LC		 2384 case Opt_data_err:
2385 if (result.uint_32 == Opt_data_err_abort)
2386 ctx_set_mount_opt(ctx, m->mount_opt);
2387 else if (result.uint_32 == Opt_data_err_ignore)
2388 ctx_clear_mount_opt(ctx, m->mount_opt);
2389 return 0;
2390 case Opt_mb_optimize_scan:
27b38686
OM		 2391 if (result.int_32 == 1) {
2392 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2393 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2394 } else if (result.int_32 == 0) {
2395 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2396 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2397 } else {
da812f61 2398 ext4_msg(NULL, KERN_WARNING,
196e402a 2399 "mb_optimize_scan should be set to 0 or 1.");
da812f61 2400 return -EINVAL;
196e402a 2401 }
ba2e524d
LC		 2402 return 0;
2403 }
2404
2405 /*
2406 * At this point we should only be getting options requiring MOPT_SET,
2407 * or MOPT_CLEAR. Anything else is a bug
2408 */
2409 if (m->token == Opt_err) {
2410 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2411 param->key);
2412 WARN_ON(1);
2413 return -EINVAL;
2414 }
2415
2416 else {
461c3af0
LC		 2417 unsigned int set = 0;
2418
2419 if ((param->type == fs_value_is_flag) ||
2420 result.uint_32 > 0)
2421 set = 1;
2422
5f3633e3 2423 if (m->flags & MOPT_CLEAR)
461c3af0 2424 set = !set;
5f3633e3 2425 else if (unlikely(!(m->flags & MOPT_SET))) {
da812f61 2426 ext4_msg(NULL, KERN_WARNING,
461c3af0
LC		 2427 "buggy handling of option %s",
2428 param->key);
5f3633e3 2429 WARN_ON(1);
da812f61 2430 return -EINVAL;
5f3633e3 2431 }
995a3ed6 2432 if (m->flags & MOPT_2) {
461c3af0 2433 if (set != 0)
6e47a3cc 2434 ctx_set_mount_opt2(ctx, m->mount_opt);
995a3ed6 2435 else
6e47a3cc 2436 ctx_clear_mount_opt2(ctx, m->mount_opt);
995a3ed6 2437 } else {
461c3af0 2438 if (set != 0)
6e47a3cc 2439 ctx_set_mount_opt(ctx, m->mount_opt);
995a3ed6 2440 else
6e47a3cc 2441 ctx_clear_mount_opt(ctx, m->mount_opt);
995a3ed6 2442 }
26092bf5 2443 }
ba2e524d 2444
02f960f8 2445 return 0;
26092bf5
TT		 2446}
2447
7edfd85b 2448static int parse_options(struct fs_context *fc, char *options)
26092bf5 2449{
461c3af0 2450 struct fs_parameter param;
461c3af0
LC		 2451 int ret;
2452 char *key;
26092bf5
TT		 2453
2454 if (!options)
7edfd85b 2455 return 0;
461c3af0
LC		 2456
2457 while ((key = strsep(&options, ",")) != NULL) {
2458 if (*key) {
2459 size_t v_len = 0;
2460 char *value = strchr(key, '=');
2461
2462 param.type = fs_value_is_flag;
2463 param.string = NULL;
2464
2465 if (value) {
2466 if (value == key)
2467 continue;
2468
2469 *value++ = 0;
2470 v_len = strlen(value);
2471 param.string = kmemdup_nul(value, v_len,
2472 GFP_KERNEL);
2473 if (!param.string)
7edfd85b 2474 return -ENOMEM;
461c3af0
LC		 2475 param.type = fs_value_is_string;
2476 }
2477
2478 param.key = key;
2479 param.size = v_len;
2480
02f960f8 2481 ret = ext4_parse_param(fc, &param);
ea7d09ad 2482 kfree(param.string);
461c3af0 2483 if (ret < 0)
7edfd85b 2484 return ret;
461c3af0 2485 }
ac27a0ec 2486 }
461c3af0 2487
7edfd85b 2488 ret = ext4_validate_options(fc);
da812f61 2489 if (ret < 0)
7edfd85b 2490 return ret;
da812f61 2491
7edfd85b
LC		 2492 return 0;
2493}
2494
2495static int parse_apply_sb_mount_options(struct super_block *sb,
2496 struct ext4_fs_context *m_ctx)
2497{
2498 struct ext4_sb_info *sbi = EXT4_SB(sb);
2499 char *s_mount_opts = NULL;
2500 struct ext4_fs_context *s_ctx = NULL;
2501 struct fs_context *fc = NULL;
2502 int ret = -ENOMEM;
2503
2504 if (!sbi->s_es->s_mount_opts[0])
e6e268cb
LC		 2505 return 0;
2506
7edfd85b
LC		 2507 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2508 sizeof(sbi->s_es->s_mount_opts),
2509 GFP_KERNEL);
2510 if (!s_mount_opts)
2511 return ret;
2512
2513 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2514 if (!fc)
2515 goto out_free;
2516
2517 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2518 if (!s_ctx)
2519 goto out_free;
2520
2521 fc->fs_private = s_ctx;
2522 fc->s_fs_info = sbi;
2523
2524 ret = parse_options(fc, s_mount_opts);
6e47a3cc 2525 if (ret < 0)
7edfd85b 2526 goto parse_failed;
e6e268cb 2527
7edfd85b
LC		 2528 ret = ext4_check_opt_consistency(fc, sb);
2529 if (ret < 0) {
2530parse_failed:
2531 ext4_msg(sb, KERN_WARNING,
2532 "failed to parse options in superblock: %s",
2533 s_mount_opts);
2534 ret = 0;
2535 goto out_free;
2536 }
2537
2538 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2539 m_ctx->journal_devnum = s_ctx->journal_devnum;
2540 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2541 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2542
85456054
EB		 2543 ext4_apply_options(fc, sb);
2544 ret = 0;
7edfd85b
LC		 2545
2546out_free:
c069db76
EB		 2547 if (fc) {
2548 ext4_fc_free(fc);
2549 kfree(fc);
2550 }
7edfd85b
LC		 2551 kfree(s_mount_opts);
2552 return ret;
4c94bff9
LC		 2553}
2554
e6e268cb
LC		 2555static void ext4_apply_quota_options(struct fs_context *fc,
2556 struct super_block *sb)
2557{
2558#ifdef CONFIG_QUOTA
6e47a3cc 2559 bool quota_feature = ext4_has_feature_quota(sb);
e6e268cb
LC		 2560 struct ext4_fs_context *ctx = fc->fs_private;
2561 struct ext4_sb_info *sbi = EXT4_SB(sb);
2562 char *qname;
2563 int i;
2564
6e47a3cc
LC		 2565 if (quota_feature)
2566 return;
2567
2568 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2569 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2570 if (!(ctx->qname_spec & (1 << i)))
2571 continue;
2572
2573 qname = ctx->s_qf_names[i]; /* May be NULL */
4c1bd5a9
LC		 2574 if (qname)
2575 set_opt(sb, QUOTA);
6e47a3cc 2576 ctx->s_qf_names[i] = NULL;
13b215a9
LC		 2577 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2578 lockdep_is_held(&sb->s_umount));
2579 if (qname)
10e4f310 2580 kfree_rcu_mightsleep(qname);
6e47a3cc 2581 }
e6e268cb 2582 }
6e47a3cc
LC		 2583
2584 if (ctx->spec & EXT4_SPEC_JQFMT)
2585 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
e6e268cb
LC		 2586#endif
2587}
2588
2589/*
2590 * Check quota settings consistency.
2591 */
2592static int ext4_check_quota_consistency(struct fs_context *fc,
2593 struct super_block *sb)
2594{
2595#ifdef CONFIG_QUOTA
2596 struct ext4_fs_context *ctx = fc->fs_private;
2597 struct ext4_sb_info *sbi = EXT4_SB(sb);
2598 bool quota_feature = ext4_has_feature_quota(sb);
2599 bool quota_loaded = sb_any_quota_loaded(sb);
6e47a3cc
LC		 2600 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2601 int quota_flags, i;
2602
2603 /*
2604 * We do the test below only for project quotas. 'usrquota' and
2605 * 'grpquota' mount options are allowed even without quota feature
2606 * to support legacy quotas in quota files.
2607 */
2608 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2609 !ext4_has_feature_project(sb)) {
2610 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2611 "Cannot enable project quota enforcement.");
2612 return -EINVAL;
2613 }
e6e268cb 2614
6e47a3cc
LC		 2615 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2616 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2617 if (quota_loaded &&
2618 ctx->mask_s_mount_opt & quota_flags &&
2619 !ctx_test_mount_opt(ctx, quota_flags))
2620 goto err_quota_change;
2621
2622 if (ctx->spec & EXT4_SPEC_JQUOTA) {
e6e268cb
LC		 2623
2624 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2625 if (!(ctx->qname_spec & (1 << i)))
2626 continue;
2627
6e47a3cc
LC		 2628 if (quota_loaded &&
2629 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
e6e268cb
LC		 2630 goto err_jquota_change;
2631
2632 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
13b215a9 2633 strcmp(get_qf_name(sb, sbi, i),
e6e268cb
LC		 2634 ctx->s_qf_names[i]) != 0)
2635 goto err_jquota_specified;
2636 }
6e47a3cc
LC		 2637
2638 if (quota_feature) {
2639 ext4_msg(NULL, KERN_INFO,
2640 "Journaled quota options ignored when "
2641 "QUOTA feature is enabled");
2642 return 0;
2643 }
e6e268cb
LC		 2644 }
2645
6e47a3cc 2646 if (ctx->spec & EXT4_SPEC_JQFMT) {
e6e268cb 2647 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
6e47a3cc 2648 goto err_jquota_change;
e6e268cb
LC		 2649 if (quota_feature) {
2650 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2651 "ignored when QUOTA feature is enabled");
2652 return 0;
2653 }
2654 }
6e47a3cc
LC		 2655
2656 /* Make sure we don't mix old and new quota format */
2657 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2658 ctx->s_qf_names[USRQUOTA]);
2659 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2660 ctx->s_qf_names[GRPQUOTA]);
2661
2662 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2663 test_opt(sb, USRQUOTA));
2664
2665 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2666 test_opt(sb, GRPQUOTA));
2667
2668 if (usr_qf_name) {
2669 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2670 usrquota = false;
2671 }
2672 if (grp_qf_name) {
2673 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2674 grpquota = false;
2675 }
2676
2677 if (usr_qf_name || grp_qf_name) {
2678 if (usrquota || grpquota) {
2679 ext4_msg(NULL, KERN_ERR, "old and new quota "
2680 "format mixing");
2681 return -EINVAL;
2682 }
2683
2684 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2685 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2686 "not specified");
2687 return -EINVAL;
2688 }
2689 }
2690
e6e268cb
LC		 2691 return 0;
2692
2693err_quota_change:
2694 ext4_msg(NULL, KERN_ERR,
2695 "Cannot change quota options when quota turned on");
2696 return -EINVAL;
2697err_jquota_change:
2698 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2699 "options when quota turned on");
2700 return -EINVAL;
2701err_jquota_specified:
2702 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2703 QTYPE2NAME(i));
2704 return -EINVAL;
e6e268cb
LC		 2705#else
2706 return 0;
2707#endif
2708}
2709
5f41fdae
EB		 2710static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2711 struct super_block *sb)
2712{
5f41fdae
EB		 2713 const struct ext4_fs_context *ctx = fc->fs_private;
2714 const struct ext4_sb_info *sbi = EXT4_SB(sb);
2715
85456054 2716 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
5f41fdae
EB		 2717 return 0;
2718
2719 if (!ext4_has_feature_encrypt(sb)) {
2720 ext4_msg(NULL, KERN_WARNING,
2721 "test_dummy_encryption requires encrypt feature");
2722 return -EINVAL;
2723 }
2724 /*
2725 * This mount option is just for testing, and it's not worthwhile to
2726 * implement the extra complexity (e.g. RCU protection) that would be
2727 * needed to allow it to be set or changed during remount. We do allow
2728 * it to be specified during remount, but only if there is no change.
2729 */
85456054
EB		 2730 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2731 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2732 &ctx->dummy_enc_policy))
2733 return 0;
5f41fdae 2734 ext4_msg(NULL, KERN_WARNING,
85456054 2735 "Can't set or change test_dummy_encryption on remount");
5f41fdae
EB		 2736 return -EINVAL;
2737 }
85456054
EB		 2738 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2739 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2740 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2741 &ctx->dummy_enc_policy))
2742 return 0;
2743 ext4_msg(NULL, KERN_WARNING,
2744 "Conflicting test_dummy_encryption options");
2745 return -EINVAL;
2746 }
7959eb19 2747 return 0;
85456054
EB		 2748}
2749
2750static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2751 struct super_block *sb)
2752{
2753 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2754 /* if already set, it was already verified to be the same */
2755 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2756 return;
2757 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2758 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2759 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
5f41fdae
EB		 2760}
2761
b6bd2435
LC		 2762static int ext4_check_opt_consistency(struct fs_context *fc,
2763 struct super_block *sb)
2764{
2765 struct ext4_fs_context *ctx = fc->fs_private;
6e47a3cc
LC		 2766 struct ext4_sb_info *sbi = fc->s_fs_info;
2767 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
5f41fdae 2768 int err;
b6bd2435
LC		 2769
2770 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2771 ext4_msg(NULL, KERN_ERR,
2772 "Mount option(s) incompatible with ext2");
2773 return -EINVAL;
2774 }
2775 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2776 ext4_msg(NULL, KERN_ERR,
2777 "Mount option(s) incompatible with ext3");
2778 return -EINVAL;
2779 }
2780
6e47a3cc
LC		 2781 if (ctx->s_want_extra_isize >
2782 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2783 ext4_msg(NULL, KERN_ERR,
2784 "Invalid want_extra_isize %d",
2785 ctx->s_want_extra_isize);
2786 return -EINVAL;
2787 }
2788
5f41fdae
EB		 2789 err = ext4_check_test_dummy_encryption(fc, sb);
2790 if (err)
2791 return err;
6e47a3cc
LC		 2792
2793 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2794 if (!sbi->s_journal) {
2795 ext4_msg(NULL, KERN_WARNING,
2796 "Remounting file system with no journal "
2797 "so ignoring journalled data option");
2798 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
4c246728
LC		 2799 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2800 test_opt(sb, DATA_FLAGS)) {
6e47a3cc
LC		 2801 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2802 "on remount");
2803 return -EINVAL;
2804 }
2805 }
2806
2807 if (is_remount) {
2808 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2809 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2810 ext4_msg(NULL, KERN_ERR, "can't mount with "
2811 "both data=journal and dax");
2812 return -EINVAL;
2813 }
2814
2815 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2816 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2817 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2818fail_dax_change_remount:
2819 ext4_msg(NULL, KERN_ERR, "can't change "
2820 "dax mount option while remounting");
2821 return -EINVAL;
2822 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2823 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2824 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2825 goto fail_dax_change_remount;
2826 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2827 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2828 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2829 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2830 goto fail_dax_change_remount;
2831 }
2832 }
2833
b6bd2435
LC		 2834 return ext4_check_quota_consistency(fc, sb);
2835}
2836
85456054 2837static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
4c94bff9 2838{
6e47a3cc 2839 struct ext4_fs_context *ctx = fc->fs_private;
da812f61 2840 struct ext4_sb_info *sbi = fc->s_fs_info;
6e47a3cc
LC		 2841
2842 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2843 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2844 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2845 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
6e47a3cc
LC		 2846 sb->s_flags &= ~ctx->mask_s_flags;
2847 sb->s_flags |= ctx->vals_s_flags;
2848
2849#define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2850 APPLY(s_commit_interval);
2851 APPLY(s_stripe);
2852 APPLY(s_max_batch_time);
2853 APPLY(s_min_batch_time);
2854 APPLY(s_want_extra_isize);
2855 APPLY(s_inode_readahead_blks);
2856 APPLY(s_max_dir_size_kb);
2857 APPLY(s_li_wait_mult);
2858 APPLY(s_resgid);
2859 APPLY(s_resuid);
2860
2861#ifdef CONFIG_EXT4_DEBUG
2862 APPLY(s_fc_debug_max_replay);
2863#endif
2864
2865 ext4_apply_quota_options(fc, sb);
85456054 2866 ext4_apply_test_dummy_encryption(ctx, sb);
6e47a3cc
LC		 2867}
2868
2869
2870static int ext4_validate_options(struct fs_context *fc)
2871{
ac27a0ec 2872#ifdef CONFIG_QUOTA
6e47a3cc 2873 struct ext4_fs_context *ctx = fc->fs_private;
4c94bff9 2874 char *usr_qf_name, *grp_qf_name;
6e47a3cc
LC		 2875
2876 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2877 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2878
33458eab 2879 if (usr_qf_name || grp_qf_name) {
6e47a3cc
LC		 2880 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2881 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
ac27a0ec 2882
6e47a3cc
LC		 2883 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2884 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
ac27a0ec 2885
6e47a3cc
LC		 2886 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2887 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
da812f61 2888 ext4_msg(NULL, KERN_ERR, "old and new quota "
6e47a3cc 2889 "format mixing");
da812f61 2890 return -EINVAL;
ac27a0ec 2891 }
ac27a0ec
DK		 2892 }
2893#endif
6e47a3cc 2894 return 1;
ac27a0ec
DK		 2895}
2896
2adf6da8
TT		 2897static inline void ext4_show_quota_options(struct seq_file *seq,
2898 struct super_block *sb)
2899{
2900#if defined(CONFIG_QUOTA)
2901 struct ext4_sb_info *sbi = EXT4_SB(sb);
33458eab 2902 char *usr_qf_name, *grp_qf_name;
2adf6da8
TT		 2903
2904 if (sbi->s_jquota_fmt) {
2905 char *fmtname = "";
2906
2907 switch (sbi->s_jquota_fmt) {
2908 case QFMT_VFS_OLD:
2909 fmtname = "vfsold";
2910 break;
2911 case QFMT_VFS_V0:
2912 fmtname = "vfsv0";
2913 break;
2914 case QFMT_VFS_V1:
2915 fmtname = "vfsv1";
2916 break;
2917 }
2918 seq_printf(seq, ",jqfmt=%s", fmtname);
2919 }
2920
33458eab
TT		 2921 rcu_read_lock();
2922 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2923 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2924 if (usr_qf_name)
2925 seq_show_option(seq, "usrjquota", usr_qf_name);
2926 if (grp_qf_name)
2927 seq_show_option(seq, "grpjquota", grp_qf_name);
2928 rcu_read_unlock();
2adf6da8
TT		 2929#endif
2930}
2931
5a916be1
TT		 2932static const char *token2str(int token)
2933{
97d8a670 2934 const struct fs_parameter_spec *spec;
5a916be1 2935
97d8a670
LC		 2936 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2937 if (spec->opt == token && !spec->type)
5a916be1 2938 break;
97d8a670 2939 return spec->name;
5a916be1
TT		 2940}
2941
2adf6da8
TT		 2942/*
2943 * Show an option if
2944 * - it's set to a non-default value OR
2945 * - if the per-sb default is different from the global default
2946 */
66acdcf4
TT		 2947static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2948 int nodefs)
2adf6da8 2949{
2adf6da8
TT		 2950 struct ext4_sb_info *sbi = EXT4_SB(sb);
2951 struct ext4_super_block *es = sbi->s_es;
e3645d72 2952 int def_errors;
5a916be1 2953 const struct mount_opts *m;
66acdcf4 2954 char sep = nodefs ? '\n' : ',';
2adf6da8 2955
66acdcf4
TT		 2956#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2957#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 2958
2959 if (sbi->s_sb_block != 1)
5a916be1
TT		 2960 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2961
2962 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2963 int want_set = m->flags & MOPT_SET;
e3645d72
ZY		 2964 int opt_2 = m->flags & MOPT_2;
2965 unsigned int mount_opt, def_mount_opt;
2966
5a916be1 2967 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
ba2e524d 2968 m->flags & MOPT_SKIP)
5a916be1 2969 continue;
e3645d72
ZY		 2970
2971 if (opt_2) {
2972 mount_opt = sbi->s_mount_opt2;
2973 def_mount_opt = sbi->s_def_mount_opt2;
2974 } else {
2975 mount_opt = sbi->s_mount_opt;
2976 def_mount_opt = sbi->s_def_mount_opt;
2977 }
2978 /* skip if same as the default */
2979 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt)))
2980 continue;
2981 /* select Opt_noFoo vs Opt_Foo */
5a916be1 2982 if ((want_set &&
e3645d72
ZY		 2983 (mount_opt & m->mount_opt) != m->mount_opt) ||
2984 (!want_set && (mount_opt & m->mount_opt)))
2985 continue;
5a916be1 2986 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 2987 }
5a916be1 2988
08cefc7a 2989 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 2990 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 2991 SEQ_OPTS_PRINT("resuid=%u",
2992 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2993 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 2994 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 2995 SEQ_OPTS_PRINT("resgid=%u",
2996 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 2997 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 2998 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2999 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 3000 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 3001 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 3002 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 3003 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 3004 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 3005 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 3006 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 3007 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 3008 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 3009 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
66acdcf4 3010 if (nodefs || sbi->s_stripe)
5a916be1 3011 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
68afa7e0 3012 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
e3645d72 3013 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5a916be1
TT		 3014 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3015 SEQ_OPTS_PUTS("data=journal");
3016 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3017 SEQ_OPTS_PUTS("data=ordered");
3018 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
3019 SEQ_OPTS_PUTS("data=writeback");
3020 }
66acdcf4
TT		 3021 if (nodefs ||
3022 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 3023 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
3024 sbi->s_inode_readahead_blks);
2adf6da8 3025
ceec0376 3026 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
66acdcf4 3027 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 3028 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 3029 if (nodefs || sbi->s_max_dir_size_kb)
3030 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
7915a861
AN		 3031 if (test_opt(sb, DATA_ERR_ABORT))
3032 SEQ_OPTS_PUTS("data_err=abort");
ed318a6c
EB		 3033
3034 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2adf6da8 3035
4f74d15f
EB		 3036 if (sb->s_flags & SB_INLINECRYPT)
3037 SEQ_OPTS_PUTS("inlinecrypt");
3038
9cb20f94
IW		 3039 if (test_opt(sb, DAX_ALWAYS)) {
3040 if (IS_EXT2_SB(sb))
3041 SEQ_OPTS_PUTS("dax");
3042 else
3043 SEQ_OPTS_PUTS("dax=always");
3044 } else if (test_opt2(sb, DAX_NEVER)) {
3045 SEQ_OPTS_PUTS("dax=never");
3046 } else if (test_opt2(sb, DAX_INODE)) {
3047 SEQ_OPTS_PUTS("dax=inode");
3048 }
3fa5d23e
OM		 3049
3050 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3051 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3052 SEQ_OPTS_PUTS("mb_optimize_scan=0");
3053 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3054 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3055 SEQ_OPTS_PUTS("mb_optimize_scan=1");
3056 }
3057
2adf6da8 3058 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 3059 return 0;
3060}
3061
66acdcf4
TT		 3062static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3063{
3064 return _ext4_show_options(seq, root->d_sb, 0);
3065}
3066
ebd173be 3067int ext4_seq_options_show(struct seq_file *seq, void *offset)
66acdcf4
TT		 3068{
3069 struct super_block *sb = seq->private;
3070 int rc;
3071
bc98a42c 3072 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
66acdcf4
TT		 3073 rc = _ext4_show_options(seq, sb, 1);
3074 seq_puts(seq, "\n");
3075 return rc;
3076}
3077
617ba13b 3078static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 3079 int read_only)
3080{
617ba13b 3081 struct ext4_sb_info *sbi = EXT4_SB(sb);
c89128a0 3082 int err = 0;
ac27a0ec 3083
617ba13b 3084 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 3085 ext4_msg(sb, KERN_ERR, "revision level too high, "
3086 "forcing read-only mode");
c89128a0 3087 err = -EROFS;
5adaccac 3088 goto done;
ac27a0ec
DK		 3089 }
3090 if (read_only)
281b5995 3091 goto done;
617ba13b 3092 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 3093 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3094 "running e2fsck is recommended");
c8b459f4 3095 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 3096 ext4_msg(sb, KERN_WARNING,
3097 "warning: mounting fs with errors, "
3098 "running e2fsck is recommended");
ed3ce80a 3099 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 3100 le16_to_cpu(es->s_mnt_count) >=
3101 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 3102 ext4_msg(sb, KERN_WARNING,
3103 "warning: maximal mount count reached, "
3104 "running e2fsck is recommended");
ac27a0ec 3105 else if (le32_to_cpu(es->s_checkinterval) &&
6a0678a7
AB		 3106 (ext4_get_tstamp(es, s_lastcheck) +
3107 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
b31e1552
ES		 3108 ext4_msg(sb, KERN_WARNING,
3109 "warning: checktime reached, "
3110 "running e2fsck is recommended");
