projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: disable synchronous transaction batching if max_batch_time==0
[GitHub/LineageOS/android_kernel_motorola_exynos9610.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
4a092d73 48#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9c191f70 62static int ext4_mballoc_ready;
9f6200bb 63
617ba13b 64static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 65 unsigned long journal_devnum);
2adf6da8 66static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 67static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 68static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
617ba13b 72static int ext4_sync_fs(struct super_block *sb, int wait);
06a407f1 73static int ext4_sync_fs_nojournal(struct super_block *sb, int wait);
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);
27dd4385 86static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
ac27a0ec 87
2035e776
TT		 88#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89static struct file_system_type ext2_fs_type = {
90 .owner = THIS_MODULE,
91 .name = "ext2",
92 .mount = ext4_mount,
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
95};
7f78e035 96MODULE_ALIAS_FS("ext2");
fa7614dd 97MODULE_ALIAS("ext2");
2035e776
TT		 98#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
99#else
100#define IS_EXT2_SB(sb) (0)
101#endif
102
103
ba69f9ab
JK		 104#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
105static struct file_system_type ext3_fs_type = {
106 .owner = THIS_MODULE,
107 .name = "ext3",
152a0836 108 .mount = ext4_mount,
ba69f9ab
JK		 109 .kill_sb = kill_block_super,
110 .fs_flags = FS_REQUIRES_DEV,
111};
7f78e035 112MODULE_ALIAS_FS("ext3");
fa7614dd 113MODULE_ALIAS("ext3");
ba69f9ab
JK		 114#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
115#else
116#define IS_EXT3_SB(sb) (0)
117#endif
bd81d8ee 118
d25425f8
DW		 119static int ext4_verify_csum_type(struct super_block *sb,
120 struct ext4_super_block *es)
121{
122 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
123 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
124 return 1;
125
126 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
127}
128
a9c47317
DW		 129static __le32 ext4_superblock_csum(struct super_block *sb,
130 struct ext4_super_block *es)
131{
132 struct ext4_sb_info *sbi = EXT4_SB(sb);
133 int offset = offsetof(struct ext4_super_block, s_checksum);
134 __u32 csum;
135
136 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
137
138 return cpu_to_le32(csum);
139}
140
c197855e
SH		 141static int ext4_superblock_csum_verify(struct super_block *sb,
142 struct ext4_super_block *es)
a9c47317
DW		 143{
144 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
145 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
146 return 1;
147
148 return es->s_checksum == ext4_superblock_csum(sb, es);
149}
150
06db49e6 151void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 152{
06db49e6
TT		 153 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
154
a9c47317
DW		 155 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
156 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
157 return;
158
159 es->s_checksum = ext4_superblock_csum(sb, es);
160}
161
9933fc0a
TT		 162void *ext4_kvmalloc(size_t size, gfp_t flags)
163{
164 void *ret;
165
8be04b93 166 ret = kmalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 167 if (!ret)
168 ret = __vmalloc(size, flags, PAGE_KERNEL);
169 return ret;
170}
171
172void *ext4_kvzalloc(size_t size, gfp_t flags)
173{
174 void *ret;
175
8be04b93 176 ret = kzalloc(size, flags | __GFP_NOWARN);
9933fc0a
TT		 177 if (!ret)
178 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
179 return ret;
180}
181
182void ext4_kvfree(void *ptr)
183{
184 if (is_vmalloc_addr(ptr))
185 vfree(ptr);
186 else
187 kfree(ptr);
188
189}
190
8fadc143
AR		 191ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
192 struct ext4_group_desc *bg)
bd81d8ee 193{
3a14589c 194 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 196 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 197}
198
8fadc143
AR		 199ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
200 struct ext4_group_desc *bg)
bd81d8ee 201{
5272f837 202 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 203 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 204 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 205}
206
8fadc143
AR		 207ext4_fsblk_t ext4_inode_table(struct super_block *sb,
208 struct ext4_group_desc *bg)
bd81d8ee 209{
5272f837 210 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 211 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 212 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 213}
214
021b65bb
TT		 215__u32 ext4_free_group_clusters(struct super_block *sb,
216 struct ext4_group_desc *bg)
560671a0
AK		 217{
218 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
219 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 220 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 221}
222
223__u32 ext4_free_inodes_count(struct super_block *sb,
224 struct ext4_group_desc *bg)
225{
226 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
227 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 228 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 229}
230
231__u32 ext4_used_dirs_count(struct super_block *sb,
232 struct ext4_group_desc *bg)
233{
234 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
235 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 236 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 237}
238
239__u32 ext4_itable_unused_count(struct super_block *sb,
240 struct ext4_group_desc *bg)
241{
242 return le16_to_cpu(bg->bg_itable_unused_lo) |
243 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 244 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 245}
246
8fadc143
AR		 247void ext4_block_bitmap_set(struct super_block *sb,
248 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 249{
3a14589c 250 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 253}
254
8fadc143
AR		 255void ext4_inode_bitmap_set(struct super_block *sb,
256 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 257{
5272f837 258 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 259 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
260 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 261}
262
8fadc143
AR		 263void ext4_inode_table_set(struct super_block *sb,
264 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 265{
5272f837 266 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 267 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
268 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 269}
270
021b65bb
TT		 271void ext4_free_group_clusters_set(struct super_block *sb,
272 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 273{
274 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
275 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
276 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
277}
278
279void ext4_free_inodes_set(struct super_block *sb,
280 struct ext4_group_desc *bg, __u32 count)
281{
282 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
283 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
284 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
285}
286
287void ext4_used_dirs_set(struct super_block *sb,
288 struct ext4_group_desc *bg, __u32 count)
289{
290 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
291 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
292 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
293}
294
295void ext4_itable_unused_set(struct super_block *sb,
296 struct ext4_group_desc *bg, __u32 count)
297{
298 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
299 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
300 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
301}
302
d3d1faf6 303
1c13d5c0
TT		 304static void __save_error_info(struct super_block *sb, const char *func,
305 unsigned int line)
306{
307 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
308
309 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
310 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
311 es->s_last_error_time = cpu_to_le32(get_seconds());
312 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
313 es->s_last_error_line = cpu_to_le32(line);
314 if (!es->s_first_error_time) {
315 es->s_first_error_time = es->s_last_error_time;
316 strncpy(es->s_first_error_func, func,
317 sizeof(es->s_first_error_func));
318 es->s_first_error_line = cpu_to_le32(line);
319 es->s_first_error_ino = es->s_last_error_ino;
320 es->s_first_error_block = es->s_last_error_block;
321 }
66e61a9e
TT		 322 /*
323 * Start the daily error reporting function if it hasn't been
324 * started already
325 */
326 if (!es->s_error_count)
327 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 328 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 329}
330
331static void save_error_info(struct super_block *sb, const char *func,
332 unsigned int line)
333{
334 __save_error_info(sb, func, line);
335 ext4_commit_super(sb, 1);
336}
337
7c2e7087
TT		 338/*
339 * The del_gendisk() function uninitializes the disk-specific data
340 * structures, including the bdi structure, without telling anyone
341 * else. Once this happens, any attempt to call mark_buffer_dirty()
342 * (for example, by ext4_commit_super), will cause a kernel OOPS.
343 * This is a kludge to prevent these oops until we can put in a proper
344 * hook in del_gendisk() to inform the VFS and file system layers.
345 */
346static int block_device_ejected(struct super_block *sb)
347{
348 struct inode *bd_inode = sb->s_bdev->bd_inode;
349 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
350
351 return bdi->dev == NULL;
352}
353
18aadd47
BJ		 354static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
355{
356 struct super_block *sb = journal->j_private;
357 struct ext4_sb_info *sbi = EXT4_SB(sb);
358 int error = is_journal_aborted(journal);
5d3ee208 359 struct ext4_journal_cb_entry *jce;
18aadd47 360
5d3ee208 361 BUG_ON(txn->t_state == T_FINISHED);
18aadd47 362 spin_lock(&sbi->s_md_lock);
5d3ee208
DM		 363 while (!list_empty(&txn->t_private_list)) {
364 jce = list_entry(txn->t_private_list.next,
365 struct ext4_journal_cb_entry, jce_list);
18aadd47
BJ		 366 list_del_init(&jce->jce_list);
367 spin_unlock(&sbi->s_md_lock);
368 jce->jce_func(sb, jce, error);
369 spin_lock(&sbi->s_md_lock);
370 }
371 spin_unlock(&sbi->s_md_lock);
372}
1c13d5c0 373
ac27a0ec
DK		 374/* Deal with the reporting of failure conditions on a filesystem such as
375 * inconsistencies detected or read IO failures.
376 *
377 * On ext2, we can store the error state of the filesystem in the
617ba13b 378 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 379 * write ordering constraints on the superblock which prevent us from
380 * writing it out straight away; and given that the journal is about to
381 * be aborted, we can't rely on the current, or future, transactions to
382 * write out the superblock safely.
383 *
dab291af 384 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 385 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 386 * that error until we've noted it down and cleared it.
387 */
388
617ba13b 389static void ext4_handle_error(struct super_block *sb)
ac27a0ec 390{
ac27a0ec
DK		 391 if (sb->s_flags & MS_RDONLY)
392 return;
393
2b2d6d01 394 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 395 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 396
4ab2f15b 397 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 398 if (journal)
dab291af 399 jbd2_journal_abort(journal, -EIO);
ac27a0ec 400 }
2b2d6d01 401 if (test_opt(sb, ERRORS_RO)) {
b31e1552 402 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
4418e141
DM		 403 /*
404 * Make sure updated value of ->s_mount_flags will be visible
405 * before ->s_flags update
406 */
407 smp_wmb();
ac27a0ec
DK		 408 sb->s_flags |= MS_RDONLY;
409 }
ac27a0ec 410 if (test_opt(sb, ERRORS_PANIC))
617ba13b 411 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 412 sb->s_id);
413}
414
efbed4dc
TT		 415#define ext4_error_ratelimit(sb) \
416 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
417 "EXT4-fs error")
418
12062ddd 419void __ext4_error(struct super_block *sb, const char *function,
c398eda0 420 unsigned int line, const char *fmt, ...)
ac27a0ec 421{
0ff2ea7d 422 struct va_format vaf;
ac27a0ec
DK		 423 va_list args;
424
efbed4dc
TT		 425 if (ext4_error_ratelimit(sb)) {
426 va_start(args, fmt);
427 vaf.fmt = fmt;
428 vaf.va = &args;
429 printk(KERN_CRIT
430 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
431 sb->s_id, function, line, current->comm, &vaf);
432 va_end(args);
433 }
f3fc0210 434 save_error_info(sb, function, line);
617ba13b 435 ext4_handle_error(sb);
ac27a0ec
DK		 436}
437
e7c96e8e
JP		 438void __ext4_error_inode(struct inode *inode, const char *function,
439 unsigned int line, ext4_fsblk_t block,
440 const char *fmt, ...)
273df556
FM		 441{
442 va_list args;
f7c21177 443 struct va_format vaf;
1c13d5c0 444 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 445
1c13d5c0
TT		 446 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
447 es->s_last_error_block = cpu_to_le64(block);
efbed4dc
TT		 448 if (ext4_error_ratelimit(inode->i_sb)) {
449 va_start(args, fmt);
450 vaf.fmt = fmt;
451 vaf.va = &args;
452 if (block)
453 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
454 "inode #%lu: block %llu: comm %s: %pV\n",
455 inode->i_sb->s_id, function, line, inode->i_ino,
456 block, current->comm, &vaf);
457 else
458 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
459 "inode #%lu: comm %s: %pV\n",
460 inode->i_sb->s_id, function, line, inode->i_ino,
461 current->comm, &vaf);
462 va_end(args);
463 }
1c13d5c0 464 save_error_info(inode->i_sb, function, line);
273df556
FM		 465 ext4_handle_error(inode->i_sb);
466}
467
e7c96e8e
JP		 468void __ext4_error_file(struct file *file, const char *function,
469 unsigned int line, ext4_fsblk_t block,
470 const char *fmt, ...)
273df556
FM		 471{
472 va_list args;
f7c21177 473 struct va_format vaf;
1c13d5c0 474 struct ext4_super_block *es;
496ad9aa 475 struct inode *inode = file_inode(file);
273df556
FM		 476 char pathname[80], *path;
477
1c13d5c0
TT		 478 es = EXT4_SB(inode->i_sb)->s_es;
479 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
efbed4dc
TT		 480 if (ext4_error_ratelimit(inode->i_sb)) {
481 path = d_path(&(file->f_path), pathname, sizeof(pathname));
482 if (IS_ERR(path))
483 path = "(unknown)";
484 va_start(args, fmt);
485 vaf.fmt = fmt;
486 vaf.va = &args;
487 if (block)
488 printk(KERN_CRIT
489 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
490 "block %llu: comm %s: path %s: %pV\n",
491 inode->i_sb->s_id, function, line, inode->i_ino,
492 block, current->comm, path, &vaf);
493 else
494 printk(KERN_CRIT
495 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
496 "comm %s: path %s: %pV\n",
497 inode->i_sb->s_id, function, line, inode->i_ino,
498 current->comm, path, &vaf);
499 va_end(args);
500 }
1c13d5c0 501 save_error_info(inode->i_sb, function, line);
273df556
FM		 502 ext4_handle_error(inode->i_sb);
503}
504
722887dd
TT		 505const char *ext4_decode_error(struct super_block *sb, int errno,
506 char nbuf[16])
ac27a0ec
DK		 507{
508 char *errstr = NULL;
509
510 switch (errno) {
511 case -EIO:
512 errstr = "IO failure";
513 break;
514 case -ENOMEM:
515 errstr = "Out of memory";
516 break;
517 case -EROFS:
78f1ddbb
TT		 518 if (!sb || (EXT4_SB(sb)->s_journal &&
519 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 520 errstr = "Journal has aborted";
521 else
522 errstr = "Readonly filesystem";
523 break;
524 default:
525 /* If the caller passed in an extra buffer for unknown
526 * errors, textualise them now. Else we just return
527 * NULL. */
528 if (nbuf) {
529 /* Check for truncated error codes... */
530 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
531 errstr = nbuf;
532 }
533 break;
534 }
535
536 return errstr;
537}
538
617ba13b 539/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 540 * automatically and invokes the appropriate error response. */
541
c398eda0
TT		 542void __ext4_std_error(struct super_block *sb, const char *function,
543 unsigned int line, int errno)
ac27a0ec
DK		 544{
545 char nbuf[16];
546 const char *errstr;
547
548 /* Special case: if the error is EROFS, and we're not already
549 * inside a transaction, then there's really no point in logging
550 * an error. */
551 if (errno == -EROFS && journal_current_handle() == NULL &&
552 (sb->s_flags & MS_RDONLY))
553 return;
554
efbed4dc
TT		 555 if (ext4_error_ratelimit(sb)) {
556 errstr = ext4_decode_error(sb, errno, nbuf);
557 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
558 sb->s_id, function, line, errstr);
559 }
ac27a0ec 560
efbed4dc 561 save_error_info(sb, function, line);
617ba13b 562 ext4_handle_error(sb);
ac27a0ec
DK		 563}
564
565/*
617ba13b 566 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 567 * abort function may be used to deal with unrecoverable failures such
568 * as journal IO errors or ENOMEM at a critical moment in log management.
569 *
570 * We unconditionally force the filesystem into an ABORT|READONLY state,
571 * unless the error response on the fs has been set to panic in which
572 * case we take the easy way out and panic immediately.
573 */
574
c67d859e 575void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 576 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 577{
578 va_list args;
579
1c13d5c0 580 save_error_info(sb, function, line);
ac27a0ec 581 va_start(args, fmt);
c398eda0
TT		 582 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
583 function, line);
ac27a0ec
DK		 584 vprintk(fmt, args);
585 printk("\n");
586 va_end(args);
587
1c13d5c0
TT		 588 if ((sb->s_flags & MS_RDONLY) == 0) {
589 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
1c13d5c0 590 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
4418e141
DM		 591 /*
592 * Make sure updated value of ->s_mount_flags will be visible
593 * before ->s_flags update
594 */
595 smp_wmb();
596 sb->s_flags |= MS_RDONLY;
1c13d5c0
TT		 597 if (EXT4_SB(sb)->s_journal)
598 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
599 save_error_info(sb, function, line);
600 }
ac27a0ec 601 if (test_opt(sb, ERRORS_PANIC))
617ba13b 602 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 603}
604
e7c96e8e
JP		 605void __ext4_msg(struct super_block *sb,
606 const char *prefix, const char *fmt, ...)
b31e1552 607{
0ff2ea7d 608 struct va_format vaf;
b31e1552
ES		 609 va_list args;
610
efbed4dc
TT		 611 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
612 return;
613
b31e1552 614 va_start(args, fmt);
0ff2ea7d
JP		 615 vaf.fmt = fmt;
616 vaf.va = &args;
617 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 618 va_end(args);
619}
620
12062ddd 621void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 622 unsigned int line, const char *fmt, ...)
ac27a0ec 623{
0ff2ea7d 624 struct va_format vaf;
ac27a0ec
DK		 625 va_list args;
626
efbed4dc
TT		 627 if (!___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
628 "EXT4-fs warning"))
629 return;
630
ac27a0ec 631 va_start(args, fmt);
0ff2ea7d
JP		 632 vaf.fmt = fmt;
633 vaf.va = &args;
634 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
635 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 636 va_end(args);
637}
638
e29136f8
TT		 639void __ext4_grp_locked_error(const char *function, unsigned int line,
640 struct super_block *sb, ext4_group_t grp,
641 unsigned long ino, ext4_fsblk_t block,
642 const char *fmt, ...)
5d1b1b3f
AK		 643__releases(bitlock)
644__acquires(bitlock)
645{
0ff2ea7d 646 struct va_format vaf;
5d1b1b3f
AK		 647 va_list args;
648 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
649
1c13d5c0
TT		 650 es->s_last_error_ino = cpu_to_le32(ino);
651 es->s_last_error_block = cpu_to_le64(block);
652 __save_error_info(sb, function, line);
0ff2ea7d 653
efbed4dc
TT		 654 if (ext4_error_ratelimit(sb)) {
655 va_start(args, fmt);
656 vaf.fmt = fmt;
657 vaf.va = &args;
658 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
659 sb->s_id, function, line, grp);
660 if (ino)
661 printk(KERN_CONT "inode %lu: ", ino);
662 if (block)
663 printk(KERN_CONT "block %llu:",
664 (unsigned long long) block);
665 printk(KERN_CONT "%pV\n", &vaf);
666 va_end(args);
667 }
5d1b1b3f
AK		 668
669 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 670 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 671 return;
672 }
1c13d5c0 673
5d1b1b3f
AK		 674 ext4_unlock_group(sb, grp);
675 ext4_handle_error(sb);
676 /*
677 * We only get here in the ERRORS_RO case; relocking the group
678 * may be dangerous, but nothing bad will happen since the
679 * filesystem will have already been marked read/only and the
680 * journal has been aborted. We return 1 as a hint to callers
681 * who might what to use the return value from
25985edc 682 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 683 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
684 * aggressively from the ext4 function in question, with a
685 * more appropriate error code.
686 */
687 ext4_lock_group(sb, grp);
688 return;
689}
690
617ba13b 691void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 692{
617ba13b 693 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 694
617ba13b 695 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 696 return;
697
12062ddd 698 ext4_warning(sb,
ac27a0ec
DK		 699 "updating to rev %d because of new feature flag, "
700 "running e2fsck is recommended",
617ba13b 701 EXT4_DYNAMIC_REV);
ac27a0ec 702
617ba13b
MC		 703 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
704 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
705 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 706 /* leave es->s_feature_*compat flags alone */
707 /* es->s_uuid will be set by e2fsck if empty */
708
709 /*
710 * The rest of the superblock fields should be zero, and if not it
711 * means they are likely already in use, so leave them alone. We
712 * can leave it up to e2fsck to clean up any inconsistencies there.
713 */
714}
715
716/*
717 * Open the external journal device
718 */
b31e1552 719static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 720{
721 struct block_device *bdev;
722 char b[BDEVNAME_SIZE];
723
d4d77629 724 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 725 if (IS_ERR(bdev))
726 goto fail;
727 return bdev;
728
729fail:
b31e1552 730 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 731 __bdevname(dev, b), PTR_ERR(bdev));
732 return NULL;
733}
734
735/*
736 * Release the journal device
737 */
4385bab1 738static void ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 739{
4385bab1 740 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 741}
742
4385bab1 743static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 744{
745 struct block_device *bdev;
ac27a0ec
DK		 746 bdev = sbi->journal_bdev;
747 if (bdev) {
4385bab1 748 ext4_blkdev_put(bdev);
ac27a0ec
DK		 749 sbi->journal_bdev = NULL;
750 }
ac27a0ec
DK		 751}
752
753static inline struct inode *orphan_list_entry(struct list_head *l)
754{
617ba13b 755 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 756}
757
617ba13b 758static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 759{
760 struct list_head *l;
761
b31e1552
ES		 762 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
763 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 764
765 printk(KERN_ERR "sb_info orphan list:\n");
766 list_for_each(l, &sbi->s_orphan) {
767 struct inode *inode = orphan_list_entry(l);
768 printk(KERN_ERR " "
769 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
770 inode->i_sb->s_id, inode->i_ino, inode,
771 inode->i_mode, inode->i_nlink,
772 NEXT_ORPHAN(inode));
773 }
774}
775
2b2d6d01 776static void ext4_put_super(struct super_block *sb)
ac27a0ec 777{
617ba13b
MC		 778 struct ext4_sb_info *sbi = EXT4_SB(sb);
779 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 780 int i, err;
ac27a0ec 781
857ac889 782 ext4_unregister_li_request(sb);
e0ccfd95
CH		 783 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
784
2e8fa54e 785 flush_workqueue(sbi->rsv_conversion_wq);
2e8fa54e 786 destroy_workqueue(sbi->rsv_conversion_wq);
4c0425ff 787
0390131b
FM		 788 if (sbi->s_journal) {
789 err = jbd2_journal_destroy(sbi->s_journal);
790 sbi->s_journal = NULL;
791 if (err < 0)
c67d859e 792 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 793 }
d4edac31 794
d3922a77 795 ext4_es_unregister_shrinker(sbi);
9105bb14 796 del_timer_sync(&sbi->s_err_report);
d4edac31
JB		 797 ext4_release_system_zone(sb);
798 ext4_mb_release(sb);
799 ext4_ext_release(sb);
800 ext4_xattr_put_super(sb);
801
ac27a0ec 802 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 803 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 804 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 805 }
58c5873a 806 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 807 ext4_commit_super(sb, 1);
808
240799cd 809 if (sbi->s_proc) {
66acdcf4 810 remove_proc_entry("options", sbi->s_proc);
9f6200bb 811 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 812 }
3197ebdb 813 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 814
815 for (i = 0; i < sbi->s_gdb_count; i++)
816 brelse(sbi->s_group_desc[i]);
f18a5f21 817 ext4_kvfree(sbi->s_group_desc);
9933fc0a 818 ext4_kvfree(sbi->s_flex_groups);
57042651 819 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 820 percpu_counter_destroy(&sbi->s_freeinodes_counter);
821 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 822 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 823 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
ac27a0ec
DK		 824 brelse(sbi->s_sbh);
825#ifdef CONFIG_QUOTA
826 for (i = 0; i < MAXQUOTAS; i++)
827 kfree(sbi->s_qf_names[i]);
828#endif
829
830 /* Debugging code just in case the in-memory inode orphan list
831 * isn't empty. The on-disk one can be non-empty if we've
832 * detected an error and taken the fs readonly, but the
833 * in-memory list had better be clean by this point. */
834 if (!list_empty(&sbi->s_orphan))
835 dump_orphan_list(sb, sbi);
836 J_ASSERT(list_empty(&sbi->s_orphan));
837
f98393a6 838 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 839 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
840 /*
841 * Invalidate the journal device's buffers. We don't want them
842 * floating about in memory - the physical journal device may
843 * hotswapped, and it breaks the `ro-after' testing code.
