]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: remove unused assignment
[mirror_ubuntu-artful-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"
6f91bc5f 48#include "ext4_extents.h"
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;
91246c00 942 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 943 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 944 INIT_LIST_HEAD(&ei->i_prealloc_list);
945 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 946 ei->i_reserved_data_blocks = 0;
947 ei->i_reserved_meta_blocks = 0;
948 ei->i_allocated_meta_blocks = 0;
9d0be502 949 ei->i_da_metadata_calc_len = 0;
7e731bc9 950 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 951 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 952#ifdef CONFIG_QUOTA
953 ei->i_reserved_quota = 0;
954#endif
8aefcd55 955 ei->jinode = NULL;
c7064ef1 956 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 957 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 958 ei->i_sync_tid = 0;
959 ei->i_datasync_tid = 0;
f7ad6d2e 960 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 961 atomic_set(&ei->i_unwritten, 0);
0b8e58a1 962
ac27a0ec
DK		 963 return &ei->vfs_inode;
964}
965
7ff9c073
TT		 966static int ext4_drop_inode(struct inode *inode)
967{
968 int drop = generic_drop_inode(inode);
969
970 trace_ext4_drop_inode(inode, drop);
971 return drop;
972}
973
fa0d7e3d
NP		 974static void ext4_i_callback(struct rcu_head *head)
975{
976 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 977 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
978}
979
617ba13b 980static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 981{
9f7dd93d 982 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 983 ext4_msg(inode->i_sb, KERN_ERR,
984 "Inode %lu (%p): orphan list check failed!",
985 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 986 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
987 EXT4_I(inode), sizeof(struct ext4_inode_info),
988 true);
989 dump_stack();
990 }
fa0d7e3d 991 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 992}
993
51cc5068 994static void init_once(void *foo)
ac27a0ec 995{
617ba13b 996 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 997
a35afb83 998 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 999#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 1000 init_rwsem(&ei->xattr_sem);
ac27a0ec 1001#endif
0e855ac8 1002 init_rwsem(&ei->i_data_sem);
a35afb83 1003 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1004}
1005
1006static int init_inodecache(void)
1007{
617ba13b
MC		 1008 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1009 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 1010 0, (SLAB_RECLAIM_ACCOUNT|
1011 SLAB_MEM_SPREAD),
20c2df83 1012 init_once);
617ba13b 1013 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1014 return -ENOMEM;
1015 return 0;
1016}
1017
1018static void destroy_inodecache(void)
1019{
8c0a8537
KS		 1020 /*
1021 * Make sure all delayed rcu free inodes are flushed before we
1022 * destroy cache.
1023 */
1024 rcu_barrier();
617ba13b 1025 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1026}
1027
0930fcc1 1028void ext4_clear_inode(struct inode *inode)
ac27a0ec 1029{
0930fcc1 1030 invalidate_inode_buffers(inode);
dbd5768f 1031 clear_inode(inode);
9f754758 1032 dquot_drop(inode);
c2ea3fde 1033 ext4_discard_preallocations(inode);
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},
03010a33 1450#ifdef CONFIG_EXT4_FS_XATTR
26092bf5
TT		 1451 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1452 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
ac27a0ec 1453#else
26092bf5
TT		 1454 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1455 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
ac27a0ec 1456#endif
03010a33 1457#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1458 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1459 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1460#else
26092bf5
TT		 1461 {Opt_acl, 0, MOPT_NOSUPPORT},
1462 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1463#endif
26092bf5
TT		 1464 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1465 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1466 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1467 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1468 MOPT_SET | MOPT_Q},
1469 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1470 MOPT_SET | MOPT_Q},
1471 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1472 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1473 {Opt_usrjquota, 0, MOPT_Q},
1474 {Opt_grpjquota, 0, MOPT_Q},
1475 {Opt_offusrjquota, 0, MOPT_Q},
1476 {Opt_offgrpjquota, 0, MOPT_Q},
1477 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1478 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1479 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1480 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1481 {Opt_err, 0, 0}
1482};
1483
1484static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1485 substring_t *args, unsigned long *journal_devnum,
1486 unsigned int *journal_ioprio, int is_remount)
1487{
1488 struct ext4_sb_info *sbi = EXT4_SB(sb);
1489 const struct mount_opts *m;
08cefc7a
EB		 1490 kuid_t uid;
1491 kgid_t gid;
26092bf5
TT		 1492 int arg = 0;
1493
57f73c2c
TT		 1494#ifdef CONFIG_QUOTA
1495 if (token == Opt_usrjquota)
1496 return set_qf_name(sb, USRQUOTA, &args[0]);
1497 else if (token == Opt_grpjquota)
1498 return set_qf_name(sb, GRPQUOTA, &args[0]);
1499 else if (token == Opt_offusrjquota)
1500 return clear_qf_name(sb, USRQUOTA);
1501 else if (token == Opt_offgrpjquota)
1502 return clear_qf_name(sb, GRPQUOTA);
1503#endif
26092bf5
TT		 1504 if (args->from && match_int(args, &arg))
1505 return -1;
1506 switch (token) {
f7048605
TT		 1507 case Opt_noacl:
1508 case Opt_nouser_xattr:
1509 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1510 break;
26092bf5
TT		 1511 case Opt_sb:
1512 return 1; /* handled by get_sb_block() */
1513 case Opt_removed:
1514 ext4_msg(sb, KERN_WARNING,
1515 "Ignoring removed %s option", opt);
1516 return 1;
1517 case Opt_resuid:
08cefc7a
EB		 1518 uid = make_kuid(current_user_ns(), arg);
1519 if (!uid_valid(uid)) {
1520 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1521 return -1;
1522 }
1523 sbi->s_resuid = uid;
26092bf5
TT		 1524 return 1;
1525 case Opt_resgid:
08cefc7a
EB		 1526 gid = make_kgid(current_user_ns(), arg);
1527 if (!gid_valid(gid)) {
1528 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1529 return -1;
1530 }
1531 sbi->s_resgid = gid;
26092bf5
TT		 1532 return 1;
1533 case Opt_abort:
1534 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1535 return 1;
1536 case Opt_i_version:
1537 sb->s_flags |= MS_I_VERSION;
1538 return 1;
1539 case Opt_journal_dev:
1540 if (is_remount) {
1541 ext4_msg(sb, KERN_ERR,
1542 "Cannot specify journal on remount");
1543 return -1;
1544 }
1545 *journal_devnum = arg;
1546 return 1;
1547 case Opt_journal_ioprio:
1548 if (arg < 0 || arg > 7)
1549 return -1;
1550 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1551 return 1;
1552 }
1553
1554 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1555 if (token != m->token)
1556 continue;
1557 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1558 return -1;
1559 if (m->flags & MOPT_EXPLICIT)
1560 set_opt2(sb, EXPLICIT_DELALLOC);
1561 if (m->flags & MOPT_CLEAR_ERR)
1562 clear_opt(sb, ERRORS_MASK);
1563 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1564 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1565 "options when quota turned on");
1566 return -1;
1567 }
1568
1569 if (m->flags & MOPT_NOSUPPORT) {
1570 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1571 } else if (token == Opt_commit) {
1572 if (arg == 0)
1573 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1574 sbi->s_commit_interval = HZ * arg;
1575 } else if (token == Opt_max_batch_time) {
1576 if (arg == 0)
1577 arg = EXT4_DEF_MAX_BATCH_TIME;
1578 sbi->s_max_batch_time = arg;
1579 } else if (token == Opt_min_batch_time) {
1580 sbi->s_min_batch_time = arg;
1581 } else if (token == Opt_inode_readahead_blks) {
1582 if (arg > (1 << 30))
1583 return -1;
1584 if (arg && !is_power_of_2(arg)) {
b31e1552 1585 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1586 "EXT4-fs: inode_readahead_blks"
1587 " must be a power of 2");
1588 return -1;
ac27a0ec 1589 }
26092bf5
TT		 1590 sbi->s_inode_readahead_blks = arg;
1591 } else if (token == Opt_init_itable) {
1592 set_opt(sb, INIT_INODE_TABLE);
1593 if (!args->from)
1594 arg = EXT4_DEF_LI_WAIT_MULT;
1595 sbi->s_li_wait_mult = arg;
df981d03
TT		 1596 } else if (token == Opt_max_dir_size_kb) {
1597 sbi->s_max_dir_size_kb = arg;
26092bf5
TT		 1598 } else if (token == Opt_stripe) {
1599 sbi->s_stripe = arg;
1600 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1601 if (is_remount) {
eb513689
TT		 1602 if (!sbi->s_journal)
1603 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1604 else if (test_opt(sb, DATA_FLAGS) !=
1605 m->mount_opt) {
b31e1552 1606 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1607 "Cannot change data mode on remount");
1608 return -1;
ac27a0ec
DK		 1609 }
1610 } else {
fd8c37ec 1611 clear_opt(sb, DATA_FLAGS);
26092bf5 1612 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1613 }
ac27a0ec 1614#ifdef CONFIG_QUOTA
26092bf5 1615 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1616 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1617 sbi->s_jquota_fmt != m->mount_opt) {
1618 ext4_msg(sb, KERN_ERR, "Cannot "
1619 "change journaled quota options "
1620 "when quota turned on");
1621 return -1;
ac27a0ec 1622 }
26092bf5 1623 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1624#endif
26092bf5
TT		 1625 } else {
1626 if (!args->from)
1627 arg = 1;
1628 if (m->flags & MOPT_CLEAR)
1629 arg = !arg;
1630 else if (unlikely(!(m->flags & MOPT_SET))) {
1631 ext4_msg(sb, KERN_WARNING,
1632 "buggy handling of option %s", opt);
1633 WARN_ON(1);
1634 return -1;
3197ebdb 1635 }
26092bf5
TT		 1636 if (arg != 0)
1637 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1638 else
26092bf5 1639 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1640 }
26092bf5
TT		 1641 return 1;
1642 }
1643 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1644 "or missing value", opt);
1645 return -1;
1646}
1647
1648static int parse_options(char *options, struct super_block *sb,
1649 unsigned long *journal_devnum,
1650 unsigned int *journal_ioprio,
1651 int is_remount)
1652{
db7e5c66 1653#ifdef CONFIG_QUOTA
26092bf5 1654 struct ext4_sb_info *sbi = EXT4_SB(sb);
db7e5c66 1655#endif
26092bf5
TT		 1656 char *p;
1657 substring_t args[MAX_OPT_ARGS];
1658 int token;
1659
1660 if (!options)
1661 return 1;
1662
1663 while ((p = strsep(&options, ",")) != NULL) {
1664 if (!*p)
1665 continue;
1666 /*
1667 * Initialize args struct so we know whether arg was
1668 * found; some options take optional arguments.
1669 */
caecd0af 1670 args[0].to = args[0].from = NULL;
26092bf5
TT		 1671 token = match_token(p, tokens, args);
1672 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1673 journal_ioprio, is_remount) < 0)
1674 return 0;
ac27a0ec
DK		 1675 }
1676#ifdef CONFIG_QUOTA
1677 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1678 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1679 clear_opt(sb, USRQUOTA);
ac27a0ec 1680
482a7425 1681 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1682 clear_opt(sb, GRPQUOTA);
ac27a0ec 1683
56c50f11 1684 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1685 ext4_msg(sb, KERN_ERR, "old and new quota "
1686 "format mixing");
ac27a0ec
DK		 1687 return 0;
1688 }
1689
1690 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1691 ext4_msg(sb, KERN_ERR, "journaled quota format "
1692 "not specified");
ac27a0ec
DK		 1693 return 0;
1694 }
1695 } else {
1696 if (sbi->s_jquota_fmt) {
b31e1552 1697 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1698 "specified with no journaling "
b31e1552 1699 "enabled");
ac27a0ec
DK		 1700 return 0;
1701 }
1702 }
1703#endif
1704 return 1;
1705}
1706
2adf6da8
TT		 1707static inline void ext4_show_quota_options(struct seq_file *seq,
1708 struct super_block *sb)
1709{
1710#if defined(CONFIG_QUOTA)
1711 struct ext4_sb_info *sbi = EXT4_SB(sb);
1712
1713 if (sbi->s_jquota_fmt) {
1714 char *fmtname = "";
1715
1716 switch (sbi->s_jquota_fmt) {
1717 case QFMT_VFS_OLD:
1718 fmtname = "vfsold";
1719 break;
1720 case QFMT_VFS_V0:
1721 fmtname = "vfsv0";
1722 break;
1723 case QFMT_VFS_V1:
1724 fmtname = "vfsv1";
1725 break;
1726 }
1727 seq_printf(seq, ",jqfmt=%s", fmtname);
1728 }
1729
1730 if (sbi->s_qf_names[USRQUOTA])
1731 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1732
1733 if (sbi->s_qf_names[GRPQUOTA])
1734 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1735
1736 if (test_opt(sb, USRQUOTA))
1737 seq_puts(seq, ",usrquota");
1738
1739 if (test_opt(sb, GRPQUOTA))
1740 seq_puts(seq, ",grpquota");
1741#endif
1742}
1743
5a916be1
TT		 1744static const char *token2str(int token)
1745{
50df9fd5 1746 const struct match_token *t;
5a916be1
TT		 1747
1748 for (t = tokens; t->token != Opt_err; t++)
1749 if (t->token == token && !strchr(t->pattern, '='))
1750 break;
1751 return t->pattern;
1752}
1753
2adf6da8
TT		 1754/*
1755 * Show an option if
1756 * - it's set to a non-default value OR
1757 * - if the per-sb default is different from the global default
1758 */
66acdcf4
TT		 1759static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1760 int nodefs)
2adf6da8 1761{
2adf6da8
TT		 1762 struct ext4_sb_info *sbi = EXT4_SB(sb);
1763 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1764 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1765 const struct mount_opts *m;
66acdcf4 1766 char sep = nodefs ? '\n' : ',';
2adf6da8 1767
66acdcf4
TT		 1768#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1769#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1770
1771 if (sbi->s_sb_block != 1)
5a916be1
TT		 1772 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1773
1774 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1775 int want_set = m->flags & MOPT_SET;
1776 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1777 (m->flags & MOPT_CLEAR_ERR))
1778 continue;
66acdcf4 1779 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1780 continue; /* skip if same as the default */
1781 if ((want_set &&
1782 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1783 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1784 continue; /* select Opt_noFoo vs Opt_Foo */
1785 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1786 }
5a916be1 1787
08cefc7a 1788 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1789 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1790 SEQ_OPTS_PRINT("resuid=%u",
1791 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1792 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1793 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1794 SEQ_OPTS_PRINT("resgid=%u",
1795 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1796 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1797 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1798 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1799 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1800 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1801 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1802 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1803 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1804 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1805 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1806 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1807 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1808 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1809 if (sb->s_flags & MS_I_VERSION)
5a916be1 1810 SEQ_OPTS_PUTS("i_version");
66acdcf4 1811 if (nodefs || sbi->s_stripe)
5a916be1 1812 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1813 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1814 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1815 SEQ_OPTS_PUTS("data=journal");
1816 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1817 SEQ_OPTS_PUTS("data=ordered");
1818 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1819 SEQ_OPTS_PUTS("data=writeback");
1820 }
66acdcf4
TT		 1821 if (nodefs ||
1822 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1823 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1824 sbi->s_inode_readahead_blks);
2adf6da8 1825
66acdcf4
TT		 1826 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1827 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1828 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1829 if (nodefs || sbi->s_max_dir_size_kb)
1830 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1831
1832 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1833 return 0;
1834}
1835
66acdcf4
TT		 1836static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1837{
1838 return _ext4_show_options(seq, root->d_sb, 0);
1839}
1840
1841static int options_seq_show(struct seq_file *seq, void *offset)
1842{
1843 struct super_block *sb = seq->private;
1844 int rc;