0b8e58a1 3111 if (!sbi->s_journal)
0390131b 3112 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 3113 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 3114 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 3115 le16_add_cpu(&es->s_mnt_count, 1);
6a0678a7 3116 ext4_update_tstamp(es, s_mtime);
02f310fc 3117 if (sbi->s_journal) {
e2b911c5 3118 ext4_set_feature_journal_needs_recovery(sb);
02f310fc
JK		 3119 if (ext4_has_feature_orphan_file(sb))
3120 ext4_set_feature_orphan_present(sb);
3121 }
ac27a0ec 3122
4392fbc4 3123 err = ext4_commit_super(sb);
281b5995 3124done:
ac27a0ec 3125 if (test_opt(sb, DEBUG))
a9df9a49 3126 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 3127 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 3128 sb->s_blocksize,
3129 sbi->s_groups_count,
617ba13b
MC		 3130 EXT4_BLOCKS_PER_GROUP(sb),
3131 EXT4_INODES_PER_GROUP(sb),
a2595b8a 3132 sbi->s_mount_opt, sbi->s_mount_opt2);
c89128a0 3133 return err;
ac27a0ec
DK		 3134}
3135
117fff10
TT		 3136int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3137{
3138 struct ext4_sb_info *sbi = EXT4_SB(sb);
7c990728 3139 struct flex_groups **old_groups, **new_groups;
37b0b6b8 3140 int size, i, j;
117fff10
TT		 3141
3142 if (!sbi->s_log_groups_per_flex)
3143 return 0;
3144
3145 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3146 if (size <= sbi->s_flex_groups_allocated)
3147 return 0;
3148
7c990728
SJS		 3149 new_groups = kvzalloc(roundup_pow_of_two(size *
3150 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
117fff10 3151 if (!new_groups) {
7c990728
SJS		 3152 ext4_msg(sb, KERN_ERR,
3153 "not enough memory for %d flex group pointers", size);
117fff10
TT		 3154 return -ENOMEM;
3155 }
7c990728
SJS		 3156 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3157 new_groups[i] = kvzalloc(roundup_pow_of_two(
3158 sizeof(struct flex_groups)),
3159 GFP_KERNEL);
3160 if (!new_groups[i]) {
37b0b6b8
DC		 3161 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3162 kvfree(new_groups[j]);
7c990728
SJS		 3163 kvfree(new_groups);
3164 ext4_msg(sb, KERN_ERR,
3165 "not enough memory for %d flex groups", size);
3166 return -ENOMEM;
3167 }
117fff10 3168 }
7c990728
SJS		 3169 rcu_read_lock();
3170 old_groups = rcu_dereference(sbi->s_flex_groups);
3171 if (old_groups)
3172 memcpy(new_groups, old_groups,
3173 (sbi->s_flex_groups_allocated *
3174 sizeof(struct flex_groups *)));
3175 rcu_read_unlock();
3176 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3177 sbi->s_flex_groups_allocated = size;
3178 if (old_groups)
3179 ext4_kvfree_array_rcu(old_groups);
117fff10
TT		 3180 return 0;
3181}
3182
772cb7c8
JS		 3183static int ext4_fill_flex_info(struct super_block *sb)
3184{
3185 struct ext4_sb_info *sbi = EXT4_SB(sb);
3186 struct ext4_group_desc *gdp = NULL;
7c990728 3187 struct flex_groups *fg;
772cb7c8 3188 ext4_group_t flex_group;
117fff10 3189 int i, err;
772cb7c8 3190
503358ae 3191 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 3192 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 3193 sbi->s_log_groups_per_flex = 0;
3194 return 1;
3195 }
3196
117fff10
TT		 3197 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3198 if (err)
9933fc0a 3199 goto failed;
772cb7c8 3200
772cb7c8 3201 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 3202 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 3203
3204 flex_group = ext4_flex_group(sbi, i);
7c990728
SJS		 3205 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3206 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
90ba983f 3207 atomic64_add(ext4_free_group_clusters(sb, gdp),
7c990728
SJS		 3208 &fg->free_clusters);
3209 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
772cb7c8
JS		 3210 }
3211
3212 return 1;
3213failed:
3214 return 0;
3215}
3216
e2b911c5 3217static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
feb0ab32 3218 struct ext4_group_desc *gdp)
717d50e4 3219{
b47820ed 3220 int offset = offsetof(struct ext4_group_desc, bg_checksum);
717d50e4 3221 __u16 crc = 0;
feb0ab32 3222 __le32 le_group = cpu_to_le32(block_group);
e2b911c5 3223 struct ext4_sb_info *sbi = EXT4_SB(sb);
717d50e4 3224
9aa5d32b 3225 if (ext4_has_metadata_csum(sbi->s_sb)) {
feb0ab32 3226 /* Use new metadata_csum algorithm */
feb0ab32 3227 __u32 csum32;
b47820ed 3228 __u16 dummy_csum = 0;
feb0ab32 3229
feb0ab32
DW		 3230 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3231 sizeof(le_group));
b47820ed
DJ		 3232 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3233 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3234 sizeof(dummy_csum));
3235 offset += sizeof(dummy_csum);
3236 if (offset < sbi->s_desc_size)
3237 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3238 sbi->s_desc_size - offset);
feb0ab32
DW		 3239
3240 crc = csum32 & 0xFFFF;
3241 goto out;
717d50e4
AD		 3242 }
3243
feb0ab32 3244 /* old crc16 code */
e2b911c5 3245 if (!ext4_has_feature_gdt_csum(sb))
813d32f9
DW		 3246 return 0;
3247
feb0ab32
DW		 3248 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3249 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3250 crc = crc16(crc, (__u8 *)gdp, offset);
3251 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3252 /* for checksum of struct ext4_group_desc do the rest...*/
4f043518 3253 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size)
feb0ab32 3254 crc = crc16(crc, (__u8 *)gdp + offset,
4f043518 3255 sbi->s_desc_size - offset);
feb0ab32
DW		 3256
3257out:
717d50e4
AD		 3258 return cpu_to_le16(crc);
3259}
3260
feb0ab32 3261int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 3262 struct ext4_group_desc *gdp)
3263{
feb0ab32 3264 if (ext4_has_group_desc_csum(sb) &&
e2b911c5 3265 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
717d50e4
AD		 3266 return 0;
3267
3268 return 1;
3269}
3270
feb0ab32
DW		 3271void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3272 struct ext4_group_desc *gdp)
3273{
3274 if (!ext4_has_group_desc_csum(sb))
3275 return;
e2b911c5 3276 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
feb0ab32
DW		 3277}
3278
ac27a0ec 3279/* Called at mount-time, super-block is locked */
bfff6873 3280static int ext4_check_descriptors(struct super_block *sb,
829fa70d 3281 ext4_fsblk_t sb_block,
bfff6873 3282 ext4_group_t *first_not_zeroed)
ac27a0ec 3283{
617ba13b
MC		 3284 struct ext4_sb_info *sbi = EXT4_SB(sb);
3285 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3286 ext4_fsblk_t last_block;
44de022c 3287 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
bd81d8ee
LV		 3288 ext4_fsblk_t block_bitmap;
3289 ext4_fsblk_t inode_bitmap;
3290 ext4_fsblk_t inode_table;
ce421581 3291 int flexbg_flag = 0;
bfff6873 3292 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 3293
e2b911c5 3294 if (ext4_has_feature_flex_bg(sb))
ce421581
JS		 3295 flexbg_flag = 1;
3296
af5bc92d 3297 ext4_debug("Checking group descriptors");
ac27a0ec 3298
197cd65a
AM		 3299 for (i = 0; i < sbi->s_groups_count; i++) {
3300 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3301
ce421581 3302 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 3303 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 3304 else
3305 last_block = first_block +
617ba13b 3306 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 3307
bfff6873
LC		 3308 if ((grp == sbi->s_groups_count) &&
3309 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3310 grp = i;
3311
8fadc143 3312 block_bitmap = ext4_block_bitmap(sb, gdp);
829fa70d
TT		 3313 if (block_bitmap == sb_block) {
3314 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3315 "Block bitmap for group %u overlaps "
3316 "superblock", i);
18db4b4e
TT		 3317 if (!sb_rdonly(sb))
3318 return 0;
829fa70d 3319 }
77260807
TT		 3320 if (block_bitmap >= sb_block + 1 &&
3321 block_bitmap <= last_bg_block) {
3322 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3323 "Block bitmap for group %u overlaps "
3324 "block group descriptors", i);
3325 if (!sb_rdonly(sb))
3326 return 0;
3327 }
2b2d6d01 3328 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 3329 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 3330 "Block bitmap for group %u not in group "
b31e1552 3331 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 3332 return 0;
3333 }
8fadc143 3334 inode_bitmap = ext4_inode_bitmap(sb, gdp);
829fa70d
TT		 3335 if (inode_bitmap == sb_block) {
3336 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3337 "Inode bitmap for group %u overlaps "
3338 "superblock", i);
18db4b4e
TT		 3339 if (!sb_rdonly(sb))
3340 return 0;
829fa70d 3341 }
77260807
TT		 3342 if (inode_bitmap >= sb_block + 1 &&
3343 inode_bitmap <= last_bg_block) {
3344 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3345 "Inode bitmap for group %u overlaps "
3346 "block group descriptors", i);
3347 if (!sb_rdonly(sb))
3348 return 0;
3349 }
2b2d6d01 3350 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 3351 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 3352 "Inode bitmap for group %u not in group "
b31e1552 3353 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 3354 return 0;
3355 }
8fadc143 3356 inode_table = ext4_inode_table(sb, gdp);
829fa70d
TT		 3357 if (inode_table == sb_block) {
3358 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3359 "Inode table for group %u overlaps "
3360 "superblock", i);
18db4b4e
TT		 3361 if (!sb_rdonly(sb))
3362 return 0;
829fa70d 3363 }
77260807
TT		 3364 if (inode_table >= sb_block + 1 &&
3365 inode_table <= last_bg_block) {
3366 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3367 "Inode table for group %u overlaps "
3368 "block group descriptors", i);
3369 if (!sb_rdonly(sb))
3370 return 0;
3371 }
bd81d8ee 3372 if (inode_table < first_block ||
2b2d6d01 3373 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 3374 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 3375 "Inode table for group %u not in group "
b31e1552 3376 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 3377 return 0;
3378 }
955ce5f5 3379 ext4_lock_group(sb, i);
feb0ab32 3380 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 3381 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3382 "Checksum for group %u failed (%u!=%u)",
e2b911c5 3383 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
b31e1552 3384 gdp)), le16_to_cpu(gdp->bg_checksum));
bc98a42c 3385 if (!sb_rdonly(sb)) {
955ce5f5 3386 ext4_unlock_group(sb, i);
8a266467 3387 return 0;
7ee1ec4c 3388 }
717d50e4 3389 }
955ce5f5 3390 ext4_unlock_group(sb, i);
ce421581
JS		 3391 if (!flexbg_flag)
3392 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 3393 }
bfff6873
LC		 3394 if (NULL != first_not_zeroed)
3395 *first_not_zeroed = grp;
ac27a0ec
DK		 3396 return 1;
3397}
3398
cd2291a4
ES		 3399/*
3400 * Maximal extent format file size.
3401 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3402 * extent format containers, within a sector_t, and within i_blocks
3403 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3404 * so that won't be a limiting factor.
3405 *
f17722f9
LC		 3406 * However there is other limiting factor. We do store extents in the form
3407 * of starting block and length, hence the resulting length of the extent
3408 * covering maximum file size must fit into on-disk format containers as
3409 * well. Given that length is always by 1 unit bigger than max unit (because
3410 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3411 *
cd2291a4
ES		 3412 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3413 */
f287a1a5 3414static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 3415{
3416 loff_t res;
3417 loff_t upper_limit = MAX_LFS_FILESIZE;
3418
72deb455
CH		 3419 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3420
3421 if (!has_huge_files) {
cd2291a4
ES		 3422 upper_limit = (1LL << 32) - 1;
3423
3424 /* total blocks in file system block size */
3425 upper_limit >>= (blkbits - 9);
3426 upper_limit <<= blkbits;
3427 }
3428
f17722f9
LC		 3429 /*
3430 * 32-bit extent-start container, ee_block. We lower the maxbytes
3431 * by one fs block, so ee_len can cover the extent of maximum file
3432 * size
3433 */
3434 res = (1LL << 32) - 1;
cd2291a4 3435 res <<= blkbits;
cd2291a4
ES		 3436
3437 /* Sanity check against vm- & vfs- imposed limits */
3438 if (res > upper_limit)
3439 res = upper_limit;
3440
3441 return res;
3442}
ac27a0ec 3443
ac27a0ec 3444/*
cd2291a4 3445 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 3446 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3447 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 3448 */
f287a1a5 3449static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 3450{
5c93e8ec 3451 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
0fc1b451 3452 int meta_blocks;
5c93e8ec 3453 unsigned int ppb = 1 << (bits - 2);
75ca6ad4
RH		 3454
3455 /*
3456 * This is calculated to be the largest file size for a dense, block
0b8e58a1
AD		 3457 * mapped file such that the file's total number of 512-byte sectors,
3458 * including data and all indirect blocks, does not exceed (2^48 - 1).
3459 *
3460 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3461 * number of 512-byte sectors of the file.
0fc1b451 3462 */
72deb455 3463 if (!has_huge_files) {
0fc1b451 3464 /*
72deb455
CH		 3465 * !has_huge_files or implies that the inode i_block field
3466 * represents total file blocks in 2^32 512-byte sectors ==
3467 * size of vfs inode i_blocks * 8
0fc1b451
AK		 3468 */
3469 upper_limit = (1LL << 32) - 1;
3470
3471 /* total blocks in file system block size */
3472 upper_limit >>= (bits - 9);
3473
3474 } else {
8180a562
AK		 3475 /*
3476 * We use 48 bit ext4_inode i_blocks
3477 * With EXT4_HUGE_FILE_FL set the i_blocks
3478 * represent total number of blocks in
3479 * file system block size
3480 */
0fc1b451
AK		 3481 upper_limit = (1LL << 48) - 1;
3482
0fc1b451
AK		 3483 }
3484
5c93e8ec
ZY		 3485 /* Compute how many blocks we can address by block tree */
3486 res += ppb;
3487 res += ppb * ppb;
3488 res += ((loff_t)ppb) * ppb * ppb;
3489 /* Compute how many metadata blocks are needed */
3490 meta_blocks = 1;
3491 meta_blocks += 1 + ppb;
3492 meta_blocks += 1 + ppb + ppb * ppb;
3493 /* Does block tree limit file size? */
3494 if (res + meta_blocks <= upper_limit)
3495 goto check_lfs;
3496
3497 res = upper_limit;
3498 /* How many metadata blocks are needed for addressing upper_limit? */
3499 upper_limit -= EXT4_NDIR_BLOCKS;
0fc1b451
AK		 3500 /* indirect blocks */
3501 meta_blocks = 1;
5c93e8ec 3502 upper_limit -= ppb;
0fc1b451 3503 /* double indirect blocks */
5c93e8ec
ZY		 3504 if (upper_limit < ppb * ppb) {
3505 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3506 res -= meta_blocks;
3507 goto check_lfs;
3508 }
3509 meta_blocks += 1 + ppb;
3510 upper_limit -= ppb * ppb;
3511 /* tripple indirect blocks for the rest */
3512 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3513 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3514 res -= meta_blocks;
3515check_lfs:
ac27a0ec 3516 res <<= bits;
0fc1b451
AK		 3517 if (res > MAX_LFS_FILESIZE)
3518 res = MAX_LFS_FILESIZE;
3519
5c93e8ec 3520 return res;
ac27a0ec
DK		 3521}
3522
617ba13b 3523static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 3524 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 3525{
617ba13b 3526 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 3527 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 3528 int has_super = 0;
3529
3530 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3531
e2b911c5 3532 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
70bbb3e0 3533 return logical_sb_block + nr + 1;
ac27a0ec 3534 bg = sbi->s_desc_per_block * nr;
617ba13b 3535 if (ext4_bg_has_super(sb, bg))
ac27a0ec 3536 has_super = 1;
0b8e58a1 3537
bd63f6b0
DW		 3538 /*
3539 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3540 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3541 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3542 * compensate.
3543 */
3544 if (sb->s_blocksize == 1024 && nr == 0 &&
49598e04 3545 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
bd63f6b0
DW		 3546 has_super++;
3547
617ba13b 3548 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 3549}
3550
c9de560d
AT		 3551/**
3552 * ext4_get_stripe_size: Get the stripe size.
3553 * @sbi: In memory super block info
3554 *
3555 * If we have specified it via mount option, then
3556 * use the mount option value. If the value specified at mount time is
3557 * greater than the blocks per group use the super block value.
3558 * If the super block value is greater than blocks per group return 0.
3559 * Allocator needs it be less than blocks per group.
3560 *
3561 */
3562static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3563{
3564 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3565 unsigned long stripe_width =
3566 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 3567 int ret;
c9de560d
AT		 3568
3569 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658 3570 ret = sbi->s_stripe;
5469d7c3 3571 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3eb08658 3572 ret = stripe_width;
5469d7c3 3573 else if (stride && stride <= sbi->s_blocks_per_group)
3eb08658
DE		 3574 ret = stride;
3575 else
3576 ret = 0;
c9de560d 3577
3eb08658
DE		 3578 /*
3579 * If the stripe width is 1, this makes no sense and
3580 * we set it to 0 to turn off stripe handling code.
3581 */
3582 if (ret <= 1)
3583 ret = 0;
c9de560d 3584
3eb08658 3585 return ret;
c9de560d 3586}
ac27a0ec 3587
a13fb1a4
ES		 3588/*
3589 * Check whether this filesystem can be mounted based on
3590 * the features present and the RDONLY/RDWR mount requested.
3591 * Returns 1 if this filesystem can be mounted as requested,
3592 * 0 if it cannot be.
3593 */
25c6d98f 3594int ext4_feature_set_ok(struct super_block *sb, int readonly)
a13fb1a4 3595{
e2b911c5 3596 if (ext4_has_unknown_ext4_incompat_features(sb)) {
a13fb1a4
ES		 3597 ext4_msg(sb, KERN_ERR,
3598 "Couldn't mount because of "
3599 "unsupported optional features (%x)",
3600 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3601 ~EXT4_FEATURE_INCOMPAT_SUPP));
3602 return 0;
3603 }
3604
5298d4bf 3605#if !IS_ENABLED(CONFIG_UNICODE)
c83ad55e
GKB		 3606 if (ext4_has_feature_casefold(sb)) {
3607 ext4_msg(sb, KERN_ERR,
3608 "Filesystem with casefold feature cannot be "
3609 "mounted without CONFIG_UNICODE");
3610 return 0;
3611 }
3612#endif
3613
a13fb1a4
ES		 3614 if (readonly)
3615 return 1;
3616
e2b911c5 3617 if (ext4_has_feature_readonly(sb)) {
2cb5cc8b 3618 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
1751e8a6 3619 sb->s_flags |= SB_RDONLY;
2cb5cc8b
DW		 3620 return 1;
3621 }
3622
a13fb1a4 3623 /* Check that feature set is OK for a read-write mount */
e2b911c5 3624 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
a13fb1a4
ES		 3625 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3626 "unsupported optional features (%x)",
3627 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3628 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3629 return 0;
3630 }
e2b911c5 3631 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
bab08ab9
TT		 3632 ext4_msg(sb, KERN_ERR,
3633 "Can't support bigalloc feature without "
3634 "extents feature\n");
3635 return 0;
3636 }
7c319d32 3637
9db176bc 3638#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
d65d87a0
TT		 3639 if (!readonly && (ext4_has_feature_quota(sb) ||
3640 ext4_has_feature_project(sb))) {
7c319d32 3641 ext4_msg(sb, KERN_ERR,
d65d87a0 3642 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
689c958c
LX		 3643 return 0;
3644 }
7c319d32 3645#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 3646 return 1;
3647}
3648
66e61a9e
TT		 3649/*
3650 * This function is called once a day if we have errors logged
3651 * on the file system
3652 */
235699a8 3653static void print_daily_error_info(struct timer_list *t)
66e61a9e 3654{
235699a8
KC		 3655 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3656 struct super_block *sb = sbi->s_sb;
3657 struct ext4_super_block *es = sbi->s_es;
66e61a9e
TT		 3658
3659 if (es->s_error_count)
ae0f78de
TT		 3660 /* fsck newer than v1.41.13 is needed to clean this condition. */
3661 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 3662 le32_to_cpu(es->s_error_count));
3663 if (es->s_first_error_time) {
6a0678a7
AB		 3664 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3665 sb->s_id,
3666 ext4_get_tstamp(es, s_first_error_time),
66e61a9e
TT		 3667 (int) sizeof(es->s_first_error_func),
3668 es->s_first_error_func,
3669 le32_to_cpu(es->s_first_error_line));
3670 if (es->s_first_error_ino)
651e1c3b 3671 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3672 le32_to_cpu(es->s_first_error_ino));
3673 if (es->s_first_error_block)
651e1c3b 3674 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3675 le64_to_cpu(es->s_first_error_block));
651e1c3b 3676 printk(KERN_CONT "\n");
66e61a9e
TT		 3677 }
3678 if (es->s_last_error_time) {
6a0678a7
AB		 3679 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3680 sb->s_id,
3681 ext4_get_tstamp(es, s_last_error_time),
66e61a9e
TT		 3682 (int) sizeof(es->s_last_error_func),
3683 es->s_last_error_func,
3684 le32_to_cpu(es->s_last_error_line));
3685 if (es->s_last_error_ino)
651e1c3b 3686 printk(KERN_CONT ": inode %u",
66e61a9e
TT		 3687 le32_to_cpu(es->s_last_error_ino));
3688 if (es->s_last_error_block)
651e1c3b 3689 printk(KERN_CONT ": block %llu", (unsigned long long)
66e61a9e 3690 le64_to_cpu(es->s_last_error_block));
651e1c3b 3691 printk(KERN_CONT "\n");
66e61a9e
TT		 3692 }
3693 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3694}
3695
bfff6873
LC		 3696/* Find next suitable group and run ext4_init_inode_table */
3697static int ext4_run_li_request(struct ext4_li_request *elr)
3698{
3699 struct ext4_group_desc *gdp = NULL;
3d392b26
TT		 3700 struct super_block *sb = elr->lr_super;
3701 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3702 ext4_group_t group = elr->lr_next_group;
3d392b26 3703 unsigned int prefetch_ios = 0;
bfff6873 3704 int ret = 0;
4f3d1e45 3705 int nr = EXT4_SB(sb)->s_mb_prefetch;
39fec688 3706 u64 start_time;
bfff6873 3707
3d392b26 3708 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
4f3d1e45
OM		 3709 elr->lr_next_group = ext4_mb_prefetch(sb, group, nr, &prefetch_ios);
3710 ext4_mb_prefetch_fini(sb, elr->lr_next_group, nr);
3711 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, nr);
3d392b26
TT		 3712 if (group >= elr->lr_next_group) {
3713 ret = 1;
3714 if (elr->lr_first_not_zeroed != ngroups &&
3715 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3716 elr->lr_next_group = elr->lr_first_not_zeroed;
3717 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3718 ret = 0;
3719 }
3720 }
3721 return ret;
3722 }
bfff6873 3723
3d392b26 3724 for (; group < ngroups; group++) {
bfff6873
LC		 3725 gdp = ext4_get_group_desc(sb, group, NULL);
3726 if (!gdp) {
3727 ret = 1;
3728 break;
3729 }
3730
3731 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3732 break;
3733 }
3734
7f511862 3735 if (group >= ngroups)
bfff6873
LC		 3736 ret = 1;
3737
3738 if (!ret) {
39fec688 3739 start_time = ktime_get_real_ns();
bfff6873
LC		 3740 ret = ext4_init_inode_table(sb, group,
3741 elr->lr_timeout ? 0 : 1);
3d392b26 3742 trace_ext4_lazy_itable_init(sb, group);
bfff6873 3743 if (elr->lr_timeout == 0) {
39fec688
SX		 3744 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3745 EXT4_SB(elr->lr_super)->s_li_wait_mult);
bfff6873
LC		 3746 }
3747 elr->lr_next_sched = jiffies + elr->lr_timeout;
3748 elr->lr_next_group = group + 1;
3749 }
bfff6873
LC		 3750 return ret;
3751}
3752
3753/*
3754 * Remove lr_request from the list_request and free the
4ed5c033 3755 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 3756 */
3757static void ext4_remove_li_request(struct ext4_li_request *elr)
3758{
bfff6873
LC		 3759 if (!elr)
3760 return;
3761
bfff6873 3762 list_del(&elr->lr_request);
3d392b26 3763 EXT4_SB(elr->lr_super)->s_li_request = NULL;
bfff6873
LC		 3764 kfree(elr);
3765}
3766
3767static void ext4_unregister_li_request(struct super_block *sb)
3768{
1bb933fb
LC		 3769 mutex_lock(&ext4_li_mtx);
3770 if (!ext4_li_info) {
3771 mutex_unlock(&ext4_li_mtx);
bfff6873 3772 return;
1bb933fb 3773 }
bfff6873
LC		 3774
3775 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 3776 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 3777 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 3778 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 3779}
3780
8f1f7453
ES		 3781static struct task_struct *ext4_lazyinit_task;
3782
bfff6873
LC		 3783/*
3784 * This is the function where ext4lazyinit thread lives. It walks
3785 * through the request list searching for next scheduled filesystem.