844 */
845 sync_blockdev(sbi->journal_bdev);
f98393a6 846 invalidate_bdev(sbi->journal_bdev);
617ba13b 847 ext4_blkdev_remove(sbi);
ac27a0ec 848 }
9c191f70
M		 849 if (sbi->s_mb_cache) {
850 ext4_xattr_destroy_cache(sbi->s_mb_cache);
851 sbi->s_mb_cache = NULL;
852 }
c5e06d10
JL		 853 if (sbi->s_mmp_tsk)
854 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 855 sb->s_fs_info = NULL;
3197ebdb
TT		 856 /*
857 * Now that we are completely done shutting down the
858 * superblock, we need to actually destroy the kobject.
859 */
3197ebdb
TT		 860 kobject_put(&sbi->s_kobj);
861 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 862 if (sbi->s_chksum_driver)
863 crypto_free_shash(sbi->s_chksum_driver);
705895b6 864 kfree(sbi->s_blockgroup_lock);
ac27a0ec 865 kfree(sbi);
ac27a0ec
DK		 866}
867
e18b890b 868static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 869
870/*
871 * Called inside transaction, so use GFP_NOFS
872 */
617ba13b 873static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 874{
617ba13b 875 struct ext4_inode_info *ei;
ac27a0ec 876
e6b4f8da 877 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 878 if (!ei)
879 return NULL;
0b8e58a1 880
ac27a0ec 881 ei->vfs_inode.i_version = 1;
202ee5df 882 spin_lock_init(&ei->i_raw_lock);
c9de560d
AT		 883 INIT_LIST_HEAD(&ei->i_prealloc_list);
884 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 885 ext4_es_init_tree(&ei->i_es_tree);
886 rwlock_init(&ei->i_es_lock);
74cd15cd
ZL		 887 INIT_LIST_HEAD(&ei->i_es_lru);
888 ei->i_es_lru_nr = 0;
d3922a77 889 ei->i_touch_when = 0;
d2a17637
MC		 890 ei->i_reserved_data_blocks = 0;
891 ei->i_reserved_meta_blocks = 0;
892 ei->i_allocated_meta_blocks = 0;
9d0be502 893 ei->i_da_metadata_calc_len = 0;
7e731bc9 894 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 895 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 896#ifdef CONFIG_QUOTA
897 ei->i_reserved_quota = 0;
898#endif
8aefcd55 899 ei->jinode = NULL;
2e8fa54e 900 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
744692dc 901 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 902 ei->i_sync_tid = 0;
903 ei->i_datasync_tid = 0;
f7ad6d2e 904 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 905 atomic_set(&ei->i_unwritten, 0);
2e8fa54e 906 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
0b8e58a1 907
ac27a0ec
DK		 908 return &ei->vfs_inode;
909}
910
7ff9c073
TT		 911static int ext4_drop_inode(struct inode *inode)
912{
913 int drop = generic_drop_inode(inode);
914
915 trace_ext4_drop_inode(inode, drop);
916 return drop;
917}
918
fa0d7e3d
NP		 919static void ext4_i_callback(struct rcu_head *head)
920{
921 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 922 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
923}
924
617ba13b 925static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 926{
9f7dd93d 927 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 928 ext4_msg(inode->i_sb, KERN_ERR,
929 "Inode %lu (%p): orphan list check failed!",
930 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 931 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
932 EXT4_I(inode), sizeof(struct ext4_inode_info),
933 true);
934 dump_stack();
935 }
fa0d7e3d 936 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 937}
938
51cc5068 939static void init_once(void *foo)
ac27a0ec 940{
617ba13b 941 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 942
a35afb83 943 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 944 init_rwsem(&ei->xattr_sem);
0e855ac8 945 init_rwsem(&ei->i_data_sem);
a35afb83 946 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 947}
948
e67bc2b3 949static int __init init_inodecache(void)
ac27a0ec 950{
617ba13b
MC		 951 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
952 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 953 0, (SLAB_RECLAIM_ACCOUNT|
954 SLAB_MEM_SPREAD),
20c2df83 955 init_once);
617ba13b 956 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 957 return -ENOMEM;
958 return 0;
959}
960
961static void destroy_inodecache(void)
962{
8c0a8537
KS		 963 /*
964 * Make sure all delayed rcu free inodes are flushed before we
965 * destroy cache.
966 */
967 rcu_barrier();
617ba13b 968 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 969}
970
0930fcc1 971void ext4_clear_inode(struct inode *inode)
ac27a0ec 972{
0930fcc1 973 invalidate_inode_buffers(inode);
dbd5768f 974 clear_inode(inode);
9f754758 975 dquot_drop(inode);
c2ea3fde 976 ext4_discard_preallocations(inode);
51865fda 977 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
74cd15cd 978 ext4_es_lru_del(inode);
8aefcd55
TT		 979 if (EXT4_I(inode)->jinode) {
980 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
981 EXT4_I(inode)->jinode);
982 jbd2_free_inode(EXT4_I(inode)->jinode);
983 EXT4_I(inode)->jinode = NULL;
984 }
ac27a0ec
DK		 985}
986
1b961ac0 987static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 988 u64 ino, u32 generation)
ac27a0ec 989{
ac27a0ec 990 struct inode *inode;
ac27a0ec 991
617ba13b 992 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 993 return ERR_PTR(-ESTALE);
617ba13b 994 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 995 return ERR_PTR(-ESTALE);
996
997 /* iget isn't really right if the inode is currently unallocated!!
998 *
617ba13b 999 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1000 * deleted, so we should be safe.
1001 *
1002 * Currently we don't know the generation for parent directory, so
1003 * a generation of 0 means "accept any"
1004 */
1d1fe1ee
DH		 1005 inode = ext4_iget(sb, ino);
1006 if (IS_ERR(inode))
1007 return ERR_CAST(inode);
1008 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1009 iput(inode);
1010 return ERR_PTR(-ESTALE);
1011 }
1b961ac0
CH		 1012
1013 return inode;
1014}
1015
1016static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1017 int fh_len, int fh_type)
1b961ac0
CH		 1018{
1019 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1020 ext4_nfs_get_inode);
1021}
1022
1023static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1024 int fh_len, int fh_type)
1b961ac0
CH		 1025{
1026 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1027 ext4_nfs_get_inode);
ac27a0ec
DK		 1028}
1029
c39a7f84
TO		 1030/*
1031 * Try to release metadata pages (indirect blocks, directories) which are
1032 * mapped via the block device. Since these pages could have journal heads
1033 * which would prevent try_to_free_buffers() from freeing them, we must use
1034 * jbd2 layer's try_to_free_buffers() function to release them.
1035 */
0b8e58a1
AD		 1036static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1037 gfp_t wait)
c39a7f84
TO		 1038{
1039 journal_t *journal = EXT4_SB(sb)->s_journal;
1040
1041 WARN_ON(PageChecked(page));
1042 if (!page_has_buffers(page))
1043 return 0;
1044 if (journal)
1045 return jbd2_journal_try_to_free_buffers(journal, page,
1046 wait & ~__GFP_WAIT);
1047 return try_to_free_buffers(page);
1048}
1049
ac27a0ec 1050#ifdef CONFIG_QUOTA
af5bc92d 1051#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1052#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1053
617ba13b
MC		 1054static int ext4_write_dquot(struct dquot *dquot);
1055static int ext4_acquire_dquot(struct dquot *dquot);
1056static int ext4_release_dquot(struct dquot *dquot);
1057static int ext4_mark_dquot_dirty(struct dquot *dquot);
1058static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1059static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1060 struct path *path);
7c319d32
AK		 1061static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1062 int format_id);
ca0e05e4 1063static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1064static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1065static int ext4_quota_on_mount(struct super_block *sb, int type);
1066static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1067 size_t len, loff_t off);
617ba13b 1068static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1069 const char *data, size_t len, loff_t off);
7c319d32
AK		 1070static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1071 unsigned int flags);
1072static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1073
61e225dc 1074static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1075 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1076 .write_dquot = ext4_write_dquot,
1077 .acquire_dquot = ext4_acquire_dquot,
1078 .release_dquot = ext4_release_dquot,
1079 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1080 .write_info = ext4_write_info,
1081 .alloc_dquot = dquot_alloc,
1082 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1083};
1084
0d54b217 1085static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1086 .quota_on = ext4_quota_on,
ca0e05e4 1087 .quota_off = ext4_quota_off,
287a8095
CH		 1088 .quota_sync = dquot_quota_sync,
1089 .get_info = dquot_get_dqinfo,
1090 .set_info = dquot_set_dqinfo,
1091 .get_dqblk = dquot_get_dqblk,
1092 .set_dqblk = dquot_set_dqblk
ac27a0ec 1093};
7c319d32
AK		 1094
1095static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1096 .quota_on_meta = ext4_quota_on_sysfile,
1097 .quota_off = ext4_quota_off_sysfile,
1098 .quota_sync = dquot_quota_sync,
1099 .get_info = dquot_get_dqinfo,
1100 .set_info = dquot_set_dqinfo,
1101 .get_dqblk = dquot_get_dqblk,
1102 .set_dqblk = dquot_set_dqblk
1103};
ac27a0ec
DK		 1104#endif
1105
ee9b6d61 1106static const struct super_operations ext4_sops = {
617ba13b
MC		 1107 .alloc_inode = ext4_alloc_inode,
1108 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1109 .write_inode = ext4_write_inode,
1110 .dirty_inode = ext4_dirty_inode,
7ff9c073 1111 .drop_inode = ext4_drop_inode,
0930fcc1 1112 .evict_inode = ext4_evict_inode,
617ba13b 1113 .put_super = ext4_put_super,
617ba13b 1114 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1115 .freeze_fs = ext4_freeze,
1116 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1117 .statfs = ext4_statfs,
1118 .remount_fs = ext4_remount,
617ba13b 1119 .show_options = ext4_show_options,
ac27a0ec 1120#ifdef CONFIG_QUOTA
617ba13b
MC		 1121 .quota_read = ext4_quota_read,
1122 .quota_write = ext4_quota_write,
ac27a0ec 1123#endif
c39a7f84 1124 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1125};
1126
9ca92389
TT		 1127static const struct super_operations ext4_nojournal_sops = {
1128 .alloc_inode = ext4_alloc_inode,
1129 .destroy_inode = ext4_destroy_inode,
1130 .write_inode = ext4_write_inode,
1131 .dirty_inode = ext4_dirty_inode,
7ff9c073 1132 .drop_inode = ext4_drop_inode,
0930fcc1 1133 .evict_inode = ext4_evict_inode,
06a407f1 1134 .sync_fs = ext4_sync_fs_nojournal,
9ca92389
TT		 1135 .put_super = ext4_put_super,
1136 .statfs = ext4_statfs,
1137 .remount_fs = ext4_remount,
9ca92389
TT		 1138 .show_options = ext4_show_options,
1139#ifdef CONFIG_QUOTA
1140 .quota_read = ext4_quota_read,
1141 .quota_write = ext4_quota_write,
1142#endif
1143 .bdev_try_to_free_page = bdev_try_to_free_page,
1144};
1145
39655164 1146static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1147 .fh_to_dentry = ext4_fh_to_dentry,
1148 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1149 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1150};
1151
1152enum {
1153 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1154 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1155 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1156 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1157 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
ad4eec61
ES		 1158 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1159 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1160 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1161 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1162 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1163 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1164 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1165 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1166 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1167 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1168 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1169 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1170 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1171 Opt_max_dir_size_kb,
ac27a0ec
DK		 1172};
1173
a447c093 1174static const match_table_t tokens = {
ac27a0ec
DK		 1175 {Opt_bsd_df, "bsddf"},
1176 {Opt_minix_df, "minixdf"},
1177 {Opt_grpid, "grpid"},
1178 {Opt_grpid, "bsdgroups"},
1179 {Opt_nogrpid, "nogrpid"},
1180 {Opt_nogrpid, "sysvgroups"},
1181 {Opt_resgid, "resgid=%u"},
1182 {Opt_resuid, "resuid=%u"},
1183 {Opt_sb, "sb=%u"},
1184 {Opt_err_cont, "errors=continue"},
1185 {Opt_err_panic, "errors=panic"},
1186 {Opt_err_ro, "errors=remount-ro"},
1187 {Opt_nouid32, "nouid32"},
ac27a0ec 1188 {Opt_debug, "debug"},
72578c33
TT		 1189 {Opt_removed, "oldalloc"},
1190 {Opt_removed, "orlov"},
ac27a0ec
DK		 1191 {Opt_user_xattr, "user_xattr"},
1192 {Opt_nouser_xattr, "nouser_xattr"},
1193 {Opt_acl, "acl"},
1194 {Opt_noacl, "noacl"},
e3bb52ae 1195 {Opt_noload, "norecovery"},
5a916be1 1196 {Opt_noload, "noload"},
72578c33
TT		 1197 {Opt_removed, "nobh"},
1198 {Opt_removed, "bh"},
ac27a0ec 1199 {Opt_commit, "commit=%u"},
30773840
TT		 1200 {Opt_min_batch_time, "min_batch_time=%u"},
1201 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1202 {Opt_journal_dev, "journal_dev=%u"},
ad4eec61 1203 {Opt_journal_path, "journal_path=%s"},
818d276c
GS		 1204 {Opt_journal_checksum, "journal_checksum"},
1205 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1206 {Opt_abort, "abort"},
1207 {Opt_data_journal, "data=journal"},
1208 {Opt_data_ordered, "data=ordered"},
1209 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1210 {Opt_data_err_abort, "data_err=abort"},
1211 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1212 {Opt_offusrjquota, "usrjquota="},
1213 {Opt_usrjquota, "usrjquota=%s"},
1214 {Opt_offgrpjquota, "grpjquota="},
1215 {Opt_grpjquota, "grpjquota=%s"},
1216 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1217 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1218 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1219 {Opt_grpquota, "grpquota"},
1220 {Opt_noquota, "noquota"},
1221 {Opt_quota, "quota"},
1222 {Opt_usrquota, "usrquota"},
1223 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1224 {Opt_barrier, "barrier"},
1225 {Opt_nobarrier, "nobarrier"},
25ec56b5 1226 {Opt_i_version, "i_version"},
c9de560d 1227 {Opt_stripe, "stripe=%u"},
64769240 1228 {Opt_delalloc, "delalloc"},
dd919b98 1229 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1230 {Opt_removed, "mblk_io_submit"},
1231 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1232 {Opt_block_validity, "block_validity"},
1233 {Opt_noblock_validity, "noblock_validity"},
240799cd 1234 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1235 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1236 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1237 {Opt_auto_da_alloc, "auto_da_alloc"},
1238 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1239 {Opt_dioread_nolock, "dioread_nolock"},
1240 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1241 {Opt_discard, "discard"},
1242 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1243 {Opt_init_itable, "init_itable=%u"},
1244 {Opt_init_itable, "init_itable"},
1245 {Opt_noinit_itable, "noinit_itable"},
df981d03 1246 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1247 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1248 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1249 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1250 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1251 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1252 {Opt_err, NULL},
ac27a0ec
DK		 1253};
1254
617ba13b 1255static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1256{
617ba13b 1257 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1258 char *options = (char *) *data;
1259
1260 if (!options || strncmp(options, "sb=", 3) != 0)
1261 return 1; /* Default location */
0b8e58a1 1262
ac27a0ec 1263 options += 3;
0b8e58a1 1264 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1265 sb_block = simple_strtoul(options, &options, 0);
1266 if (*options && *options != ',') {
4776004f 1267 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1268 (char *) *data);
1269 return 1;
1270 }
1271 if (*options == ',')
1272 options++;
1273 *data = (void *) options;
0b8e58a1 1274
ac27a0ec
DK		 1275 return sb_block;
1276}
1277
b3881f74 1278#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1279static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1280 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1281
56c50f11
DM		 1282#ifdef CONFIG_QUOTA
1283static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1284{
1285 struct ext4_sb_info *sbi = EXT4_SB(sb);
1286 char *qname;
03dafb5f 1287 int ret = -1;
56c50f11
DM		 1288
1289 if (sb_any_quota_loaded(sb) &&
1290 !sbi->s_qf_names[qtype]) {
1291 ext4_msg(sb, KERN_ERR,
1292 "Cannot change journaled "
1293 "quota options when quota turned on");
57f73c2c 1294 return -1;
56c50f11 1295 }
262b4662
JK		 1296 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1297 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1298 "when QUOTA feature is enabled");
1299 return -1;
1300 }
56c50f11
DM		 1301 qname = match_strdup(args);
1302 if (!qname) {
1303 ext4_msg(sb, KERN_ERR,
1304 "Not enough memory for storing quotafile name");
57f73c2c 1305 return -1;
56c50f11 1306 }
03dafb5f
CG		 1307 if (sbi->s_qf_names[qtype]) {
1308 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1309 ret = 1;
1310 else
1311 ext4_msg(sb, KERN_ERR,
1312 "%s quota file already specified",
1313 QTYPE2NAME(qtype));
1314 goto errout;
56c50f11 1315 }
03dafb5f 1316 if (strchr(qname, '/')) {
56c50f11
DM		 1317 ext4_msg(sb, KERN_ERR,
1318 "quotafile must be on filesystem root");
03dafb5f 1319 goto errout;
56c50f11 1320 }
03dafb5f 1321 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1322 set_opt(sb, QUOTA);
56c50f11 1323 return 1;
03dafb5f
CG		 1324errout:
1325 kfree(qname);
1326 return ret;
56c50f11
DM		 1327}
1328
1329static int clear_qf_name(struct super_block *sb, int qtype)
1330{
1331
1332 struct ext4_sb_info *sbi = EXT4_SB(sb);
1333
1334 if (sb_any_quota_loaded(sb) &&
1335 sbi->s_qf_names[qtype]) {
1336 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1337 " when quota turned on");
57f73c2c 1338 return -1;
56c50f11 1339 }
03dafb5f 1340 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1341 sbi->s_qf_names[qtype] = NULL;
1342 return 1;
1343}
1344#endif
1345
26092bf5
TT		 1346#define MOPT_SET 0x0001
1347#define MOPT_CLEAR 0x0002
1348#define MOPT_NOSUPPORT 0x0004
1349#define MOPT_EXPLICIT 0x0008
1350#define MOPT_CLEAR_ERR 0x0010
1351#define MOPT_GTE0 0x0020
ac27a0ec 1352#ifdef CONFIG_QUOTA
26092bf5
TT		 1353#define MOPT_Q 0
1354#define MOPT_QFMT 0x0040
1355#else
1356#define MOPT_Q MOPT_NOSUPPORT
1357#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1358#endif
26092bf5 1359#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1360#define MOPT_NO_EXT2 0x0100
1361#define MOPT_NO_EXT3 0x0200
1362#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
ad4eec61 1363#define MOPT_STRING 0x0400
26092bf5
TT		 1364
1365static const struct mount_opts {
1366 int token;
1367 int mount_opt;
1368 int flags;
1369} ext4_mount_opts[] = {
1370 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1371 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1372 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1373 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1374 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1375 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1376 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1377 MOPT_EXT4_ONLY | MOPT_SET},
1378 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1379 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1380 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1381 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1382 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1383 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1384 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
59d9fa5c 1385 MOPT_EXT4_ONLY | MOPT_CLEAR},
8dc0aa8c
TT		 1386 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1387 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1388 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1389 EXT4_MOUNT_JOURNAL_CHECKSUM),
1390 MOPT_EXT4_ONLY | MOPT_SET},
1391 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1392 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1393 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1395 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1396 MOPT_NO_EXT2 | MOPT_SET},
1397 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1398 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1399 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1400 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1401 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1402 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1403 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1404 {Opt_commit, 0, MOPT_GTE0},
1405 {Opt_max_batch_time, 0, MOPT_GTE0},
1406 {Opt_min_batch_time, 0, MOPT_GTE0},
1407 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1408 {Opt_init_itable, 0, MOPT_GTE0},
1409 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1410 {Opt_resuid, 0, MOPT_GTE0},
1411 {Opt_resgid, 0, MOPT_GTE0},
1412 {Opt_journal_dev, 0, MOPT_GTE0},
ad4eec61 1413 {Opt_journal_path, 0, MOPT_STRING},
0efb3b23 1414 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1415 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1416 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1417 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1418 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1419 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1420 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1421#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1422 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1423 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1424#else
26092bf5
TT		 1425 {Opt_acl, 0, MOPT_NOSUPPORT},
1426 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1427#endif
26092bf5
TT		 1428 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1429 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1430 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1431 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1432 MOPT_SET | MOPT_Q},
1433 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1434 MOPT_SET | MOPT_Q},
1435 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1436 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1437 {Opt_usrjquota, 0, MOPT_Q},
1438 {Opt_grpjquota, 0, MOPT_Q},
1439 {Opt_offusrjquota, 0, MOPT_Q},
1440 {Opt_offgrpjquota, 0, MOPT_Q},
1441 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1442 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1443 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1444 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1445 {Opt_err, 0, 0}
1446};
1447
1448static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1449 substring_t *args, unsigned long *journal_devnum,
1450 unsigned int *journal_ioprio, int is_remount)
1451{
1452 struct ext4_sb_info *sbi = EXT4_SB(sb);
1453 const struct mount_opts *m;
08cefc7a
EB		 1454 kuid_t uid;
1455 kgid_t gid;
26092bf5
TT		 1456 int arg = 0;
1457
57f73c2c
TT		 1458#ifdef CONFIG_QUOTA
1459 if (token == Opt_usrjquota)
1460 return set_qf_name(sb, USRQUOTA, &args[0]);
1461 else if (token == Opt_grpjquota)
1462 return set_qf_name(sb, GRPQUOTA, &args[0]);
1463 else if (token == Opt_offusrjquota)
1464 return clear_qf_name(sb, USRQUOTA);
1465 else if (token == Opt_offgrpjquota)
1466 return clear_qf_name(sb, GRPQUOTA);
1467#endif
