]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: do not try to write superblock on ro remount w/o journal
[mirror_ubuntu-bionic-kernel.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
4a092d73 48#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 72static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 73 char nbuf[16]);
2b2d6d01
TT		 74static int ext4_remount(struct super_block *sb, int *flags, char *data);
75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 76static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 77static int ext4_freeze(struct super_block *sb);
152a0836
AV		 78static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
2035e776
TT		 80static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 82static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 83static void ext4_destroy_lazyinit_thread(void);
84static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 85static void ext4_clear_request_list(void);
ac27a0ec 86
2035e776
TT		 87#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94};
95#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96#else
97#define IS_EXT2_SB(sb) (0)
98#endif
99
100
ba69f9ab
JK		 101#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
104 .name = "ext3",
152a0836 105 .mount = ext4_mount,
ba69f9ab
JK		 106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
108};
109#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110#else
111#define IS_EXT3_SB(sb) (0)
112#endif
bd81d8ee 113
d25425f8
DW		 114static int ext4_verify_csum_type(struct super_block *sb,
115 struct ext4_super_block *es)
116{
117 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
118 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
119 return 1;
120
121 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
122}
123
a9c47317
DW		 124static __le32 ext4_superblock_csum(struct super_block *sb,
125 struct ext4_super_block *es)
126{
127 struct ext4_sb_info *sbi = EXT4_SB(sb);
128 int offset = offsetof(struct ext4_super_block, s_checksum);
129 __u32 csum;
130
131 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
132
133 return cpu_to_le32(csum);
134}
135
136int ext4_superblock_csum_verify(struct super_block *sb,
137 struct ext4_super_block *es)
138{
139 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
140 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
141 return 1;
142
143 return es->s_checksum == ext4_superblock_csum(sb, es);
144}
145
06db49e6 146void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 147{
06db49e6
TT		 148 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
149
a9c47317
DW		 150 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
151 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
152 return;
153
154 es->s_checksum = ext4_superblock_csum(sb, es);
155}
156
9933fc0a
TT		 157void *ext4_kvmalloc(size_t size, gfp_t flags)
158{
159 void *ret;
160
161 ret = kmalloc(size, flags);
162 if (!ret)
163 ret = __vmalloc(size, flags, PAGE_KERNEL);
164 return ret;
165}
166
167void *ext4_kvzalloc(size_t size, gfp_t flags)
168{
169 void *ret;
170
db9481c0 171 ret = kzalloc(size, flags);
9933fc0a
TT		 172 if (!ret)
173 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
174 return ret;
175}
176
177void ext4_kvfree(void *ptr)
178{
179 if (is_vmalloc_addr(ptr))
180 vfree(ptr);
181 else
182 kfree(ptr);
183
184}
185
8fadc143
AR		 186ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
bd81d8ee 188{
3a14589c 189 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 191 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 192}
193
8fadc143
AR		 194ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
195 struct ext4_group_desc *bg)
bd81d8ee 196{
5272f837 197 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 200}
201
8fadc143
AR		 202ext4_fsblk_t ext4_inode_table(struct super_block *sb,
203 struct ext4_group_desc *bg)
bd81d8ee 204{
5272f837 205 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 207 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 208}
209
021b65bb
TT		 210__u32 ext4_free_group_clusters(struct super_block *sb,
211 struct ext4_group_desc *bg)
560671a0
AK		 212{
213 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 215 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 216}
217
218__u32 ext4_free_inodes_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220{
221 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 223 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 224}
225
226__u32 ext4_used_dirs_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
228{
229 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 231 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 232}
233
234__u32 ext4_itable_unused_count(struct super_block *sb,
235 struct ext4_group_desc *bg)
236{
237 return le16_to_cpu(bg->bg_itable_unused_lo) |
238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 239 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 240}
241
8fadc143
AR		 242void ext4_block_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 244{
3a14589c 245 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 248}
249
8fadc143
AR		 250void ext4_inode_bitmap_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 252{
5272f837 253 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 256}
257
8fadc143
AR		 258void ext4_inode_table_set(struct super_block *sb,
259 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 260{
5272f837 261 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 264}
265
021b65bb
TT		 266void ext4_free_group_clusters_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 268{
269 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
272}
273
274void ext4_free_inodes_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276{
277 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
280}
281
282void ext4_used_dirs_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284{
285 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
288}
289
290void ext4_itable_unused_set(struct super_block *sb,
291 struct ext4_group_desc *bg, __u32 count)
292{
293 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
296}
297
d3d1faf6
CW		 298
299/* Just increment the non-pointer handle value */
300static handle_t *ext4_get_nojournal(void)
301{
302 handle_t *handle = current->journal_info;
303 unsigned long ref_cnt = (unsigned long)handle;
304
305 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
306
307 ref_cnt++;
308 handle = (handle_t *)ref_cnt;
309
310 current->journal_info = handle;
311 return handle;
312}
313
314
315/* Decrement the non-pointer handle value */
316static void ext4_put_nojournal(handle_t *handle)
317{
318 unsigned long ref_cnt = (unsigned long)handle;
319
320 BUG_ON(ref_cnt == 0);
321
322 ref_cnt--;
323 handle = (handle_t *)ref_cnt;
324
325 current->journal_info = handle;
326}
327
ac27a0ec 328/*
dab291af 329 * Wrappers for jbd2_journal_start/end.
ac27a0ec 330 */
617ba13b 331handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 332{
333 journal_t *journal;
334
12706394 335 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 336 if (sb->s_flags & MS_RDONLY)
337 return ERR_PTR(-EROFS);
338
8e8ad8a5 339 WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE);
617ba13b 340 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 341 if (!journal)
342 return ext4_get_nojournal();
343 /*
344 * Special case here: if the journal has aborted behind our
345 * backs (eg. EIO in the commit thread), then we still need to
346 * take the FS itself readonly cleanly.
347 */
348 if (is_journal_aborted(journal)) {
349 ext4_abort(sb, "Detected aborted journal");
350 return ERR_PTR(-EROFS);
ac27a0ec 351 }
be4f27d3 352 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 353}
354
c398eda0 355int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 356{
357 struct super_block *sb;
358 int err;
359 int rc;
360
0390131b 361 if (!ext4_handle_valid(handle)) {
d3d1faf6 362 ext4_put_nojournal(handle);
0390131b
FM		 363 return 0;
364 }
ac27a0ec
DK		 365 sb = handle->h_transaction->t_journal->j_private;
366 err = handle->h_err;
dab291af 367 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 368
369 if (!err)
370 err = rc;
371 if (err)
c398eda0 372 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 373 return err;
374}
375
90c7201b
TT		 376void ext4_journal_abort_handle(const char *caller, unsigned int line,
377 const char *err_fn, struct buffer_head *bh,
378 handle_t *handle, int err)
ac27a0ec
DK		 379{
380 char nbuf[16];
617ba13b 381 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 382
0390131b
FM		 383 BUG_ON(!ext4_handle_valid(handle));
384
ac27a0ec
DK		 385 if (bh)
386 BUFFER_TRACE(bh, "abort");
387
388 if (!handle->h_err)
389 handle->h_err = err;
390
391 if (is_handle_aborted(handle))
392 return;
393
92b97816 394 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
90c7201b 395 caller, line, errstr, err_fn);
ac27a0ec 396
dab291af 397 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 398}
399
1c13d5c0
TT		 400static void __save_error_info(struct super_block *sb, const char *func,
401 unsigned int line)
402{
403 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
404
405 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
406 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
407 es->s_last_error_time = cpu_to_le32(get_seconds());
408 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
409 es->s_last_error_line = cpu_to_le32(line);
410 if (!es->s_first_error_time) {
411 es->s_first_error_time = es->s_last_error_time;
412 strncpy(es->s_first_error_func, func,
413 sizeof(es->s_first_error_func));
414 es->s_first_error_line = cpu_to_le32(line);
415 es->s_first_error_ino = es->s_last_error_ino;
416 es->s_first_error_block = es->s_last_error_block;
417 }
66e61a9e
TT		 418 /*
419 * Start the daily error reporting function if it hasn't been
420 * started already
421 */
422 if (!es->s_error_count)
423 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 424 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 425}
426
427static void save_error_info(struct super_block *sb, const char *func,
428 unsigned int line)
429{
430 __save_error_info(sb, func, line);
431 ext4_commit_super(sb, 1);
432}
433
7c2e7087
TT		 434/*
435 * The del_gendisk() function uninitializes the disk-specific data
436 * structures, including the bdi structure, without telling anyone
437 * else. Once this happens, any attempt to call mark_buffer_dirty()
438 * (for example, by ext4_commit_super), will cause a kernel OOPS.
439 * This is a kludge to prevent these oops until we can put in a proper
440 * hook in del_gendisk() to inform the VFS and file system layers.
441 */
442static int block_device_ejected(struct super_block *sb)
443{
444 struct inode *bd_inode = sb->s_bdev->bd_inode;
445 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
446
447 return bdi->dev == NULL;
448}
449
18aadd47
BJ		 450static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
451{
452 struct super_block *sb = journal->j_private;
453 struct ext4_sb_info *sbi = EXT4_SB(sb);
454 int error = is_journal_aborted(journal);
455 struct ext4_journal_cb_entry *jce, *tmp;
456
457 spin_lock(&sbi->s_md_lock);
458 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
459 list_del_init(&jce->jce_list);
460 spin_unlock(&sbi->s_md_lock);
461 jce->jce_func(sb, jce, error);
462 spin_lock(&sbi->s_md_lock);
463 }
464 spin_unlock(&sbi->s_md_lock);
465}
1c13d5c0 466
ac27a0ec
DK		 467/* Deal with the reporting of failure conditions on a filesystem such as
468 * inconsistencies detected or read IO failures.
469 *
470 * On ext2, we can store the error state of the filesystem in the
617ba13b 471 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 472 * write ordering constraints on the superblock which prevent us from
473 * writing it out straight away; and given that the journal is about to
474 * be aborted, we can't rely on the current, or future, transactions to
475 * write out the superblock safely.
476 *
dab291af 477 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 478 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 479 * that error until we've noted it down and cleared it.
480 */
481
617ba13b 482static void ext4_handle_error(struct super_block *sb)
ac27a0ec 483{
ac27a0ec
DK		 484 if (sb->s_flags & MS_RDONLY)
485 return;
486
2b2d6d01 487 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 488 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 489
4ab2f15b 490 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 491 if (journal)
dab291af 492 jbd2_journal_abort(journal, -EIO);
ac27a0ec 493 }
2b2d6d01 494 if (test_opt(sb, ERRORS_RO)) {
b31e1552 495 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 496 sb->s_flags |= MS_RDONLY;
497 }
ac27a0ec 498 if (test_opt(sb, ERRORS_PANIC))
617ba13b 499 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 500 sb->s_id);
501}
502
12062ddd 503void __ext4_error(struct super_block *sb, const char *function,
c398eda0 504 unsigned int line, const char *fmt, ...)
ac27a0ec 505{
0ff2ea7d 506 struct va_format vaf;
ac27a0ec
DK		 507 va_list args;
508
509 va_start(args, fmt);
0ff2ea7d
JP		 510 vaf.fmt = fmt;
511 vaf.va = &args;
512 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
513 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec 514 va_end(args);
f3fc0210 515 save_error_info(sb, function, line);
ac27a0ec 516
617ba13b 517 ext4_handle_error(sb);
ac27a0ec
DK		 518}
519
c398eda0
TT		 520void ext4_error_inode(struct inode *inode, const char *function,
521 unsigned int line, ext4_fsblk_t block,
273df556
FM		 522 const char *fmt, ...)
523{
524 va_list args;
f7c21177 525 struct va_format vaf;
1c13d5c0 526 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 527
1c13d5c0
TT		 528 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
529 es->s_last_error_block = cpu_to_le64(block);
530 save_error_info(inode->i_sb, function, line);
273df556 531 va_start(args, fmt);
f7c21177
TT		 532 vaf.fmt = fmt;
533 vaf.va = &args;
c398eda0 534 if (block)
d9ee81da
JP		 535 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
536 "inode #%lu: block %llu: comm %s: %pV\n",
537 inode->i_sb->s_id, function, line, inode->i_ino,
538 block, current->comm, &vaf);
539 else
540 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
541 "inode #%lu: comm %s: %pV\n",
542 inode->i_sb->s_id, function, line, inode->i_ino,
543 current->comm, &vaf);
273df556
FM		 544 va_end(args);
545
546 ext4_handle_error(inode->i_sb);
547}
548
c398eda0 549void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 550 unsigned int line, ext4_fsblk_t block,
551 const char *fmt, ...)
273df556
FM		 552{
553 va_list args;
f7c21177 554 struct va_format vaf;
1c13d5c0 555 struct ext4_super_block *es;
273df556
FM		 556 struct inode *inode = file->f_dentry->d_inode;
557 char pathname[80], *path;
558
1c13d5c0
TT		 559 es = EXT4_SB(inode->i_sb)->s_es;
560 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
561 save_error_info(inode->i_sb, function, line);
273df556 562 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 563 if (IS_ERR(path))
273df556 564 path = "(unknown)";
f7c21177
TT		 565 va_start(args, fmt);
566 vaf.fmt = fmt;
567 vaf.va = &args;
d9ee81da
JP		 568 if (block)
569 printk(KERN_CRIT
570 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
571 "block %llu: comm %s: path %s: %pV\n",
572 inode->i_sb->s_id, function, line, inode->i_ino,
573 block, current->comm, path, &vaf);
574 else
575 printk(KERN_CRIT
576 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
577 "comm %s: path %s: %pV\n",
578 inode->i_sb->s_id, function, line, inode->i_ino,
579 current->comm, path, &vaf);
273df556
FM		 580 va_end(args);
581
582 ext4_handle_error(inode->i_sb);
583}
584
2b2d6d01 585static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 586 char nbuf[16])
587{
588 char *errstr = NULL;
589
590 switch (errno) {
591 case -EIO:
592 errstr = "IO failure";
593 break;
594 case -ENOMEM:
595 errstr = "Out of memory";
596 break;
597 case -EROFS:
78f1ddbb
TT		 598 if (!sb || (EXT4_SB(sb)->s_journal &&
599 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 600 errstr = "Journal has aborted";
601 else
602 errstr = "Readonly filesystem";
603 break;
604 default:
605 /* If the caller passed in an extra buffer for unknown
606 * errors, textualise them now. Else we just return
607 * NULL. */
608 if (nbuf) {
609 /* Check for truncated error codes... */
610 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
611 errstr = nbuf;
612 }
613 break;
614 }
615
616 return errstr;
617}
618
617ba13b 619/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 620 * automatically and invokes the appropriate error response. */
621
c398eda0
TT		 622void __ext4_std_error(struct super_block *sb, const char *function,
623 unsigned int line, int errno)
ac27a0ec
DK		 624{
625 char nbuf[16];
626 const char *errstr;
627
628 /* Special case: if the error is EROFS, and we're not already
629 * inside a transaction, then there's really no point in logging
630 * an error. */
631 if (errno == -EROFS && journal_current_handle() == NULL &&
632 (sb->s_flags & MS_RDONLY))
633 return;
634
617ba13b 635 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 636 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
637 sb->s_id, function, line, errstr);
1c13d5c0 638 save_error_info(sb, function, line);
ac27a0ec 639
617ba13b 640 ext4_handle_error(sb);
ac27a0ec
DK		 641}
642
643/*
617ba13b 644 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 645 * abort function may be used to deal with unrecoverable failures such
646 * as journal IO errors or ENOMEM at a critical moment in log management.
647 *
648 * We unconditionally force the filesystem into an ABORT|READONLY state,
649 * unless the error response on the fs has been set to panic in which
650 * case we take the easy way out and panic immediately.
651 */
652
c67d859e 653void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 654 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 655{
656 va_list args;
657
1c13d5c0 658 save_error_info(sb, function, line);
ac27a0ec 659 va_start(args, fmt);
c398eda0
TT		 660 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
661 function, line);
ac27a0ec
DK		 662 vprintk(fmt, args);
663 printk("\n");
664 va_end(args);
665
1c13d5c0
TT		 666 if ((sb->s_flags & MS_RDONLY) == 0) {
667 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
668 sb->s_flags |= MS_RDONLY;
669 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
670 if (EXT4_SB(sb)->s_journal)
671 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
672 save_error_info(sb, function, line);
673 }
ac27a0ec 674 if (test_opt(sb, ERRORS_PANIC))
617ba13b 675 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 676}
677
0ff2ea7d 678void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 679{
0ff2ea7d 680 struct va_format vaf;
b31e1552
ES		 681 va_list args;
682
683 va_start(args, fmt);
0ff2ea7d
JP		 684 vaf.fmt = fmt;
685 vaf.va = &args;
686 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 687 va_end(args);
688}
689
12062ddd 690void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 691 unsigned int line, const char *fmt, ...)
ac27a0ec 692{
0ff2ea7d 693 struct va_format vaf;
ac27a0ec
DK		 694 va_list args;
695
696 va_start(args, fmt);
0ff2ea7d
JP		 697 vaf.fmt = fmt;
698 vaf.va = &args;
699 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
700 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 701 va_end(args);
702}
703
e29136f8
TT		 704void __ext4_grp_locked_error(const char *function, unsigned int line,
705 struct super_block *sb, ext4_group_t grp,
706 unsigned long ino, ext4_fsblk_t block,
707 const char *fmt, ...)
5d1b1b3f
AK		 708__releases(bitlock)
709__acquires(bitlock)
710{
0ff2ea7d 711 struct va_format vaf;
5d1b1b3f
AK		 712 va_list args;
713 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
714
1c13d5c0
TT		 715 es->s_last_error_ino = cpu_to_le32(ino);
716 es->s_last_error_block = cpu_to_le64(block);
717 __save_error_info(sb, function, line);
0ff2ea7d 718
5d1b1b3f 719 va_start(args, fmt);
0ff2ea7d
JP		 720
721 vaf.fmt = fmt;
722 vaf.va = &args;
21149d61 723 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 724 sb->s_id, function, line, grp);
725 if (ino)
0ff2ea7d 726 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 727 if (block)
0ff2ea7d
JP		 728 printk(KERN_CONT "block %llu:", (unsigned long long) block);
729 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 730 va_end(args);
731
732 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 733 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 734 return;
735 }
1c13d5c0 736
5d1b1b3f
AK		 737 ext4_unlock_group(sb, grp);
738 ext4_handle_error(sb);
739 /*
740 * We only get here in the ERRORS_RO case; relocking the group
741 * may be dangerous, but nothing bad will happen since the
742 * filesystem will have already been marked read/only and the
743 * journal has been aborted. We return 1 as a hint to callers
744 * who might what to use the return value from
25985edc 745 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 746 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
747 * aggressively from the ext4 function in question, with a
748 * more appropriate error code.
749 */
750 ext4_lock_group(sb, grp);
751 return;
752}
753
617ba13b 754void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 755{
617ba13b 756 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 757
617ba13b 758 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 759 return;
760
12062ddd 761 ext4_warning(sb,
ac27a0ec
DK		 762 "updating to rev %d because of new feature flag, "
763 "running e2fsck is recommended",
617ba13b 764 EXT4_DYNAMIC_REV);
ac27a0ec 765
617ba13b
MC		 766 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
767 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
768 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 769 /* leave es->s_feature_*compat flags alone */
770 /* es->s_uuid will be set by e2fsck if empty */
771
772 /*
773 * The rest of the superblock fields should be zero, and if not it
774 * means they are likely already in use, so leave them alone. We
775 * can leave it up to e2fsck to clean up any inconsistencies there.