3786 * When such a fs is found, run the lazy initialization request
3787 * (ext4_rn_li_request) and keep track of the time spend in this
3788 * function. Based on that time we compute next schedule time of
3789 * the request. When walking through the list is complete, compute
3790 * next waking time and put itself into sleep.
3791 */
3792static int ext4_lazyinit_thread(void *arg)
3793{
c30365b9 3794 struct ext4_lazy_init *eli = arg;
bfff6873
LC		 3795 struct list_head *pos, *n;
3796 struct ext4_li_request *elr;
4ed5c033 3797 unsigned long next_wakeup, cur;
bfff6873
LC		 3798
3799 BUG_ON(NULL == eli);
3b575495 3800 set_freezable();
bfff6873 3801
bfff6873
LC		 3802cont_thread:
3803 while (true) {
3804 next_wakeup = MAX_JIFFY_OFFSET;
3805
3806 mutex_lock(&eli->li_list_mtx);
3807 if (list_empty(&eli->li_request_list)) {
3808 mutex_unlock(&eli->li_list_mtx);
3809 goto exit_thread;
3810 }
bfff6873 3811 list_for_each_safe(pos, n, &eli->li_request_list) {
e22834f0
DM		 3812 int err = 0;
3813 int progress = 0;
bfff6873
LC		 3814 elr = list_entry(pos, struct ext4_li_request,
3815 lr_request);
3816
e22834f0
DM		 3817 if (time_before(jiffies, elr->lr_next_sched)) {
3818 if (time_before(elr->lr_next_sched, next_wakeup))
3819 next_wakeup = elr->lr_next_sched;
3820 continue;
3821 }
3822 if (down_read_trylock(&elr->lr_super->s_umount)) {
3823 if (sb_start_write_trylock(elr->lr_super)) {
3824 progress = 1;
3825 /*
3826 * We hold sb->s_umount, sb can not
3827 * be removed from the list, it is
3828 * now safe to drop li_list_mtx
3829 */
3830 mutex_unlock(&eli->li_list_mtx);
3831 err = ext4_run_li_request(elr);
3832 sb_end_write(elr->lr_super);
3833 mutex_lock(&eli->li_list_mtx);
3834 n = pos->next;
b2c78cd0 3835 }
e22834f0
DM		 3836 up_read((&elr->lr_super->s_umount));
3837 }
3838 /* error, remove the lazy_init job */
3839 if (err) {
3840 ext4_remove_li_request(elr);
3841 continue;
3842 }
3843 if (!progress) {
3844 elr->lr_next_sched = jiffies +
8032bf12 3845 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
bfff6873 3846 }
bfff6873
LC		 3847 if (time_before(elr->lr_next_sched, next_wakeup))
3848 next_wakeup = elr->lr_next_sched;
3849 }
3850 mutex_unlock(&eli->li_list_mtx);
3851
a0acae0e 3852 try_to_freeze();
bfff6873 3853
4ed5c033
LC		 3854 cur = jiffies;
3855 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 3856 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 3857 cond_resched();
3858 continue;
3859 }
3860
4ed5c033
LC		 3861 schedule_timeout_interruptible(next_wakeup - cur);
3862
8f1f7453
ES		 3863 if (kthread_should_stop()) {
3864 ext4_clear_request_list();
3865 goto exit_thread;
3866 }
bfff6873
LC		 3867 }
3868
3869exit_thread:
3870 /*
3871 * It looks like the request list is empty, but we need
3872 * to check it under the li_list_mtx lock, to prevent any
3873 * additions into it, and of course we should lock ext4_li_mtx
3874 * to atomically free the list and ext4_li_info, because at
3875 * this point another ext4 filesystem could be registering
3876 * new one.
3877 */
3878 mutex_lock(&ext4_li_mtx);
3879 mutex_lock(&eli->li_list_mtx);
3880 if (!list_empty(&eli->li_request_list)) {
3881 mutex_unlock(&eli->li_list_mtx);
3882 mutex_unlock(&ext4_li_mtx);
3883 goto cont_thread;
3884 }
3885 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3886 kfree(ext4_li_info);
3887 ext4_li_info = NULL;
3888 mutex_unlock(&ext4_li_mtx);
3889
3890 return 0;
3891}
3892
3893static void ext4_clear_request_list(void)
3894{
3895 struct list_head *pos, *n;
3896 struct ext4_li_request *elr;
3897
3898 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3899 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3900 elr = list_entry(pos, struct ext4_li_request,
3901 lr_request);
3902 ext4_remove_li_request(elr);
3903 }
3904 mutex_unlock(&ext4_li_info->li_list_mtx);
3905}
3906
3907static int ext4_run_lazyinit_thread(void)
3908{
8f1f7453
ES		 3909 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3910 ext4_li_info, "ext4lazyinit");
3911 if (IS_ERR(ext4_lazyinit_task)) {
3912 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3913 ext4_clear_request_list();
bfff6873
LC		 3914 kfree(ext4_li_info);
3915 ext4_li_info = NULL;
92b97816 3916 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3917 "initialization thread\n",
3918 err);
3919 return err;
3920 }
3921 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3922 return 0;
3923}
3924
3925/*
3926 * Check whether it make sense to run itable init. thread or not.
3927 * If there is at least one uninitialized inode table, return
3928 * corresponding group number, else the loop goes through all
3929 * groups and return total number of groups.
3930 */
3931static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3932{
3933 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3934 struct ext4_group_desc *gdp = NULL;
3935
8844618d
TT		 3936 if (!ext4_has_group_desc_csum(sb))
3937 return ngroups;
3938
bfff6873
LC		 3939 for (group = 0; group < ngroups; group++) {
3940 gdp = ext4_get_group_desc(sb, group, NULL);
3941 if (!gdp)
3942 continue;
3943
50122847 3944 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
bfff6873
LC		 3945 break;
3946 }
3947
3948 return group;
3949}
3950
3951static int ext4_li_info_new(void)
3952{
3953 struct ext4_lazy_init *eli = NULL;
3954
3955 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3956 if (!eli)
3957 return -ENOMEM;
3958
bfff6873
LC		 3959 INIT_LIST_HEAD(&eli->li_request_list);
3960 mutex_init(&eli->li_list_mtx);
3961
bfff6873
LC		 3962 eli->li_state |= EXT4_LAZYINIT_QUIT;
3963
3964 ext4_li_info = eli;
3965
3966 return 0;
3967}
3968
3969static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3970 ext4_group_t start)
3971{
bfff6873 3972 struct ext4_li_request *elr;
bfff6873
LC		 3973
3974 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3975 if (!elr)
3976 return NULL;
3977
3978 elr->lr_super = sb;
3d392b26 3979 elr->lr_first_not_zeroed = start;
21175ca4 3980 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3d392b26
TT		 3981 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3982 elr->lr_next_group = start;
21175ca4
HS		 3983 } else {
3984 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3d392b26 3985 }
bfff6873
LC		 3986
3987 /*
3988 * Randomize first schedule time of the request to
3989 * spread the inode table initialization requests
3990 * better.
3991 */
8032bf12 3992 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
bfff6873
LC		 3993 return elr;
3994}
3995
7f511862
TT		 3996int ext4_register_li_request(struct super_block *sb,
3997 ext4_group_t first_not_zeroed)
bfff6873
LC		 3998{
3999 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 4000 struct ext4_li_request *elr = NULL;
49598e04 4001 ext4_group_t ngroups = sbi->s_groups_count;
6c5a6cb9 4002 int ret = 0;
bfff6873 4003
7f511862 4004 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 4005 if (sbi->s_li_request != NULL) {
4006 /*
4007 * Reset timeout so it can be computed again, because
4008 * s_li_wait_mult might have changed.
4009 */
4010 sbi->s_li_request->lr_timeout = 0;
7f511862 4011 goto out;
51ce6511 4012 }
bfff6873 4013
426d15ad
JT		 4014 if (sb_rdonly(sb) ||
4015 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
4016 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
7f511862 4017 goto out;
bfff6873
LC		 4018
4019 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 4020 if (!elr) {
4021 ret = -ENOMEM;
4022 goto out;
4023 }
bfff6873
LC		 4024
4025 if (NULL == ext4_li_info) {
4026 ret = ext4_li_info_new();
4027 if (ret)
4028 goto out;
4029 }
4030
4031 mutex_lock(&ext4_li_info->li_list_mtx);
4032 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
4033 mutex_unlock(&ext4_li_info->li_list_mtx);
4034
4035 sbi->s_li_request = elr;
46e4690b
TM		 4036 /*
4037 * set elr to NULL here since it has been inserted to
4038 * the request_list and the removal and free of it is
4039 * handled by ext4_clear_request_list from now on.
4040 */
4041 elr = NULL;
bfff6873
LC		 4042
4043 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4044 ret = ext4_run_lazyinit_thread();
4045 if (ret)
4046 goto out;
4047 }
bfff6873 4048out:
beed5ecb
NK		 4049 mutex_unlock(&ext4_li_mtx);
4050 if (ret)
bfff6873 4051 kfree(elr);
bfff6873
LC		 4052 return ret;
4053}
4054
4055/*
4056 * We do not need to lock anything since this is called on
4057 * module unload.
4058 */
4059static void ext4_destroy_lazyinit_thread(void)
4060{
4061 /*
4062 * If thread exited earlier
4063 * there's nothing to be done.
4064 */
8f1f7453 4065 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 4066 return;
4067
8f1f7453 4068 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 4069}
4070
25ed6e8a
DW		 4071static int set_journal_csum_feature_set(struct super_block *sb)
4072{
4073 int ret = 1;
4074 int compat, incompat;
4075 struct ext4_sb_info *sbi = EXT4_SB(sb);
4076
9aa5d32b 4077 if (ext4_has_metadata_csum(sb)) {
db9ee220 4078 /* journal checksum v3 */
25ed6e8a 4079 compat = 0;
db9ee220 4080 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
25ed6e8a
DW		 4081 } else {
4082 /* journal checksum v1 */
4083 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4084 incompat = 0;
4085 }
4086
feb8c6d3
DW		 4087 jbd2_journal_clear_features(sbi->s_journal,
4088 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4089 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4090 JBD2_FEATURE_INCOMPAT_CSUM_V2);
25ed6e8a
DW		 4091 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4092 ret = jbd2_journal_set_features(sbi->s_journal,
4093 compat, 0,
4094 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4095 incompat);
4096 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4097 ret = jbd2_journal_set_features(sbi->s_journal,
4098 compat, 0,
4099 incompat);
4100 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4101 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4102 } else {
feb8c6d3
DW		 4103 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4104 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
25ed6e8a
DW		 4105 }
4106
4107 return ret;
4108}
4109
952fc18e
TT		 4110/*
4111 * Note: calculating the overhead so we can be compatible with
4112 * historical BSD practice is quite difficult in the face of
4113 * clusters/bigalloc. This is because multiple metadata blocks from
4114 * different block group can end up in the same allocation cluster.
4115 * Calculating the exact overhead in the face of clustered allocation
4116 * requires either O(all block bitmaps) in memory or O(number of block
4117 * groups**2) in time. We will still calculate the superblock for
4118 * older file systems --- and if we come across with a bigalloc file
4119 * system with zero in s_overhead_clusters the estimate will be close to
4120 * correct especially for very large cluster sizes --- but for newer
4121 * file systems, it's better to calculate this figure once at mkfs
4122 * time, and store it in the superblock. If the superblock value is
4123 * present (even for non-bigalloc file systems), we will use it.
4124 */
4125static int count_overhead(struct super_block *sb, ext4_group_t grp,
4126 char *buf)
4127{
4128 struct ext4_sb_info *sbi = EXT4_SB(sb);
4129 struct ext4_group_desc *gdp;
4130 ext4_fsblk_t first_block, last_block, b;
4131 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4132 int s, j, count = 0;
10b01ee9 4133 int has_super = ext4_bg_has_super(sb, grp);
952fc18e 4134
e2b911c5 4135 if (!ext4_has_feature_bigalloc(sb))
10b01ee9
TT		 4136 return (has_super + ext4_bg_num_gdb(sb, grp) +
4137 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
0548bbb8
TT		 4138 sbi->s_itb_per_group + 2);
4139
952fc18e
TT		 4140 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4141 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4142 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4143 for (i = 0; i < ngroups; i++) {
4144 gdp = ext4_get_group_desc(sb, i, NULL);
4145 b = ext4_block_bitmap(sb, gdp);
4146 if (b >= first_block && b <= last_block) {
4147 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4148 count++;
4149 }
4150 b = ext4_inode_bitmap(sb, gdp);
4151 if (b >= first_block && b <= last_block) {
4152 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4153 count++;
4154 }
4155 b = ext4_inode_table(sb, gdp);
4156 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4157 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4158 int c = EXT4_B2C(sbi, b - first_block);
4159 ext4_set_bit(c, buf);
4160 count++;
4161 }
4162 if (i != grp)
4163 continue;
4164 s = 0;
4165 if (ext4_bg_has_super(sb, grp)) {
4166 ext4_set_bit(s++, buf);
4167 count++;
4168 }
c48ae41b
TT		 4169 j = ext4_bg_num_gdb(sb, grp);
4170 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4171 ext4_error(sb, "Invalid number of block group "
4172 "descriptor blocks: %d", j);
4173 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
952fc18e 4174 }
c48ae41b
TT		 4175 count += j;
4176 for (; j > 0; j--)
4177 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
952fc18e
TT		 4178 }
4179 if (!count)
4180 return 0;
4181 return EXT4_CLUSTERS_PER_GROUP(sb) -
4182 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4183}
4184
4185/*
4186 * Compute the overhead and stash it in sbi->s_overhead
4187 */
4188int ext4_calculate_overhead(struct super_block *sb)
4189{
4190 struct ext4_sb_info *sbi = EXT4_SB(sb);
4191 struct ext4_super_block *es = sbi->s_es;
3c816ded
EW		 4192 struct inode *j_inode;
4193 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
952fc18e
TT		 4194 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4195 ext4_fsblk_t overhead = 0;
4fdb5543 4196 char *buf = (char *) get_zeroed_page(GFP_NOFS);
952fc18e 4197
952fc18e
TT		 4198 if (!buf)
4199 return -ENOMEM;
4200
4201 /*
4202 * Compute the overhead (FS structures). This is constant
4203 * for a given filesystem unless the number of block groups
4204 * changes so we cache the previous value until it does.
4205 */
4206
4207 /*
4208 * All of the blocks before first_data_block are overhead
4209 */
4210 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4211
4212 /*
4213 * Add the overhead found in each block group
4214 */
4215 for (i = 0; i < ngroups; i++) {
4216 int blks;
4217
4218 blks = count_overhead(sb, i, buf);
4219 overhead += blks;
4220 if (blks)
4221 memset(buf, 0, PAGE_SIZE);
4222 cond_resched();
4223 }
3c816ded
EW		 4224
4225 /*
4226 * Add the internal journal blocks whether the journal has been
4227 * loaded or not
4228 */
61ead714 4229 if (sbi->s_journal && !sbi->s_journal_bdev_file)
ede7dc7f 4230 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
f1eec3b0
RH		 4231 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4232 /* j_inum for internal journal is non-zero */
3c816ded 4233 j_inode = ext4_get_journal_inode(sb, j_inum);
ee5c8071 4234 if (!IS_ERR(j_inode)) {
3c816ded
EW		 4235 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4236 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4237 iput(j_inode);
4238 } else {
4239 ext4_msg(sb, KERN_ERR, "can't get journal size");
4240 }
4241 }
952fc18e
TT		 4242 sbi->s_overhead = overhead;
4243 smp_wmb();
4244 free_page((unsigned long) buf);
4245 return 0;
4246}
4247
b5799018 4248static void ext4_set_resv_clusters(struct super_block *sb)
27dd4385
LC		 4249{
4250 ext4_fsblk_t resv_clusters;
b5799018 4251 struct ext4_sb_info *sbi = EXT4_SB(sb);
27dd4385 4252
30fac0f7
JK		 4253 /*
4254 * There's no need to reserve anything when we aren't using extents.
4255 * The space estimates are exact, there are no unwritten extents,
4256 * hole punching doesn't need new metadata... This is needed especially
4257 * to keep ext2/3 backward compatibility.
4258 */
e2b911c5 4259 if (!ext4_has_feature_extents(sb))
b5799018 4260 return;
27dd4385
LC		 4261 /*
4262 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4263 * This should cover the situations where we can not afford to run
4264 * out of space like for example punch hole, or converting
556615dc 4265 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 4266 * allocation would require 1, or 2 blocks, higher numbers are
4267 * very rare.