26092bf5 1468 switch (token) {
f7048605
TT		 1469 case Opt_noacl:
1470 case Opt_nouser_xattr:
1471 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1472 break;
26092bf5
TT		 1473 case Opt_sb:
1474 return 1; /* handled by get_sb_block() */
1475 case Opt_removed:
5f3633e3 1476 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1477 return 1;
26092bf5
TT		 1478 case Opt_abort:
1479 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1480 return 1;
1481 case Opt_i_version:
1482 sb->s_flags |= MS_I_VERSION;
1483 return 1;
26092bf5
TT		 1484 }
1485
5f3633e3
JK		 1486 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1487 if (token == m->token)
1488 break;
1489
1490 if (m->token == Opt_err) {
1491 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1492 "or missing value", opt);
1493 return -1;
1494 }
1495
8dc0aa8c
TT		 1496 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1497 ext4_msg(sb, KERN_ERR,
1498 "Mount option \"%s\" incompatible with ext2", opt);
1499 return -1;
1500 }
1501 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1502 ext4_msg(sb, KERN_ERR,
1503 "Mount option \"%s\" incompatible with ext3", opt);
1504 return -1;
1505 }
1506
ad4eec61 1507 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
5f3633e3
JK		 1508 return -1;
1509 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1510 return -1;
1511 if (m->flags & MOPT_EXPLICIT)
1512 set_opt2(sb, EXPLICIT_DELALLOC);
1513 if (m->flags & MOPT_CLEAR_ERR)
1514 clear_opt(sb, ERRORS_MASK);
1515 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1516 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1517 "options when quota turned on");
1518 return -1;
1519 }
1520
1521 if (m->flags & MOPT_NOSUPPORT) {
1522 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1523 } else if (token == Opt_commit) {
1524 if (arg == 0)
1525 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1526 sbi->s_commit_interval = HZ * arg;
1527 } else if (token == Opt_max_batch_time) {
5f3633e3
JK		 1528 sbi->s_max_batch_time = arg;
1529 } else if (token == Opt_min_batch_time) {
1530 sbi->s_min_batch_time = arg;
1531 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1532 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1533 ext4_msg(sb, KERN_ERR,
1534 "EXT4-fs: inode_readahead_blks must be "
1535 "0 or a power of 2 smaller than 2^31");
26092bf5 1536 return -1;
5f3633e3
JK		 1537 }
1538 sbi->s_inode_readahead_blks = arg;
1539 } else if (token == Opt_init_itable) {
1540 set_opt(sb, INIT_INODE_TABLE);
1541 if (!args->from)
1542 arg = EXT4_DEF_LI_WAIT_MULT;
1543 sbi->s_li_wait_mult = arg;
1544 } else if (token == Opt_max_dir_size_kb) {
1545 sbi->s_max_dir_size_kb = arg;
1546 } else if (token == Opt_stripe) {
1547 sbi->s_stripe = arg;
1548 } else if (token == Opt_resuid) {
1549 uid = make_kuid(current_user_ns(), arg);
1550 if (!uid_valid(uid)) {
1551 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1552 return -1;
1553 }
5f3633e3
JK		 1554 sbi->s_resuid = uid;
1555 } else if (token == Opt_resgid) {
1556 gid = make_kgid(current_user_ns(), arg);
1557 if (!gid_valid(gid)) {
1558 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1559 return -1;
1560 }
1561 sbi->s_resgid = gid;
1562 } else if (token == Opt_journal_dev) {
1563 if (is_remount) {
1564 ext4_msg(sb, KERN_ERR,
1565 "Cannot specify journal on remount");
1566 return -1;
1567 }
1568 *journal_devnum = arg;
ad4eec61
ES		 1569 } else if (token == Opt_journal_path) {
1570 char *journal_path;
1571 struct inode *journal_inode;
1572 struct path path;
1573 int error;
1574
1575 if (is_remount) {
1576 ext4_msg(sb, KERN_ERR,
1577 "Cannot specify journal on remount");
1578 return -1;
1579 }
1580 journal_path = match_strdup(&args[0]);
1581 if (!journal_path) {
1582 ext4_msg(sb, KERN_ERR, "error: could not dup "
1583 "journal device string");
1584 return -1;
1585 }
1586
1587 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1588 if (error) {
1589 ext4_msg(sb, KERN_ERR, "error: could not find "
1590 "journal device path: error %d", error);
1591 kfree(journal_path);
1592 return -1;
1593 }
1594
1595 journal_inode = path.dentry->d_inode;
1596 if (!S_ISBLK(journal_inode->i_mode)) {
1597 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1598 "is not a block device", journal_path);
1599 path_put(&path);
1600 kfree(journal_path);
1601 return -1;
1602 }
1603
1604 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1605 path_put(&path);
1606 kfree(journal_path);
5f3633e3
JK		 1607 } else if (token == Opt_journal_ioprio) {
1608 if (arg > 7) {
1609 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1610 " (must be 0-7)");
1611 return -1;
1612 }
1613 *journal_ioprio =
1614 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1615 } else if (m->flags & MOPT_DATAJ) {
1616 if (is_remount) {
1617 if (!sbi->s_journal)
1618 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1619 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1620 ext4_msg(sb, KERN_ERR,
26092bf5 1621 "Cannot change data mode on remount");
26092bf5 1622 return -1;
ac27a0ec 1623 }
26092bf5 1624 } else {
5f3633e3
JK		 1625 clear_opt(sb, DATA_FLAGS);
1626 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1627 }
5f3633e3
JK		 1628#ifdef CONFIG_QUOTA
1629 } else if (m->flags & MOPT_QFMT) {
1630 if (sb_any_quota_loaded(sb) &&
1631 sbi->s_jquota_fmt != m->mount_opt) {
1632 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1633 "quota options when quota turned on");
1634 return -1;
1635 }
262b4662
JK		 1636 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1637 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1638 ext4_msg(sb, KERN_ERR,
1639 "Cannot set journaled quota options "
1640 "when QUOTA feature is enabled");
1641 return -1;
1642 }
5f3633e3
JK		 1643 sbi->s_jquota_fmt = m->mount_opt;
1644#endif
1645 } else {
1646 if (!args->from)
1647 arg = 1;
1648 if (m->flags & MOPT_CLEAR)
1649 arg = !arg;
1650 else if (unlikely(!(m->flags & MOPT_SET))) {
1651 ext4_msg(sb, KERN_WARNING,
1652 "buggy handling of option %s", opt);
1653 WARN_ON(1);
1654 return -1;
1655 }
1656 if (arg != 0)
1657 sbi->s_mount_opt |= m->mount_opt;
1658 else
1659 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1660 }
5f3633e3 1661 return 1;
26092bf5
TT		 1662}
1663
1664static int parse_options(char *options, struct super_block *sb,
1665 unsigned long *journal_devnum,
1666 unsigned int *journal_ioprio,
1667 int is_remount)
1668{
1669 struct ext4_sb_info *sbi = EXT4_SB(sb);
1670 char *p;
1671 substring_t args[MAX_OPT_ARGS];
1672 int token;
1673
1674 if (!options)
1675 return 1;
1676
1677 while ((p = strsep(&options, ",")) != NULL) {
1678 if (!*p)
1679 continue;
1680 /*
1681 * Initialize args struct so we know whether arg was
1682 * found; some options take optional arguments.
1683 */
caecd0af 1684 args[0].to = args[0].from = NULL;
26092bf5
TT		 1685 token = match_token(p, tokens, args);
1686 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1687 journal_ioprio, is_remount) < 0)
1688 return 0;
ac27a0ec
DK		 1689 }
1690#ifdef CONFIG_QUOTA
262b4662
JK		 1691 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1692 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1693 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1694 "feature is enabled");
1695 return 0;
1696 }
ac27a0ec 1697 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1698 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1699 clear_opt(sb, USRQUOTA);
ac27a0ec 1700
482a7425 1701 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1702 clear_opt(sb, GRPQUOTA);
ac27a0ec 1703
56c50f11 1704 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1705 ext4_msg(sb, KERN_ERR, "old and new quota "
1706 "format mixing");
ac27a0ec
DK		 1707 return 0;
1708 }
1709
1710 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1711 ext4_msg(sb, KERN_ERR, "journaled quota format "
1712 "not specified");
ac27a0ec
DK		 1713 return 0;
1714 }
1715 } else {
1716 if (sbi->s_jquota_fmt) {
b31e1552 1717 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1718 "specified with no journaling "
b31e1552 1719 "enabled");
ac27a0ec
DK		 1720 return 0;
1721 }
1722 }
1723#endif
261cb20c
JK		 1724 if (test_opt(sb, DIOREAD_NOLOCK)) {
1725 int blocksize =
1726 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1727
1728 if (blocksize < PAGE_CACHE_SIZE) {
1729 ext4_msg(sb, KERN_ERR, "can't mount with "
1730 "dioread_nolock if block size != PAGE_SIZE");
1731 return 0;
1732 }
1733 }
ac27a0ec
DK		 1734 return 1;
1735}
1736
2adf6da8
TT		 1737static inline void ext4_show_quota_options(struct seq_file *seq,
1738 struct super_block *sb)
1739{
1740#if defined(CONFIG_QUOTA)
1741 struct ext4_sb_info *sbi = EXT4_SB(sb);
1742
1743 if (sbi->s_jquota_fmt) {
1744 char *fmtname = "";
1745
1746 switch (sbi->s_jquota_fmt) {
1747 case QFMT_VFS_OLD:
1748 fmtname = "vfsold";
1749 break;
1750 case QFMT_VFS_V0:
1751 fmtname = "vfsv0";
1752 break;
1753 case QFMT_VFS_V1:
1754 fmtname = "vfsv1";
1755 break;
1756 }
1757 seq_printf(seq, ",jqfmt=%s", fmtname);
1758 }
1759
1760 if (sbi->s_qf_names[USRQUOTA])
1761 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1762
1763 if (sbi->s_qf_names[GRPQUOTA])
1764 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
2adf6da8
TT		 1765#endif
1766}
1767
5a916be1
TT		 1768static const char *token2str(int token)
1769{
50df9fd5 1770 const struct match_token *t;
5a916be1
TT		 1771
1772 for (t = tokens; t->token != Opt_err; t++)
1773 if (t->token == token && !strchr(t->pattern, '='))
1774 break;
1775 return t->pattern;
1776}
1777
2adf6da8
TT		 1778/*
1779 * Show an option if
1780 * - it's set to a non-default value OR
1781 * - if the per-sb default is different from the global default
1782 */
66acdcf4
TT		 1783static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1784 int nodefs)
2adf6da8 1785{
2adf6da8
TT		 1786 struct ext4_sb_info *sbi = EXT4_SB(sb);
1787 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1788 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1789 const struct mount_opts *m;
66acdcf4 1790 char sep = nodefs ? '\n' : ',';
2adf6da8 1791
66acdcf4
TT		 1792#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1793#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1794
1795 if (sbi->s_sb_block != 1)
5a916be1
TT		 1796 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1797
1798 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1799 int want_set = m->flags & MOPT_SET;
1800 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1801 (m->flags & MOPT_CLEAR_ERR))
1802 continue;
66acdcf4 1803 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1804 continue; /* skip if same as the default */
1805 if ((want_set &&
1806 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1807 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1808 continue; /* select Opt_noFoo vs Opt_Foo */
1809 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1810 }
5a916be1 1811
08cefc7a 1812 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1813 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1814 SEQ_OPTS_PRINT("resuid=%u",
1815 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1816 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1817 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1818 SEQ_OPTS_PRINT("resgid=%u",
1819 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1820 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1821 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1822 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1823 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1824 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1825 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1826 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1827 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1828 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1829 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1830 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1831 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1832 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1833 if (sb->s_flags & MS_I_VERSION)
5a916be1 1834 SEQ_OPTS_PUTS("i_version");
66acdcf4 1835 if (nodefs || sbi->s_stripe)
5a916be1 1836 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1837 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1838 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1839 SEQ_OPTS_PUTS("data=journal");
1840 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1841 SEQ_OPTS_PUTS("data=ordered");
1842 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1843 SEQ_OPTS_PUTS("data=writeback");
1844 }
66acdcf4
TT		 1845 if (nodefs ||
1846 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1847 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1848 sbi->s_inode_readahead_blks);
2adf6da8 1849
66acdcf4
TT		 1850 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1851 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1852 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1853 if (nodefs || sbi->s_max_dir_size_kb)
1854 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1855
1856 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1857 return 0;
1858}
1859
66acdcf4
TT		 1860static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1861{
1862 return _ext4_show_options(seq, root->d_sb, 0);
1863}
1864
1865static int options_seq_show(struct seq_file *seq, void *offset)
1866{
1867 struct super_block *sb = seq->private;
1868 int rc;
1869
1870 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1871 rc = _ext4_show_options(seq, sb, 1);
1872 seq_puts(seq, "\n");
1873 return rc;
1874}
1875
1876static int options_open_fs(struct inode *inode, struct file *file)
1877{
d9dda78b 1878 return single_open(file, options_seq_show, PDE_DATA(inode));
66acdcf4
TT		 1879}
1880
1881static const struct file_operations ext4_seq_options_fops = {
1882 .owner = THIS_MODULE,
1883 .open = options_open_fs,
1884 .read = seq_read,
1885 .llseek = seq_lseek,
1886 .release = single_release,
1887};
1888
617ba13b 1889static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1890 int read_only)
1891{
617ba13b 1892 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1893 int res = 0;
1894
617ba13b 1895 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1896 ext4_msg(sb, KERN_ERR, "revision level too high, "
1897 "forcing read-only mode");
ac27a0ec
DK		 1898 res = MS_RDONLY;
1899 }
1900 if (read_only)
281b5995 1901 goto done;
617ba13b 1902 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1903 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1904 "running e2fsck is recommended");
c8b459f4 1905 else if (sbi->s_mount_state & EXT4_ERROR_FS)
b31e1552
ES		 1906 ext4_msg(sb, KERN_WARNING,
1907 "warning: mounting fs with errors, "
1908 "running e2fsck is recommended");
ed3ce80a 1909 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1910 le16_to_cpu(es->s_mnt_count) >=
1911 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1912 ext4_msg(sb, KERN_WARNING,
1913 "warning: maximal mount count reached, "
1914 "running e2fsck is recommended");
ac27a0ec
DK		 1915 else if (le32_to_cpu(es->s_checkinterval) &&
1916 (le32_to_cpu(es->s_lastcheck) +
1917 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1918 ext4_msg(sb, KERN_WARNING,
1919 "warning: checktime reached, "
1920 "running e2fsck is recommended");
0b8e58a1 1921 if (!sbi->s_journal)
0390131b 1922 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1923 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1924 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1925 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1926 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1927 ext4_update_dynamic_rev(sb);
0390131b
FM		 1928 if (sbi->s_journal)
1929 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1930
e2d67052 1931 ext4_commit_super(sb, 1);
281b5995 1932done:
ac27a0ec 1933 if (test_opt(sb, DEBUG))
a9df9a49 1934 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1935 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1936 sb->s_blocksize,
1937 sbi->s_groups_count,
617ba13b
MC		 1938 EXT4_BLOCKS_PER_GROUP(sb),
1939 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1940 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1941
7abc52c2 1942 cleancache_init_fs(sb);
ac27a0ec
DK		 1943 return res;
1944}
1945
117fff10
TT		 1946int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1947{
1948 struct ext4_sb_info *sbi = EXT4_SB(sb);
1949 struct flex_groups *new_groups;
1950 int size;
1951
1952 if (!sbi->s_log_groups_per_flex)
1953 return 0;
1954
1955 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1956 if (size <= sbi->s_flex_groups_allocated)
1957 return 0;
1958
1959 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1960 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1961 if (!new_groups) {
1962 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1963 size / (int) sizeof(struct flex_groups));
1964 return -ENOMEM;
1965 }
1966
1967 if (sbi->s_flex_groups) {
1968 memcpy(new_groups, sbi->s_flex_groups,
1969 (sbi->s_flex_groups_allocated *
1970 sizeof(struct flex_groups)));
1971 ext4_kvfree(sbi->s_flex_groups);
1972 }
1973 sbi->s_flex_groups = new_groups;
1974 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1975 return 0;
1976}
1977
772cb7c8
JS		 1978static int ext4_fill_flex_info(struct super_block *sb)
1979{
1980 struct ext4_sb_info *sbi = EXT4_SB(sb);
1981 struct ext4_group_desc *gdp = NULL;
772cb7c8 1982 ext4_group_t flex_group;
117fff10 1983 int i, err;
772cb7c8 1984
503358ae 1985 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1986 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1987 sbi->s_log_groups_per_flex = 0;
1988 return 1;
1989 }
1990
117fff10
TT		 1991 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1992 if (err)
9933fc0a 1993 goto failed;
772cb7c8 1994
772cb7c8 1995 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1996 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1997
1998 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1999 atomic_add(ext4_free_inodes_count(sb, gdp),
2000 &sbi->s_flex_groups[flex_group].free_inodes);
90ba983f
TT		 2001 atomic64_add(ext4_free_group_clusters(sb, gdp),
2002 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2003 atomic_add(ext4_used_dirs_count(sb, gdp),
2004 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2005 }
2006
2007 return 1;
2008failed:
2009 return 0;
2010}
2011
feb0ab32
DW		 2012static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2013 struct ext4_group_desc *gdp)
717d50e4 2014{
feb0ab32 2015 int offset;
717d50e4 2016 __u16 crc = 0;
feb0ab32 2017 __le32 le_group = cpu_to_le32(block_group);
717d50e4 2018
feb0ab32
DW		 2019 if ((sbi->s_es->s_feature_ro_compat &
2020 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2021 /* Use new metadata_csum algorithm */
d6a77105 2022 __le16 save_csum;
feb0ab32
DW		 2023 __u32 csum32;
2024
d6a77105 2025 save_csum = gdp->bg_checksum;
feb0ab32
DW		 2026 gdp->bg_checksum = 0;
2027 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2028 sizeof(le_group));
2029 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2030 sbi->s_desc_size);
d6a77105 2031 gdp->bg_checksum = save_csum;
feb0ab32
DW		 2032
2033 crc = csum32 & 0xFFFF;
2034 goto out;
717d50e4
AD		 2035 }
2036
feb0ab32
DW		 2037 /* old crc16 code */
2038 offset = offsetof(struct ext4_group_desc, bg_checksum);
2039
2040 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2041 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2042 crc = crc16(crc, (__u8 *)gdp, offset);
2043 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2044 /* for checksum of struct ext4_group_desc do the rest...*/
2045 if ((sbi->s_es->s_feature_incompat &
2046 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2047 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2048 crc = crc16(crc, (__u8 *)gdp + offset,
2049 le16_to_cpu(sbi->s_es->s_desc_size) -
2050 offset);
2051
2052out:
717d50e4
AD		 2053 return cpu_to_le16(crc);
2054}
2055
feb0ab32 2056int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2057 struct ext4_group_desc *gdp)
2058{
feb0ab32
DW		 2059 if (ext4_has_group_desc_csum(sb) &&
2060 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2061 block_group, gdp)))
717d50e4
AD		 2062 return 0;
2063
2064 return 1;
2065}
2066
feb0ab32
DW		 2067void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2068 struct ext4_group_desc *gdp)
2069{
2070 if (!ext4_has_group_desc_csum(sb))
2071 return;
2072 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2073}
2074
ac27a0ec 2075/* Called at mount-time, super-block is locked */
bfff6873
LC		 2076static int ext4_check_descriptors(struct super_block *sb,
2077 ext4_group_t *first_not_zeroed)
ac27a0ec 2078{
617ba13b
MC		 2079 struct ext4_sb_info *sbi = EXT4_SB(sb);
2080 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2081 ext4_fsblk_t last_block;
bd81d8ee
LV		 2082 ext4_fsblk_t block_bitmap;
2083 ext4_fsblk_t inode_bitmap;
2084 ext4_fsblk_t inode_table;
ce421581 2085 int flexbg_flag = 0;
bfff6873 2086 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2087
ce421581
JS		 2088 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2089 flexbg_flag = 1;
2090
af5bc92d 2091 ext4_debug("Checking group descriptors");
ac27a0ec 2092
197cd65a
AM		 2093 for (i = 0; i < sbi->s_groups_count; i++) {
2094 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2095
ce421581 2096 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2097 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2098 else
2099 last_block = first_block +
617ba13b 2100 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2101
bfff6873
LC		 2102 if ((grp == sbi->s_groups_count) &&
2103 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2104 grp = i;
2105
8fadc143 2106 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2107 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2108 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2109 "Block bitmap for group %u not in group "
b31e1552 2110 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2111 return 0;
2112 }
8fadc143 2113 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2114 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2115 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2116 "Inode bitmap for group %u not in group "
b31e1552 2117 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2118 return 0;
2119 }
8fadc143 2120 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2121 if (inode_table < first_block ||
2b2d6d01 2122 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2123 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2124 "Inode table for group %u not in group "
b31e1552 2125 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2126 return 0;
2127 }
955ce5f5 2128 ext4_lock_group(sb, i);
feb0ab32 2129 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2130 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2131 "Checksum for group %u failed (%u!=%u)",
2132 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2133 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2134 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2135 ext4_unlock_group(sb, i);
8a266467 2136 return 0;
7ee1ec4c 2137 }
717d50e4 2138 }
955ce5f5 2139 ext4_unlock_group(sb, i);
ce421581
JS		 2140 if (!flexbg_flag)
2141 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2142 }
bfff6873
LC		 2143 if (NULL != first_not_zeroed)
2144 *first_not_zeroed = grp;
ac27a0ec 2145
5dee5437
TT		 2146 ext4_free_blocks_count_set(sbi->s_es,
2147 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2148 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2149 return 1;
2150}
2151
617ba13b 2152/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2153 * the superblock) which were deleted from all directories, but held open by
2154 * a process at the time of a crash. We walk the list and try to delete these
2155 * inodes at recovery time (only with a read-write filesystem).