1845
1846 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1847 rc = _ext4_show_options(seq, sb, 1);
1848 seq_puts(seq, "\n");
1849 return rc;
1850}
1851
1852static int options_open_fs(struct inode *inode, struct file *file)
1853{
1854 return single_open(file, options_seq_show, PDE(inode)->data);
1855}
1856
1857static const struct file_operations ext4_seq_options_fops = {
1858 .owner = THIS_MODULE,
1859 .open = options_open_fs,
1860 .read = seq_read,
1861 .llseek = seq_lseek,
1862 .release = single_release,
1863};
1864
617ba13b 1865static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1866 int read_only)
1867{
617ba13b 1868 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1869 int res = 0;
1870
617ba13b 1871 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1872 ext4_msg(sb, KERN_ERR, "revision level too high, "
1873 "forcing read-only mode");
ac27a0ec
DK		 1874 res = MS_RDONLY;
1875 }
1876 if (read_only)
281b5995 1877 goto done;
617ba13b 1878 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1879 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1880 "running e2fsck is recommended");
617ba13b 1881 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1882 ext4_msg(sb, KERN_WARNING,
1883 "warning: mounting fs with errors, "
1884 "running e2fsck is recommended");
ed3ce80a 1885 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1886 le16_to_cpu(es->s_mnt_count) >=
1887 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1888 ext4_msg(sb, KERN_WARNING,
1889 "warning: maximal mount count reached, "
1890 "running e2fsck is recommended");
ac27a0ec
DK		 1891 else if (le32_to_cpu(es->s_checkinterval) &&
1892 (le32_to_cpu(es->s_lastcheck) +
1893 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1894 ext4_msg(sb, KERN_WARNING,
1895 "warning: checktime reached, "
1896 "running e2fsck is recommended");
0b8e58a1 1897 if (!sbi->s_journal)
0390131b 1898 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1899 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1900 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1901 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1902 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1903 ext4_update_dynamic_rev(sb);
0390131b
FM		 1904 if (sbi->s_journal)
1905 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1906
e2d67052 1907 ext4_commit_super(sb, 1);
281b5995 1908done:
ac27a0ec 1909 if (test_opt(sb, DEBUG))
a9df9a49 1910 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1911 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1912 sb->s_blocksize,
1913 sbi->s_groups_count,
617ba13b
MC		 1914 EXT4_BLOCKS_PER_GROUP(sb),
1915 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1916 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1917
7abc52c2 1918 cleancache_init_fs(sb);
ac27a0ec
DK		 1919 return res;
1920}
1921
117fff10
TT		 1922int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1923{
1924 struct ext4_sb_info *sbi = EXT4_SB(sb);
1925 struct flex_groups *new_groups;
1926 int size;
1927
1928 if (!sbi->s_log_groups_per_flex)
1929 return 0;
1930
1931 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1932 if (size <= sbi->s_flex_groups_allocated)
1933 return 0;
1934
1935 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1936 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1937 if (!new_groups) {
1938 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1939 size / (int) sizeof(struct flex_groups));
1940 return -ENOMEM;
1941 }
1942
1943 if (sbi->s_flex_groups) {
1944 memcpy(new_groups, sbi->s_flex_groups,
1945 (sbi->s_flex_groups_allocated *
1946 sizeof(struct flex_groups)));
1947 ext4_kvfree(sbi->s_flex_groups);
1948 }
1949 sbi->s_flex_groups = new_groups;
1950 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1951 return 0;
1952}
1953
772cb7c8
JS		 1954static int ext4_fill_flex_info(struct super_block *sb)
1955{
1956 struct ext4_sb_info *sbi = EXT4_SB(sb);
1957 struct ext4_group_desc *gdp = NULL;
772cb7c8 1958 ext4_group_t flex_group;
d50f2ab6 1959 unsigned int groups_per_flex = 0;
117fff10 1960 int i, err;
772cb7c8 1961
503358ae 1962 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1963 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1964 sbi->s_log_groups_per_flex = 0;
1965 return 1;
1966 }
76495ec1 1967 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
772cb7c8 1968
117fff10
TT		 1969 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1970 if (err)
9933fc0a 1971 goto failed;
772cb7c8 1972
772cb7c8 1973 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1974 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1975
1976 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1977 atomic_add(ext4_free_inodes_count(sb, gdp),
1978 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1979 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1980 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1981 atomic_add(ext4_used_dirs_count(sb, gdp),
1982 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1983 }
1984
1985 return 1;
1986failed:
1987 return 0;
1988}
1989
feb0ab32
DW		 1990static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1991 struct ext4_group_desc *gdp)
717d50e4 1992{
feb0ab32 1993 int offset;
717d50e4 1994 __u16 crc = 0;
feb0ab32 1995 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1996
feb0ab32
DW		 1997 if ((sbi->s_es->s_feature_ro_compat &
1998 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1999 /* Use new metadata_csum algorithm */
2000 __u16 old_csum;
2001 __u32 csum32;
2002
2003 old_csum = gdp->bg_checksum;
2004 gdp->bg_checksum = 0;
2005 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2006 sizeof(le_group));
2007 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2008 sbi->s_desc_size);
2009 gdp->bg_checksum = old_csum;
2010
2011 crc = csum32 & 0xFFFF;
2012 goto out;
717d50e4
AD		 2013 }
2014
feb0ab32
DW		 2015 /* old crc16 code */
2016 offset = offsetof(struct ext4_group_desc, bg_checksum);
2017
2018 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2019 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2020 crc = crc16(crc, (__u8 *)gdp, offset);
2021 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2022 /* for checksum of struct ext4_group_desc do the rest...*/
2023 if ((sbi->s_es->s_feature_incompat &
2024 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2025 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2026 crc = crc16(crc, (__u8 *)gdp + offset,
2027 le16_to_cpu(sbi->s_es->s_desc_size) -
2028 offset);
2029
2030out:
717d50e4
AD		 2031 return cpu_to_le16(crc);
2032}
2033
feb0ab32 2034int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2035 struct ext4_group_desc *gdp)
2036{
feb0ab32
DW		 2037 if (ext4_has_group_desc_csum(sb) &&
2038 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2039 block_group, gdp)))
717d50e4
AD		 2040 return 0;
2041
2042 return 1;
2043}
2044
feb0ab32
DW		 2045void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2046 struct ext4_group_desc *gdp)
2047{
2048 if (!ext4_has_group_desc_csum(sb))
2049 return;
2050 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2051}
2052
ac27a0ec 2053/* Called at mount-time, super-block is locked */
bfff6873
LC		 2054static int ext4_check_descriptors(struct super_block *sb,
2055 ext4_group_t *first_not_zeroed)
ac27a0ec 2056{
617ba13b
MC		 2057 struct ext4_sb_info *sbi = EXT4_SB(sb);
2058 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2059 ext4_fsblk_t last_block;
bd81d8ee
LV		 2060 ext4_fsblk_t block_bitmap;
2061 ext4_fsblk_t inode_bitmap;
2062 ext4_fsblk_t inode_table;
ce421581 2063 int flexbg_flag = 0;
bfff6873 2064 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2065
ce421581
JS		 2066 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2067 flexbg_flag = 1;
2068
af5bc92d 2069 ext4_debug("Checking group descriptors");
ac27a0ec 2070
197cd65a
AM		 2071 for (i = 0; i < sbi->s_groups_count; i++) {
2072 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2073
ce421581 2074 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2075 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2076 else
2077 last_block = first_block +
617ba13b 2078 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2079
bfff6873
LC		 2080 if ((grp == sbi->s_groups_count) &&
2081 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2082 grp = i;
2083
8fadc143 2084 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2085 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2086 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2087 "Block bitmap for group %u not in group "
b31e1552 2088 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2089 return 0;
2090 }
8fadc143 2091 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2092 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2093 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2094 "Inode bitmap for group %u not in group "
b31e1552 2095 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2096 return 0;
2097 }
8fadc143 2098 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2099 if (inode_table < first_block ||
2b2d6d01 2100 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2101 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2102 "Inode table for group %u not in group "
b31e1552 2103 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2104 return 0;
2105 }
955ce5f5 2106 ext4_lock_group(sb, i);
feb0ab32 2107 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2108 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2109 "Checksum for group %u failed (%u!=%u)",
2110 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2111 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2112 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2113 ext4_unlock_group(sb, i);
8a266467 2114 return 0;
7ee1ec4c 2115 }
717d50e4 2116 }
955ce5f5 2117 ext4_unlock_group(sb, i);
ce421581
JS		 2118 if (!flexbg_flag)
2119 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2120 }
bfff6873
LC		 2121 if (NULL != first_not_zeroed)
2122 *first_not_zeroed = grp;
ac27a0ec 2123
5dee5437
TT		 2124 ext4_free_blocks_count_set(sbi->s_es,
2125 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2126 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2127 return 1;
2128}
2129
617ba13b 2130/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2131 * the superblock) which were deleted from all directories, but held open by
2132 * a process at the time of a crash. We walk the list and try to delete these
2133 * inodes at recovery time (only with a read-write filesystem).
2134 *
2135 * In order to keep the orphan inode chain consistent during traversal (in
2136 * case of crash during recovery), we link each inode into the superblock
2137 * orphan list_head and handle it the same way as an inode deletion during
2138 * normal operation (which journals the operations for us).
2139 *
2140 * We only do an iget() and an iput() on each inode, which is very safe if we
2141 * accidentally point at an in-use or already deleted inode. The worst that
2142 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2143 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2144 * e2fsck was run on this filesystem, and it must have already done the orphan
2145 * inode cleanup for us, so we can safely abort without any further action.
2146 */
2b2d6d01
TT		 2147static void ext4_orphan_cleanup(struct super_block *sb,
2148 struct ext4_super_block *es)
ac27a0ec
DK		 2149{
2150 unsigned int s_flags = sb->s_flags;
2151 int nr_orphans = 0, nr_truncates = 0;
2152#ifdef CONFIG_QUOTA
2153 int i;
2154#endif
2155 if (!es->s_last_orphan) {
2156 jbd_debug(4, "no orphan inodes to clean up\n");
2157 return;
2158 }
2159
a8f48a95 2160 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2161 ext4_msg(sb, KERN_ERR, "write access "
2162 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2163 return;
2164 }
2165
d39195c3
AG		 2166 /* Check if feature set would not allow a r/w mount */
2167 if (!ext4_feature_set_ok(sb, 0)) {
2168 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2169 "unknown ROCOMPAT features");
2170 return;
2171 }
2172
617ba13b 2173 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2174 /* don't clear list on RO mount w/ errors */
2175 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2176 jbd_debug(1, "Errors on filesystem, "
2177 "clearing orphan list.\n");
c25f9bc6
ES		 2178 es->s_last_orphan = 0;
2179 }
ac27a0ec
DK		 2180 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2181 return;
2182 }
2183
2184 if (s_flags & MS_RDONLY) {
b31e1552 2185 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2186 sb->s_flags &= ~MS_RDONLY;
2187 }
2188#ifdef CONFIG_QUOTA
2189 /* Needed for iput() to work correctly and not trash data */
2190 sb->s_flags |= MS_ACTIVE;
2191 /* Turn on quotas so that they are updated correctly */
2192 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2193 if (EXT4_SB(sb)->s_qf_names[i]) {
2194 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2195 if (ret < 0)
b31e1552
ES		 2196 ext4_msg(sb, KERN_ERR,
2197 "Cannot turn on journaled "
2198 "quota: error %d", ret);
ac27a0ec
DK		 2199 }
2200 }
2201#endif
2202
2203 while (es->s_last_orphan) {
2204 struct inode *inode;
2205
97bd42b9
JB		 2206 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2207 if (IS_ERR(inode)) {
ac27a0ec
DK		 2208 es->s_last_orphan = 0;
2209 break;
2210 }
2211
617ba13b 2212 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2213 dquot_initialize(inode);
ac27a0ec 2214 if (inode->i_nlink) {
b31e1552
ES		 2215 ext4_msg(sb, KERN_DEBUG,
2216 "%s: truncating inode %lu to %lld bytes",
46e665e9 2217 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2218 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2219 inode->i_ino, inode->i_size);
617ba13b 2220 ext4_truncate(inode);
ac27a0ec
DK		 2221 nr_truncates++;
2222 } else {
b31e1552
ES		 2223 ext4_msg(sb, KERN_DEBUG,
2224 "%s: deleting unreferenced inode %lu",
46e665e9 2225 __func__, inode->i_ino);
ac27a0ec
DK		 2226 jbd_debug(2, "deleting unreferenced inode %lu\n",
2227 inode->i_ino);
2228 nr_orphans++;
2229 }
2230 iput(inode); /* The delete magic happens here! */
2231 }
2232
2b2d6d01 2233#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2234
2235 if (nr_orphans)
b31e1552
ES		 2236 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2237 PLURAL(nr_orphans));
ac27a0ec 2238 if (nr_truncates)
b31e1552
ES		 2239 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2240 PLURAL(nr_truncates));
ac27a0ec
DK		 2241#ifdef CONFIG_QUOTA
2242 /* Turn quotas off */
2243 for (i = 0; i < MAXQUOTAS; i++) {
2244 if (sb_dqopt(sb)->files[i])
287a8095 2245 dquot_quota_off(sb, i);
ac27a0ec
DK		 2246 }
2247#endif
2248 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2249}
0b8e58a1 2250
cd2291a4
ES		 2251/*
2252 * Maximal extent format file size.
2253 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2254 * extent format containers, within a sector_t, and within i_blocks
2255 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2256 * so that won't be a limiting factor.
2257 *
f17722f9
LC		 2258 * However there is other limiting factor. We do store extents in the form
2259 * of starting block and length, hence the resulting length of the extent
2260 * covering maximum file size must fit into on-disk format containers as
2261 * well. Given that length is always by 1 unit bigger than max unit (because
2262 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2263 *
cd2291a4
ES		 2264 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2265 */
f287a1a5 2266static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2267{
2268 loff_t res;
2269 loff_t upper_limit = MAX_LFS_FILESIZE;
2270
2271 /* small i_blocks in vfs inode? */
f287a1a5 2272 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2273 /*
90c699a9 2274 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2275 * i_block represent total blocks in 512 bytes
2276 * 32 == size of vfs inode i_blocks * 8
2277 */
2278 upper_limit = (1LL << 32) - 1;
2279
2280 /* total blocks in file system block size */
2281 upper_limit >>= (blkbits - 9);
2282 upper_limit <<= blkbits;
2283 }
2284
f17722f9
LC		 2285 /*
2286 * 32-bit extent-start container, ee_block. We lower the maxbytes
2287 * by one fs block, so ee_len can cover the extent of maximum file
2288 * size
2289 */
2290 res = (1LL << 32) - 1;
cd2291a4 2291 res <<= blkbits;
cd2291a4
ES		 2292
2293 /* Sanity check against vm- & vfs- imposed limits */
2294 if (res > upper_limit)
2295 res = upper_limit;
2296
2297 return res;
2298}
ac27a0ec 2299
ac27a0ec 2300/*
cd2291a4 2301 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2302 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2303 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2304 */
f287a1a5 2305static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2306{
617ba13b 2307 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2308 int meta_blocks;
2309 loff_t upper_limit;
0b8e58a1
AD		 2310 /* This is calculated to be the largest file size for a dense, block
2311 * mapped file such that the file's total number of 512-byte sectors,
2312 * including data and all indirect blocks, does not exceed (2^48 - 1).