4268 */
b5799018
TT		 4269 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4270 sbi->s_cluster_bits);
27dd4385
LC		 4271
4272 do_div(resv_clusters, 50);
4273 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4274
b5799018 4275 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
27dd4385
LC		 4276}
4277
ca9b404f
RA		 4278static const char *ext4_quota_mode(struct super_block *sb)
4279{
4280#ifdef CONFIG_QUOTA
4281 if (!ext4_quota_capable(sb))
4282 return "none";
4283
4284 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4285 return "journalled";
4286 else
4287 return "writeback";
4288#else
4289 return "disabled";
4290#endif
4291}
4292
188c299e
JK		 4293static void ext4_setup_csum_trigger(struct super_block *sb,
4294 enum ext4_journal_trigger_type type,
4295 void (*trigger)(
4296 struct jbd2_buffer_trigger_type *type,
4297 struct buffer_head *bh,
4298 void *mapped_data,
4299 size_t size))
4300{
4301 struct ext4_sb_info *sbi = EXT4_SB(sb);
4302
4303 sbi->s_journal_triggers[type].sb = sb;
4304 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4305}
4306
7edfd85b
LC		 4307static void ext4_free_sbi(struct ext4_sb_info *sbi)
4308{
4309 if (!sbi)
4310 return;
4311
4312 kfree(sbi->s_blockgroup_lock);
8012b866 4313 fs_put_dax(sbi->s_daxdev, NULL);
7edfd85b
LC		 4314 kfree(sbi);
4315}
4316
4317static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4318{
4319 struct ext4_sb_info *sbi;
4320
4321 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4322 if (!sbi)
4323 return NULL;
4324
8012b866
SR		 4325 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4326 NULL, NULL);
7edfd85b
LC		 4327
4328 sbi->s_blockgroup_lock =
4329 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4330
4331 if (!sbi->s_blockgroup_lock)
4332 goto err_out;
4333
4334 sb->s_fs_info = sbi;
4335 sbi->s_sb = sb;
4336 return sbi;
4337err_out:
8012b866 4338 fs_put_dax(sbi->s_daxdev, NULL);
7edfd85b
LC		 4339 kfree(sbi);
4340 return NULL;
4341}
4342
5f6d662d
JY		 4343static void ext4_set_def_opts(struct super_block *sb,
4344 struct ext4_super_block *es)
4345{
4346 unsigned long def_mount_opts;
4347
4348 /* Set defaults before we parse the mount options */
4349 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4350 set_opt(sb, INIT_INODE_TABLE);
4351 if (def_mount_opts & EXT4_DEFM_DEBUG)
4352 set_opt(sb, DEBUG);
4353 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4354 set_opt(sb, GRPID);
4355 if (def_mount_opts & EXT4_DEFM_UID16)
4356 set_opt(sb, NO_UID32);
4357 /* xattr user namespace & acls are now defaulted on */
4358 set_opt(sb, XATTR_USER);
4359#ifdef CONFIG_EXT4_FS_POSIX_ACL
4360 set_opt(sb, POSIX_ACL);
4361#endif
4362 if (ext4_has_feature_fast_commit(sb))
4363 set_opt2(sb, JOURNAL_FAST_COMMIT);
4364 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4365 if (ext4_has_metadata_csum(sb))
4366 set_opt(sb, JOURNAL_CHECKSUM);
4367
4368 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4369 set_opt(sb, JOURNAL_DATA);
4370 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4371 set_opt(sb, ORDERED_DATA);
4372 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4373 set_opt(sb, WRITEBACK_DATA);
4374
4375 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4376 set_opt(sb, ERRORS_PANIC);
4377 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4378 set_opt(sb, ERRORS_CONT);
4379 else
4380 set_opt(sb, ERRORS_RO);
4381 /* block_validity enabled by default; disable with noblock_validity */
4382 set_opt(sb, BLOCK_VALIDITY);
4383 if (def_mount_opts & EXT4_DEFM_DISCARD)
4384 set_opt(sb, DISCARD);
4385
4386 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4387 set_opt(sb, BARRIER);
4388
4389 /*
4390 * enable delayed allocation by default
4391 * Use -o nodelalloc to turn it off
4392 */
4393 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4394 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4395 set_opt(sb, DELALLOC);
3df11e27 4396
e89fdcc4 4397 if (sb->s_blocksize <= PAGE_SIZE)
3df11e27 4398 set_opt(sb, DIOREAD_NOLOCK);
5f6d662d
JY		 4399}
4400
c8267c51 4401static int ext4_handle_clustersize(struct super_block *sb)
4a8557b0
JY		 4402{
4403 struct ext4_sb_info *sbi = EXT4_SB(sb);
4404 struct ext4_super_block *es = sbi->s_es;
4405 int clustersize;
4406
4407 /* Handle clustersize */
4408 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4409 if (ext4_has_feature_bigalloc(sb)) {
c8267c51 4410 if (clustersize < sb->s_blocksize) {
4a8557b0
JY		 4411 ext4_msg(sb, KERN_ERR,
4412 "cluster size (%d) smaller than "
c8267c51 4413 "block size (%lu)", clustersize, sb->s_blocksize);
4a8557b0
JY		 4414 return -EINVAL;
4415 }
4416 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4417 le32_to_cpu(es->s_log_block_size);
4a8557b0 4418 } else {
c8267c51 4419 if (clustersize != sb->s_blocksize) {
4a8557b0
JY		 4420 ext4_msg(sb, KERN_ERR,
4421 "fragment/cluster size (%d) != "
c8267c51 4422 "block size (%lu)", clustersize, sb->s_blocksize);
4a8557b0
JY		 4423 return -EINVAL;
4424 }
c8267c51 4425 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4a8557b0
JY		 4426 ext4_msg(sb, KERN_ERR,
4427 "#blocks per group too big: %lu",
4428 sbi->s_blocks_per_group);
4429 return -EINVAL;
4430 }
4a8557b0
JY		 4431 sbi->s_cluster_bits = 0;
4432 }
40da553f
JK		 4433 sbi->s_clusters_per_group = le32_to_cpu(es->s_clusters_per_group);
4434 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4435 ext4_msg(sb, KERN_ERR, "#clusters per group too big: %lu",
4436 sbi->s_clusters_per_group);
4437 return -EINVAL;
4438 }
4439 if (sbi->s_blocks_per_group !=
4440 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4441 ext4_msg(sb, KERN_ERR,
4442 "blocks per group (%lu) and clusters per group (%lu) inconsistent",
4443 sbi->s_blocks_per_group, sbi->s_clusters_per_group);
4444 return -EINVAL;
4445 }
c8267c51 4446 sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4a8557b0
JY		 4447
4448 /* Do we have standard group size of clustersize * 8 blocks ? */
4449 if (sbi->s_blocks_per_group == clustersize << 3)
4450 set_opt2(sb, STD_GROUP_SIZE);
4451
4452 return 0;
4453}
4454
f7314a67
JY		 4455static void ext4_fast_commit_init(struct super_block *sb)
4456{
4457 struct ext4_sb_info *sbi = EXT4_SB(sb);
4458
4459 /* Initialize fast commit stuff */
4460 atomic_set(&sbi->s_fc_subtid, 0);
4461 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4462 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4463 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4464 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4465 sbi->s_fc_bytes = 0;
4466 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4467 sbi->s_fc_ineligible_tid = 0;
4468 spin_lock_init(&sbi->s_fc_lock);
4469 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4470 sbi->s_fc_replay_state.fc_regions = NULL;
4471 sbi->s_fc_replay_state.fc_regions_size = 0;
4472 sbi->s_fc_replay_state.fc_regions_used = 0;
4473 sbi->s_fc_replay_state.fc_regions_valid = 0;
4474 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4475 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4476 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4477}
4478
0e495f7c 4479static int ext4_inode_info_init(struct super_block *sb,
c8267c51 4480 struct ext4_super_block *es)
0e495f7c
JY		 4481{
4482 struct ext4_sb_info *sbi = EXT4_SB(sb);
4483
4484 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4485 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4486 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4487 } else {
4488 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4489 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4490 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4491 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4492 sbi->s_first_ino);
4493 return -EINVAL;
4494 }
4495 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4496 (!is_power_of_2(sbi->s_inode_size)) ||
c8267c51 4497 (sbi->s_inode_size > sb->s_blocksize)) {
0e495f7c
JY		 4498 ext4_msg(sb, KERN_ERR,
4499 "unsupported inode size: %d",
4500 sbi->s_inode_size);
c8267c51 4501 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
0e495f7c
JY		 4502 return -EINVAL;
4503 }
4504 /*
4505 * i_atime_extra is the last extra field available for
4506 * [acm]times in struct ext4_inode. Checking for that
4507 * field should suffice to ensure we have extra space
4508 * for all three.
4509 */
4510 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4511 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4512 sb->s_time_gran = 1;
4513 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4514 } else {
4515 sb->s_time_gran = NSEC_PER_SEC;
4516 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4517 }
4518 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4519 }
4520
4521 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4522 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4523 EXT4_GOOD_OLD_INODE_SIZE;
4524 if (ext4_has_feature_extra_isize(sb)) {
4525 unsigned v, max = (sbi->s_inode_size -
4526 EXT4_GOOD_OLD_INODE_SIZE);
4527
4528 v = le16_to_cpu(es->s_want_extra_isize);
4529 if (v > max) {
4530 ext4_msg(sb, KERN_ERR,
4531 "bad s_want_extra_isize: %d", v);
4532 return -EINVAL;
4533 }
4534 if (sbi->s_want_extra_isize < v)
4535 sbi->s_want_extra_isize = v;
4536
4537 v = le16_to_cpu(es->s_min_extra_isize);
4538 if (v > max) {
4539 ext4_msg(sb, KERN_ERR,
4540 "bad s_min_extra_isize: %d", v);
4541 return -EINVAL;
4542 }
4543 if (sbi->s_want_extra_isize < v)
4544 sbi->s_want_extra_isize = v;
4545 }
4546 }
4547
4548 return 0;
4549}
4550
39c135b0
JY		 4551#if IS_ENABLED(CONFIG_UNICODE)
4552static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4553{
4554 const struct ext4_sb_encodings *encoding_info;
4555 struct unicode_map *encoding;
4556 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4557
4558 if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4559 return 0;
4560
4561 encoding_info = ext4_sb_read_encoding(es);
4562 if (!encoding_info) {
4563 ext4_msg(sb, KERN_ERR,
4564 "Encoding requested by superblock is unknown");
4565 return -EINVAL;
4566 }
4567
4568 encoding = utf8_load(encoding_info->version);
4569 if (IS_ERR(encoding)) {
4570 ext4_msg(sb, KERN_ERR,
4571 "can't mount with superblock charset: %s-%u.%u.%u "
4572 "not supported by the kernel. flags: 0x%x.",
4573 encoding_info->name,
4574 unicode_major(encoding_info->version),
4575 unicode_minor(encoding_info->version),
4576 unicode_rev(encoding_info->version),
4577 encoding_flags);
4578 return -EINVAL;
4579 }
4580 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4581 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4582 unicode_major(encoding_info->version),
4583 unicode_minor(encoding_info->version),
4584 unicode_rev(encoding_info->version),
4585 encoding_flags);
4586
4587 sb->s_encoding = encoding;
4588 sb->s_encoding_flags = encoding_flags;
4589
4590 return 0;
4591}
4592#else
4593static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4594{
4595 return 0;
4596}
4597#endif
4598
b26458d1
JY		 4599static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4600{
4601 struct ext4_sb_info *sbi = EXT4_SB(sb);
4602
4603 /* Warn if metadata_csum and gdt_csum are both set. */
4604 if (ext4_has_feature_metadata_csum(sb) &&
4605 ext4_has_feature_gdt_csum(sb))
4606 ext4_warning(sb, "metadata_csum and uninit_bg are "
4607 "redundant flags; please run fsck.");
4608
4609 /* Check for a known checksum algorithm */
4610 if (!ext4_verify_csum_type(sb, es)) {
4611 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4612 "unknown checksum algorithm.");
4613 return -EINVAL;
4614 }
4615 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4616 ext4_orphan_file_block_trigger);
4617
4618 /* Load the checksum driver */
4619 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4620 if (IS_ERR(sbi->s_chksum_driver)) {
4621 int ret = PTR_ERR(sbi->s_chksum_driver);
4622 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4623 sbi->s_chksum_driver = NULL;
4624 return ret;
4625 }
4626
4627 /* Check superblock checksum */
4628 if (!ext4_superblock_csum_verify(sb, es)) {
4629 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4630 "invalid superblock checksum. Run e2fsck?");
4631 return -EFSBADCRC;
4632 }
4633
4634 /* Precompute checksum seed for all metadata */
4635 if (ext4_has_feature_csum_seed(sb))
4636 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4637 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4638 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4639 sizeof(es->s_uuid));
4640 return 0;
4641}
4642
d7f3542b
JY		 4643static int ext4_check_feature_compatibility(struct super_block *sb,
4644 struct ext4_super_block *es,
4645 int silent)
4646{
54902099
JY		 4647 struct ext4_sb_info *sbi = EXT4_SB(sb);
4648
d7f3542b
JY		 4649 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4650 (ext4_has_compat_features(sb) ||
4651 ext4_has_ro_compat_features(sb) ||
4652 ext4_has_incompat_features(sb)))
4653 ext4_msg(sb, KERN_WARNING,
4654 "feature flags set on rev 0 fs, "
4655 "running e2fsck is recommended");
4656
4657 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4658 set_opt2(sb, HURD_COMPAT);
4659 if (ext4_has_feature_64bit(sb)) {
4660 ext4_msg(sb, KERN_ERR,
4661 "The Hurd can't support 64-bit file systems");
4662 return -EINVAL;
4663 }
4664
4665 /*
4666 * ea_inode feature uses l_i_version field which is not
4667 * available in HURD_COMPAT mode.
4668 */
4669 if (ext4_has_feature_ea_inode(sb)) {
4670 ext4_msg(sb, KERN_ERR,
4671 "ea_inode feature is not supported for Hurd");
4672 return -EINVAL;
4673 }
4674 }
4675
4676 if (IS_EXT2_SB(sb)) {
4677 if (ext2_feature_set_ok(sb))
4678 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4679 "using the ext4 subsystem");
4680 else {
4681 /*
4682 * If we're probing be silent, if this looks like
4683 * it's actually an ext[34] filesystem.
4684 */
4685 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4686 return -EINVAL;
4687 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4688 "to feature incompatibilities");
4689 return -EINVAL;
4690 }
4691 }
4692
4693 if (IS_EXT3_SB(sb)) {
4694 if (ext3_feature_set_ok(sb))
4695 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4696 "using the ext4 subsystem");
4697 else {
4698 /*
4699 * If we're probing be silent, if this looks like
4700 * it's actually an ext4 filesystem.
4701 */
4702 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4703 return -EINVAL;
4704 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4705 "to feature incompatibilities");
4706 return -EINVAL;
4707 }
4708 }
4709
4710 /*
4711 * Check feature flags regardless of the revision level, since we
4712 * previously didn't change the revision level when setting the flags,
4713 * so there is a chance incompat flags are set on a rev 0 filesystem.
4714 */
4715 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4716 return -EINVAL;
4717
54902099
JY		 4718 if (sbi->s_daxdev) {
4719 if (sb->s_blocksize == PAGE_SIZE)
4720 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4721 else
4722 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
4723 }
4724
4725 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4726 if (ext4_has_feature_inline_data(sb)) {
4727 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4728 " that may contain inline data");
4729 return -EINVAL;
4730 }
4731 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4732 ext4_msg(sb, KERN_ERR,
4733 "DAX unsupported by block device.");
4734 return -EINVAL;
4735 }
4736 }
4737
4738 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4739 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4740 es->s_encryption_level);
4741 return -EINVAL;
4742 }
4743
d7f3542b
JY		 4744 return 0;
4745}
4746
68e62439 4747static int ext4_check_geometry(struct super_block *sb,
bc62dbf9
JY		 4748 struct ext4_super_block *es)
4749{
4750 struct ext4_sb_info *sbi = EXT4_SB(sb);
4751 __u64 blocks_count;
269e9226
JY		 4752 int err;
4753
4754 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
4755 ext4_msg(sb, KERN_ERR,
4756 "Number of reserved GDT blocks insanely large: %d",
4757 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4758 return -EINVAL;
4759 }
4760 /*
4761 * Test whether we have more sectors than will fit in sector_t,
4762 * and whether the max offset is addressable by the page cache.
4763 */
4764 err = generic_check_addressable(sb->s_blocksize_bits,
4765 ext4_blocks_count(es));
4766 if (err) {
4767 ext4_msg(sb, KERN_ERR, "filesystem"
4768 " too large to mount safely on this system");
4769 return err;
4770 }
bc62dbf9
JY		 4771
4772 /* check blocks count against device size */
4773 blocks_count = sb_bdev_nr_blocks(sb);
4774 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4775 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4776 "exceeds size of device (%llu blocks)",
4777 ext4_blocks_count(es), blocks_count);
4778 return -EINVAL;
4779 }
4780
4781 /*
4782 * It makes no sense for the first data block to be beyond the end
4783 * of the filesystem.
4784 */
4785 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4786 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4787 "block %u is beyond end of filesystem (%llu)",
4788 le32_to_cpu(es->s_first_data_block),
4789 ext4_blocks_count(es));
4790 return -EINVAL;
4791 }
4792 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4793 (sbi->s_cluster_ratio == 1)) {
4794 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4795 "block is 0 with a 1k block and cluster size");
4796 return -EINVAL;
4797 }
4798
4799 blocks_count = (ext4_blocks_count(es) -
4800 le32_to_cpu(es->s_first_data_block) +
4801 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4802 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4803 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4804 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4805 "(block count %llu, first data block %u, "
4806 "blocks per group %lu)", blocks_count,
4807 ext4_blocks_count(es),
4808 le32_to_cpu(es->s_first_data_block),
4809 EXT4_BLOCKS_PER_GROUP(sb));
4810 return -EINVAL;
4811 }
4812 sbi->s_groups_count = blocks_count;
4813 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4814 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4815 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4816 le32_to_cpu(es->s_inodes_count)) {
4817 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4818 le32_to_cpu(es->s_inodes_count),
4819 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4820 return -EINVAL;
4821 }
4822
4823 return 0;
4824}
4825
a4e6a511
JY		 4826static int ext4_group_desc_init(struct super_block *sb,
4827 struct ext4_super_block *es,
4828 ext4_fsblk_t logical_sb_block,
4829 ext4_group_t *first_not_zeroed)
4830{
4831 struct ext4_sb_info *sbi = EXT4_SB(sb);
4832 unsigned int db_count;
4833 ext4_fsblk_t block;
a4e6a511
JY		 4834 int i;
4835
4836 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4837 EXT4_DESC_PER_BLOCK(sb);
4838 if (ext4_has_feature_meta_bg(sb)) {
4839 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4840 ext4_msg(sb, KERN_WARNING,
4841 "first meta block group too large: %u "
4842 "(group descriptor block count %u)",
4843 le32_to_cpu(es->s_first_meta_bg), db_count);
4844 return -EINVAL;
4845 }
4846 }
4847 rcu_assign_pointer(sbi->s_group_desc,
4848 kvmalloc_array(db_count,
4849 sizeof(struct buffer_head *),
4850 GFP_KERNEL));
4851 if (sbi->s_group_desc == NULL) {
4852 ext4_msg(sb, KERN_ERR, "not enough memory");
4853 return -ENOMEM;
4854 }
4855
4856 bgl_lock_init(sbi->s_blockgroup_lock);
4857
4858 /* Pre-read the descriptors into the buffer cache */
4859 for (i = 0; i < db_count; i++) {
4860 block = descriptor_loc(sb, logical_sb_block, i);
4861 ext4_sb_breadahead_unmovable(sb, block);
4862 }
4863
4864 for (i = 0; i < db_count; i++) {
4865 struct buffer_head *bh;
4866
4867 block = descriptor_loc(sb, logical_sb_block, i);
4868 bh = ext4_sb_bread_unmovable(sb, block);
4869 if (IS_ERR(bh)) {
4870 ext4_msg(sb, KERN_ERR,
4871 "can't read group descriptor %d", i);
4872 sbi->s_gdb_count = i;
172e344e 4873 return PTR_ERR(bh);
a4e6a511
JY		 4874 }
4875 rcu_read_lock();
4876 rcu_dereference(sbi->s_group_desc)[i] = bh;
4877 rcu_read_unlock();
4878 }
4879 sbi->s_gdb_count = db_count;
4880 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) {
4881 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
172e344e 4882 return -EFSCORRUPTED;
a4e6a511 4883 }
172e344e 4884
a4e6a511 4885 return 0;
a4e6a511
JY		 4886}
4887
9c1dd22d
JY		 4888static int ext4_load_and_init_journal(struct super_block *sb,
4889 struct ext4_super_block *es,
4890 struct ext4_fs_context *ctx)
4891{
4892 struct ext4_sb_info *sbi = EXT4_SB(sb);
4893 int err;
4894
4895 err = ext4_load_journal(sb, es, ctx->journal_devnum);
4896 if (err)
4897 return err;
4898
4899 if (ext4_has_feature_64bit(sb) &&
4900 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4901 JBD2_FEATURE_INCOMPAT_64BIT)) {
4902 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4903 goto out;
4904 }
4905
4906 if (!set_journal_csum_feature_set(sb)) {
4907 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4908 "feature set");
4909 goto out;
4910 }
4911
4912 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4913 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4914 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4915 ext4_msg(sb, KERN_ERR,
4916 "Failed to set fast commit journal feature");
4917 goto out;
4918 }
4919
4920 /* We have now updated the journal if required, so we can
4921 * validate the data journaling mode. */
4922 switch (test_opt(sb, DATA_FLAGS)) {
4923 case 0:
4924 /* No mode set, assume a default based on the journal
4925 * capabilities: ORDERED_DATA if the journal can
4926 * cope, else JOURNAL_DATA
4927 */
4928 if (jbd2_journal_check_available_features
4929 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4930 set_opt(sb, ORDERED_DATA);
4931 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4932 } else {
4933 set_opt(sb, JOURNAL_DATA);
4934 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4935 }
4936 break;
4937
4938 case EXT4_MOUNT_ORDERED_DATA:
4939 case EXT4_MOUNT_WRITEBACK_DATA:
4940 if (!jbd2_journal_check_available_features
4941 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4942 ext4_msg(sb, KERN_ERR, "Journal does not support "
4943 "requested data journaling mode");
4944 goto out;
4945 }
4946 break;
4947 default:
4948 break;
4949 }
4950
4951 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4952 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4953 ext4_msg(sb, KERN_ERR, "can't mount with "
4954 "journal_async_commit in data=ordered mode");
4955 goto out;
4956 }
4957
4958 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
4959
4960 sbi->s_journal->j_submit_inode_data_buffers =
4961 ext4_journal_submit_inode_data_buffers;
4962 sbi->s_journal->j_finish_inode_data_buffers =
4963 ext4_journal_finish_inode_data_buffers;
4964
4965 return 0;
4966
4967out:
bb15cea2
TT		 4968 /* flush s_sb_upd_work before destroying the journal. */
4969 flush_work(&sbi->s_sb_upd_work);
9c1dd22d
JY		 4970 jbd2_journal_destroy(sbi->s_journal);
4971 sbi->s_journal = NULL;
9f2a1d9f 4972 return -EINVAL;
9c1dd22d
JY		 4973}
4974
68e62439 4975static int ext4_check_journal_data_mode(struct super_block *sb)
a5991e53
JY		 4976{
4977 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4978 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4979 "data=journal disables delayed allocation, "
4980 "dioread_nolock, O_DIRECT and fast_commit support!\n");
4981 /* can't mount with both data=journal and dioread_nolock. */
4982 clear_opt(sb, DIOREAD_NOLOCK);
4983 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4984 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4985 ext4_msg(sb, KERN_ERR, "can't mount with "
4986 "both data=journal and delalloc");
4987 return -EINVAL;
4988 }
4989 if (test_opt(sb, DAX_ALWAYS)) {
4990 ext4_msg(sb, KERN_ERR, "can't mount with "
4991 "both data=journal and dax");
4992 return -EINVAL;
4993 }
4994 if (ext4_has_feature_encrypt(sb)) {
4995 ext4_msg(sb, KERN_WARNING,
4996 "encrypted files will use data=ordered "
4997 "instead of data journaling mode");
4998 }
4999 if (test_opt(sb, DELALLOC))
5000 clear_opt(sb, DELALLOC);
5001 } else {
5002 sb->s_iflags |= SB_I_CGROUPWB;
5003 }
5004
5005 return 0;
5006}
5007
a7a79c29
JY		 5008static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
5009 int silent)
ac27a0ec 5010{
7edfd85b 5011 struct ext4_sb_info *sbi = EXT4_SB(sb);
a7a79c29 5012 struct ext4_super_block *es;
70bbb3e0 5013 ext4_fsblk_t logical_sb_block;
ac27a0ec 5014 unsigned long offset = 0;
a7a79c29
JY		 5015 struct buffer_head *bh;
5016 int ret = -EINVAL;
4a8557b0 5017 int blocksize;
705895b6 5018
617ba13b 5019 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 5020 if (!blocksize) {
b31e1552 5021 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
a7a79c29 5022 return -EINVAL;
ac27a0ec
DK		 5023 }
5024
5025 /*
617ba13b 5026 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 5027 * block sizes. We need to calculate the offset from buffer start.
5028 */
617ba13b 5029 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
7edfd85b 5030 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
70bbb3e0 5031 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 5032 } else {
7edfd85b 5033 logical_sb_block = sbi->s_sb_block;
ac27a0ec
DK		 5034 }
5035
8394a6ab 5036 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5037 if (IS_ERR(bh)) {
b31e1552 5038 ext4_msg(sb, KERN_ERR, "unable to read superblock");
a7a79c29 5039 return PTR_ERR(bh);
ac27a0ec
DK		 5040 }
5041 /*
5042 * Note: s_es must be initialized as soon as possible because
617ba13b 5043 * some ext4 macro-instructions depend on its value
ac27a0ec 5044 */
2716b802 5045 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 5046 sbi->s_es = es;
5047 sb->s_magic = le16_to_cpu(es->s_magic);
a5fc5119
JY		 5048 if (sb->s_magic != EXT4_SUPER_MAGIC) {
5049 if (!silent)
5050 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
a7a79c29 5051 goto out;
a5fc5119
JY		 5052 }
5053
a7a79c29
JY		 5054 if (le32_to_cpu(es->s_log_block_size) >
5055 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5056 ext4_msg(sb, KERN_ERR,
5057 "Invalid log block size: %u",
5058 le32_to_cpu(es->s_log_block_size));
5059 goto out;
5060 }
5061 if (le32_to_cpu(es->s_log_cluster_size) >
5062 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5063 ext4_msg(sb, KERN_ERR,
5064 "Invalid log cluster size: %u",
5065 le32_to_cpu(es->s_log_cluster_size));
5066 goto out;
5067 }
5068
5069 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
5070
5071 /*
5072 * If the default block size is not the same as the real block size,
5073 * we need to reload it.