2156 *
2157 * In order to keep the orphan inode chain consistent during traversal (in
2158 * case of crash during recovery), we link each inode into the superblock
2159 * orphan list_head and handle it the same way as an inode deletion during
2160 * normal operation (which journals the operations for us).
2161 *
2162 * We only do an iget() and an iput() on each inode, which is very safe if we
2163 * accidentally point at an in-use or already deleted inode. The worst that
2164 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2165 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2166 * e2fsck was run on this filesystem, and it must have already done the orphan
2167 * inode cleanup for us, so we can safely abort without any further action.
2168 */
2b2d6d01
TT		 2169static void ext4_orphan_cleanup(struct super_block *sb,
2170 struct ext4_super_block *es)
ac27a0ec
DK		 2171{
2172 unsigned int s_flags = sb->s_flags;
2173 int nr_orphans = 0, nr_truncates = 0;
2174#ifdef CONFIG_QUOTA
2175 int i;
2176#endif
2177 if (!es->s_last_orphan) {
2178 jbd_debug(4, "no orphan inodes to clean up\n");
2179 return;
2180 }
2181
a8f48a95 2182 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2183 ext4_msg(sb, KERN_ERR, "write access "
2184 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2185 return;
2186 }
2187
d39195c3
AG		 2188 /* Check if feature set would not allow a r/w mount */
2189 if (!ext4_feature_set_ok(sb, 0)) {
2190 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2191 "unknown ROCOMPAT features");
2192 return;
2193 }
2194
617ba13b 2195 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2196 /* don't clear list on RO mount w/ errors */
2197 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2198 jbd_debug(1, "Errors on filesystem, "
2199 "clearing orphan list.\n");
c25f9bc6
ES		 2200 es->s_last_orphan = 0;
2201 }
ac27a0ec
DK		 2202 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2203 return;
2204 }
2205
2206 if (s_flags & MS_RDONLY) {
b31e1552 2207 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2208 sb->s_flags &= ~MS_RDONLY;
2209 }
2210#ifdef CONFIG_QUOTA
2211 /* Needed for iput() to work correctly and not trash data */
2212 sb->s_flags |= MS_ACTIVE;
2213 /* Turn on quotas so that they are updated correctly */
2214 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2215 if (EXT4_SB(sb)->s_qf_names[i]) {
2216 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2217 if (ret < 0)
b31e1552
ES		 2218 ext4_msg(sb, KERN_ERR,
2219 "Cannot turn on journaled "
2220 "quota: error %d", ret);
ac27a0ec
DK		 2221 }
2222 }
2223#endif
2224
2225 while (es->s_last_orphan) {
2226 struct inode *inode;
2227
97bd42b9
JB		 2228 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2229 if (IS_ERR(inode)) {
ac27a0ec
DK		 2230 es->s_last_orphan = 0;
2231 break;
2232 }
2233
617ba13b 2234 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2235 dquot_initialize(inode);
ac27a0ec 2236 if (inode->i_nlink) {
566370a2
PT		 2237 if (test_opt(sb, DEBUG))
2238 ext4_msg(sb, KERN_DEBUG,
2239 "%s: truncating inode %lu to %lld bytes",
2240 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2241 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2242 inode->i_ino, inode->i_size);
721e3eba 2243 mutex_lock(&inode->i_mutex);
55f252c9 2244 truncate_inode_pages(inode->i_mapping, inode->i_size);
617ba13b 2245 ext4_truncate(inode);
721e3eba 2246 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2247 nr_truncates++;
2248 } else {
566370a2
PT		 2249 if (test_opt(sb, DEBUG))
2250 ext4_msg(sb, KERN_DEBUG,
2251 "%s: deleting unreferenced inode %lu",
2252 __func__, inode->i_ino);
ac27a0ec
DK		 2253 jbd_debug(2, "deleting unreferenced inode %lu\n",
2254 inode->i_ino);
2255 nr_orphans++;
2256 }
2257 iput(inode); /* The delete magic happens here! */
2258 }
2259
2b2d6d01 2260#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2261
2262 if (nr_orphans)
b31e1552
ES		 2263 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2264 PLURAL(nr_orphans));
ac27a0ec 2265 if (nr_truncates)
b31e1552
ES		 2266 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2267 PLURAL(nr_truncates));
ac27a0ec
DK		 2268#ifdef CONFIG_QUOTA
2269 /* Turn quotas off */
2270 for (i = 0; i < MAXQUOTAS; i++) {
2271 if (sb_dqopt(sb)->files[i])
287a8095 2272 dquot_quota_off(sb, i);
ac27a0ec
DK		 2273 }
2274#endif
2275 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2276}
0b8e58a1 2277
cd2291a4
ES		 2278/*
2279 * Maximal extent format file size.
2280 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2281 * extent format containers, within a sector_t, and within i_blocks
2282 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2283 * so that won't be a limiting factor.
2284 *
f17722f9
LC		 2285 * However there is other limiting factor. We do store extents in the form
2286 * of starting block and length, hence the resulting length of the extent
2287 * covering maximum file size must fit into on-disk format containers as
2288 * well. Given that length is always by 1 unit bigger than max unit (because
2289 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2290 *
cd2291a4
ES		 2291 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2292 */
f287a1a5 2293static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2294{
2295 loff_t res;
2296 loff_t upper_limit = MAX_LFS_FILESIZE;
2297
2298 /* small i_blocks in vfs inode? */
f287a1a5 2299 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2300 /*
90c699a9 2301 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2302 * i_block represent total blocks in 512 bytes
2303 * 32 == size of vfs inode i_blocks * 8
2304 */
2305 upper_limit = (1LL << 32) - 1;
2306
2307 /* total blocks in file system block size */
2308 upper_limit >>= (blkbits - 9);
2309 upper_limit <<= blkbits;
2310 }
2311
f17722f9
LC		 2312 /*
2313 * 32-bit extent-start container, ee_block. We lower the maxbytes
2314 * by one fs block, so ee_len can cover the extent of maximum file
2315 * size
2316 */
2317 res = (1LL << 32) - 1;
cd2291a4 2318 res <<= blkbits;
cd2291a4
ES		 2319
2320 /* Sanity check against vm- & vfs- imposed limits */
2321 if (res > upper_limit)
2322 res = upper_limit;
2323
2324 return res;
2325}
ac27a0ec 2326
ac27a0ec 2327/*
cd2291a4 2328 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2329 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2330 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2331 */
f287a1a5 2332static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2333{
617ba13b 2334 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2335 int meta_blocks;
2336 loff_t upper_limit;
0b8e58a1
AD		 2337 /* This is calculated to be the largest file size for a dense, block
2338 * mapped file such that the file's total number of 512-byte sectors,
2339 * including data and all indirect blocks, does not exceed (2^48 - 1).
2340 *
2341 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2342 * number of 512-byte sectors of the file.
0fc1b451
AK		 2343 */
2344
f287a1a5 2345 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2346 /*
90c699a9 2347 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2348 * the inode i_block field represents total file blocks in
2349 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2350 */
2351 upper_limit = (1LL << 32) - 1;
2352
2353 /* total blocks in file system block size */
2354 upper_limit >>= (bits - 9);
2355
2356 } else {
8180a562
AK		 2357 /*
2358 * We use 48 bit ext4_inode i_blocks
2359 * With EXT4_HUGE_FILE_FL set the i_blocks
2360 * represent total number of blocks in
2361 * file system block size
2362 */
0fc1b451
AK		 2363 upper_limit = (1LL << 48) - 1;
2364
0fc1b451
AK		 2365 }
2366
2367 /* indirect blocks */
2368 meta_blocks = 1;
2369 /* double indirect blocks */
2370 meta_blocks += 1 + (1LL << (bits-2));
2371 /* tripple indirect blocks */
2372 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2373
2374 upper_limit -= meta_blocks;
2375 upper_limit <<= bits;
ac27a0ec
DK		 2376
2377 res += 1LL << (bits-2);
2378 res += 1LL << (2*(bits-2));
2379 res += 1LL << (3*(bits-2));
2380 res <<= bits;
2381 if (res > upper_limit)
2382 res = upper_limit;
0fc1b451
AK		 2383
2384 if (res > MAX_LFS_FILESIZE)
2385 res = MAX_LFS_FILESIZE;
2386
ac27a0ec
DK		 2387 return res;
2388}
2389
617ba13b 2390static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2391 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2392{
617ba13b 2393 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2394 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2395 int has_super = 0;
2396
2397 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2398
617ba13b 2399 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2400 nr < first_meta_bg)
70bbb3e0 2401 return logical_sb_block + nr + 1;
ac27a0ec 2402 bg = sbi->s_desc_per_block * nr;
617ba13b 2403 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2404 has_super = 1;
0b8e58a1 2405
bd63f6b0
DW		 2406 /*
2407 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2408 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2409 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2410 * compensate.
2411 */
2412 if (sb->s_blocksize == 1024 && nr == 0 &&
2413 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2414 has_super++;
2415
617ba13b 2416 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2417}
2418
c9de560d
AT		 2419/**
2420 * ext4_get_stripe_size: Get the stripe size.
2421 * @sbi: In memory super block info
2422 *
2423 * If we have specified it via mount option, then
2424 * use the mount option value. If the value specified at mount time is
2425 * greater than the blocks per group use the super block value.
2426 * If the super block value is greater than blocks per group return 0.
2427 * Allocator needs it be less than blocks per group.
2428 *
2429 */
2430static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2431{
2432 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2433 unsigned long stripe_width =
2434 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2435 int ret;
c9de560d
AT		 2436
2437 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2438 ret = sbi->s_stripe;
2439 else if (stripe_width <= sbi->s_blocks_per_group)
2440 ret = stripe_width;
2441 else if (stride <= sbi->s_blocks_per_group)
2442 ret = stride;
2443 else
2444 ret = 0;
c9de560d 2445
3eb08658
DE		 2446 /*
2447 * If the stripe width is 1, this makes no sense and
2448 * we set it to 0 to turn off stripe handling code.
2449 */
2450 if (ret <= 1)
2451 ret = 0;
c9de560d 2452
3eb08658 2453 return ret;
c9de560d 2454}
ac27a0ec 2455
3197ebdb
TT		 2456/* sysfs supprt */
2457
2458struct ext4_attr {
2459 struct attribute attr;
2460 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2461 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb 2462 const char *, size_t);
f2d50a65
JK		 2463 union {
2464 int offset;
2465 int deprecated_val;
2466 } u;
3197ebdb
TT		 2467};
2468
27dd4385
LC		 2469static int parse_strtoull(const char *buf,
2470 unsigned long long max, unsigned long long *value)
2471{
2472 int ret;
2473
2474 ret = kstrtoull(skip_spaces(buf), 0, value);
2475 if (!ret && *value > max)
2476 ret = -EINVAL;
2477 return ret;
2478}
2479
3197ebdb
TT		 2480static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2481 struct ext4_sb_info *sbi,
2482 char *buf)
2483{
2484 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2485 (s64) EXT4_C2B(sbi,
2486 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2487}
2488
2489static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2490 struct ext4_sb_info *sbi, char *buf)
2491{
2492 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2493
f613dfcb
TT		 2494 if (!sb->s_bdev->bd_part)
2495 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2496 return snprintf(buf, PAGE_SIZE, "%lu\n",
2497 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2498 sbi->s_sectors_written_start) >> 1);
2499}
2500
2501static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2502 struct ext4_sb_info *sbi, char *buf)
2503{
2504 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2505
f613dfcb
TT		 2506 if (!sb->s_bdev->bd_part)
2507 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2508 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2509 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2510 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2511 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2512}
2513
2514static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2515 struct ext4_sb_info *sbi,
2516 const char *buf, size_t count)
2517{
2518 unsigned long t;
e1091b15 2519 int ret;
3197ebdb 2520
e1091b15
LC		 2521 ret = kstrtoul(skip_spaces(buf), 0, &t);
2522 if (ret)
2523 return ret;
3197ebdb 2524
e1091b15 2525 if (t && (!is_power_of_2(t) || t > 0x40000000))
3197ebdb
TT		 2526 return -EINVAL;
2527
2528 sbi->s_inode_readahead_blks = t;
2529 return count;
2530}
2531
2532static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2533 struct ext4_sb_info *sbi, char *buf)
3197ebdb 2534{
f2d50a65 2535 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb
TT		 2536
2537 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2538}
2539
2540static ssize_t sbi_ui_store(struct ext4_attr *a,
2541 struct ext4_sb_info *sbi,
2542 const char *buf, size_t count)
2543{
f2d50a65 2544 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
3197ebdb 2545 unsigned long t;
e1091b15 2546 int ret;
3197ebdb 2547
e1091b15
LC		 2548 ret = kstrtoul(skip_spaces(buf), 0, &t);
2549 if (ret)
2550 return ret;
3197ebdb
TT		 2551 *ui = t;
2552 return count;
2553}
2554
27dd4385
LC		 2555static ssize_t reserved_clusters_show(struct ext4_attr *a,
2556 struct ext4_sb_info *sbi, char *buf)
2557{
2558 return snprintf(buf, PAGE_SIZE, "%llu\n",
2559 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2560}
2561
2562static ssize_t reserved_clusters_store(struct ext4_attr *a,
2563 struct ext4_sb_info *sbi,
2564 const char *buf, size_t count)
2565{
2566 unsigned long long val;
2567 int ret;
2568
2569 if (parse_strtoull(buf, -1ULL, &val))
2570 return -EINVAL;
2571 ret = ext4_reserve_clusters(sbi, val);
2572
2573 return ret ? ret : count;
2574}
2575
2c0544b2
TT		 2576static ssize_t trigger_test_error(struct ext4_attr *a,
2577 struct ext4_sb_info *sbi,
2578 const char *buf, size_t count)
2579{
2580 int len = count;
2581
2582 if (!capable(CAP_SYS_ADMIN))
2583 return -EPERM;
2584
2585 if (len && buf[len-1] == '\n')
2586 len--;
2587
2588 if (len)
2589 ext4_error(sbi->s_sb, "%.*s", len, buf);
2590 return count;
2591}
2592
f2d50a65
JK		 2593static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2594 struct ext4_sb_info *sbi, char *buf)
2595{
2596 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2597}
2598
3197ebdb
TT		 2599#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2600static struct ext4_attr ext4_attr_##_name = { \
2601 .attr = {.name = __stringify(_name), .mode = _mode }, \
2602 .show = _show, \
2603 .store = _store, \
f2d50a65
JK		 2604 .u = { \
2605 .offset = offsetof(struct ext4_sb_info, _elname),\
2606 }, \
3197ebdb
TT		 2607}
2608#define EXT4_ATTR(name, mode, show, store) \
2609static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2610
857ac889 2611#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2612#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2613#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2614#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2615 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2616#define ATTR_LIST(name) &ext4_attr_##name.attr
f2d50a65
JK		 2617#define EXT4_DEPRECATED_ATTR(_name, _val) \
2618static struct ext4_attr ext4_attr_##_name = { \
2619 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2620 .show = sbi_deprecated_show, \
2621 .u = { \
2622 .deprecated_val = _val, \
2623 }, \
2624}
3197ebdb
TT		 2625
2626EXT4_RO_ATTR(delayed_allocation_blocks);
2627EXT4_RO_ATTR(session_write_kbytes);
2628EXT4_RO_ATTR(lifetime_write_kbytes);
27dd4385 2629EXT4_RW_ATTR(reserved_clusters);
3197ebdb
TT		 2630EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2631 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2632EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2633EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2634EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2635EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2636EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2637EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2638EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
f2d50a65 2639EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
67a5da56 2640EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2641EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
efbed4dc
TT		 2642EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2643EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2644EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2645EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2646EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2647EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
3197ebdb
TT		 2648
2649static struct attribute *ext4_attrs[] = {
2650 ATTR_LIST(delayed_allocation_blocks),
2651 ATTR_LIST(session_write_kbytes),
2652 ATTR_LIST(lifetime_write_kbytes),
27dd4385 2653 ATTR_LIST(reserved_clusters),
3197ebdb 2654 ATTR_LIST(inode_readahead_blks),
11013911 2655 ATTR_LIST(inode_goal),
3197ebdb
TT		 2656 ATTR_LIST(mb_stats),
2657 ATTR_LIST(mb_max_to_scan),
2658 ATTR_LIST(mb_min_to_scan),
2659 ATTR_LIST(mb_order2_req),
2660 ATTR_LIST(mb_stream_req),
2661 ATTR_LIST(mb_group_prealloc),
55138e0b 2662 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2663 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2664 ATTR_LIST(trigger_fs_error),
efbed4dc
TT		 2665 ATTR_LIST(err_ratelimit_interval_ms),
2666 ATTR_LIST(err_ratelimit_burst),
2667 ATTR_LIST(warning_ratelimit_interval_ms),
2668 ATTR_LIST(warning_ratelimit_burst),
2669 ATTR_LIST(msg_ratelimit_interval_ms),
2670 ATTR_LIST(msg_ratelimit_burst),
3197ebdb
TT		 2671 NULL,
2672};
2673
857ac889
LC		 2674/* Features this copy of ext4 supports */
2675EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2676EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2677EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2678
2679static struct attribute *ext4_feat_attrs[] = {
2680 ATTR_LIST(lazy_itable_init),
27ee40df 2681 ATTR_LIST(batched_discard),
5e7bbef1 2682 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2683 NULL,
2684};
2685
3197ebdb
TT		 2686static ssize_t ext4_attr_show(struct kobject *kobj,
2687 struct attribute *attr, char *buf)
2688{
2689 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2690 s_kobj);
2691 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2692
2693 return a->show ? a->show(a, sbi, buf) : 0;
2694}
2695
2696static ssize_t ext4_attr_store(struct kobject *kobj,
2697 struct attribute *attr,
2698 const char *buf, size_t len)
2699{
2700 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2701 s_kobj);
2702 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2703
2704 return a->store ? a->store(a, sbi, buf, len) : 0;
2705}
2706
2707static void ext4_sb_release(struct kobject *kobj)
2708{
2709 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2710 s_kobj);
2711 complete(&sbi->s_kobj_unregister);
2712}
2713
52cf25d0 2714static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2715 .show = ext4_attr_show,
2716 .store = ext4_attr_store,
2717};
2718
2719static struct kobj_type ext4_ktype = {
2720 .default_attrs = ext4_attrs,
2721 .sysfs_ops = &ext4_attr_ops,
2722 .release = ext4_sb_release,
2723};
2724
857ac889
LC		 2725static void ext4_feat_release(struct kobject *kobj)
2726{
2727 complete(&ext4_feat->f_kobj_unregister);
2728}
2729
2730static struct kobj_type ext4_feat_ktype = {
2731 .default_attrs = ext4_feat_attrs,
2732 .sysfs_ops = &ext4_attr_ops,
2733 .release = ext4_feat_release,
2734};
2735
a13fb1a4
ES		 2736/*
2737 * Check whether this filesystem can be mounted based on
2738 * the features present and the RDONLY/RDWR mount requested.
2739 * Returns 1 if this filesystem can be mounted as requested,
2740 * 0 if it cannot be.