776 */
777}
778
779/*
780 * Open the external journal device
781 */
b31e1552 782static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 783{
784 struct block_device *bdev;
785 char b[BDEVNAME_SIZE];
786
d4d77629 787 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 788 if (IS_ERR(bdev))
789 goto fail;
790 return bdev;
791
792fail:
b31e1552 793 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 794 __bdevname(dev, b), PTR_ERR(bdev));
795 return NULL;
796}
797
798/*
799 * Release the journal device
800 */
617ba13b 801static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 802{
e525fd89 803 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 804}
805
617ba13b 806static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 807{
808 struct block_device *bdev;
809 int ret = -ENODEV;
810
811 bdev = sbi->journal_bdev;
812 if (bdev) {
617ba13b 813 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 814 sbi->journal_bdev = NULL;
815 }
816 return ret;
817}
818
819static inline struct inode *orphan_list_entry(struct list_head *l)
820{
617ba13b 821 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 822}
823
617ba13b 824static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 825{
826 struct list_head *l;
827
b31e1552
ES		 828 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
829 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 830
831 printk(KERN_ERR "sb_info orphan list:\n");
832 list_for_each(l, &sbi->s_orphan) {
833 struct inode *inode = orphan_list_entry(l);
834 printk(KERN_ERR " "
835 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
836 inode->i_sb->s_id, inode->i_ino, inode,
837 inode->i_mode, inode->i_nlink,
838 NEXT_ORPHAN(inode));
839 }
840}
841
2b2d6d01 842static void ext4_put_super(struct super_block *sb)
ac27a0ec 843{
617ba13b
MC		 844 struct ext4_sb_info *sbi = EXT4_SB(sb);
845 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 846 int i, err;
ac27a0ec 847
857ac889 848 ext4_unregister_li_request(sb);
e0ccfd95
CH		 849 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
850
4c0425ff
MC		 851 flush_workqueue(sbi->dio_unwritten_wq);
852 destroy_workqueue(sbi->dio_unwritten_wq);
853
0390131b
FM		 854 if (sbi->s_journal) {
855 err = jbd2_journal_destroy(sbi->s_journal);
856 sbi->s_journal = NULL;
857 if (err < 0)
c67d859e 858 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 859 }
d4edac31 860
a1c6c569 861 del_timer(&sbi->s_err_report);
d4edac31
JB		 862 ext4_release_system_zone(sb);
863 ext4_mb_release(sb);
864 ext4_ext_release(sb);
865 ext4_xattr_put_super(sb);
866
ac27a0ec 867 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 868 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 869 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 870 }
58c5873a 871 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 872 ext4_commit_super(sb, 1);
873
240799cd 874 if (sbi->s_proc) {
66acdcf4 875 remove_proc_entry("options", sbi->s_proc);
9f6200bb 876 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 877 }
3197ebdb 878 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 879
880 for (i = 0; i < sbi->s_gdb_count; i++)
881 brelse(sbi->s_group_desc[i]);
f18a5f21 882 ext4_kvfree(sbi->s_group_desc);
9933fc0a 883 ext4_kvfree(sbi->s_flex_groups);
57042651 884 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 885 percpu_counter_destroy(&sbi->s_freeinodes_counter);
886 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 887 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 888 brelse(sbi->s_sbh);
889#ifdef CONFIG_QUOTA
890 for (i = 0; i < MAXQUOTAS; i++)
891 kfree(sbi->s_qf_names[i]);
892#endif
893
894 /* Debugging code just in case the in-memory inode orphan list
895 * isn't empty. The on-disk one can be non-empty if we've
896 * detected an error and taken the fs readonly, but the
897 * in-memory list had better be clean by this point. */
898 if (!list_empty(&sbi->s_orphan))
899 dump_orphan_list(sb, sbi);
900 J_ASSERT(list_empty(&sbi->s_orphan));
901
f98393a6 902 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 903 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
904 /*
905 * Invalidate the journal device's buffers. We don't want them
906 * floating about in memory - the physical journal device may
907 * hotswapped, and it breaks the `ro-after' testing code.
908 */
909 sync_blockdev(sbi->journal_bdev);
f98393a6 910 invalidate_bdev(sbi->journal_bdev);
617ba13b 911 ext4_blkdev_remove(sbi);
ac27a0ec 912 }
c5e06d10
JL		 913 if (sbi->s_mmp_tsk)
914 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 915 sb->s_fs_info = NULL;
3197ebdb
TT		 916 /*
917 * Now that we are completely done shutting down the
918 * superblock, we need to actually destroy the kobject.
919 */
3197ebdb
TT		 920 kobject_put(&sbi->s_kobj);
921 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 922 if (sbi->s_chksum_driver)
923 crypto_free_shash(sbi->s_chksum_driver);
705895b6 924 kfree(sbi->s_blockgroup_lock);
ac27a0ec 925 kfree(sbi);
ac27a0ec
DK		 926}
927
e18b890b 928static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 929
930/*
931 * Called inside transaction, so use GFP_NOFS
932 */
617ba13b 933static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 934{
617ba13b 935 struct ext4_inode_info *ei;
ac27a0ec 936
e6b4f8da 937 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 938 if (!ei)
939 return NULL;
0b8e58a1 940
ac27a0ec 941 ei->vfs_inode.i_version = 1;
a86c6181 942 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 943 INIT_LIST_HEAD(&ei->i_prealloc_list);
944 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 945 ext4_es_init_tree(&ei->i_es_tree);
946 rwlock_init(&ei->i_es_lock);
d2a17637
MC		 947 ei->i_reserved_data_blocks = 0;
948 ei->i_reserved_meta_blocks = 0;
949 ei->i_allocated_meta_blocks = 0;
9d0be502 950 ei->i_da_metadata_calc_len = 0;
7e731bc9 951 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 952 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 953#ifdef CONFIG_QUOTA
954 ei->i_reserved_quota = 0;
955#endif
8aefcd55 956 ei->jinode = NULL;
c7064ef1 957 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 958 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 959 ei->i_sync_tid = 0;
960 ei->i_datasync_tid = 0;
f7ad6d2e 961 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 962 atomic_set(&ei->i_unwritten, 0);
0b8e58a1 963
ac27a0ec
DK		 964 return &ei->vfs_inode;
965}
966
7ff9c073
TT		 967static int ext4_drop_inode(struct inode *inode)
968{
969 int drop = generic_drop_inode(inode);
970
971 trace_ext4_drop_inode(inode, drop);
972 return drop;
973}
974
fa0d7e3d
NP		 975static void ext4_i_callback(struct rcu_head *head)
976{
977 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 978 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
979}
980
617ba13b 981static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 982{
9f7dd93d 983 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 984 ext4_msg(inode->i_sb, KERN_ERR,
985 "Inode %lu (%p): orphan list check failed!",
986 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 987 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
988 EXT4_I(inode), sizeof(struct ext4_inode_info),
989 true);
990 dump_stack();
991 }
fa0d7e3d 992 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 993}
994
51cc5068 995static void init_once(void *foo)
ac27a0ec 996{
617ba13b 997 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 998
a35afb83 999 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1000 init_rwsem(&ei->xattr_sem);
0e855ac8 1001 init_rwsem(&ei->i_data_sem);
a35afb83 1002 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1003}
1004
1005static int init_inodecache(void)
1006{
617ba13b
MC		 1007 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1008 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 1009 0, (SLAB_RECLAIM_ACCOUNT|
1010 SLAB_MEM_SPREAD),
20c2df83 1011 init_once);
617ba13b 1012 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1013 return -ENOMEM;
1014 return 0;
1015}
1016
1017static void destroy_inodecache(void)
1018{
8c0a8537
KS		 1019 /*
1020 * Make sure all delayed rcu free inodes are flushed before we
1021 * destroy cache.
1022 */
1023 rcu_barrier();
617ba13b 1024 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1025}
1026
0930fcc1 1027void ext4_clear_inode(struct inode *inode)
ac27a0ec 1028{
0930fcc1 1029 invalidate_inode_buffers(inode);
dbd5768f 1030 clear_inode(inode);
9f754758 1031 dquot_drop(inode);
c2ea3fde 1032 ext4_discard_preallocations(inode);
51865fda 1033 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1034 if (EXT4_I(inode)->jinode) {
1035 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1036 EXT4_I(inode)->jinode);
1037 jbd2_free_inode(EXT4_I(inode)->jinode);
1038 EXT4_I(inode)->jinode = NULL;
1039 }
ac27a0ec
DK		 1040}
1041
1b961ac0 1042static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1043 u64 ino, u32 generation)
ac27a0ec 1044{
ac27a0ec 1045 struct inode *inode;
ac27a0ec 1046
617ba13b 1047 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1048 return ERR_PTR(-ESTALE);
617ba13b 1049 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1050 return ERR_PTR(-ESTALE);
1051
1052 /* iget isn't really right if the inode is currently unallocated!!
1053 *
617ba13b 1054 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1055 * deleted, so we should be safe.
1056 *
1057 * Currently we don't know the generation for parent directory, so
1058 * a generation of 0 means "accept any"
1059 */
1d1fe1ee
DH		 1060 inode = ext4_iget(sb, ino);
1061 if (IS_ERR(inode))
1062 return ERR_CAST(inode);
1063 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1064 iput(inode);
1065 return ERR_PTR(-ESTALE);
1066 }
1b961ac0
CH		 1067
1068 return inode;
1069}
1070
1071static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1072 int fh_len, int fh_type)
1b961ac0
CH		 1073{
1074 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1075 ext4_nfs_get_inode);
1076}
1077
1078static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1079 int fh_len, int fh_type)
1b961ac0
CH		 1080{
1081 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1082 ext4_nfs_get_inode);
ac27a0ec
DK		 1083}
1084
c39a7f84
TO		 1085/*
1086 * Try to release metadata pages (indirect blocks, directories) which are
1087 * mapped via the block device. Since these pages could have journal heads
1088 * which would prevent try_to_free_buffers() from freeing them, we must use
1089 * jbd2 layer's try_to_free_buffers() function to release them.
1090 */
0b8e58a1
AD		 1091static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1092 gfp_t wait)
c39a7f84
TO		 1093{
1094 journal_t *journal = EXT4_SB(sb)->s_journal;
1095
1096 WARN_ON(PageChecked(page));
1097 if (!page_has_buffers(page))
1098 return 0;
1099 if (journal)
1100 return jbd2_journal_try_to_free_buffers(journal, page,
1101 wait & ~__GFP_WAIT);
1102 return try_to_free_buffers(page);
1103}
1104
ac27a0ec 1105#ifdef CONFIG_QUOTA
af5bc92d 1106#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1107#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1108
617ba13b
MC		 1109static int ext4_write_dquot(struct dquot *dquot);
1110static int ext4_acquire_dquot(struct dquot *dquot);
1111static int ext4_release_dquot(struct dquot *dquot);
1112static int ext4_mark_dquot_dirty(struct dquot *dquot);
1113static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1114static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1115 struct path *path);
7c319d32
AK		 1116static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1117 int format_id);
ca0e05e4 1118static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1119static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1120static int ext4_quota_on_mount(struct super_block *sb, int type);
1121static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1122 size_t len, loff_t off);
617ba13b 1123static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1124 const char *data, size_t len, loff_t off);
7c319d32
AK		 1125static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1126 unsigned int flags);
1127static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1128
61e225dc 1129static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1130 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1131 .write_dquot = ext4_write_dquot,
1132 .acquire_dquot = ext4_acquire_dquot,
1133 .release_dquot = ext4_release_dquot,
1134 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1135 .write_info = ext4_write_info,
1136 .alloc_dquot = dquot_alloc,
1137 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1138};
1139
0d54b217 1140static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1141 .quota_on = ext4_quota_on,
ca0e05e4 1142 .quota_off = ext4_quota_off,
287a8095
CH		 1143 .quota_sync = dquot_quota_sync,
1144 .get_info = dquot_get_dqinfo,
1145 .set_info = dquot_set_dqinfo,
1146 .get_dqblk = dquot_get_dqblk,
1147 .set_dqblk = dquot_set_dqblk
ac27a0ec 1148};
7c319d32
AK		 1149
1150static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1151 .quota_on_meta = ext4_quota_on_sysfile,
1152 .quota_off = ext4_quota_off_sysfile,
1153 .quota_sync = dquot_quota_sync,
1154 .get_info = dquot_get_dqinfo,
1155 .set_info = dquot_set_dqinfo,
1156 .get_dqblk = dquot_get_dqblk,
1157 .set_dqblk = dquot_set_dqblk
1158};
ac27a0ec
DK		 1159#endif
1160
ee9b6d61 1161static const struct super_operations ext4_sops = {
617ba13b
MC		 1162 .alloc_inode = ext4_alloc_inode,
1163 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1164 .write_inode = ext4_write_inode,
1165 .dirty_inode = ext4_dirty_inode,
7ff9c073 1166 .drop_inode = ext4_drop_inode,
0930fcc1 1167 .evict_inode = ext4_evict_inode,
617ba13b 1168 .put_super = ext4_put_super,
617ba13b 1169 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1170 .freeze_fs = ext4_freeze,
1171 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1172 .statfs = ext4_statfs,
1173 .remount_fs = ext4_remount,
617ba13b 1174 .show_options = ext4_show_options,
ac27a0ec 1175#ifdef CONFIG_QUOTA
617ba13b
MC		 1176 .quota_read = ext4_quota_read,
1177 .quota_write = ext4_quota_write,
ac27a0ec 1178#endif
c39a7f84 1179 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1180};
1181
9ca92389
TT		 1182static const struct super_operations ext4_nojournal_sops = {
1183 .alloc_inode = ext4_alloc_inode,
1184 .destroy_inode = ext4_destroy_inode,
1185 .write_inode = ext4_write_inode,
1186 .dirty_inode = ext4_dirty_inode,
7ff9c073 1187 .drop_inode = ext4_drop_inode,
0930fcc1 1188 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1189 .put_super = ext4_put_super,
1190 .statfs = ext4_statfs,
1191 .remount_fs = ext4_remount,
9ca92389
TT		 1192 .show_options = ext4_show_options,
1193#ifdef CONFIG_QUOTA
1194 .quota_read = ext4_quota_read,
1195 .quota_write = ext4_quota_write,
1196#endif
1197 .bdev_try_to_free_page = bdev_try_to_free_page,
1198};
1199
39655164 1200static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1201 .fh_to_dentry = ext4_fh_to_dentry,
1202 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1203 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1204};
1205
1206enum {
1207 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1208 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1209 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1210 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1211 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1212 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1213 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1214 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1215 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1216 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1217 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1218 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1219 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1220 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1221 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1222 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1223 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1224 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1225 Opt_max_dir_size_kb,
ac27a0ec
DK		 1226};
1227
a447c093 1228static const match_table_t tokens = {
ac27a0ec
DK		 1229 {Opt_bsd_df, "bsddf"},
1230 {Opt_minix_df, "minixdf"},
1231 {Opt_grpid, "grpid"},
1232 {Opt_grpid, "bsdgroups"},
1233 {Opt_nogrpid, "nogrpid"},
1234 {Opt_nogrpid, "sysvgroups"},
1235 {Opt_resgid, "resgid=%u"},
1236 {Opt_resuid, "resuid=%u"},
1237 {Opt_sb, "sb=%u"},
1238 {Opt_err_cont, "errors=continue"},
1239 {Opt_err_panic, "errors=panic"},
1240 {Opt_err_ro, "errors=remount-ro"},
1241 {Opt_nouid32, "nouid32"},
ac27a0ec 1242 {Opt_debug, "debug"},
72578c33
TT		 1243 {Opt_removed, "oldalloc"},
1244 {Opt_removed, "orlov"},
ac27a0ec
DK		 1245 {Opt_user_xattr, "user_xattr"},
1246 {Opt_nouser_xattr, "nouser_xattr"},
1247 {Opt_acl, "acl"},
1248 {Opt_noacl, "noacl"},
e3bb52ae 1249 {Opt_noload, "norecovery"},
5a916be1 1250 {Opt_noload, "noload"},
72578c33
TT		 1251 {Opt_removed, "nobh"},
1252 {Opt_removed, "bh"},
ac27a0ec 1253 {Opt_commit, "commit=%u"},
30773840
TT		 1254 {Opt_min_batch_time, "min_batch_time=%u"},
1255 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1256 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1257 {Opt_journal_checksum, "journal_checksum"},
1258 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1259 {Opt_abort, "abort"},
1260 {Opt_data_journal, "data=journal"},
1261 {Opt_data_ordered, "data=ordered"},
1262 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1263 {Opt_data_err_abort, "data_err=abort"},
1264 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1265 {Opt_offusrjquota, "usrjquota="},
1266 {Opt_usrjquota, "usrjquota=%s"},
1267 {Opt_offgrpjquota, "grpjquota="},
1268 {Opt_grpjquota, "grpjquota=%s"},
1269 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1270 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1271 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1272 {Opt_grpquota, "grpquota"},
1273 {Opt_noquota, "noquota"},
1274 {Opt_quota, "quota"},
1275 {Opt_usrquota, "usrquota"},
1276 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1277 {Opt_barrier, "barrier"},
1278 {Opt_nobarrier, "nobarrier"},
25ec56b5 1279 {Opt_i_version, "i_version"},
c9de560d 1280 {Opt_stripe, "stripe=%u"},
64769240 1281 {Opt_delalloc, "delalloc"},
dd919b98 1282 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1283 {Opt_mblk_io_submit, "mblk_io_submit"},
1284 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1285 {Opt_block_validity, "block_validity"},
1286 {Opt_noblock_validity, "noblock_validity"},
240799cd 1287 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1288 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1289 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1290 {Opt_auto_da_alloc, "auto_da_alloc"},
1291 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1292 {Opt_dioread_nolock, "dioread_nolock"},
1293 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1294 {Opt_discard, "discard"},
1295 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1296 {Opt_init_itable, "init_itable=%u"},
1297 {Opt_init_itable, "init_itable"},
1298 {Opt_noinit_itable, "noinit_itable"},
df981d03 1299 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1300 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1301 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1302 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1303 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1304 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1305 {Opt_err, NULL},
ac27a0ec
DK		 1306};
1307
617ba13b 1308static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1309{
617ba13b 1310 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1311 char *options = (char *) *data;
1312
1313 if (!options || strncmp(options, "sb=", 3) != 0)
1314 return 1; /* Default location */
0b8e58a1 1315
ac27a0ec 1316 options += 3;
0b8e58a1 1317 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1318 sb_block = simple_strtoul(options, &options, 0);
1319 if (*options && *options != ',') {
4776004f 1320 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1321 (char *) *data);
1322 return 1;
1323 }
1324 if (*options == ',')
1325 options++;
1326 *data = (void *) options;
0b8e58a1 1327
ac27a0ec
DK		 1328 return sb_block;
1329}
1330
b3881f74 1331#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1332static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1333 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1334
56c50f11
DM		 1335#ifdef CONFIG_QUOTA
1336static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1337{
1338 struct ext4_sb_info *sbi = EXT4_SB(sb);
1339 char *qname;
1340
1341 if (sb_any_quota_loaded(sb) &&
1342 !sbi->s_qf_names[qtype]) {
1343 ext4_msg(sb, KERN_ERR,
1344 "Cannot change journaled "
1345 "quota options when quota turned on");
57f73c2c 1346 return -1;
56c50f11
DM		 1347 }
1348 qname = match_strdup(args);
1349 if (!qname) {
1350 ext4_msg(sb, KERN_ERR,
1351 "Not enough memory for storing quotafile name");
57f73c2c 1352 return -1;
56c50f11
DM		 1353 }
1354 if (sbi->s_qf_names[qtype] &&
1355 strcmp(sbi->s_qf_names[qtype], qname)) {
1356 ext4_msg(sb, KERN_ERR,
1357 "%s quota file already specified", QTYPE2NAME(qtype));
1358 kfree(qname);
57f73c2c 1359 return -1;
56c50f11
DM		 1360 }
1361 sbi->s_qf_names[qtype] = qname;
1362 if (strchr(sbi->s_qf_names[qtype], '/')) {
1363 ext4_msg(sb, KERN_ERR,
1364 "quotafile must be on filesystem root");
1365 kfree(sbi->s_qf_names[qtype]);
1366 sbi->s_qf_names[qtype] = NULL;
57f73c2c 1367 return -1;
56c50f11 1368 }
fd8c37ec 1369 set_opt(sb, QUOTA);
56c50f11
DM		 1370 return 1;
1371}
1372
1373static int clear_qf_name(struct super_block *sb, int qtype)
1374{
1375
1376 struct ext4_sb_info *sbi = EXT4_SB(sb);
1377
1378 if (sb_any_quota_loaded(sb) &&
1379 sbi->s_qf_names[qtype]) {
1380 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1381 " when quota turned on");
57f73c2c 1382 return -1;
56c50f11
DM		 1383 }
1384 /*
1385 * The space will be released later when all options are confirmed
1386 * to be correct
1387 */
1388 sbi->s_qf_names[qtype] = NULL;
1389 return 1;
1390}
1391#endif
1392
26092bf5
TT		 1393#define MOPT_SET 0x0001
1394#define MOPT_CLEAR 0x0002
1395#define MOPT_NOSUPPORT 0x0004
1396#define MOPT_EXPLICIT 0x0008
1397#define MOPT_CLEAR_ERR 0x0010
1398#define MOPT_GTE0 0x0020
ac27a0ec 1399#ifdef CONFIG_QUOTA
26092bf5
TT		 1400#define MOPT_Q 0
1401#define MOPT_QFMT 0x0040
1402#else
1403#define MOPT_Q MOPT_NOSUPPORT
1404#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1405#endif
26092bf5
TT		 1406#define MOPT_DATAJ 0x0080
1407
1408static const struct mount_opts {
1409 int token;
1410 int mount_opt;
1411 int flags;
1412} ext4_mount_opts[] = {
1413 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1414 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1415 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1416 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1417 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1418 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1419 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1420 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1421 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1422 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1423 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1424 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1425 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1426 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1427 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1428 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1429 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1430 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1431 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1432 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1433 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1434 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1435 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1436 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1437 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1438 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1439 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1440 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1441 {Opt_commit, 0, MOPT_GTE0},
1442 {Opt_max_batch_time, 0, MOPT_GTE0},
1443 {Opt_min_batch_time, 0, MOPT_GTE0},