2313 *
2314 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2315 * number of 512-byte sectors of the file.
0fc1b451
AK		 2316 */
2317
f287a1a5 2318 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2319 /*
90c699a9 2320 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2321 * the inode i_block field represents total file blocks in
2322 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2323 */
2324 upper_limit = (1LL << 32) - 1;
2325
2326 /* total blocks in file system block size */
2327 upper_limit >>= (bits - 9);
2328
2329 } else {
8180a562
AK		 2330 /*
2331 * We use 48 bit ext4_inode i_blocks
2332 * With EXT4_HUGE_FILE_FL set the i_blocks
2333 * represent total number of blocks in
2334 * file system block size
2335 */
0fc1b451
AK		 2336 upper_limit = (1LL << 48) - 1;
2337
0fc1b451
AK		 2338 }
2339
2340 /* indirect blocks */
2341 meta_blocks = 1;
2342 /* double indirect blocks */
2343 meta_blocks += 1 + (1LL << (bits-2));
2344 /* tripple indirect blocks */
2345 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2346
2347 upper_limit -= meta_blocks;
2348 upper_limit <<= bits;
ac27a0ec
DK		 2349
2350 res += 1LL << (bits-2);
2351 res += 1LL << (2*(bits-2));
2352 res += 1LL << (3*(bits-2));
2353 res <<= bits;
2354 if (res > upper_limit)
2355 res = upper_limit;
0fc1b451
AK		 2356
2357 if (res > MAX_LFS_FILESIZE)
2358 res = MAX_LFS_FILESIZE;
2359
ac27a0ec
DK		 2360 return res;
2361}
2362
617ba13b 2363static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2364 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2365{
617ba13b 2366 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2367 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2368 int has_super = 0;
2369
2370 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2371
617ba13b 2372 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2373 nr < first_meta_bg)
70bbb3e0 2374 return logical_sb_block + nr + 1;
ac27a0ec 2375 bg = sbi->s_desc_per_block * nr;
617ba13b 2376 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2377 has_super = 1;
0b8e58a1 2378
617ba13b 2379 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2380}
2381
c9de560d
AT		 2382/**
2383 * ext4_get_stripe_size: Get the stripe size.
2384 * @sbi: In memory super block info
2385 *
2386 * If we have specified it via mount option, then
2387 * use the mount option value. If the value specified at mount time is
2388 * greater than the blocks per group use the super block value.
2389 * If the super block value is greater than blocks per group return 0.
2390 * Allocator needs it be less than blocks per group.
2391 *
2392 */
2393static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2394{
2395 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2396 unsigned long stripe_width =
2397 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2398 int ret;
c9de560d
AT		 2399
2400 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2401 ret = sbi->s_stripe;
2402 else if (stripe_width <= sbi->s_blocks_per_group)
2403 ret = stripe_width;
2404 else if (stride <= sbi->s_blocks_per_group)
2405 ret = stride;
2406 else
2407 ret = 0;
c9de560d 2408
3eb08658
DE		 2409 /*
2410 * If the stripe width is 1, this makes no sense and
2411 * we set it to 0 to turn off stripe handling code.
2412 */
2413 if (ret <= 1)
2414 ret = 0;
c9de560d 2415
3eb08658 2416 return ret;
c9de560d 2417}
ac27a0ec 2418
3197ebdb
TT		 2419/* sysfs supprt */
2420
2421struct ext4_attr {
2422 struct attribute attr;
2423 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2424 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2425 const char *, size_t);
2426 int offset;
2427};
2428
2429static int parse_strtoul(const char *buf,
2430 unsigned long max, unsigned long *value)
2431{
2432 char *endp;
2433
e7d2860b
AGR		 2434 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2435 endp = skip_spaces(endp);
3197ebdb
TT		 2436 if (*endp || *value > max)
2437 return -EINVAL;
2438
2439 return 0;
2440}
2441
2442static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2443 struct ext4_sb_info *sbi,
2444 char *buf)
2445{
2446 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2447 (s64) EXT4_C2B(sbi,
2448 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2449}
2450
2451static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2452 struct ext4_sb_info *sbi, char *buf)
2453{
2454 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2455
f613dfcb
TT		 2456 if (!sb->s_bdev->bd_part)
2457 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2458 return snprintf(buf, PAGE_SIZE, "%lu\n",
2459 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2460 sbi->s_sectors_written_start) >> 1);
2461}
2462
2463static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2464 struct ext4_sb_info *sbi, char *buf)
2465{
2466 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2467
f613dfcb
TT		 2468 if (!sb->s_bdev->bd_part)
2469 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2470 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2471 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2472 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2473 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2474}
2475
2476static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2477 struct ext4_sb_info *sbi,
2478 const char *buf, size_t count)
2479{
2480 unsigned long t;
2481
2482 if (parse_strtoul(buf, 0x40000000, &t))
2483 return -EINVAL;
2484
5dbd571d 2485 if (t && !is_power_of_2(t))
3197ebdb
TT		 2486 return -EINVAL;
2487
2488 sbi->s_inode_readahead_blks = t;
2489 return count;
2490}
2491
2492static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2493 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2494{
2495 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2496
2497 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2498}
2499
2500static ssize_t sbi_ui_store(struct ext4_attr *a,
2501 struct ext4_sb_info *sbi,
2502 const char *buf, size_t count)
2503{
2504 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2505 unsigned long t;
2506
2507 if (parse_strtoul(buf, 0xffffffff, &t))
2508 return -EINVAL;
2509 *ui = t;
2510 return count;
2511}
2512
2c0544b2
TT		 2513static ssize_t trigger_test_error(struct ext4_attr *a,
2514 struct ext4_sb_info *sbi,
2515 const char *buf, size_t count)
2516{
2517 int len = count;
2518
2519 if (!capable(CAP_SYS_ADMIN))
2520 return -EPERM;
2521
2522 if (len && buf[len-1] == '\n')
2523 len--;
2524
2525 if (len)
2526 ext4_error(sbi->s_sb, "%.*s", len, buf);
2527 return count;
2528}
2529
3197ebdb
TT		 2530#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2531static struct ext4_attr ext4_attr_##_name = { \
2532 .attr = {.name = __stringify(_name), .mode = _mode }, \
2533 .show = _show, \
2534 .store = _store, \
2535 .offset = offsetof(struct ext4_sb_info, _elname), \
2536}
2537#define EXT4_ATTR(name, mode, show, store) \
2538static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2539
857ac889 2540#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2541#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2542#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2543#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2544 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2545#define ATTR_LIST(name) &ext4_attr_##name.attr
2546
2547EXT4_RO_ATTR(delayed_allocation_blocks);
2548EXT4_RO_ATTR(session_write_kbytes);
2549EXT4_RO_ATTR(lifetime_write_kbytes);
2550EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2551 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2552EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2553EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2554EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2555EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2556EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2557EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2558EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2559EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
67a5da56 2560EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2561EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2562
2563static struct attribute *ext4_attrs[] = {
2564 ATTR_LIST(delayed_allocation_blocks),
2565 ATTR_LIST(session_write_kbytes),
2566 ATTR_LIST(lifetime_write_kbytes),
2567 ATTR_LIST(inode_readahead_blks),
11013911 2568 ATTR_LIST(inode_goal),
3197ebdb
TT		 2569 ATTR_LIST(mb_stats),
2570 ATTR_LIST(mb_max_to_scan),
2571 ATTR_LIST(mb_min_to_scan),
2572 ATTR_LIST(mb_order2_req),
2573 ATTR_LIST(mb_stream_req),
2574 ATTR_LIST(mb_group_prealloc),
55138e0b 2575 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2576 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2577 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2578 NULL,
2579};
2580
857ac889
LC		 2581/* Features this copy of ext4 supports */
2582EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2583EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2584EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2585
2586static struct attribute *ext4_feat_attrs[] = {
2587 ATTR_LIST(lazy_itable_init),
27ee40df 2588 ATTR_LIST(batched_discard),
5e7bbef1 2589 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2590 NULL,
2591};
2592
3197ebdb
TT		 2593static ssize_t ext4_attr_show(struct kobject *kobj,
2594 struct attribute *attr, char *buf)
2595{
2596 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2597 s_kobj);
2598 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2599
2600 return a->show ? a->show(a, sbi, buf) : 0;
2601}
2602
2603static ssize_t ext4_attr_store(struct kobject *kobj,
2604 struct attribute *attr,
2605 const char *buf, size_t len)
2606{
2607 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2608 s_kobj);
2609 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2610
2611 return a->store ? a->store(a, sbi, buf, len) : 0;
2612}
2613
2614static void ext4_sb_release(struct kobject *kobj)
2615{
2616 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2617 s_kobj);
2618 complete(&sbi->s_kobj_unregister);
2619}
2620
52cf25d0 2621static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2622 .show = ext4_attr_show,
2623 .store = ext4_attr_store,
2624};
2625
2626static struct kobj_type ext4_ktype = {
2627 .default_attrs = ext4_attrs,
2628 .sysfs_ops = &ext4_attr_ops,
2629 .release = ext4_sb_release,
2630};
2631
857ac889
LC		 2632static void ext4_feat_release(struct kobject *kobj)
2633{
2634 complete(&ext4_feat->f_kobj_unregister);
2635}
2636
2637static struct kobj_type ext4_feat_ktype = {
2638 .default_attrs = ext4_feat_attrs,
2639 .sysfs_ops = &ext4_attr_ops,
2640 .release = ext4_feat_release,
2641};
2642
a13fb1a4
ES		 2643/*
2644 * Check whether this filesystem can be mounted based on
2645 * the features present and the RDONLY/RDWR mount requested.
2646 * Returns 1 if this filesystem can be mounted as requested,
2647 * 0 if it cannot be.
2648 */
2649static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2650{
2651 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2652 ext4_msg(sb, KERN_ERR,
2653 "Couldn't mount because of "
2654 "unsupported optional features (%x)",
2655 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2656 ~EXT4_FEATURE_INCOMPAT_SUPP));
2657 return 0;
2658 }
2659
2660 if (readonly)
2661 return 1;
2662
2663 /* Check that feature set is OK for a read-write mount */
2664 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2665 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2666 "unsupported optional features (%x)",
2667 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2668 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2669 return 0;
2670 }
2671 /*
2672 * Large file size enabled file system can only be mounted
2673 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2674 */
2675 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2676 if (sizeof(blkcnt_t) < sizeof(u64)) {
2677 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2678 "cannot be mounted RDWR without "
2679 "CONFIG_LBDAF");
2680 return 0;
2681 }
2682 }
bab08ab9
TT		 2683 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2684 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2685 ext4_msg(sb, KERN_ERR,
2686 "Can't support bigalloc feature without "
2687 "extents feature\n");
2688 return 0;
2689 }
7c319d32
AK		 2690
2691#ifndef CONFIG_QUOTA
2692 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2693 !readonly) {
2694 ext4_msg(sb, KERN_ERR,
2695 "Filesystem with quota feature cannot be mounted RDWR "
2696 "without CONFIG_QUOTA");
2697 return 0;
2698 }
2699#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2700 return 1;
2701}
2702
66e61a9e
TT		 2703/*
2704 * This function is called once a day if we have errors logged
2705 * on the file system
2706 */
2707static void print_daily_error_info(unsigned long arg)
2708{
2709 struct super_block *sb = (struct super_block *) arg;
2710 struct ext4_sb_info *sbi;
2711 struct ext4_super_block *es;
2712
2713 sbi = EXT4_SB(sb);
2714 es = sbi->s_es;
2715
2716 if (es->s_error_count)
2717 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2718 le32_to_cpu(es->s_error_count));
2719 if (es->s_first_error_time) {
2720 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2721 sb->s_id, le32_to_cpu(es->s_first_error_time),
2722 (int) sizeof(es->s_first_error_func),
2723 es->s_first_error_func,
2724 le32_to_cpu(es->s_first_error_line));
2725 if (es->s_first_error_ino)
2726 printk(": inode %u",
2727 le32_to_cpu(es->s_first_error_ino));
2728 if (es->s_first_error_block)
2729 printk(": block %llu", (unsigned long long)
2730 le64_to_cpu(es->s_first_error_block));
2731 printk("\n");
2732 }
2733 if (es->s_last_error_time) {
2734 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2735 sb->s_id, le32_to_cpu(es->s_last_error_time),
2736 (int) sizeof(es->s_last_error_func),
2737 es->s_last_error_func,
2738 le32_to_cpu(es->s_last_error_line));
2739 if (es->s_last_error_ino)
2740 printk(": inode %u",
2741 le32_to_cpu(es->s_last_error_ino));
2742 if (es->s_last_error_block)
2743 printk(": block %llu", (unsigned long long)
2744 le64_to_cpu(es->s_last_error_block));
2745 printk("\n");
2746 }
2747 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2748}
2749
bfff6873
LC		 2750/* Find next suitable group and run ext4_init_inode_table */
2751static int ext4_run_li_request(struct ext4_li_request *elr)
2752{
2753 struct ext4_group_desc *gdp = NULL;
2754 ext4_group_t group, ngroups;
2755 struct super_block *sb;
2756 unsigned long timeout = 0;
2757 int ret = 0;
2758
2759 sb = elr->lr_super;
2760 ngroups = EXT4_SB(sb)->s_groups_count;
2761
8e8ad8a5 2762 sb_start_write(sb);
bfff6873
LC		 2763 for (group = elr->lr_next_group; group < ngroups; group++) {
2764 gdp = ext4_get_group_desc(sb, group, NULL);
2765 if (!gdp) {
2766 ret = 1;
2767 break;
2768 }
2769
2770 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2771 break;
2772 }
2773
2774 if (group == ngroups)
2775 ret = 1;
2776
2777 if (!ret) {
2778 timeout = jiffies;
2779 ret = ext4_init_inode_table(sb, group,
2780 elr->lr_timeout ? 0 : 1);
2781 if (elr->lr_timeout == 0) {
51ce6511
LC		 2782 timeout = (jiffies - timeout) *
2783 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2784 elr->lr_timeout = timeout;
2785 }
2786 elr->lr_next_sched = jiffies + elr->lr_timeout;
2787 elr->lr_next_group = group + 1;
2788 }
8e8ad8a5 2789 sb_end_write(sb);
bfff6873
LC		 2790
2791 return ret;
2792}
2793
2794/*
2795 * Remove lr_request from the list_request and free the
4ed5c033 2796 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2797 */
2798static void ext4_remove_li_request(struct ext4_li_request *elr)
2799{
2800 struct ext4_sb_info *sbi;
2801
2802 if (!elr)
2803 return;
2804
2805 sbi = elr->lr_sbi;
2806
2807 list_del(&elr->lr_request);
2808 sbi->s_li_request = NULL;
2809 kfree(elr);
2810}
2811
2812static void ext4_unregister_li_request(struct super_block *sb)
2813{
1bb933fb
LC		 2814 mutex_lock(&ext4_li_mtx);
2815 if (!ext4_li_info) {
2816 mutex_unlock(&ext4_li_mtx);
bfff6873 2817 return;
1bb933fb 2818 }
bfff6873
LC		 2819
2820 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2821 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2822 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2823 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2824}
2825
8f1f7453
ES		 2826static struct task_struct *ext4_lazyinit_task;
2827
bfff6873
LC		 2828/*
2829 * This is the function where ext4lazyinit thread lives. It walks
2830 * through the request list searching for next scheduled filesystem.
2831 * When such a fs is found, run the lazy initialization request
2832 * (ext4_rn_li_request) and keep track of the time spend in this
2833 * function. Based on that time we compute next schedule time of
2834 * the request. When walking through the list is complete, compute
2835 * next waking time and put itself into sleep.