2741 */
2742static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2743{
2744 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2745 ext4_msg(sb, KERN_ERR,
2746 "Couldn't mount because of "
2747 "unsupported optional features (%x)",
2748 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2749 ~EXT4_FEATURE_INCOMPAT_SUPP));
2750 return 0;
2751 }
2752
2753 if (readonly)
2754 return 1;
2755
2756 /* Check that feature set is OK for a read-write mount */
2757 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2758 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2759 "unsupported optional features (%x)",
2760 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2761 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2762 return 0;
2763 }
2764 /*
2765 * Large file size enabled file system can only be mounted
2766 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2767 */
2768 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2769 if (sizeof(blkcnt_t) < sizeof(u64)) {
2770 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2771 "cannot be mounted RDWR without "
2772 "CONFIG_LBDAF");
2773 return 0;
2774 }
2775 }
bab08ab9
TT		 2776 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2777 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2778 ext4_msg(sb, KERN_ERR,
2779 "Can't support bigalloc feature without "
2780 "extents feature\n");
2781 return 0;
2782 }
7c319d32
AK		 2783
2784#ifndef CONFIG_QUOTA
2785 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2786 !readonly) {
2787 ext4_msg(sb, KERN_ERR,
2788 "Filesystem with quota feature cannot be mounted RDWR "
2789 "without CONFIG_QUOTA");
2790 return 0;
2791 }
2792#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2793 return 1;
2794}
2795
66e61a9e
TT		 2796/*
2797 * This function is called once a day if we have errors logged
2798 * on the file system
2799 */
2800static void print_daily_error_info(unsigned long arg)
2801{
2802 struct super_block *sb = (struct super_block *) arg;
2803 struct ext4_sb_info *sbi;
2804 struct ext4_super_block *es;
2805
2806 sbi = EXT4_SB(sb);
2807 es = sbi->s_es;
2808
2809 if (es->s_error_count)
ae0f78de
TT		 2810 /* fsck newer than v1.41.13 is needed to clean this condition. */
2811 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
66e61a9e
TT		 2812 le32_to_cpu(es->s_error_count));
2813 if (es->s_first_error_time) {
ae0f78de 2814 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
66e61a9e
TT		 2815 sb->s_id, le32_to_cpu(es->s_first_error_time),
2816 (int) sizeof(es->s_first_error_func),
2817 es->s_first_error_func,
2818 le32_to_cpu(es->s_first_error_line));
2819 if (es->s_first_error_ino)
2820 printk(": inode %u",
2821 le32_to_cpu(es->s_first_error_ino));
2822 if (es->s_first_error_block)
2823 printk(": block %llu", (unsigned long long)
2824 le64_to_cpu(es->s_first_error_block));
2825 printk("\n");
2826 }
2827 if (es->s_last_error_time) {
ae0f78de 2828 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
66e61a9e
TT		 2829 sb->s_id, le32_to_cpu(es->s_last_error_time),
2830 (int) sizeof(es->s_last_error_func),
2831 es->s_last_error_func,
2832 le32_to_cpu(es->s_last_error_line));
2833 if (es->s_last_error_ino)
2834 printk(": inode %u",
2835 le32_to_cpu(es->s_last_error_ino));
2836 if (es->s_last_error_block)
2837 printk(": block %llu", (unsigned long long)
2838 le64_to_cpu(es->s_last_error_block));
2839 printk("\n");
2840 }
2841 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2842}
2843
bfff6873
LC		 2844/* Find next suitable group and run ext4_init_inode_table */
2845static int ext4_run_li_request(struct ext4_li_request *elr)
2846{
2847 struct ext4_group_desc *gdp = NULL;
2848 ext4_group_t group, ngroups;
2849 struct super_block *sb;
2850 unsigned long timeout = 0;
2851 int ret = 0;
2852
2853 sb = elr->lr_super;
2854 ngroups = EXT4_SB(sb)->s_groups_count;
2855
8e8ad8a5 2856 sb_start_write(sb);
bfff6873
LC		 2857 for (group = elr->lr_next_group; group < ngroups; group++) {
2858 gdp = ext4_get_group_desc(sb, group, NULL);
2859 if (!gdp) {
2860 ret = 1;
2861 break;
2862 }
2863
2864 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2865 break;
2866 }
2867
7f511862 2868 if (group >= ngroups)
bfff6873
LC		 2869 ret = 1;
2870
2871 if (!ret) {
2872 timeout = jiffies;
2873 ret = ext4_init_inode_table(sb, group,
2874 elr->lr_timeout ? 0 : 1);
2875 if (elr->lr_timeout == 0) {
51ce6511
LC		 2876 timeout = (jiffies - timeout) *
2877 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2878 elr->lr_timeout = timeout;
2879 }
2880 elr->lr_next_sched = jiffies + elr->lr_timeout;
2881 elr->lr_next_group = group + 1;
2882 }
8e8ad8a5 2883 sb_end_write(sb);
bfff6873
LC		 2884
2885 return ret;
2886}
2887
2888/*
2889 * Remove lr_request from the list_request and free the
4ed5c033 2890 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2891 */
2892static void ext4_remove_li_request(struct ext4_li_request *elr)
2893{
2894 struct ext4_sb_info *sbi;
2895
2896 if (!elr)
2897 return;
2898
2899 sbi = elr->lr_sbi;
2900
2901 list_del(&elr->lr_request);
2902 sbi->s_li_request = NULL;
2903 kfree(elr);
2904}
2905
2906static void ext4_unregister_li_request(struct super_block *sb)
2907{
1bb933fb
LC		 2908 mutex_lock(&ext4_li_mtx);
2909 if (!ext4_li_info) {
2910 mutex_unlock(&ext4_li_mtx);
bfff6873 2911 return;
1bb933fb 2912 }
bfff6873
LC		 2913
2914 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2915 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2916 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2917 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2918}
2919
8f1f7453
ES		 2920static struct task_struct *ext4_lazyinit_task;
2921
bfff6873
LC		 2922/*
2923 * This is the function where ext4lazyinit thread lives. It walks
2924 * through the request list searching for next scheduled filesystem.
2925 * When such a fs is found, run the lazy initialization request
2926 * (ext4_rn_li_request) and keep track of the time spend in this
2927 * function. Based on that time we compute next schedule time of
2928 * the request. When walking through the list is complete, compute
2929 * next waking time and put itself into sleep.
2930 */
2931static int ext4_lazyinit_thread(void *arg)
2932{
2933 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2934 struct list_head *pos, *n;
2935 struct ext4_li_request *elr;
4ed5c033 2936 unsigned long next_wakeup, cur;
bfff6873
LC		 2937
2938 BUG_ON(NULL == eli);
2939
bfff6873
LC		 2940cont_thread:
2941 while (true) {
2942 next_wakeup = MAX_JIFFY_OFFSET;
2943
2944 mutex_lock(&eli->li_list_mtx);
2945 if (list_empty(&eli->li_request_list)) {
2946 mutex_unlock(&eli->li_list_mtx);
2947 goto exit_thread;
2948 }
2949
2950 list_for_each_safe(pos, n, &eli->li_request_list) {
2951 elr = list_entry(pos, struct ext4_li_request,
2952 lr_request);
2953
b2c78cd0
TT		 2954 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2955 if (ext4_run_li_request(elr) != 0) {
2956 /* error, remove the lazy_init job */
2957 ext4_remove_li_request(elr);
2958 continue;
2959 }
bfff6873
LC		 2960 }
2961
2962 if (time_before(elr->lr_next_sched, next_wakeup))
2963 next_wakeup = elr->lr_next_sched;
2964 }
2965 mutex_unlock(&eli->li_list_mtx);
2966
a0acae0e 2967 try_to_freeze();
bfff6873 2968
4ed5c033
LC		 2969 cur = jiffies;
2970 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2971 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2972 cond_resched();
2973 continue;
2974 }
2975
4ed5c033
LC		 2976 schedule_timeout_interruptible(next_wakeup - cur);
2977
8f1f7453
ES		 2978 if (kthread_should_stop()) {
2979 ext4_clear_request_list();
2980 goto exit_thread;
2981 }
bfff6873
LC		 2982 }
2983
2984exit_thread:
2985 /*
2986 * It looks like the request list is empty, but we need
2987 * to check it under the li_list_mtx lock, to prevent any
2988 * additions into it, and of course we should lock ext4_li_mtx
2989 * to atomically free the list and ext4_li_info, because at
2990 * this point another ext4 filesystem could be registering
2991 * new one.
2992 */
2993 mutex_lock(&ext4_li_mtx);
2994 mutex_lock(&eli->li_list_mtx);
2995 if (!list_empty(&eli->li_request_list)) {
2996 mutex_unlock(&eli->li_list_mtx);
2997 mutex_unlock(&ext4_li_mtx);
2998 goto cont_thread;
2999 }
3000 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 3001 kfree(ext4_li_info);
3002 ext4_li_info = NULL;
3003 mutex_unlock(&ext4_li_mtx);
3004
3005 return 0;
3006}
3007
3008static void ext4_clear_request_list(void)
3009{
3010 struct list_head *pos, *n;
3011 struct ext4_li_request *elr;
3012
3013 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 3014 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3015 elr = list_entry(pos, struct ext4_li_request,
3016 lr_request);
3017 ext4_remove_li_request(elr);
3018 }
3019 mutex_unlock(&ext4_li_info->li_list_mtx);
3020}
3021
3022static int ext4_run_lazyinit_thread(void)
3023{
8f1f7453
ES		 3024 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3025 ext4_li_info, "ext4lazyinit");
3026 if (IS_ERR(ext4_lazyinit_task)) {
3027 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 3028 ext4_clear_request_list();
bfff6873
LC		 3029 kfree(ext4_li_info);
3030 ext4_li_info = NULL;
92b97816 3031 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 3032 "initialization thread\n",
3033 err);
3034 return err;
3035 }
3036 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 3037 return 0;
3038}
3039
3040/*
3041 * Check whether it make sense to run itable init. thread or not.
3042 * If there is at least one uninitialized inode table, return
3043 * corresponding group number, else the loop goes through all
3044 * groups and return total number of groups.
3045 */
3046static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3047{
3048 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3049 struct ext4_group_desc *gdp = NULL;
3050
3051 for (group = 0; group < ngroups; group++) {
3052 gdp = ext4_get_group_desc(sb, group, NULL);
3053 if (!gdp)
3054 continue;
3055
3056 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3057 break;
3058 }
3059
3060 return group;
3061}
3062
3063static int ext4_li_info_new(void)
3064{
3065 struct ext4_lazy_init *eli = NULL;
3066
3067 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3068 if (!eli)
3069 return -ENOMEM;
3070
bfff6873
LC		 3071 INIT_LIST_HEAD(&eli->li_request_list);
3072 mutex_init(&eli->li_list_mtx);
3073
bfff6873
LC		 3074 eli->li_state |= EXT4_LAZYINIT_QUIT;
3075
3076 ext4_li_info = eli;
3077
3078 return 0;
3079}
3080
3081static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3082 ext4_group_t start)
3083{
3084 struct ext4_sb_info *sbi = EXT4_SB(sb);
3085 struct ext4_li_request *elr;
bfff6873
LC		 3086
3087 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3088 if (!elr)
3089 return NULL;
3090
3091 elr->lr_super = sb;
3092 elr->lr_sbi = sbi;
3093 elr->lr_next_group = start;
3094
3095 /*
3096 * Randomize first schedule time of the request to
3097 * spread the inode table initialization requests
3098 * better.
3099 */
dd1f723b
TT		 3100 elr->lr_next_sched = jiffies + (prandom_u32() %
3101 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
bfff6873
LC		 3102 return elr;
3103}
3104
7f511862
TT		 3105int ext4_register_li_request(struct super_block *sb,
3106 ext4_group_t first_not_zeroed)
bfff6873
LC		 3107{
3108 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3109 struct ext4_li_request *elr = NULL;
bfff6873 3110 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3111 int ret = 0;
bfff6873 3112
7f511862 3113 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3114 if (sbi->s_li_request != NULL) {
3115 /*
3116 * Reset timeout so it can be computed again, because
3117 * s_li_wait_mult might have changed.
3118 */
3119 sbi->s_li_request->lr_timeout = 0;
7f511862 3120 goto out;
51ce6511 3121 }
bfff6873
LC		 3122
3123 if (first_not_zeroed == ngroups ||
3124 (sb->s_flags & MS_RDONLY) ||
55ff3840 3125 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3126 goto out;
bfff6873
LC		 3127
3128 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3129 if (!elr) {
3130 ret = -ENOMEM;
3131 goto out;
3132 }
bfff6873
LC		 3133
3134 if (NULL == ext4_li_info) {
3135 ret = ext4_li_info_new();
3136 if (ret)
3137 goto out;
3138 }
3139
3140 mutex_lock(&ext4_li_info->li_list_mtx);
3141 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3142 mutex_unlock(&ext4_li_info->li_list_mtx);
3143
3144 sbi->s_li_request = elr;
46e4690b
TM		 3145 /*
3146 * set elr to NULL here since it has been inserted to
3147 * the request_list and the removal and free of it is
3148 * handled by ext4_clear_request_list from now on.
3149 */
3150 elr = NULL;
bfff6873
LC		 3151
3152 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3153 ret = ext4_run_lazyinit_thread();
3154 if (ret)
3155 goto out;
3156 }
bfff6873 3157out:
beed5ecb
NK		 3158 mutex_unlock(&ext4_li_mtx);
3159 if (ret)
bfff6873 3160 kfree(elr);
bfff6873
LC		 3161 return ret;
3162}
3163
3164/*
3165 * We do not need to lock anything since this is called on
3166 * module unload.
3167 */
3168static void ext4_destroy_lazyinit_thread(void)
3169{
3170 /*
3171 * If thread exited earlier
3172 * there's nothing to be done.
3173 */
8f1f7453 3174 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3175 return;
3176
8f1f7453 3177 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3178}
3179
25ed6e8a
DW		 3180static int set_journal_csum_feature_set(struct super_block *sb)
3181{
3182 int ret = 1;
3183 int compat, incompat;
3184 struct ext4_sb_info *sbi = EXT4_SB(sb);
3185
3186 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3187 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3188 /* journal checksum v2 */
3189 compat = 0;
3190 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3191 } else {
3192 /* journal checksum v1 */
3193 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3194 incompat = 0;
3195 }
3196
3197 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3198 ret = jbd2_journal_set_features(sbi->s_journal,
3199 compat, 0,
3200 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3201 incompat);
3202 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3203 ret = jbd2_journal_set_features(sbi->s_journal,
3204 compat, 0,
3205 incompat);
3206 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3207 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3208 } else {
3209 jbd2_journal_clear_features(sbi->s_journal,
3210 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3211 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3212 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3213 }
3214
3215 return ret;
3216}
3217
952fc18e
TT		 3218/*
3219 * Note: calculating the overhead so we can be compatible with
3220 * historical BSD practice is quite difficult in the face of
3221 * clusters/bigalloc. This is because multiple metadata blocks from
3222 * different block group can end up in the same allocation cluster.
3223 * Calculating the exact overhead in the face of clustered allocation
3224 * requires either O(all block bitmaps) in memory or O(number of block
3225 * groups**2) in time. We will still calculate the superblock for
3226 * older file systems --- and if we come across with a bigalloc file
3227 * system with zero in s_overhead_clusters the estimate will be close to
3228 * correct especially for very large cluster sizes --- but for newer
3229 * file systems, it's better to calculate this figure once at mkfs
3230 * time, and store it in the superblock. If the superblock value is
3231 * present (even for non-bigalloc file systems), we will use it.
3232 */
3233static int count_overhead(struct super_block *sb, ext4_group_t grp,
3234 char *buf)
3235{
3236 struct ext4_sb_info *sbi = EXT4_SB(sb);
3237 struct ext4_group_desc *gdp;
3238 ext4_fsblk_t first_block, last_block, b;
3239 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3240 int s, j, count = 0;
3241
0548bbb8
TT		 3242 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3243 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3244 sbi->s_itb_per_group + 2);
3245
952fc18e
TT		 3246 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3247 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3248 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3249 for (i = 0; i < ngroups; i++) {
3250 gdp = ext4_get_group_desc(sb, i, NULL);
3251 b = ext4_block_bitmap(sb, gdp);
3252 if (b >= first_block && b <= last_block) {
3253 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3254 count++;
3255 }
3256 b = ext4_inode_bitmap(sb, gdp);
3257 if (b >= first_block && b <= last_block) {
3258 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3259 count++;
3260 }
3261 b = ext4_inode_table(sb, gdp);
3262 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3263 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3264 int c = EXT4_B2C(sbi, b - first_block);
3265 ext4_set_bit(c, buf);
3266 count++;
3267 }
3268 if (i != grp)
3269 continue;
3270 s = 0;
3271 if (ext4_bg_has_super(sb, grp)) {
3272 ext4_set_bit(s++, buf);
3273 count++;
3274 }
3275 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3276 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3277 count++;
3278 }
3279 }
3280 if (!count)
3281 return 0;
3282 return EXT4_CLUSTERS_PER_GROUP(sb) -
3283 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3284}
3285
3286/*
3287 * Compute the overhead and stash it in sbi->s_overhead
3288 */
3289int ext4_calculate_overhead(struct super_block *sb)
3290{
3291 struct ext4_sb_info *sbi = EXT4_SB(sb);
3292 struct ext4_super_block *es = sbi->s_es;
3293 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3294 ext4_fsblk_t overhead = 0;
3295 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3296
952fc18e
TT		 3297 if (!buf)
3298 return -ENOMEM;
3299
3300 /*
3301 * Compute the overhead (FS structures). This is constant
3302 * for a given filesystem unless the number of block groups
3303 * changes so we cache the previous value until it does.
3304 */
3305
3306 /*
3307 * All of the blocks before first_data_block are overhead
3308 */
3309 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3310
3311 /*
3312 * Add the overhead found in each block group
3313 */
3314 for (i = 0; i < ngroups; i++) {
3315 int blks;
3316
3317 blks = count_overhead(sb, i, buf);
3318 overhead += blks;
3319 if (blks)
3320 memset(buf, 0, PAGE_SIZE);
3321 cond_resched();
3322 }
0875a2b4
ES		 3323 /* Add the journal blocks as well */
3324 if (sbi->s_journal)
810da240 3325 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
0875a2b4 3326
952fc18e
TT		 3327 sbi->s_overhead = overhead;
3328 smp_wmb();
3329 free_page((unsigned long) buf);
3330 return 0;
3331}
3332
27dd4385 3333
30fac0f7 3334static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
27dd4385
LC		 3335{
3336 ext4_fsblk_t resv_clusters;
3337
30fac0f7
JK		 3338 /*
3339 * There's no need to reserve anything when we aren't using extents.
3340 * The space estimates are exact, there are no unwritten extents,
3341 * hole punching doesn't need new metadata... This is needed especially
3342 * to keep ext2/3 backward compatibility.
3343 */
3344 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3345 return 0;
27dd4385
LC		 3346 /*
3347 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3348 * This should cover the situations where we can not afford to run
3349 * out of space like for example punch hole, or converting
556615dc 3350 * unwritten extents in delalloc path. In most cases such
27dd4385
LC		 3351 * allocation would require 1, or 2 blocks, higher numbers are
3352 * very rare.
3353 */
30fac0f7
JK		 3354 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3355 EXT4_SB(sb)->s_cluster_bits;
27dd4385
LC		 3356
3357 do_div(resv_clusters, 50);
3358 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3359
3360 return resv_clusters;
3361}
3362
3363
3364static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3365{
3366 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3367 sbi->s_cluster_bits;
3368
3369 if (count >= clusters)
3370 return -EINVAL;
3371
3372 atomic64_set(&sbi->s_resv_clusters, count);
3373 return 0;
3374}
3375
2b2d6d01 3376static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3377{
d4c402d9 3378 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3379 struct buffer_head *bh;
617ba13b
MC		 3380 struct ext4_super_block *es = NULL;
3381 struct ext4_sb_info *sbi;
3382 ext4_fsblk_t block;
3383 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3384 ext4_fsblk_t logical_sb_block;
ac27a0ec 3385 unsigned long offset = 0;
ac27a0ec
DK		 3386 unsigned long journal_devnum = 0;
3387 unsigned long def_mount_opts;
3388 struct inode *root;
9f6200bb 3389 char *cp;
0390131b 3390 const char *descr;
dcc7dae3 3391 int ret = -ENOMEM;
281b5995 3392 int blocksize, clustersize;
4ec11028
TT		 3393 unsigned int db_count;
3394 unsigned int i;
281b5995 3395 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3396 __u64 blocks_count;
07aa2ea1 3397 int err = 0;
b3881f74 3398 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3399 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3400
3401 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3402 if (!sbi)
dcc7dae3 3403 goto out_free_orig;
705895b6
PE		 3404
3405 sbi->s_blockgroup_lock =
3406 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3407 if (!sbi->s_blockgroup_lock) {
3408 kfree(sbi);
dcc7dae3 3409 goto out_free_orig;
705895b6 3410 }
ac27a0ec 3411 sb->s_fs_info = sbi;
2c0544b2 3412 sbi->s_sb = sb;
240799cd 3413 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3414 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3415 if (sb->s_bdev->bd_part)
3416 sbi->s_sectors_written_start =
3417 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3418
9f6200bb
TT		 3419 /* Cleanup superblock name */
3420 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3421 *cp = '!';
3422
07aa2ea1 3423 /* -EINVAL is default */
dcc7dae3 3424 ret = -EINVAL;
617ba13b 3425 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3426 if (!blocksize) {
b31e1552 3427 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3428 goto out_fail;
3429 }
3430
3431 /*
617ba13b 3432 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3433 * block sizes. We need to calculate the offset from buffer start.
3434 */
617ba13b 3435 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3436 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3437 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3438 } else {
70bbb3e0 3439 logical_sb_block = sb_block;
ac27a0ec
DK		 3440 }
3441
70bbb3e0 3442 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3443 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3444 goto out_fail;
3445 }
3446 /*
3447 * Note: s_es must be initialized as soon as possible because
617ba13b 3448 * some ext4 macro-instructions depend on its value
ac27a0ec 3449 */
2716b802 3450 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3451 sbi->s_es = es;
3452 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3453 if (sb->s_magic != EXT4_SUPER_MAGIC)
3454 goto cantfind_ext4;
afc32f7e 3455 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3456
feb0ab32
DW		 3457 /* Warn if metadata_csum and gdt_csum are both set. */
3458 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3459 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3460 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3461 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3462 "redundant flags; please run fsck.");
3463
d25425f8
DW		 3464 /* Check for a known checksum algorithm */
3465 if (!ext4_verify_csum_type(sb, es)) {
3466 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3467 "unknown checksum algorithm.");
3468 silent = 1;
3469 goto cantfind_ext4;
3470 }
3471
0441984a
DW		 3472 /* Load the checksum driver */
3473 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3474 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3475 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3476 if (IS_ERR(sbi->s_chksum_driver)) {
3477 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3478 ret = PTR_ERR(sbi->s_chksum_driver);
3479 sbi->s_chksum_driver = NULL;
3480 goto failed_mount;
3481 }
3482 }
3483
a9c47317
DW		 3484 /* Check superblock checksum */
3485 if (!ext4_superblock_csum_verify(sb, es)) {
3486 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3487 "invalid superblock checksum. Run e2fsck?");
3488 silent = 1;
3489 goto cantfind_ext4;
3490 }
3491
3492 /* Precompute checksum seed for all metadata */
3493 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3494 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3495 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3496 sizeof(es->s_uuid));
3497
ac27a0ec
DK		 3498 /* Set defaults before we parse the mount options */
3499 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3500 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3501 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3502 set_opt(sb, DEBUG);
87f26807 3503 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3504 set_opt(sb, GRPID);
617ba13b 3505 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3506 set_opt(sb, NO_UID32);
ea663336 3507 /* xattr user namespace & acls are now defaulted on */
ea663336 3508 set_opt(sb, XATTR_USER);
03010a33 3509#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3510 set_opt(sb, POSIX_ACL);
2e7842b8 3511#endif
617ba13b 3512 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3513 set_opt(sb, JOURNAL_DATA);
617ba13b 3514 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3515 set_opt(sb, ORDERED_DATA);
617ba13b 3516 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3517 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3518
3519 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3520 set_opt(sb, ERRORS_PANIC);
bb4f397a 3521 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3522 set_opt(sb, ERRORS_CONT);
bb4f397a 3523 else
fd8c37ec 3524 set_opt(sb, ERRORS_RO);
8b67f04a 3525 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3526 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3527 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3528 set_opt(sb, DISCARD);
ac27a0ec 3529
08cefc7a
EB		 3530 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3531 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3532 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3533 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3534 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3535
8b67f04a 3536 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3537 set_opt(sb, BARRIER);
ac27a0ec 3538
dd919b98
AK		 3539 /*
3540 * enable delayed allocation by default
3541 * Use -o nodelalloc to turn it off
3542 */
bc0b75f7 3543 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3544 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3545 set_opt(sb, DELALLOC);
dd919b98 3546
51ce6511
LC		 3547 /*
3548 * set default s_li_wait_mult for lazyinit, for the case there is
3549 * no mount option specified.