1444 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1445 {Opt_init_itable, 0, MOPT_GTE0},
1446 {Opt_stripe, 0, MOPT_GTE0},
1447 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1448 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1449 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
26092bf5
TT		 1450 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1451 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1452#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1453 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1454 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1455#else
26092bf5
TT		 1456 {Opt_acl, 0, MOPT_NOSUPPORT},
1457 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1458#endif
26092bf5
TT		 1459 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1460 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1461 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1462 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1463 MOPT_SET | MOPT_Q},
1464 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1465 MOPT_SET | MOPT_Q},
1466 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1467 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1468 {Opt_usrjquota, 0, MOPT_Q},
1469 {Opt_grpjquota, 0, MOPT_Q},
1470 {Opt_offusrjquota, 0, MOPT_Q},
1471 {Opt_offgrpjquota, 0, MOPT_Q},
1472 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1473 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1474 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1475 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1476 {Opt_err, 0, 0}
1477};
1478
1479static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1480 substring_t *args, unsigned long *journal_devnum,
1481 unsigned int *journal_ioprio, int is_remount)
1482{
1483 struct ext4_sb_info *sbi = EXT4_SB(sb);
1484 const struct mount_opts *m;
08cefc7a
EB		 1485 kuid_t uid;
1486 kgid_t gid;
26092bf5
TT		 1487 int arg = 0;
1488
57f73c2c
TT		 1489#ifdef CONFIG_QUOTA
1490 if (token == Opt_usrjquota)
1491 return set_qf_name(sb, USRQUOTA, &args[0]);
1492 else if (token == Opt_grpjquota)
1493 return set_qf_name(sb, GRPQUOTA, &args[0]);
1494 else if (token == Opt_offusrjquota)
1495 return clear_qf_name(sb, USRQUOTA);
1496 else if (token == Opt_offgrpjquota)
1497 return clear_qf_name(sb, GRPQUOTA);
1498#endif
26092bf5
TT		 1499 if (args->from && match_int(args, &arg))
1500 return -1;
1501 switch (token) {
f7048605
TT		 1502 case Opt_noacl:
1503 case Opt_nouser_xattr:
1504 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1505 break;
26092bf5
TT		 1506 case Opt_sb:
1507 return 1; /* handled by get_sb_block() */
1508 case Opt_removed:
1509 ext4_msg(sb, KERN_WARNING,
1510 "Ignoring removed %s option", opt);
1511 return 1;
1512 case Opt_resuid:
08cefc7a
EB		 1513 uid = make_kuid(current_user_ns(), arg);
1514 if (!uid_valid(uid)) {
1515 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1516 return -1;
1517 }
1518 sbi->s_resuid = uid;
26092bf5
TT		 1519 return 1;
1520 case Opt_resgid:
08cefc7a
EB		 1521 gid = make_kgid(current_user_ns(), arg);
1522 if (!gid_valid(gid)) {
1523 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1524 return -1;
1525 }
1526 sbi->s_resgid = gid;
26092bf5
TT		 1527 return 1;
1528 case Opt_abort:
1529 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1530 return 1;
1531 case Opt_i_version:
1532 sb->s_flags |= MS_I_VERSION;
1533 return 1;
1534 case Opt_journal_dev:
1535 if (is_remount) {
1536 ext4_msg(sb, KERN_ERR,
1537 "Cannot specify journal on remount");
1538 return -1;
1539 }
1540 *journal_devnum = arg;
1541 return 1;
1542 case Opt_journal_ioprio:
1543 if (arg < 0 || arg > 7)
1544 return -1;
1545 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1546 return 1;
1547 }
1548
1549 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1550 if (token != m->token)
1551 continue;
1552 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1553 return -1;
1554 if (m->flags & MOPT_EXPLICIT)
1555 set_opt2(sb, EXPLICIT_DELALLOC);
1556 if (m->flags & MOPT_CLEAR_ERR)
1557 clear_opt(sb, ERRORS_MASK);
1558 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1559 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1560 "options when quota turned on");
1561 return -1;
1562 }
1563
1564 if (m->flags & MOPT_NOSUPPORT) {
1565 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1566 } else if (token == Opt_commit) {
1567 if (arg == 0)
1568 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1569 sbi->s_commit_interval = HZ * arg;
1570 } else if (token == Opt_max_batch_time) {
1571 if (arg == 0)
1572 arg = EXT4_DEF_MAX_BATCH_TIME;
1573 sbi->s_max_batch_time = arg;
1574 } else if (token == Opt_min_batch_time) {
1575 sbi->s_min_batch_time = arg;
1576 } else if (token == Opt_inode_readahead_blks) {
1577 if (arg > (1 << 30))
1578 return -1;
1579 if (arg && !is_power_of_2(arg)) {
b31e1552 1580 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1581 "EXT4-fs: inode_readahead_blks"
1582 " must be a power of 2");
1583 return -1;
ac27a0ec 1584 }
26092bf5
TT		 1585 sbi->s_inode_readahead_blks = arg;
1586 } else if (token == Opt_init_itable) {
1587 set_opt(sb, INIT_INODE_TABLE);
1588 if (!args->from)
1589 arg = EXT4_DEF_LI_WAIT_MULT;
1590 sbi->s_li_wait_mult = arg;
df981d03
TT		 1591 } else if (token == Opt_max_dir_size_kb) {
1592 sbi->s_max_dir_size_kb = arg;
26092bf5
TT		 1593 } else if (token == Opt_stripe) {
1594 sbi->s_stripe = arg;
1595 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1596 if (is_remount) {
eb513689
TT		 1597 if (!sbi->s_journal)
1598 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1599 else if (test_opt(sb, DATA_FLAGS) !=
1600 m->mount_opt) {
b31e1552 1601 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1602 "Cannot change data mode on remount");
1603 return -1;
ac27a0ec
DK		 1604 }
1605 } else {
fd8c37ec 1606 clear_opt(sb, DATA_FLAGS);
26092bf5 1607 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1608 }
ac27a0ec 1609#ifdef CONFIG_QUOTA
26092bf5 1610 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1611 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1612 sbi->s_jquota_fmt != m->mount_opt) {
1613 ext4_msg(sb, KERN_ERR, "Cannot "
1614 "change journaled quota options "
1615 "when quota turned on");
1616 return -1;
ac27a0ec 1617 }
26092bf5 1618 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1619#endif
26092bf5
TT		 1620 } else {
1621 if (!args->from)
1622 arg = 1;
1623 if (m->flags & MOPT_CLEAR)
1624 arg = !arg;
1625 else if (unlikely(!(m->flags & MOPT_SET))) {
1626 ext4_msg(sb, KERN_WARNING,
1627 "buggy handling of option %s", opt);
1628 WARN_ON(1);
1629 return -1;
3197ebdb 1630 }
26092bf5
TT		 1631 if (arg != 0)
1632 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1633 else
26092bf5 1634 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1635 }
26092bf5
TT		 1636 return 1;
1637 }
1638 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1639 "or missing value", opt);
1640 return -1;
1641}
1642
1643static int parse_options(char *options, struct super_block *sb,
1644 unsigned long *journal_devnum,
1645 unsigned int *journal_ioprio,
1646 int is_remount)
1647{
1648 struct ext4_sb_info *sbi = EXT4_SB(sb);
1649 char *p;
1650 substring_t args[MAX_OPT_ARGS];
1651 int token;
1652
1653 if (!options)
1654 return 1;
1655
1656 while ((p = strsep(&options, ",")) != NULL) {
1657 if (!*p)
1658 continue;
1659 /*
1660 * Initialize args struct so we know whether arg was
1661 * found; some options take optional arguments.
1662 */
caecd0af 1663 args[0].to = args[0].from = NULL;
26092bf5
TT		 1664 token = match_token(p, tokens, args);
1665 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1666 journal_ioprio, is_remount) < 0)
1667 return 0;
ac27a0ec
DK		 1668 }
1669#ifdef CONFIG_QUOTA
1670 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1671 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1672 clear_opt(sb, USRQUOTA);
ac27a0ec 1673
482a7425 1674 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1675 clear_opt(sb, GRPQUOTA);
ac27a0ec 1676
56c50f11 1677 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1678 ext4_msg(sb, KERN_ERR, "old and new quota "
1679 "format mixing");
ac27a0ec
DK		 1680 return 0;
1681 }
1682
1683 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1684 ext4_msg(sb, KERN_ERR, "journaled quota format "
1685 "not specified");
ac27a0ec
DK		 1686 return 0;
1687 }
1688 } else {
1689 if (sbi->s_jquota_fmt) {
b31e1552 1690 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1691 "specified with no journaling "
b31e1552 1692 "enabled");
ac27a0ec
DK		 1693 return 0;
1694 }
1695 }
1696#endif
261cb20c
JK		 1697 if (test_opt(sb, DIOREAD_NOLOCK)) {
1698 int blocksize =
1699 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1700
1701 if (blocksize < PAGE_CACHE_SIZE) {
1702 ext4_msg(sb, KERN_ERR, "can't mount with "
1703 "dioread_nolock if block size != PAGE_SIZE");
1704 return 0;
1705 }
1706 }
ac27a0ec
DK		 1707 return 1;
1708}
1709
2adf6da8
TT		 1710static inline void ext4_show_quota_options(struct seq_file *seq,
1711 struct super_block *sb)
1712{
1713#if defined(CONFIG_QUOTA)
1714 struct ext4_sb_info *sbi = EXT4_SB(sb);
1715
1716 if (sbi->s_jquota_fmt) {
1717 char *fmtname = "";
1718
1719 switch (sbi->s_jquota_fmt) {
1720 case QFMT_VFS_OLD:
1721 fmtname = "vfsold";
1722 break;
1723 case QFMT_VFS_V0:
1724 fmtname = "vfsv0";
1725 break;
1726 case QFMT_VFS_V1:
1727 fmtname = "vfsv1";
1728 break;
1729 }
1730 seq_printf(seq, ",jqfmt=%s", fmtname);
1731 }
1732
1733 if (sbi->s_qf_names[USRQUOTA])
1734 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1735
1736 if (sbi->s_qf_names[GRPQUOTA])
1737 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1738
1739 if (test_opt(sb, USRQUOTA))
1740 seq_puts(seq, ",usrquota");
1741
1742 if (test_opt(sb, GRPQUOTA))
1743 seq_puts(seq, ",grpquota");
1744#endif
1745}
1746
5a916be1
TT		 1747static const char *token2str(int token)
1748{
50df9fd5 1749 const struct match_token *t;
5a916be1
TT		 1750
1751 for (t = tokens; t->token != Opt_err; t++)
1752 if (t->token == token && !strchr(t->pattern, '='))
1753 break;
1754 return t->pattern;
1755}
1756
2adf6da8
TT		 1757/*
1758 * Show an option if
1759 * - it's set to a non-default value OR
1760 * - if the per-sb default is different from the global default
1761 */
66acdcf4
TT		 1762static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1763 int nodefs)
2adf6da8 1764{
2adf6da8
TT		 1765 struct ext4_sb_info *sbi = EXT4_SB(sb);
1766 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1767 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1768 const struct mount_opts *m;
66acdcf4 1769 char sep = nodefs ? '\n' : ',';
2adf6da8 1770
66acdcf4
TT		 1771#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1772#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1773
1774 if (sbi->s_sb_block != 1)
5a916be1
TT		 1775 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1776
1777 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1778 int want_set = m->flags & MOPT_SET;
1779 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1780 (m->flags & MOPT_CLEAR_ERR))
1781 continue;
66acdcf4 1782 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1783 continue; /* skip if same as the default */
1784 if ((want_set &&
1785 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1786 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1787 continue; /* select Opt_noFoo vs Opt_Foo */
1788 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1789 }
5a916be1 1790
08cefc7a 1791 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1792 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1793 SEQ_OPTS_PRINT("resuid=%u",
1794 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1795 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1796 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1797 SEQ_OPTS_PRINT("resgid=%u",
1798 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1799 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1800 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1801 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1802 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1803 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1804 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1805 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1806 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1807 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1808 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1809 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1810 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1811 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1812 if (sb->s_flags & MS_I_VERSION)
5a916be1 1813 SEQ_OPTS_PUTS("i_version");
66acdcf4 1814 if (nodefs || sbi->s_stripe)
5a916be1 1815 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1816 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1817 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1818 SEQ_OPTS_PUTS("data=journal");
1819 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1820 SEQ_OPTS_PUTS("data=ordered");
1821 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1822 SEQ_OPTS_PUTS("data=writeback");
1823 }
66acdcf4
TT		 1824 if (nodefs ||
1825 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1826 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1827 sbi->s_inode_readahead_blks);
2adf6da8 1828
66acdcf4
TT		 1829 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1830 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1831 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1832 if (nodefs || sbi->s_max_dir_size_kb)
1833 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1834
1835 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1836 return 0;
1837}
1838
66acdcf4
TT		 1839static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1840{
1841 return _ext4_show_options(seq, root->d_sb, 0);
1842}
1843
1844static int options_seq_show(struct seq_file *seq, void *offset)
1845{
1846 struct super_block *sb = seq->private;
1847 int rc;
1848
1849 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1850 rc = _ext4_show_options(seq, sb, 1);
1851 seq_puts(seq, "\n");
1852 return rc;
1853}
1854
1855static int options_open_fs(struct inode *inode, struct file *file)
1856{
1857 return single_open(file, options_seq_show, PDE(inode)->data);
1858}
1859
1860static const struct file_operations ext4_seq_options_fops = {
1861 .owner = THIS_MODULE,
1862 .open = options_open_fs,
1863 .read = seq_read,
1864 .llseek = seq_lseek,
1865 .release = single_release,
1866};
1867
617ba13b 1868static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1869 int read_only)
1870{
617ba13b 1871 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1872 int res = 0;
1873
617ba13b 1874 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1875 ext4_msg(sb, KERN_ERR, "revision level too high, "
1876 "forcing read-only mode");
ac27a0ec
DK		 1877 res = MS_RDONLY;
1878 }
1879 if (read_only)
281b5995 1880 goto done;
617ba13b 1881 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1882 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1883 "running e2fsck is recommended");
617ba13b 1884 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1885 ext4_msg(sb, KERN_WARNING,
1886 "warning: mounting fs with errors, "
1887 "running e2fsck is recommended");
ed3ce80a 1888 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1889 le16_to_cpu(es->s_mnt_count) >=
1890 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1891 ext4_msg(sb, KERN_WARNING,
1892 "warning: maximal mount count reached, "
1893 "running e2fsck is recommended");
ac27a0ec
DK		 1894 else if (le32_to_cpu(es->s_checkinterval) &&
1895 (le32_to_cpu(es->s_lastcheck) +
1896 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1897 ext4_msg(sb, KERN_WARNING,
1898 "warning: checktime reached, "
1899 "running e2fsck is recommended");
0b8e58a1 1900 if (!sbi->s_journal)
0390131b 1901 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1902 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1903 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1904 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1905 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1906 ext4_update_dynamic_rev(sb);
0390131b
FM		 1907 if (sbi->s_journal)
1908 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1909
e2d67052 1910 ext4_commit_super(sb, 1);
281b5995 1911done:
ac27a0ec 1912 if (test_opt(sb, DEBUG))
a9df9a49 1913 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1914 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1915 sb->s_blocksize,
1916 sbi->s_groups_count,
617ba13b
MC		 1917 EXT4_BLOCKS_PER_GROUP(sb),
1918 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1919 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1920
7abc52c2 1921 cleancache_init_fs(sb);
ac27a0ec
DK		 1922 return res;
1923}
1924
117fff10
TT		 1925int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1926{
1927 struct ext4_sb_info *sbi = EXT4_SB(sb);
1928 struct flex_groups *new_groups;
1929 int size;
1930
1931 if (!sbi->s_log_groups_per_flex)
1932 return 0;
1933
1934 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1935 if (size <= sbi->s_flex_groups_allocated)
1936 return 0;
1937
1938 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1939 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1940 if (!new_groups) {
1941 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1942 size / (int) sizeof(struct flex_groups));
1943 return -ENOMEM;
1944 }
1945
1946 if (sbi->s_flex_groups) {
1947 memcpy(new_groups, sbi->s_flex_groups,
1948 (sbi->s_flex_groups_allocated *
1949 sizeof(struct flex_groups)));
1950 ext4_kvfree(sbi->s_flex_groups);
1951 }
1952 sbi->s_flex_groups = new_groups;
1953 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1954 return 0;
1955}
1956
772cb7c8
JS		 1957static int ext4_fill_flex_info(struct super_block *sb)
1958{
1959 struct ext4_sb_info *sbi = EXT4_SB(sb);
1960 struct ext4_group_desc *gdp = NULL;
772cb7c8 1961 ext4_group_t flex_group;
d50f2ab6 1962 unsigned int groups_per_flex = 0;
117fff10 1963 int i, err;
772cb7c8 1964
503358ae 1965 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1966 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1967 sbi->s_log_groups_per_flex = 0;
1968 return 1;
1969 }
76495ec1 1970 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
772cb7c8 1971
117fff10
TT		 1972 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1973 if (err)
9933fc0a 1974 goto failed;
772cb7c8 1975
772cb7c8 1976 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1977 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1978
1979 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1980 atomic_add(ext4_free_inodes_count(sb, gdp),
1981 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1982 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1983 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1984 atomic_add(ext4_used_dirs_count(sb, gdp),
1985 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1986 }
1987
1988 return 1;
1989failed:
1990 return 0;
1991}
1992
feb0ab32
DW		 1993static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1994 struct ext4_group_desc *gdp)
717d50e4 1995{
feb0ab32 1996 int offset;
717d50e4 1997 __u16 crc = 0;
feb0ab32 1998 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1999
feb0ab32
DW		 2000 if ((sbi->s_es->s_feature_ro_compat &
2001 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2002 /* Use new metadata_csum algorithm */
2003 __u16 old_csum;
2004 __u32 csum32;
2005
2006 old_csum = gdp->bg_checksum;
2007 gdp->bg_checksum = 0;
2008 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2009 sizeof(le_group));
2010 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2011 sbi->s_desc_size);
2012 gdp->bg_checksum = old_csum;
2013
2014 crc = csum32 & 0xFFFF;
2015 goto out;
717d50e4
AD		 2016 }
2017
feb0ab32
DW		 2018 /* old crc16 code */
2019 offset = offsetof(struct ext4_group_desc, bg_checksum);
2020
2021 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2022 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2023 crc = crc16(crc, (__u8 *)gdp, offset);
2024 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2025 /* for checksum of struct ext4_group_desc do the rest...*/
2026 if ((sbi->s_es->s_feature_incompat &
2027 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2028 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2029 crc = crc16(crc, (__u8 *)gdp + offset,
2030 le16_to_cpu(sbi->s_es->s_desc_size) -
2031 offset);
2032
2033out:
717d50e4
AD		 2034 return cpu_to_le16(crc);
2035}
2036
feb0ab32 2037int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2038 struct ext4_group_desc *gdp)
2039{
feb0ab32
DW		 2040 if (ext4_has_group_desc_csum(sb) &&
2041 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2042 block_group, gdp)))
717d50e4
AD		 2043 return 0;
2044
2045 return 1;
2046}
2047
feb0ab32
DW		 2048void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2049 struct ext4_group_desc *gdp)
2050{
2051 if (!ext4_has_group_desc_csum(sb))
2052 return;
2053 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2054}
2055
ac27a0ec 2056/* Called at mount-time, super-block is locked */
bfff6873
LC		 2057static int ext4_check_descriptors(struct super_block *sb,
2058 ext4_group_t *first_not_zeroed)
ac27a0ec 2059{
617ba13b
MC		 2060 struct ext4_sb_info *sbi = EXT4_SB(sb);
2061 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2062 ext4_fsblk_t last_block;
bd81d8ee
LV		 2063 ext4_fsblk_t block_bitmap;
2064 ext4_fsblk_t inode_bitmap;
2065 ext4_fsblk_t inode_table;
ce421581 2066 int flexbg_flag = 0;
bfff6873 2067 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2068
ce421581
JS		 2069 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2070 flexbg_flag = 1;
2071
af5bc92d 2072 ext4_debug("Checking group descriptors");
ac27a0ec 2073
197cd65a
AM		 2074 for (i = 0; i < sbi->s_groups_count; i++) {
2075 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2076
ce421581 2077 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2078 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2079 else
2080 last_block = first_block +
617ba13b 2081 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2082
bfff6873
LC		 2083 if ((grp == sbi->s_groups_count) &&
2084 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2085 grp = i;
2086
8fadc143 2087 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2088 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2089 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2090 "Block bitmap for group %u not in group "
b31e1552 2091 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2092 return 0;
2093 }
8fadc143 2094 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2095 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2096 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2097 "Inode bitmap for group %u not in group "
b31e1552 2098 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2099 return 0;
2100 }
8fadc143 2101 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2102 if (inode_table < first_block ||
2b2d6d01 2103 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2104 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2105 "Inode table for group %u not in group "
b31e1552 2106 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2107 return 0;
2108 }
955ce5f5 2109 ext4_lock_group(sb, i);
feb0ab32 2110 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2111 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2112 "Checksum for group %u failed (%u!=%u)",
2113 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2114 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2115 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2116 ext4_unlock_group(sb, i);
8a266467 2117 return 0;
7ee1ec4c 2118 }
717d50e4 2119 }
955ce5f5 2120 ext4_unlock_group(sb, i);
ce421581
JS		 2121 if (!flexbg_flag)
2122 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2123 }
bfff6873
LC		 2124 if (NULL != first_not_zeroed)
2125 *first_not_zeroed = grp;
ac27a0ec 2126
5dee5437
TT		 2127 ext4_free_blocks_count_set(sbi->s_es,
2128 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2129 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2130 return 1;
2131}
2132
617ba13b 2133/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2134 * the superblock) which were deleted from all directories, but held open by
2135 * a process at the time of a crash. We walk the list and try to delete these
2136 * inodes at recovery time (only with a read-write filesystem).