5074 */
5075 if (sb->s_blocksize == blocksize) {
5076 *lsb = logical_sb_block;
5077 sbi->s_sbh = bh;
5078 return 0;
5079 }
5080
5081 /*
5082 * bh must be released before kill_bdev(), otherwise
5083 * it won't be freed and its page also. kill_bdev()
5084 * is called by sb_set_blocksize().
5085 */
5086 brelse(bh);
5087 /* Validate the filesystem blocksize */
5088 if (!sb_set_blocksize(sb, blocksize)) {
5089 ext4_msg(sb, KERN_ERR, "bad block size %d",
5090 blocksize);
5091 bh = NULL;
5092 goto out;
5093 }
5094
5095 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5096 offset = do_div(logical_sb_block, blocksize);
5097 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5098 if (IS_ERR(bh)) {
5099 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5100 ret = PTR_ERR(bh);
5101 bh = NULL;
5102 goto out;
5103 }
5104 es = (struct ext4_super_block *)(bh->b_data + offset);
5105 sbi->s_es = es;
5106 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5107 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5108 goto out;
5109 }
5110 *lsb = logical_sb_block;
5111 sbi->s_sbh = bh;
5112 return 0;
5113out:
5114 brelse(bh);
5115 return ret;
5116}
5117
db9345d9
JY		 5118static void ext4_hash_info_init(struct super_block *sb)
5119{
5120 struct ext4_sb_info *sbi = EXT4_SB(sb);
5121 struct ext4_super_block *es = sbi->s_es;
5122 unsigned int i;
5123
5124 for (i = 0; i < 4; i++)
5125 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5126
5127 sbi->s_def_hash_version = es->s_def_hash_version;
5128 if (ext4_has_feature_dir_index(sb)) {
5129 i = le32_to_cpu(es->s_flags);
5130 if (i & EXT2_FLAGS_UNSIGNED_HASH)
5131 sbi->s_hash_unsigned = 3;
5132 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
5133#ifdef __CHAR_UNSIGNED__
5134 if (!sb_rdonly(sb))
5135 es->s_flags |=
5136 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5137 sbi->s_hash_unsigned = 3;
5138#else
5139 if (!sb_rdonly(sb))
5140 es->s_flags |=
5141 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
5142#endif
5143 }
5144 }
5145}
5146
107d2be9
JY		 5147static int ext4_block_group_meta_init(struct super_block *sb, int silent)
5148{
5149 struct ext4_sb_info *sbi = EXT4_SB(sb);
5150 struct ext4_super_block *es = sbi->s_es;
5151 int has_huge_files;
5152
5153 has_huge_files = ext4_has_feature_huge_file(sb);
5154 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
5155 has_huge_files);
5156 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
5157
5158 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
5159 if (ext4_has_feature_64bit(sb)) {
5160 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
5161 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
5162 !is_power_of_2(sbi->s_desc_size)) {
5163 ext4_msg(sb, KERN_ERR,
5164 "unsupported descriptor size %lu",
5165 sbi->s_desc_size);
5166 return -EINVAL;
5167 }
5168 } else
5169 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
5170
5171 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
5172 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
5173
5174 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
5175 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5176 if (!silent)
5177 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5178 return -EINVAL;
5179 }
5180 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
5181 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
5182 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
5183 sbi->s_inodes_per_group);
5184 return -EINVAL;
5185 }
5186 sbi->s_itb_per_group = sbi->s_inodes_per_group /
5187 sbi->s_inodes_per_block;
5188 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
5189 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
5190 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5191 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
5192
5193 return 0;
5194}
5195
a7a79c29
JY		 5196static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5197{
5198 struct ext4_super_block *es = NULL;
5199 struct ext4_sb_info *sbi = EXT4_SB(sb);
a7a79c29
JY		 5200 ext4_fsblk_t logical_sb_block;
5201 struct inode *root;
107d2be9 5202 int needs_recovery;
d4fab7b2 5203 int err;
a7a79c29
JY		 5204 ext4_group_t first_not_zeroed;
5205 struct ext4_fs_context *ctx = fc->fs_private;
5206 int silent = fc->sb_flags & SB_SILENT;
5207
5208 /* Set defaults for the variables that will be set during parsing */
5209 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5210 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5211
5212 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5213 sbi->s_sectors_written_start =
5214 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5215
a7a79c29
JY		 5216 err = ext4_load_super(sb, &logical_sb_block, silent);
5217 if (err)
5218 goto out_fail;
5219
5220 es = sbi->s_es;
afc32f7e 5221 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 5222
b26458d1
JY		 5223 err = ext4_init_metadata_csum(sb, es);
5224 if (err)
a5fc5119 5225 goto failed_mount;
a9c47317 5226
5f6d662d 5227 ext4_set_def_opts(sb, es);
ac27a0ec 5228
08cefc7a
EB		 5229 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
5230 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 5231 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5232 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5233 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 5234
51ce6511
LC		 5235 /*
5236 * set default s_li_wait_mult for lazyinit, for the case there is
5237 * no mount option specified.
5238 */
5239 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5240
d4fab7b2
TT		 5241 err = ext4_inode_info_init(sb, es);
5242 if (err)
0e495f7c 5243 goto failed_mount;
9803387c 5244
7edfd85b
LC		 5245 err = parse_apply_sb_mount_options(sb, ctx);
5246 if (err < 0)
5247 goto failed_mount;
5248
5a916be1 5249 sbi->s_def_mount_opt = sbi->s_mount_opt;
e3645d72 5250 sbi->s_def_mount_opt2 = sbi->s_mount_opt2;
7edfd85b
LC		 5251
5252 err = ext4_check_opt_consistency(fc, sb);
5253 if (err < 0)
5254 goto failed_mount;
5255
85456054 5256 ext4_apply_options(fc, sb);
ac27a0ec 5257
d4fab7b2
TT		 5258 err = ext4_encoding_init(sb, es);
5259 if (err)
39c135b0 5260 goto failed_mount;
c83ad55e 5261
d4fab7b2
TT		 5262 err = ext4_check_journal_data_mode(sb);
5263 if (err)
a5991e53 5264 goto failed_mount;
56889787 5265
1751e8a6
LT		 5266 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5267 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec 5268
1ff20307
JL		 5269 /* i_version is always enabled now */
5270 sb->s_flags |= SB_I_VERSION;
5271
d4fab7b2
TT		 5272 err = ext4_check_feature_compatibility(sb, es, silent);
5273 if (err)
ac27a0ec 5274 goto failed_mount;
a13fb1a4 5275
d4fab7b2
TT		 5276 err = ext4_block_group_meta_init(sb, silent);
5277 if (err)
5b9554dc 5278 goto failed_mount;
5b9554dc 5279
db9345d9 5280 ext4_hash_info_init(sb);
ac27a0ec 5281
d4fab7b2
TT		 5282 err = ext4_handle_clustersize(sb);
5283 if (err)
4a8557b0 5284 goto failed_mount;
960fd856 5285
d4fab7b2
TT		 5286 err = ext4_check_geometry(sb, es);
5287 if (err)
0f2ddca6 5288 goto failed_mount;
0f2ddca6 5289
235699a8 5290 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
c92dc856 5291 spin_lock_init(&sbi->s_error_lock);
bb15cea2 5292 INIT_WORK(&sbi->s_sb_upd_work, update_super_work);
04496411 5293
172e344e
YB		 5294 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed);
5295 if (err)
5296 goto failed_mount3;
5297
d4fab7b2
TT		 5298 err = ext4_es_register_shrinker(sbi);
5299 if (err)
ce7e010a 5300 goto failed_mount3;
ce7e010a 5301
c9de560d 5302 sbi->s_stripe = ext4_get_stripe_size(sbi);
c3defd99
KS		 5303 /*
5304 * It's hard to get stripe aligned blocks if stripe is not aligned with
5305 * cluster, just disable stripe and alert user to simpfy code and avoid
5306 * stripe aligned allocation which will rarely successes.
5307 */
5308 if (sbi->s_stripe > 0 && sbi->s_cluster_ratio > 1 &&
5309 sbi->s_stripe % sbi->s_cluster_ratio != 0) {
5310 ext4_msg(sb, KERN_WARNING,
5311 "stripe (%lu) is not aligned with cluster size (%u), "
5312 "stripe is disabled",
5313 sbi->s_stripe, sbi->s_cluster_ratio);
5314 sbi->s_stripe = 0;
5315 }
67a5da56 5316 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 5317
f9ae9cf5
TT		 5318 /*
5319 * set up enough so that it can read an inode
5320 */
f6e63f90 5321 sb->s_op = &ext4_sops;
617ba13b
MC		 5322 sb->s_export_op = &ext4_export_ops;
5323 sb->s_xattr = ext4_xattr_handlers;
643fa961 5324#ifdef CONFIG_FS_ENCRYPTION
a7550b30 5325 sb->s_cop = &ext4_cryptops;
ffcc4182 5326#endif
c93d8f88
EB		 5327#ifdef CONFIG_FS_VERITY
5328 sb->s_vop = &ext4_verityops;
5329#endif
ac27a0ec 5330#ifdef CONFIG_QUOTA
617ba13b 5331 sb->dq_op = &ext4_quota_operations;
e2b911c5 5332 if (ext4_has_feature_quota(sb))
1fa5efe3 5333 sb->s_qcop = &dquot_quotactl_sysfile_ops;
262b4662
JK		 5334 else
5335 sb->s_qcop = &ext4_qctl_operations;
689c958c 5336 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
ac27a0ec 5337#endif
a4af51ce 5338 super_set_uuid(sb, es->s_uuid, sizeof(es->s_uuid));
fb092d40 5339 super_set_sysfs_name_bdev(sb);
f2fa2ffc 5340
ac27a0ec 5341 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 5342 mutex_init(&sbi->s_orphan_lock);
ac27a0ec 5343
f7314a67 5344 ext4_fast_commit_init(sb);
aa75f4d3 5345
ac27a0ec
DK		 5346 sb->s_root = NULL;
5347
5348 needs_recovery = (es->s_last_orphan != 0 ||
02f310fc 5349 ext4_has_feature_orphan_present(sb) ||
e2b911c5 5350 ext4_has_feature_journal_needs_recovery(sb));
ac27a0ec 5351
3b50d501
TT		 5352 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) {
5353 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block));
5354 if (err)
50460fe8 5355 goto failed_mount3a;
3b50d501 5356 }
c5e06d10 5357
d4fab7b2 5358 err = -EINVAL;
ac27a0ec
DK		 5359 /*
5360 * The first inode we look at is the journal inode. Don't try
5361 * root first: it may be modified in the journal!
5362 */
e2b911c5 5363 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
9c1dd22d 5364 err = ext4_load_and_init_journal(sb, es, ctx);
4753d8a2 5365 if (err)
50460fe8 5366 goto failed_mount3a;
bc98a42c 5367 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
e2b911c5 5368 ext4_has_feature_journal_needs_recovery(sb)) {
b31e1552
ES		 5369 ext4_msg(sb, KERN_ERR, "required journal recovery "
5370 "suppressed and not mounted read-only");
43bd6f1b 5371 goto failed_mount3a;
ac27a0ec 5372 } else {
1e381f60 5373 /* Nojournal mode, all journal mount options are illegal */
89481b5f 5374 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1e381f60 5375 ext4_msg(sb, KERN_ERR, "can't mount with "
89481b5f 5376 "journal_async_commit, fs mounted w/o journal");
43bd6f1b 5377 goto failed_mount3a;
1e381f60 5378 }
89481b5f
BL		 5379
5380 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
1e381f60 5381 ext4_msg(sb, KERN_ERR, "can't mount with "
89481b5f 5382 "journal_checksum, fs mounted w/o journal");
43bd6f1b 5383 goto failed_mount3a;
1e381f60
DM		 5384 }
5385 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5386 ext4_msg(sb, KERN_ERR, "can't mount with "
5387 "commit=%lu, fs mounted w/o journal",
5388 sbi->s_commit_interval / HZ);
43bd6f1b 5389 goto failed_mount3a;
1e381f60
DM		 5390 }
5391 if (EXT4_MOUNT_DATA_FLAGS &
5392 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5393 ext4_msg(sb, KERN_ERR, "can't mount with "
5394 "data=, fs mounted w/o journal");
43bd6f1b 5395 goto failed_mount3a;
1e381f60 5396 }
50b29d8f 5397 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
1e381f60 5398 clear_opt(sb, JOURNAL_CHECKSUM);
fd8c37ec 5399 clear_opt(sb, DATA_FLAGS);
995a3ed6 5400 clear_opt2(sb, JOURNAL_FAST_COMMIT);
0390131b
FM		 5401 sbi->s_journal = NULL;
5402 needs_recovery = 0;
ac27a0ec
DK		 5403 }
5404
cdb7ee4c
TE		 5405 if (!test_opt(sb, NO_MBCACHE)) {
5406 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5407 if (!sbi->s_ea_block_cache) {
dec214d0 5408 ext4_msg(sb, KERN_ERR,
cdb7ee4c 5409 "Failed to create ea_block_cache");
d4fab7b2 5410 err = -EINVAL;
dec214d0
TE		 5411 goto failed_mount_wq;
5412 }
cdb7ee4c
TE		 5413
5414 if (ext4_has_feature_ea_inode(sb)) {
5415 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5416 if (!sbi->s_ea_inode_cache) {
5417 ext4_msg(sb, KERN_ERR,
5418 "Failed to create ea_inode_cache");
d4fab7b2 5419 err = -EINVAL;
cdb7ee4c
TE		 5420 goto failed_mount_wq;
5421 }
5422 }
9c191f70
M		 5423 }
5424
952fc18e
TT		 5425 /*
5426 * Get the # of file system overhead blocks from the
5427 * superblock if present.
5428 */
85d825db
TT		 5429 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5430 /* ignore the precalculated value if it is ridiculous */
5431 if (sbi->s_overhead > ext4_blocks_count(es))
5432 sbi->s_overhead = 0;
5433 /*
5434 * If the bigalloc feature is not enabled recalculating the
5435 * overhead doesn't take long, so we might as well just redo
5436 * it to make sure we are using the correct value.
5437 */
5438 if (!ext4_has_feature_bigalloc(sb))
5439 sbi->s_overhead = 0;
5440 if (sbi->s_overhead == 0) {
07aa2ea1
LC		 5441 err = ext4_calculate_overhead(sb);
5442 if (err)
952fc18e
TT		 5443 goto failed_mount_wq;
5444 }
5445
fd89d5f2
TH		 5446 /*
5447 * The maximum number of concurrent works can be high and
5448 * concurrency isn't really necessary. Limit it to 1.
5449 */
2e8fa54e
JK		 5450 EXT4_SB(sb)->rsv_conversion_wq =
5451 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5452 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5453 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
d4fab7b2 5454 err = -ENOMEM;
2e8fa54e
JK		 5455 goto failed_mount4;
5456 }
5457
ac27a0ec 5458 /*
dab291af 5459 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 5460 * so we can safely mount the rest of the filesystem now.
5461 */
5462
8a363970 5463 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
1d1fe1ee 5464 if (IS_ERR(root)) {
b31e1552 5465 ext4_msg(sb, KERN_ERR, "get root inode failed");
d4fab7b2 5466 err = PTR_ERR(root);
32a9bb57 5467 root = NULL;
ac27a0ec
DK		 5468 goto failed_mount4;
5469 }
5470 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 5471 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 5472 iput(root);
d4fab7b2 5473 err = -EFSCORRUPTED;
ac27a0ec
DK		 5474 goto failed_mount4;
5475 }
b886ee3e 5476
04aa5f4e 5477 generic_set_sb_d_ops(sb);
48fde701 5478 sb->s_root = d_make_root(root);
1d1fe1ee 5479 if (!sb->s_root) {
b31e1552 5480 ext4_msg(sb, KERN_ERR, "get root dentry failed");
d4fab7b2 5481 err = -ENOMEM;
1d1fe1ee
DH		 5482 goto failed_mount4;
5483 }
ac27a0ec 5484
d4fab7b2
TT		 5485 err = ext4_setup_super(sb, es, sb_rdonly(sb));
5486 if (err == -EROFS) {
1751e8a6 5487 sb->s_flags |= SB_RDONLY;
d4fab7b2 5488 } else if (err)
c89128a0 5489 goto failed_mount4a;
ef7f3835 5490
b5799018 5491 ext4_set_resv_clusters(sb);
27dd4385 5492
0f5bde1d
JK		 5493 if (test_opt(sb, BLOCK_VALIDITY)) {
5494 err = ext4_setup_system_zone(sb);
5495 if (err) {
5496 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5497 "zone (%d)", err);
5498 goto failed_mount4a;
5499 }
f9ae9cf5 5500 }
8016e29f 5501 ext4_fc_replay_cleanup(sb);
f9ae9cf5
TT		 5502
5503 ext4_ext_init(sb);
196e402a
HS		 5504
5505 /*
5506 * Enable optimize_scan if number of groups is > threshold. This can be
5507 * turned off by passing "mb_optimize_scan=0". This can also be
5508 * turned on forcefully by passing "mb_optimize_scan=1".
5509 */
27b38686
OM		 5510 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5511 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5512 set_opt2(sb, MB_OPTIMIZE_SCAN);
5513 else
5514 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5515 }
196e402a 5516
f9ae9cf5
TT		 5517 err = ext4_mb_init(sb);
5518 if (err) {
5519 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5520 err);
dcf2d804 5521 goto failed_mount5;
c2774d84
AK		 5522 }
5523
027f14f5
TT		 5524 /*
5525 * We can only set up the journal commit callback once
5526 * mballoc is initialized
5527 */
5528 if (sbi->s_journal)
5529 sbi->s_journal->j_commit_callback =
5530 ext4_journal_commit_callback;
5531
d5e72c4e
TT		 5532 err = ext4_percpu_param_init(sbi);
5533 if (err)
d5e03cbb 5534 goto failed_mount6;
d5e03cbb 5535
e2b911c5 5536 if (ext4_has_feature_flex_bg(sb))
d5e03cbb
TT		 5537 if (!ext4_fill_flex_info(sb)) {
5538 ext4_msg(sb, KERN_ERR,
5539 "unable to initialize "
5540 "flex_bg meta info!");
d4fab7b2 5541 err = -ENOMEM;
d5e03cbb
TT		 5542 goto failed_mount6;
5543 }
5544
bfff6873
LC		 5545 err = ext4_register_li_request(sb, first_not_zeroed);
5546 if (err)
dcf2d804 5547 goto failed_mount6;
bfff6873 5548
02f310fc
JK		 5549 err = ext4_init_orphan_info(sb);
5550 if (err)
b4b4fda3 5551 goto failed_mount7;
9b2ff357
JK		 5552#ifdef CONFIG_QUOTA
5553 /* Enable quota usage during mount. */
bc98a42c 5554 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
9b2ff357
JK		 5555 err = ext4_enable_quotas(sb);
5556 if (err)
b4b4fda3 5557 goto failed_mount8;
9b2ff357
JK		 5558 }
5559#endif /* CONFIG_QUOTA */
5560
bc71726c 5561 /*
5562 * Save the original bdev mapping's wb_err value which could be
5563 * used to detect the metadata async write error.
5564 */
5565 spin_lock_init(&sbi->s_bdev_wb_lock);
9704a322
ZX		 5566 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5567 &sbi->s_bdev_wb_err);
617ba13b
MC		 5568 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5569 ext4_orphan_cleanup(sb, es);
5570 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
9b6641dd
YB		 5571 /*
5572 * Update the checksum after updating free space/inode counters and
5573 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5574 * checksum in the buffer cache until it is written out and
5575 * e2fsprogs programs trying to open a file system immediately
5576 * after it is mounted can fail.