3550 */
3551 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3552
8b67f04a 3553 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3554 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3555 ext4_msg(sb, KERN_WARNING,
3556 "failed to parse options in superblock: %s",
3557 sbi->s_es->s_mount_opts);
3558 }
5a916be1 3559 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3560 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3561 &journal_ioprio, 0))
ac27a0ec
DK		 3562 goto failed_mount;
3563
56889787
TT		 3564 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3565 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3566 "with data=journal disables delayed "
3567 "allocation and O_DIRECT support!\n");
3568 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3569 ext4_msg(sb, KERN_ERR, "can't mount with "
3570 "both data=journal and delalloc");
3571 goto failed_mount;
3572 }
3573 if (test_opt(sb, DIOREAD_NOLOCK)) {
3574 ext4_msg(sb, KERN_ERR, "can't mount with "
6ae6514b 3575 "both data=journal and dioread_nolock");
56889787
TT		 3576 goto failed_mount;
3577 }
3578 if (test_opt(sb, DELALLOC))
3579 clear_opt(sb, DELALLOC);
3580 }
3581
ac27a0ec 3582 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3583 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3584
617ba13b
MC		 3585 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3586 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3587 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3588 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3589 ext4_msg(sb, KERN_WARNING,
3590 "feature flags set on rev 0 fs, "
3591 "running e2fsck is recommended");
469108ff 3592
ed3654eb
TT		 3593 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3594 set_opt2(sb, HURD_COMPAT);
3595 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3596 EXT4_FEATURE_INCOMPAT_64BIT)) {
3597 ext4_msg(sb, KERN_ERR,
3598 "The Hurd can't support 64-bit file systems");
3599 goto failed_mount;
3600 }
3601 }
3602
2035e776
TT		 3603 if (IS_EXT2_SB(sb)) {
3604 if (ext2_feature_set_ok(sb))
3605 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3606 "using the ext4 subsystem");
3607 else {
3608 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3609 "to feature incompatibilities");
3610 goto failed_mount;
3611 }
3612 }
3613
3614 if (IS_EXT3_SB(sb)) {
3615 if (ext3_feature_set_ok(sb))
3616 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3617 "using the ext4 subsystem");
3618 else {
3619 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3620 "to feature incompatibilities");
3621 goto failed_mount;
3622 }
3623 }
3624
ac27a0ec
DK		 3625 /*
3626 * Check feature flags regardless of the revision level, since we
3627 * previously didn't change the revision level when setting the flags,
3628 * so there is a chance incompat flags are set on a rev 0 filesystem.
3629 */
a13fb1a4 3630 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3631 goto failed_mount;
a13fb1a4 3632
261cb20c 3633 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3634 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3635 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3636 ext4_msg(sb, KERN_ERR,
3637 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3638 goto failed_mount;
3639 }
3640
ac27a0ec 3641 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3642 /* Validate the filesystem blocksize */
3643 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3644 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3645 blocksize);
ac27a0ec
DK		 3646 goto failed_mount;
3647 }
3648
2b2d6d01 3649 brelse(bh);
70bbb3e0
AM		 3650 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3651 offset = do_div(logical_sb_block, blocksize);
3652 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3653 if (!bh) {
b31e1552
ES		 3654 ext4_msg(sb, KERN_ERR,
3655 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3656 goto failed_mount;
3657 }
2716b802 3658 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3659 sbi->s_es = es;
617ba13b 3660 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3661 ext4_msg(sb, KERN_ERR,
3662 "Magic mismatch, very weird!");
ac27a0ec
DK		 3663 goto failed_mount;
3664 }
3665 }
3666
a13fb1a4
ES		 3667 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3668 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3669 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3670 has_huge_files);
3671 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3672
617ba13b
MC		 3673 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3674 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3675 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3676 } else {
3677 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3678 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3679 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3680 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3681 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3682 ext4_msg(sb, KERN_ERR,
3683 "unsupported inode size: %d",
2b2d6d01 3684 sbi->s_inode_size);
ac27a0ec
DK		 3685 goto failed_mount;
3686 }
ef7f3835
KS		 3687 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3688 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3689 }
0b8e58a1 3690
0d1ee42f
AR		 3691 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3692 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3693 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3694 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3695 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3696 ext4_msg(sb, KERN_ERR,
3697 "unsupported descriptor size %lu",
0d1ee42f
AR		 3698 sbi->s_desc_size);
3699 goto failed_mount;
3700 }
3701 } else
3702 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3703
ac27a0ec 3704 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3705 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3706 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3707 goto cantfind_ext4;
0b8e58a1 3708
617ba13b 3709 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3710 if (sbi->s_inodes_per_block == 0)
617ba13b 3711 goto cantfind_ext4;
ac27a0ec
DK		 3712 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3713 sbi->s_inodes_per_block;
0d1ee42f 3714 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3715 sbi->s_sbh = bh;
3716 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3717 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3718 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3719
2b2d6d01 3720 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3721 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3722 sbi->s_def_hash_version = es->s_def_hash_version;
23301410
TT		 3723 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3724 i = le32_to_cpu(es->s_flags);
3725 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3726 sbi->s_hash_unsigned = 3;
3727 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
f99b2589 3728#ifdef __CHAR_UNSIGNED__
23301410
TT		 3729 if (!(sb->s_flags & MS_RDONLY))
3730 es->s_flags |=
3731 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3732 sbi->s_hash_unsigned = 3;
f99b2589 3733#else
23301410
TT		 3734 if (!(sb->s_flags & MS_RDONLY))
3735 es->s_flags |=
3736 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
f99b2589 3737#endif
23301410 3738 }
f99b2589 3739 }
ac27a0ec 3740
281b5995
TT		 3741 /* Handle clustersize */
3742 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3743 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3744 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3745 if (has_bigalloc) {
3746 if (clustersize < blocksize) {
3747 ext4_msg(sb, KERN_ERR,
3748 "cluster size (%d) smaller than "
3749 "block size (%d)", clustersize, blocksize);
3750 goto failed_mount;
3751 }
3752 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3753 le32_to_cpu(es->s_log_block_size);
3754 sbi->s_clusters_per_group =
3755 le32_to_cpu(es->s_clusters_per_group);
3756 if (sbi->s_clusters_per_group > blocksize * 8) {
3757 ext4_msg(sb, KERN_ERR,
3758 "#clusters per group too big: %lu",
3759 sbi->s_clusters_per_group);
3760 goto failed_mount;
3761 }
3762 if (sbi->s_blocks_per_group !=
3763 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3764 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3765 "clusters per group (%lu) inconsistent",
3766 sbi->s_blocks_per_group,
3767 sbi->s_clusters_per_group);
3768 goto failed_mount;
3769 }
3770 } else {
3771 if (clustersize != blocksize) {
3772 ext4_warning(sb, "fragment/cluster size (%d) != "
3773 "block size (%d)", clustersize,
3774 blocksize);
3775 clustersize = blocksize;
3776 }
3777 if (sbi->s_blocks_per_group > blocksize * 8) {
3778 ext4_msg(sb, KERN_ERR,
3779 "#blocks per group too big: %lu",
3780 sbi->s_blocks_per_group);
3781 goto failed_mount;
3782 }
3783 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3784 sbi->s_cluster_bits = 0;
ac27a0ec 3785 }
281b5995
TT		 3786 sbi->s_cluster_ratio = clustersize / blocksize;
3787
ac27a0ec 3788 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3789 ext4_msg(sb, KERN_ERR,
3790 "#inodes per group too big: %lu",
2b2d6d01 3791 sbi->s_inodes_per_group);
ac27a0ec
DK		 3792 goto failed_mount;
3793 }
3794
960fd856
TT		 3795 /* Do we have standard group size of clustersize * 8 blocks ? */
3796 if (sbi->s_blocks_per_group == clustersize << 3)
3797 set_opt2(sb, STD_GROUP_SIZE);
3798
bf43d84b
ES		 3799 /*
3800 * Test whether we have more sectors than will fit in sector_t,
3801 * and whether the max offset is addressable by the page cache.
3802 */
5a9ae68a 3803 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3804 ext4_blocks_count(es));
5a9ae68a 3805 if (err) {
b31e1552 3806 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3807 " too large to mount safely on this system");
ac27a0ec 3808 if (sizeof(sector_t) < 8)
90c699a9 3809 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3810 goto failed_mount;
3811 }
3812
617ba13b
MC		 3813 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3814 goto cantfind_ext4;
e7c95593 3815
0f2ddca6
FTN		 3816 /* check blocks count against device size */
3817 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3818 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3819 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3820 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3821 ext4_blocks_count(es), blocks_count);
3822 goto failed_mount;
3823 }
3824
0b8e58a1
AD		 3825 /*
3826 * It makes no sense for the first data block to be beyond the end
3827 * of the filesystem.
3828 */
3829 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3830 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3831 "block %u is beyond end of filesystem (%llu)",
3832 le32_to_cpu(es->s_first_data_block),
3833 ext4_blocks_count(es));
e7c95593
ES		 3834 goto failed_mount;
3835 }
bd81d8ee
LV		 3836 blocks_count = (ext4_blocks_count(es) -
3837 le32_to_cpu(es->s_first_data_block) +
3838 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3839 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3840 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3841 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3842 "(block count %llu, first data block %u, "
b31e1552 3843 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3844 ext4_blocks_count(es),
3845 le32_to_cpu(es->s_first_data_block),
3846 EXT4_BLOCKS_PER_GROUP(sb));
3847 goto failed_mount;
3848 }
bd81d8ee 3849 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3850 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3851 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3852 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3853 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3854 sbi->s_group_desc = ext4_kvmalloc(db_count *
3855 sizeof(struct buffer_head *),
3856 GFP_KERNEL);
ac27a0ec 3857 if (sbi->s_group_desc == NULL) {
b31e1552 3858 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3859 ret = -ENOMEM;
ac27a0ec
DK		 3860 goto failed_mount;
3861 }
3862
9f6200bb
TT		 3863 if (ext4_proc_root)
3864 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3865
66acdcf4
TT		 3866 if (sbi->s_proc)
3867 proc_create_data("options", S_IRUGO, sbi->s_proc,
3868 &ext4_seq_options_fops, sb);
3869
705895b6 3870 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3871
3872 for (i = 0; i < db_count; i++) {
70bbb3e0 3873 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3874 sbi->s_group_desc[i] = sb_bread(sb, block);
3875 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3876 ext4_msg(sb, KERN_ERR,
3877 "can't read group descriptor %d", i);
ac27a0ec
DK		 3878 db_count = i;
3879 goto failed_mount2;
3880 }
3881 }
00764937
AK		 3882
3883 /*
3884 * set up enough so that it can read an inode,
3885 * and create new inode for buddy allocator
3886 */
3887 sbi->s_gdb_count = db_count;
3888 if (!test_opt(sb, NOLOAD) &&
3889 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3890 sb->s_op = &ext4_sops;
3891 else
3892 sb->s_op = &ext4_nojournal_sops;
3893
3894 ext4_ext_init(sb);
3895 err = ext4_mb_init(sb);
3896 if (err) {
3897 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3898 err);
3899 goto failed_mount2;
3900 }
3901
bfff6873 3902 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3903 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
00764937 3904 goto failed_mount2a;
ac27a0ec 3905 }
772cb7c8
JS		 3906 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3907 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3908 ext4_msg(sb, KERN_ERR,
3909 "unable to initialize "
3910 "flex_bg meta info!");
00764937 3911 goto failed_mount2a;
772cb7c8
JS		 3912 }
3913
ac27a0ec
DK		 3914 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3915 spin_lock_init(&sbi->s_next_gen_lock);
3916
04496411
TM		 3917 init_timer(&sbi->s_err_report);
3918 sbi->s_err_report.function = print_daily_error_info;
3919 sbi->s_err_report.data = (unsigned long) sb;
3920
a75ae78f 3921 /* Register extent status tree shrinker */
d3922a77 3922 ext4_es_register_shrinker(sbi);
a75ae78f 3923
57042651 3924 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3925 ext4_count_free_clusters(sb));
ce7e010a
TT		 3926 if (!err) {
3927 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3928 ext4_count_free_inodes(sb));
3929 }
3930 if (!err) {
3931 err = percpu_counter_init(&sbi->s_dirs_counter,
3932 ext4_count_dirs(sb));
3933 }
3934 if (!err) {
57042651 3935 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a 3936 }
1ac6466f
TT		 3937 if (!err) {
3938 err = percpu_counter_init(&sbi->s_extent_cache_cnt, 0);
3939 }
ce7e010a
TT		 3940 if (err) {
3941 ext4_msg(sb, KERN_ERR, "insufficient memory");
3942 goto failed_mount3;
3943 }
3944
c9de560d 3945 sbi->s_stripe = ext4_get_stripe_size(sbi);
67a5da56 3946 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3947
617ba13b
MC		 3948 sb->s_export_op = &ext4_export_ops;
3949 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3950#ifdef CONFIG_QUOTA
617ba13b 3951 sb->dq_op = &ext4_quota_operations;
262b4662 3952 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
7c319d32 3953 sb->s_qcop = &ext4_qctl_sysfile_operations;
262b4662
JK		 3954 else
3955 sb->s_qcop = &ext4_qctl_operations;
ac27a0ec 3956#endif
f2fa2ffc
AK		 3957 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3958
ac27a0ec 3959 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3960 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3961
3962 sb->s_root = NULL;
3963
3964 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3965 EXT4_HAS_INCOMPAT_FEATURE(sb,
3966 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3967
c5e06d10
JL		 3968 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3969 !(sb->s_flags & MS_RDONLY))
3970 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3971 goto failed_mount3;
3972
ac27a0ec
DK		 3973 /*
3974 * The first inode we look at is the journal inode. Don't try
3975 * root first: it may be modified in the journal!
3976 */
3977 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3978 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3979 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3980 goto failed_mount3;
0390131b
FM		 3981 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3982 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3983 ext4_msg(sb, KERN_ERR, "required journal recovery "
3984 "suppressed and not mounted read-only");
744692dc 3985 goto failed_mount_wq;
ac27a0ec 3986 } else {
fd8c37ec 3987 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3988 sbi->s_journal = NULL;
3989 needs_recovery = 0;
3990 goto no_journal;
ac27a0ec
DK		 3991 }
3992
f32aaf2d 3993 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3994 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3995 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3996 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3997 goto failed_mount_wq;
eb40a09c
JS		 3998 }
3999
25ed6e8a
DW		 4000 if (!set_journal_csum_feature_set(sb)) {
4001 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4002 "feature set");
4003 goto failed_mount_wq;
d4da6c9c 4004 }
818d276c 4005
ac27a0ec
DK		 4006 /* We have now updated the journal if required, so we can
4007 * validate the data journaling mode. */
4008 switch (test_opt(sb, DATA_FLAGS)) {
4009 case 0:
4010 /* No mode set, assume a default based on the journal
63f57933
AM		 4011 * capabilities: ORDERED_DATA if the journal can
4012 * cope, else JOURNAL_DATA
4013 */
dab291af
MC		 4014 if (jbd2_journal_check_available_features
4015 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 4016 set_opt(sb, ORDERED_DATA);
ac27a0ec 4017 else
fd8c37ec 4018 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 4019 break;
4020
617ba13b
MC		 4021 case EXT4_MOUNT_ORDERED_DATA:
4022 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 4023 if (!jbd2_journal_check_available_features
4024 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 4025 ext4_msg(sb, KERN_ERR, "Journal does not support "
4026 "requested data journaling mode");
744692dc 4027 goto failed_mount_wq;
ac27a0ec
DK		 4028 }
4029 default:
4030 break;
4031 }
b3881f74 4032 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 4033
18aadd47
BJ		 4034 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4035
ce7e010a
TT		 4036 /*
4037 * The journal may have updated the bg summary counts, so we
4038 * need to update the global counters.
4039 */
57042651 4040 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 4041 ext4_count_free_clusters(sb));
ce7e010a
TT		 4042 percpu_counter_set(&sbi->s_freeinodes_counter,
4043 ext4_count_free_inodes(sb));
4044 percpu_counter_set(&sbi->s_dirs_counter,
4045 ext4_count_dirs(sb));
57042651 4046 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 4047
ce7e010a 4048no_journal:
9c191f70
M		 4049 if (ext4_mballoc_ready) {
4050 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4051 if (!sbi->s_mb_cache) {
4052 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4053 goto failed_mount_wq;
4054 }
4055 }
4056
952fc18e
TT		 4057 /*
4058 * Get the # of file system overhead blocks from the
4059 * superblock if present.
4060 */
4061 if (es->s_overhead_clusters)
4062 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4063 else {
07aa2ea1
LC		 4064 err = ext4_calculate_overhead(sb);
4065 if (err)
952fc18e
TT		 4066 goto failed_mount_wq;
4067 }
4068
fd89d5f2
TH		 4069 /*
4070 * The maximum number of concurrent works can be high and
4071 * concurrency isn't really necessary. Limit it to 1.
4072 */
2e8fa54e
JK		 4073 EXT4_SB(sb)->rsv_conversion_wq =
4074 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4075 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4076 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
07aa2ea1 4077 ret = -ENOMEM;
2e8fa54e
JK		 4078 goto failed_mount4;
4079 }
4080
ac27a0ec 4081 /*
dab291af 4082 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 4083 * so we can safely mount the rest of the filesystem now.
4084 */
4085
1d1fe1ee
DH		 4086 root = ext4_iget(sb, EXT4_ROOT_INO);
4087 if (IS_ERR(root)) {
b31e1552 4088 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 4089 ret = PTR_ERR(root);
32a9bb57 4090 root = NULL;
ac27a0ec
DK		 4091 goto failed_mount4;
4092 }
4093 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 4094 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 4095 iput(root);
ac27a0ec
DK		 4096 goto failed_mount4;
4097 }
48fde701 4098 sb->s_root = d_make_root(root);
1d1fe1ee 4099 if (!sb->s_root) {
b31e1552 4100 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 4101 ret = -ENOMEM;
4102 goto failed_mount4;
4103 }
ac27a0ec 4104
7e84b621
ES		 4105 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4106 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 4107
4108 /* determine the minimum size of new large inodes, if present */
4109 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4110 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4111 EXT4_GOOD_OLD_INODE_SIZE;
4112 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4113 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4114 if (sbi->s_want_extra_isize <
4115 le16_to_cpu(es->s_want_extra_isize))
4116 sbi->s_want_extra_isize =
4117 le16_to_cpu(es->s_want_extra_isize);
4118 if (sbi->s_want_extra_isize <
4119 le16_to_cpu(es->s_min_extra_isize))
4120 sbi->s_want_extra_isize =
4121 le16_to_cpu(es->s_min_extra_isize);
4122 }
4123 }
4124 /* Check if enough inode space is available */
4125 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4126 sbi->s_inode_size) {
4127 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4128 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 4129 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4130 "available");
ef7f3835
KS		 4131 }
4132
30fac0f7 4133 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
27dd4385
LC		 4134 if (err) {
4135 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
30fac0f7 4136 "reserved pool", ext4_calculate_resv_clusters(sb));
00764937 4137 goto failed_mount5;
27dd4385
LC		 4138 }
4139
6fd058f7
TT		 4140 err = ext4_setup_system_zone(sb);
4141 if (err) {
b31e1552 4142 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 4143 "zone (%d)", err);
dcf2d804 4144 goto failed_mount5;
c2774d84
AK		 4145 }
4146
bfff6873
LC		 4147 err = ext4_register_li_request(sb, first_not_zeroed);
4148 if (err)
dcf2d804 4149 goto failed_mount6;
bfff6873 4150
3197ebdb
TT		 4151 sbi->s_kobj.kset = ext4_kset;
4152 init_completion(&sbi->s_kobj_unregister);
4153 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4154 "%s", sb->s_id);
dcf2d804
TM		 4155 if (err)
4156 goto failed_mount7;
3197ebdb 4157
9b2ff357
JK		 4158#ifdef CONFIG_QUOTA
4159 /* Enable quota usage during mount. */
4160 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4161 !(sb->s_flags & MS_RDONLY)) {
4162 err = ext4_enable_quotas(sb);
4163 if (err)
4164 goto failed_mount8;
4165 }
4166#endif /* CONFIG_QUOTA */
4167
617ba13b
MC		 4168 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4169 ext4_orphan_cleanup(sb, es);
4170 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4171 if (needs_recovery) {
b31e1552 4172 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4173 ext4_mark_recovery_complete(sb, es);
4174 }
4175 if (EXT4_SB(sb)->s_journal) {
4176 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4177 descr = " journalled data mode";
4178 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4179 descr = " ordered data mode";
4180 else
4181 descr = " writeback data mode";
4182 } else
4183 descr = "out journal";
4184
79add3a3
LC		 4185 if (test_opt(sb, DISCARD)) {
4186 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4187 if (!blk_queue_discard(q))
4188 ext4_msg(sb, KERN_WARNING,
4189 "mounting with \"discard\" option, but "
4190 "the device does not support discard");
4191 }
4192
d4c402d9 4193 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4194 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4195 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4196
66e61a9e
TT		 4197 if (es->s_error_count)
4198 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4199
efbed4dc
TT		 4200 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4201 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4202 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4203 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4204
d4c402d9 4205 kfree(orig_data);
ac27a0ec
DK		 4206 return 0;
4207
617ba13b 4208cantfind_ext4:
ac27a0ec 4209 if (!silent)
b31e1552 4210 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4211 goto failed_mount;
4212
72ba7450
TT		 4213#ifdef CONFIG_QUOTA
4214failed_mount8:
4215 kobject_del(&sbi->s_kobj);
4216#endif
dcf2d804
TM		 4217failed_mount7:
4218 ext4_unregister_li_request(sb);
4219failed_mount6:
dcf2d804 4220 ext4_release_system_zone(sb);
00764937 4221failed_mount5:
94bf608a 4222 dput(sb->s_root);
32a9bb57 4223 sb->s_root = NULL;
94bf608a 4224failed_mount4:
b31e1552 4225 ext4_msg(sb, KERN_ERR, "mount failed");
2e8fa54e
JK		 4226 if (EXT4_SB(sb)->rsv_conversion_wq)
4227 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4c0425ff 4228failed_mount_wq:
0390131b
FM		 4229 if (sbi->s_journal) {
4230 jbd2_journal_destroy(sbi->s_journal);
4231 sbi->s_journal = NULL;
4232 }
ac27a0ec 4233failed_mount3:
d3922a77 4234 ext4_es_unregister_shrinker(sbi);
9105bb14 4235 del_timer_sync(&sbi->s_err_report);
9933fc0a
TT		 4236 if (sbi->s_flex_groups)
4237 ext4_kvfree(sbi->s_flex_groups);
57042651 4238 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4239 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4240 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4241 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1ac6466f 4242 percpu_counter_destroy(&sbi->s_extent_cache_cnt);
c5e06d10
JL		 4243 if (sbi->s_mmp_tsk)
4244 kthread_stop(sbi->s_mmp_tsk);
00764937
AK		 4245failed_mount2a:
4246 ext4_mb_release(sb);
ac27a0ec
DK		 4247failed_mount2:
4248 for (i = 0; i < db_count; i++)
4249 brelse(sbi->s_group_desc[i]);
f18a5f21 4250 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4251failed_mount:
00764937 4252 ext4_ext_release(sb);
0441984a
DW		 4253 if (sbi->s_chksum_driver)
4254 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4255 if (sbi->s_proc) {
66acdcf4 4256 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4257 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4258 }
ac27a0ec
DK		 4259#ifdef CONFIG_QUOTA
4260 for (i = 0; i < MAXQUOTAS; i++)
4261 kfree(sbi->s_qf_names[i]);
4262#endif
617ba13b 4263 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4264 brelse(bh);
4265out_fail:
4266 sb->s_fs_info = NULL;
f6830165 4267 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4268 kfree(sbi);
dcc7dae3 4269out_free_orig:
d4c402d9 4270 kfree(orig_data);
07aa2ea1 4271 return err ? err : ret;
ac27a0ec
DK		 4272}
4273
4274/*
4275 * Setup any per-fs journal parameters now. We'll do this both on
4276 * initial mount, once the journal has been initialised but before we've
4277 * done any recovery; and again on any subsequent remount.