2137 *
2138 * In order to keep the orphan inode chain consistent during traversal (in
2139 * case of crash during recovery), we link each inode into the superblock
2140 * orphan list_head and handle it the same way as an inode deletion during
2141 * normal operation (which journals the operations for us).
2142 *
2143 * We only do an iget() and an iput() on each inode, which is very safe if we
2144 * accidentally point at an in-use or already deleted inode. The worst that
2145 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2146 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2147 * e2fsck was run on this filesystem, and it must have already done the orphan
2148 * inode cleanup for us, so we can safely abort without any further action.
2149 */
2b2d6d01
TT		 2150static void ext4_orphan_cleanup(struct super_block *sb,
2151 struct ext4_super_block *es)
ac27a0ec
DK		 2152{
2153 unsigned int s_flags = sb->s_flags;
2154 int nr_orphans = 0, nr_truncates = 0;
2155#ifdef CONFIG_QUOTA
2156 int i;
2157#endif
2158 if (!es->s_last_orphan) {
2159 jbd_debug(4, "no orphan inodes to clean up\n");
2160 return;
2161 }
2162
a8f48a95 2163 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2164 ext4_msg(sb, KERN_ERR, "write access "
2165 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2166 return;
2167 }
2168
d39195c3
AG		 2169 /* Check if feature set would not allow a r/w mount */
2170 if (!ext4_feature_set_ok(sb, 0)) {
2171 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2172 "unknown ROCOMPAT features");
2173 return;
2174 }
2175
617ba13b 2176 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2177 /* don't clear list on RO mount w/ errors */
2178 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2179 jbd_debug(1, "Errors on filesystem, "
2180 "clearing orphan list.\n");
c25f9bc6
ES		 2181 es->s_last_orphan = 0;
2182 }
ac27a0ec
DK		 2183 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2184 return;
2185 }
2186
2187 if (s_flags & MS_RDONLY) {
b31e1552 2188 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2189 sb->s_flags &= ~MS_RDONLY;
2190 }
2191#ifdef CONFIG_QUOTA
2192 /* Needed for iput() to work correctly and not trash data */
2193 sb->s_flags |= MS_ACTIVE;
2194 /* Turn on quotas so that they are updated correctly */
2195 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2196 if (EXT4_SB(sb)->s_qf_names[i]) {
2197 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2198 if (ret < 0)
b31e1552
ES		 2199 ext4_msg(sb, KERN_ERR,
2200 "Cannot turn on journaled "
2201 "quota: error %d", ret);
ac27a0ec
DK		 2202 }
2203 }
2204#endif
2205
2206 while (es->s_last_orphan) {
2207 struct inode *inode;
2208
97bd42b9
JB		 2209 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2210 if (IS_ERR(inode)) {
ac27a0ec
DK		 2211 es->s_last_orphan = 0;
2212 break;
2213 }
2214
617ba13b 2215 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2216 dquot_initialize(inode);
ac27a0ec 2217 if (inode->i_nlink) {
b31e1552
ES		 2218 ext4_msg(sb, KERN_DEBUG,
2219 "%s: truncating inode %lu to %lld bytes",
46e665e9 2220 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2221 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2222 inode->i_ino, inode->i_size);
617ba13b 2223 ext4_truncate(inode);
ac27a0ec
DK		 2224 nr_truncates++;
2225 } else {
b31e1552
ES		 2226 ext4_msg(sb, KERN_DEBUG,
2227 "%s: deleting unreferenced inode %lu",
46e665e9 2228 __func__, inode->i_ino);
ac27a0ec
DK		 2229 jbd_debug(2, "deleting unreferenced inode %lu\n",
2230 inode->i_ino);
2231 nr_orphans++;
2232 }
2233 iput(inode); /* The delete magic happens here! */
2234 }
2235
2b2d6d01 2236#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2237
2238 if (nr_orphans)
b31e1552
ES		 2239 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2240 PLURAL(nr_orphans));
ac27a0ec 2241 if (nr_truncates)
b31e1552
ES		 2242 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2243 PLURAL(nr_truncates));
ac27a0ec
DK		 2244#ifdef CONFIG_QUOTA
2245 /* Turn quotas off */
2246 for (i = 0; i < MAXQUOTAS; i++) {
2247 if (sb_dqopt(sb)->files[i])
287a8095 2248 dquot_quota_off(sb, i);
ac27a0ec
DK		 2249 }
2250#endif
2251 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2252}
0b8e58a1 2253
cd2291a4
ES		 2254/*
2255 * Maximal extent format file size.
2256 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2257 * extent format containers, within a sector_t, and within i_blocks
2258 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2259 * so that won't be a limiting factor.
2260 *
f17722f9
LC		 2261 * However there is other limiting factor. We do store extents in the form
2262 * of starting block and length, hence the resulting length of the extent
2263 * covering maximum file size must fit into on-disk format containers as
2264 * well. Given that length is always by 1 unit bigger than max unit (because
2265 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2266 *
cd2291a4
ES		 2267 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2268 */
f287a1a5 2269static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2270{
2271 loff_t res;
2272 loff_t upper_limit = MAX_LFS_FILESIZE;
2273
2274 /* small i_blocks in vfs inode? */
f287a1a5 2275 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2276 /*
90c699a9 2277 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2278 * i_block represent total blocks in 512 bytes
2279 * 32 == size of vfs inode i_blocks * 8
2280 */
2281 upper_limit = (1LL << 32) - 1;
2282
2283 /* total blocks in file system block size */
2284 upper_limit >>= (blkbits - 9);
2285 upper_limit <<= blkbits;
2286 }
2287
f17722f9
LC		 2288 /*
2289 * 32-bit extent-start container, ee_block. We lower the maxbytes
2290 * by one fs block, so ee_len can cover the extent of maximum file
2291 * size
2292 */
2293 res = (1LL << 32) - 1;
cd2291a4 2294 res <<= blkbits;
cd2291a4
ES		 2295
2296 /* Sanity check against vm- & vfs- imposed limits */
2297 if (res > upper_limit)
2298 res = upper_limit;
2299
2300 return res;
2301}
ac27a0ec 2302
ac27a0ec 2303/*
cd2291a4 2304 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2305 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2306 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2307 */
f287a1a5 2308static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2309{
617ba13b 2310 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2311 int meta_blocks;
2312 loff_t upper_limit;
0b8e58a1
AD		 2313 /* This is calculated to be the largest file size for a dense, block
2314 * mapped file such that the file's total number of 512-byte sectors,
2315 * including data and all indirect blocks, does not exceed (2^48 - 1).
2316 *
2317 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2318 * number of 512-byte sectors of the file.
0fc1b451
AK		 2319 */
2320
f287a1a5 2321 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2322 /*
90c699a9 2323 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2324 * the inode i_block field represents total file blocks in
2325 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2326 */
2327 upper_limit = (1LL << 32) - 1;
2328
2329 /* total blocks in file system block size */
2330 upper_limit >>= (bits - 9);
2331
2332 } else {
8180a562
AK		 2333 /*
2334 * We use 48 bit ext4_inode i_blocks
2335 * With EXT4_HUGE_FILE_FL set the i_blocks
2336 * represent total number of blocks in
2337 * file system block size
2338 */
0fc1b451
AK		 2339 upper_limit = (1LL << 48) - 1;
2340
0fc1b451
AK		 2341 }
2342
2343 /* indirect blocks */
2344 meta_blocks = 1;
2345 /* double indirect blocks */
2346 meta_blocks += 1 + (1LL << (bits-2));
2347 /* tripple indirect blocks */
2348 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2349
2350 upper_limit -= meta_blocks;
2351 upper_limit <<= bits;
ac27a0ec
DK		 2352
2353 res += 1LL << (bits-2);
2354 res += 1LL << (2*(bits-2));
2355 res += 1LL << (3*(bits-2));
2356 res <<= bits;
2357 if (res > upper_limit)
2358 res = upper_limit;
0fc1b451
AK		 2359
2360 if (res > MAX_LFS_FILESIZE)
2361 res = MAX_LFS_FILESIZE;
2362
ac27a0ec
DK		 2363 return res;
2364}
2365
617ba13b 2366static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2367 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2368{
617ba13b 2369 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2370 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2371 int has_super = 0;
2372
2373 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2374
617ba13b 2375 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2376 nr < first_meta_bg)
70bbb3e0 2377 return logical_sb_block + nr + 1;
ac27a0ec 2378 bg = sbi->s_desc_per_block * nr;
617ba13b 2379 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2380 has_super = 1;
0b8e58a1 2381
617ba13b 2382 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2383}
2384
c9de560d
AT		 2385/**
2386 * ext4_get_stripe_size: Get the stripe size.
2387 * @sbi: In memory super block info
2388 *
2389 * If we have specified it via mount option, then
2390 * use the mount option value. If the value specified at mount time is
2391 * greater than the blocks per group use the super block value.
2392 * If the super block value is greater than blocks per group return 0.
2393 * Allocator needs it be less than blocks per group.
2394 *
2395 */
2396static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2397{
2398 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2399 unsigned long stripe_width =
2400 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2401 int ret;
c9de560d
AT		 2402
2403 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2404 ret = sbi->s_stripe;
2405 else if (stripe_width <= sbi->s_blocks_per_group)
2406 ret = stripe_width;
2407 else if (stride <= sbi->s_blocks_per_group)
2408 ret = stride;
2409 else
2410 ret = 0;
c9de560d 2411
3eb08658
DE		 2412 /*
2413 * If the stripe width is 1, this makes no sense and
2414 * we set it to 0 to turn off stripe handling code.
2415 */
2416 if (ret <= 1)
2417 ret = 0;
c9de560d 2418
3eb08658 2419 return ret;
c9de560d 2420}
ac27a0ec 2421
3197ebdb
TT		 2422/* sysfs supprt */
2423
2424struct ext4_attr {
2425 struct attribute attr;
2426 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2427 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2428 const char *, size_t);
2429 int offset;
2430};
2431
2432static int parse_strtoul(const char *buf,
2433 unsigned long max, unsigned long *value)
2434{
2435 char *endp;
2436
e7d2860b
AGR		 2437 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2438 endp = skip_spaces(endp);
3197ebdb
TT		 2439 if (*endp || *value > max)
2440 return -EINVAL;
2441
2442 return 0;
2443}
2444
2445static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2446 struct ext4_sb_info *sbi,
2447 char *buf)
2448{
2449 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2450 (s64) EXT4_C2B(sbi,
2451 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2452}
2453
2454static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2455 struct ext4_sb_info *sbi, char *buf)
2456{
2457 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2458
f613dfcb
TT		 2459 if (!sb->s_bdev->bd_part)
2460 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2461 return snprintf(buf, PAGE_SIZE, "%lu\n",
2462 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2463 sbi->s_sectors_written_start) >> 1);
2464}
2465
2466static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2467 struct ext4_sb_info *sbi, char *buf)
2468{
2469 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2470
f613dfcb
TT		 2471 if (!sb->s_bdev->bd_part)
2472 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2473 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2474 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2475 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2476 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2477}
2478
2479static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2480 struct ext4_sb_info *sbi,
2481 const char *buf, size_t count)
2482{
2483 unsigned long t;
2484
2485 if (parse_strtoul(buf, 0x40000000, &t))
2486 return -EINVAL;
2487
5dbd571d 2488 if (t && !is_power_of_2(t))
3197ebdb
TT		 2489 return -EINVAL;
2490
2491 sbi->s_inode_readahead_blks = t;
2492 return count;
2493}
2494
2495static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2496 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2497{
2498 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2499
2500 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2501}
2502
2503static ssize_t sbi_ui_store(struct ext4_attr *a,
2504 struct ext4_sb_info *sbi,
2505 const char *buf, size_t count)
2506{
2507 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2508 unsigned long t;
2509
2510 if (parse_strtoul(buf, 0xffffffff, &t))
2511 return -EINVAL;
2512 *ui = t;
2513 return count;
2514}
2515
2c0544b2
TT		 2516static ssize_t trigger_test_error(struct ext4_attr *a,
2517 struct ext4_sb_info *sbi,
2518 const char *buf, size_t count)
2519{
2520 int len = count;
2521
2522 if (!capable(CAP_SYS_ADMIN))
2523 return -EPERM;
2524
2525 if (len && buf[len-1] == '\n')
2526 len--;
2527
2528 if (len)
2529 ext4_error(sbi->s_sb, "%.*s", len, buf);
2530 return count;
2531}
2532
3197ebdb
TT		 2533#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2534static struct ext4_attr ext4_attr_##_name = { \
2535 .attr = {.name = __stringify(_name), .mode = _mode }, \
2536 .show = _show, \
2537 .store = _store, \
2538 .offset = offsetof(struct ext4_sb_info, _elname), \
2539}
2540#define EXT4_ATTR(name, mode, show, store) \
2541static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2542
857ac889 2543#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2544#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2545#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2546#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2547 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2548#define ATTR_LIST(name) &ext4_attr_##name.attr
2549
2550EXT4_RO_ATTR(delayed_allocation_blocks);
2551EXT4_RO_ATTR(session_write_kbytes);
2552EXT4_RO_ATTR(lifetime_write_kbytes);
2553EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2554 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2555EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2556EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2557EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2558EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2559EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2560EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2561EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2562EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
67a5da56 2563EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2564EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2565
2566static struct attribute *ext4_attrs[] = {
2567 ATTR_LIST(delayed_allocation_blocks),
2568 ATTR_LIST(session_write_kbytes),
2569 ATTR_LIST(lifetime_write_kbytes),
2570 ATTR_LIST(inode_readahead_blks),
11013911 2571 ATTR_LIST(inode_goal),
3197ebdb
TT		 2572 ATTR_LIST(mb_stats),
2573 ATTR_LIST(mb_max_to_scan),
2574 ATTR_LIST(mb_min_to_scan),
2575 ATTR_LIST(mb_order2_req),
2576 ATTR_LIST(mb_stream_req),
2577 ATTR_LIST(mb_group_prealloc),
55138e0b 2578 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2579 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2580 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2581 NULL,
2582};
2583
857ac889
LC		 2584/* Features this copy of ext4 supports */
2585EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2586EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2587EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2588
2589static struct attribute *ext4_feat_attrs[] = {
2590 ATTR_LIST(lazy_itable_init),
27ee40df 2591 ATTR_LIST(batched_discard),
5e7bbef1 2592 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2593 NULL,
2594};
2595
3197ebdb
TT		 2596static ssize_t ext4_attr_show(struct kobject *kobj,
2597 struct attribute *attr, char *buf)
2598{
2599 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2600 s_kobj);
2601 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2602
2603 return a->show ? a->show(a, sbi, buf) : 0;
2604}
2605
2606static ssize_t ext4_attr_store(struct kobject *kobj,
2607 struct attribute *attr,
2608 const char *buf, size_t len)
2609{
2610 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2611 s_kobj);
2612 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2613
2614 return a->store ? a->store(a, sbi, buf, len) : 0;
2615}
2616
2617static void ext4_sb_release(struct kobject *kobj)
2618{
2619 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2620 s_kobj);
2621 complete(&sbi->s_kobj_unregister);
2622}
2623
52cf25d0 2624static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2625 .show = ext4_attr_show,
2626 .store = ext4_attr_store,
2627};
2628
2629static struct kobj_type ext4_ktype = {
2630 .default_attrs = ext4_attrs,
2631 .sysfs_ops = &ext4_attr_ops,
2632 .release = ext4_sb_release,
2633};
2634
857ac889
LC		 2635static void ext4_feat_release(struct kobject *kobj)
2636{
2637 complete(&ext4_feat->f_kobj_unregister);
2638}
2639
2640static struct kobj_type ext4_feat_ktype = {
2641 .default_attrs = ext4_feat_attrs,
2642 .sysfs_ops = &ext4_attr_ops,
2643 .release = ext4_feat_release,
2644};
2645
a13fb1a4
ES		 2646/*
2647 * Check whether this filesystem can be mounted based on
2648 * the features present and the RDONLY/RDWR mount requested.
2649 * Returns 1 if this filesystem can be mounted as requested,
2650 * 0 if it cannot be.