2836 */
2837static int ext4_lazyinit_thread(void *arg)
2838{
2839 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2840 struct list_head *pos, *n;
2841 struct ext4_li_request *elr;
4ed5c033 2842 unsigned long next_wakeup, cur;
bfff6873
LC		 2843
2844 BUG_ON(NULL == eli);
2845
bfff6873
LC		 2846cont_thread:
2847 while (true) {
2848 next_wakeup = MAX_JIFFY_OFFSET;
2849
2850 mutex_lock(&eli->li_list_mtx);
2851 if (list_empty(&eli->li_request_list)) {
2852 mutex_unlock(&eli->li_list_mtx);
2853 goto exit_thread;
2854 }
2855
2856 list_for_each_safe(pos, n, &eli->li_request_list) {
2857 elr = list_entry(pos, struct ext4_li_request,
2858 lr_request);
2859
b2c78cd0
TT		 2860 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2861 if (ext4_run_li_request(elr) != 0) {
2862 /* error, remove the lazy_init job */
2863 ext4_remove_li_request(elr);
2864 continue;
2865 }
bfff6873
LC		 2866 }
2867
2868 if (time_before(elr->lr_next_sched, next_wakeup))
2869 next_wakeup = elr->lr_next_sched;
2870 }
2871 mutex_unlock(&eli->li_list_mtx);
2872
a0acae0e 2873 try_to_freeze();
bfff6873 2874
4ed5c033
LC		 2875 cur = jiffies;
2876 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2877 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2878 cond_resched();
2879 continue;
2880 }
2881
4ed5c033
LC		 2882 schedule_timeout_interruptible(next_wakeup - cur);
2883
8f1f7453
ES		 2884 if (kthread_should_stop()) {
2885 ext4_clear_request_list();
2886 goto exit_thread;
2887 }
bfff6873
LC		 2888 }
2889
2890exit_thread:
2891 /*
2892 * It looks like the request list is empty, but we need
2893 * to check it under the li_list_mtx lock, to prevent any
2894 * additions into it, and of course we should lock ext4_li_mtx
2895 * to atomically free the list and ext4_li_info, because at
2896 * this point another ext4 filesystem could be registering
2897 * new one.
2898 */
2899 mutex_lock(&ext4_li_mtx);
2900 mutex_lock(&eli->li_list_mtx);
2901 if (!list_empty(&eli->li_request_list)) {
2902 mutex_unlock(&eli->li_list_mtx);
2903 mutex_unlock(&ext4_li_mtx);
2904 goto cont_thread;
2905 }
2906 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2907 kfree(ext4_li_info);
2908 ext4_li_info = NULL;
2909 mutex_unlock(&ext4_li_mtx);
2910
2911 return 0;
2912}
2913
2914static void ext4_clear_request_list(void)
2915{
2916 struct list_head *pos, *n;
2917 struct ext4_li_request *elr;
2918
2919 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2920 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2921 elr = list_entry(pos, struct ext4_li_request,
2922 lr_request);
2923 ext4_remove_li_request(elr);
2924 }
2925 mutex_unlock(&ext4_li_info->li_list_mtx);
2926}
2927
2928static int ext4_run_lazyinit_thread(void)
2929{
8f1f7453
ES		 2930 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2931 ext4_li_info, "ext4lazyinit");
2932 if (IS_ERR(ext4_lazyinit_task)) {
2933 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2934 ext4_clear_request_list();
bfff6873
LC		 2935 kfree(ext4_li_info);
2936 ext4_li_info = NULL;
92b97816 2937 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2938 "initialization thread\n",
2939 err);
2940 return err;
2941 }
2942 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2943 return 0;
2944}
2945
2946/*
2947 * Check whether it make sense to run itable init. thread or not.
2948 * If there is at least one uninitialized inode table, return
2949 * corresponding group number, else the loop goes through all
2950 * groups and return total number of groups.
2951 */
2952static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2953{
2954 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2955 struct ext4_group_desc *gdp = NULL;
2956
2957 for (group = 0; group < ngroups; group++) {
2958 gdp = ext4_get_group_desc(sb, group, NULL);
2959 if (!gdp)
2960 continue;
2961
2962 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2963 break;
2964 }
2965
2966 return group;
2967}
2968
2969static int ext4_li_info_new(void)
2970{
2971 struct ext4_lazy_init *eli = NULL;
2972
2973 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2974 if (!eli)
2975 return -ENOMEM;
2976
bfff6873
LC		 2977 INIT_LIST_HEAD(&eli->li_request_list);
2978 mutex_init(&eli->li_list_mtx);
2979
bfff6873
LC		 2980 eli->li_state |= EXT4_LAZYINIT_QUIT;
2981
2982 ext4_li_info = eli;
2983
2984 return 0;
2985}
2986
2987static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2988 ext4_group_t start)
2989{
2990 struct ext4_sb_info *sbi = EXT4_SB(sb);
2991 struct ext4_li_request *elr;
2992 unsigned long rnd;
2993
2994 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2995 if (!elr)
2996 return NULL;
2997
2998 elr->lr_super = sb;
2999 elr->lr_sbi = sbi;
3000 elr->lr_next_group = start;
3001
3002 /*
3003 * Randomize first schedule time of the request to
3004 * spread the inode table initialization requests
3005 * better.
3006 */
3007 get_random_bytes(&rnd, sizeof(rnd));
3008 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3009 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3010
3011 return elr;
3012}
3013
3014static int ext4_register_li_request(struct super_block *sb,
3015 ext4_group_t first_not_zeroed)
3016{
3017 struct ext4_sb_info *sbi = EXT4_SB(sb);
3018 struct ext4_li_request *elr;
3019 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3020 int ret = 0;
bfff6873 3021
51ce6511
LC		 3022 if (sbi->s_li_request != NULL) {
3023 /*
3024 * Reset timeout so it can be computed again, because
3025 * s_li_wait_mult might have changed.
3026 */
3027 sbi->s_li_request->lr_timeout = 0;
beed5ecb 3028 return 0;
51ce6511 3029 }
bfff6873
LC		 3030
3031 if (first_not_zeroed == ngroups ||
3032 (sb->s_flags & MS_RDONLY) ||
55ff3840 3033 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 3034 return 0;
bfff6873
LC		 3035
3036 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 3037 if (!elr)
3038 return -ENOMEM;
bfff6873
LC		 3039
3040 mutex_lock(&ext4_li_mtx);
3041
3042 if (NULL == ext4_li_info) {
3043 ret = ext4_li_info_new();
3044 if (ret)
3045 goto out;
3046 }
3047
3048 mutex_lock(&ext4_li_info->li_list_mtx);
3049 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3050 mutex_unlock(&ext4_li_info->li_list_mtx);
3051
3052 sbi->s_li_request = elr;
46e4690b
TM		 3053 /*
3054 * set elr to NULL here since it has been inserted to
3055 * the request_list and the removal and free of it is
3056 * handled by ext4_clear_request_list from now on.
3057 */
3058 elr = NULL;
bfff6873
LC		 3059
3060 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3061 ret = ext4_run_lazyinit_thread();
3062 if (ret)
3063 goto out;
3064 }
bfff6873 3065out:
beed5ecb
NK		 3066 mutex_unlock(&ext4_li_mtx);
3067 if (ret)
bfff6873 3068 kfree(elr);
bfff6873
LC		 3069 return ret;
3070}
3071
3072/*
3073 * We do not need to lock anything since this is called on
3074 * module unload.
3075 */
3076static void ext4_destroy_lazyinit_thread(void)
3077{
3078 /*
3079 * If thread exited earlier
3080 * there's nothing to be done.
3081 */
8f1f7453 3082 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3083 return;
3084
8f1f7453 3085 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3086}
3087
25ed6e8a
DW		 3088static int set_journal_csum_feature_set(struct super_block *sb)
3089{
3090 int ret = 1;
3091 int compat, incompat;
3092 struct ext4_sb_info *sbi = EXT4_SB(sb);
3093
3094 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3095 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3096 /* journal checksum v2 */
3097 compat = 0;
3098 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3099 } else {
3100 /* journal checksum v1 */
3101 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3102 incompat = 0;
3103 }
3104
3105 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3106 ret = jbd2_journal_set_features(sbi->s_journal,
3107 compat, 0,
3108 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3109 incompat);
3110 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3111 ret = jbd2_journal_set_features(sbi->s_journal,
3112 compat, 0,
3113 incompat);
3114 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3115 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3116 } else {
3117 jbd2_journal_clear_features(sbi->s_journal,
3118 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3119 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3120 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3121 }
3122
3123 return ret;
3124}
3125
952fc18e
TT		 3126/*
3127 * Note: calculating the overhead so we can be compatible with
3128 * historical BSD practice is quite difficult in the face of
3129 * clusters/bigalloc. This is because multiple metadata blocks from
3130 * different block group can end up in the same allocation cluster.
3131 * Calculating the exact overhead in the face of clustered allocation
3132 * requires either O(all block bitmaps) in memory or O(number of block
3133 * groups**2) in time. We will still calculate the superblock for
3134 * older file systems --- and if we come across with a bigalloc file
3135 * system with zero in s_overhead_clusters the estimate will be close to
3136 * correct especially for very large cluster sizes --- but for newer
3137 * file systems, it's better to calculate this figure once at mkfs
3138 * time, and store it in the superblock. If the superblock value is
3139 * present (even for non-bigalloc file systems), we will use it.
3140 */
3141static int count_overhead(struct super_block *sb, ext4_group_t grp,
3142 char *buf)
3143{
3144 struct ext4_sb_info *sbi = EXT4_SB(sb);
3145 struct ext4_group_desc *gdp;
3146 ext4_fsblk_t first_block, last_block, b;
3147 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3148 int s, j, count = 0;
3149
0548bbb8
TT		 3150 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3151 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3152 sbi->s_itb_per_group + 2);
3153
952fc18e
TT		 3154 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3155 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3156 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3157 for (i = 0; i < ngroups; i++) {
3158 gdp = ext4_get_group_desc(sb, i, NULL);
3159 b = ext4_block_bitmap(sb, gdp);
3160 if (b >= first_block && b <= last_block) {
3161 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3162 count++;
3163 }
3164 b = ext4_inode_bitmap(sb, gdp);
3165 if (b >= first_block && b <= last_block) {
3166 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3167 count++;
3168 }
3169 b = ext4_inode_table(sb, gdp);
3170 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3171 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3172 int c = EXT4_B2C(sbi, b - first_block);
3173 ext4_set_bit(c, buf);
3174 count++;
3175 }
3176 if (i != grp)
3177 continue;
3178 s = 0;
3179 if (ext4_bg_has_super(sb, grp)) {
3180 ext4_set_bit(s++, buf);
3181 count++;
3182 }
3183 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3184 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3185 count++;
3186 }
3187 }
3188 if (!count)
3189 return 0;
3190 return EXT4_CLUSTERS_PER_GROUP(sb) -
3191 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3192}
3193
3194/*
3195 * Compute the overhead and stash it in sbi->s_overhead
3196 */
3197int ext4_calculate_overhead(struct super_block *sb)
3198{
3199 struct ext4_sb_info *sbi = EXT4_SB(sb);
3200 struct ext4_super_block *es = sbi->s_es;
3201 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3202 ext4_fsblk_t overhead = 0;
3203 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3204
3205 memset(buf, 0, PAGE_SIZE);
3206 if (!buf)
3207 return -ENOMEM;
3208
3209 /*
3210 * Compute the overhead (FS structures). This is constant
3211 * for a given filesystem unless the number of block groups
3212 * changes so we cache the previous value until it does.
3213 */
3214
3215 /*
3216 * All of the blocks before first_data_block are overhead
3217 */
3218 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3219
3220 /*
3221 * Add the overhead found in each block group
3222 */
3223 for (i = 0; i < ngroups; i++) {
3224 int blks;
3225
3226 blks = count_overhead(sb, i, buf);
3227 overhead += blks;
3228 if (blks)
3229 memset(buf, 0, PAGE_SIZE);
3230 cond_resched();
3231 }
3232 sbi->s_overhead = overhead;
3233 smp_wmb();
3234 free_page((unsigned long) buf);
3235 return 0;
3236}
3237
2b2d6d01 3238static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3239{
d4c402d9 3240 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3241 struct buffer_head *bh;
617ba13b
MC		 3242 struct ext4_super_block *es = NULL;
3243 struct ext4_sb_info *sbi;
3244 ext4_fsblk_t block;
3245 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3246 ext4_fsblk_t logical_sb_block;
ac27a0ec 3247 unsigned long offset = 0;
ac27a0ec
DK		 3248 unsigned long journal_devnum = 0;
3249 unsigned long def_mount_opts;
3250 struct inode *root;
9f6200bb 3251 char *cp;
0390131b 3252 const char *descr;
dcc7dae3 3253 int ret = -ENOMEM;
281b5995 3254 int blocksize, clustersize;
4ec11028
TT		 3255 unsigned int db_count;
3256 unsigned int i;
281b5995 3257 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3258 __u64 blocks_count;
833f4077 3259 int err;
b3881f74 3260 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3261 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3262
3263 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3264 if (!sbi)
dcc7dae3 3265 goto out_free_orig;
705895b6
PE		 3266
3267 sbi->s_blockgroup_lock =
3268 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3269 if (!sbi->s_blockgroup_lock) {
3270 kfree(sbi);
dcc7dae3 3271 goto out_free_orig;
705895b6 3272 }
ac27a0ec 3273 sb->s_fs_info = sbi;
2c0544b2 3274 sbi->s_sb = sb;
240799cd 3275 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3276 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3277 if (sb->s_bdev->bd_part)
3278 sbi->s_sectors_written_start =
3279 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3280
9f6200bb
TT		 3281 /* Cleanup superblock name */
3282 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3283 *cp = '!';
3284
dcc7dae3 3285 ret = -EINVAL;
617ba13b 3286 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3287 if (!blocksize) {
b31e1552 3288 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3289 goto out_fail;
3290 }
3291
3292 /*
617ba13b 3293 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3294 * block sizes. We need to calculate the offset from buffer start.