4278 */
617ba13b 4279static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4280{
617ba13b 4281 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4282
30773840
TT		 4283 journal->j_commit_interval = sbi->s_commit_interval;
4284 journal->j_min_batch_time = sbi->s_min_batch_time;
4285 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4286
a931da6a 4287 write_lock(&journal->j_state_lock);
ac27a0ec 4288 if (test_opt(sb, BARRIER))
dab291af 4289 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4290 else
dab291af 4291 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4292 if (test_opt(sb, DATA_ERR_ABORT))
4293 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4294 else
4295 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4296 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4297}
4298
617ba13b 4299static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4300 unsigned int journal_inum)
4301{
4302 struct inode *journal_inode;
4303 journal_t *journal;
4304
0390131b
FM		 4305 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4306
ac27a0ec
DK		 4307 /* First, test for the existence of a valid inode on disk. Bad
4308 * things happen if we iget() an unused inode, as the subsequent
4309 * iput() will try to delete it. */
4310
1d1fe1ee
DH		 4311 journal_inode = ext4_iget(sb, journal_inum);
4312 if (IS_ERR(journal_inode)) {
b31e1552 4313 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4314 return NULL;
4315 }
4316 if (!journal_inode->i_nlink) {
4317 make_bad_inode(journal_inode);
4318 iput(journal_inode);
b31e1552 4319 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4320 return NULL;
4321 }
4322
e5f8eab8 4323 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4324 journal_inode, journal_inode->i_size);
1d1fe1ee 4325 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4326 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4327 iput(journal_inode);
4328 return NULL;
4329 }
4330
dab291af 4331 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4332 if (!journal) {
b31e1552 4333 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4334 iput(journal_inode);
4335 return NULL;
4336 }
4337 journal->j_private = sb;
617ba13b 4338 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4339 return journal;
4340}
4341
617ba13b 4342static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4343 dev_t j_dev)
4344{
2b2d6d01 4345 struct buffer_head *bh;
ac27a0ec 4346 journal_t *journal;
617ba13b
MC		 4347 ext4_fsblk_t start;
4348 ext4_fsblk_t len;
ac27a0ec 4349 int hblock, blocksize;
617ba13b 4350 ext4_fsblk_t sb_block;
ac27a0ec 4351 unsigned long offset;
2b2d6d01 4352 struct ext4_super_block *es;
ac27a0ec
DK		 4353 struct block_device *bdev;
4354
0390131b
FM		 4355 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4356
b31e1552 4357 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4358 if (bdev == NULL)
4359 return NULL;
4360
ac27a0ec 4361 blocksize = sb->s_blocksize;
e1defc4f 4362 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4363 if (blocksize < hblock) {
b31e1552
ES		 4364 ext4_msg(sb, KERN_ERR,
4365 "blocksize too small for journal device");
ac27a0ec
DK		 4366 goto out_bdev;
4367 }
4368
617ba13b
MC		 4369 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4370 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4371 set_blocksize(bdev, blocksize);
4372 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4373 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4374 "external journal");
ac27a0ec
DK		 4375 goto out_bdev;
4376 }
4377
2716b802 4378 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4379 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4380 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4381 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4382 ext4_msg(sb, KERN_ERR, "external journal has "
4383 "bad superblock");
ac27a0ec
DK		 4384 brelse(bh);
4385 goto out_bdev;
4386 }
4387
617ba13b 4388 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4389 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4390 brelse(bh);
4391 goto out_bdev;
4392 }
4393
bd81d8ee 4394 len = ext4_blocks_count(es);
ac27a0ec
DK		 4395 start = sb_block + 1;
4396 brelse(bh); /* we're done with the superblock */
4397
dab291af 4398 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4399 start, len, blocksize);
4400 if (!journal) {
b31e1552 4401 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4402 goto out_bdev;
4403 }
4404 journal->j_private = sb;
9f203507 4405 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
ac27a0ec
DK		 4406 wait_on_buffer(journal->j_sb_buffer);
4407 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4408 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4409 goto out_journal;
4410 }
4411 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4412 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4413 "user (unsupported) - %d",
ac27a0ec
DK		 4414 be32_to_cpu(journal->j_superblock->s_nr_users));
4415 goto out_journal;
4416 }
617ba13b
MC		 4417 EXT4_SB(sb)->journal_bdev = bdev;
4418 ext4_init_journal_params(sb, journal);
ac27a0ec 4419 return journal;
0b8e58a1 4420
ac27a0ec 4421out_journal:
dab291af 4422 jbd2_journal_destroy(journal);
ac27a0ec 4423out_bdev:
617ba13b 4424 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4425 return NULL;
4426}
4427
617ba13b
MC		 4428static int ext4_load_journal(struct super_block *sb,
4429 struct ext4_super_block *es,
ac27a0ec
DK		 4430 unsigned long journal_devnum)
4431{
4432 journal_t *journal;
4433 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4434 dev_t journal_dev;
4435 int err = 0;
4436 int really_read_only;
4437
0390131b
FM		 4438 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4439
ac27a0ec
DK		 4440 if (journal_devnum &&
4441 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4442 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4443 "numbers have changed");
ac27a0ec
DK		 4444 journal_dev = new_decode_dev(journal_devnum);
4445 } else
4446 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4447
4448 really_read_only = bdev_read_only(sb->s_bdev);
4449
4450 /*
4451 * Are we loading a blank journal or performing recovery after a
4452 * crash? For recovery, we need to check in advance whether we
4453 * can get read-write access to the device.
4454 */
617ba13b 4455 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4456 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4457 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4458 "required on readonly filesystem");
ac27a0ec 4459 if (really_read_only) {
b31e1552
ES		 4460 ext4_msg(sb, KERN_ERR, "write access "
4461 "unavailable, cannot proceed");
ac27a0ec
DK		 4462 return -EROFS;
4463 }
b31e1552
ES		 4464 ext4_msg(sb, KERN_INFO, "write access will "
4465 "be enabled during recovery");
ac27a0ec
DK		 4466 }
4467 }
4468
4469 if (journal_inum && journal_dev) {
b31e1552
ES		 4470 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4471 "and inode journals!");
ac27a0ec
DK		 4472 return -EINVAL;
4473 }
4474
4475 if (journal_inum) {
617ba13b 4476 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4477 return -EINVAL;
4478 } else {
617ba13b 4479 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4480 return -EINVAL;
4481 }
4482
90576c0b 4483 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4484 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4485
617ba13b 4486 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4487 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4488 if (!err) {
4489 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4490 if (save)
4491 memcpy(save, ((char *) es) +
4492 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4493 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4494 if (save)
4495 memcpy(((char *) es) + EXT4_S_ERR_START,
4496 save, EXT4_S_ERR_LEN);
4497 kfree(save);
4498 }
ac27a0ec
DK		 4499
4500 if (err) {
b31e1552 4501 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4502 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4503 return err;
4504 }
4505
617ba13b
MC		 4506 EXT4_SB(sb)->s_journal = journal;
4507 ext4_clear_journal_err(sb, es);
ac27a0ec 4508
c41303ce 4509 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4510 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4511 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4512
4513 /* Make sure we flush the recovery flag to disk. */
e2d67052 4514 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4515 }
4516
4517 return 0;
4518}
4519
e2d67052 4520static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4521{
e2d67052 4522 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4523 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4524 int error = 0;
ac27a0ec 4525
7c2e7087 4526 if (!sbh || block_device_ejected(sb))
c4be0c1d 4527 return error;
914258bf
TT		 4528 if (buffer_write_io_error(sbh)) {
4529 /*
4530 * Oh, dear. A previous attempt to write the
4531 * superblock failed. This could happen because the
4532 * USB device was yanked out. Or it could happen to
4533 * be a transient write error and maybe the block will
4534 * be remapped. Nothing we can do but to retry the
4535 * write and hope for the best.
4536 */
b31e1552
ES		 4537 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4538 "superblock detected");
914258bf
TT		 4539 clear_buffer_write_io_error(sbh);
4540 set_buffer_uptodate(sbh);
4541 }
71290b36
TT		 4542 /*
4543 * If the file system is mounted read-only, don't update the
4544 * superblock write time. This avoids updating the superblock
4545 * write time when we are mounting the root file system
4546 * read/only but we need to replay the journal; at that point,
4547 * for people who are east of GMT and who make their clock
4548 * tick in localtime for Windows bug-for-bug compatibility,
4549 * the clock is set in the future, and this will cause e2fsck
4550 * to complain and force a full file system check.
4551 */
4552 if (!(sb->s_flags & MS_RDONLY))
4553 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4554 if (sb->s_bdev->bd_part)
4555 es->s_kbytes_written =
4556 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4557 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4558 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4559 else
4560 es->s_kbytes_written =
4561 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4562 ext4_free_blocks_count_set(es,
4563 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4564 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4565 es->s_free_inodes_count =
4566 cpu_to_le32(percpu_counter_sum_positive(
4567 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4568 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4569 ext4_superblock_csum_set(sb);
ac27a0ec 4570 mark_buffer_dirty(sbh);
914258bf 4571 if (sync) {
c4be0c1d
TS		 4572 error = sync_dirty_buffer(sbh);
4573 if (error)
4574 return error;
4575
4576 error = buffer_write_io_error(sbh);
4577 if (error) {
b31e1552
ES		 4578 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4579 "superblock");
914258bf
TT		 4580 clear_buffer_write_io_error(sbh);
4581 set_buffer_uptodate(sbh);
4582 }
4583 }
c4be0c1d 4584 return error;
ac27a0ec
DK		 4585}
4586
ac27a0ec
DK		 4587/*
4588 * Have we just finished recovery? If so, and if we are mounting (or
4589 * remounting) the filesystem readonly, then we will end up with a
4590 * consistent fs on disk. Record that fact.
4591 */
2b2d6d01
TT		 4592static void ext4_mark_recovery_complete(struct super_block *sb,
4593 struct ext4_super_block *es)
ac27a0ec 4594{
617ba13b 4595 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4596
0390131b
FM		 4597 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4598 BUG_ON(journal != NULL);
4599 return;
4600 }
dab291af 4601 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4602 if (jbd2_journal_flush(journal) < 0)
4603 goto out;
4604
617ba13b 4605 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4606 sb->s_flags & MS_RDONLY) {
617ba13b 4607 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4608 ext4_commit_super(sb, 1);
ac27a0ec 4609 }
7ffe1ea8
HK		 4610
4611out:
dab291af 4612 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4613}
4614
4615/*
4616 * If we are mounting (or read-write remounting) a filesystem whose journal
4617 * has recorded an error from a previous lifetime, move that error to the
4618 * main filesystem now.
4619 */
2b2d6d01
TT		 4620static void ext4_clear_journal_err(struct super_block *sb,
4621 struct ext4_super_block *es)
ac27a0ec
DK		 4622{
4623 journal_t *journal;
4624 int j_errno;
4625 const char *errstr;
4626
0390131b
FM		 4627 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4628
617ba13b 4629 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4630
4631 /*
4632 * Now check for any error status which may have been recorded in the
617ba13b 4633 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4634 */
4635
dab291af 4636 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4637 if (j_errno) {
4638 char nbuf[16];
4639
617ba13b 4640 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4641 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4642 "from previous mount: %s", errstr);
12062ddd 4643 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4644
617ba13b
MC		 4645 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4646 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4647 ext4_commit_super(sb, 1);
ac27a0ec 4648
dab291af 4649 jbd2_journal_clear_err(journal);
d796c52e 4650 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4651 }
4652}
4653
4654/*
4655 * Force the running and committing transactions to commit,
4656 * and wait on the commit.
4657 */
617ba13b 4658int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4659{
4660 journal_t *journal;
ac27a0ec
DK		 4661
4662 if (sb->s_flags & MS_RDONLY)
4663 return 0;
4664
617ba13b 4665 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4666 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4667}
4668
617ba13b 4669static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4670{
14ce0cb4 4671 int ret = 0;
9eddacf9 4672 tid_t target;
06a407f1 4673 bool needs_barrier = false;
8d5d02e6 4674 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4675
9bffad1e 4676 trace_ext4_sync_fs(sb, wait);
2e8fa54e 4677 flush_workqueue(sbi->rsv_conversion_wq);
a1177825
JK		 4678 /*
4679 * Writeback quota in non-journalled quota case - journalled quota has
4680 * no dirty dquots
4681 */
4682 dquot_writeback_dquots(sb, -1);
06a407f1
DM		 4683 /*
4684 * Data writeback is possible w/o journal transaction, so barrier must
4685 * being sent at the end of the function. But we can skip it if
4686 * transaction_commit will do it for us.
4687 */
4688 target = jbd2_get_latest_transaction(sbi->s_journal);
4689 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4690 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4691 needs_barrier = true;
4692
8d5d02e6 4693 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4694 if (wait)
06a407f1
DM		 4695 ret = jbd2_log_wait_commit(sbi->s_journal, target);
4696 }
4697 if (needs_barrier) {
4698 int err;
4699 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4700 if (!ret)
4701 ret = err;
0390131b 4702 }
06a407f1
DM		 4703
4704 return ret;
4705}
4706
4707static int ext4_sync_fs_nojournal(struct super_block *sb, int wait)
4708{
4709 int ret = 0;
4710
4711 trace_ext4_sync_fs(sb, wait);
4712 flush_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
06a407f1
DM		 4713 dquot_writeback_dquots(sb, -1);
4714 if (wait && test_opt(sb, BARRIER))
4715 ret = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4716
14ce0cb4 4717 return ret;
ac27a0ec
DK		 4718}
4719
4720/*
4721 * LVM calls this function before a (read-only) snapshot is created. This
4722 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4723 *
4724 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4725 * state independently. It relies on upper layer to stop all data & metadata
4726 * modifications.
ac27a0ec 4727 */
c4be0c1d 4728static int ext4_freeze(struct super_block *sb)
ac27a0ec 4729{
c4be0c1d
TS		 4730 int error = 0;
4731 journal_t *journal;
ac27a0ec 4732
9ca92389
TT		 4733 if (sb->s_flags & MS_RDONLY)
4734 return 0;
ac27a0ec 4735
9ca92389 4736 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4737
9ca92389
TT		 4738 /* Now we set up the journal barrier. */
4739 jbd2_journal_lock_updates(journal);
ac27a0ec 4740
9ca92389
TT		 4741 /*
4742 * Don't clear the needs_recovery flag if we failed to flush
4743 * the journal.
4744 */
4745 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4746 if (error < 0)
4747 goto out;
9ca92389
TT		 4748
4749 /* Journal blocked and flushed, clear needs_recovery flag. */
4750 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4751 error = ext4_commit_super(sb, 1);
6b0310fb 4752out:
8e8ad8a5 4753 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4754 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4755 return error;
ac27a0ec
DK		 4756}
4757
4758/*
4759 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4760 * flag here, even though the filesystem is not technically dirty yet.
4761 */
c4be0c1d 4762static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4763{
9ca92389
TT		 4764 if (sb->s_flags & MS_RDONLY)
4765 return 0;
4766
9ca92389
TT		 4767 /* Reset the needs_recovery flag before the fs is unlocked. */
4768 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4769 ext4_commit_super(sb, 1);
c4be0c1d 4770 return 0;
ac27a0ec
DK		 4771}
4772
673c6100
TT		 4773/*
4774 * Structure to save mount options for ext4_remount's benefit
4775 */
4776struct ext4_mount_options {
4777 unsigned long s_mount_opt;
a2595b8a 4778 unsigned long s_mount_opt2;
08cefc7a
EB		 4779 kuid_t s_resuid;
4780 kgid_t s_resgid;
673c6100
TT		 4781 unsigned long s_commit_interval;
4782 u32 s_min_batch_time, s_max_batch_time;
4783#ifdef CONFIG_QUOTA
4784 int s_jquota_fmt;
4785 char *s_qf_names[MAXQUOTAS];
4786#endif
4787};
4788
2b2d6d01 4789static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4790{
2b2d6d01 4791 struct ext4_super_block *es;
617ba13b 4792 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4793 unsigned long old_sb_flags;
617ba13b 4794 struct ext4_mount_options old_opts;
c79d967d 4795 int enable_quota = 0;
8a266467 4796 ext4_group_t g;
b3881f74 4797 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4798 int err = 0;
ac27a0ec 4799#ifdef CONFIG_QUOTA
03dafb5f 4800 int i, j;
ac27a0ec 4801#endif
d4c402d9 4802 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4803
4804 /* Store the original options */
4805 old_sb_flags = sb->s_flags;
4806 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4807 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4808 old_opts.s_resuid = sbi->s_resuid;
4809 old_opts.s_resgid = sbi->s_resgid;
4810 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4811 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4812 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4813#ifdef CONFIG_QUOTA
4814 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4815 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f
CG		 4816 if (sbi->s_qf_names[i]) {
4817 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4818 GFP_KERNEL);
4819 if (!old_opts.s_qf_names[i]) {
4820 for (j = 0; j < i; j++)
4821 kfree(old_opts.s_qf_names[j]);
3e36a163 4822 kfree(orig_data);
03dafb5f
CG		 4823 return -ENOMEM;
4824 }
4825 } else
4826 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4827#endif
b3881f74
TT		 4828 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4829 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4830
4831 /*
4832 * Allow the "check" option to be passed as a remount option.
4833 */
661aa520 4834 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4835 err = -EINVAL;
4836 goto restore_opts;
4837 }
4838
6ae6514b
PS		 4839 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4840 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4841 ext4_msg(sb, KERN_ERR, "can't mount with "
4842 "both data=journal and delalloc");
4843 err = -EINVAL;
4844 goto restore_opts;
4845 }
4846 if (test_opt(sb, DIOREAD_NOLOCK)) {
4847 ext4_msg(sb, KERN_ERR, "can't mount with "
4848 "both data=journal and dioread_nolock");
4849 err = -EINVAL;
4850 goto restore_opts;
4851 }
4852 }
4853
4ab2f15b 4854 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4855 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4856
4857 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4858 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4859
4860 es = sbi->s_es;
4861
b3881f74 4862 if (sbi->s_journal) {
0390131b 4863 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4864 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4865 }
ac27a0ec 4866
661aa520 4867 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4868 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4869 err = -EROFS;
4870 goto restore_opts;
4871 }
4872
4873 if (*flags & MS_RDONLY) {
38c03b34
TT		 4874 err = sync_filesystem(sb);
4875 if (err < 0)
4876 goto restore_opts;
0f0dd62f
CH		 4877 err = dquot_suspend(sb, -1);
4878 if (err < 0)
c79d967d 4879 goto restore_opts;
c79d967d 4880
ac27a0ec
DK		 4881 /*
4882 * First of all, the unconditional stuff we have to do
4883 * to disable replay of the journal when we next remount
4884 */
4885 sb->s_flags |= MS_RDONLY;
4886
4887 /*
4888 * OK, test if we are remounting a valid rw partition
4889 * readonly, and if so set the rdonly flag and then
4890 * mark the partition as valid again.