5577 */
5578 ext4_superblock_csum_set(sb);
0390131b 5579 if (needs_recovery) {
b31e1552 5580 ext4_msg(sb, KERN_INFO, "recovery complete");
11215630
JK		 5581 err = ext4_mark_recovery_complete(sb, es);
5582 if (err)
b4b4fda3 5583 goto failed_mount9;
0390131b 5584 }
0390131b 5585
70200574
CH		 5586 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5587 ext4_msg(sb, KERN_WARNING,
5588 "mounting with \"discard\" option, but the device does not support discard");
79add3a3 5589
66e61a9e
TT		 5590 if (es->s_error_count)
5591 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 5592
efbed4dc
TT		 5593 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5594 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5595 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5596 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
1cf006ed
DM		 5597 atomic_set(&sbi->s_warning_count, 0);
5598 atomic_set(&sbi->s_msg_count, 0);
efbed4dc 5599
b4b4fda3
BL		 5600 /* Register sysfs after all initializations are complete. */
5601 err = ext4_register_sysfs(sb);
5602 if (err)
5603 goto failed_mount9;
5604
ac27a0ec
DK		 5605 return 0;
5606
b4b4fda3 5607failed_mount9:
f3c1c42e 5608 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
b4b4fda3 5609failed_mount8: __maybe_unused
02f310fc 5610 ext4_release_orphan_info(sb);
dcf2d804
TM		 5611failed_mount7:
5612 ext4_unregister_li_request(sb);
5613failed_mount6:
f9ae9cf5 5614 ext4_mb_release(sb);
dcbf8758 5615 ext4_flex_groups_free(sbi);
1f79467c 5616 ext4_percpu_param_destroy(sbi);
00764937 5617failed_mount5:
f9ae9cf5
TT		 5618 ext4_ext_release(sb);
5619 ext4_release_system_zone(sb);
5620failed_mount4a:
94bf608a 5621 dput(sb->s_root);
32a9bb57 5622 sb->s_root = NULL;
94bf608a 5623failed_mount4:
b31e1552 5624 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 5625 if (EXT4_SB(sb)->rsv_conversion_wq)
5626 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 5627failed_mount_wq:
50c15df6
CX		 5628 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5629 sbi->s_ea_inode_cache = NULL;
5630
5631 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5632 sbi->s_ea_block_cache = NULL;
5633
0390131b 5634 if (sbi->s_journal) {
bb15cea2
TT		 5635 /* flush s_sb_upd_work before journal destroy. */
5636 flush_work(&sbi->s_sb_upd_work);
0390131b
FM		 5637 jbd2_journal_destroy(sbi->s_journal);
5638 sbi->s_journal = NULL;
5639 }
50460fe8 5640failed_mount3a:
d3922a77 5641 ext4_es_unregister_shrinker(sbi);
eb68d0e2 5642failed_mount3:
bb15cea2
TT		 5643 /* flush s_sb_upd_work before sbi destroy */
5644 flush_work(&sbi->s_sb_upd_work);
2a4ae3bc 5645 del_timer_sync(&sbi->s_err_report);
618f0031 5646 ext4_stop_mmpd(sbi);
a4e6a511 5647 ext4_group_desc_free(sbi);
ac27a0ec 5648failed_mount:
0441984a
DW		 5649 if (sbi->s_chksum_driver)
5650 crypto_free_shash(sbi->s_chksum_driver);
c83ad55e 5651
5298d4bf 5652#if IS_ENABLED(CONFIG_UNICODE)
f8f4acb6 5653 utf8_unload(sb->s_encoding);
c83ad55e
GKB		 5654#endif
5655
ac27a0ec 5656#ifdef CONFIG_QUOTA
856dd6c5 5657 for (unsigned int i = 0; i < EXT4_MAXQUOTAS; i++)
0ba33fac 5658 kfree(get_qf_name(sb, sbi, i));
ac27a0ec 5659#endif
ac4acb1f 5660 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
a7a79c29 5661 brelse(sbi->s_sbh);
61ead714
CB		 5662 if (sbi->s_journal_bdev_file) {
5663 invalidate_bdev(file_bdev(sbi->s_journal_bdev_file));
22650a99 5664 bdev_fput(sbi->s_journal_bdev_file);
1489dffd 5665 }
ac27a0ec 5666out_fail:
26fb5290 5667 invalidate_bdev(sb->s_bdev);
ac27a0ec 5668 sb->s_fs_info = NULL;
d4fab7b2 5669 return err;
ac27a0ec
DK		 5670}
5671
cebe85d5 5672static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
7edfd85b 5673{
cebe85d5 5674 struct ext4_fs_context *ctx = fc->fs_private;
7edfd85b 5675 struct ext4_sb_info *sbi;
7edfd85b 5676 const char *descr;
cebe85d5 5677 int ret;
7edfd85b
LC		 5678
5679 sbi = ext4_alloc_sbi(sb);
cebe85d5 5680 if (!sbi)
7c268d4c 5681 return -ENOMEM;
7edfd85b 5682
cebe85d5
LC		 5683 fc->s_fs_info = sbi;
5684
5685 /* Cleanup superblock name */
5686 strreplace(sb->s_id, '/', '!');
7edfd85b
LC		 5687
5688 sbi->s_sb_block = 1; /* Default super block location */
cebe85d5
LC		 5689 if (ctx->spec & EXT4_SPEC_s_sb_block)
5690 sbi->s_sb_block = ctx->s_sb_block;
7edfd85b 5691
960e0ab6 5692 ret = __ext4_fill_super(fc, sb);
7edfd85b
LC		 5693 if (ret < 0)
5694 goto free_sbi;
5695
cebe85d5 5696 if (sbi->s_journal) {
7edfd85b
LC		 5697 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5698 descr = " journalled data mode";
5699 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5700 descr = " ordered data mode";
5701 else
5702 descr = " writeback data mode";
5703 } else
5704 descr = "out journal";
5705
5706 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
6dcc98fb
TT		 5707 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU %s with%s. "
5708 "Quota mode: %s.", &sb->s_uuid,
5709 sb_rdonly(sb) ? "ro" : "r/w", descr,
bb0fbc78 5710 ext4_quota_mode(sb));
cebe85d5 5711
eb705421 5712 /* Update the s_overhead_clusters if necessary */
827891a3 5713 ext4_update_overhead(sb, false);
7edfd85b 5714 return 0;
cebe85d5 5715
7edfd85b
LC		 5716free_sbi:
5717 ext4_free_sbi(sbi);
cebe85d5 5718 fc->s_fs_info = NULL;
7edfd85b
LC		 5719 return ret;
5720}
5721
cebe85d5
LC		 5722static int ext4_get_tree(struct fs_context *fc)
5723{
5724 return get_tree_bdev(fc, ext4_fill_super);
5725}
5726
ac27a0ec
DK		 5727/*
5728 * Setup any per-fs journal parameters now. We'll do this both on
5729 * initial mount, once the journal has been initialised but before we've
5730 * done any recovery; and again on any subsequent remount.
5731 */
617ba13b 5732static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 5733{
617ba13b 5734 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 5735
30773840
TT		 5736 journal->j_commit_interval = sbi->s_commit_interval;
5737 journal->j_min_batch_time = sbi->s_min_batch_time;
5738 journal->j_max_batch_time = sbi->s_max_batch_time;
6866d7b3 5739 ext4_fc_init(sb, journal);
ac27a0ec 5740
a931da6a 5741 write_lock(&journal->j_state_lock);
ac27a0ec 5742 if (test_opt(sb, BARRIER))
dab291af 5743 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 5744 else
dab291af 5745 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 5746 if (test_opt(sb, DATA_ERR_ABORT))
5747 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5748 else
5749 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
72945058
ZY		 5750 /*
5751 * Always enable journal cycle record option, letting the journal
5752 * records log transactions continuously between each mount.
5753 */
5754 journal->j_flags |= JBD2_CYCLE_RECORD;
a931da6a 5755 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 5756}
5757
c6cb7e77
EW		 5758static struct inode *ext4_get_journal_inode(struct super_block *sb,
5759 unsigned int journal_inum)
ac27a0ec
DK		 5760{
5761 struct inode *journal_inode;
ac27a0ec 5762
c6cb7e77
EW		 5763 /*
5764 * Test for the existence of a valid inode on disk. Bad things
5765 * happen if we iget() an unused inode, as the subsequent iput()
5766 * will try to delete it.
5767 */
8a363970 5768 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
1d1fe1ee 5769 if (IS_ERR(journal_inode)) {
b31e1552 5770 ext4_msg(sb, KERN_ERR, "no journal found");
ee5c8071 5771 return ERR_CAST(journal_inode);
ac27a0ec
DK		 5772 }
5773 if (!journal_inode->i_nlink) {
5774 make_bad_inode(journal_inode);
5775 iput(journal_inode);
b31e1552 5776 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ee5c8071 5777 return ERR_PTR(-EFSCORRUPTED);
ac27a0ec 5778 }
105c78e1 5779 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
b31e1552 5780 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec 5781 iput(journal_inode);
ee5c8071 5782 return ERR_PTR(-EFSCORRUPTED);
ac27a0ec 5783 }
bc74e6a3
ZY		 5784
5785 ext4_debug("Journal inode found at %p: %lld bytes\n",
5786 journal_inode, journal_inode->i_size);
c6cb7e77
EW		 5787 return journal_inode;
5788}
5789
62913ae9
TT		 5790static int ext4_journal_bmap(journal_t *journal, sector_t *block)
5791{
5792 struct ext4_map_blocks map;
5793 int ret;
5794
5795 if (journal->j_inode == NULL)
5796 return 0;
5797
5798 map.m_lblk = *block;
5799 map.m_len = 1;
5800 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0);
5801 if (ret <= 0) {
5802 ext4_msg(journal->j_inode->i_sb, KERN_CRIT,
5803 "journal bmap failed: block %llu ret %d\n",
5804 *block, ret);
5805 jbd2_journal_abort(journal, ret ? ret : -EIO);
5806 return ret;
5807 }
5808 *block = map.m_pblk;
5809 return 0;
5810}
5811
ee5c8071
ZY		 5812static journal_t *ext4_open_inode_journal(struct super_block *sb,
5813 unsigned int journal_inum)
c6cb7e77
EW		 5814{
5815 struct inode *journal_inode;
5816 journal_t *journal;
5817
c6cb7e77 5818 journal_inode = ext4_get_journal_inode(sb, journal_inum);
ee5c8071
ZY		 5819 if (IS_ERR(journal_inode))
5820 return ERR_CAST(journal_inode);
ac27a0ec 5821
dab291af 5822 journal = jbd2_journal_init_inode(journal_inode);
8e6cf5fb 5823 if (IS_ERR(journal)) {
b31e1552 5824 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec 5825 iput(journal_inode);
ee5c8071 5826 return ERR_CAST(journal);
ac27a0ec
DK		 5827 }
5828 journal->j_private = sb;
62913ae9 5829 journal->j_bmap = ext4_journal_bmap;
617ba13b 5830 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 5831 return journal;
5832}
5833
61ead714 5834static struct file *ext4_get_journal_blkdev(struct super_block *sb,
bc74e6a3
ZY		 5835 dev_t j_dev, ext4_fsblk_t *j_start,
5836 ext4_fsblk_t *j_len)
ac27a0ec 5837{
2b2d6d01 5838 struct buffer_head *bh;
bc74e6a3 5839 struct block_device *bdev;
61ead714 5840 struct file *bdev_file;
ac27a0ec 5841 int hblock, blocksize;
617ba13b 5842 ext4_fsblk_t sb_block;
ac27a0ec 5843 unsigned long offset;
2b2d6d01 5844 struct ext4_super_block *es;
ee5c8071 5845 int errno;
0390131b 5846
61ead714 5847 bdev_file = bdev_file_open_by_dev(j_dev,
afde134b
JK		 5848 BLK_OPEN_READ | BLK_OPEN_WRITE | BLK_OPEN_RESTRICT_WRITES,
5849 sb, &fs_holder_ops);
61ead714 5850 if (IS_ERR(bdev_file)) {
bc74e6a3
ZY		 5851 ext4_msg(sb, KERN_ERR,
5852 "failed to open journal device unknown-block(%u,%u) %ld",
61ead714
CB		 5853 MAJOR(j_dev), MINOR(j_dev), PTR_ERR(bdev_file));
5854 return bdev_file;
bc74e6a3 5855 }
ac27a0ec 5856
61ead714 5857 bdev = file_bdev(bdev_file);
ac27a0ec 5858 blocksize = sb->s_blocksize;
e1defc4f 5859 hblock = bdev_logical_block_size(bdev);
ac27a0ec 5860 if (blocksize < hblock) {
b31e1552
ES		 5861 ext4_msg(sb, KERN_ERR,
5862 "blocksize too small for journal device");
ee5c8071 5863 errno = -EINVAL;
ac27a0ec
DK		 5864 goto out_bdev;
5865 }
5866
617ba13b
MC		 5867 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5868 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec 5869 set_blocksize(bdev, blocksize);
bc74e6a3
ZY		 5870 bh = __bread(bdev, sb_block, blocksize);
5871 if (!bh) {
b31e1552
ES		 5872 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5873 "external journal");
ee5c8071 5874 errno = -EINVAL;
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)) {
bc74e6a3 5882 ext4_msg(sb, KERN_ERR, "external journal has bad superblock");
ee5c8071 5883 errno = -EFSCORRUPTED;
bc74e6a3 5884 goto out_bh;
ac27a0ec
DK		 5885 }
5886
df4763be
DW		 5887 if ((le32_to_cpu(es->s_feature_ro_compat) &
5888 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5889 es->s_checksum != ext4_superblock_csum(sb, es)) {
bc74e6a3 5890 ext4_msg(sb, KERN_ERR, "external journal has corrupt superblock");
ee5c8071 5891 errno = -EFSCORRUPTED;
bc74e6a3 5892 goto out_bh;
df4763be
DW		 5893 }
5894
617ba13b 5895 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 5896 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ee5c8071 5897 errno = -EFSCORRUPTED;
bc74e6a3 5898 goto out_bh;
ac27a0ec
DK		 5899 }
5900
bc74e6a3
ZY		 5901 *j_start = sb_block + 1;
5902 *j_len = ext4_blocks_count(es);
5903 brelse(bh);
61ead714 5904 return bdev_file;
ac27a0ec 5905
bc74e6a3
ZY		 5906out_bh:
5907 brelse(bh);
5908out_bdev:
22650a99 5909 bdev_fput(bdev_file);
ee5c8071 5910 return ERR_PTR(errno);
bc74e6a3
ZY		 5911}
5912
ee5c8071
ZY		 5913static journal_t *ext4_open_dev_journal(struct super_block *sb,
5914 dev_t j_dev)
bc74e6a3
ZY		 5915{
5916 journal_t *journal;
5917 ext4_fsblk_t j_start;
5918 ext4_fsblk_t j_len;
61ead714 5919 struct file *bdev_file;
ee5c8071 5920 int errno = 0;
bc74e6a3 5921
61ead714
CB		 5922 bdev_file = ext4_get_journal_blkdev(sb, j_dev, &j_start, &j_len);
5923 if (IS_ERR(bdev_file))
5924 return ERR_CAST(bdev_file);
bc74e6a3 5925
61ead714 5926 journal = jbd2_journal_init_dev(file_bdev(bdev_file), sb->s_bdev, j_start,
bc74e6a3 5927 j_len, sb->s_blocksize);
8e6cf5fb 5928 if (IS_ERR(journal)) {
b31e1552 5929 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ee5c8071 5930 errno = PTR_ERR(journal);
ac27a0ec
DK		 5931 goto out_bdev;
5932 }
ac27a0ec 5933 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 5934 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5935 "user (unsupported) - %d",
ac27a0ec 5936 be32_to_cpu(journal->j_superblock->s_nr_users));
ee5c8071 5937 errno = -EINVAL;
ac27a0ec
DK		 5938 goto out_journal;
5939 }
bc74e6a3 5940 journal->j_private = sb;
61ead714 5941 EXT4_SB(sb)->s_journal_bdev_file = bdev_file;
617ba13b 5942 ext4_init_journal_params(sb, journal);
ac27a0ec 5943 return journal;
0b8e58a1 5944
ac27a0ec 5945out_journal:
dab291af 5946 jbd2_journal_destroy(journal);
ac27a0ec 5947out_bdev:
22650a99 5948 bdev_fput(bdev_file);
ee5c8071 5949 return ERR_PTR(errno);
ac27a0ec
DK		 5950}
5951
617ba13b
MC		 5952static int ext4_load_journal(struct super_block *sb,
5953 struct ext4_super_block *es,
ac27a0ec
DK		 5954 unsigned long journal_devnum)
5955{
5956 journal_t *journal;
5957 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5958 dev_t journal_dev;
5959 int err = 0;
5960 int really_read_only;
273108fa 5961 int journal_dev_ro;
ac27a0ec 5962
11215630
JK		 5963 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5964 return -EFSCORRUPTED;
0390131b 5965
ac27a0ec
DK		 5966 if (journal_devnum &&
5967 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 5968 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5969 "numbers have changed");
ac27a0ec
DK		 5970 journal_dev = new_decode_dev(journal_devnum);
5971 } else
5972 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5973
273108fa
LC		 5974 if (journal_inum && journal_dev) {
5975 ext4_msg(sb, KERN_ERR,
5976 "filesystem has both journal inode and journal device!");
5977 return -EINVAL;
5978 }
5979
5980 if (journal_inum) {
ee5c8071
ZY		 5981 journal = ext4_open_inode_journal(sb, journal_inum);
5982 if (IS_ERR(journal))
5983 return PTR_ERR(journal);
273108fa 5984 } else {
ee5c8071
ZY		 5985 journal = ext4_open_dev_journal(sb, journal_dev);
5986 if (IS_ERR(journal))
5987 return PTR_ERR(journal);
273108fa
LC		 5988 }
5989
5990 journal_dev_ro = bdev_read_only(journal->j_dev);
5991 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5992
5993 if (journal_dev_ro && !sb_rdonly(sb)) {
5994 ext4_msg(sb, KERN_ERR,
5995 "journal device read-only, try mounting with '-o ro'");
5996 err = -EROFS;
5997 goto err_out;
5998 }
ac27a0ec
DK		 5999
6000 /*
6001 * Are we loading a blank journal or performing recovery after a
6002 * crash? For recovery, we need to check in advance whether we
6003 * can get read-write access to the device.
6004 */
e2b911c5 6005 if (ext4_has_feature_journal_needs_recovery(sb)) {
bc98a42c 6006 if (sb_rdonly(sb)) {
b31e1552
ES		 6007 ext4_msg(sb, KERN_INFO, "INFO: recovery "
6008 "required on readonly filesystem");
ac27a0ec 6009 if (really_read_only) {
b31e1552 6010 ext4_msg(sb, KERN_ERR, "write access "
d98bf8cd
SR		 6011 "unavailable, cannot proceed "
6012 "(try mounting with noload)");
273108fa
LC		 6013 err = -EROFS;
6014 goto err_out;
ac27a0ec 6015 }
b31e1552
ES		 6016 ext4_msg(sb, KERN_INFO, "write access will "
6017 "be enabled during recovery");
ac27a0ec
DK		 6018 }
6019 }
6020
90576c0b 6021 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 6022 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 6023
e2b911c5 6024 if (!ext4_has_feature_journal_needs_recovery(sb))
dab291af 6025 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 6026 if (!err) {
6027 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
2ef6c32a
TT		 6028 __le16 orig_state;
6029 bool changed = false;
eee00237 6030
1c13d5c0
TT		 6031 if (save)
6032 memcpy(save, ((char *) es) +
6033 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 6034 err = jbd2_journal_load(journal);
2ef6c32a
TT		 6035 if (save && memcmp(((char *) es) + EXT4_S_ERR_START,
6036 save, EXT4_S_ERR_LEN)) {
1c13d5c0
TT		 6037 memcpy(((char *) es) + EXT4_S_ERR_START,
6038 save, EXT4_S_ERR_LEN);
2ef6c32a
TT		 6039 changed = true;
6040 }
1c13d5c0 6041 kfree(save);
2ef6c32a 6042 orig_state = es->s_state;
eee00237
YB		 6043 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state &
6044 EXT4_ERROR_FS);
2ef6c32a
TT		 6045 if (orig_state != es->s_state)
6046 changed = true;
eee00237 6047 /* Write out restored error information to the superblock */
2ef6c32a 6048 if (changed && !really_read_only) {
eee00237
YB		 6049 int err2;
6050 err2 = ext4_commit_super(sb);
6051 err = err ? : err2;
6052 }
1c13d5c0 6053 }
ac27a0ec
DK		 6054
6055 if (err) {
b31e1552 6056 ext4_msg(sb, KERN_ERR, "error loading journal");
273108fa 6057 goto err_out;
ac27a0ec
DK		 6058 }
6059
617ba13b 6060 EXT4_SB(sb)->s_journal = journal;
11215630
JK		 6061 err = ext4_clear_journal_err(sb, es);
6062 if (err) {
6063 EXT4_SB(sb)->s_journal = NULL;
6064 jbd2_journal_destroy(journal);
6065 return err;
6066 }
ac27a0ec 6067
c41303ce 6068 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 6069 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
6070 es->s_journal_dev = cpu_to_le32(journal_devnum);
3039d8b8
BL		 6071 ext4_commit_super(sb);
6072 }
6073 if (!really_read_only && journal_inum &&
6074 journal_inum != le32_to_cpu(es->s_journal_inum)) {
6075 es->s_journal_inum = cpu_to_le32(journal_inum);
4392fbc4 6076 ext4_commit_super(sb);
ac27a0ec
DK		 6077 }
6078
6079 return 0;
273108fa
LC		 6080
6081err_out:
6082 jbd2_journal_destroy(journal);
6083 return err;
ac27a0ec
DK		 6084}
6085
2d01ddc8
JK		 6086/* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
6087static void ext4_update_super(struct super_block *sb)
ac27a0ec 6088{
c92dc856 6089 struct ext4_sb_info *sbi = EXT4_SB(sb);
e92ad03f
JK		 6090 struct ext4_super_block *es = sbi->s_es;
6091 struct buffer_head *sbh = sbi->s_sbh;
a17712c8 6092
05c2c00f 6093 lock_buffer(sbh);
71290b36
TT		 6094 /*
6095 * If the file system is mounted read-only, don't update the
6096 * superblock write time. This avoids updating the superblock
6097 * write time when we are mounting the root file system
6098 * read/only but we need to replay the journal; at that point,
6099 * for people who are east of GMT and who make their clock
6100 * tick in localtime for Windows bug-for-bug compatibility,
6101 * the clock is set in the future, and this will cause e2fsck
6102 * to complain and force a full file system check.
6103 */
d5d020b3 6104 if (!sb_rdonly(sb))
6a0678a7 6105 ext4_update_tstamp(es, s_wtime);
8446fe92 6106 es->s_kbytes_written =
0bc9bc1d 6107 cpu_to_le64(sbi->s_kbytes_written +
8446fe92 6108 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
0bc9bc1d 6109 sbi->s_sectors_written_start) >> 1));
e92ad03f 6110 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
d5e03cbb 6111 ext4_free_blocks_count_set(es,
e92ad03f
JK		 6112 EXT4_C2B(sbi, percpu_counter_sum_positive(
6113 &sbi->s_freeclusters_counter)));
6114 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
d5e03cbb
TT		 6115 es->s_free_inodes_count =
6116 cpu_to_le32(percpu_counter_sum_positive(
e92ad03f 6117 &sbi->s_freeinodes_counter));
c92dc856
JK		 6118 /* Copy error information to the on-disk superblock */
6119 spin_lock(&sbi->s_error_lock);
6120 if (sbi->s_add_error_count > 0) {
6121 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6122 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6123 __ext4_update_tstamp(&es->s_first_error_time,
6124 &es->s_first_error_time_hi,
6125 sbi->s_first_error_time);
744a5638
JS		 6126 strtomem_pad(es->s_first_error_func,
6127 sbi->s_first_error_func, 0);
c92dc856
JK		 6128 es->s_first_error_line =
6129 cpu_to_le32(sbi->s_first_error_line);
6130 es->s_first_error_ino =
6131 cpu_to_le32(sbi->s_first_error_ino);
6132 es->s_first_error_block =
6133 cpu_to_le64(sbi->s_first_error_block);
6134 es->s_first_error_errcode =
6135 ext4_errno_to_code(sbi->s_first_error_code);
6136 }
6137 __ext4_update_tstamp(&es->s_last_error_time,
6138 &es->s_last_error_time_hi,
6139 sbi->s_last_error_time);
744a5638 6140 strtomem_pad(es->s_last_error_func, sbi->s_last_error_func, 0);
c92dc856
JK		 6141 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6142 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6143 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6144 es->s_last_error_errcode =
6145 ext4_errno_to_code(sbi->s_last_error_code);
6146 /*
6147 * Start the daily error reporting function if it hasn't been
6148 * started already
6149 */
6150 if (!es->s_error_count)
6151 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6152 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6153 sbi->s_add_error_count = 0;
6154 }
6155 spin_unlock(&sbi->s_error_lock);
6156
06db49e6 6157 ext4_superblock_csum_set(sb);
2d01ddc8
JK		 6158 unlock_buffer(sbh);
6159}
6160
6161static int ext4_commit_super(struct super_block *sb)
6162{
6163 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
2d01ddc8 6164
f88f1466
FC		 6165 if (!sbh)
6166 return -EINVAL;
6167 if (block_device_ejected(sb))
6168 return -ENODEV;
2d01ddc8
JK		 6169
6170 ext4_update_super(sb);
6171
15baa7dc
ZY		 6172 lock_buffer(sbh);
6173 /* Buffer got discarded which means block device got invalidated */
6174 if (!buffer_mapped(sbh)) {
6175 unlock_buffer(sbh);
6176 return -EIO;
6177 }
6178
e8680786 6179 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
4743f839
PKS		 6180 /*
6181 * Oh, dear. A previous attempt to write the
6182 * superblock failed. This could happen because the
6183 * USB device was yanked out. Or it could happen to
6184 * be a transient write error and maybe the block will
6185 * be remapped. Nothing we can do but to retry the
6186 * write and hope for the best.