2651 */
2652static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2653{
2654 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2655 ext4_msg(sb, KERN_ERR,
2656 "Couldn't mount because of "
2657 "unsupported optional features (%x)",
2658 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2659 ~EXT4_FEATURE_INCOMPAT_SUPP));
2660 return 0;
2661 }
2662
2663 if (readonly)
2664 return 1;
2665
2666 /* Check that feature set is OK for a read-write mount */
2667 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2668 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2669 "unsupported optional features (%x)",
2670 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2671 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2672 return 0;
2673 }
2674 /*
2675 * Large file size enabled file system can only be mounted
2676 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2677 */
2678 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2679 if (sizeof(blkcnt_t) < sizeof(u64)) {
2680 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2681 "cannot be mounted RDWR without "
2682 "CONFIG_LBDAF");
2683 return 0;
2684 }
2685 }
bab08ab9
TT		 2686 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2687 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2688 ext4_msg(sb, KERN_ERR,
2689 "Can't support bigalloc feature without "
2690 "extents feature\n");
2691 return 0;
2692 }
7c319d32
AK		 2693
2694#ifndef CONFIG_QUOTA
2695 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2696 !readonly) {
2697 ext4_msg(sb, KERN_ERR,
2698 "Filesystem with quota feature cannot be mounted RDWR "
2699 "without CONFIG_QUOTA");
2700 return 0;
2701 }
2702#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2703 return 1;
2704}
2705
66e61a9e
TT		 2706/*
2707 * This function is called once a day if we have errors logged
2708 * on the file system
2709 */
2710static void print_daily_error_info(unsigned long arg)
2711{
2712 struct super_block *sb = (struct super_block *) arg;
2713 struct ext4_sb_info *sbi;
2714 struct ext4_super_block *es;
2715
2716 sbi = EXT4_SB(sb);
2717 es = sbi->s_es;
2718
2719 if (es->s_error_count)
2720 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2721 le32_to_cpu(es->s_error_count));
2722 if (es->s_first_error_time) {
2723 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2724 sb->s_id, le32_to_cpu(es->s_first_error_time),
2725 (int) sizeof(es->s_first_error_func),
2726 es->s_first_error_func,
2727 le32_to_cpu(es->s_first_error_line));
2728 if (es->s_first_error_ino)
2729 printk(": inode %u",
2730 le32_to_cpu(es->s_first_error_ino));
2731 if (es->s_first_error_block)
2732 printk(": block %llu", (unsigned long long)
2733 le64_to_cpu(es->s_first_error_block));
2734 printk("\n");
2735 }
2736 if (es->s_last_error_time) {
2737 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2738 sb->s_id, le32_to_cpu(es->s_last_error_time),
2739 (int) sizeof(es->s_last_error_func),
2740 es->s_last_error_func,
2741 le32_to_cpu(es->s_last_error_line));
2742 if (es->s_last_error_ino)
2743 printk(": inode %u",
2744 le32_to_cpu(es->s_last_error_ino));
2745 if (es->s_last_error_block)
2746 printk(": block %llu", (unsigned long long)
2747 le64_to_cpu(es->s_last_error_block));
2748 printk("\n");
2749 }
2750 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2751}
2752
bfff6873
LC		 2753/* Find next suitable group and run ext4_init_inode_table */
2754static int ext4_run_li_request(struct ext4_li_request *elr)
2755{
2756 struct ext4_group_desc *gdp = NULL;
2757 ext4_group_t group, ngroups;
2758 struct super_block *sb;
2759 unsigned long timeout = 0;
2760 int ret = 0;
2761
2762 sb = elr->lr_super;
2763 ngroups = EXT4_SB(sb)->s_groups_count;
2764
8e8ad8a5 2765 sb_start_write(sb);
bfff6873
LC		 2766 for (group = elr->lr_next_group; group < ngroups; group++) {
2767 gdp = ext4_get_group_desc(sb, group, NULL);
2768 if (!gdp) {
2769 ret = 1;
2770 break;
2771 }
2772
2773 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2774 break;
2775 }
2776
2777 if (group == ngroups)
2778 ret = 1;
2779
2780 if (!ret) {
2781 timeout = jiffies;
2782 ret = ext4_init_inode_table(sb, group,
2783 elr->lr_timeout ? 0 : 1);
2784 if (elr->lr_timeout == 0) {
51ce6511
LC		 2785 timeout = (jiffies - timeout) *
2786 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2787 elr->lr_timeout = timeout;
2788 }
2789 elr->lr_next_sched = jiffies + elr->lr_timeout;
2790 elr->lr_next_group = group + 1;
2791 }
8e8ad8a5 2792 sb_end_write(sb);
bfff6873
LC		 2793
2794 return ret;
2795}
2796
2797/*
2798 * Remove lr_request from the list_request and free the
4ed5c033 2799 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2800 */
2801static void ext4_remove_li_request(struct ext4_li_request *elr)
2802{
2803 struct ext4_sb_info *sbi;
2804
2805 if (!elr)
2806 return;
2807
2808 sbi = elr->lr_sbi;
2809
2810 list_del(&elr->lr_request);
2811 sbi->s_li_request = NULL;
2812 kfree(elr);
2813}
2814
2815static void ext4_unregister_li_request(struct super_block *sb)
2816{
1bb933fb
LC		 2817 mutex_lock(&ext4_li_mtx);
2818 if (!ext4_li_info) {
2819 mutex_unlock(&ext4_li_mtx);
bfff6873 2820 return;
1bb933fb 2821 }
bfff6873
LC		 2822
2823 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2824 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2825 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2826 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2827}
2828
8f1f7453
ES		 2829static struct task_struct *ext4_lazyinit_task;
2830
bfff6873
LC		 2831/*
2832 * This is the function where ext4lazyinit thread lives. It walks
2833 * through the request list searching for next scheduled filesystem.
2834 * When such a fs is found, run the lazy initialization request
2835 * (ext4_rn_li_request) and keep track of the time spend in this
2836 * function. Based on that time we compute next schedule time of
2837 * the request. When walking through the list is complete, compute
2838 * next waking time and put itself into sleep.
2839 */
2840static int ext4_lazyinit_thread(void *arg)
2841{
2842 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2843 struct list_head *pos, *n;
2844 struct ext4_li_request *elr;
4ed5c033 2845 unsigned long next_wakeup, cur;
bfff6873
LC		 2846
2847 BUG_ON(NULL == eli);
2848
bfff6873
LC		 2849cont_thread:
2850 while (true) {
2851 next_wakeup = MAX_JIFFY_OFFSET;
2852
2853 mutex_lock(&eli->li_list_mtx);
2854 if (list_empty(&eli->li_request_list)) {
2855 mutex_unlock(&eli->li_list_mtx);
2856 goto exit_thread;
2857 }
2858
2859 list_for_each_safe(pos, n, &eli->li_request_list) {
2860 elr = list_entry(pos, struct ext4_li_request,
2861 lr_request);
2862
b2c78cd0
TT		 2863 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2864 if (ext4_run_li_request(elr) != 0) {
2865 /* error, remove the lazy_init job */
2866 ext4_remove_li_request(elr);
2867 continue;
2868 }
bfff6873
LC		 2869 }
2870
2871 if (time_before(elr->lr_next_sched, next_wakeup))
2872 next_wakeup = elr->lr_next_sched;
2873 }
2874 mutex_unlock(&eli->li_list_mtx);
2875
a0acae0e 2876 try_to_freeze();
bfff6873 2877
4ed5c033
LC		 2878 cur = jiffies;
2879 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2880 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2881 cond_resched();
2882 continue;
2883 }
2884
4ed5c033
LC		 2885 schedule_timeout_interruptible(next_wakeup - cur);
2886
8f1f7453
ES		 2887 if (kthread_should_stop()) {
2888 ext4_clear_request_list();
2889 goto exit_thread;
2890 }
bfff6873
LC		 2891 }
2892
2893exit_thread:
2894 /*
2895 * It looks like the request list is empty, but we need
2896 * to check it under the li_list_mtx lock, to prevent any
2897 * additions into it, and of course we should lock ext4_li_mtx
2898 * to atomically free the list and ext4_li_info, because at
2899 * this point another ext4 filesystem could be registering
2900 * new one.
2901 */
2902 mutex_lock(&ext4_li_mtx);
2903 mutex_lock(&eli->li_list_mtx);
2904 if (!list_empty(&eli->li_request_list)) {
2905 mutex_unlock(&eli->li_list_mtx);
2906 mutex_unlock(&ext4_li_mtx);
2907 goto cont_thread;
2908 }
2909 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2910 kfree(ext4_li_info);
2911 ext4_li_info = NULL;
2912 mutex_unlock(&ext4_li_mtx);
2913
2914 return 0;
2915}
2916
2917static void ext4_clear_request_list(void)
2918{
2919 struct list_head *pos, *n;
2920 struct ext4_li_request *elr;
2921
2922 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2923 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2924 elr = list_entry(pos, struct ext4_li_request,
2925 lr_request);
2926 ext4_remove_li_request(elr);
2927 }
2928 mutex_unlock(&ext4_li_info->li_list_mtx);
2929}
2930
2931static int ext4_run_lazyinit_thread(void)
2932{
8f1f7453
ES		 2933 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2934 ext4_li_info, "ext4lazyinit");
2935 if (IS_ERR(ext4_lazyinit_task)) {
2936 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2937 ext4_clear_request_list();
bfff6873
LC		 2938 kfree(ext4_li_info);
2939 ext4_li_info = NULL;
92b97816 2940 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2941 "initialization thread\n",
2942 err);
2943 return err;
2944 }
2945 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2946 return 0;
2947}
2948
2949/*
2950 * Check whether it make sense to run itable init. thread or not.
2951 * If there is at least one uninitialized inode table, return
2952 * corresponding group number, else the loop goes through all
2953 * groups and return total number of groups.
2954 */
2955static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2956{
2957 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2958 struct ext4_group_desc *gdp = NULL;
2959
2960 for (group = 0; group < ngroups; group++) {
2961 gdp = ext4_get_group_desc(sb, group, NULL);
2962 if (!gdp)
2963 continue;
2964
2965 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2966 break;
2967 }
2968
2969 return group;
2970}
2971
2972static int ext4_li_info_new(void)
2973{
2974 struct ext4_lazy_init *eli = NULL;
2975
2976 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2977 if (!eli)
2978 return -ENOMEM;
2979
bfff6873
LC		 2980 INIT_LIST_HEAD(&eli->li_request_list);
2981 mutex_init(&eli->li_list_mtx);
2982
bfff6873
LC		 2983 eli->li_state |= EXT4_LAZYINIT_QUIT;
2984
2985 ext4_li_info = eli;
2986
2987 return 0;
2988}
2989
2990static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2991 ext4_group_t start)
2992{
2993 struct ext4_sb_info *sbi = EXT4_SB(sb);
2994 struct ext4_li_request *elr;
2995 unsigned long rnd;
2996
2997 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2998 if (!elr)
2999 return NULL;
3000
3001 elr->lr_super = sb;
3002 elr->lr_sbi = sbi;
3003 elr->lr_next_group = start;
3004
3005 /*
3006 * Randomize first schedule time of the request to
3007 * spread the inode table initialization requests
3008 * better.
3009 */
3010 get_random_bytes(&rnd, sizeof(rnd));
3011 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3012 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3013
3014 return elr;
3015}
3016
3017static int ext4_register_li_request(struct super_block *sb,
3018 ext4_group_t first_not_zeroed)
3019{
3020 struct ext4_sb_info *sbi = EXT4_SB(sb);
3021 struct ext4_li_request *elr;
3022 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3023 int ret = 0;
bfff6873 3024
51ce6511
LC		 3025 if (sbi->s_li_request != NULL) {
3026 /*
3027 * Reset timeout so it can be computed again, because
3028 * s_li_wait_mult might have changed.
3029 */
3030 sbi->s_li_request->lr_timeout = 0;
beed5ecb 3031 return 0;
51ce6511 3032 }
bfff6873
LC		 3033
3034 if (first_not_zeroed == ngroups ||
3035 (sb->s_flags & MS_RDONLY) ||
55ff3840 3036 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 3037 return 0;
bfff6873
LC		 3038
3039 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 3040 if (!elr)
3041 return -ENOMEM;
bfff6873
LC		 3042
3043 mutex_lock(&ext4_li_mtx);
3044
3045 if (NULL == ext4_li_info) {
3046 ret = ext4_li_info_new();
3047 if (ret)
3048 goto out;
3049 }
3050
3051 mutex_lock(&ext4_li_info->li_list_mtx);
3052 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3053 mutex_unlock(&ext4_li_info->li_list_mtx);
3054
3055 sbi->s_li_request = elr;
46e4690b
TM		 3056 /*
3057 * set elr to NULL here since it has been inserted to
3058 * the request_list and the removal and free of it is
3059 * handled by ext4_clear_request_list from now on.
3060 */
3061 elr = NULL;
bfff6873
LC		 3062
3063 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3064 ret = ext4_run_lazyinit_thread();
3065 if (ret)
3066 goto out;
3067 }
bfff6873 3068out:
beed5ecb
NK		 3069 mutex_unlock(&ext4_li_mtx);
3070 if (ret)
bfff6873 3071 kfree(elr);
bfff6873
LC		 3072 return ret;
3073}
3074
3075/*
3076 * We do not need to lock anything since this is called on
3077 * module unload.
3078 */
3079static void ext4_destroy_lazyinit_thread(void)
3080{
3081 /*
3082 * If thread exited earlier
3083 * there's nothing to be done.
3084 */
8f1f7453 3085 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3086 return;
3087
8f1f7453 3088 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3089}
3090
25ed6e8a
DW		 3091static int set_journal_csum_feature_set(struct super_block *sb)
3092{
3093 int ret = 1;
3094 int compat, incompat;
3095 struct ext4_sb_info *sbi = EXT4_SB(sb);
3096
3097 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3098 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3099 /* journal checksum v2 */
3100 compat = 0;
3101 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3102 } else {
3103 /* journal checksum v1 */
3104 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3105 incompat = 0;
3106 }
3107
3108 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3109 ret = jbd2_journal_set_features(sbi->s_journal,
3110 compat, 0,
3111 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3112 incompat);
3113 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3114 ret = jbd2_journal_set_features(sbi->s_journal,
3115 compat, 0,
3116 incompat);
3117 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3118 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3119 } else {
3120 jbd2_journal_clear_features(sbi->s_journal,
3121 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3122 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3123 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3124 }
3125
3126 return ret;
3127}
3128
952fc18e
TT		 3129/*
3130 * Note: calculating the overhead so we can be compatible with
3131 * historical BSD practice is quite difficult in the face of
3132 * clusters/bigalloc. This is because multiple metadata blocks from
3133 * different block group can end up in the same allocation cluster.
3134 * Calculating the exact overhead in the face of clustered allocation
3135 * requires either O(all block bitmaps) in memory or O(number of block
3136 * groups**2) in time. We will still calculate the superblock for
3137 * older file systems --- and if we come across with a bigalloc file
3138 * system with zero in s_overhead_clusters the estimate will be close to
3139 * correct especially for very large cluster sizes --- but for newer
3140 * file systems, it's better to calculate this figure once at mkfs
3141 * time, and store it in the superblock. If the superblock value is
3142 * present (even for non-bigalloc file systems), we will use it.
3143 */
3144static int count_overhead(struct super_block *sb, ext4_group_t grp,
3145 char *buf)
3146{
3147 struct ext4_sb_info *sbi = EXT4_SB(sb);
3148 struct ext4_group_desc *gdp;
3149 ext4_fsblk_t first_block, last_block, b;
3150 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3151 int s, j, count = 0;
3152
0548bbb8
TT		 3153 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3154 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3155 sbi->s_itb_per_group + 2);
3156
952fc18e
TT		 3157 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3158 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3159 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3160 for (i = 0; i < ngroups; i++) {
3161 gdp = ext4_get_group_desc(sb, i, NULL);
3162 b = ext4_block_bitmap(sb, gdp);
3163 if (b >= first_block && b <= last_block) {
3164 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3165 count++;
3166 }
3167 b = ext4_inode_bitmap(sb, gdp);
3168 if (b >= first_block && b <= last_block) {
3169 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3170 count++;
3171 }
3172 b = ext4_inode_table(sb, gdp);
3173 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3174 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3175 int c = EXT4_B2C(sbi, b - first_block);
3176 ext4_set_bit(c, buf);
3177 count++;
3178 }
3179 if (i != grp)
3180 continue;
3181 s = 0;
3182 if (ext4_bg_has_super(sb, grp)) {
3183 ext4_set_bit(s++, buf);
3184 count++;
3185 }
3186 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3187 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3188 count++;
3189 }
3190 }
3191 if (!count)
3192 return 0;
3193 return EXT4_CLUSTERS_PER_GROUP(sb) -
3194 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3195}
3196
3197/*
3198 * Compute the overhead and stash it in sbi->s_overhead
3199 */
3200int ext4_calculate_overhead(struct super_block *sb)
3201{
3202 struct ext4_sb_info *sbi = EXT4_SB(sb);
3203 struct ext4_super_block *es = sbi->s_es;
3204 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3205 ext4_fsblk_t overhead = 0;
3206 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3207
952fc18e
TT		 3208 if (!buf)
3209 return -ENOMEM;
3210
3211 /*
3212 * Compute the overhead (FS structures). This is constant
3213 * for a given filesystem unless the number of block groups
3214 * changes so we cache the previous value until it does.
3215 */
3216
3217 /*
3218 * All of the blocks before first_data_block are overhead
3219 */
3220 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3221
3222 /*
3223 * Add the overhead found in each block group
3224 */
3225 for (i = 0; i < ngroups; i++) {
3226 int blks;
3227
3228 blks = count_overhead(sb, i, buf);
3229 overhead += blks;
3230 if (blks)
3231 memset(buf, 0, PAGE_SIZE);
3232 cond_resched();
3233 }
0875a2b4
ES		 3234 /* Add the journal blocks as well */
3235 if (sbi->s_journal)
3236 overhead += EXT4_B2C(sbi, sbi->s_journal->j_maxlen);
3237
952fc18e
TT		 3238 sbi->s_overhead = overhead;
3239 smp_wmb();
3240 free_page((unsigned long) buf);
3241 return 0;
3242}
3243
2b2d6d01 3244static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3245{
d4c402d9 3246 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3247 struct buffer_head *bh;
617ba13b
MC		 3248 struct ext4_super_block *es = NULL;
3249 struct ext4_sb_info *sbi;
3250 ext4_fsblk_t block;
3251 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3252 ext4_fsblk_t logical_sb_block;
ac27a0ec 3253 unsigned long offset = 0;
ac27a0ec
DK		 3254 unsigned long journal_devnum = 0;
3255 unsigned long def_mount_opts;
3256 struct inode *root;
9f6200bb 3257 char *cp;
0390131b 3258 const char *descr;
dcc7dae3 3259 int ret = -ENOMEM;
281b5995 3260 int blocksize, clustersize;
4ec11028
TT		 3261 unsigned int db_count;
3262 unsigned int i;
281b5995 3263 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3264 __u64 blocks_count;
07aa2ea1 3265 int err = 0;
b3881f74 3266 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3267 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3268
3269 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3270 if (!sbi)
dcc7dae3 3271 goto out_free_orig;
705895b6
PE		 3272
3273 sbi->s_blockgroup_lock =
3274 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3275 if (!sbi->s_blockgroup_lock) {
3276 kfree(sbi);
dcc7dae3 3277 goto out_free_orig;
705895b6 3278 }
ac27a0ec 3279 sb->s_fs_info = sbi;
2c0544b2 3280 sbi->s_sb = sb;
240799cd 3281 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3282 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3283 if (sb->s_bdev->bd_part)
3284 sbi->s_sectors_written_start =
3285 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3286
9f6200bb
TT		 3287 /* Cleanup superblock name */
3288 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3289 *cp = '!';
3290
07aa2ea1 3291 /* -EINVAL is default */
dcc7dae3 3292 ret = -EINVAL;
617ba13b 3293 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3294 if (!blocksize) {
b31e1552 3295 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3296 goto out_fail;
3297 }
3298
3299 /*
617ba13b 3300 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3301 * block sizes. We need to calculate the offset from buffer start.
3302 */
617ba13b 3303 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3304 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3305 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3306 } else {
70bbb3e0 3307 logical_sb_block = sb_block;
ac27a0ec
DK		 3308 }
3309
70bbb3e0 3310 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3311 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3312 goto out_fail;
3313 }
3314 /*
3315 * Note: s_es must be initialized as soon as possible because
617ba13b 3316 * some ext4 macro-instructions depend on its value
ac27a0ec 3317 */
2716b802 3318 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3319 sbi->s_es = es;
3320 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3321 if (sb->s_magic != EXT4_SUPER_MAGIC)
3322 goto cantfind_ext4;
afc32f7e 3323 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3324
feb0ab32
DW		 3325 /* Warn if metadata_csum and gdt_csum are both set. */
3326 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3327 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3328 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3329 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3330 "redundant flags; please run fsck.");
3331
d25425f8
DW		 3332 /* Check for a known checksum algorithm */
3333 if (!ext4_verify_csum_type(sb, es)) {
3334 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3335 "unknown checksum algorithm.");
3336 silent = 1;
3337 goto cantfind_ext4;
3338 }
3339
0441984a
DW		 3340 /* Load the checksum driver */
3341 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3342 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3343 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3344 if (IS_ERR(sbi->s_chksum_driver)) {
3345 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3346 ret = PTR_ERR(sbi->s_chksum_driver);
3347 sbi->s_chksum_driver = NULL;
3348 goto failed_mount;
3349 }
3350 }
3351
a9c47317
DW		 3352 /* Check superblock checksum */
3353 if (!ext4_superblock_csum_verify(sb, es)) {
3354 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3355 "invalid superblock checksum. Run e2fsck?");
3356 silent = 1;
3357 goto cantfind_ext4;
3358 }
3359
3360 /* Precompute checksum seed for all metadata */
3361 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3362 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3363 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3364 sizeof(es->s_uuid));
3365
ac27a0ec
DK		 3366 /* Set defaults before we parse the mount options */
3367 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3368 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3369 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3370 set_opt(sb, DEBUG);
87f26807 3371 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3372 set_opt(sb, GRPID);
617ba13b 3373 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3374 set_opt(sb, NO_UID32);
ea663336 3375 /* xattr user namespace & acls are now defaulted on */
ea663336 3376 set_opt(sb, XATTR_USER);
03010a33 3377#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3378 set_opt(sb, POSIX_ACL);
2e7842b8 3379#endif
6fd7a467 3380 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3381 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3382 set_opt(sb, JOURNAL_DATA);
617ba13b 3383 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3384 set_opt(sb, ORDERED_DATA);
617ba13b 3385 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3386 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3387
3388 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3389 set_opt(sb, ERRORS_PANIC);
bb4f397a 3390 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3391 set_opt(sb, ERRORS_CONT);
bb4f397a 3392 else
fd8c37ec 3393 set_opt(sb, ERRORS_RO);
8b67f04a 3394 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3395 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3396 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3397 set_opt(sb, DISCARD);
ac27a0ec 3398
08cefc7a
EB		 3399 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3400 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3401 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3402 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3403 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3404
8b67f04a 3405 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3406 set_opt(sb, BARRIER);
ac27a0ec 3407
dd919b98
AK		 3408 /*
3409 * enable delayed allocation by default
3410 * Use -o nodelalloc to turn it off
3411 */
bc0b75f7 3412 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3413 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3414 set_opt(sb, DELALLOC);
dd919b98 3415
51ce6511
LC		 3416 /*
3417 * set default s_li_wait_mult for lazyinit, for the case there is
3418 * no mount option specified.