3295 */
617ba13b 3296 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3297 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3298 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3299 } else {
70bbb3e0 3300 logical_sb_block = sb_block;
ac27a0ec
DK		 3301 }
3302
70bbb3e0 3303 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3304 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3305 goto out_fail;
3306 }
3307 /*
3308 * Note: s_es must be initialized as soon as possible because
617ba13b 3309 * some ext4 macro-instructions depend on its value
ac27a0ec 3310 */
2716b802 3311 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3312 sbi->s_es = es;
3313 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3314 if (sb->s_magic != EXT4_SUPER_MAGIC)
3315 goto cantfind_ext4;
afc32f7e 3316 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3317
feb0ab32
DW		 3318 /* Warn if metadata_csum and gdt_csum are both set. */
3319 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3320 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3321 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3322 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3323 "redundant flags; please run fsck.");
3324
d25425f8
DW		 3325 /* Check for a known checksum algorithm */
3326 if (!ext4_verify_csum_type(sb, es)) {
3327 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3328 "unknown checksum algorithm.");
3329 silent = 1;
3330 goto cantfind_ext4;
3331 }
3332
0441984a
DW		 3333 /* Load the checksum driver */
3334 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3335 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3336 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3337 if (IS_ERR(sbi->s_chksum_driver)) {
3338 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3339 ret = PTR_ERR(sbi->s_chksum_driver);
3340 sbi->s_chksum_driver = NULL;
3341 goto failed_mount;
3342 }
3343 }
3344
a9c47317
DW		 3345 /* Check superblock checksum */
3346 if (!ext4_superblock_csum_verify(sb, es)) {
3347 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3348 "invalid superblock checksum. Run e2fsck?");
3349 silent = 1;
3350 goto cantfind_ext4;
3351 }
3352
3353 /* Precompute checksum seed for all metadata */
3354 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3355 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3356 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3357 sizeof(es->s_uuid));
3358
ac27a0ec
DK		 3359 /* Set defaults before we parse the mount options */
3360 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3361 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3362 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3363 set_opt(sb, DEBUG);
87f26807 3364 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3365 set_opt(sb, GRPID);
617ba13b 3366 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3367 set_opt(sb, NO_UID32);
ea663336 3368 /* xattr user namespace & acls are now defaulted on */
03010a33 3369#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3370 set_opt(sb, XATTR_USER);
2e7842b8 3371#endif
03010a33 3372#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3373 set_opt(sb, POSIX_ACL);
2e7842b8 3374#endif
6fd7a467 3375 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3376 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3377 set_opt(sb, JOURNAL_DATA);
617ba13b 3378 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3379 set_opt(sb, ORDERED_DATA);
617ba13b 3380 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3381 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3382
3383 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3384 set_opt(sb, ERRORS_PANIC);
bb4f397a 3385 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3386 set_opt(sb, ERRORS_CONT);
bb4f397a 3387 else
fd8c37ec 3388 set_opt(sb, ERRORS_RO);
8b67f04a 3389 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3390 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3391 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3392 set_opt(sb, DISCARD);
ac27a0ec 3393
08cefc7a
EB		 3394 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3395 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3396 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3397 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3398 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3399
8b67f04a 3400 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3401 set_opt(sb, BARRIER);
ac27a0ec 3402
dd919b98
AK		 3403 /*
3404 * enable delayed allocation by default
3405 * Use -o nodelalloc to turn it off
3406 */
bc0b75f7 3407 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3408 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3409 set_opt(sb, DELALLOC);
dd919b98 3410
51ce6511
LC		 3411 /*
3412 * set default s_li_wait_mult for lazyinit, for the case there is
3413 * no mount option specified.
3414 */
3415 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3416
8b67f04a 3417 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3418 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3419 ext4_msg(sb, KERN_WARNING,
3420 "failed to parse options in superblock: %s",
3421 sbi->s_es->s_mount_opts);
3422 }
5a916be1 3423 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3424 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3425 &journal_ioprio, 0))
ac27a0ec
DK		 3426 goto failed_mount;
3427
56889787
TT		 3428 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3429 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3430 "with data=journal disables delayed "
3431 "allocation and O_DIRECT support!\n");
3432 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3433 ext4_msg(sb, KERN_ERR, "can't mount with "
3434 "both data=journal and delalloc");
3435 goto failed_mount;
3436 }
3437 if (test_opt(sb, DIOREAD_NOLOCK)) {
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, DELALLOC))
3443 clear_opt(sb, DELALLOC);
3444 }
3445
3446 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3447 if (test_opt(sb, DIOREAD_NOLOCK)) {
3448 if (blocksize < PAGE_SIZE) {
3449 ext4_msg(sb, KERN_ERR, "can't mount with "
3450 "dioread_nolock if block size != PAGE_SIZE");
3451 goto failed_mount;
3452 }
3453 }
3454
ac27a0ec 3455 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3456 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3457
617ba13b
MC		 3458 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3459 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3460 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3461 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3462 ext4_msg(sb, KERN_WARNING,
3463 "feature flags set on rev 0 fs, "
3464 "running e2fsck is recommended");
469108ff 3465
2035e776
TT		 3466 if (IS_EXT2_SB(sb)) {
3467 if (ext2_feature_set_ok(sb))
3468 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3469 "using the ext4 subsystem");
3470 else {
3471 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3472 "to feature incompatibilities");
3473 goto failed_mount;
3474 }
3475 }
3476
3477 if (IS_EXT3_SB(sb)) {
3478 if (ext3_feature_set_ok(sb))
3479 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3480 "using the ext4 subsystem");
3481 else {
3482 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3483 "to feature incompatibilities");
3484 goto failed_mount;
3485 }
3486 }
3487
ac27a0ec
DK		 3488 /*
3489 * Check feature flags regardless of the revision level, since we
3490 * previously didn't change the revision level when setting the flags,
3491 * so there is a chance incompat flags are set on a rev 0 filesystem.
3492 */
a13fb1a4 3493 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3494 goto failed_mount;
a13fb1a4 3495
617ba13b
MC		 3496 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3497 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3498 ext4_msg(sb, KERN_ERR,
3499 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3500 goto failed_mount;
3501 }
3502
ac27a0ec 3503 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3504 /* Validate the filesystem blocksize */
3505 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3506 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3507 blocksize);
ac27a0ec
DK		 3508 goto failed_mount;
3509 }
3510
2b2d6d01 3511 brelse(bh);
70bbb3e0
AM		 3512 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3513 offset = do_div(logical_sb_block, blocksize);
3514 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3515 if (!bh) {
b31e1552
ES		 3516 ext4_msg(sb, KERN_ERR,
3517 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3518 goto failed_mount;
3519 }
2716b802 3520 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3521 sbi->s_es = es;
617ba13b 3522 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3523 ext4_msg(sb, KERN_ERR,
3524 "Magic mismatch, very weird!");
ac27a0ec
DK		 3525 goto failed_mount;
3526 }
3527 }
3528
a13fb1a4
ES		 3529 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3530 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3531 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3532 has_huge_files);
3533 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3534
617ba13b
MC		 3535 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3536 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3537 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3538 } else {
3539 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3540 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3541 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3542 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3543 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3544 ext4_msg(sb, KERN_ERR,
3545 "unsupported inode size: %d",
2b2d6d01 3546 sbi->s_inode_size);
ac27a0ec
DK		 3547 goto failed_mount;
3548 }
ef7f3835
KS		 3549 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3550 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3551 }
0b8e58a1 3552
0d1ee42f
AR		 3553 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3554 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3555 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3556 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3557 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3558 ext4_msg(sb, KERN_ERR,
3559 "unsupported descriptor size %lu",
0d1ee42f
AR		 3560 sbi->s_desc_size);
3561 goto failed_mount;
3562 }
3563 } else
3564 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3565
ac27a0ec 3566 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3567 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3568 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3569 goto cantfind_ext4;
0b8e58a1 3570
617ba13b 3571 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3572 if (sbi->s_inodes_per_block == 0)
617ba13b 3573 goto cantfind_ext4;
ac27a0ec
DK		 3574 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3575 sbi->s_inodes_per_block;
0d1ee42f 3576 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3577 sbi->s_sbh = bh;
3578 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3579 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3580 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3581
2b2d6d01 3582 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3583 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3584 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3585 i = le32_to_cpu(es->s_flags);
3586 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3587 sbi->s_hash_unsigned = 3;
3588 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3589#ifdef __CHAR_UNSIGNED__
3590 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3591 sbi->s_hash_unsigned = 3;
3592#else
3593 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3594#endif
f99b2589 3595 }
ac27a0ec 3596
281b5995
TT		 3597 /* Handle clustersize */
3598 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3599 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3600 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3601 if (has_bigalloc) {
3602 if (clustersize < blocksize) {
3603 ext4_msg(sb, KERN_ERR,
3604 "cluster size (%d) smaller than "
3605 "block size (%d)", clustersize, blocksize);
3606 goto failed_mount;
3607 }
3608 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3609 le32_to_cpu(es->s_log_block_size);
3610 sbi->s_clusters_per_group =
3611 le32_to_cpu(es->s_clusters_per_group);
3612 if (sbi->s_clusters_per_group > blocksize * 8) {
3613 ext4_msg(sb, KERN_ERR,
3614 "#clusters per group too big: %lu",
3615 sbi->s_clusters_per_group);
3616 goto failed_mount;
3617 }
3618 if (sbi->s_blocks_per_group !=
3619 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3620 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3621 "clusters per group (%lu) inconsistent",
3622 sbi->s_blocks_per_group,
3623 sbi->s_clusters_per_group);
3624 goto failed_mount;
3625 }
3626 } else {
3627 if (clustersize != blocksize) {
3628 ext4_warning(sb, "fragment/cluster size (%d) != "
3629 "block size (%d)", clustersize,
3630 blocksize);
3631 clustersize = blocksize;
3632 }
3633 if (sbi->s_blocks_per_group > blocksize * 8) {
3634 ext4_msg(sb, KERN_ERR,
3635 "#blocks per group too big: %lu",
3636 sbi->s_blocks_per_group);
3637 goto failed_mount;
3638 }
3639 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3640 sbi->s_cluster_bits = 0;
ac27a0ec 3641 }
281b5995
TT		 3642 sbi->s_cluster_ratio = clustersize / blocksize;
3643
ac27a0ec 3644 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3645 ext4_msg(sb, KERN_ERR,
3646 "#inodes per group too big: %lu",
2b2d6d01 3647 sbi->s_inodes_per_group);
ac27a0ec
DK		 3648 goto failed_mount;
3649 }
3650
bf43d84b
ES		 3651 /*
3652 * Test whether we have more sectors than will fit in sector_t,
3653 * and whether the max offset is addressable by the page cache.
3654 */
5a9ae68a 3655 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3656 ext4_blocks_count(es));
5a9ae68a 3657 if (err) {
b31e1552 3658 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3659 " too large to mount safely on this system");
ac27a0ec 3660 if (sizeof(sector_t) < 8)
90c699a9 3661 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3662 ret = err;
ac27a0ec
DK		 3663 goto failed_mount;
3664 }
3665
617ba13b
MC		 3666 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3667 goto cantfind_ext4;
e7c95593 3668
0f2ddca6
FTN		 3669 /* check blocks count against device size */
3670 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3671 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3672 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3673 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3674 ext4_blocks_count(es), blocks_count);
3675 goto failed_mount;
3676 }
3677
0b8e58a1
AD		 3678 /*
3679 * It makes no sense for the first data block to be beyond the end
3680 * of the filesystem.
3681 */
3682 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3683 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3684 "block %u is beyond end of filesystem (%llu)",
3685 le32_to_cpu(es->s_first_data_block),
3686 ext4_blocks_count(es));
e7c95593
ES		 3687 goto failed_mount;
3688 }
bd81d8ee
LV		 3689 blocks_count = (ext4_blocks_count(es) -
3690 le32_to_cpu(es->s_first_data_block) +
3691 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3692 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3693 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3694 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3695 "(block count %llu, first data block %u, "
b31e1552 3696 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3697 ext4_blocks_count(es),
3698 le32_to_cpu(es->s_first_data_block),
3699 EXT4_BLOCKS_PER_GROUP(sb));
3700 goto failed_mount;
3701 }
bd81d8ee 3702 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3703 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3704 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3705 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3706 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3707 sbi->s_group_desc = ext4_kvmalloc(db_count *
3708 sizeof(struct buffer_head *),
3709 GFP_KERNEL);
ac27a0ec 3710 if (sbi->s_group_desc == NULL) {
b31e1552 3711 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3712 ret = -ENOMEM;
ac27a0ec
DK		 3713 goto failed_mount;
3714 }
3715
9f6200bb
TT		 3716 if (ext4_proc_root)
3717 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3718
66acdcf4
TT		 3719 if (sbi->s_proc)
3720 proc_create_data("options", S_IRUGO, sbi->s_proc,
3721 &ext4_seq_options_fops, sb);
3722
705895b6 3723 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3724
3725 for (i = 0; i < db_count; i++) {
70bbb3e0 3726 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3727 sbi->s_group_desc[i] = sb_bread(sb, block);
3728 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3729 ext4_msg(sb, KERN_ERR,
3730 "can't read group descriptor %d", i);
ac27a0ec
DK		 3731 db_count = i;
3732 goto failed_mount2;
3733 }
3734 }
bfff6873 3735 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3736 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3737 goto failed_mount2;
3738 }
772cb7c8
JS		 3739 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3740 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3741 ext4_msg(sb, KERN_ERR,
3742 "unable to initialize "
3743 "flex_bg meta info!");
772cb7c8
JS		 3744 goto failed_mount2;
3745 }
3746
ac27a0ec
DK		 3747 sbi->s_gdb_count = db_count;
3748 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3749 spin_lock_init(&sbi->s_next_gen_lock);
3750
04496411
TM		 3751 init_timer(&sbi->s_err_report);
3752 sbi->s_err_report.function = print_daily_error_info;
3753 sbi->s_err_report.data = (unsigned long) sb;
3754
57042651 3755 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3756 ext4_count_free_clusters(sb));
ce7e010a
TT		 3757 if (!err) {
3758 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3759 ext4_count_free_inodes(sb));
3760 }
3761 if (!err) {
3762 err = percpu_counter_init(&sbi->s_dirs_counter,
3763 ext4_count_dirs(sb));
3764 }
3765 if (!err) {
57042651 3766 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3767 }
3768 if (err) {
3769 ext4_msg(sb, KERN_ERR, "insufficient memory");
2cde417d 3770 ret = err;
ce7e010a
TT		 3771 goto failed_mount3;
3772 }
3773
c9de560d 3774 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3775 sbi->s_max_writeback_mb_bump = 128;
67a5da56 3776 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3777
ac27a0ec
DK		 3778 /*
3779 * set up enough so that it can read an inode
3780 */
9ca92389
TT		 3781 if (!test_opt(sb, NOLOAD) &&
3782 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3783 sb->s_op = &ext4_sops;
3784 else
3785 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3786 sb->s_export_op = &ext4_export_ops;
3787 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3788#ifdef CONFIG_QUOTA
617ba13b
MC		 3789 sb->s_qcop = &ext4_qctl_operations;
3790 sb->dq_op = &ext4_quota_operations;
7c319d32
AK		 3791
3792 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3793 /* Use qctl operations for hidden quota files. */
3794 sb->s_qcop = &ext4_qctl_sysfile_operations;
3795 }
ac27a0ec 3796#endif
f2fa2ffc
AK		 3797 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3798
ac27a0ec 3799 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3800 mutex_init(&sbi->s_orphan_lock);
8f82f840 3801 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3802
3803 sb->s_root = NULL;
3804
3805 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3806 EXT4_HAS_INCOMPAT_FEATURE(sb,
3807 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3808
c5e06d10
JL		 3809 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3810 !(sb->s_flags & MS_RDONLY))
3811 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3812 goto failed_mount3;
3813
ac27a0ec
DK		 3814 /*
3815 * The first inode we look at is the journal inode. Don't try
3816 * root first: it may be modified in the journal!