4891 */
617ba13b
MC		 4892 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4893 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4894 es->s_state = cpu_to_le16(sbi->s_mount_state);
4895
a63c9eb2 4896 if (sbi->s_journal)
0390131b 4897 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4898 } else {
a13fb1a4
ES		 4899 /* Make sure we can mount this feature set readwrite */
4900 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4901 err = -EROFS;
4902 goto restore_opts;
4903 }
8a266467
TT		 4904 /*
4905 * Make sure the group descriptor checksums
0b8e58a1 4906 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4907 */
4908 for (g = 0; g < sbi->s_groups_count; g++) {
4909 struct ext4_group_desc *gdp =
4910 ext4_get_group_desc(sb, g, NULL);
4911
feb0ab32 4912 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4913 ext4_msg(sb, KERN_ERR,
4914 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4915 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4916 le16_to_cpu(gdp->bg_checksum));
4917 err = -EINVAL;
4918 goto restore_opts;
4919 }
4920 }
4921
ead6596b
ES		 4922 /*
4923 * If we have an unprocessed orphan list hanging
4924 * around from a previously readonly bdev mount,
4925 * require a full umount/remount for now.
4926 */
4927 if (es->s_last_orphan) {
b31e1552 4928 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4929 "remount RDWR because of unprocessed "
4930 "orphan inode list. Please "
b31e1552 4931 "umount/remount instead");
ead6596b
ES		 4932 err = -EINVAL;
4933 goto restore_opts;
4934 }
4935
ac27a0ec
DK		 4936 /*
4937 * Mounting a RDONLY partition read-write, so reread
4938 * and store the current valid flag. (It may have
4939 * been changed by e2fsck since we originally mounted
4940 * the partition.)
4941 */
0390131b
FM		 4942 if (sbi->s_journal)
4943 ext4_clear_journal_err(sb, es);
ac27a0ec 4944 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4945 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4946 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4947 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4948 EXT4_FEATURE_INCOMPAT_MMP))
4949 if (ext4_multi_mount_protect(sb,
4950 le64_to_cpu(es->s_mmp_block))) {
4951 err = -EROFS;
4952 goto restore_opts;
4953 }
c79d967d 4954 enable_quota = 1;
ac27a0ec
DK		 4955 }
4956 }
bfff6873
LC		 4957
4958 /*
4959 * Reinitialize lazy itable initialization thread based on
4960 * current settings
4961 */
4962 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4963 ext4_unregister_li_request(sb);
4964 else {
4965 ext4_group_t first_not_zeroed;
4966 first_not_zeroed = ext4_has_uninit_itable(sb);
4967 ext4_register_li_request(sb, first_not_zeroed);
4968 }
4969
6fd058f7 4970 ext4_setup_system_zone(sb);
d096ad0f 4971 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4972 ext4_commit_super(sb, 1);
0390131b 4973
ac27a0ec
DK		 4974#ifdef CONFIG_QUOTA
4975 /* Release old quota file names */
4976 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f 4977 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4978 if (enable_quota) {
4979 if (sb_any_quota_suspended(sb))
4980 dquot_resume(sb, -1);
4981 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4982 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4983 err = ext4_enable_quotas(sb);
07724f98 4984 if (err)
7c319d32 4985 goto restore_opts;
7c319d32
AK		 4986 }
4987 }
ac27a0ec 4988#endif
d4c402d9
CW		 4989
4990 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4991 kfree(orig_data);
ac27a0ec 4992 return 0;
0b8e58a1 4993
ac27a0ec
DK		 4994restore_opts:
4995 sb->s_flags = old_sb_flags;
4996 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4997 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4998 sbi->s_resuid = old_opts.s_resuid;
4999 sbi->s_resgid = old_opts.s_resgid;
5000 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 5001 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5002 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 5003#ifdef CONFIG_QUOTA
5004 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5005 for (i = 0; i < MAXQUOTAS; i++) {
03dafb5f 5006 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 5007 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5008 }
5009#endif
d4c402d9 5010 kfree(orig_data);
ac27a0ec
DK		 5011 return err;
5012}
5013
2b2d6d01 5014static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 5015{
5016 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 5017 struct ext4_sb_info *sbi = EXT4_SB(sb);
5018 struct ext4_super_block *es = sbi->s_es;
27dd4385 5019 ext4_fsblk_t overhead = 0, resv_blocks;
960cc398 5020 u64 fsid;
d02a9391 5021 s64 bfree;
27dd4385 5022 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
ac27a0ec 5023
952fc18e
TT		 5024 if (!test_opt(sb, MINIX_DF))
5025 overhead = sbi->s_overhead;
ac27a0ec 5026
617ba13b 5027 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 5028 buf->f_bsize = sb->s_blocksize;
b72f78cb 5029 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 5030 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5031 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 5032 /* prevent underflow in case that few free space is available */
57042651 5033 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
27dd4385
LC		 5034 buf->f_bavail = buf->f_bfree -
5035 (ext4_r_blocks_count(es) + resv_blocks);
5036 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
ac27a0ec
DK		 5037 buf->f_bavail = 0;
5038 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 5039 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 5040 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 5041 fsid = le64_to_cpup((void *)es->s_uuid) ^
5042 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5043 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5044 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 5045
ac27a0ec
DK		 5046 return 0;
5047}
5048
0b8e58a1
AD		 5049/* Helper function for writing quotas on sync - we need to start transaction
5050 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 5051 * Process 1 Process 2
617ba13b 5052 * ext4_create() quota_sync()
a269eb18 5053 * jbd2_journal_start() write_dquot()
871a2931 5054 * dquot_initialize() down(dqio_mutex)
dab291af 5055 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 5056 *
5057 */
5058
5059#ifdef CONFIG_QUOTA
5060
5061static inline struct inode *dquot_to_inode(struct dquot *dquot)
5062{
4c376dca 5063 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 5064}
5065
617ba13b 5066static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 5067{
5068 int ret, err;
5069 handle_t *handle;
5070 struct inode *inode;
5071
5072 inode = dquot_to_inode(dquot);
9924a92a 5073 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
0b8e58a1 5074 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5075 if (IS_ERR(handle))
5076 return PTR_ERR(handle);
5077 ret = dquot_commit(dquot);
617ba13b 5078 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5079 if (!ret)
5080 ret = err;
5081 return ret;
5082}
5083
617ba13b 5084static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 5085{
5086 int ret, err;
5087 handle_t *handle;
5088
9924a92a 5089 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5090 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 5091 if (IS_ERR(handle))
5092 return PTR_ERR(handle);
5093 ret = dquot_acquire(dquot);
617ba13b 5094 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5095 if (!ret)
5096 ret = err;
5097 return ret;
5098}
5099
617ba13b 5100static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 5101{
5102 int ret, err;
5103 handle_t *handle;
5104
9924a92a 5105 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
0b8e58a1 5106 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 5107 if (IS_ERR(handle)) {
5108 /* Release dquot anyway to avoid endless cycle in dqput() */
5109 dquot_release(dquot);
ac27a0ec 5110 return PTR_ERR(handle);
9c3013e9 5111 }
ac27a0ec 5112 ret = dquot_release(dquot);
617ba13b 5113 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5114 if (!ret)
5115 ret = err;
5116 return ret;
5117}
5118
617ba13b 5119static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 5120{
262b4662
JK		 5121 struct super_block *sb = dquot->dq_sb;
5122 struct ext4_sb_info *sbi = EXT4_SB(sb);
5123
2c8be6b2 5124 /* Are we journaling quotas? */
262b4662
JK		 5125 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5126 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
ac27a0ec 5127 dquot_mark_dquot_dirty(dquot);
617ba13b 5128 return ext4_write_dquot(dquot);
ac27a0ec
DK		 5129 } else {
5130 return dquot_mark_dquot_dirty(dquot);
5131 }
5132}
5133
617ba13b 5134static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 5135{
5136 int ret, err;
5137 handle_t *handle;
5138
5139 /* Data block + inode block */
9924a92a 5140 handle = ext4_journal_start(sb->s_root->d_inode, EXT4_HT_QUOTA, 2);
ac27a0ec
DK		 5141 if (IS_ERR(handle))
5142 return PTR_ERR(handle);
5143 ret = dquot_commit_info(sb, type);
617ba13b 5144 err = ext4_journal_stop(handle);
ac27a0ec
DK		 5145 if (!ret)
5146 ret = err;
5147 return ret;
5148}
5149
5150/*
5151 * Turn on quotas during mount time - we need to find
5152 * the quota file and such...
5153 */
617ba13b 5154static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 5155{
287a8095
CH		 5156 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5157 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 5158}
5159
5160/*
5161 * Standard function to be called on quota_on
5162 */
617ba13b 5163static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 5164 struct path *path)
ac27a0ec
DK		 5165{
5166 int err;
ac27a0ec
DK		 5167
5168 if (!test_opt(sb, QUOTA))
5169 return -EINVAL;
0623543b 5170
ac27a0ec 5171 /* Quotafile not on the same filesystem? */
d8c9584e 5172 if (path->dentry->d_sb != sb)
ac27a0ec 5173 return -EXDEV;
0623543b
JK		 5174 /* Journaling quota? */
5175 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 5176 /* Quotafile not in fs root? */
f00c9e44 5177 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 5178 ext4_msg(sb, KERN_WARNING,
5179 "Quota file not on filesystem root. "
5180 "Journaled quota will not work");
2b2d6d01 5181 }
0623543b
JK		 5182
5183 /*
5184 * When we journal data on quota file, we have to flush journal to see
5185 * all updates to the file when we bypass pagecache...
5186 */
0390131b 5187 if (EXT4_SB(sb)->s_journal &&
f00c9e44 5188 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 5189 /*
5190 * We don't need to lock updates but journal_flush() could
5191 * otherwise be livelocked...
5192 */
5193 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 5194 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 5195 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 5196 if (err)
7ffe1ea8 5197 return err;
0623543b
JK		 5198 }
5199
f00c9e44 5200 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 5201}
5202
7c319d32
AK		 5203static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5204 unsigned int flags)
5205{
5206 int err;
5207 struct inode *qf_inode;
5208 unsigned long qf_inums[MAXQUOTAS] = {
5209 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5210 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5211 };
5212
5213 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5214
5215 if (!qf_inums[type])
5216 return -EPERM;
5217
5218 qf_inode = ext4_iget(sb, qf_inums[type]);
5219 if (IS_ERR(qf_inode)) {
5220 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5221 return PTR_ERR(qf_inode);
5222 }
5223
bcb13850
JK		 5224 /* Don't account quota for quota files to avoid recursion */
5225 qf_inode->i_flags |= S_NOQUOTA;
7c319d32
AK		 5226 err = dquot_enable(qf_inode, type, format_id, flags);
5227 iput(qf_inode);
5228
5229 return err;
5230}
5231
5232/* Enable usage tracking for all quota types. */
5233static int ext4_enable_quotas(struct super_block *sb)
5234{
5235 int type, err = 0;
5236 unsigned long qf_inums[MAXQUOTAS] = {
5237 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5238 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5239 };
5240
5241 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5242 for (type = 0; type < MAXQUOTAS; type++) {
5243 if (qf_inums[type]) {
5244 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5245 DQUOT_USAGE_ENABLED);
5246 if (err) {
5247 ext4_warning(sb,
72ba7450
TT		 5248 "Failed to enable quota tracking "
5249 "(type=%d, err=%d). Please run "
5250 "e2fsck to fix.", type, err);
7c319d32
AK		 5251 return err;
5252 }
5253 }
5254 }
5255 return 0;
5256}
5257
5258/*
5259 * quota_on function that is used when QUOTA feature is set.
5260 */
5261static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5262 int format_id)
5263{
5264 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5265 return -EINVAL;
5266
5267 /*
5268 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5269 */
5270 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5271}
5272
ca0e05e4
DM		 5273static int ext4_quota_off(struct super_block *sb, int type)
5274{
21f97697
JK		 5275 struct inode *inode = sb_dqopt(sb)->files[type];
5276 handle_t *handle;
5277
87009d86
DM		 5278 /* Force all delayed allocation blocks to be allocated.
5279 * Caller already holds s_umount sem */
5280 if (test_opt(sb, DELALLOC))
ca0e05e4 5281 sync_filesystem(sb);
ca0e05e4 5282
0b268590
AG		 5283 if (!inode)
5284 goto out;
5285
21f97697
JK		 5286 /* Update modification times of quota files when userspace can
5287 * start looking at them */
9924a92a 5288 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
21f97697
JK		 5289 if (IS_ERR(handle))
5290 goto out;
5291 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5292 ext4_mark_inode_dirty(handle, inode);
5293 ext4_journal_stop(handle);
5294
5295out:
ca0e05e4
DM		 5296 return dquot_quota_off(sb, type);
5297}
5298
7c319d32
AK		 5299/*
5300 * quota_off function that is used when QUOTA feature is set.
5301 */
5302static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5303{
5304 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5305 return -EINVAL;
5306
5307 /* Disable only the limits. */
5308 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5309}
5310
ac27a0ec
DK		 5311/* Read data from quotafile - avoid pagecache and such because we cannot afford
5312 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5313 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5314 * we don't have to be afraid of races */
617ba13b 5315static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5316 size_t len, loff_t off)
5317{
5318 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5319 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5320 int err = 0;
5321 int offset = off & (sb->s_blocksize - 1);
5322 int tocopy;
5323 size_t toread;
5324 struct buffer_head *bh;
5325 loff_t i_size = i_size_read(inode);
5326
5327 if (off > i_size)
5328 return 0;
5329 if (off+len > i_size)
5330 len = i_size-off;
5331 toread = len;
5332 while (toread > 0) {
5333 tocopy = sb->s_blocksize - offset < toread ?
5334 sb->s_blocksize - offset : toread;
617ba13b 5335 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5336 if (err)
5337 return err;
5338 if (!bh) /* A hole? */
5339 memset(data, 0, tocopy);
5340 else
5341 memcpy(data, bh->b_data+offset, tocopy);
5342 brelse(bh);
5343 offset = 0;
5344 toread -= tocopy;
5345 data += tocopy;
5346 blk++;
5347 }
5348 return len;
5349}
5350
5351/* Write to quotafile (we know the transaction is already started and has
5352 * enough credits) */
617ba13b 5353static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5354 const char *data, size_t len, loff_t off)
5355{
5356 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5357 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5358 int err = 0;
5359 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5360 struct buffer_head *bh;
5361 handle_t *handle = journal_current_handle();
5362
0390131b 5363 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5364 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5365 " cancelled because transaction is not started",
9c3013e9
JK		 5366 (unsigned long long)off, (unsigned long long)len);
5367 return -EIO;
5368 }
67eeb568
DM		 5369 /*
5370 * Since we account only one data block in transaction credits,
5371 * then it is impossible to cross a block boundary.
5372 */
5373 if (sb->s_blocksize - offset < len) {
5374 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5375 " cancelled because not block aligned",
5376 (unsigned long long)off, (unsigned long long)len);
5377 return -EIO;
5378 }
5379
67eeb568
DM		 5380 bh = ext4_bread(handle, inode, blk, 1, &err);
5381 if (!bh)
5382 goto out;
5d601255 5383 BUFFER_TRACE(bh, "get write access");
62d2b5f2
JK		 5384 err = ext4_journal_get_write_access(handle, bh);
5385 if (err) {
5386 brelse(bh);
5387 goto out;
ac27a0ec 5388 }
67eeb568
DM		 5389 lock_buffer(bh);
5390 memcpy(bh->b_data+offset, data, len);
5391 flush_dcache_page(bh->b_page);
5392 unlock_buffer(bh);
62d2b5f2 5393 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5394 brelse(bh);
ac27a0ec 5395out:
0b7f7cef 5396 if (err)
ac27a0ec 5397 return err;
67eeb568
DM		 5398 if (inode->i_size < off + len) {
5399 i_size_write(inode, off + len);
617ba13b 5400 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5401 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5402 }
67eeb568 5403 return len;
ac27a0ec
DK		 5404}
5405
5406#endif
5407
152a0836
AV		 5408static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5409 const char *dev_name, void *data)
ac27a0ec 5410{
152a0836 5411 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5412}
5413
37f328eb 5414#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5415static inline void register_as_ext2(void)
5416{
5417 int err = register_filesystem(&ext2_fs_type);
5418 if (err)
5419 printk(KERN_WARNING
5420 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5421}
5422
5423static inline void unregister_as_ext2(void)
5424{
5425 unregister_filesystem(&ext2_fs_type);
5426}
2035e776
TT		 5427
5428static inline int ext2_feature_set_ok(struct super_block *sb)
5429{
5430 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5431 return 0;
5432 if (sb->s_flags & MS_RDONLY)
5433 return 1;
5434 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5435 return 0;
5436 return 1;
5437}
24b58424
TT		 5438#else
5439static inline void register_as_ext2(void) { }
5440static inline void unregister_as_ext2(void) { }
2035e776 5441static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5442#endif
5443
37f328eb 5444#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5445static inline void register_as_ext3(void)
5446{
5447 int err = register_filesystem(&ext3_fs_type);
5448 if (err)
5449 printk(KERN_WARNING
5450 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5451}
5452
5453static inline void unregister_as_ext3(void)
5454{
5455 unregister_filesystem(&ext3_fs_type);
5456}
2035e776
TT		 5457
5458static inline int ext3_feature_set_ok(struct super_block *sb)
5459{
5460 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5461 return 0;
5462 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5463 return 0;
5464 if (sb->s_flags & MS_RDONLY)
5465 return 1;
5466 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5467 return 0;
5468 return 1;
5469}
24b58424
TT		 5470#else
5471static inline void register_as_ext3(void) { }
5472static inline void unregister_as_ext3(void) { }
2035e776 5473static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5474#endif
5475
03010a33
TT		 5476static struct file_system_type ext4_fs_type = {
5477 .owner = THIS_MODULE,
5478 .name = "ext4",
152a0836 5479 .mount = ext4_mount,
03010a33
TT		 5480 .kill_sb = kill_block_super,
5481 .fs_flags = FS_REQUIRES_DEV,
5482};
7f78e035 5483MODULE_ALIAS_FS("ext4");
03010a33 5484
8f021222 5485static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5486{
5487 struct ext4_features *ef;
5488 int ret = -ENOMEM;
5489
5490 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5491 if (!ef)
5492 goto out;
5493
5494 ef->f_kobj.kset = ext4_kset;
5495 init_completion(&ef->f_kobj_unregister);
5496 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5497 "features");
5498 if (ret) {
5499 kfree(ef);
5500 goto out;
5501 }
5502
5503 ext4_feat = ef;
5504 ret = 0;
5505out:
5506 return ret;
5507}
5508
8f021222
LC		 5509static void ext4_exit_feat_adverts(void)
5510{
5511 kobject_put(&ext4_feat->f_kobj);
5512 wait_for_completion(&ext4_feat->f_kobj_unregister);
5513 kfree(ext4_feat);
5514}
5515
e9e3bcec
ES		 5516/* Shared across all ext4 file systems */
5517wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5518struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5519
5dabfc78 5520static int __init ext4_init_fs(void)
ac27a0ec 5521{
e9e3bcec 5522 int i, err;
c9de560d 5523
07c0c5d8
AV		 5524 ext4_li_info = NULL;
5525 mutex_init(&ext4_li_mtx);
5526
9a4c8019 5527 /* Build-time check for flags consistency */
12e9b892 5528 ext4_check_flag_values();
e9e3bcec
ES		 5529
5530 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5531 mutex_init(&ext4__aio_mutex[i]);
5532 init_waitqueue_head(&ext4__ioend_wq[i]);
5533 }
5534
51865fda 5535 err = ext4_init_es();
6fd058f7
TT		 5536 if (err)
5537 return err;
51865fda
ZL		 5538
5539 err = ext4_init_pageio();
5540 if (err)
5541 goto out7;
5542
5dabfc78 5543 err = ext4_init_system_zone();
bd2d0210 5544 if (err)
d44651d0 5545 goto out6;
3197ebdb 5546 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5547 if (!ext4_kset) {
5548 err = -ENOMEM;
dd68314c 5549 goto out5;
0e376b1e 5550 }
d44651d0 5551 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5552
5553 err = ext4_init_feat_adverts();
dd68314c
TT		 5554 if (err)
5555 goto out4;
857ac889 5556
5dabfc78 5557 err = ext4_init_mballoc();
c9de560d
AT		 5558 if (err)
5559 goto out2;
9c191f70
M		 5560 else
5561 ext4_mballoc_ready = 1;
ac27a0ec
DK		 5562 err = init_inodecache();
5563 if (err)
5564 goto out1;
24b58424 5565 register_as_ext3();
2035e776 5566 register_as_ext2();
03010a33 5567 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5568 if (err)
5569 goto out;
bfff6873 5570
ac27a0ec
DK		 5571 return 0;
5572out:
24b58424
TT		 5573 unregister_as_ext2();
5574 unregister_as_ext3();
ac27a0ec
DK		 5575 destroy_inodecache();
5576out1:
9c191f70 5577 ext4_mballoc_ready = 0;
5dabfc78 5578 ext4_exit_mballoc();
9c191f70 5579out2:
8f021222 5580 ext4_exit_feat_adverts();
dd68314c 5581out4:
d44651d0
FJ		 5582 if (ext4_proc_root)
5583 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5584 kset_unregister(ext4_kset);
d44651d0 5585out5:
5dabfc78 5586 ext4_exit_system_zone();
d44651d0 5587out6:
5dabfc78 5588 ext4_exit_pageio();
51865fda
ZL		 5589out7:
5590 ext4_exit_es();
5591
ac27a0ec
DK		 5592 return err;
5593}
5594
5dabfc78 5595static void __exit ext4_exit_fs(void)
ac27a0ec 5596{
bfff6873 5597 ext4_destroy_lazyinit_thread();
24b58424
TT		 5598 unregister_as_ext2();
5599 unregister_as_ext3();
03010a33 5600 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5601 destroy_inodecache();
5dabfc78 5602 ext4_exit_mballoc();
8f021222 5603 ext4_exit_feat_adverts();
9f6200bb 5604 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5605 kset_unregister(ext4_kset);
5dabfc78
TT		 5606 ext4_exit_system_zone();
5607 ext4_exit_pageio();
dd12ed14 5608 ext4_exit_es();
ac27a0ec
DK		 5609}
5610
5611MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5612MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5613MODULE_LICENSE("GPL");
5dabfc78
TT		 5614module_init(ext4_init_fs)
5615module_exit(ext4_exit_fs)