3419 */
3420 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3421
8b67f04a 3422 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3423 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3424 ext4_msg(sb, KERN_WARNING,
3425 "failed to parse options in superblock: %s",
3426 sbi->s_es->s_mount_opts);
3427 }
5a916be1 3428 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3429 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3430 &journal_ioprio, 0))
ac27a0ec
DK		 3431 goto failed_mount;
3432
56889787
TT		 3433 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3434 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3435 "with data=journal disables delayed "
3436 "allocation and O_DIRECT support!\n");
3437 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3438 ext4_msg(sb, KERN_ERR, "can't mount with "
3439 "both data=journal and delalloc");
3440 goto failed_mount;
3441 }
3442 if (test_opt(sb, DIOREAD_NOLOCK)) {
3443 ext4_msg(sb, KERN_ERR, "can't mount with "
3444 "both data=journal and delalloc");
3445 goto failed_mount;
3446 }
3447 if (test_opt(sb, DELALLOC))
3448 clear_opt(sb, DELALLOC);
3449 }
3450
ac27a0ec 3451 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3452 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3453
617ba13b
MC		 3454 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3455 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3456 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3457 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3458 ext4_msg(sb, KERN_WARNING,
3459 "feature flags set on rev 0 fs, "
3460 "running e2fsck is recommended");
469108ff 3461
2035e776
TT		 3462 if (IS_EXT2_SB(sb)) {
3463 if (ext2_feature_set_ok(sb))
3464 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3465 "using the ext4 subsystem");
3466 else {
3467 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3468 "to feature incompatibilities");
3469 goto failed_mount;
3470 }
3471 }
3472
3473 if (IS_EXT3_SB(sb)) {
3474 if (ext3_feature_set_ok(sb))
3475 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3476 "using the ext4 subsystem");
3477 else {
3478 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3479 "to feature incompatibilities");
3480 goto failed_mount;
3481 }
3482 }
3483
ac27a0ec
DK		 3484 /*
3485 * Check feature flags regardless of the revision level, since we
3486 * previously didn't change the revision level when setting the flags,
3487 * so there is a chance incompat flags are set on a rev 0 filesystem.
3488 */
a13fb1a4 3489 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3490 goto failed_mount;
a13fb1a4 3491
261cb20c 3492 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3493 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3494 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3495 ext4_msg(sb, KERN_ERR,
3496 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3497 goto failed_mount;
3498 }
3499
ac27a0ec 3500 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3501 /* Validate the filesystem blocksize */
3502 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3503 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3504 blocksize);
ac27a0ec
DK		 3505 goto failed_mount;
3506 }
3507
2b2d6d01 3508 brelse(bh);
70bbb3e0
AM		 3509 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3510 offset = do_div(logical_sb_block, blocksize);
3511 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3512 if (!bh) {
b31e1552
ES		 3513 ext4_msg(sb, KERN_ERR,
3514 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3515 goto failed_mount;
3516 }
2716b802 3517 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3518 sbi->s_es = es;
617ba13b 3519 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3520 ext4_msg(sb, KERN_ERR,
3521 "Magic mismatch, very weird!");
ac27a0ec
DK		 3522 goto failed_mount;
3523 }
3524 }
3525
a13fb1a4
ES		 3526 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3527 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3528 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3529 has_huge_files);
3530 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3531
617ba13b
MC		 3532 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3533 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3534 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3535 } else {
3536 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3537 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3538 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3539 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3540 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3541 ext4_msg(sb, KERN_ERR,
3542 "unsupported inode size: %d",
2b2d6d01 3543 sbi->s_inode_size);
ac27a0ec
DK		 3544 goto failed_mount;
3545 }
ef7f3835
KS		 3546 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3547 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3548 }
0b8e58a1 3549
0d1ee42f
AR		 3550 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3551 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3552 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3553 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3554 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3555 ext4_msg(sb, KERN_ERR,
3556 "unsupported descriptor size %lu",
0d1ee42f
AR		 3557 sbi->s_desc_size);
3558 goto failed_mount;
3559 }
3560 } else
3561 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3562
ac27a0ec 3563 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3564 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3565 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3566 goto cantfind_ext4;
0b8e58a1 3567
617ba13b 3568 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3569 if (sbi->s_inodes_per_block == 0)
617ba13b 3570 goto cantfind_ext4;
ac27a0ec
DK		 3571 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3572 sbi->s_inodes_per_block;
0d1ee42f 3573 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3574 sbi->s_sbh = bh;
3575 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3576 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3577 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3578
2b2d6d01 3579 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3580 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3581 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3582 i = le32_to_cpu(es->s_flags);
3583 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3584 sbi->s_hash_unsigned = 3;
3585 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3586#ifdef __CHAR_UNSIGNED__
3587 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3588 sbi->s_hash_unsigned = 3;
3589#else
3590 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3591#endif
f99b2589 3592 }
ac27a0ec 3593
281b5995
TT		 3594 /* Handle clustersize */
3595 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3596 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3597 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3598 if (has_bigalloc) {
3599 if (clustersize < blocksize) {
3600 ext4_msg(sb, KERN_ERR,
3601 "cluster size (%d) smaller than "
3602 "block size (%d)", clustersize, blocksize);
3603 goto failed_mount;
3604 }
3605 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3606 le32_to_cpu(es->s_log_block_size);
3607 sbi->s_clusters_per_group =
3608 le32_to_cpu(es->s_clusters_per_group);
3609 if (sbi->s_clusters_per_group > blocksize * 8) {
3610 ext4_msg(sb, KERN_ERR,
3611 "#clusters per group too big: %lu",
3612 sbi->s_clusters_per_group);
3613 goto failed_mount;
3614 }
3615 if (sbi->s_blocks_per_group !=
3616 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3617 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3618 "clusters per group (%lu) inconsistent",
3619 sbi->s_blocks_per_group,
3620 sbi->s_clusters_per_group);
3621 goto failed_mount;
3622 }
3623 } else {
3624 if (clustersize != blocksize) {
3625 ext4_warning(sb, "fragment/cluster size (%d) != "
3626 "block size (%d)", clustersize,
3627 blocksize);
3628 clustersize = blocksize;
3629 }
3630 if (sbi->s_blocks_per_group > blocksize * 8) {
3631 ext4_msg(sb, KERN_ERR,
3632 "#blocks per group too big: %lu",
3633 sbi->s_blocks_per_group);
3634 goto failed_mount;
3635 }
3636 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3637 sbi->s_cluster_bits = 0;
ac27a0ec 3638 }
281b5995
TT		 3639 sbi->s_cluster_ratio = clustersize / blocksize;
3640
ac27a0ec 3641 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3642 ext4_msg(sb, KERN_ERR,
3643 "#inodes per group too big: %lu",
2b2d6d01 3644 sbi->s_inodes_per_group);
ac27a0ec
DK		 3645 goto failed_mount;
3646 }
3647
bf43d84b
ES		 3648 /*
3649 * Test whether we have more sectors than will fit in sector_t,
3650 * and whether the max offset is addressable by the page cache.
3651 */
5a9ae68a 3652 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3653 ext4_blocks_count(es));
5a9ae68a 3654 if (err) {
b31e1552 3655 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3656 " too large to mount safely on this system");
ac27a0ec 3657 if (sizeof(sector_t) < 8)
90c699a9 3658 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3659 goto failed_mount;
3660 }
3661
617ba13b
MC		 3662 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3663 goto cantfind_ext4;
e7c95593 3664
0f2ddca6
FTN		 3665 /* check blocks count against device size */
3666 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3667 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3668 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3669 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3670 ext4_blocks_count(es), blocks_count);
3671 goto failed_mount;
3672 }
3673
0b8e58a1
AD		 3674 /*
3675 * It makes no sense for the first data block to be beyond the end
3676 * of the filesystem.
3677 */
3678 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3679 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3680 "block %u is beyond end of filesystem (%llu)",
3681 le32_to_cpu(es->s_first_data_block),
3682 ext4_blocks_count(es));
e7c95593
ES		 3683 goto failed_mount;
3684 }
bd81d8ee
LV		 3685 blocks_count = (ext4_blocks_count(es) -
3686 le32_to_cpu(es->s_first_data_block) +
3687 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3688 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3689 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3690 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3691 "(block count %llu, first data block %u, "
b31e1552 3692 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3693 ext4_blocks_count(es),
3694 le32_to_cpu(es->s_first_data_block),
3695 EXT4_BLOCKS_PER_GROUP(sb));
3696 goto failed_mount;
3697 }
bd81d8ee 3698 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3699 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3700 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3701 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3702 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3703 sbi->s_group_desc = ext4_kvmalloc(db_count *
3704 sizeof(struct buffer_head *),
3705 GFP_KERNEL);
ac27a0ec 3706 if (sbi->s_group_desc == NULL) {
b31e1552 3707 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3708 ret = -ENOMEM;
ac27a0ec
DK		 3709 goto failed_mount;
3710 }
3711
9f6200bb
TT		 3712 if (ext4_proc_root)
3713 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3714
66acdcf4
TT		 3715 if (sbi->s_proc)
3716 proc_create_data("options", S_IRUGO, sbi->s_proc,
3717 &ext4_seq_options_fops, sb);
3718
705895b6 3719 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3720
3721 for (i = 0; i < db_count; i++) {
70bbb3e0 3722 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3723 sbi->s_group_desc[i] = sb_bread(sb, block);
3724 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3725 ext4_msg(sb, KERN_ERR,
3726 "can't read group descriptor %d", i);
ac27a0ec
DK		 3727 db_count = i;
3728 goto failed_mount2;
3729 }
3730 }
bfff6873 3731 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3732 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3733 goto failed_mount2;
3734 }
772cb7c8
JS		 3735 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3736 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3737 ext4_msg(sb, KERN_ERR,
3738 "unable to initialize "
3739 "flex_bg meta info!");
772cb7c8
JS		 3740 goto failed_mount2;
3741 }
3742
ac27a0ec
DK		 3743 sbi->s_gdb_count = db_count;
3744 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3745 spin_lock_init(&sbi->s_next_gen_lock);
3746
04496411
TM		 3747 init_timer(&sbi->s_err_report);
3748 sbi->s_err_report.function = print_daily_error_info;
3749 sbi->s_err_report.data = (unsigned long) sb;
3750
57042651 3751 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3752 ext4_count_free_clusters(sb));
ce7e010a
TT		 3753 if (!err) {
3754 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3755 ext4_count_free_inodes(sb));
3756 }
3757 if (!err) {
3758 err = percpu_counter_init(&sbi->s_dirs_counter,
3759 ext4_count_dirs(sb));
3760 }
3761 if (!err) {
57042651 3762 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3763 }
3764 if (err) {
3765 ext4_msg(sb, KERN_ERR, "insufficient memory");
3766 goto failed_mount3;
3767 }
3768
c9de560d 3769 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3770 sbi->s_max_writeback_mb_bump = 128;
67a5da56 3771 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3772
ac27a0ec
DK		 3773 /*
3774 * set up enough so that it can read an inode
3775 */
9ca92389
TT		 3776 if (!test_opt(sb, NOLOAD) &&
3777 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3778 sb->s_op = &ext4_sops;
3779 else
3780 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3781 sb->s_export_op = &ext4_export_ops;
3782 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3783#ifdef CONFIG_QUOTA
617ba13b
MC		 3784 sb->s_qcop = &ext4_qctl_operations;
3785 sb->dq_op = &ext4_quota_operations;
7c319d32
AK		 3786
3787 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3788 /* Use qctl operations for hidden quota files. */
3789 sb->s_qcop = &ext4_qctl_sysfile_operations;
3790 }
ac27a0ec 3791#endif
f2fa2ffc
AK		 3792 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3793
ac27a0ec 3794 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3795 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3796
3797 sb->s_root = NULL;
3798
3799 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3800 EXT4_HAS_INCOMPAT_FEATURE(sb,
3801 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3802
c5e06d10
JL		 3803 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3804 !(sb->s_flags & MS_RDONLY))
3805 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3806 goto failed_mount3;
3807
ac27a0ec
DK		 3808 /*
3809 * The first inode we look at is the journal inode. Don't try
3810 * root first: it may be modified in the journal!
3811 */
3812 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3813 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3814 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3815 goto failed_mount3;
0390131b
FM		 3816 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3817 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3818 ext4_msg(sb, KERN_ERR, "required journal recovery "
3819 "suppressed and not mounted read-only");
744692dc 3820 goto failed_mount_wq;
ac27a0ec 3821 } else {
fd8c37ec 3822 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3823 sbi->s_journal = NULL;
3824 needs_recovery = 0;
3825 goto no_journal;
ac27a0ec
DK		 3826 }
3827
f32aaf2d 3828 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3829 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3830 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3831 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3832 goto failed_mount_wq;
eb40a09c
JS		 3833 }
3834
25ed6e8a
DW		 3835 if (!set_journal_csum_feature_set(sb)) {
3836 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3837 "feature set");
3838 goto failed_mount_wq;
d4da6c9c 3839 }
818d276c 3840
ac27a0ec
DK		 3841 /* We have now updated the journal if required, so we can
3842 * validate the data journaling mode. */
3843 switch (test_opt(sb, DATA_FLAGS)) {
3844 case 0:
3845 /* No mode set, assume a default based on the journal
63f57933
AM		 3846 * capabilities: ORDERED_DATA if the journal can
3847 * cope, else JOURNAL_DATA
3848 */
dab291af
MC		 3849 if (jbd2_journal_check_available_features
3850 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3851 set_opt(sb, ORDERED_DATA);
ac27a0ec 3852 else
fd8c37ec 3853 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3854 break;
3855
617ba13b
MC		 3856 case EXT4_MOUNT_ORDERED_DATA:
3857 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3858 if (!jbd2_journal_check_available_features
3859 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3860 ext4_msg(sb, KERN_ERR, "Journal does not support "
3861 "requested data journaling mode");
744692dc 3862 goto failed_mount_wq;
ac27a0ec
DK		 3863 }
3864 default:
3865 break;
3866 }
b3881f74 3867 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3868
18aadd47
BJ		 3869 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3870
ce7e010a
TT		 3871 /*
3872 * The journal may have updated the bg summary counts, so we
3873 * need to update the global counters.
3874 */
57042651 3875 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3876 ext4_count_free_clusters(sb));
ce7e010a
TT		 3877 percpu_counter_set(&sbi->s_freeinodes_counter,
3878 ext4_count_free_inodes(sb));
3879 percpu_counter_set(&sbi->s_dirs_counter,
3880 ext4_count_dirs(sb));
57042651 3881 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3882
ce7e010a 3883no_journal:
952fc18e
TT		 3884 /*
3885 * Get the # of file system overhead blocks from the
3886 * superblock if present.
3887 */
3888 if (es->s_overhead_clusters)
3889 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3890 else {
07aa2ea1
LC		 3891 err = ext4_calculate_overhead(sb);
3892 if (err)
952fc18e
TT		 3893 goto failed_mount_wq;
3894 }
3895
fd89d5f2
TH		 3896 /*
3897 * The maximum number of concurrent works can be high and
3898 * concurrency isn't really necessary. Limit it to 1.
3899 */
3900 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3901 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3902 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3903 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
07aa2ea1 3904 ret = -ENOMEM;
4c0425ff
MC		 3905 goto failed_mount_wq;
3906 }
3907
ac27a0ec 3908 /*
dab291af 3909 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3910 * so we can safely mount the rest of the filesystem now.
3911 */
3912
1d1fe1ee
DH		 3913 root = ext4_iget(sb, EXT4_ROOT_INO);
3914 if (IS_ERR(root)) {
b31e1552 3915 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3916 ret = PTR_ERR(root);
32a9bb57 3917 root = NULL;
ac27a0ec
DK		 3918 goto failed_mount4;
3919 }
3920 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3921 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3922 iput(root);
ac27a0ec
DK		 3923 goto failed_mount4;
3924 }
48fde701 3925 sb->s_root = d_make_root(root);
1d1fe1ee 3926 if (!sb->s_root) {
b31e1552 3927 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3928 ret = -ENOMEM;
3929 goto failed_mount4;
3930 }
ac27a0ec 3931
7e84b621
ES		 3932 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3933 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3934
3935 /* determine the minimum size of new large inodes, if present */
3936 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3937 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3938 EXT4_GOOD_OLD_INODE_SIZE;
3939 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3940 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3941 if (sbi->s_want_extra_isize <
3942 le16_to_cpu(es->s_want_extra_isize))
3943 sbi->s_want_extra_isize =
3944 le16_to_cpu(es->s_want_extra_isize);
3945 if (sbi->s_want_extra_isize <
3946 le16_to_cpu(es->s_min_extra_isize))
3947 sbi->s_want_extra_isize =
3948 le16_to_cpu(es->s_min_extra_isize);
3949 }
3950 }
3951 /* Check if enough inode space is available */
3952 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3953 sbi->s_inode_size) {
3954 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3955 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3956 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3957 "available");
ef7f3835
KS		 3958 }
3959
6fd058f7
TT		 3960 err = ext4_setup_system_zone(sb);
3961 if (err) {
b31e1552 3962 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3963 "zone (%d)", err);
94bf608a 3964 goto failed_mount4a;
6fd058f7
TT		 3965 }
3966
c2774d84 3967 ext4_ext_init(sb);
9d99012f 3968 err = ext4_mb_init(sb);
c2774d84 3969 if (err) {
421f91d2 3970 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3971 err);
dcf2d804 3972 goto failed_mount5;
c2774d84
AK		 3973 }
3974
bfff6873
LC		 3975 err = ext4_register_li_request(sb, first_not_zeroed);
3976 if (err)
dcf2d804 3977 goto failed_mount6;
bfff6873 3978
3197ebdb
TT		 3979 sbi->s_kobj.kset = ext4_kset;
3980 init_completion(&sbi->s_kobj_unregister);
3981 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3982 "%s", sb->s_id);
dcf2d804
TM		 3983 if (err)
3984 goto failed_mount7;
3197ebdb 3985
617ba13b
MC		 3986 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3987 ext4_orphan_cleanup(sb, es);
3988 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3989 if (needs_recovery) {
b31e1552 3990 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3991 ext4_mark_recovery_complete(sb, es);
3992 }
3993 if (EXT4_SB(sb)->s_journal) {
3994 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3995 descr = " journalled data mode";
3996 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3997 descr = " ordered data mode";
3998 else
3999 descr = " writeback data mode";
4000 } else
4001 descr = "out journal";
4002
7c319d32
AK		 4003#ifdef CONFIG_QUOTA
4004 /* Enable quota usage during mount. */
4005 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4006 !(sb->s_flags & MS_RDONLY)) {
07aa2ea1
LC		 4007 err = ext4_enable_quotas(sb);
4008 if (err)
7c319d32
AK		 4009 goto failed_mount7;
4010 }
4011#endif /* CONFIG_QUOTA */
4012
79add3a3
LC		 4013 if (test_opt(sb, DISCARD)) {
4014 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4015 if (!blk_queue_discard(q))
4016 ext4_msg(sb, KERN_WARNING,
4017 "mounting with \"discard\" option, but "
4018 "the device does not support discard");
4019 }
4020
d4c402d9 4021 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4022 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4023 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4024
66e61a9e
TT		 4025 if (es->s_error_count)
4026 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4027
d4c402d9 4028 kfree(orig_data);
ac27a0ec
DK		 4029 return 0;
4030
617ba13b 4031cantfind_ext4:
ac27a0ec 4032 if (!silent)
b31e1552 4033 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4034 goto failed_mount;
4035
dcf2d804
TM		 4036failed_mount7:
4037 ext4_unregister_li_request(sb);
4038failed_mount6:
dcf2d804 4039 ext4_mb_release(sb);
94bf608a
AV		 4040failed_mount5:
4041 ext4_ext_release(sb);
dcf2d804 4042 ext4_release_system_zone(sb);
94bf608a
AV		 4043failed_mount4a:
4044 dput(sb->s_root);
32a9bb57 4045 sb->s_root = NULL;
94bf608a 4046failed_mount4:
b31e1552 4047 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 4048 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4049failed_mount_wq:
0390131b
FM		 4050 if (sbi->s_journal) {
4051 jbd2_journal_destroy(sbi->s_journal);
4052 sbi->s_journal = NULL;
4053 }
ac27a0ec 4054failed_mount3:
04496411 4055 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4056 if (sbi->s_flex_groups)
4057 ext4_kvfree(sbi->s_flex_groups);
57042651 4058 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4059 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4060 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4061 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 4062 if (sbi->s_mmp_tsk)
4063 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4064failed_mount2:
4065 for (i = 0; i < db_count; i++)
4066 brelse(sbi->s_group_desc[i]);
f18a5f21 4067 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4068failed_mount:
0441984a
DW		 4069 if (sbi->s_chksum_driver)
4070 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4071 if (sbi->s_proc) {
66acdcf4 4072 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4073 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4074 }
ac27a0ec
DK		 4075#ifdef CONFIG_QUOTA
4076 for (i = 0; i < MAXQUOTAS; i++)
4077 kfree(sbi->s_qf_names[i]);
4078#endif
617ba13b 4079 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4080 brelse(bh);
4081out_fail:
4082 sb->s_fs_info = NULL;
f6830165 4083 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4084 kfree(sbi);
dcc7dae3 4085out_free_orig:
d4c402d9 4086 kfree(orig_data);
07aa2ea1 4087 return err ? err : ret;
ac27a0ec
DK		 4088}
4089
4090/*
4091 * Setup any per-fs journal parameters now. We'll do this both on
4092 * initial mount, once the journal has been initialised but before we've
4093 * done any recovery; and again on any subsequent remount.