3817 */
3818 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3819 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3820 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3821 goto failed_mount3;
0390131b
FM		 3822 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3823 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3824 ext4_msg(sb, KERN_ERR, "required journal recovery "
3825 "suppressed and not mounted read-only");
744692dc 3826 goto failed_mount_wq;
ac27a0ec 3827 } else {
fd8c37ec 3828 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3829 sbi->s_journal = NULL;
3830 needs_recovery = 0;
3831 goto no_journal;
ac27a0ec
DK		 3832 }
3833
f32aaf2d 3834 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3835 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3836 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3837 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3838 goto failed_mount_wq;
eb40a09c
JS		 3839 }
3840
25ed6e8a
DW		 3841 if (!set_journal_csum_feature_set(sb)) {
3842 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3843 "feature set");
3844 goto failed_mount_wq;
d4da6c9c 3845 }
818d276c 3846
ac27a0ec
DK		 3847 /* We have now updated the journal if required, so we can
3848 * validate the data journaling mode. */
3849 switch (test_opt(sb, DATA_FLAGS)) {
3850 case 0:
3851 /* No mode set, assume a default based on the journal
63f57933
AM		 3852 * capabilities: ORDERED_DATA if the journal can
3853 * cope, else JOURNAL_DATA
3854 */
dab291af
MC		 3855 if (jbd2_journal_check_available_features
3856 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3857 set_opt(sb, ORDERED_DATA);
ac27a0ec 3858 else
fd8c37ec 3859 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3860 break;
3861
617ba13b
MC		 3862 case EXT4_MOUNT_ORDERED_DATA:
3863 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3864 if (!jbd2_journal_check_available_features
3865 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3866 ext4_msg(sb, KERN_ERR, "Journal does not support "
3867 "requested data journaling mode");
744692dc 3868 goto failed_mount_wq;
ac27a0ec
DK		 3869 }
3870 default:
3871 break;
3872 }
b3881f74 3873 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3874
18aadd47
BJ		 3875 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3876
ce7e010a
TT		 3877 /*
3878 * The journal may have updated the bg summary counts, so we
3879 * need to update the global counters.
3880 */
57042651 3881 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3882 ext4_count_free_clusters(sb));
ce7e010a
TT		 3883 percpu_counter_set(&sbi->s_freeinodes_counter,
3884 ext4_count_free_inodes(sb));
3885 percpu_counter_set(&sbi->s_dirs_counter,
3886 ext4_count_dirs(sb));
57042651 3887 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3888
ce7e010a 3889no_journal:
952fc18e
TT		 3890 /*
3891 * Get the # of file system overhead blocks from the
3892 * superblock if present.
3893 */
3894 if (es->s_overhead_clusters)
3895 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3896 else {
3897 ret = ext4_calculate_overhead(sb);
3898 if (ret)
3899 goto failed_mount_wq;
3900 }
3901
fd89d5f2
TH		 3902 /*
3903 * The maximum number of concurrent works can be high and
3904 * concurrency isn't really necessary. Limit it to 1.
3905 */
3906 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3907 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3908 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3909 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3910 goto failed_mount_wq;
3911 }
3912
ac27a0ec 3913 /*
dab291af 3914 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3915 * so we can safely mount the rest of the filesystem now.
3916 */
3917
1d1fe1ee
DH		 3918 root = ext4_iget(sb, EXT4_ROOT_INO);
3919 if (IS_ERR(root)) {
b31e1552 3920 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3921 ret = PTR_ERR(root);
32a9bb57 3922 root = NULL;
ac27a0ec
DK		 3923 goto failed_mount4;
3924 }
3925 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3926 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3927 iput(root);
ac27a0ec
DK		 3928 goto failed_mount4;
3929 }
48fde701 3930 sb->s_root = d_make_root(root);
1d1fe1ee 3931 if (!sb->s_root) {
b31e1552 3932 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3933 ret = -ENOMEM;
3934 goto failed_mount4;
3935 }
ac27a0ec 3936
7e84b621
ES		 3937 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3938 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3939
3940 /* determine the minimum size of new large inodes, if present */
3941 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3942 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3943 EXT4_GOOD_OLD_INODE_SIZE;
3944 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3945 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3946 if (sbi->s_want_extra_isize <
3947 le16_to_cpu(es->s_want_extra_isize))
3948 sbi->s_want_extra_isize =
3949 le16_to_cpu(es->s_want_extra_isize);
3950 if (sbi->s_want_extra_isize <
3951 le16_to_cpu(es->s_min_extra_isize))
3952 sbi->s_want_extra_isize =
3953 le16_to_cpu(es->s_min_extra_isize);
3954 }
3955 }
3956 /* Check if enough inode space is available */
3957 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3958 sbi->s_inode_size) {
3959 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3960 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3961 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3962 "available");
ef7f3835
KS		 3963 }
3964
6fd058f7
TT		 3965 err = ext4_setup_system_zone(sb);
3966 if (err) {
b31e1552 3967 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3968 "zone (%d)", err);
94bf608a 3969 goto failed_mount4a;
6fd058f7
TT		 3970 }
3971
c2774d84 3972 ext4_ext_init(sb);
9d99012f 3973 err = ext4_mb_init(sb);
c2774d84 3974 if (err) {
421f91d2 3975 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3976 err);
dcf2d804 3977 goto failed_mount5;
c2774d84
AK		 3978 }
3979
bfff6873
LC		 3980 err = ext4_register_li_request(sb, first_not_zeroed);
3981 if (err)
dcf2d804 3982 goto failed_mount6;
bfff6873 3983
3197ebdb
TT		 3984 sbi->s_kobj.kset = ext4_kset;
3985 init_completion(&sbi->s_kobj_unregister);
3986 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3987 "%s", sb->s_id);
dcf2d804
TM		 3988 if (err)
3989 goto failed_mount7;
3197ebdb 3990
617ba13b
MC		 3991 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3992 ext4_orphan_cleanup(sb, es);
3993 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3994 if (needs_recovery) {
b31e1552 3995 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3996 ext4_mark_recovery_complete(sb, es);
3997 }
3998 if (EXT4_SB(sb)->s_journal) {
3999 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4000 descr = " journalled data mode";
4001 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4002 descr = " ordered data mode";
4003 else
4004 descr = " writeback data mode";
4005 } else
4006 descr = "out journal";
4007
7c319d32
AK		 4008#ifdef CONFIG_QUOTA
4009 /* Enable quota usage during mount. */
4010 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4011 !(sb->s_flags & MS_RDONLY)) {
4012 ret = ext4_enable_quotas(sb);
4013 if (ret)
4014 goto failed_mount7;
4015 }
4016#endif /* CONFIG_QUOTA */
4017
d4c402d9 4018 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4019 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4020 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4021
66e61a9e
TT		 4022 if (es->s_error_count)
4023 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4024
d4c402d9 4025 kfree(orig_data);
ac27a0ec
DK		 4026 return 0;
4027
617ba13b 4028cantfind_ext4:
ac27a0ec 4029 if (!silent)
b31e1552 4030 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4031 goto failed_mount;
4032
dcf2d804
TM		 4033failed_mount7:
4034 ext4_unregister_li_request(sb);
4035failed_mount6:
dcf2d804 4036 ext4_mb_release(sb);
94bf608a
AV		 4037failed_mount5:
4038 ext4_ext_release(sb);
dcf2d804 4039 ext4_release_system_zone(sb);
94bf608a
AV		 4040failed_mount4a:
4041 dput(sb->s_root);
32a9bb57 4042 sb->s_root = NULL;
94bf608a 4043failed_mount4:
b31e1552 4044 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 4045 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4046failed_mount_wq:
0390131b
FM		 4047 if (sbi->s_journal) {
4048 jbd2_journal_destroy(sbi->s_journal);
4049 sbi->s_journal = NULL;
4050 }
ac27a0ec 4051failed_mount3:
04496411 4052 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4053 if (sbi->s_flex_groups)
4054 ext4_kvfree(sbi->s_flex_groups);
57042651 4055 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4056 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4057 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4058 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 4059 if (sbi->s_mmp_tsk)
4060 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4061failed_mount2:
4062 for (i = 0; i < db_count; i++)
4063 brelse(sbi->s_group_desc[i]);
f18a5f21 4064 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4065failed_mount:
0441984a
DW		 4066 if (sbi->s_chksum_driver)
4067 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4068 if (sbi->s_proc) {
66acdcf4 4069 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4070 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4071 }
ac27a0ec
DK		 4072#ifdef CONFIG_QUOTA
4073 for (i = 0; i < MAXQUOTAS; i++)
4074 kfree(sbi->s_qf_names[i]);
4075#endif
617ba13b 4076 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4077 brelse(bh);
4078out_fail:
4079 sb->s_fs_info = NULL;
f6830165 4080 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4081 kfree(sbi);
dcc7dae3 4082out_free_orig:
d4c402d9 4083 kfree(orig_data);
1d1fe1ee 4084 return ret;
ac27a0ec
DK		 4085}
4086
4087/*
4088 * Setup any per-fs journal parameters now. We'll do this both on
4089 * initial mount, once the journal has been initialised but before we've
4090 * done any recovery; and again on any subsequent remount.
4091 */
617ba13b 4092static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4093{
617ba13b 4094 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4095
30773840
TT		 4096 journal->j_commit_interval = sbi->s_commit_interval;
4097 journal->j_min_batch_time = sbi->s_min_batch_time;
4098 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4099
a931da6a 4100 write_lock(&journal->j_state_lock);
ac27a0ec 4101 if (test_opt(sb, BARRIER))
dab291af 4102 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4103 else
dab291af 4104 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4105 if (test_opt(sb, DATA_ERR_ABORT))
4106 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4107 else
4108 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4109 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4110}
4111
617ba13b 4112static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4113 unsigned int journal_inum)
4114{
4115 struct inode *journal_inode;
4116 journal_t *journal;
4117
0390131b
FM		 4118 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4119
ac27a0ec
DK		 4120 /* First, test for the existence of a valid inode on disk. Bad
4121 * things happen if we iget() an unused inode, as the subsequent
4122 * iput() will try to delete it. */
4123
1d1fe1ee
DH		 4124 journal_inode = ext4_iget(sb, journal_inum);
4125 if (IS_ERR(journal_inode)) {
b31e1552 4126 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4127 return NULL;
4128 }
4129 if (!journal_inode->i_nlink) {
4130 make_bad_inode(journal_inode);
4131 iput(journal_inode);
b31e1552 4132 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4133 return NULL;
4134 }
4135
e5f8eab8 4136 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4137 journal_inode, journal_inode->i_size);
1d1fe1ee 4138 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4139 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4140 iput(journal_inode);
4141 return NULL;
4142 }
4143
dab291af 4144 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4145 if (!journal) {
b31e1552 4146 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4147 iput(journal_inode);
4148 return NULL;
4149 }
4150 journal->j_private = sb;
617ba13b 4151 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4152 return journal;
4153}
4154
617ba13b 4155static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4156 dev_t j_dev)
4157{
2b2d6d01 4158 struct buffer_head *bh;
ac27a0ec 4159 journal_t *journal;
617ba13b
MC		 4160 ext4_fsblk_t start;
4161 ext4_fsblk_t len;
ac27a0ec 4162 int hblock, blocksize;
617ba13b 4163 ext4_fsblk_t sb_block;
ac27a0ec 4164 unsigned long offset;
2b2d6d01 4165 struct ext4_super_block *es;
ac27a0ec
DK		 4166 struct block_device *bdev;
4167
0390131b
FM		 4168 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4169
b31e1552 4170 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4171 if (bdev == NULL)
4172 return NULL;
4173
ac27a0ec 4174 blocksize = sb->s_blocksize;
e1defc4f 4175 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4176 if (blocksize < hblock) {
b31e1552
ES		 4177 ext4_msg(sb, KERN_ERR,
4178 "blocksize too small for journal device");
ac27a0ec
DK		 4179 goto out_bdev;
4180 }
4181
617ba13b
MC		 4182 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4183 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4184 set_blocksize(bdev, blocksize);
4185 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4186 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4187 "external journal");
ac27a0ec
DK		 4188 goto out_bdev;
4189 }
4190
2716b802 4191 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4192 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4193 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4194 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4195 ext4_msg(sb, KERN_ERR, "external journal has "
4196 "bad superblock");
ac27a0ec
DK		 4197 brelse(bh);
4198 goto out_bdev;
4199 }
4200
617ba13b 4201 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4202 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4203 brelse(bh);
4204 goto out_bdev;
4205 }
4206
bd81d8ee 4207 len = ext4_blocks_count(es);
ac27a0ec
DK		 4208 start = sb_block + 1;
4209 brelse(bh); /* we're done with the superblock */
4210
dab291af 4211 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4212 start, len, blocksize);
4213 if (!journal) {
b31e1552 4214 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4215 goto out_bdev;
4216 }
4217 journal->j_private = sb;
4218 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4219 wait_on_buffer(journal->j_sb_buffer);
4220 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4221 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4222 goto out_journal;
4223 }
4224 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4225 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4226 "user (unsupported) - %d",
ac27a0ec
DK		 4227 be32_to_cpu(journal->j_superblock->s_nr_users));
4228 goto out_journal;
4229 }
617ba13b
MC		 4230 EXT4_SB(sb)->journal_bdev = bdev;
4231 ext4_init_journal_params(sb, journal);
ac27a0ec 4232 return journal;
0b8e58a1 4233
ac27a0ec 4234out_journal:
dab291af 4235 jbd2_journal_destroy(journal);
ac27a0ec 4236out_bdev:
617ba13b 4237 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4238 return NULL;
4239}
4240
617ba13b
MC		 4241static int ext4_load_journal(struct super_block *sb,
4242 struct ext4_super_block *es,
ac27a0ec
DK		 4243 unsigned long journal_devnum)
4244{
4245 journal_t *journal;
4246 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4247 dev_t journal_dev;
4248 int err = 0;
4249 int really_read_only;
4250
0390131b
FM		 4251 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4252
ac27a0ec
DK		 4253 if (journal_devnum &&
4254 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4255 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4256 "numbers have changed");
ac27a0ec
DK		 4257 journal_dev = new_decode_dev(journal_devnum);
4258 } else
4259 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4260
4261 really_read_only = bdev_read_only(sb->s_bdev);
4262
4263 /*
4264 * Are we loading a blank journal or performing recovery after a
4265 * crash? For recovery, we need to check in advance whether we
4266 * can get read-write access to the device.
4267 */
617ba13b 4268 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4269 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4270 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4271 "required on readonly filesystem");
ac27a0ec 4272 if (really_read_only) {
b31e1552
ES		 4273 ext4_msg(sb, KERN_ERR, "write access "
4274 "unavailable, cannot proceed");
ac27a0ec
DK		 4275 return -EROFS;
4276 }
b31e1552
ES		 4277 ext4_msg(sb, KERN_INFO, "write access will "
4278 "be enabled during recovery");
ac27a0ec
DK		 4279 }
4280 }
4281
4282 if (journal_inum && journal_dev) {
b31e1552
ES		 4283 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4284 "and inode journals!");
ac27a0ec
DK		 4285 return -EINVAL;
4286 }
4287
4288 if (journal_inum) {
617ba13b 4289 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4290 return -EINVAL;
4291 } else {
617ba13b 4292 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4293 return -EINVAL;
4294 }
4295
90576c0b 4296 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4297 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4298
617ba13b 4299 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4300 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4301 if (!err) {
4302 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4303 if (save)
4304 memcpy(save, ((char *) es) +
4305 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4306 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4307 if (save)
4308 memcpy(((char *) es) + EXT4_S_ERR_START,
4309 save, EXT4_S_ERR_LEN);
4310 kfree(save);
4311 }
ac27a0ec
DK		 4312
4313 if (err) {
b31e1552 4314 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4315 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4316 return err;
4317 }
4318
617ba13b
MC		 4319 EXT4_SB(sb)->s_journal = journal;
4320 ext4_clear_journal_err(sb, es);
ac27a0ec 4321
c41303ce 4322 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4323 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4324 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4325
4326 /* Make sure we flush the recovery flag to disk. */
e2d67052 4327 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4328 }
4329
4330 return 0;
4331}
4332
e2d67052 4333static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4334{
e2d67052 4335 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4336 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4337 int error = 0;
ac27a0ec 4338
7c2e7087 4339 if (!sbh || block_device_ejected(sb))
c4be0c1d 4340 return error;
914258bf
TT		 4341 if (buffer_write_io_error(sbh)) {
4342 /*
4343 * Oh, dear. A previous attempt to write the
4344 * superblock failed. This could happen because the
4345 * USB device was yanked out. Or it could happen to
4346 * be a transient write error and maybe the block will
4347 * be remapped. Nothing we can do but to retry the
4348 * write and hope for the best.