6187 */
6188 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6189 "superblock detected");
6190 clear_buffer_write_io_error(sbh);
6191 set_buffer_uptodate(sbh);
6192 }
15baa7dc
ZY		 6193 get_bh(sbh);
6194 /* Clear potential dirty bit if it was journalled update */
6195 clear_buffer_dirty(sbh);
6196 sbh->b_end_io = end_buffer_write_sync;
1420c4a5
BVA		 6197 submit_bh(REQ_OP_WRITE | REQ_SYNC |
6198 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
15baa7dc 6199 wait_on_buffer(sbh);
4392fbc4
JK		 6200 if (buffer_write_io_error(sbh)) {
6201 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6202 "superblock");
6203 clear_buffer_write_io_error(sbh);
6204 set_buffer_uptodate(sbh);
15baa7dc 6205 return -EIO;
914258bf 6206 }
15baa7dc 6207 return 0;
ac27a0ec
DK		 6208}
6209
ac27a0ec
DK		 6210/*
6211 * Have we just finished recovery? If so, and if we are mounting (or
6212 * remounting) the filesystem readonly, then we will end up with a
6213 * consistent fs on disk. Record that fact.
6214 */
11215630
JK		 6215static int ext4_mark_recovery_complete(struct super_block *sb,
6216 struct ext4_super_block *es)
ac27a0ec 6217{
11215630 6218 int err;
617ba13b 6219 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 6220
e2b911c5 6221 if (!ext4_has_feature_journal(sb)) {
11215630
JK		 6222 if (journal != NULL) {
6223 ext4_error(sb, "Journal got removed while the fs was "
6224 "mounted!");
6225 return -EFSCORRUPTED;
6226 }
6227 return 0;
0390131b 6228 }
dab291af 6229 jbd2_journal_lock_updates(journal);
01d5d965 6230 err = jbd2_journal_flush(journal, 0);
11215630 6231 if (err < 0)
7ffe1ea8
HK		 6232 goto out;
6233
02f310fc
JK		 6234 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6235 ext4_has_feature_orphan_present(sb))) {
6236 if (!ext4_orphan_file_empty(sb)) {
6237 ext4_error(sb, "Orphan file not empty on read-only fs.");
6238 err = -EFSCORRUPTED;
6239 goto out;
6240 }
e2b911c5 6241 ext4_clear_feature_journal_needs_recovery(sb);
02f310fc 6242 ext4_clear_feature_orphan_present(sb);
4392fbc4 6243 ext4_commit_super(sb);
ac27a0ec 6244 }
7ffe1ea8 6245out:
dab291af 6246 jbd2_journal_unlock_updates(journal);
11215630 6247 return err;
ac27a0ec
DK		 6248}
6249
6250/*
6251 * If we are mounting (or read-write remounting) a filesystem whose journal
6252 * has recorded an error from a previous lifetime, move that error to the
6253 * main filesystem now.
6254 */
11215630 6255static int ext4_clear_journal_err(struct super_block *sb,
2b2d6d01 6256 struct ext4_super_block *es)
ac27a0ec
DK		 6257{
6258 journal_t *journal;
6259 int j_errno;
6260 const char *errstr;
6261
11215630
JK		 6262 if (!ext4_has_feature_journal(sb)) {
6263 ext4_error(sb, "Journal got removed while the fs was mounted!");
6264 return -EFSCORRUPTED;
6265 }
0390131b 6266
617ba13b 6267 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 6268
6269 /*
6270 * Now check for any error status which may have been recorded in the
617ba13b 6271 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 6272 */
6273
dab291af 6274 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 6275 if (j_errno) {
6276 char nbuf[16];
6277
617ba13b 6278 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 6279 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 6280 "from previous mount: %s", errstr);
ac27a0ec 6281
617ba13b
MC		 6282 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6283 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
f57886ca
YB		 6284 j_errno = ext4_commit_super(sb);
6285 if (j_errno)
6286 return j_errno;
6287 ext4_warning(sb, "Marked fs in need of filesystem check.");
ac27a0ec 6288
dab291af 6289 jbd2_journal_clear_err(journal);
d796c52e 6290 jbd2_journal_update_sb_errno(journal);
ac27a0ec 6291 }
11215630 6292 return 0;
ac27a0ec
DK		 6293}
6294
6295/*
6296 * Force the running and committing transactions to commit,
6297 * and wait on the commit.
6298 */
617ba13b 6299int ext4_force_commit(struct super_block *sb)
ac27a0ec 6300{
889860e4 6301 return ext4_journal_force_commit(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 6302}
6303
617ba13b 6304static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 6305{
14ce0cb4 6306 int ret = 0;
9eddacf9 6307 tid_t target;
06a407f1 6308 bool needs_barrier = false;
8d5d02e6 6309 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 6310
eb8ab444 6311 if (unlikely(ext4_forced_shutdown(sb)))
0db1ff22
TT		 6312 return 0;
6313
9bffad1e 6314 trace_ext4_sync_fs(sb, wait);
2e8fa54e 6315 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 6316 /*
6317 * Writeback quota in non-journalled quota case - journalled quota has
6318 * no dirty dquots
6319 */
6320 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 6321 /*
6322 * Data writeback is possible w/o journal transaction, so barrier must
6323 * being sent at the end of the function. But we can skip it if
6324 * transaction_commit will do it for us.
6325 */
bda32530
TT		 6326 if (sbi->s_journal) {
6327 target = jbd2_get_latest_transaction(sbi->s_journal);
6328 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6329 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6330 needs_barrier = true;
6331
6332 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6333 if (wait)
6334 ret = jbd2_log_wait_commit(sbi->s_journal,
6335 target);
6336 }
6337 } else if (wait && test_opt(sb, BARRIER))
06a407f1 6338 needs_barrier = true;
06a407f1
DM		 6339 if (needs_barrier) {
6340 int err;
c6bf3f0e 6341 err = blkdev_issue_flush(sb->s_bdev);
06a407f1
DM		 6342 if (!ret)
6343 ret = err;
0390131b 6344 }
06a407f1
DM		 6345
6346 return ret;
6347}
6348
ac27a0ec
DK		 6349/*
6350 * LVM calls this function before a (read-only) snapshot is created. This
6351 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 6352 *
6353 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 6354 * state independently. It relies on upper layer to stop all data & metadata
6355 * modifications.
ac27a0ec 6356 */
c4be0c1d 6357static int ext4_freeze(struct super_block *sb)
ac27a0ec 6358{
c4be0c1d 6359 int error = 0;
98175720 6360 journal_t *journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 6361
bb044576
TT		 6362 if (journal) {
6363 /* Now we set up the journal barrier. */
6364 jbd2_journal_lock_updates(journal);
ac27a0ec 6365
bb044576
TT		 6366 /*
6367 * Don't clear the needs_recovery flag if we failed to
6368 * flush the journal.
6369 */
01d5d965 6370 error = jbd2_journal_flush(journal, 0);
bb044576
TT		 6371 if (error < 0)
6372 goto out;
c642dc9e
ES		 6373
6374 /* Journal blocked and flushed, clear needs_recovery flag. */
e2b911c5 6375 ext4_clear_feature_journal_needs_recovery(sb);
02f310fc
JK		 6376 if (ext4_orphan_file_empty(sb))
6377 ext4_clear_feature_orphan_present(sb);
bb044576 6378 }
9ca92389 6379
4392fbc4 6380 error = ext4_commit_super(sb);
6b0310fb 6381out:
bb044576
TT		 6382 if (journal)
6383 /* we rely on upper layer to stop further updates */
6384 jbd2_journal_unlock_updates(journal);
6b0310fb 6385 return error;
ac27a0ec
DK		 6386}
6387
6388/*
6389 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6390 * flag here, even though the filesystem is not technically dirty yet.
6391 */
c4be0c1d 6392static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 6393{
eb8ab444 6394 if (ext4_forced_shutdown(sb))
9ca92389
TT		 6395 return 0;
6396
c642dc9e
ES		 6397 if (EXT4_SB(sb)->s_journal) {
6398 /* Reset the needs_recovery flag before the fs is unlocked. */
e2b911c5 6399 ext4_set_feature_journal_needs_recovery(sb);
02f310fc
JK		 6400 if (ext4_has_feature_orphan_file(sb))
6401 ext4_set_feature_orphan_present(sb);
c642dc9e
ES		 6402 }
6403
4392fbc4 6404 ext4_commit_super(sb);
c4be0c1d 6405 return 0;
ac27a0ec
DK		 6406}
6407
673c6100
TT		 6408/*
6409 * Structure to save mount options for ext4_remount's benefit
6410 */
6411struct ext4_mount_options {
6412 unsigned long s_mount_opt;
a2595b8a 6413 unsigned long s_mount_opt2;
08cefc7a
EB		 6414 kuid_t s_resuid;
6415 kgid_t s_resgid;
673c6100
TT		 6416 unsigned long s_commit_interval;
6417 u32 s_min_batch_time, s_max_batch_time;
6418#ifdef CONFIG_QUOTA
6419 int s_jquota_fmt;
a2d4a646 6420 char *s_qf_names[EXT4_MAXQUOTAS];
673c6100
TT		 6421#endif
6422};
6423
960e0ab6 6424static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
ac27a0ec 6425{
7edfd85b 6426 struct ext4_fs_context *ctx = fc->fs_private;
2b2d6d01 6427 struct ext4_super_block *es;
617ba13b 6428 struct ext4_sb_info *sbi = EXT4_SB(sb);
960e0ab6 6429 unsigned long old_sb_flags;
617ba13b 6430 struct ext4_mount_options old_opts;
8a266467 6431 ext4_group_t g;
c5e06d10 6432 int err = 0;
745f17a4 6433 int alloc_ctx;
ac27a0ec 6434#ifdef CONFIG_QUOTA
3bbef91b 6435 int enable_quota = 0;
03dafb5f 6436 int i, j;
33458eab 6437 char *to_free[EXT4_MAXQUOTAS];
ac27a0ec
DK		 6438#endif
6439
21ac738e 6440
ac27a0ec
DK		 6441 /* Store the original options */
6442 old_sb_flags = sb->s_flags;
6443 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 6444 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 6445 old_opts.s_resuid = sbi->s_resuid;
6446 old_opts.s_resgid = sbi->s_resgid;
6447 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 6448 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6449 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 6450#ifdef CONFIG_QUOTA
6451 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
a2d4a646 6452 for (i = 0; i < EXT4_MAXQUOTAS; i++)
03dafb5f 6453 if (sbi->s_qf_names[i]) {
33458eab
TT		 6454 char *qf_name = get_qf_name(sb, sbi, i);
6455
6456 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
03dafb5f
CG		 6457 if (!old_opts.s_qf_names[i]) {
6458 for (j = 0; j < i; j++)
6459 kfree(old_opts.s_qf_names[j]);
6460 return -ENOMEM;
6461 }
6462 } else
6463 old_opts.s_qf_names[i] = NULL;
ac27a0ec 6464#endif
e4e58e5d
OM		 6465 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6466 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6467 ctx->journal_ioprio =
6468 sbi->s_journal->j_task->io_context->ioprio;
6469 else
6470 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6471
6472 }
ac27a0ec 6473
745f17a4
BL		 6474 /*
6475 * Changing the DIOREAD_NOLOCK or DELALLOC mount options may cause
6476 * two calls to ext4_should_dioread_nolock() to return inconsistent
6477 * values, triggering WARN_ON in ext4_add_complete_io(). we grab
6478 * here s_writepages_rwsem to avoid race between writepages ops and
6479 * remount.
6480 */
6481 alloc_ctx = ext4_writepages_down_write(sb);
7edfd85b 6482 ext4_apply_options(fc, sb);
745f17a4 6483 ext4_writepages_up_write(sb, alloc_ctx);
ac27a0ec 6484
6b992ff2 6485 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
c6d3d56d
DW		 6486 test_opt(sb, JOURNAL_CHECKSUM)) {
6487 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
2d5b86e0
ES		 6488 "during remount not supported; ignoring");
6489 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6b992ff2
DW		 6490 }
6491
6ae6514b
PS		 6492 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6493 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6494 ext4_msg(sb, KERN_ERR, "can't mount with "
6495 "both data=journal and delalloc");
6496 err = -EINVAL;
6497 goto restore_opts;
6498 }
6499 if (test_opt(sb, DIOREAD_NOLOCK)) {
6500 ext4_msg(sb, KERN_ERR, "can't mount with "
6501 "both data=journal and dioread_nolock");
6502 err = -EINVAL;
6503 goto restore_opts;
6504 }
ab04df78
JK		 6505 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6506 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6507 ext4_msg(sb, KERN_ERR, "can't mount with "
6508 "journal_async_commit in data=ordered mode");
6509 err = -EINVAL;
6510 goto restore_opts;
6511 }
923ae0ff
RZ		 6512 }
6513
cdb7ee4c
TE		 6514 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6515 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6516 err = -EINVAL;
6517 goto restore_opts;
6518 }
6519
22b8d707 6520 if (test_opt2(sb, ABORT))
124e7c61 6521 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
ac27a0ec 6522
1751e8a6
LT		 6523 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6524 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
ac27a0ec
DK		 6525
6526 es = sbi->s_es;
6527
b3881f74 6528 if (sbi->s_journal) {
0390131b 6529 ext4_init_journal_params(sb, sbi->s_journal);
7edfd85b 6530 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
b3881f74 6531 }
ac27a0ec 6532
c92dc856 6533 /* Flush outstanding errors before changing fs state */
bb15cea2 6534 flush_work(&sbi->s_sb_upd_work);
c92dc856 6535
960e0ab6 6536 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
95257987 6537 if (ext4_forced_shutdown(sb)) {
ac27a0ec
DK		 6538 err = -EROFS;
6539 goto restore_opts;
6540 }
6541
960e0ab6 6542 if (fc->sb_flags & SB_RDONLY) {
38c03b34
TT		 6543 err = sync_filesystem(sb);
6544 if (err < 0)
6545 goto restore_opts;
0f0dd62f
CH		 6546 err = dquot_suspend(sb, -1);
6547 if (err < 0)
c79d967d 6548 goto restore_opts;
c79d967d 6549
ac27a0ec
DK		 6550 /*
6551 * First of all, the unconditional stuff we have to do
6552 * to disable replay of the journal when we next remount
6553 */
1751e8a6 6554 sb->s_flags |= SB_RDONLY;
ac27a0ec
DK		 6555
6556 /*
6557 * OK, test if we are remounting a valid rw partition
6558 * readonly, and if so set the rdonly flag and then
6559 * mark the partition as valid again.
6560 */
617ba13b
MC		 6561 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6562 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 6563 es->s_state = cpu_to_le16(sbi->s_mount_state);
6564
11215630
JK		 6565 if (sbi->s_journal) {
6566 /*
6567 * We let remount-ro finish even if marking fs
6568 * as clean failed...
6569 */
0390131b 6570 ext4_mark_recovery_complete(sb, es);
11215630 6571 }
ac27a0ec 6572 } else {
a13fb1a4 6573 /* Make sure we can mount this feature set readwrite */
e2b911c5 6574 if (ext4_has_feature_readonly(sb) ||
2cb5cc8b 6575 !ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 6576 err = -EROFS;
6577 goto restore_opts;
6578 }
8a266467
TT		 6579 /*
6580 * Make sure the group descriptor checksums
0b8e58a1 6581 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 6582 */
6583 for (g = 0; g < sbi->s_groups_count; g++) {
6584 struct ext4_group_desc *gdp =
6585 ext4_get_group_desc(sb, g, NULL);
6586
feb0ab32 6587 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 6588 ext4_msg(sb, KERN_ERR,
6589 "ext4_remount: Checksum for group %u failed (%u!=%u)",
e2b911c5 6590 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
8a266467 6591 le16_to_cpu(gdp->bg_checksum));
6a797d27 6592 err = -EFSBADCRC;
8a266467
TT		 6593 goto restore_opts;
6594 }
6595 }
6596
ead6596b
ES		 6597 /*
6598 * If we have an unprocessed orphan list hanging
6599 * around from a previously readonly bdev mount,
6600 * require a full umount/remount for now.
6601 */
02f310fc 6602 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
b31e1552 6603 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 6604 "remount RDWR because of unprocessed "
6605 "orphan inode list. Please "
b31e1552 6606 "umount/remount instead");
ead6596b
ES		 6607 err = -EINVAL;
6608 goto restore_opts;
6609 }
6610
ac27a0ec
DK		 6611 /*
6612 * Mounting a RDONLY partition read-write, so reread
6613 * and store the current valid flag. (It may have
6614 * been changed by e2fsck since we originally mounted
6615 * the partition.)
6616 */
11215630
JK		 6617 if (sbi->s_journal) {
6618 err = ext4_clear_journal_err(sb, es);
6619 if (err)
6620 goto restore_opts;
6621 }
c878bea3
TT		 6622 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6623 ~EXT4_FC_REPLAY);
c89128a0
JK		 6624
6625 err = ext4_setup_super(sb, es, 0);
6626 if (err)
6627 goto restore_opts;
6628
1b292439 6629 sb->s_flags &= ~SB_RDONLY;
3b50d501
TT		 6630 if (ext4_has_feature_mmp(sb)) {
6631 err = ext4_multi_mount_protect(sb,
6632 le64_to_cpu(es->s_mmp_block));
6633 if (err)
c5e06d10 6634 goto restore_opts;
3b50d501 6635 }
3bbef91b 6636#ifdef CONFIG_QUOTA
c79d967d 6637 enable_quota = 1;
3bbef91b 6638#endif
ac27a0ec
DK		 6639 }
6640 }
bfff6873 6641
0f5bde1d
JK		 6642 /*
6643 * Handle creation of system zone data early because it can fail.
6644 * Releasing of existing data is done when we are sure remount will
6645 * succeed.
6646 */
dd0db94f 6647 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
0f5bde1d
JK		 6648 err = ext4_setup_system_zone(sb);
6649 if (err)
6650 goto restore_opts;
6651 }
d176b1f6 6652
c89128a0 6653 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
4392fbc4 6654 err = ext4_commit_super(sb);
c89128a0
JK		 6655 if (err)
6656 goto restore_opts;
6657 }
0390131b 6658
ac27a0ec 6659#ifdef CONFIG_QUOTA
7c319d32
AK		 6660 if (enable_quota) {
6661 if (sb_any_quota_suspended(sb))
6662 dquot_resume(sb, -1);
e2b911c5 6663 else if (ext4_has_feature_quota(sb)) {
7c319d32 6664 err = ext4_enable_quotas(sb);
07724f98 6665 if (err)
7c319d32 6666 goto restore_opts;
7c319d32
AK		 6667 }
6668 }
4c0b4818
TT		 6669 /* Release old quota file names */
6670 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6671 kfree(old_opts.s_qf_names[i]);
ac27a0ec 6672#endif
dd0db94f 6673 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
0f5bde1d 6674 ext4_release_system_zone(sb);
d4c402d9 6675
eb1f822c
TT		 6676 /*
6677 * Reinitialize lazy itable initialization thread based on
6678 * current settings
6679 */
6680 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6681 ext4_unregister_li_request(sb);
6682 else {
6683 ext4_group_t first_not_zeroed;
6684 first_not_zeroed = ext4_has_uninit_itable(sb);
6685 ext4_register_li_request(sb, first_not_zeroed);
6686 }
6687
61bb4a1c
TT		 6688 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6689 ext4_stop_mmpd(sbi);
6690
ac27a0ec 6691 return 0;
0b8e58a1 6692
ac27a0ec 6693restore_opts:
4c0b4818
TT		 6694 /*
6695 * If there was a failing r/w to ro transition, we may need to
6696 * re-enable quota
6697 */
d5d020b3 6698 if (sb_rdonly(sb) && !(old_sb_flags & SB_RDONLY) &&
4c0b4818
TT		 6699 sb_any_quota_suspended(sb))
6700 dquot_resume(sb, -1);
745f17a4
BL		 6701
6702 alloc_ctx = ext4_writepages_down_write(sb);
ac27a0ec
DK		 6703 sb->s_flags = old_sb_flags;
6704 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 6705 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 6706 sbi->s_resuid = old_opts.s_resuid;
6707 sbi->s_resgid = old_opts.s_resgid;
6708 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 6709 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6710 sbi->s_max_batch_time = old_opts.s_max_batch_time;
745f17a4
BL		 6711 ext4_writepages_up_write(sb, alloc_ctx);
6712
dd0db94f 6713 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
0f5bde1d 6714 ext4_release_system_zone(sb);
ac27a0ec
DK		 6715#ifdef CONFIG_QUOTA
6716 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
a2d4a646 6717 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
33458eab
TT		 6718 to_free[i] = get_qf_name(sb, sbi, i);
6719 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
ac27a0ec 6720 }
33458eab
TT		 6721 synchronize_rcu();
6722 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6723 kfree(to_free[i]);
ac27a0ec 6724#endif
61bb4a1c
TT		 6725 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6726 ext4_stop_mmpd(sbi);
ac27a0ec
DK		 6727 return err;
6728}
6729
cebe85d5 6730static int ext4_reconfigure(struct fs_context *fc)
7edfd85b 6731{
cebe85d5 6732 struct super_block *sb = fc->root->d_sb;
7edfd85b
LC		 6733 int ret;
6734
cebe85d5 6735 fc->s_fs_info = EXT4_SB(sb);
7edfd85b 6736
cebe85d5 6737 ret = ext4_check_opt_consistency(fc, sb);
7edfd85b 6738 if (ret < 0)
cebe85d5 6739 return ret;
7edfd85b 6740
960e0ab6 6741 ret = __ext4_remount(fc, sb);
7edfd85b 6742 if (ret < 0)
cebe85d5 6743 return ret;
7edfd85b 6744
6dcc98fb
TT		 6745 ext4_msg(sb, KERN_INFO, "re-mounted %pU %s. Quota mode: %s.",
6746 &sb->s_uuid, sb_rdonly(sb) ? "ro" : "r/w",
6747 ext4_quota_mode(sb));
7edfd85b 6748
7edfd85b 6749 return 0;
7edfd85b
LC		 6750}
6751
689c958c
LX		 6752#ifdef CONFIG_QUOTA
6753static int ext4_statfs_project(struct super_block *sb,
6754 kprojid_t projid, struct kstatfs *buf)
6755{
6756 struct kqid qid;
6757 struct dquot *dquot;
6758 u64 limit;
6759 u64 curblock;
6760
6761 qid = make_kqid_projid(projid);
6762 dquot = dqget(sb, qid);
6763 if (IS_ERR(dquot))
6764 return PTR_ERR(dquot);
7b9ca4c6 6765 spin_lock(&dquot->dq_dqb_lock);
689c958c 6766
a08fe66e
CX		 6767 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6768 dquot->dq_dqb.dqb_bhardlimit);
57c32ea4
CX		 6769 limit >>= sb->s_blocksize_bits;
6770
689c958c 6771 if (limit && buf->f_blocks > limit) {
f06925c7
KK		 6772 curblock = (dquot->dq_dqb.dqb_curspace +
6773 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
689c958c
LX		 6774 buf->f_blocks = limit;
6775 buf->f_bfree = buf->f_bavail =
6776 (buf->f_blocks > curblock) ?