4094 */
617ba13b 4095static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4096{
617ba13b 4097 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4098
30773840
TT		 4099 journal->j_commit_interval = sbi->s_commit_interval;
4100 journal->j_min_batch_time = sbi->s_min_batch_time;
4101 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4102
a931da6a 4103 write_lock(&journal->j_state_lock);
ac27a0ec 4104 if (test_opt(sb, BARRIER))
dab291af 4105 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4106 else
dab291af 4107 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4108 if (test_opt(sb, DATA_ERR_ABORT))
4109 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4110 else
4111 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4112 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4113}
4114
617ba13b 4115static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4116 unsigned int journal_inum)
4117{
4118 struct inode *journal_inode;
4119 journal_t *journal;
4120
0390131b
FM		 4121 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4122
ac27a0ec
DK		 4123 /* First, test for the existence of a valid inode on disk. Bad
4124 * things happen if we iget() an unused inode, as the subsequent
4125 * iput() will try to delete it. */
4126
1d1fe1ee
DH		 4127 journal_inode = ext4_iget(sb, journal_inum);
4128 if (IS_ERR(journal_inode)) {
b31e1552 4129 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4130 return NULL;
4131 }
4132 if (!journal_inode->i_nlink) {
4133 make_bad_inode(journal_inode);
4134 iput(journal_inode);
b31e1552 4135 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4136 return NULL;
4137 }
4138
e5f8eab8 4139 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4140 journal_inode, journal_inode->i_size);
1d1fe1ee 4141 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4142 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4143 iput(journal_inode);
4144 return NULL;
4145 }
4146
dab291af 4147 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4148 if (!journal) {
b31e1552 4149 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4150 iput(journal_inode);
4151 return NULL;
4152 }
4153 journal->j_private = sb;
617ba13b 4154 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4155 return journal;
4156}
4157
617ba13b 4158static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4159 dev_t j_dev)
4160{
2b2d6d01 4161 struct buffer_head *bh;
ac27a0ec 4162 journal_t *journal;
617ba13b
MC		 4163 ext4_fsblk_t start;
4164 ext4_fsblk_t len;
ac27a0ec 4165 int hblock, blocksize;
617ba13b 4166 ext4_fsblk_t sb_block;
ac27a0ec 4167 unsigned long offset;
2b2d6d01 4168 struct ext4_super_block *es;
ac27a0ec
DK		 4169 struct block_device *bdev;
4170
0390131b
FM		 4171 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4172
b31e1552 4173 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4174 if (bdev == NULL)
4175 return NULL;
4176
ac27a0ec 4177 blocksize = sb->s_blocksize;
e1defc4f 4178 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4179 if (blocksize < hblock) {
b31e1552
ES		 4180 ext4_msg(sb, KERN_ERR,
4181 "blocksize too small for journal device");
ac27a0ec
DK		 4182 goto out_bdev;
4183 }
4184
617ba13b
MC		 4185 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4186 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4187 set_blocksize(bdev, blocksize);
4188 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4189 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4190 "external journal");
ac27a0ec
DK		 4191 goto out_bdev;
4192 }
4193
2716b802 4194 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4195 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4196 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4197 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4198 ext4_msg(sb, KERN_ERR, "external journal has "
4199 "bad superblock");
ac27a0ec
DK		 4200 brelse(bh);
4201 goto out_bdev;
4202 }
4203
617ba13b 4204 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4205 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4206 brelse(bh);
4207 goto out_bdev;
4208 }
4209
bd81d8ee 4210 len = ext4_blocks_count(es);
ac27a0ec
DK		 4211 start = sb_block + 1;
4212 brelse(bh); /* we're done with the superblock */
4213
dab291af 4214 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4215 start, len, blocksize);
4216 if (!journal) {
b31e1552 4217 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4218 goto out_bdev;
4219 }
4220 journal->j_private = sb;
4221 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4222 wait_on_buffer(journal->j_sb_buffer);
4223 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4224 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4225 goto out_journal;
4226 }
4227 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4228 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4229 "user (unsupported) - %d",
ac27a0ec
DK		 4230 be32_to_cpu(journal->j_superblock->s_nr_users));
4231 goto out_journal;
4232 }
617ba13b
MC		 4233 EXT4_SB(sb)->journal_bdev = bdev;
4234 ext4_init_journal_params(sb, journal);
ac27a0ec 4235 return journal;
0b8e58a1 4236
ac27a0ec 4237out_journal:
dab291af 4238 jbd2_journal_destroy(journal);
ac27a0ec 4239out_bdev:
617ba13b 4240 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4241 return NULL;
4242}
4243
617ba13b
MC		 4244static int ext4_load_journal(struct super_block *sb,
4245 struct ext4_super_block *es,
ac27a0ec
DK		 4246 unsigned long journal_devnum)
4247{
4248 journal_t *journal;
4249 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4250 dev_t journal_dev;
4251 int err = 0;
4252 int really_read_only;
4253
0390131b
FM		 4254 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4255
ac27a0ec
DK		 4256 if (journal_devnum &&
4257 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4258 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4259 "numbers have changed");
ac27a0ec
DK		 4260 journal_dev = new_decode_dev(journal_devnum);
4261 } else
4262 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4263
4264 really_read_only = bdev_read_only(sb->s_bdev);
4265
4266 /*
4267 * Are we loading a blank journal or performing recovery after a
4268 * crash? For recovery, we need to check in advance whether we
4269 * can get read-write access to the device.
4270 */
617ba13b 4271 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4272 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4273 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4274 "required on readonly filesystem");
ac27a0ec 4275 if (really_read_only) {
b31e1552
ES		 4276 ext4_msg(sb, KERN_ERR, "write access "
4277 "unavailable, cannot proceed");
ac27a0ec
DK		 4278 return -EROFS;
4279 }
b31e1552
ES		 4280 ext4_msg(sb, KERN_INFO, "write access will "
4281 "be enabled during recovery");
ac27a0ec
DK		 4282 }
4283 }
4284
4285 if (journal_inum && journal_dev) {
b31e1552
ES		 4286 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4287 "and inode journals!");
ac27a0ec
DK		 4288 return -EINVAL;
4289 }
4290
4291 if (journal_inum) {
617ba13b 4292 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4293 return -EINVAL;
4294 } else {
617ba13b 4295 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4296 return -EINVAL;
4297 }
4298
90576c0b 4299 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4300 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4301
617ba13b 4302 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4303 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4304 if (!err) {
4305 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4306 if (save)
4307 memcpy(save, ((char *) es) +
4308 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4309 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4310 if (save)
4311 memcpy(((char *) es) + EXT4_S_ERR_START,
4312 save, EXT4_S_ERR_LEN);
4313 kfree(save);
4314 }
ac27a0ec
DK		 4315
4316 if (err) {
b31e1552 4317 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4318 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4319 return err;
4320 }
4321
617ba13b
MC		 4322 EXT4_SB(sb)->s_journal = journal;
4323 ext4_clear_journal_err(sb, es);
ac27a0ec 4324
c41303ce 4325 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4326 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4327 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4328
4329 /* Make sure we flush the recovery flag to disk. */
e2d67052 4330 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4331 }
4332
4333 return 0;
4334}
4335
e2d67052 4336static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4337{
e2d67052 4338 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4339 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4340 int error = 0;
ac27a0ec 4341
7c2e7087 4342 if (!sbh || block_device_ejected(sb))
c4be0c1d 4343 return error;
914258bf
TT		 4344 if (buffer_write_io_error(sbh)) {
4345 /*
4346 * Oh, dear. A previous attempt to write the
4347 * superblock failed. This could happen because the
4348 * USB device was yanked out. Or it could happen to
4349 * be a transient write error and maybe the block will
4350 * be remapped. Nothing we can do but to retry the
4351 * write and hope for the best.
4352 */
b31e1552
ES		 4353 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4354 "superblock detected");
914258bf
TT		 4355 clear_buffer_write_io_error(sbh);
4356 set_buffer_uptodate(sbh);
4357 }
71290b36
TT		 4358 /*
4359 * If the file system is mounted read-only, don't update the
4360 * superblock write time. This avoids updating the superblock
4361 * write time when we are mounting the root file system
4362 * read/only but we need to replay the journal; at that point,
4363 * for people who are east of GMT and who make their clock
4364 * tick in localtime for Windows bug-for-bug compatibility,
4365 * the clock is set in the future, and this will cause e2fsck
4366 * to complain and force a full file system check.
4367 */
4368 if (!(sb->s_flags & MS_RDONLY))
4369 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4370 if (sb->s_bdev->bd_part)
4371 es->s_kbytes_written =
4372 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4373 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4374 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4375 else
4376 es->s_kbytes_written =
4377 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4378 ext4_free_blocks_count_set(es,
4379 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4380 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4381 es->s_free_inodes_count =
4382 cpu_to_le32(percpu_counter_sum_positive(
4383 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4384 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4385 ext4_superblock_csum_set(sb);
ac27a0ec 4386 mark_buffer_dirty(sbh);
914258bf 4387 if (sync) {
c4be0c1d
TS		 4388 error = sync_dirty_buffer(sbh);
4389 if (error)
4390 return error;
4391
4392 error = buffer_write_io_error(sbh);
4393 if (error) {
b31e1552
ES		 4394 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4395 "superblock");
914258bf
TT		 4396 clear_buffer_write_io_error(sbh);
4397 set_buffer_uptodate(sbh);
4398 }
4399 }
c4be0c1d 4400 return error;
ac27a0ec
DK		 4401}
4402
ac27a0ec
DK		 4403/*
4404 * Have we just finished recovery? If so, and if we are mounting (or
4405 * remounting) the filesystem readonly, then we will end up with a
4406 * consistent fs on disk. Record that fact.
4407 */
2b2d6d01
TT		 4408static void ext4_mark_recovery_complete(struct super_block *sb,
4409 struct ext4_super_block *es)
ac27a0ec 4410{
617ba13b 4411 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4412
0390131b
FM		 4413 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4414 BUG_ON(journal != NULL);
4415 return;
4416 }
dab291af 4417 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4418 if (jbd2_journal_flush(journal) < 0)
4419 goto out;
4420
617ba13b 4421 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4422 sb->s_flags & MS_RDONLY) {
617ba13b 4423 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4424 ext4_commit_super(sb, 1);
ac27a0ec 4425 }
7ffe1ea8
HK		 4426
4427out:
dab291af 4428 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4429}
4430
4431/*
4432 * If we are mounting (or read-write remounting) a filesystem whose journal
4433 * has recorded an error from a previous lifetime, move that error to the
4434 * main filesystem now.
4435 */
2b2d6d01
TT		 4436static void ext4_clear_journal_err(struct super_block *sb,
4437 struct ext4_super_block *es)
ac27a0ec
DK		 4438{
4439 journal_t *journal;
4440 int j_errno;
4441 const char *errstr;
4442
0390131b
FM		 4443 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4444
617ba13b 4445 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4446
4447 /*
4448 * Now check for any error status which may have been recorded in the
617ba13b 4449 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4450 */
4451
dab291af 4452 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4453 if (j_errno) {
4454 char nbuf[16];
4455
617ba13b 4456 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4457 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4458 "from previous mount: %s", errstr);
12062ddd 4459 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4460
617ba13b
MC		 4461 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4462 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4463 ext4_commit_super(sb, 1);
ac27a0ec 4464
dab291af 4465 jbd2_journal_clear_err(journal);
d796c52e 4466 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4467 }
4468}
4469
4470/*
4471 * Force the running and committing transactions to commit,
4472 * and wait on the commit.
4473 */
617ba13b 4474int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4475{
4476 journal_t *journal;
0390131b 4477 int ret = 0;
ac27a0ec
DK		 4478
4479 if (sb->s_flags & MS_RDONLY)
4480 return 0;
4481
617ba13b 4482 journal = EXT4_SB(sb)->s_journal;
8e8ad8a5 4483 if (journal)
0390131b 4484 ret = ext4_journal_force_commit(journal);
0390131b 4485
ac27a0ec
DK		 4486 return ret;
4487}
4488
617ba13b 4489static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4490{
14ce0cb4 4491 int ret = 0;
9eddacf9 4492 tid_t target;
8d5d02e6 4493 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4494
9bffad1e 4495 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4496 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4497 /*
4498 * Writeback quota in non-journalled quota case - journalled quota has
4499 * no dirty dquots
4500 */
4501 dquot_writeback_dquots(sb, -1);
8d5d02e6 4502 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4503 if (wait)
8d5d02e6 4504 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4505 }
14ce0cb4 4506 return ret;
ac27a0ec
DK		 4507}
4508
4509/*
4510 * LVM calls this function before a (read-only) snapshot is created. This
4511 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4512 *
4513 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4514 * state independently. It relies on upper layer to stop all data & metadata
4515 * modifications.
ac27a0ec 4516 */
c4be0c1d 4517static int ext4_freeze(struct super_block *sb)
ac27a0ec 4518{
c4be0c1d
TS		 4519 int error = 0;
4520 journal_t *journal;
ac27a0ec 4521
9ca92389
TT		 4522 if (sb->s_flags & MS_RDONLY)
4523 return 0;
ac27a0ec 4524
9ca92389 4525 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4526
9ca92389
TT		 4527 /* Now we set up the journal barrier. */
4528 jbd2_journal_lock_updates(journal);
ac27a0ec 4529
9ca92389
TT		 4530 /*
4531 * Don't clear the needs_recovery flag if we failed to flush
4532 * the journal.
4533 */
4534 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4535 if (error < 0)
4536 goto out;
9ca92389
TT		 4537
4538 /* Journal blocked and flushed, clear needs_recovery flag. */
4539 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4540 error = ext4_commit_super(sb, 1);
6b0310fb 4541out:
8e8ad8a5 4542 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4543 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4544 return error;
ac27a0ec
DK		 4545}
4546
4547/*
4548 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4549 * flag here, even though the filesystem is not technically dirty yet.
4550 */
c4be0c1d 4551static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4552{
9ca92389
TT		 4553 if (sb->s_flags & MS_RDONLY)
4554 return 0;
4555
9ca92389
TT		 4556 /* Reset the needs_recovery flag before the fs is unlocked. */
4557 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4558 ext4_commit_super(sb, 1);
c4be0c1d 4559 return 0;
ac27a0ec
DK		 4560}
4561
673c6100
TT		 4562/*
4563 * Structure to save mount options for ext4_remount's benefit
4564 */
4565struct ext4_mount_options {
4566 unsigned long s_mount_opt;
a2595b8a 4567 unsigned long s_mount_opt2;
08cefc7a
EB		 4568 kuid_t s_resuid;
4569 kgid_t s_resgid;
673c6100
TT		 4570 unsigned long s_commit_interval;
4571 u32 s_min_batch_time, s_max_batch_time;
4572#ifdef CONFIG_QUOTA
4573 int s_jquota_fmt;
4574 char *s_qf_names[MAXQUOTAS];
4575#endif
4576};
4577
2b2d6d01 4578static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4579{
2b2d6d01 4580 struct ext4_super_block *es;
617ba13b 4581 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4582 unsigned long old_sb_flags;
617ba13b 4583 struct ext4_mount_options old_opts;
c79d967d 4584 int enable_quota = 0;
8a266467 4585 ext4_group_t g;
b3881f74 4586 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4587 int err = 0;
ac27a0ec
DK		 4588#ifdef CONFIG_QUOTA
4589 int i;
4590#endif
d4c402d9 4591 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4592
4593 /* Store the original options */
4594 old_sb_flags = sb->s_flags;
4595 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4596 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4597 old_opts.s_resuid = sbi->s_resuid;
4598 old_opts.s_resgid = sbi->s_resgid;
4599 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4600 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4601 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4602#ifdef CONFIG_QUOTA
4603 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4604 for (i = 0; i < MAXQUOTAS; i++)
4605 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4606#endif
b3881f74
TT		 4607 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4608 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4609
4610 /*
4611 * Allow the "check" option to be passed as a remount option.
4612 */
661aa520 4613 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4614 err = -EINVAL;
4615 goto restore_opts;
4616 }
4617
4ab2f15b 4618 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4619 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4620
4621 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4622 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4623
4624 es = sbi->s_es;
4625
b3881f74 4626 if (sbi->s_journal) {
0390131b 4627 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4628 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4629 }
ac27a0ec 4630
661aa520 4631 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4632 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4633 err = -EROFS;
4634 goto restore_opts;
4635 }
4636
4637 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4638 err = dquot_suspend(sb, -1);
4639 if (err < 0)
c79d967d 4640 goto restore_opts;
c79d967d 4641
ac27a0ec
DK		 4642 /*
4643 * First of all, the unconditional stuff we have to do
4644 * to disable replay of the journal when we next remount
4645 */
4646 sb->s_flags |= MS_RDONLY;
4647
4648 /*
4649 * OK, test if we are remounting a valid rw partition
4650 * readonly, and if so set the rdonly flag and then
4651 * mark the partition as valid again.
4652 */
617ba13b
MC		 4653 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4654 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4655 es->s_state = cpu_to_le16(sbi->s_mount_state);
4656
a63c9eb2 4657 if (sbi->s_journal)
0390131b 4658 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4659 } else {
a13fb1a4
ES		 4660 /* Make sure we can mount this feature set readwrite */
4661 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4662 err = -EROFS;
4663 goto restore_opts;
4664 }
8a266467
TT		 4665 /*
4666 * Make sure the group descriptor checksums
0b8e58a1 4667 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4668 */
4669 for (g = 0; g < sbi->s_groups_count; g++) {
4670 struct ext4_group_desc *gdp =
4671 ext4_get_group_desc(sb, g, NULL);
4672
feb0ab32 4673 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4674 ext4_msg(sb, KERN_ERR,
4675 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4676 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4677 le16_to_cpu(gdp->bg_checksum));
4678 err = -EINVAL;
4679 goto restore_opts;
4680 }
4681 }
4682
ead6596b
ES		 4683 /*
4684 * If we have an unprocessed orphan list hanging
4685 * around from a previously readonly bdev mount,
4686 * require a full umount/remount for now.