4349 */
b31e1552
ES		 4350 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4351 "superblock detected");
914258bf
TT		 4352 clear_buffer_write_io_error(sbh);
4353 set_buffer_uptodate(sbh);
4354 }
71290b36
TT		 4355 /*
4356 * If the file system is mounted read-only, don't update the
4357 * superblock write time. This avoids updating the superblock
4358 * write time when we are mounting the root file system
4359 * read/only but we need to replay the journal; at that point,
4360 * for people who are east of GMT and who make their clock
4361 * tick in localtime for Windows bug-for-bug compatibility,
4362 * the clock is set in the future, and this will cause e2fsck
4363 * to complain and force a full file system check.
4364 */
4365 if (!(sb->s_flags & MS_RDONLY))
4366 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4367 if (sb->s_bdev->bd_part)
4368 es->s_kbytes_written =
4369 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4370 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4371 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4372 else
4373 es->s_kbytes_written =
4374 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4375 ext4_free_blocks_count_set(es,
4376 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4377 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4378 es->s_free_inodes_count =
4379 cpu_to_le32(percpu_counter_sum_positive(
4380 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4381 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4382 ext4_superblock_csum_set(sb);
ac27a0ec 4383 mark_buffer_dirty(sbh);
914258bf 4384 if (sync) {
c4be0c1d
TS		 4385 error = sync_dirty_buffer(sbh);
4386 if (error)
4387 return error;
4388
4389 error = buffer_write_io_error(sbh);
4390 if (error) {
b31e1552
ES		 4391 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4392 "superblock");
914258bf
TT		 4393 clear_buffer_write_io_error(sbh);
4394 set_buffer_uptodate(sbh);
4395 }
4396 }
c4be0c1d 4397 return error;
ac27a0ec
DK		 4398}
4399
ac27a0ec
DK		 4400/*
4401 * Have we just finished recovery? If so, and if we are mounting (or
4402 * remounting) the filesystem readonly, then we will end up with a
4403 * consistent fs on disk. Record that fact.
4404 */
2b2d6d01
TT		 4405static void ext4_mark_recovery_complete(struct super_block *sb,
4406 struct ext4_super_block *es)
ac27a0ec 4407{
617ba13b 4408 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4409
0390131b
FM		 4410 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4411 BUG_ON(journal != NULL);
4412 return;
4413 }
dab291af 4414 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4415 if (jbd2_journal_flush(journal) < 0)
4416 goto out;
4417
617ba13b 4418 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4419 sb->s_flags & MS_RDONLY) {
617ba13b 4420 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4421 ext4_commit_super(sb, 1);
ac27a0ec 4422 }
7ffe1ea8
HK		 4423
4424out:
dab291af 4425 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4426}
4427
4428/*
4429 * If we are mounting (or read-write remounting) a filesystem whose journal
4430 * has recorded an error from a previous lifetime, move that error to the
4431 * main filesystem now.
4432 */
2b2d6d01
TT		 4433static void ext4_clear_journal_err(struct super_block *sb,
4434 struct ext4_super_block *es)
ac27a0ec
DK		 4435{
4436 journal_t *journal;
4437 int j_errno;
4438 const char *errstr;
4439
0390131b
FM		 4440 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4441
617ba13b 4442 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4443
4444 /*
4445 * Now check for any error status which may have been recorded in the
617ba13b 4446 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4447 */
4448
dab291af 4449 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4450 if (j_errno) {
4451 char nbuf[16];
4452
617ba13b 4453 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4454 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4455 "from previous mount: %s", errstr);
12062ddd 4456 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4457
617ba13b
MC		 4458 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4459 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4460 ext4_commit_super(sb, 1);
ac27a0ec 4461
dab291af 4462 jbd2_journal_clear_err(journal);
d796c52e 4463 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4464 }
4465}
4466
4467/*
4468 * Force the running and committing transactions to commit,
4469 * and wait on the commit.
4470 */
617ba13b 4471int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4472{
4473 journal_t *journal;
0390131b 4474 int ret = 0;
ac27a0ec
DK		 4475
4476 if (sb->s_flags & MS_RDONLY)
4477 return 0;
4478
617ba13b 4479 journal = EXT4_SB(sb)->s_journal;
8e8ad8a5 4480 if (journal)
0390131b 4481 ret = ext4_journal_force_commit(journal);
0390131b 4482
ac27a0ec
DK		 4483 return ret;
4484}
4485
617ba13b 4486static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4487{
14ce0cb4 4488 int ret = 0;
9eddacf9 4489 tid_t target;
8d5d02e6 4490 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4491
9bffad1e 4492 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4493 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4494 /*
4495 * Writeback quota in non-journalled quota case - journalled quota has
4496 * no dirty dquots
4497 */
4498 dquot_writeback_dquots(sb, -1);
8d5d02e6 4499 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4500 if (wait)
8d5d02e6 4501 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4502 }
14ce0cb4 4503 return ret;
ac27a0ec
DK		 4504}
4505
4506/*
4507 * LVM calls this function before a (read-only) snapshot is created. This
4508 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4509 *
4510 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4511 * state independently. It relies on upper layer to stop all data & metadata
4512 * modifications.
ac27a0ec 4513 */
c4be0c1d 4514static int ext4_freeze(struct super_block *sb)
ac27a0ec 4515{
c4be0c1d
TS		 4516 int error = 0;
4517 journal_t *journal;
ac27a0ec 4518
9ca92389
TT		 4519 if (sb->s_flags & MS_RDONLY)
4520 return 0;
ac27a0ec 4521
9ca92389 4522 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4523
9ca92389
TT		 4524 /* Now we set up the journal barrier. */
4525 jbd2_journal_lock_updates(journal);
ac27a0ec 4526
9ca92389
TT		 4527 /*
4528 * Don't clear the needs_recovery flag if we failed to flush
4529 * the journal.
4530 */
4531 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4532 if (error < 0)
4533 goto out;
9ca92389
TT		 4534
4535 /* Journal blocked and flushed, clear needs_recovery flag. */
4536 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4537 error = ext4_commit_super(sb, 1);
6b0310fb 4538out:
8e8ad8a5 4539 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4540 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4541 return error;
ac27a0ec
DK		 4542}
4543
4544/*
4545 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4546 * flag here, even though the filesystem is not technically dirty yet.
4547 */
c4be0c1d 4548static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4549{
9ca92389
TT		 4550 if (sb->s_flags & MS_RDONLY)
4551 return 0;
4552
9ca92389
TT		 4553 /* Reset the needs_recovery flag before the fs is unlocked. */
4554 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4555 ext4_commit_super(sb, 1);
c4be0c1d 4556 return 0;
ac27a0ec
DK		 4557}
4558
673c6100
TT		 4559/*
4560 * Structure to save mount options for ext4_remount's benefit
4561 */
4562struct ext4_mount_options {
4563 unsigned long s_mount_opt;
a2595b8a 4564 unsigned long s_mount_opt2;
08cefc7a
EB		 4565 kuid_t s_resuid;
4566 kgid_t s_resgid;
673c6100
TT		 4567 unsigned long s_commit_interval;
4568 u32 s_min_batch_time, s_max_batch_time;
4569#ifdef CONFIG_QUOTA
4570 int s_jquota_fmt;
4571 char *s_qf_names[MAXQUOTAS];
4572#endif
4573};
4574
2b2d6d01 4575static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4576{
2b2d6d01 4577 struct ext4_super_block *es;
617ba13b 4578 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4579 unsigned long old_sb_flags;
617ba13b 4580 struct ext4_mount_options old_opts;
c79d967d 4581 int enable_quota = 0;
8a266467 4582 ext4_group_t g;
b3881f74 4583 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4584 int err = 0;
ac27a0ec
DK		 4585#ifdef CONFIG_QUOTA
4586 int i;
4587#endif
d4c402d9 4588 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4589
4590 /* Store the original options */
4591 old_sb_flags = sb->s_flags;
4592 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4593 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4594 old_opts.s_resuid = sbi->s_resuid;
4595 old_opts.s_resgid = sbi->s_resgid;
4596 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4597 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4598 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4599#ifdef CONFIG_QUOTA
4600 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4601 for (i = 0; i < MAXQUOTAS; i++)
4602 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4603#endif
b3881f74
TT		 4604 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4605 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4606
4607 /*
4608 * Allow the "check" option to be passed as a remount option.
4609 */
661aa520 4610 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4611 err = -EINVAL;
4612 goto restore_opts;
4613 }
4614
4ab2f15b 4615 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4616 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4617
4618 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4619 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4620
4621 es = sbi->s_es;
4622
b3881f74 4623 if (sbi->s_journal) {
0390131b 4624 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4625 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4626 }
ac27a0ec 4627
661aa520 4628 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4629 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4630 err = -EROFS;
4631 goto restore_opts;
4632 }
4633
4634 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4635 err = dquot_suspend(sb, -1);
4636 if (err < 0)
c79d967d 4637 goto restore_opts;
c79d967d 4638
ac27a0ec
DK		 4639 /*
4640 * First of all, the unconditional stuff we have to do
4641 * to disable replay of the journal when we next remount
4642 */
4643 sb->s_flags |= MS_RDONLY;
4644
4645 /*
4646 * OK, test if we are remounting a valid rw partition
4647 * readonly, and if so set the rdonly flag and then
4648 * mark the partition as valid again.
4649 */
617ba13b
MC		 4650 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4651 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4652 es->s_state = cpu_to_le16(sbi->s_mount_state);
4653
a63c9eb2 4654 if (sbi->s_journal)
0390131b 4655 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4656 } else {
a13fb1a4
ES		 4657 /* Make sure we can mount this feature set readwrite */
4658 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4659 err = -EROFS;
4660 goto restore_opts;
4661 }
8a266467
TT		 4662 /*
4663 * Make sure the group descriptor checksums
0b8e58a1 4664 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4665 */
4666 for (g = 0; g < sbi->s_groups_count; g++) {
4667 struct ext4_group_desc *gdp =
4668 ext4_get_group_desc(sb, g, NULL);
4669
feb0ab32 4670 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4671 ext4_msg(sb, KERN_ERR,
4672 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4673 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4674 le16_to_cpu(gdp->bg_checksum));
4675 err = -EINVAL;
4676 goto restore_opts;
4677 }
4678 }
4679
ead6596b
ES		 4680 /*
4681 * If we have an unprocessed orphan list hanging
4682 * around from a previously readonly bdev mount,
4683 * require a full umount/remount for now.
4684 */
4685 if (es->s_last_orphan) {
b31e1552 4686 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4687 "remount RDWR because of unprocessed "
4688 "orphan inode list. Please "
b31e1552 4689 "umount/remount instead");
ead6596b
ES		 4690 err = -EINVAL;
4691 goto restore_opts;
4692 }
4693
ac27a0ec
DK		 4694 /*
4695 * Mounting a RDONLY partition read-write, so reread
4696 * and store the current valid flag. (It may have
4697 * been changed by e2fsck since we originally mounted
4698 * the partition.)
4699 */
0390131b
FM		 4700 if (sbi->s_journal)
4701 ext4_clear_journal_err(sb, es);
ac27a0ec 4702 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4703 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4704 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4705 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4706 EXT4_FEATURE_INCOMPAT_MMP))
4707 if (ext4_multi_mount_protect(sb,
4708 le64_to_cpu(es->s_mmp_block))) {
4709 err = -EROFS;
4710 goto restore_opts;
4711 }
c79d967d 4712 enable_quota = 1;
ac27a0ec
DK		 4713 }
4714 }
bfff6873
LC		 4715
4716 /*
4717 * Reinitialize lazy itable initialization thread based on
4718 * current settings
4719 */
4720 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4721 ext4_unregister_li_request(sb);
4722 else {
4723 ext4_group_t first_not_zeroed;
4724 first_not_zeroed = ext4_has_uninit_itable(sb);
4725 ext4_register_li_request(sb, first_not_zeroed);
4726 }
4727
6fd058f7 4728 ext4_setup_system_zone(sb);
0390131b 4729 if (sbi->s_journal == NULL)
e2d67052 4730 ext4_commit_super(sb, 1);
0390131b 4731
ac27a0ec
DK		 4732#ifdef CONFIG_QUOTA
4733 /* Release old quota file names */
4734 for (i = 0; i < MAXQUOTAS; i++)
4735 if (old_opts.s_qf_names[i] &&
4736 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4737 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4738 if (enable_quota) {
4739 if (sb_any_quota_suspended(sb))
4740 dquot_resume(sb, -1);
4741 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4742 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4743 err = ext4_enable_quotas(sb);
07724f98 4744 if (err)
7c319d32 4745 goto restore_opts;
7c319d32
AK		 4746 }
4747 }
ac27a0ec 4748#endif
d4c402d9
CW		 4749
4750 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4751 kfree(orig_data);
ac27a0ec 4752 return 0;
0b8e58a1 4753
ac27a0ec
DK		 4754restore_opts:
4755 sb->s_flags = old_sb_flags;
4756 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4757 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4758 sbi->s_resuid = old_opts.s_resuid;
4759 sbi->s_resgid = old_opts.s_resgid;
4760 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4761 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4762 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4763#ifdef CONFIG_QUOTA
4764 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4765 for (i = 0; i < MAXQUOTAS; i++) {
4766 if (sbi->s_qf_names[i] &&
4767 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4768 kfree(sbi->s_qf_names[i]);
4769 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4770 }
4771#endif
d4c402d9 4772 kfree(orig_data);
ac27a0ec
DK		 4773 return err;
4774}
4775
2b2d6d01 4776static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4777{
4778 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4779 struct ext4_sb_info *sbi = EXT4_SB(sb);
4780 struct ext4_super_block *es = sbi->s_es;
952fc18e 4781 ext4_fsblk_t overhead = 0;
960cc398 4782 u64 fsid;
d02a9391 4783 s64 bfree;
ac27a0ec 4784
952fc18e
TT		 4785 if (!test_opt(sb, MINIX_DF))
4786 overhead = sbi->s_overhead;
ac27a0ec 4787
617ba13b 4788 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4789 buf->f_bsize = sb->s_blocksize;
b72f78cb 4790 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4791 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4792 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4793 /* prevent underflow in case that few free space is available */
57042651 4794 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4795 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4796 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4797 buf->f_bavail = 0;
4798 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4799 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4800 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4801 fsid = le64_to_cpup((void *)es->s_uuid) ^
4802 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4803 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4804 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4805
ac27a0ec
DK		 4806 return 0;
4807}
4808
0b8e58a1
AD		 4809/* Helper function for writing quotas on sync - we need to start transaction
4810 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4811 * Process 1 Process 2
617ba13b 4812 * ext4_create() quota_sync()
a269eb18 4813 * jbd2_journal_start() write_dquot()
871a2931 4814 * dquot_initialize() down(dqio_mutex)
dab291af 4815 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4816 *
4817 */
4818
4819#ifdef CONFIG_QUOTA
4820
4821static inline struct inode *dquot_to_inode(struct dquot *dquot)
4822{
4c376dca 4823 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4824}
4825
617ba13b 4826static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4827{
4828 int ret, err;
4829 handle_t *handle;
4830 struct inode *inode;
4831
4832 inode = dquot_to_inode(dquot);
617ba13b 4833 handle = ext4_journal_start(inode,
0b8e58a1 4834 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4835 if (IS_ERR(handle))
4836 return PTR_ERR(handle);
4837 ret = dquot_commit(dquot);
617ba13b 4838 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4839 if (!ret)
4840 ret = err;
4841 return ret;
4842}
4843
617ba13b 4844static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4845{
4846 int ret, err;
4847 handle_t *handle;
4848
617ba13b 4849 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4850 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4851 if (IS_ERR(handle))
4852 return PTR_ERR(handle);
4853 ret = dquot_acquire(dquot);
617ba13b 4854 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4855 if (!ret)
4856 ret = err;
4857 return ret;
4858}
4859
617ba13b 4860static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4861{
4862 int ret, err;
4863 handle_t *handle;
4864
617ba13b 4865 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4866 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4867 if (IS_ERR(handle)) {
4868 /* Release dquot anyway to avoid endless cycle in dqput() */
4869 dquot_release(dquot);
ac27a0ec 4870 return PTR_ERR(handle);
9c3013e9 4871 }
ac27a0ec 4872 ret = dquot_release(dquot);
617ba13b 4873 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4874 if (!ret)
4875 ret = err;
4876 return ret;
4877}
4878
617ba13b 4879static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4880{
2c8be6b2 4881 /* Are we journaling quotas? */
617ba13b
MC		 4882 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4883 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4884 dquot_mark_dquot_dirty(dquot);
617ba13b 4885 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4886 } else {
4887 return dquot_mark_dquot_dirty(dquot);
4888 }
4889}
4890
617ba13b 4891static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4892{
4893 int ret, err;
4894 handle_t *handle;
4895
4896 /* Data block + inode block */
617ba13b 4897 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4898 if (IS_ERR(handle))
4899 return PTR_ERR(handle);
4900 ret = dquot_commit_info(sb, type);
617ba13b 4901 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4902 if (!ret)
4903 ret = err;
4904 return ret;
4905}
4906
4907/*
4908 * Turn on quotas during mount time - we need to find
4909 * the quota file and such...