6777 (buf->f_blocks - curblock) : 0;
6778 }
6779
a08fe66e
CX		 6780 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6781 dquot->dq_dqb.dqb_ihardlimit);
689c958c
LX		 6782 if (limit && buf->f_files > limit) {
6783 buf->f_files = limit;
6784 buf->f_ffree =
6785 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6786 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6787 }
6788
7b9ca4c6 6789 spin_unlock(&dquot->dq_dqb_lock);
689c958c
LX		 6790 dqput(dquot);
6791 return 0;
6792}
6793#endif
6794
2b2d6d01 6795static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 6796{
6797 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 6798 struct ext4_sb_info *sbi = EXT4_SB(sb);
6799 struct ext4_super_block *es = sbi->s_es;
27dd4385 6800 ext4_fsblk_t overhead = 0, resv_blocks;
d02a9391 6801 s64 bfree;
27dd4385 6802 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 6803
952fc18e
TT		 6804 if (!test_opt(sb, MINIX_DF))
6805 overhead = sbi->s_overhead;
ac27a0ec 6806
617ba13b 6807 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 6808 buf->f_bsize = sb->s_blocksize;
b72f78cb 6809 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 6810 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6811 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 6812 /* prevent underflow in case that few free space is available */
57042651 6813 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 6814 buf->f_bavail = buf->f_bfree -
6815 (ext4_r_blocks_count(es) + resv_blocks);
6816 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 6817 buf->f_bavail = 0;
6818 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 6819 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 6820 buf->f_namelen = EXT4_NAME_LEN;
9591c3a3 6821 buf->f_fsid = uuid_to_fsid(es->s_uuid);
0b8e58a1 6822
689c958c
LX		 6823#ifdef CONFIG_QUOTA
6824 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6825 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6826 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6827#endif
ac27a0ec
DK		 6828 return 0;
6829}
6830
ac27a0ec
DK		 6831
6832#ifdef CONFIG_QUOTA
6833
bc8230ee
JK		 6834/*
6835 * Helper functions so that transaction is started before we acquire dqio_sem
6836 * to keep correct lock ordering of transaction > dqio_sem
6837 */
ac27a0ec
DK		 6838static inline struct inode *dquot_to_inode(struct dquot *dquot)
6839{
4c376dca 6840 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 6841}
6842
617ba13b 6843static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 6844{
6845 int ret, err;
6846 handle_t *handle;
6847 struct inode *inode;
6848
6849 inode = dquot_to_inode(dquot);
9924a92a 6850 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 6851 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 6852 if (IS_ERR(handle))
6853 return PTR_ERR(handle);
6854 ret = dquot_commit(dquot);
d8b945fa
YB		 6855 if (ret < 0)
6856 ext4_error_err(dquot->dq_sb, -ret,
6857 "Failed to commit dquot type %d",
6858 dquot->dq_id.type);
617ba13b 6859 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6860 if (!ret)
6861 ret = err;
6862 return ret;
6863}
6864
617ba13b 6865static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 6866{
6867 int ret, err;
6868 handle_t *handle;
6869
9924a92a 6870 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 6871 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 6872 if (IS_ERR(handle))
6873 return PTR_ERR(handle);
6874 ret = dquot_acquire(dquot);
d8b945fa
YB		 6875 if (ret < 0)
6876 ext4_error_err(dquot->dq_sb, -ret,
6877 "Failed to acquire dquot type %d",
6878 dquot->dq_id.type);
617ba13b 6879 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6880 if (!ret)
6881 ret = err;
6882 return ret;
6883}
6884
617ba13b 6885static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 6886{
6887 int ret, err;
6888 handle_t *handle;
6889
9924a92a 6890 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 6891 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 6892 if (IS_ERR(handle)) {
6893 /* Release dquot anyway to avoid endless cycle in dqput() */
6894 dquot_release(dquot);
ac27a0ec 6895 return PTR_ERR(handle);
9c3013e9 6896 }
ac27a0ec 6897 ret = dquot_release(dquot);
d8b945fa
YB		 6898 if (ret < 0)
6899 ext4_error_err(dquot->dq_sb, -ret,
6900 "Failed to release dquot type %d",
6901 dquot->dq_id.type);
617ba13b 6902 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6903 if (!ret)
6904 ret = err;
6905 return ret;
6906}
6907
617ba13b 6908static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 6909{
262b4662 6910 struct super_block *sb = dquot->dq_sb;
262b4662 6911
f177ee08 6912 if (ext4_is_quota_journalled(sb)) {
ac27a0ec 6913 dquot_mark_dquot_dirty(dquot);
617ba13b 6914 return ext4_write_dquot(dquot);
ac27a0ec
DK		 6915 } else {
6916 return dquot_mark_dquot_dirty(dquot);
6917 }
6918}
6919
617ba13b 6920static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 6921{
6922 int ret, err;
6923 handle_t *handle;
6924
6925 /* Data block + inode block */
f9c1f248 6926 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 6927 if (IS_ERR(handle))
6928 return PTR_ERR(handle);
6929 ret = dquot_commit_info(sb, type);
617ba13b 6930 err = ext4_journal_stop(handle);
ac27a0ec
DK		 6931 if (!ret)
6932 ret = err;
6933 return ret;
6934}
6935
daf647d2
TT		 6936static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6937{
6938 struct ext4_inode_info *ei = EXT4_I(inode);
6939
6940 /* The first argument of lockdep_set_subclass has to be
6941 * *exactly* the same as the argument to init_rwsem() --- in
6942 * this case, in init_once() --- or lockdep gets unhappy
6943 * because the name of the lock is set using the
6944 * stringification of the argument to init_rwsem().
6945 */
6946 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6947 lockdep_set_subclass(&ei->i_data_sem, subclass);
6948}
6949
ac27a0ec
DK		 6950/*
6951 * Standard function to be called on quota_on
6952 */
617ba13b 6953static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8c54ca9c 6954 const struct path *path)
ac27a0ec
DK		 6955{
6956 int err;
ac27a0ec
DK		 6957
6958 if (!test_opt(sb, QUOTA))
6959 return -EINVAL;
0623543b 6960
ac27a0ec 6961 /* Quotafile not on the same filesystem? */
d8c9584e 6962 if (path->dentry->d_sb != sb)
ac27a0ec 6963 return -EXDEV;
e0770e91
JK		 6964
6965 /* Quota already enabled for this file? */
6966 if (IS_NOQUOTA(d_inode(path->dentry)))
6967 return -EBUSY;
6968
0623543b
JK		 6969 /* Journaling quota? */
6970 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 6971 /* Quotafile not in fs root? */
f00c9e44 6972 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 6973 ext4_msg(sb, KERN_WARNING,
6974 "Quota file not on filesystem root. "
6975 "Journaled quota will not work");
91389240
JK		 6976 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6977 } else {
6978 /*
6979 * Clear the flag just in case mount options changed since
6980 * last time.
6981 */
6982 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
2b2d6d01 6983 }
0623543b 6984
daf647d2
TT		 6985 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6986 err = dquot_quota_on(sb, type, format_id, path);
15fc69bb 6987 if (!err) {
957153fc
JK		 6988 struct inode *inode = d_inode(path->dentry);
6989 handle_t *handle;
6990
61a92987
JK		 6991 /*
6992 * Set inode flags to prevent userspace from messing with quota
6993 * files. If this fails, we return success anyway since quotas
6994 * are already enabled and this is not a hard failure.
6995 */
957153fc
JK		 6996 inode_lock(inode);
6997 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6998 if (IS_ERR(handle))
6999 goto unlock_inode;
7000 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
7001 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
7002 S_NOATIME | S_IMMUTABLE);
4209ae12 7003 err = ext4_mark_inode_dirty(handle, inode);
957153fc
JK		 7004 ext4_journal_stop(handle);
7005 unlock_inode:
7006 inode_unlock(inode);
15fc69bb
JK		 7007 if (err)
7008 dquot_quota_off(sb, type);
957153fc 7009 }
15fc69bb
JK		 7010 if (err)
7011 lockdep_set_quota_inode(path->dentry->d_inode,
7012 I_DATA_SEM_NORMAL);
daf647d2 7013 return err;
ac27a0ec
DK		 7014}
7015
07342ec2
BL		 7016static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
7017{
7018 switch (type) {
7019 case USRQUOTA:
7020 return qf_inum == EXT4_USR_QUOTA_INO;
7021 case GRPQUOTA:
7022 return qf_inum == EXT4_GRP_QUOTA_INO;
7023 case PRJQUOTA:
7024 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
7025 default:
7026 BUG();
7027 }
7028}
7029
7c319d32
AK		 7030static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
7031 unsigned int flags)
7032{
7033 int err;
7034 struct inode *qf_inode;
a2d4a646 7035 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 7036 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 7037 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7038 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32
AK		 7039 };
7040
e2b911c5 7041 BUG_ON(!ext4_has_feature_quota(sb));
7c319d32
AK		 7042
7043 if (!qf_inums[type])
7044 return -EPERM;
7045
07342ec2
BL		 7046 if (!ext4_check_quota_inum(type, qf_inums[type])) {
7047 ext4_error(sb, "Bad quota inum: %lu, type: %d",
7048 qf_inums[type], type);
7049 return -EUCLEAN;
7050 }
7051
8a363970 7052 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7c319d32 7053 if (IS_ERR(qf_inode)) {
07342ec2
BL		 7054 ext4_error(sb, "Bad quota inode: %lu, type: %d",
7055 qf_inums[type], type);
7c319d32
AK		 7056 return PTR_ERR(qf_inode);
7057 }
7058
bcb13850
JK		 7059 /* Don't account quota for quota files to avoid recursion */
7060 qf_inode->i_flags |= S_NOQUOTA;
daf647d2 7061 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7212b95e 7062 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
daf647d2
TT		 7063 if (err)
7064 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
61157b24 7065 iput(qf_inode);
7c319d32
AK		 7066
7067 return err;
7068}
7069
7070/* Enable usage tracking for all quota types. */
25c6d98f 7071int ext4_enable_quotas(struct super_block *sb)
7c319d32
AK		 7072{
7073 int type, err = 0;
a2d4a646 7074 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7c319d32 7075 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
689c958c
LX		 7076 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7077 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7c319d32 7078 };
49da9392
JK		 7079 bool quota_mopt[EXT4_MAXQUOTAS] = {
7080 test_opt(sb, USRQUOTA),
7081 test_opt(sb, GRPQUOTA),
7082 test_opt(sb, PRJQUOTA),
7083 };
7c319d32 7084
91389240 7085 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
a2d4a646 7086 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7c319d32
AK		 7087 if (qf_inums[type]) {
7088 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
49da9392
JK		 7089 DQUOT_USAGE_ENABLED |
7090 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7c319d32
AK		 7091 if (err) {
7092 ext4_warning(sb,
72ba7450 7093 "Failed to enable quota tracking "
07342ec2
BL		 7094 "(type=%d, err=%d, ino=%lu). "
7095 "Please run e2fsck to fix.", type,
7096 err, qf_inums[type]);
7f144fd0 7097
f3c1c42e 7098 ext4_quotas_off(sb, type);
7c319d32
AK		 7099 return err;
7100 }
7101 }
7102 }
7103 return 0;
7104}
7105
ca0e05e4
DM		 7106static int ext4_quota_off(struct super_block *sb, int type)
7107{
21f97697
JK		 7108 struct inode *inode = sb_dqopt(sb)->files[type];
7109 handle_t *handle;
957153fc 7110 int err;
21f97697 7111
87009d86
DM		 7112 /* Force all delayed allocation blocks to be allocated.
7113 * Caller already holds s_umount sem */
7114 if (test_opt(sb, DELALLOC))
ca0e05e4 7115 sync_filesystem(sb);
ca0e05e4 7116
957153fc 7117 if (!inode || !igrab(inode))
0b268590
AG		 7118 goto out;
7119
957153fc 7120 err = dquot_quota_off(sb, type);
964edf66 7121 if (err || ext4_has_feature_quota(sb))
957153fc 7122 goto out_put;
e0e985f3
JK		 7123 /*
7124 * When the filesystem was remounted read-only first, we cannot cleanup
7125 * inode flags here. Bad luck but people should be using QUOTA feature
7126 * these days anyway.
7127 */
7128 if (sb_rdonly(sb))
7129 goto out_put;
957153fc
JK		 7130
7131 inode_lock(inode);
61a92987
JK		 7132 /*
7133 * Update modification times of quota files when userspace can
7134 * start looking at them. If we fail, we return success anyway since
7135 * this is not a hard failure and quotas are already disabled.
7136 */
9924a92a 7137 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
4209ae12
HS		 7138 if (IS_ERR(handle)) {
7139 err = PTR_ERR(handle);
957153fc 7140 goto out_unlock;
4209ae12 7141 }
957153fc
JK		 7142 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7143 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
b898ab23 7144 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
4209ae12 7145 err = ext4_mark_inode_dirty(handle, inode);
21f97697 7146 ext4_journal_stop(handle);
957153fc
JK		 7147out_unlock:
7148 inode_unlock(inode);
7149out_put:
964edf66 7150 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
957153fc
JK		 7151 iput(inode);
7152 return err;
21f97697 7153out:
ca0e05e4
DM		 7154 return dquot_quota_off(sb, type);
7155}
7156
ac27a0ec
DK		 7157/* Read data from quotafile - avoid pagecache and such because we cannot afford
7158 * acquiring the locks... As quota files are never truncated and quota code
25985edc 7159 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 7160 * we don't have to be afraid of races */
617ba13b 7161static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 7162 size_t len, loff_t off)
7163{
7164 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 7165 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 7166 int offset = off & (sb->s_blocksize - 1);
7167 int tocopy;
7168 size_t toread;
7169 struct buffer_head *bh;
7170 loff_t i_size = i_size_read(inode);
7171
7172 if (off > i_size)
7173 return 0;
7174 if (off+len > i_size)
7175 len = i_size-off;
7176 toread = len;
7177 while (toread > 0) {
66267814 7178 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
1c215028
TT		 7179 bh = ext4_bread(NULL, inode, blk, 0);
7180 if (IS_ERR(bh))
7181 return PTR_ERR(bh);
ac27a0ec
DK		 7182 if (!bh) /* A hole? */
7183 memset(data, 0, tocopy);
7184 else
7185 memcpy(data, bh->b_data+offset, tocopy);
7186 brelse(bh);
7187 offset = 0;
7188 toread -= tocopy;
7189 data += tocopy;
7190 blk++;
7191 }
7192 return len;
7193}
7194
7195/* Write to quotafile (we know the transaction is already started and has
7196 * enough credits) */
617ba13b 7197static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 7198 const char *data, size_t len, loff_t off)
7199{
7200 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 7201 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4209ae12 7202 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
c5e298ae 7203 int retries = 0;
ac27a0ec
DK		 7204 struct buffer_head *bh;
7205 handle_t *handle = journal_current_handle();
7206
380a0091 7207 if (!handle) {
b31e1552
ES		 7208 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7209 " cancelled because transaction is not started",
9c3013e9
JK		 7210 (unsigned long long)off, (unsigned long long)len);
7211 return -EIO;
7212 }
67eeb568
DM		 7213 /*
7214 * Since we account only one data block in transaction credits,
7215 * then it is impossible to cross a block boundary.
7216 */
7217 if (sb->s_blocksize - offset < len) {
7218 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7219 " cancelled because not block aligned",
7220 (unsigned long long)off, (unsigned long long)len);
7221 return -EIO;
7222 }
7223
c5e298ae
TT		 7224 do {
7225 bh = ext4_bread(handle, inode, blk,
7226 EXT4_GET_BLOCKS_CREATE |
7227 EXT4_GET_BLOCKS_METADATA_NOFAIL);
45586c70 7228 } while (PTR_ERR(bh) == -ENOSPC &&
c5e298ae 7229 ext4_should_retry_alloc(inode->i_sb, &retries));
1c215028
TT		 7230 if (IS_ERR(bh))
7231 return PTR_ERR(bh);
67eeb568
DM		 7232 if (!bh)
7233 goto out;
5d601255 7234 BUFFER_TRACE(bh, "get write access");
188c299e 7235 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
62d2b5f2
JK		 7236 if (err) {
7237 brelse(bh);
1c215028 7238 return err;
ac27a0ec 7239 }
67eeb568
DM		 7240 lock_buffer(bh);
7241 memcpy(bh->b_data+offset, data, len);
7242 flush_dcache_page(bh->b_page);
7243 unlock_buffer(bh);
62d2b5f2 7244 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 7245 brelse(bh);
ac27a0ec 7246out:
67eeb568
DM		 7247 if (inode->i_size < off + len) {
7248 i_size_write(inode, off + len);
617ba13b 7249 EXT4_I(inode)->i_disksize = inode->i_size;
4209ae12
HS		 7250 err2 = ext4_mark_inode_dirty(handle, inode);
7251 if (unlikely(err2 && !err))
7252 err = err2;
ac27a0ec 7253 }
4209ae12 7254 return err ? err : len;
ac27a0ec 7255}
ac27a0ec
DK		 7256#endif
7257
c290ea01 7258#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
24b58424
TT		 7259static inline void register_as_ext2(void)
7260{
7261 int err = register_filesystem(&ext2_fs_type);
7262 if (err)
7263 printk(KERN_WARNING
7264 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7265}
7266
7267static inline void unregister_as_ext2(void)
7268{
7269 unregister_filesystem(&ext2_fs_type);
7270}
2035e776
TT		 7271
7272static inline int ext2_feature_set_ok(struct super_block *sb)
7273{
e2b911c5 7274 if (ext4_has_unknown_ext2_incompat_features(sb))
2035e776 7275 return 0;
bc98a42c 7276 if (sb_rdonly(sb))
2035e776 7277 return 1;
e2b911c5 7278 if (ext4_has_unknown_ext2_ro_compat_features(sb))
2035e776
TT		 7279 return 0;
7280 return 1;
7281}
24b58424
TT		 7282#else
7283static inline void register_as_ext2(void) { }
7284static inline void unregister_as_ext2(void) { }
2035e776 7285static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 7286#endif
7287
24b58424
TT		 7288static inline void register_as_ext3(void)
7289{
7290 int err = register_filesystem(&ext3_fs_type);
7291 if (err)
7292 printk(KERN_WARNING
7293 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7294}
7295
7296static inline void unregister_as_ext3(void)
7297{
7298 unregister_filesystem(&ext3_fs_type);
7299}
2035e776
TT		 7300
7301static inline int ext3_feature_set_ok(struct super_block *sb)
7302{
e2b911c5 7303 if (ext4_has_unknown_ext3_incompat_features(sb))
2035e776 7304 return 0;
e2b911c5 7305 if (!ext4_has_feature_journal(sb))
2035e776 7306 return 0;
bc98a42c 7307 if (sb_rdonly(sb))
2035e776 7308 return 1;
e2b911c5 7309 if (ext4_has_unknown_ext3_ro_compat_features(sb))
2035e776
TT		 7310 return 0;
7311 return 1;
7312}
24b58424 7313
1489dffd
CH		 7314static void ext4_kill_sb(struct super_block *sb)
7315{
7316 struct ext4_sb_info *sbi = EXT4_SB(sb);
61ead714 7317 struct file *bdev_file = sbi ? sbi->s_journal_bdev_file : NULL;
1489dffd
CH		 7318
7319 kill_block_super(sb);
7320
61ead714 7321 if (bdev_file)
22650a99 7322 bdev_fput(bdev_file);
1489dffd
CH		 7323}
7324
03010a33 7325static struct file_system_type ext4_fs_type = {
cebe85d5
LC		 7326 .owner = THIS_MODULE,
7327 .name = "ext4",
7328 .init_fs_context = ext4_init_fs_context,
7329 .parameters = ext4_param_specs,
1489dffd 7330 .kill_sb = ext4_kill_sb,
50ec1d72 7331 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
03010a33 7332};
7f78e035 7333MODULE_ALIAS_FS("ext4");
03010a33 7334
e9e3bcec
ES		 7335/* Shared across all ext4 file systems */
7336wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
e9e3bcec 7337
5dabfc78 7338static int __init ext4_init_fs(void)
ac27a0ec 7339{
e9e3bcec 7340 int i, err;
c9de560d 7341
e294a537 7342 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
07c0c5d8 7343 ext4_li_info = NULL;
07c0c5d8 7344
9a4c8019 7345 /* Build-time check for flags consistency */
12e9b892 7346 ext4_check_flag_values();
e9e3bcec 7347
e142d052 7348 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
e9e3bcec 7349 init_waitqueue_head(&ext4__ioend_wq[i]);
e9e3bcec 7350
51865fda 7351 err = ext4_init_es();
6fd058f7
TT		 7352 if (err)
7353 return err;
51865fda 7354
1dc0aa46 7355 err = ext4_init_pending();
22cfe4b4
EB		 7356 if (err)
7357 goto out7;
7358
7359 err = ext4_init_post_read_processing();
1dc0aa46
EW		 7360 if (err)
7361 goto out6;
7362
51865fda
ZL		 7363 err = ext4_init_pageio();
7364 if (err)
b5799018 7365 goto out5;
51865fda 7366
5dabfc78 7367 err = ext4_init_system_zone();
bd2d0210 7368 if (err)
b5799018 7369 goto out4;
857ac889 7370
b5799018 7371 err = ext4_init_sysfs();
dd68314c 7372 if (err)
b5799018 7373 goto out3;
857ac889 7374
5dabfc78 7375 err = ext4_init_mballoc();
c9de560d
AT		 7376 if (err)
7377 goto out2;
ac27a0ec
DK		 7378 err = init_inodecache();
7379 if (err)
7380 goto out1;
aa75f4d3
HS		 7381
7382 err = ext4_fc_init_dentry_cache();
7383 if (err)
7384 goto out05;
7385
24b58424 7386 register_as_ext3();
2035e776 7387 register_as_ext2();
03010a33 7388 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 7389 if (err)
7390 goto out;
bfff6873 7391
ac27a0ec
DK		 7392 return 0;
7393out:
24b58424
TT		 7394 unregister_as_ext2();
7395 unregister_as_ext3();
ab047d51 7396 ext4_fc_destroy_dentry_cache();
aa75f4d3 7397out05:
ac27a0ec
DK		 7398 destroy_inodecache();
7399out1:
5dabfc78 7400 ext4_exit_mballoc();
9c191f70 7401out2:
b5799018
TT		 7402 ext4_exit_sysfs();
7403out3:
5dabfc78 7404 ext4_exit_system_zone();
b5799018 7405out4:
5dabfc78 7406 ext4_exit_pageio();
b5799018 7407out5:
22cfe4b4 7408 ext4_exit_post_read_processing();
1dc0aa46 7409out6:
22cfe4b4
EB		 7410 ext4_exit_pending();
7411out7:
51865fda
ZL		 7412 ext4_exit_es();
7413
ac27a0ec
DK		 7414 return err;
7415}
7416
5dabfc78 7417static void __exit ext4_exit_fs(void)
ac27a0ec 7418{
bfff6873 7419 ext4_destroy_lazyinit_thread();
24b58424
TT		 7420 unregister_as_ext2();
7421 unregister_as_ext3();
03010a33 7422 unregister_filesystem(&ext4_fs_type);
ab047d51 7423 ext4_fc_destroy_dentry_cache();
ac27a0ec 7424 destroy_inodecache();
5dabfc78 7425 ext4_exit_mballoc();
b5799018 7426 ext4_exit_sysfs();
5dabfc78
TT		 7427 ext4_exit_system_zone();
7428 ext4_exit_pageio();
22cfe4b4 7429 ext4_exit_post_read_processing();
dd12ed14 7430 ext4_exit_es();
1dc0aa46 7431 ext4_exit_pending();
ac27a0ec
DK		 7432}
7433
7434MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 7435MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 7436MODULE_LICENSE("GPL");
7ef79ad5 7437MODULE_SOFTDEP("pre: crc32c");
5dabfc78
TT		 7438module_init(ext4_init_fs)
7439module_exit(ext4_exit_fs)
Linux 6.x block layer and io_uring tree(s)