4687 */
4688 if (es->s_last_orphan) {
b31e1552 4689 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4690 "remount RDWR because of unprocessed "
4691 "orphan inode list. Please "
b31e1552 4692 "umount/remount instead");
ead6596b
ES		 4693 err = -EINVAL;
4694 goto restore_opts;
4695 }
4696
ac27a0ec
DK		 4697 /*
4698 * Mounting a RDONLY partition read-write, so reread
4699 * and store the current valid flag. (It may have
4700 * been changed by e2fsck since we originally mounted
4701 * the partition.)
4702 */
0390131b
FM		 4703 if (sbi->s_journal)
4704 ext4_clear_journal_err(sb, es);
ac27a0ec 4705 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4706 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4707 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4708 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4709 EXT4_FEATURE_INCOMPAT_MMP))
4710 if (ext4_multi_mount_protect(sb,
4711 le64_to_cpu(es->s_mmp_block))) {
4712 err = -EROFS;
4713 goto restore_opts;
4714 }
c79d967d 4715 enable_quota = 1;
ac27a0ec
DK		 4716 }
4717 }
bfff6873
LC		 4718
4719 /*
4720 * Reinitialize lazy itable initialization thread based on
4721 * current settings
4722 */
4723 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4724 ext4_unregister_li_request(sb);
4725 else {
4726 ext4_group_t first_not_zeroed;
4727 first_not_zeroed = ext4_has_uninit_itable(sb);
4728 ext4_register_li_request(sb, first_not_zeroed);
4729 }
4730
6fd058f7 4731 ext4_setup_system_zone(sb);
d096ad0f 4732 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4733 ext4_commit_super(sb, 1);
0390131b 4734
ac27a0ec
DK		 4735#ifdef CONFIG_QUOTA
4736 /* Release old quota file names */
4737 for (i = 0; i < MAXQUOTAS; i++)
4738 if (old_opts.s_qf_names[i] &&
4739 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4740 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4741 if (enable_quota) {
4742 if (sb_any_quota_suspended(sb))
4743 dquot_resume(sb, -1);
4744 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4745 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4746 err = ext4_enable_quotas(sb);
07724f98 4747 if (err)
7c319d32 4748 goto restore_opts;
7c319d32
AK		 4749 }
4750 }
ac27a0ec 4751#endif
d4c402d9
CW		 4752
4753 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4754 kfree(orig_data);
ac27a0ec 4755 return 0;
0b8e58a1 4756
ac27a0ec
DK		 4757restore_opts:
4758 sb->s_flags = old_sb_flags;
4759 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4760 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4761 sbi->s_resuid = old_opts.s_resuid;
4762 sbi->s_resgid = old_opts.s_resgid;
4763 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4764 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4765 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4766#ifdef CONFIG_QUOTA
4767 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4768 for (i = 0; i < MAXQUOTAS; i++) {
4769 if (sbi->s_qf_names[i] &&
4770 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4771 kfree(sbi->s_qf_names[i]);
4772 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4773 }
4774#endif
d4c402d9 4775 kfree(orig_data);
ac27a0ec
DK		 4776 return err;
4777}
4778
2b2d6d01 4779static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4780{
4781 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4782 struct ext4_sb_info *sbi = EXT4_SB(sb);
4783 struct ext4_super_block *es = sbi->s_es;
952fc18e 4784 ext4_fsblk_t overhead = 0;
960cc398 4785 u64 fsid;
d02a9391 4786 s64 bfree;
ac27a0ec 4787
952fc18e
TT		 4788 if (!test_opt(sb, MINIX_DF))
4789 overhead = sbi->s_overhead;
ac27a0ec 4790
617ba13b 4791 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4792 buf->f_bsize = sb->s_blocksize;
b72f78cb 4793 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4794 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4795 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4796 /* prevent underflow in case that few free space is available */
57042651 4797 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4798 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4799 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4800 buf->f_bavail = 0;
4801 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4802 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4803 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4804 fsid = le64_to_cpup((void *)es->s_uuid) ^
4805 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4806 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4807 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4808
ac27a0ec
DK		 4809 return 0;
4810}
4811
0b8e58a1
AD		 4812/* Helper function for writing quotas on sync - we need to start transaction
4813 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4814 * Process 1 Process 2
617ba13b 4815 * ext4_create() quota_sync()
a269eb18 4816 * jbd2_journal_start() write_dquot()
871a2931 4817 * dquot_initialize() down(dqio_mutex)
dab291af 4818 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4819 *
4820 */
4821
4822#ifdef CONFIG_QUOTA
4823
4824static inline struct inode *dquot_to_inode(struct dquot *dquot)
4825{
4c376dca 4826 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4827}
4828
617ba13b 4829static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4830{
4831 int ret, err;
4832 handle_t *handle;
4833 struct inode *inode;
4834
4835 inode = dquot_to_inode(dquot);
617ba13b 4836 handle = ext4_journal_start(inode,
0b8e58a1 4837 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4838 if (IS_ERR(handle))
4839 return PTR_ERR(handle);
4840 ret = dquot_commit(dquot);
617ba13b 4841 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4842 if (!ret)
4843 ret = err;
4844 return ret;
4845}
4846
617ba13b 4847static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4848{
4849 int ret, err;
4850 handle_t *handle;
4851
617ba13b 4852 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4853 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4854 if (IS_ERR(handle))
4855 return PTR_ERR(handle);
4856 ret = dquot_acquire(dquot);
617ba13b 4857 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4858 if (!ret)
4859 ret = err;
4860 return ret;
4861}
4862
617ba13b 4863static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4864{
4865 int ret, err;
4866 handle_t *handle;
4867
617ba13b 4868 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4869 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4870 if (IS_ERR(handle)) {
4871 /* Release dquot anyway to avoid endless cycle in dqput() */
4872 dquot_release(dquot);
ac27a0ec 4873 return PTR_ERR(handle);
9c3013e9 4874 }
ac27a0ec 4875 ret = dquot_release(dquot);
617ba13b 4876 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4877 if (!ret)
4878 ret = err;
4879 return ret;
4880}
4881
617ba13b 4882static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4883{
2c8be6b2 4884 /* Are we journaling quotas? */
617ba13b
MC		 4885 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4886 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4887 dquot_mark_dquot_dirty(dquot);
617ba13b 4888 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4889 } else {
4890 return dquot_mark_dquot_dirty(dquot);
4891 }
4892}
4893
617ba13b 4894static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4895{
4896 int ret, err;
4897 handle_t *handle;
4898
4899 /* Data block + inode block */
617ba13b 4900 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4901 if (IS_ERR(handle))
4902 return PTR_ERR(handle);
4903 ret = dquot_commit_info(sb, type);
617ba13b 4904 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4905 if (!ret)
4906 ret = err;
4907 return ret;
4908}
4909
4910/*
4911 * Turn on quotas during mount time - we need to find
4912 * the quota file and such...
4913 */
617ba13b 4914static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4915{
287a8095
CH		 4916 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4917 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4918}
4919
4920/*
4921 * Standard function to be called on quota_on
4922 */
617ba13b 4923static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4924 struct path *path)
ac27a0ec
DK		 4925{
4926 int err;
ac27a0ec
DK		 4927
4928 if (!test_opt(sb, QUOTA))
4929 return -EINVAL;
0623543b 4930
ac27a0ec 4931 /* Quotafile not on the same filesystem? */
d8c9584e 4932 if (path->dentry->d_sb != sb)
ac27a0ec 4933 return -EXDEV;
0623543b
JK		 4934 /* Journaling quota? */
4935 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4936 /* Quotafile not in fs root? */
f00c9e44 4937 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4938 ext4_msg(sb, KERN_WARNING,
4939 "Quota file not on filesystem root. "
4940 "Journaled quota will not work");
2b2d6d01 4941 }
0623543b
JK		 4942
4943 /*
4944 * When we journal data on quota file, we have to flush journal to see
4945 * all updates to the file when we bypass pagecache...
4946 */
0390131b 4947 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4948 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4949 /*
4950 * We don't need to lock updates but journal_flush() could
4951 * otherwise be livelocked...
4952 */
4953 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4954 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4955 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4956 if (err)
7ffe1ea8 4957 return err;
0623543b
JK		 4958 }
4959
f00c9e44 4960 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4961}
4962
7c319d32
AK		 4963static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4964 unsigned int flags)
4965{
4966 int err;
4967 struct inode *qf_inode;
4968 unsigned long qf_inums[MAXQUOTAS] = {
4969 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4970 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4971 };
4972
4973 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4974
4975 if (!qf_inums[type])
4976 return -EPERM;
4977
4978 qf_inode = ext4_iget(sb, qf_inums[type]);
4979 if (IS_ERR(qf_inode)) {
4980 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4981 return PTR_ERR(qf_inode);
4982 }
4983
4984 err = dquot_enable(qf_inode, type, format_id, flags);
4985 iput(qf_inode);
4986
4987 return err;
4988}
4989
4990/* Enable usage tracking for all quota types. */
4991static int ext4_enable_quotas(struct super_block *sb)
4992{
4993 int type, err = 0;
4994 unsigned long qf_inums[MAXQUOTAS] = {
4995 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4996 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4997 };
4998
4999 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5000 for (type = 0; type < MAXQUOTAS; type++) {
5001 if (qf_inums[type]) {
5002 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5003 DQUOT_USAGE_ENABLED);
5004 if (err) {
5005 ext4_warning(sb,
5006 "Failed to enable quota (type=%d) "
5007 "tracking. Please run e2fsck to fix.",
5008 type);
5009 return err;
5010 }
5011 }
5012 }
5013 return 0;
5014}
5015
5016/*
5017 * quota_on function that is used when QUOTA feature is set.
5018 */
5019static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5020 int format_id)
5021{
5022 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5023 return -EINVAL;
5024
5025 /*
5026 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5027 */
5028 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5029}
5030
ca0e05e4
DM		 5031static int ext4_quota_off(struct super_block *sb, int type)
5032{
21f97697
JK		 5033 struct inode *inode = sb_dqopt(sb)->files[type];
5034 handle_t *handle;
5035
87009d86
DM		 5036 /* Force all delayed allocation blocks to be allocated.
5037 * Caller already holds s_umount sem */
5038 if (test_opt(sb, DELALLOC))
ca0e05e4 5039 sync_filesystem(sb);
ca0e05e4 5040
0b268590
AG		 5041 if (!inode)
5042 goto out;
5043
21f97697
JK		 5044 /* Update modification times of quota files when userspace can
5045 * start looking at them */
5046 handle = ext4_journal_start(inode, 1);
5047 if (IS_ERR(handle))
5048 goto out;
5049 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5050 ext4_mark_inode_dirty(handle, inode);
5051 ext4_journal_stop(handle);
5052
5053out:
ca0e05e4
DM		 5054 return dquot_quota_off(sb, type);
5055}
5056
7c319d32
AK		 5057/*
5058 * quota_off function that is used when QUOTA feature is set.
5059 */
5060static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5061{
5062 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5063 return -EINVAL;
5064
5065 /* Disable only the limits. */
5066 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5067}
5068
ac27a0ec
DK		 5069/* Read data from quotafile - avoid pagecache and such because we cannot afford
5070 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5071 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5072 * we don't have to be afraid of races */
617ba13b 5073static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5074 size_t len, loff_t off)
5075{
5076 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5077 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5078 int err = 0;
5079 int offset = off & (sb->s_blocksize - 1);
5080 int tocopy;
5081 size_t toread;
5082 struct buffer_head *bh;
5083 loff_t i_size = i_size_read(inode);
5084
5085 if (off > i_size)
5086 return 0;
5087 if (off+len > i_size)
5088 len = i_size-off;
5089 toread = len;
5090 while (toread > 0) {
5091 tocopy = sb->s_blocksize - offset < toread ?
5092 sb->s_blocksize - offset : toread;
617ba13b 5093 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5094 if (err)
5095 return err;
5096 if (!bh) /* A hole? */
5097 memset(data, 0, tocopy);
5098 else
5099 memcpy(data, bh->b_data+offset, tocopy);
5100 brelse(bh);
5101 offset = 0;
5102 toread -= tocopy;
5103 data += tocopy;
5104 blk++;
5105 }
5106 return len;
5107}
5108
5109/* Write to quotafile (we know the transaction is already started and has
5110 * enough credits) */
617ba13b 5111static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5112 const char *data, size_t len, loff_t off)
5113{
5114 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5115 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5116 int err = 0;
5117 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5118 struct buffer_head *bh;
5119 handle_t *handle = journal_current_handle();
5120
0390131b 5121 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5122 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5123 " cancelled because transaction is not started",
9c3013e9
JK		 5124 (unsigned long long)off, (unsigned long long)len);
5125 return -EIO;
5126 }
67eeb568
DM		 5127 /*
5128 * Since we account only one data block in transaction credits,
5129 * then it is impossible to cross a block boundary.
5130 */
5131 if (sb->s_blocksize - offset < len) {
5132 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5133 " cancelled because not block aligned",
5134 (unsigned long long)off, (unsigned long long)len);
5135 return -EIO;
5136 }
5137
67eeb568
DM		 5138 bh = ext4_bread(handle, inode, blk, 1, &err);
5139 if (!bh)
5140 goto out;
62d2b5f2
JK		 5141 err = ext4_journal_get_write_access(handle, bh);
5142 if (err) {
5143 brelse(bh);
5144 goto out;
ac27a0ec 5145 }
67eeb568
DM		 5146 lock_buffer(bh);
5147 memcpy(bh->b_data+offset, data, len);
5148 flush_dcache_page(bh->b_page);
5149 unlock_buffer(bh);
62d2b5f2 5150 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5151 brelse(bh);
ac27a0ec 5152out:
0b7f7cef 5153 if (err)
ac27a0ec 5154 return err;
67eeb568
DM		 5155 if (inode->i_size < off + len) {
5156 i_size_write(inode, off + len);
617ba13b 5157 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5158 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5159 }
67eeb568 5160 return len;
ac27a0ec
DK		 5161}
5162
5163#endif
5164
152a0836
AV		 5165static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5166 const char *dev_name, void *data)
ac27a0ec 5167{
152a0836 5168 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5169}
5170
37f328eb 5171#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5172static inline void register_as_ext2(void)
5173{
5174 int err = register_filesystem(&ext2_fs_type);
5175 if (err)
5176 printk(KERN_WARNING
5177 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5178}
5179
5180static inline void unregister_as_ext2(void)
5181{
5182 unregister_filesystem(&ext2_fs_type);
5183}
2035e776
TT		 5184
5185static inline int ext2_feature_set_ok(struct super_block *sb)
5186{
5187 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5188 return 0;
5189 if (sb->s_flags & MS_RDONLY)
5190 return 1;
5191 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5192 return 0;
5193 return 1;
5194}
51b7e3c9 5195MODULE_ALIAS("ext2");
24b58424
TT		 5196#else
5197static inline void register_as_ext2(void) { }
5198static inline void unregister_as_ext2(void) { }
2035e776 5199static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5200#endif
5201
37f328eb 5202#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5203static inline void register_as_ext3(void)
5204{
5205 int err = register_filesystem(&ext3_fs_type);
5206 if (err)
5207 printk(KERN_WARNING
5208 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5209}
5210
5211static inline void unregister_as_ext3(void)
5212{
5213 unregister_filesystem(&ext3_fs_type);
5214}
2035e776
TT		 5215
5216static inline int ext3_feature_set_ok(struct super_block *sb)
5217{
5218 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5219 return 0;
5220 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5221 return 0;
5222 if (sb->s_flags & MS_RDONLY)
5223 return 1;
5224 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5225 return 0;
5226 return 1;
5227}
51b7e3c9 5228MODULE_ALIAS("ext3");
24b58424
TT		 5229#else
5230static inline void register_as_ext3(void) { }
5231static inline void unregister_as_ext3(void) { }
2035e776 5232static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5233#endif
5234
03010a33
TT		 5235static struct file_system_type ext4_fs_type = {
5236 .owner = THIS_MODULE,
5237 .name = "ext4",
152a0836 5238 .mount = ext4_mount,
03010a33
TT		 5239 .kill_sb = kill_block_super,
5240 .fs_flags = FS_REQUIRES_DEV,
5241};
5242
8f021222 5243static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5244{
5245 struct ext4_features *ef;
5246 int ret = -ENOMEM;
5247
5248 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5249 if (!ef)
5250 goto out;
5251
5252 ef->f_kobj.kset = ext4_kset;
5253 init_completion(&ef->f_kobj_unregister);
5254 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5255 "features");
5256 if (ret) {
5257 kfree(ef);
5258 goto out;
5259 }
5260
5261 ext4_feat = ef;
5262 ret = 0;
5263out:
5264 return ret;
5265}
5266
8f021222
LC		 5267static void ext4_exit_feat_adverts(void)
5268{
5269 kobject_put(&ext4_feat->f_kobj);
5270 wait_for_completion(&ext4_feat->f_kobj_unregister);
5271 kfree(ext4_feat);
5272}
5273
e9e3bcec
ES		 5274/* Shared across all ext4 file systems */
5275wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5276struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5277
5dabfc78 5278static int __init ext4_init_fs(void)
ac27a0ec 5279{
e9e3bcec 5280 int i, err;
c9de560d 5281
07c0c5d8
AV		 5282 ext4_li_info = NULL;
5283 mutex_init(&ext4_li_mtx);
5284
9a4c8019 5285 /* Build-time check for flags consistency */
12e9b892 5286 ext4_check_flag_values();
e9e3bcec
ES		 5287
5288 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5289 mutex_init(&ext4__aio_mutex[i]);
5290 init_waitqueue_head(&ext4__ioend_wq[i]);
5291 }
5292
51865fda 5293 err = ext4_init_es();
6fd058f7
TT		 5294 if (err)
5295 return err;
51865fda
ZL		 5296
5297 err = ext4_init_pageio();
5298 if (err)
5299 goto out7;
5300
5dabfc78 5301 err = ext4_init_system_zone();
bd2d0210 5302 if (err)
d44651d0 5303 goto out6;
3197ebdb 5304 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5305 if (!ext4_kset) {
5306 err = -ENOMEM;
dd68314c 5307 goto out5;
0e376b1e 5308 }
d44651d0 5309 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5310
5311 err = ext4_init_feat_adverts();
dd68314c
TT		 5312 if (err)
5313 goto out4;
857ac889 5314
5dabfc78 5315 err = ext4_init_mballoc();
ac27a0ec 5316 if (err)
6fd058f7 5317 goto out3;
c9de560d 5318
5dabfc78 5319 err = ext4_init_xattr();
c9de560d
AT		 5320 if (err)
5321 goto out2;
ac27a0ec
DK		 5322 err = init_inodecache();
5323 if (err)
5324 goto out1;
24b58424 5325 register_as_ext3();
2035e776 5326 register_as_ext2();
03010a33 5327 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5328 if (err)
5329 goto out;
bfff6873 5330
ac27a0ec
DK		 5331 return 0;
5332out:
24b58424
TT		 5333 unregister_as_ext2();
5334 unregister_as_ext3();
ac27a0ec
DK		 5335 destroy_inodecache();
5336out1:
5dabfc78 5337 ext4_exit_xattr();
c9de560d 5338out2:
5dabfc78 5339 ext4_exit_mballoc();
6fd058f7 5340out3:
8f021222 5341 ext4_exit_feat_adverts();
dd68314c 5342out4:
d44651d0
FJ		 5343 if (ext4_proc_root)
5344 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5345 kset_unregister(ext4_kset);
d44651d0 5346out5:
5dabfc78 5347 ext4_exit_system_zone();
d44651d0 5348out6:
5dabfc78 5349 ext4_exit_pageio();
51865fda
ZL		 5350out7:
5351 ext4_exit_es();
5352
ac27a0ec
DK		 5353 return err;
5354}
5355
5dabfc78 5356static void __exit ext4_exit_fs(void)
ac27a0ec 5357{
bfff6873 5358 ext4_destroy_lazyinit_thread();
24b58424
TT		 5359 unregister_as_ext2();
5360 unregister_as_ext3();
03010a33 5361 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5362 destroy_inodecache();
5dabfc78
TT		 5363 ext4_exit_xattr();
5364 ext4_exit_mballoc();
8f021222 5365 ext4_exit_feat_adverts();
9f6200bb 5366 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5367 kset_unregister(ext4_kset);
5dabfc78
TT		 5368 ext4_exit_system_zone();
5369 ext4_exit_pageio();
ac27a0ec
DK		 5370}
5371
5372MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5373MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5374MODULE_LICENSE("GPL");
5dabfc78
TT		 5375module_init(ext4_init_fs)
5376module_exit(ext4_exit_fs)
ubuntu-bionic-kernel mirror