4910 */
617ba13b 4911static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4912{
287a8095
CH		 4913 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4914 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4915}
4916
4917/*
4918 * Standard function to be called on quota_on
4919 */
617ba13b 4920static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4921 struct path *path)
ac27a0ec
DK		 4922{
4923 int err;
ac27a0ec
DK		 4924
4925 if (!test_opt(sb, QUOTA))
4926 return -EINVAL;
0623543b 4927
ac27a0ec 4928 /* Quotafile not on the same filesystem? */
d8c9584e 4929 if (path->dentry->d_sb != sb)
ac27a0ec 4930 return -EXDEV;
0623543b
JK		 4931 /* Journaling quota? */
4932 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4933 /* Quotafile not in fs root? */
f00c9e44 4934 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4935 ext4_msg(sb, KERN_WARNING,
4936 "Quota file not on filesystem root. "
4937 "Journaled quota will not work");
2b2d6d01 4938 }
0623543b
JK		 4939
4940 /*
4941 * When we journal data on quota file, we have to flush journal to see
4942 * all updates to the file when we bypass pagecache...
4943 */
0390131b 4944 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4945 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4946 /*
4947 * We don't need to lock updates but journal_flush() could
4948 * otherwise be livelocked...
4949 */
4950 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4951 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4952 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4953 if (err)
7ffe1ea8 4954 return err;
0623543b
JK		 4955 }
4956
f00c9e44 4957 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4958}
4959
7c319d32
AK		 4960static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4961 unsigned int flags)
4962{
4963 int err;
4964 struct inode *qf_inode;
4965 unsigned long qf_inums[MAXQUOTAS] = {
4966 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4967 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4968 };
4969
4970 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4971
4972 if (!qf_inums[type])
4973 return -EPERM;
4974
4975 qf_inode = ext4_iget(sb, qf_inums[type]);
4976 if (IS_ERR(qf_inode)) {
4977 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4978 return PTR_ERR(qf_inode);
4979 }
4980
4981 err = dquot_enable(qf_inode, type, format_id, flags);
4982 iput(qf_inode);
4983
4984 return err;
4985}
4986
4987/* Enable usage tracking for all quota types. */
4988static int ext4_enable_quotas(struct super_block *sb)
4989{
4990 int type, err = 0;
4991 unsigned long qf_inums[MAXQUOTAS] = {
4992 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4993 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4994 };
4995
4996 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4997 for (type = 0; type < MAXQUOTAS; type++) {
4998 if (qf_inums[type]) {
4999 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5000 DQUOT_USAGE_ENABLED);
5001 if (err) {
5002 ext4_warning(sb,
5003 "Failed to enable quota (type=%d) "
5004 "tracking. Please run e2fsck to fix.",
5005 type);
5006 return err;
5007 }
5008 }
5009 }
5010 return 0;
5011}
5012
5013/*
5014 * quota_on function that is used when QUOTA feature is set.
5015 */
5016static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5017 int format_id)
5018{
5019 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5020 return -EINVAL;
5021
5022 /*
5023 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5024 */
5025 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5026}
5027
ca0e05e4
DM		 5028static int ext4_quota_off(struct super_block *sb, int type)
5029{
21f97697
JK		 5030 struct inode *inode = sb_dqopt(sb)->files[type];
5031 handle_t *handle;
5032
87009d86
DM		 5033 /* Force all delayed allocation blocks to be allocated.
5034 * Caller already holds s_umount sem */
5035 if (test_opt(sb, DELALLOC))
ca0e05e4 5036 sync_filesystem(sb);
ca0e05e4 5037
0b268590
AG		 5038 if (!inode)
5039 goto out;
5040
21f97697
JK		 5041 /* Update modification times of quota files when userspace can
5042 * start looking at them */
5043 handle = ext4_journal_start(inode, 1);
5044 if (IS_ERR(handle))
5045 goto out;
5046 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5047 ext4_mark_inode_dirty(handle, inode);
5048 ext4_journal_stop(handle);
5049
5050out:
ca0e05e4
DM		 5051 return dquot_quota_off(sb, type);
5052}
5053
7c319d32
AK		 5054/*
5055 * quota_off function that is used when QUOTA feature is set.
5056 */
5057static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5058{
5059 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5060 return -EINVAL;
5061
5062 /* Disable only the limits. */
5063 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5064}
5065
ac27a0ec
DK		 5066/* Read data from quotafile - avoid pagecache and such because we cannot afford
5067 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5068 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5069 * we don't have to be afraid of races */
617ba13b 5070static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5071 size_t len, loff_t off)
5072{
5073 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5074 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5075 int err = 0;
5076 int offset = off & (sb->s_blocksize - 1);
5077 int tocopy;
5078 size_t toread;
5079 struct buffer_head *bh;
5080 loff_t i_size = i_size_read(inode);
5081
5082 if (off > i_size)
5083 return 0;
5084 if (off+len > i_size)
5085 len = i_size-off;
5086 toread = len;
5087 while (toread > 0) {
5088 tocopy = sb->s_blocksize - offset < toread ?
5089 sb->s_blocksize - offset : toread;
617ba13b 5090 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5091 if (err)
5092 return err;
5093 if (!bh) /* A hole? */
5094 memset(data, 0, tocopy);
5095 else
5096 memcpy(data, bh->b_data+offset, tocopy);
5097 brelse(bh);
5098 offset = 0;
5099 toread -= tocopy;
5100 data += tocopy;
5101 blk++;
5102 }
5103 return len;
5104}
5105
5106/* Write to quotafile (we know the transaction is already started and has
5107 * enough credits) */
617ba13b 5108static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5109 const char *data, size_t len, loff_t off)
5110{
5111 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5112 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5113 int err = 0;
5114 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5115 struct buffer_head *bh;
5116 handle_t *handle = journal_current_handle();
5117
0390131b 5118 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5119 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5120 " cancelled because transaction is not started",
9c3013e9
JK		 5121 (unsigned long long)off, (unsigned long long)len);
5122 return -EIO;
5123 }
67eeb568
DM		 5124 /*
5125 * Since we account only one data block in transaction credits,
5126 * then it is impossible to cross a block boundary.
5127 */
5128 if (sb->s_blocksize - offset < len) {
5129 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5130 " cancelled because not block aligned",
5131 (unsigned long long)off, (unsigned long long)len);
5132 return -EIO;
5133 }
5134
67eeb568
DM		 5135 bh = ext4_bread(handle, inode, blk, 1, &err);
5136 if (!bh)
5137 goto out;
62d2b5f2
JK		 5138 err = ext4_journal_get_write_access(handle, bh);
5139 if (err) {
5140 brelse(bh);
5141 goto out;
ac27a0ec 5142 }
67eeb568
DM		 5143 lock_buffer(bh);
5144 memcpy(bh->b_data+offset, data, len);
5145 flush_dcache_page(bh->b_page);
5146 unlock_buffer(bh);
62d2b5f2 5147 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5148 brelse(bh);
ac27a0ec 5149out:
0b7f7cef 5150 if (err)
ac27a0ec 5151 return err;
67eeb568
DM		 5152 if (inode->i_size < off + len) {
5153 i_size_write(inode, off + len);
617ba13b 5154 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5155 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5156 }
67eeb568 5157 return len;
ac27a0ec
DK		 5158}
5159
5160#endif
5161
152a0836
AV		 5162static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5163 const char *dev_name, void *data)
ac27a0ec 5164{
152a0836 5165 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5166}
5167
37f328eb 5168#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5169static inline void register_as_ext2(void)
5170{
5171 int err = register_filesystem(&ext2_fs_type);
5172 if (err)
5173 printk(KERN_WARNING
5174 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5175}
5176
5177static inline void unregister_as_ext2(void)
5178{
5179 unregister_filesystem(&ext2_fs_type);
5180}
2035e776
TT		 5181
5182static inline int ext2_feature_set_ok(struct super_block *sb)
5183{
5184 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5185 return 0;
5186 if (sb->s_flags & MS_RDONLY)
5187 return 1;
5188 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5189 return 0;
5190 return 1;
5191}
51b7e3c9 5192MODULE_ALIAS("ext2");
24b58424
TT		 5193#else
5194static inline void register_as_ext2(void) { }
5195static inline void unregister_as_ext2(void) { }
2035e776 5196static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5197#endif
5198
37f328eb 5199#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5200static inline void register_as_ext3(void)
5201{
5202 int err = register_filesystem(&ext3_fs_type);
5203 if (err)
5204 printk(KERN_WARNING
5205 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5206}
5207
5208static inline void unregister_as_ext3(void)
5209{
5210 unregister_filesystem(&ext3_fs_type);
5211}
2035e776
TT		 5212
5213static inline int ext3_feature_set_ok(struct super_block *sb)
5214{
5215 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5216 return 0;
5217 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5218 return 0;
5219 if (sb->s_flags & MS_RDONLY)
5220 return 1;
5221 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5222 return 0;
5223 return 1;
5224}
51b7e3c9 5225MODULE_ALIAS("ext3");
24b58424
TT		 5226#else
5227static inline void register_as_ext3(void) { }
5228static inline void unregister_as_ext3(void) { }
2035e776 5229static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5230#endif
5231
03010a33
TT		 5232static struct file_system_type ext4_fs_type = {
5233 .owner = THIS_MODULE,
5234 .name = "ext4",
152a0836 5235 .mount = ext4_mount,
03010a33
TT		 5236 .kill_sb = kill_block_super,
5237 .fs_flags = FS_REQUIRES_DEV,
5238};
5239
8f021222 5240static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5241{
5242 struct ext4_features *ef;
5243 int ret = -ENOMEM;
5244
5245 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5246 if (!ef)
5247 goto out;
5248
5249 ef->f_kobj.kset = ext4_kset;
5250 init_completion(&ef->f_kobj_unregister);
5251 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5252 "features");
5253 if (ret) {
5254 kfree(ef);
5255 goto out;
5256 }
5257
5258 ext4_feat = ef;
5259 ret = 0;
5260out:
5261 return ret;
5262}
5263
8f021222
LC		 5264static void ext4_exit_feat_adverts(void)
5265{
5266 kobject_put(&ext4_feat->f_kobj);
5267 wait_for_completion(&ext4_feat->f_kobj_unregister);
5268 kfree(ext4_feat);
5269}
5270
e9e3bcec
ES		 5271/* Shared across all ext4 file systems */
5272wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5273struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5274
5dabfc78 5275static int __init ext4_init_fs(void)
ac27a0ec 5276{
e9e3bcec 5277 int i, err;
c9de560d 5278
07c0c5d8
AV		 5279 ext4_li_info = NULL;
5280 mutex_init(&ext4_li_mtx);
5281
12e9b892 5282 ext4_check_flag_values();
e9e3bcec
ES		 5283
5284 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5285 mutex_init(&ext4__aio_mutex[i]);
5286 init_waitqueue_head(&ext4__ioend_wq[i]);
5287 }
5288
5dabfc78 5289 err = ext4_init_pageio();
6fd058f7
TT		 5290 if (err)
5291 return err;
5dabfc78 5292 err = ext4_init_system_zone();
bd2d0210 5293 if (err)
d44651d0 5294 goto out6;
3197ebdb 5295 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5296 if (!ext4_kset) {
5297 err = -ENOMEM;
dd68314c 5298 goto out5;
0e376b1e 5299 }
d44651d0 5300 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5301
5302 err = ext4_init_feat_adverts();
dd68314c
TT		 5303 if (err)
5304 goto out4;
857ac889 5305
5dabfc78 5306 err = ext4_init_mballoc();
ac27a0ec 5307 if (err)
6fd058f7 5308 goto out3;
c9de560d 5309
5dabfc78 5310 err = ext4_init_xattr();
c9de560d
AT		 5311 if (err)
5312 goto out2;
ac27a0ec
DK		 5313 err = init_inodecache();
5314 if (err)
5315 goto out1;
24b58424 5316 register_as_ext3();
2035e776 5317 register_as_ext2();
03010a33 5318 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5319 if (err)
5320 goto out;
bfff6873 5321
ac27a0ec
DK		 5322 return 0;
5323out:
24b58424
TT		 5324 unregister_as_ext2();
5325 unregister_as_ext3();
ac27a0ec
DK		 5326 destroy_inodecache();
5327out1:
5dabfc78 5328 ext4_exit_xattr();
c9de560d 5329out2:
5dabfc78 5330 ext4_exit_mballoc();
6fd058f7 5331out3:
8f021222 5332 ext4_exit_feat_adverts();
dd68314c 5333out4:
d44651d0
FJ		 5334 if (ext4_proc_root)
5335 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5336 kset_unregister(ext4_kset);
d44651d0 5337out5:
5dabfc78 5338 ext4_exit_system_zone();
d44651d0 5339out6:
5dabfc78 5340 ext4_exit_pageio();
ac27a0ec
DK		 5341 return err;
5342}
5343
5dabfc78 5344static void __exit ext4_exit_fs(void)
ac27a0ec 5345{
bfff6873 5346 ext4_destroy_lazyinit_thread();
24b58424
TT		 5347 unregister_as_ext2();
5348 unregister_as_ext3();
03010a33 5349 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5350 destroy_inodecache();
5dabfc78
TT		 5351 ext4_exit_xattr();
5352 ext4_exit_mballoc();
8f021222 5353 ext4_exit_feat_adverts();
9f6200bb 5354 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5355 kset_unregister(ext4_kset);
5dabfc78
TT		 5356 ext4_exit_system_zone();
5357 ext4_exit_pageio();
ac27a0ec
DK		 5358}
5359
5360MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5361MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5362MODULE_LICENSE("GPL");
5dabfc78
TT		 5363module_init(ext4_init_fs)
5364module_exit(ext4_exit_fs)
Ubuntu Artful kernel mirror